Da-Wen Sun (Editor) - Computational Fluid Dynamics in Food Processing-CRC Press (2018)

Computational Fluid
Dynamics in Food Processing

Second Edition
Contemporary Food Engineering
Series Editor
Professor Da-Wen Sun, Director
Food Refrigeration & Computerized Food Technology
National University of Ireland, Dublin
(University College Dublin)
Dublin, Ireland
http://www.ucd.ie/sun/
Emerging Technologies for Food Quality and Food Safety Evaluation, edited by Yong-Jin Cho
and Sukwon Kang
Operations in Food Refrigeration, edited by Rodolfo H. Mascheroni
Advances in Food Extrusion Technology, edited by Medeni Maskan and Aylin Altan
Modified Atmosphere and Active Packaging Technologies, edited by Ioannis Arvanitoyannis
Juice Processing: Quality, Safety and Value-Added Opportunities, edited by Victor Falguera
and Albert Ibarz
Physical Properties of Foods: Novel Measurement Techniques and Applications, edited by
Ignacio Arana
Fermentation Processes Engineering in the Food Industry, edited by Carlos Ricardo Soccol,
Ashok Pandey, and Christian Larroche
Engineering Aspects of Cereal and Cereal-Based Products, edited by Raquel de Pinho Ferreira
Guine, and Paula Maria dos Reis Correia
Enhancing Extraction Processes in the Food Industry, edited by Nikolai Lebovka, Eugene
Vorobiev, and Farid Chemat
Thermal Food Processing: New Technologies and Quality Issues, Second Edition, edited by
Da-Wen Sun
Advances in Fruit Processing Technologies, edited by Sueli Rodrigues and Fabiano Andre
Narciso Fernandes
Biosensors in Food Processing, Safety, and Quality Control, edited by Mehmet Mutlu
Edible Oils: Extraction, Processing, and Applications, edited by Smain Chemat
Engineering Aspects of Membrane Separation and Application in Food Processing, edited by
Robert W. Field, Erika Bekassy-Molnar, Frank Lipnizki, and Gyula Vatai
Engineering Aspects of Food Emulsification and Homogenization, edited by Marilyn Rayner
and Petr Dejmek
Advances in Meat Processing Technology, edited by Alaa El-Din A. Bekhit
Engineering Aspects of Food Biotechnology, edited by Jose A. Teixeira and Antonio A. Vicente
High Pressure Processing of Fruit and Vegetable Juices, edited by Milan Houška and Filipa
Vinagre Marques da Silva
Trends in Fish Processing Technologies, edited by Daniela Borda, Anca I. Nicolau, and Peter Raspor
Food Biofortification Technologies, edited by Agnieszka Saeid
Advances in Postharvest Fruit and Vegetable Technology, edited by Ron B.H. Wills and John
Golding
Computational Fluid Dynamics in Food Processing, Second Edition, edited by Da-Wen Sun
For more information about this series, please visit: https://www.crcpress.com/

Contemporary-Food-Engineering/book-series/CRCCONFOOENG
Computational Fluid
Dynamics in Food Processing
Second Edition
Edited by
Da-Wen Sun
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2019 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works
Printed on acid-free paper
International Standard Book Number-13: 978-1-138-56831-0 (Hardback)
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to
publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materi-
als or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material
reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained.
If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in
any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, micro-
filming, and recording, or in any information storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www
.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-
8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that
have been granted a photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identifi-
cation and explanation without intent to infringe.
Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com
and the CRC Press Web site at

http://www.crcpress.com
Contents
Series Preface.....................................................................................................................................ix
Preface to the Second Edition............................................................................................................xi
Editor............................................................................................................................................................xiii
Contributors................................................................................................................................................. xv
Section I CFD Applications in Cold Chain Facilities
Chapter 1 CFD Aided Retail Cabinets Design..............................................................................3

Giovanni Cortella
Chapter 2 CFD Optimization of Perturbed Air Curtains for Refrigerated Display Cabinets..... 23
Jean Moureh
Chapter 3 CFD Modeling to Improve the Performance of Industrial Cooling of Large

Beef Carcasses..................................................................................................................61
Mulugeta A. Delele, Kumsa D. Kuffi, Bart Nicolai, and Pieter Verboven
Chapter 4 CFD Modeling of Heat and Mass Transfer in a Hydrofluidization System

During Food Chilling and Freezing............................................................................ 87
Juan Manuel Peralta and Susana E. Zorrilla
Chapter 5 Improving the Performance of a Partially Loaded Cold Store by CFD.....................105

Fumihiko Tanaka and Fumina Tanaka
Chapter 6 CFD Investigation of Fresh Produce Cooling Processes and Effects

of Package Stacking.......................................................................................................121
Umezuruike Linus Opara, Alemayehu Ambaw, and Tarl Berry
Chapter 7 Optimization of Ventilation Ports of Packaging for Fresh Produce Using CFD...... 149
Hiroaki Kitazawa
Chapter 8 Optimization of Horticultural Carton Vent Hole Design by CFD..............................169

Umezuruike Linus Opara, Tarl Berry, and Alemayehu Ambaw
v
vi Contents
Section II CFD Applications in Thermal

Processing and Heat Exchangers
Chapter 9 Three-Dimensional CFD Modeling of Continuous Industrial Baking Process........ 193

Weibiao Zhou and Nantawan Therdthai
Chapter 10 Improving Bread-Baking Process Under Different Oven Load Conditions

by CFD Modeling..................................................................................................... 225
N. Chhanwal, J. A. Moses, and C. Anandharamakrishnan
Chapter 11 CFD Applications of Food Packaging Sterilization and Filling............................... 243

Giuseppe Vignali and Filippo Ferrari
Chapter 12 CFD Analysis of Food Pasteurization Processes.........................................................261

Padma Ishwarya and C. Anandharamakrishnan
Chapter 13 CFD Determination of F Values During Thermal Processing of Still Cans............ 289
Won Byong Yoon and Hyeon Woo Park
Chapter 14 CFD Modeling of Thermal Processing of Particulate Foods.................................... 305

Massimiliano Rinaldi, Matteo Cordioli, and Davide Barbanti
Chapter 15 CFD Modeling of Natural-Convection Heating Processes....................................... 319

Ferruh Erdogdu, Huseyin Topcam, Fabrizio Sarghini, and Francesco Marra
Chapter 16 CFD Modeling of Convective Drying of Cylindrical Fruit Slices..............................339

Alexandros Vouros, Dimitrios Tzempelikos, Dimitrios Mitrakos,
and Andronikos Filios
Chapter 17 CFD Modeling of Convection Flow in Pan Cooking................................................ 365

Yvan Llave and Noboru Sakai
Chapter 18 CFD Analysis of Thermal Processing of Intact Eggs................................................. 389

Behzad Abbasnezhad, Mohsen Dalvi-Isfahan, and Nasser Hamdami
Chapter 19 Applications of CFD for Optimization of Cabinet Dryers........................................ 415

Yasaman Amanlou and Mohammad Hadi Khoshtaghaza
Contents vii
Section III CFD Applications in Other Food Processes
Chapter 20 CFD Design and Optimization of Biosensors in the Food Industry..........................439

Agnese Piovesan, Jeroen Lammertyn, Bart Nicolai, and Pieter Verboven
Chapter 21 Analysis and Simulation of Pasta Dough Extrusion Process by CFD....................... 463
Fabrizio Sarghini
Chapter 22 Computational Modeling of Radio Frequency Thawing of Frozen Food Products..... 487
Francesco Marra, Tesfaye Faye Bedane, Oriana Casaburi, Ozan Altin,
Rahmi Uyar, and Ferruh Erdogdu
Chapter 23 CFD Application for the Evaluation of Food Texture...............................................509

Won Byong Yoon and Hwabin Jung
Chapter 24 CFD Study of Top-Spray Fluidized Bed Coating Process........................................ 531

Wasan Duangkhamchan, Frederik Ronsse, and Jan G. Pieters
Chapter 25 Operation of Biofilm Reactors for the Food Industry Using CFD............................ 561
Luciana C. Gomes, João Miranda, and Filipe J. Mergulhão
Index............................................................................................................................................... 591
http://taylorandfrancis.com
Series Preface
CONTEMPORARY FOOD ENGINEERING
Food engineering is the multidisciplinary field of applied physical sciences combined with the
knowledge of product properties. Food engineers provide the technological knowledge transfer
essential to the cost-effective production and commercialization of food products and services.
In particular, food engineers develop and design processes and equipment in order to convert raw
agricultural materials and ingredients into safe, convenient, and nutritious consumer food products.
However, food engineering topics are continuously undergoing changes to meet diverse consumer
demands, and the subject is being rapidly developed to reflect market needs.
In the development of food engineering, one of the many challenges is to employ modern tools
and knowledge, such as computational materials science and nanotechnology, to develop new prod-
ucts and processes. Simultaneously, improving quality, safety, and security remain critical issues
in food engineering study. New packaging materials and techniques are being developed to pro-
vide more protection to foods, and novel preservation technologies are emerging to enhance food
security and defense. Additionally, process control and automation regularly appear among the top
priorities identified in food engineering. Advanced monitoring and control systems are developed to
facilitate automation and flexible food manufacturing. Furthermore, energy saving and minimiza-
tion of environmental problems continue to be important food engineering issues, and significant
progress is being made in waste management, efficient utilization of energy, and reduction of efflu-
ents and emissions in food production.
Consisting of edited books, the Contemporary Food Engineering book series attempts to address
some of the recent developments in food engineering. Advances in classical unit operations in engi-
neering applied to food manufacturing are covered as well as such topics as progress in the transport
and storage of liquid and solid foods; heating, chilling, and freezing of foods; mass transfer in foods;
chemical and biochemical aspects of food engineering and the use of kinetic analysis; dehydra-
tion, thermal processing, nonthermal processing, extrusion, liquid food concentration, membrane
processes and applications of membranes in food processing; shelf-life, electronic indicators in
inventory management, and sustainable technologies in food processing; and packaging, cleaning,
and sanitation. The books are aimed at professional food scientists, academics researching food
engineering problems, and graduate level students.
The editors of the books are leading engineers and scientists from many parts of the world. All
the editors were asked to present their books in a manner that will address the market need and
pinpoint the cutting-edge technologies in food engineering. Furthermore, all contributions are writ-
ten by internationally renowned experts who have both academic and professional credentials. All
authors have attempted to provide critical, comprehensive, and readily accessible information on
the art and science of a relevant topic in each chapter, with reference lists to be used by readers for
further information. Therefore, each book can serve as an essential reference source to students and
researchers in universities and research institutions.
Da-Wen Sun
Series Editor
ix
Preface to the Second Edition
Computational Fluid Dynamics (CFD) uses applied mathematics, physics and computational soft-
ware to provide quick and efficient simulation and visualization of fluid flow and heat transfer. CFD
analysis aims to solve the governing equations describing fluid flow (i.e., the continuity equation
and the set of Navier–Stokes equations) and any additional conservation equations, such as energy
balance, in order to predict the profiles of velocity, temperature, pressure, and other parameters.
As many processes in the food industry involve fluid flow and heat and mass transfer, CFD pro-
vides a powerful early-stage simulation tool for gaining a qualitative and quantitative assessment
of the performance of food processing, allowing engineers to test concepts all the way through the
development of a process or system. By experimenting and analyzing the significance and effects
of various design parameters and working conditions on the computer, better understanding of the
dynamics and the underlying physics of a food process or phenomenon can be achieved, leading to
the optimization of existing and new processes and the overcoming of the need to test the design
with each modification. Therefore, the food industry has used CFD widely in its continual quest for
process and product improvement.
As the first book in the area, the first edition of Computational Fluid Dynamics in Food
Processing was published in 2007, with the main aims to present a comprehensive review of CFD
applications for the food industry and pinpoint the research and development trends in the develop-
ment of the technology, to provide the engineer and technologist working in research, development,
and operations in the food industry with critical, comprehensive and readily accessible information
on the art and science of CFD, and to serve as an essential reference source to undergraduate and
postgraduate students and researchers in universities and research institutions. This will continue
to be the purpose of this second edition. In the second edition, in order to reflect the most recent
research and development trends in the technology, only a few original chapters are updated with
latest developments. Therefore, this new edition mostly contains new chapters covering the analysis
and optimization of cold chain facilities, simulation of thermal processing and modeling of heat
exchangers, and CFD applications in other food processes.
xi
Editor
Born in Southern China, Professor Da-Wen Sun is a global
authority in food engineering research and education. He is a
Member of the Royal Irish Academy, the highest academic hon-
our in Ireland; a Member of Academia Europaea (The Academy
of Europe), one of the most prestigious academies in the world; a
Foreign Member of the Polish Academy of Sciences, the highest
lifetime honour bestowed by the Polish government; a Fellow
of the International Academy of Food Science and Technology;
a Fellow of the International Academy of Agricultural and
Biosystems Engineering; and a Full Member (Academician)
of the International Academy of Refrigeration. He is also
the founder and Editor-in-Chief of Food and Bioprocess
Technology, one of the most prestigious food science and technology journals; Series Editor of
Contemporary Food Engineering book series with already over 50 volumes published; and the
Founder and President of the International Academy of Agricultural and Biosystems Engineering
(iAABE). In addition, he served as the President the International Commission of Agricultural and
Biosystems Engineering (CIGR), the world largest organization in the field, in 2013–2014, and is
now Honorary President of CIGR. He has significantly contributed to the field of food engineering
as a researcher, as an academic authority and as an educator.
His main research activities include cooling, drying and refrigeration processes and systems,
quality and safety of food products, bioprocess simulation and optimization, and computer vision/
image processing and hyperspectral imaging technologies. Especially, his many scholarly works
have become standard reference materials for researchers in the areas of hyperspectral imaging,
computer vision, ultrasound assisted freezing, vacuum cooling, computational fluid dynamics mod-
eling, etc. Results of his work have been published in over 900 papers including more than 500
peer-reviewed journal papers indexed by Web of Science, with an average citation of over 35 per
paper (Web of Science h-index = 91, SCOPUS h-index = 94). Among them, 50 papers have been
selected by Thomson Reuters’s Essential Science IndicatorsSM as Highly Cited/Hot Papers in Field,
ranking him No. 2 in the world in Agricultural Sciences. He has also edited 17 authoritative books.
In addition, Professor Sun has been named Highly Cited Researcher in the last three consecutive
years (2015–2017) by Clarivate Analytics (formerly Thomson Reuters).
He received a first class BSc Honours and MSc in Mechanical Engineering, and a PhD in
Chemical Engineering in China before working in various universities in Europe. He became the
first Chinese national to be permanently employed in an Irish University when he was appointed
College Lecturer at National University of Ireland, Dublin (University College Dublin) in 1995, and
was then continuously promoted in the shortest possible time to Associate Professor, Professor, and
Full Professor. Dr. Sun is now the Full Professor of Food and Biosystems Engineering and Director
of the Food Refrigeration and Computerised Food Technology Research Group at University
College Dublin (UCD).
As a leading educator in food engineering, Professor Sun has significantly contributed to the field
of food engineering. He has trained many PhD students, who have made their own c ontributions
to the industry and academia. He has also given lectures on advances in food engineering on a
regular basis in academic institutions internationally and delivered keynote speeches at interna-
tional conferences. In recognition of his significant contribution to Food Engineering worldwide
and for his outstanding leadership in the field, the International Commission of Agricultural and
Biosystems Engineering (CIGR) awarded him the “CIGR Merit Award” in 2000, in 2006, and
again in 2016. The Institution of Mechanical Engineers (IMechE) based in the UK named him
xiii
xiv Editor
“Food Engineer of the Year 2004.” In 2007 he was presented with the only “AFST(I) Fellow
Award” in that year by the Association of Food Scientists and Technologists (India). In 2008 he was
awarded “CIGR Recognition Award” in honour of his distinguished achievements as the top one
percent of Agricultural Engineering scientists in the world. In 2010 he received the “CIGR Fellow
Award”; the title of Fellow is the highest honour in CIGR, and is conferred to individuals who have
made sustained, outstanding contributions worldwide. In March 2013, he was presented with the
“You Bring Charm to the World Award” by Hong Kong-based Phoenix Satellite Television, with
other award recipients including the 2012 Nobel Laureate in Literature and the Chinese Astronaut
Team for Shenzhou IX Spaceship. In July 2013 he received “The Frozen Food Foundation Freezing
Research Award” from the International Association for Food Protection (IAFP) for his significant
contributions to enhancing the field of food freezing technologies, the first time that this presti-
gious award was presented to a scientist outside the USA. In June 2015 he was presented with the
“IAEF Lifetime Achievement Award”, this IAEF (International Association of Engineering and
Food) award highlights the lifetime contribution of a prominent engineer in the field of food, and
in February 2018, he was conferred with the honorary doctorate degree by Universidad Privada del
Norte in Peru.
Contributors
Behzad Abbasnezhad Oriana Casaburi
Food Science and Technology Department Department of Industrial Engineering
College of Agriculture University of Salerno
Isfahan University of Technology Fisciano, Italy
Isfahan, Iran
N. Chhanwal
Ozan Altin Food Engineering Department
Department of Food Engineering Institute of Chemical Technology
Ankara University Mumbai, India
Ankara, Turkey
Matteo Cordioli
Yasaman Amanlou Department of Food Science
Biosystems Engineering Department University of Parma
Tarbiat Modares University Parma, Italy
Tehran, Iran
Giovanni Cortella
Alemayehu Ambaw Polytechnic Department of Engineering and
Postharvest Technology Research Laboratory Architecture
South African Research Chair in Postharvest University of Udine
Technology Udine, Italy
Stellenbosch University
Stellenbosch, South Africa Mohsen Dalvi-Isfahan
Faculty of Agriculture
C. Anandharamakrishnan Department of Food Science and Technology
Director, Indian Institute of Food Processing Jahrom University
Technology (IIFPT) Jahrom, Iran
Thanjavur, India
Mulugeta A. Delele
Davide Barbanti BIOSYST-MeBioS
Department of Food Science University of Leuven
University of Parma Heverlee, Belgium
Parma, Italy
Wasan Duangkhamchan
Tesfaye Faye Bedane Functional Foods, Department of Food
Department of Industrial Engineering Technology and Nutrition
University of Salerno Faculty of Technology
Fisciano, Italy Mahasarakham University
Maha Sarakham, Thailand
Tarl Berry
Postharvest Technology Research Laboratory Ferruh Erdogdu
South African Research Chair in Postharvest Department of Food Engineering
Technology Ankara University
Stellenbosch University Ankara, Turkey
Stellenbosch, South Africa
xv
xvi Contributors
Filippo Ferrari Jeroen Lammertyn

Department of Engineering and Architecture Faculty of Bioscience Engineering
University of Parma Department of Biosystems
Parma, Italy Mechatronics-Biostatistics-Sensors Division
Katholieke Universiteit Leuven
Andronikos Filios Leuven, Belgium
Department of Mechanical Engineering
University of West Attica Yvan Llave
Athens, Greece Department of Agro–Food Science
Niigata Agro–Food University
Luciana C. Gomes Niigata, Japan
LEPABE
Department of Chemical Engineering Francesco Marra
Faculty of Engineering Dipartimento di Ingegneria Industriale
University of Porto Universita degli studi di Salerno
Porto, Portugal Fisciano, Italy
Nasser Hamdami Filipe J. Mergulhão

Food Science and Technology Department LEPABE
College of Agriculture Department of Chemical Engineering
Isfahan University of Technology Faculty of Engineering
Isfahan, Iran University of Porto
Porto, Portugal
Padma Ishwarya
Indian Institute of Food Processing Technology
João Miranda
Thanjavur, India
CEFT
Hwabin Jung Department of Chemical Engineering
Department of Food Science and Faculty of Engineering
Biotechnology University of Porto
College of Agricultural and Life Science Porto, Portugal
Kangwon National University
Chuncheon, South Korea Dimitrios Mitrakos
Greek Atomic Energy Commission
Mohammad Hadi Khoshtaghaza Athens, Greece
Biosystems Engineering Department
Tarbiat Modares University J. A. Moses
Tehran, Iran Computational Modeling and Nanoscale
Processing Unit
Hiroaki Kitazawa Indian Institute of Food Processing Technology
Postharvest Science and Technology Unit Thanjavur, India
Division of Food Processing and Distribution
Research Jean Moureh
Food Research Institute Chargé de Recherche
National Agriculture and Food Research HDR/Researcher, Irstea
Organization (NARO) Unité de Recherche Génie des Procédés
Tsukuba, Japan Frigorifiques
Antony, France
Kumsa D. Kuffi
BIOSYST-MeBioS
University of Leuven
Heverlee, Belgium
Contributors xvii
Bart Nicolai Noboru Sakai

Faculty of Bioscience Engineering Department of Food Science and Technology
Department of Biosystems Tokyo University of Marine Science and
Mechatronics-Biostatistics-Sensors Division Technology
Katholieke Universiteit Leuven Tokyo, Japan
Leuven, Belgium
Fabrizio Sarghini
Umezuruike Linus Opara University of Naples Federico II
Postharvest Technology Research Laboratory Department of Agricultural Sciences
South African Research Chair in Postharvest Portici, Italy
Technology
Stellenbosch University Fumihiko Tanaka
Stellenbosch, South Africa Department of Bio-production Environmental
Sciences
Hyeon Woo Park Faculty of Agriculture
Department of Food Science and Kyushu University
Biotechnology Fukuoka, Japan
College of Agricultural and Life Science
Kangwon National University Fumina Tanaka
Chuncheon, South Korea Department of Bio-production Environmental
Sciences
Juan Manuel Peralta Faculty of Agriculture
Instituto de Desarrollo Tecnológico para la Kyushu University
Industria Química (INTEC) Fukuoka, Japan
Universidad Nacional del Litoral – CONICET
Santa Fe, Argentina Nantawan Therdthai
Department of Product Development
Jan G. Pieters Faculty of Agro-Industry
Department of Biosystems Engineering Kasetsart University
Faculty of Bioscience Engineering Bangkok, Thailand
Ghent University
Gent, Belgium Huseyin Topcam
Department of Food Engineering
Agnese Piovesan Ankara University
Faculty of Bioscience Engineering Ankara, Turkey
Department of Biosystems
Mechatronics-Biostatistics-Sensors Division Dimitrios Tzempelikos
Katholieke Universiteit Leuven Department of Mechanical Engineering
Leuven, Belgium University of West Attica
Athens, Greece
Massimiliano Rinaldi
Department of Food Science Rahmi Uyar
University of Parma Department of Food Engineering
Parma, Italy Siirt University
Siirt, Turkey
Frederik Ronsse
Department of Biosystems Engineering Pieter Verboven
Faculty of Bioscience Engineering BIOSYST-MeBioS
Ghent University University of Leuven
Gent, Belgium Heverlee, Belgium
xviii Contributors
Giuseppe Vignali Weibiao Zhou

Department of Industrial Engineering Food Science and Technology Programme
University of Parma Department of Chemistry
Parco Area delle Scienze National University of Singapore
Parma, Italy Singapore
Alexandros Vouros Susana E. Zorrilla

Department of Mechanical Engineering Instituto de Desarrollo Tecnológico para la
Educators Industria Química (INTEC)
School of Pedagogical and Technological Universidad Nacional del Litoral – CONICET
Education (ASPETE) Santa Fe, Argentina
Athens, Greece
Won Byong Yoon

Department of Food Science and
Biotechnology
College of Agricultural and Life Science
Kangwon National University
Chuncheon, South Korea
Section I
CFD Applications in Cold
Chain Facilities
1 CFD Aided Retail Cabinets
Design
Giovanni Cortella
CONTENTS
1.1 Introduction........................................................................................................................... 4
1.2 The Retail Cabinet.....................................................................................................................4
1.2.1 Classification..................................................................................................................4
1.2.1.1 Storage Temperature.................................................................................... 4
1.2.1.2 Geometry..................................................................................................................5
1.2.1.3 Refrigeration Equipment................................................................................. 6
1.2.1.4 Air Circulation................................................................................................ 7
1.2.1.5 Energy Consumption.................................................................................................. 7
1.2.2 Standardized Temperature Tests....................................................................................7
1.2.3 Air Curtains................................................................................................................... 9
1.3 Applications of CFD to Display Cabinets.......................................................................................10
1.3.1 Modeling Product Temperature Distribution.............................................................. 10
1.3.2 Modeling Airflow..................................................................................................................11
1.3.2.1 Air Curtains...................................................................................................... 12
1.3.2.2 Shelves.........................................................................................................................12
1.3.2.3 Evaporator and Rear Ducts...............................................................................13
1.3.3 Modeling the Influence of Air Humidity..................................................................... 13
1.3.4 Modeling Interactions with the Ambient Conditions.....................................................13
1.3.4.1 Radiation....................................................................................................... 13
1.3.4.2 Ambient Air Movement....................................................................................13
1.3.5 Glass Doors Fogging and Defogging.................................................................................14
1.3.6 Humidification............................................................................................................. 14
1.4 CFD Codes......................................................................................................................................... 15
1.4.1 Methodology................................................................................................................ 15
1.4.1.1 Preprocessing................................................................................................ 15
1.4.1.2 Solving....................................................................................................................16
1.4.1.3 Postprocessing.............................................................................................. 16
1.4.2 Turbulence Models................................................................................................................16
1.4.3 Mass Transfer.............................................................................................................. 16
1.4.4 Validation..................................................................................................................... 17
1.5 Conclusions.............................................................................................................................. 18
References......................................................................................................................................... 18
3
4 Computational Fluid Dynamics in Food Processing
1.1 INTRODUCTION
In retail stores, refrigerated cabinets are used to display perishable food. For merchandizing pur-
poses, the main function of such equipment is the effective display of products to make them visible
and easily accessible to customers. At the same time, food should be maintained at the prescribed
temperature and preserved from radiant heat. The safety and quality of perishable foodstuffs are
strongly affected by inappropriate storage temperature and by uneven temperature fluctuations,
which are regrettably encountered to a large extent in display cabinets [1–3]. For this reason, from
the point of view of storage conditions, retail cabinets are considered to be one of the weakest links
in the cold chain, and only the typical short residence time of food in such appliances reduces
the risk of quality loss. It is therefore essential that the efficacy of retail cabinets in terms of food
preservation is improved. The people who can play an important role in this improvement are the
manufacturer, the person in charge of installation and maintenance, and the shop manager.
The cabinet is certified by the manufacturer to comply with the testing standards currently in
place for a specified climate class defined by ambient temperature and relative humidity. Actually,
the performance of display cabinets in terms of food temperature is strongly affected by ambient
conditions, particularly air velocity and direction, and radiative heat load [4–7]. For this reason, par-
ticular care is necessary for the installation, and furthermore, accurate maintenance and operation
are essential to accomplish correct food storage conditions.
The manufacturer is the only person who can take advantage of Computational Fluid Dynamics
(CFD). Thus, the contents of this chapter will mainly focus on problems related to the design of
retail cabinets, giving only some suggestions about installation and proper use.
1.2 THE RETAIL CABINET

The main features of a retail cabinet can be summarized with the following statements:
• Food should be displayed in the most efficient way to promote selling;

• Correct food storage temperature should be ensured, with temperature fluctuations reduced
as much as possible.
The preservation and the display functions are contrasting requirements, because the best way
to protect the product from temperature fluctuations is to protect it from the shop environment, thus
keeping it out of sight of customers. Furthermore, the manufacturer must operate an optimization
process aiming to fulfill another important requirement, which is low energy consumption. In fact,
while retaining the preserving function, a better display function usually requires higher energy
consumption. Low energy consumption is becoming increasingly important, and the manufacturer
should make huge efforts to comply with all these requirements.
1.2.1 Classification
Retail cabinets are classified according to various criteria [4–6,8]. Among them, the most important
are the storage temperature and the cabinet geometry, which are the key factors for the choice of
the most suitable unit. Another important classification can be made according to the kind of air
circulation, which is crucial for certain food. Finally, a further classification can be made according
to energy consumption.
1.2.1.1 Storage Temperature
Retail cabinets are intended to host almost every kind of perishable food, from frozen food at −18°C
to some kinds of fruit and pastry at +10°C. For this reason they are usually classified as “low tem-
perature” cabinets in the case of storage and display of frozen food, and “medium temperature”
CFD Aided Retail Cabinets Design 5
cabinets in the case of storage and display of chilled food. A more detailed classification is defined
in the testing standards currently in place and will be discussed in the section 3.II.A.
1.2.1.2 Geometry
As regards geometry, retail cabinets can be [4]:
• Closed (in the presence of doors or sliding covers/glasses) or open;

• Vertical multideck (Figure 1.1), horizontal single deck (Figure 1.2) or horizontal serve-over
counters; or
• A combination of these (e.g., a horizontal open-top cabinet combined with a vertical multi-
deck closed cabinet).
Of course, the various geometries of cabinets are not suitable for all foods and temperatures. As
an example, open cabinets are not suitable for frozen food, due to heat infiltration from the ambient;
FIGURE 1.1 Vertical, multi-deck display cabinet.

FIGURE 1.2 Horizontal display cabinet.
if it is necessary to use open cabinets, then horizontal open-top cabinets are to be preferred, while
vertical open cabinets are to be avoided.
1.2.1.3 Refrigeration Equipment
Display cabinets are designed to maintain the correct storage temperature of food. Therefore, they
are not capable of reducing the temperature of products if they are too warm when loaded [4,6].
However, the cabinets’ refrigeration equipment is forced to perform heavy duty because of the
huge amount of heat due to air infiltration, radiative heat transfer and product manipulation by the
customers.
Depending on the refrigerating equipment, display cabinets can be classified as incorporated
condensing units and remote condensing units.
In the cabinets with incorporated condensing units, also named “stand alone,” the whole vapor
compression refrigerating equipment is contained within the cabinet, which only needs a power
supply connection and a drainage piping.
In the remote condensing units the cabinet is connected to a refrigerating unit that usually sup-
plies several cabinets, both at low and medium temperatures. The remote condensing units can be
further distinguished depending on the refrigerating system as a compression-type refrigerating
system or an indirect type-refrigerating system.
In the cabinets of the first category only the expansion valve and the evaporator are contained
in the cabinet, which is fed with liquid refrigerant from a centralized refrigerator. Because the
compression-type refrigeration system is complex and affords limited flexibility under changes to
supermarket layout, it is typically used only in medium to large size stores. However, this configura-
tion is preferable because of its enhanced energy efficiency.
The recent issue with halocarbon refrigerants is pushing interest toward employing systems of
the second category, in which a secondary refrigerant circulating system is installed between a
central refrigerating system (usually placed in an outbuilding) and the cabinets. This configura-
tion reduces dramatically the amount of refrigerant circulated [9] and allows the use of toxic or
flammable refrigerants with lower environmental impact (e.g., ammonia) at the expense of a more
complicated circuitry [10]. This secondary refrigerant can be a single-phase mixture of water and
antifreeze additives or a two-phase fluid like ice slurry or carbon dioxide. The reduction of the
amount of refrigerant circulated is also pushed by the issue of environmental impact of refrigerants
[11,12]. For this reason, new low-GWP (Greenhouse Warming Potential) are being adopted both in
incorporated and remote condensing units, leading to the need for a re-design of the evaporators. In
this case, CFD can be effectively employed in the design of such heat exchangers.
1.2.1.4 Air Circulation
A further classification can be made with reference to air circulation inside the cabinet. Cold air
distribution can be ensured by forced or natural circulation, the choice depending mostly on the
kind of foodstuff. As a general rule, forced air circulation is preferable because it is much more
effective in transferring the refrigerating power, thus enabling the correct operation of almost every
kind of display cabinet. Natural air convection should be preferred for the display of unwrapped
sensitive food like meat, pastry and ice cream, where water loss coupled with heat transfer on the
food surface can give rise to significant quality damage due to dehydration. It is mostly used on
horizontal serve-over units, where air stratification helps reduce warm air infiltration inside the
load volume.
1.2.1.5 Energy Consumption
Supermarkets are intensive users of energy in all countries. Electricity consumption in large super-
markets represents a substantial share (about 4%) of the national electric energy use in the United
States and in France. A large part of this consumption, varying from 50% to 70%, is due to air con-
ditioning and refrigeration cases [13,14]. In the United States, typical supermarkets with approxi-
mately 3700 to 5600 m2 of sales area consume about 2 to 3 million kWh annually for total store
energy use [15]. The national average electricity intensity (the annual electricity use divided by the
size of the facility) of a grocery store in the United States is about 565 kWh m−2 per year [1,16],
and 400 kWh m−2 per year for Europe [13]. These figures are a real challenge for energy savings,
and the supermarket chains are spending a substantial proportion of money on the yearly energy
consumption compared to the investment costs. For this reason the evaluation and certification of
the energy consumption of display cabinets is becoming an essential step in the future development
of such equipment. Legislation is considering this issue, and in many countries laws are in force for
a compulsory energy labelling of display cabinets.
1.2.2 Standardized Temperature Tests

Various testing standards for retail display cabinets are currently in place (e.g., the EN Standard
23953 [17,18] in Europe and the ASHRAE Standards 72 [19] in the United States). The objective of
such standards is to specify “requirements for the construction, characteristics and performance
of refrigerated display cabinets used in the sale and display of foodstuffs,…. to specify test condi-
tions and methods for checking that the requirements have been satisfied, as well as classification
of the cabinets, their marking and the list of their characteristics to be declared by the manufac-
turer” [18].
Apart from the requirements about construction, the main scope of these standards includes clas-
sifying retail cabinets as a function of their storage temperature and giving instructions for measur-
ing their energy consumption.
Specific conditions for the “temperature test” are thus defined in the standards. The cabinet is
loaded with packages made of a specified composition of water, cellulose and additives. It is placed
in a test room (Figure 1.3) where air temperature and humidity, air velocity and radiant heat are
controlled. The temperature of a certain number of “measure packages” is recorded over a period
of 24 hours after having reached steady state. Various temperature classes are identified, depending
on the load temperature measured during such tests.
As an example, the European standard identifies the temperature classes through the definition
of the highest and lowest temperatures of the warmest and coldest packages, as reported in Table 1.1
[18]. The test room conditions are identified through the definition of the ambient psychrometric
conditions (climate classes), as reported in Table 1.2 [18]. For all of them, air velocity and radiant
heat are the same. In particular, air velocity shall lie between 0.1 and 0.2 m/s parallel to the plane
of the cabinet display opening and to the longitudinal axis. Radiant heat shall be controlled by the
definition of the wall temperature and emissivity, and the level of illumination.
FIGURE 1.3 A display cabinet during a standardized temperature test.
TABLE 1.1
Temperature Classes According to the EN 23953 Standard
Highest Temperature of the Lowest Temperature of the Lowest Temperature of the
Class Warmest Package (°C) Coldest Package (°C) Warmest Package (°C)
L1 −15 – −18
L2 −12 – −18
L3 −12 – −15
M0 +4 −1 −
M1 +5 −1 −
M2 +7 −1 −
H1 +10 +1 −
H2 +10 −1 −
S Special classification
It should be noted that compliance of a retail cabinet with the standards does not mean that the
correct storage temperature will be kept during normal operation in the retail store, even if the cabi-
net has been set up and situated in accordance with the recommendations of the manufacturer (“nor-
mal conditions of use”). This is mostly due to the dissimilar thermal properties of foodstuffs and test
packages, and to some differences in ambient conditions (i.e., air velocity, air temperature, radiative
heat load). The EN standard 23953-2 [18] highlights this argument in the Annex B “Comparison
between laboratory and in-store conditions.” It states thus: “The complete range of various climate
TABLE 1.2
Climate Classes According to the EN 23953 Standard
Test Room Dry Bulb Temperature Relative Humidity
Climate Class (°C) (%)
0 20 50
1 16 80
2 22 65
3 25 60
4 30 55
5 40 40
6 27 70
7 35 75
8 24 55
conditions and various ways of loading in stores cannot be simulated in the laboratory. For these
reasons, specific climate classes and loading are defined for tests in the laboratory to classify cabi-
nets and to make comparisons. For open refrigerated display cabinets, test results in laboratory
cannot be directly transposed in stores.”
For this reason it is crucial that the cabinet is installed and operated with awareness. For the
same reason, designers can make great use of CFD because they can check different configurations,
thereby saving the huge amount of time required for the standardized tests (at least a couple of days
each), and they can predict the performance of the apparatus at operating conditions different from
the standardized ones.
1.2.3 Air Curtains
In an open cabinet, refrigerated air curtains are established at the cabinet opening when cooling of
the load is achieved via forced air distribution. The reason for the choice of this position is the need
to reduce heat transfer from the external environment by creating a barrier between the load volume
and the external ambience. Heat transfer through the solid walls surrounding the load volume can
be effectively cut by means of adequate insulating material. On the contrary, load is subject at the
opening to both radiative and convective heat transfer from the ambient. Both cause heating of the
food surface.
Radiative heat transfer takes place between the load surface and the room walls, lights and all
other objects surrounding the opening. It plays an important role, since emissivity of the food pack-
aging and of the surrounding objects is usually high (in the range 0.8–0.9). It has been measured
that the temperature of the exposed surface of frozen food can increase up to 5–10 K due to the
absorption of radiant heat. Radiant heat can be reduced by using low emissivity materials for food
packaging, high efficiency (low temperature) lights in the environment, and shielding coatings on
the glass door surfaces, if any [20].
Convective heat transfer is due to the temperature difference between the load volume and the
environment. Air movement caused by natural convection unavoidably causes infiltration of warm
air through the opening, which is enhanced in the presence of even slight air movements in the
room. The air curtain is capable of effectively restraining the convective heat transfer and the warm
air infiltration, in the meantime reducing the surface heating due to radiation.
In the case of open cabinets, more than one air curtains are used, the temperature of the exter-
nal air curtain being higher than that of the internal one for the sake of a better flow stability. Air
from the curtains is then extracted through a grille and forced by fans through a finned cooling coil
where heat is removed. The surface temperature of the cooling coil is usually below the dew point
temperature of air; thus water condensation takes place. In the frequent case of surface temperature
of the coil below 0°C, frost formation takes place, which requires cyclical coil defrosting operations.
Finally, refrigerated air is supplied to a plenum and then through a honeycomb to the supply grille,
thus creating the air curtain or curtains.
In the case of closed cabinets, usually only one air curtain is established, which flows close to the
internal surface of the door. When the door is open, the air curtain helps prevent warm air infiltra-
tion. When the door is closed, the air curtain extracts heat from the load volume and particularly
from the food surface, which is still subject to radiant heat. Air curtains will be discussed more in
detail in the section 3.III.B.1 and in Chapter 2.
1.3 APPLICATIONS OF CFD TO DISPLAY CABINETS

CFD is a successful tool for the designer of display cabinets, who can take advantage of this tool to
improve load temperature distribution, predict the flow pattern of air and its efficacy, reduce warm
air entrainment and improve product refrigeration [21]. As previously mentioned, the main concern
arises from the necessity to ensure an effective display, while preserving the optimal storage condi-
tions and achieving the lowest energy consumption. The manufacturer must face this challenge and
find the solution that best fits such requirements. Air movement inside the cabinet plays the key role
in this challenge, essentially because it is in charge of product refrigeration. Thorough comprehen-
sion of the phenomena associated with airflows in display cabinets is itself a difficult task, due to
the various interdependent factors that act simultaneously. Often a trial and error process has to be
established, requiring numerous experimental tests that entail spending a huge amount of time and
money. Numerical modeling performed by a skilled person can be a viable alternative, once its reli-
ability has been validated against experimental data. Sensitivity analyses can be easily executed,
efforts can be directed to optimize the most critical components, and improvements in the per-
formance of the whole cabinet can be achieved in a much shorter time than through experimental
testing. Furthermore, the performance of the cabinet at different ambient conditions (e.g., tempera-
ture, humidity, air velocity) or operating conditions (e.g., load arrangement) can be predicted with
sufficient accuracy, thus leading to a better awareness of the possible performance of the unit in the
actual conditions at the retail store.
In the following, the most important applications of CFD to retail cabinets are briefly described
and the possible advantages of numerical modeling are discussed.
1.3.1 Modeling Product Temperature Distribution

Product temperature inside a display cabinet suffers from an uneven distribution, both from left
to right and from back to front of the shelves. Temperature differences up to about 5 K for chilled
food and about 10 K for frozen food can be encountered, which can be unacceptable. The differ-
ence from the left to the right side is often due to uneven air distribution, and will be discussed
later on. The main reasons for the difference from back to front are the proximity of the cooling
coil to the back/bottom of the load volume and the effect of radiant heat on the front surface.
Radiant heat can account for up to 12% [22] of the total load, and therefore cannot be neglected.
Furthermore, as radiation is concentrated on the surface, it leads to a significant local temperature
increase. Experimental tests [23] showed a reduction of up to 10 K of the surface temperature of the
upper layer of products in a horizontal frozen food cabinet thanks to the application of a low emis-
sivity shield. This is because food packaging has an average emissivity of about 0.9, which is also
the value required by the EN standard for the test packages.
The main problem when simulating the product temperature distribution resides in the necessity
to adopt a transient state model. In fact, heat exchanges through radiation and with the cooling coil
are both time-dependent phenomena, linked to the shop opening time, to the presence of covers or
night curtains, to the on-off cycling of the refrigerating equipment and to the defrosting operations.
FIGURE 1.4 Temperature of the warmest and of the coldest packages during a temperature test.
Figure 1.4 reports as an example the temperature of the warmest and the coldest test packages
measured at different locations in a frozen food cabinet. In addition to the influence of the location,
temperature fluctuation due to the defrosting operations is also clearly visible. The defrosting opera-
tions can be performed by heating the coil or simply by switching off the refrigerating equipment.
The missed refrigerating power leads to a step increase in food temperature, which requires a few
hours to completely recover.
For this reason, modeling the product temperature distribution inside a cabinet should be per-
formed via a transient state simulation adopting a reliable model of radiation. In this case, it is not
convenient to model concurrently the airflow pattern in the air curtains, because of the great dispar-
ity between the time constants of the two phenomena. A satisfactory transient model of the airflow
should require a time step of a few hundredths of a second, whereas for the food temperature a few
minutes could be enough. When modeling the load temperature distribution, the air curtain could
be simulated as a convective boundary condition with an average convective coefficient evaluated
by means of previous simulations or through classical correlations. In this case, the CFD model
becomes a much simpler coupled conduction-convection model that can also be easily solved with
in-house codes [23,24].
1.3.2 Modeling Airflow
From the point of view of the designer, the effective simulation of the airflow pattern inside the
whole display cabinet is the most interesting result. However, there are too many factors regarding
the geometry of the cabinet, the operating conditions of the cabinet, and the ambient conditions
thatinteract and influence the performance of the unit. This would require a very complex model
with an almost unpredictable accuracy that probably could not fit the actual operating conditions
well [25].
A solitary paper is present in the literature with a simulation of a whole vertical open chilled cab-
inet in 3D, with the purpose of obtaining steady-state temperature distribution in the product [26].
The whole cabinet, including the load, the air curtains, the air ducts and a portion of the ambient,
is included in the domain, and the authors claim reasonable agreement with the load temperature
distribution obtained from spot thermocouple measurements and infrared camera images.
It is much more effective to split the whole flow course into a few sections and set up simpli-
fied models where some variables can be disregarded, after verification by means of a sensitivity
analysis. With such models it will become almost impossible to closely reproduce the operation of
the whole cabinet; however, they will be much more effective for a quick comparison of various
configurations [25,27–29].
The most widely used simplified models relate to the air curtain, the air distribution between two
shelves, the airflow at the evaporator and the airflow in the rear ducts. In all of these, load surface
can be considered as an adiabatic surface at the storage temperature. The effect of radiation on the
load surface can be considered through a suitable increase in the food surface temperature.
Usually simulations are performed for a cabinet fully loaded, because this operating condition is
required for the standard test and is the most common in supermarkets.
1.3.2.1 Air Curtains
The function of the air curtain has already been introduced, and will be further detailed in Chapters
4 and 5. It is indeed the most investigated part of the display cabinet, because of its crucial influence
in the performance of the unit, in terms of both product temperature and energy consumption. This
is mostly due to the warm air infiltration, which accounts for 60% to 75% of the total refrigeration
load [4,30]. In fact, as soon as the air curtain leaves the air discharge, entrainment of warm air takes
place due to the increase in the width of the curtain. Because of the necessity to respect the mass
balance, a portion of the flow rate will be lost at the air return and will overspill at the bottom of the
cabinet, thus causing the so-called “cold feet effect.”
The induction factor α is defined for a single air curtain as the ratio of the mass flow rate of ambi-
ent air entrained to the total mass flow rate at the return grille [4,29].
m ambient t −t
α= ≅ return discharge (1.1)

mreturn tambient − tdischarge

It can be computed also in the case of multiple air curtains, through the calculation of the average
values of air temperature weighed with the respective mass flow rates and of back panel flow [31].
The induction factor is commonly considered for the evaluation of the effectiveness of the
air curtain, especially because the variables required for its calculation are easily measurable.
It has been found to depend upon several factors, among which the most important are the
initial turbulence intensity, the Reynolds number and the velocity profile at the air discharge
[32].
Early simulations of the air curtains were performed using in-house codes in 2D domains. As an
example, Cortella [33] utilized a finite element code based on the streamfunction-vorticity formula-
tion, with a turbulence model similar to an LES procedure. Transient simulations were performed
in a 20,000 grid points domain of a vertical open cabinet for chilled food. The induction factor and
the refrigerating power were found to be in good agreement with experimental values, and some
suggestions could be given to enhance curtain stability. Ge and Tassou [34] also used an in-house
code based on finite differences, which took into account the moisture content in the air curtain. The
authors derived from this model some correlations for the estimation of the heat flow rate and of the
return air temperature at various conditions.
The employment of commercial codes made simulations much easier, because the main prob-
lems of computational efficiency and user friendliness of in-house codes were overcome. In the last
decade many authors published 2D simulations of air curtains in horizontal and vertical cabinets,
with the aim of predicting the air flow pattern and evaluating the curtain efficacy at various condi-
tions [29,35–41,42], including the airflow from the back panel [43]. A 3D simulation of air curtains
has been developed by D’Agaro et al. [29] and will be discussed more in detail in 3.3.4.
In all the studies mentioned here, CFD has proven to be a successful tool for the optimization of
the air curtain.
1.3.2.2 Shelves
In open vertical cabinets, air curtains are deeply influenced by the shape, length and loading of
shelves.
In regard to loading, the best air curtain efficacy is encountered when the cabinet is fully
loaded, which is the standard test and design condition and the most common condition in
retail stores. In regard to geometry, at the design stage a lot of experimental tests or CFD
simulations are required to minimize air curtain disruption at the front of the shelves [44].
Some interest is also raised from the employment of air guiding strips (airfoils) at the edge of
the shelves to guide the air curtain and reduce the disruption effect [45]. They showed to be
effective to a large extent with some geometries and are worth further investigation through
CFD simulations.
1.3.2.3 Evaporator and Rear Ducts

Uneven distribution of air at the evaporator ducts is crucial because it could lead to an uneven air
curtain velocity at the air discharge, thus causing differences in product temperature from left to
right in the load volume. Foster et al. [44] investigated the flow of air as it exited the evaporator and
entered the rear duct. The effects of a dead space were identified and modifications were suggested
in order to reduce the formation of vortices and to improve air distribution in the back plenum.
Similar simulations can be performed on different geometries of cabinets, and not only the shape
of the rear ducts but also the position of the evaporator and of the fans can be investigated [46–51].
More details are given in Chapter 4.
1.3.3 Modeling the Influence of Air Humidity

Commercial codes give the user the opportunity to include moisture content in the air flow
models, in order to investigate both heat and mass entrainment in the cabinet. Some authors did
use this feature when performing their simulations [26,34,35]. Actually, presence of humidity
is crucial for the cabinet performance, because humidity entrainment leads to performance
detriment due to evaporator frosting. However, the increase in computer power requirements
due to the inclusion of moisture content in the CFD model can be avoided by estimating the
latent heat from a mass balance on the water vapor content of the air curtain, once the induc-
tion factor and the humidity ratio of the ambient air are known, and assuming that air is
saturated at the evaporator outlet. More details about the numerical methodology are given in
Section 3.4.3.
1.3.4 Modeling Interactions with the Ambient Conditions

Display cabinets operation, especially for open cabinets, is crucially influenced by ambient condi-
tions [52]. Radiant heat, ambient air velocity and direction are the most important variables that
must be considered when designing such units.
1.3.4.1 Radiation
In the previous section the effect of radiant heat has been described, and some suggestions have
been given on how to reduce radiant heat gain on the load surface. In regard to simulations, it has
been clarified that radiation has to be accounted for only when simulations of the load temperature
distributions are being performed, whereas it is unnecessary for the evaluation of airflow patterns
when load surface can be considered as a constant temperature surface.
1.3.4.2 Ambient Air Movement

Air velocity in the ambient environment and its direction are also crucial for the performance of
the cabinet, and for this reason standard tests prescribe both of them. Furthermore, air flow visu-
alizations performed on the air curtains during cabinet testing showed that 3D effects take place
and can be significant, even with still air in the ambient. For these reasons, we realize that it is
necessary to investigate more thoroughly the 3D effects in the air curtains. Typically almost all
papers in the literature describe 2D simulations, for the sake of CPU time and memory requirement
reduction, assuming that simulations are performed on the median section and that end effects can
be negligible. In fact, especially for short length cabinets, end effects can be significant, and lead
to uneven air curtain velocity and food temperature distribution. D’Agaro et al. [29] performed 3D
simulations on a 2.44 m long vertical cabinet for frozen food and investigated the effect of longi-
tudinal air movement. The authors report that 3D flow structures that may originate from slow air
movements in the ambient are responsible for 20% decay in the performance of the unit between
a 2 m and a 1 m long cabinet, thus underlining the importance of 3D simulations in the design of
short to medium cabinets.
Another important interaction between the cabinet and the ambient is the accumulation of cold
air that overspills the return air grille and accumulates on the floor in front of the unit. This situation
is named “cold feet effect” because of the unpleasant sensation on the customer who walks close to
vertical open cabinets. Some authors [53] tried to simulate the whole sales area of a store, but the
model was too complex and time-consuming. For this reason they moved to a simplified model of
a chilled aisle, simulated on the three symmetry planes, and investigated different ventilation and
heating strategies. More details on the topic are reported in Chapter 4.
1.3.5 Glass Doors Fogging and Defogging

On closed display cabinets with doors, mist deposition occurs on the internal side of the door each
time it is open, especially for frozen food cabinets. In the case of transparent doors, a quick defog-
ging must be achieved to recover product visibility through the glass. For this purpose, an air cur-
tain is established flowing along the internal glass surface, and an electric heater is sometimes
embedded in the multiglazed door. Demisting time can be estimated through the application of
CFD. D’Agaro et al. [54] coupled a CFD commercial code with an in-house code for the evaluation
of the air flow pattern and of the water layer evolution, respectively. More details on the numeri-
cal procedure are given in Section 3.4.3. The computation enables the prediction of the water layer
height during condensation when the door is open and during evaporation in the presence of electri-
cal heaters. Various parameters both of the refrigerated cabinet (e.g., geometry, air curtain velocity
and temperature) and of the glass door (e.g., geometry, global heat transfer coefficient, and presence
of electrical heaters) can be considered. The model also takes into account condensation as a thin
film or a collection of water droplets. Results showed the model to be reliable for the evaluation of
the entity of defogging time reduction that might be expected with different solutions.
Mist formation can also occur at the external side of the glass door at high relative humidity
ambient conditions. Usually electrical heaters are employed to limit the occurrence by increasing
the glass surface temperature. More recently some investigations considered the chance to take
advantage of the air conditioning supply flow to heat up the glass without electrical heaters.
1.3.6 Humidification
Unwrapped food products like fruits and vegetables are subject to dehydration when displayed
in open cabinets. This is due to air dehumidification that takes place on the cooling coil surface,
especially when forced convection cabinets are employed. The importance of relative humidity on
the shelf life of products is well known, and all efforts are made to limit quality decay and weight
loss when possible. In display cabinets, air dehumidification can be limited by choosing appropri-
ate air velocity and temperature at the design stage; however, the solution is not fully satisfactory.
Another approach is derived from the air conditioning plants, and from some vegetable refriger-
ated storage rooms, where air humidification is sometimes exploited through water spray. This
technique has the disadvantage of possible bacteria growth, and therefore requires strict control
from the microbiological point of view. Tests have been performed on humidification equipment
that uses water mist sprayed over the products’ surface above each shelf and the well [55]. Results
were encouraging, since weight loss was reduced at the expense of a slightly higher refrigeration
load requirement.
This device is already commercially available and supplied upon request by display cabinet
manufacturers; however, it has not yet been thoroughly investigated. Commercial codes are avail-
able to treat heat and mass transfer, and some research has been performed in the similar field of
weight loss during blast chilling [56]. Thus in the near future some research work will be probably
devoted to this interesting topic.
1.4 CFD CODES
Computational Fluid Dynamics (CFD) is based on the solution of the governing flow (i.e., the con-
tinuity and the Navier–Stokes) equations of the energy conservation equations, and sometimes on
the conservation of other factors (e.g., water moisture). It has become popular only recently, when
the availability of more powerful and affordable computers made it possible to investigate practical
problems that were previously too computationally expensive.
When applied to display cabinets, CFD can model fluid flow, conductive, convective and radia-
tive heat transfer, and moisture transfer.
In addition to numerous in-house codes, there are a number of commercial codes that can now
cope with a high level of complexity. However, most of them are general purpose software designed
for use in many different research fields. Therefore, sometimes robustness is enhanced to the detri-
ment of accuracy, which still needs to be improved [57].
1.4.1 Methodology
Every CFD simulation can be split into three consecutive phases: preprocessing, solving, postpro-
cessing [57].
1.4.1.1 Preprocessing
Preprocessing starts with the choice of the computational domain to be simulated, includes the
mesh generation and the definition of material properties, and ends with the application of bound-
ary conditions. It is a crucial phase for obtaining reliable results. The users must be fully aware of
the physics of the problem, because in this phase they set up the model of their practical problem.
The choice of the domain to be investigated needs to be carefully considered in order to include
all possible effects on the object of investigation. As an example, CFD simulation of an open display
cabinet requires a portion of the external ambient to be included in the computational domain, in
order to evaluate correctly the warm air entrainment and the cold air overspill.
Two dimensional or three dimensional domains can be considered for display cabinets. Until
now, almost all the simulations of air curtains have been performed in 2D, because of the neces-
sity to reduce the computational load. However, D’Agaro et al. [29] have shown that in short-length
cabinets the end effects cannot be disregarded, thus demonstrating that in certain cases 3D simula-
tions could be necessary. For the simulation of air ducts (e.g., rear ducts, evaporator, and fans) 3D
simulations are usually indispensable, because of their complicated geometry.
The dimension of the elements in the grid influences the level of accuracy of the solution. Usually
the dimension must be reduced in the portions of domain where an accurate solution is required (e.g.,
in the presence of turbulence or of high velocity gradients) or close to solid boundaries, where the
requirements of the turbulence model must be satisfied. Of course the computational time increases
with the number of elements in the domain, thus suggesting the limiting of grid refinements to those
areas where it is strictly necessary. For an accurate simulation, it should be necessary to check
that the solution is not “grid dependent” (i.e., it does not depend on the dimension of the elements).
Otherwise, mesh refinements are still required.
Once all the properties of the various solid and fluid substances have been identified, the bound-
ary conditions must be defined. This is another important step in the simulation, theeffect of which
on the solution can be significant. Usually it is necessary to introduce assumptions at this step,
because the available boundary conditions rarely match the actual conditions. In many cases, differ-
ent boundary conditions must be checked, the results compared and a sensitivity analysis performed
for the identification of the best conditions. More specific information about different boundary
conditions used to simulate display cabinets is reported in Chapter 4.
1.4.1.2 Solving
The solution of the governing equations requires their discretization and an iterative process
to obtain an approximation of the value of each variable at specific points in the domain.
Calculation is stopped when the residuals in the calculation of the balance of one or more prop-
erties are below a specified value, and the solution is said to converge. Reaching convergence is
not trivial, and the choice of the threshold value of residuals is not easy. Some suggestions are
given in Chapter 4.
Another main concern about solving in CFD simulation is the choice between steady-state or
transient calculations. The choice depends on the phenomena to be investigated, and some sugges-
tions have been given in the previous sections depending on the object of the simulation. In general,
transient simulations should be performed only in the case where the time evolution of a phenom-
enon is under investigation, because convergence must be reached at each time step and therefore
they are much more time-consuming.
1.4.1.3 Postprocessing
When the solution reaches convergence, a distribution of the values of all the variables throughout
the whole domain is produced. Such values must be processed to obtain visualizations and some
require numerical results (e.g., the induction factor, the refrigerating power, etc.). The postprocess-
ing phase is thus essential for the evaluation of the simulation results and an important tool for their
most thorough understanding. In fact, the postprocessor also performs calculations and balances,
thus giving further precious information.
1.4.2 Turbulence Models
Turbulence models must be adopted to take into account the turbulence effects, which cannot be
evaluated through a direct simulation. In fact, direct simulation of turbulence in large domains
as those used for display cabinets would require a really huge amount of memory and CPU time.
There are many turbulence models available, and unfortunately the choice of the model signifi-
cantly affects the results. Literature can be helpful in this respect, but experience is fundamental.
The basic turbulence models are the so-called “two equations models,” which are the default
choice for many commercial codes. Among these, the k-ε and the RNG k-ε models are the most
widely used. Although easy to implement, they require the previous evaluation of the turbulence
kinetic energy and dissipation rate, which is a matter of difficult measurements or experience.
Furthermore, they are not considered as the best choice because of the poor accuracy sometimes
encountered.
Other models are those based on the Reynolds stresses and the Large Eddy Simulation (LES).
The former was found to be accurate [48] even if it required a quite fine mesh. The latter was suc-
cessfully used by Cortella et al. [33] in the framework of a streamfunction-vorticity in-house code.
Some more suggestions, particularly on the initial turbulence intensity, will be given in Chapter 4,
while a deeper discussion of turbulence models is left to specific literature.
1.4.3 Mass Transfer
It has already been pointed out in Section 3.3.3 that the moisture content of the air is a critical factor
influencing the performance of display cabinets, because humidity entrainment involves an addi-
tional latent heat that shall be removed at the cooling coil. Furthermore, also in the case of chilled
cabinet, the surface temperature of the cooling coil is often below 0°C, thus leading to frost forma-
tion and the need for cyclical defrosting operations.
The evaluation of the latent heat does not strictly require a coupled heat and mass transfer simu-
lation. The induction factor can be once again considered from the evaluation of the water mass
balance in the air curtain
m ambient x − xdischarge
α= ≅ return (1.2)

mreturn x ambient − x discharge

where x is the humidity ratio (kgvap/kgdry air). The ambient humidity ratio is known, while the dis-
charge humidity ratio can be easily estimated by assuming saturation for the discharge air [2,29].
On the contrary, there are some particular cases where coupled heat and mass transfer must be
simulated, as in the simulation of door fogging and defogging in closed display cabinets. This topic
has already been introduced in Section 3.3.5. In the following, more details are given on the numeri-
cal procedure, which involves coupling of two different codes for the dry heat and flow and for the
mass transfer [54]. The problem is split into two phases, the first regarding the dew deposition on the
internal face of the glass when the door is open, and the second regarding the defogging operation
once the door is closed.
In regard to the heat transfer, the evaluation of the local heat transfer coefficient for the air flow
on the internal face of the glass is performed with a single phase solver. In fact, the phase change
during evaporation is limited to a thin layer on the solid surface, and therefore it is accounted for
through proper boundary conditions. Conductive heat transfer through the multiglazed glass can be
solved by means of a steady state network of thermal resistances, which can also take into account
radiation in the air cavities, thus leading to a very simple conduction problem easily solvable with
simple in-house codes. Finally, the external heat transfer coefficient can be effectively estimated by
means of the classical empirical correlations for steady state natural convection on a vertical plate.
The different domains and solvers are coupled through an exchange of boundary conditions, thus
leading to a more flexible algorithm.
In regard to the mass transfer, an in-house code was used, where the latent heat contribution due
to condensation or evaporation appeared as a heat source or sink, respectively, placed at the inter-
face between the solid and the fluid domain. Thermal and mass balances can be established at the
interface and solved taking advantage of the heat and mass transfer analogy. A detailed description
of this procedure is given in [54]. It is interesting to note that in this model the water layer can be
considered as a continuous film or as a number of droplets, the geometry of which changes dur-
ing the condensation and evaporation processes. Furthermore, the effect of electric heaters can be
accounted for in order to speed up the demisting process. The simulation showed to be reliable when
compared to experimental tests, especially using the droplet model.
1.4.4 Validation
It is common opinion that CFD simulations must be validated, and this is especially true in the case
of display cabinets, where the simulation of such a complex problem requires a number of assump-
tions to set up the model.
Usually validation is performed against experimental tests in controlled conditions, like those
in accordance with the standards in force. The variables that can be compared are essentially load
temperature, air temperature and air velocity [59].
Measurements of load temperature can be easily performed, and the effect of radiative heat
transfer can be highlighted by means of infrared thermal imaging systems.
Measurements of air temperature can be easily performed at the discharge and return grille,
although it is quite complicated to measure air temperature along the air curtain. Infrared systems
cannot visualize air temperature, because it is transparent under infrared radiation. Some images
taken with infrared cameras and able to show air temperature are actually infrared visualizations
of the cabinet end wall temperature.
Accurate measurements of air velocity are difficult, especially along the air curtain. At the dis-
charge and return grille, the use of hot wire anemometers allows for a sufficiently accurate evalu-
ation of the velocity distribution, even if important information about direction and turbulence is
lost. Anemometers placed along the air curtain are susceptible to disturbing the airflow pattern,
thus providing an incorrect evaluation. Axell [60] reported a strong influence of the distance and
shape of the sensors, and claimed to have measured values in “good qualitative agreement” with the
numerical results.
Much more reliable results can be obtained using the Particle Image Velocimetry (PIV), which
is a very accurate method for flow pattern measurements. Air has to be seeded, the field of investi-
gation is lighted with a laser sheet and a number of subsequent images are taken by a digital video
camera placed perpendicular to the laser plane. Processing of the images allows for the identifica-
tion of the movement of each seeding particle, thus leading to the complete flow pattern recognition.
Rather than for CFD validation, this is a valuable tool for the adoption of the best CFD boundary
conditions, especially at the discharge air (air velocity and direction, turbulence intensity), which
are crucial to obtain reliable results.
1.5 CONCLUSIONS
Retail display cabinet design can take great advantage of CFD, in terms of both time and money
savings. It is rather difficult to replicate experimental results, due to flow complexity and to difficul-
ties in reproducing the actual ambient conditions. However, CFD is very effective for sensitivity
analyses, and can be very helpful to compare the performance at different operating conditions and
find the optimal design of the unit.
In the near future CFD will surely grow in popularity, and probably it will be much easier to
perform even complicated simulations, due to the increasing computational power. 3D simulations
will be more affordable, and larger computational domains will permit a more thorough evaluation
of the influence of ambient air, as an example.
Nevertheless, a lot of critical assumptions or choices must be made to perform a CFD simulation,
and the outcomes will always depend on the judgment of the operator. Furthermore, commercial
codes tend to improve robustness at the expense of accuracy every time convergence is difficult,
thus leading to incorrect results. For these reasons CFD must be always operated by skilled people,
the results accurately assessed and, whenever possible, boundary conditions and results of some
reference cases should be validated by comparison with experimental tests.
REFERENCES
1. Spiess W.E.L., Boehme T., Wolf W., Quality changes during distribution of deep-frozen and chilled
foods: Distribution chain situation and modeling considerations, In: Food Storage Stability, Taub I.A.
and Singh R.P., Eds, CRC Press, Boca Raton, FL, USA, 399–417, 1997.
2. Morelli E., Noel V., Rosset P., Poumeyrol G., Performance and conditions of use of refrigerated display
cabinets among producer/vendors of foodstuffs, Food Control, 26: 363–368, 2012.
3. Sergelidis D., Abrahim A., Sarimvei A., Panoulis C., Karaioannoglou P., Genigeorgis C., Temperature
distribution and prevalence of Listeria spp. in domestic, retail and industrial refrigerators in Greece,
International Journal of Food Microbiology, 34(2): 171–177, 1997.
4. Rigot G., Meubles et Vitrines Frigorifiques, Pyc Edition, Paris, 1990.
5. Gac A., Gautherin W., Le Froid dans les Magasins de Vente de Denreés Périssables, Pyc Edition, Paris,
France, 1987.
6. American Society of Heating, Refrigerating and Air Conditioning Engineers, Handbook 2006
Refrigeration, ch. 46, Retail Food Store Refrigeration and Equipment, ASHRAE, Atlanta, 2006.
7. Billiard F., Gautherin W., Heat Balance of an open type freezer food display cabinet, In: Proceedings of
the International Conference “Cold Chain Refrigeration Equipment by Design,” International Institute
of Refrigeration Comm. B1, B2, D1, D2/3, Palmerston North, New Zealand, 322–332, 1993.
8. IIR-International Institute of Refrigeration, Recommendations for the Processing and Handling of
Frozen Foods, IIR/IIF Paris, 1986.
9. Sharma V., Fricke B., Bansal P., Supermarket refrigeration system charge reduction using cascade sys-
tems, In: Proceedings of the 11th IIR Gustav Lorentzen Conference on Natural Refrigerants: Natural
Refrigerants and Environmental Protection, GL 2014, pp. 956–962, 2014.
10. Sawalha S., Perales Cabrejas C., Likitthammanit M., Rogstam J., Nilsson P.O., Experimental investiga-
tion of NH3/CO2 cascade system and comparison with R-404A system for supermarket refrigeration,
In: Proceedings of the 22nd International Congress of Refrigeration, IIR–IIF, Beijing, 2007.
11. Gullo P., Elmegaard B., Cortella G., Energy and environmental performance assessment of R744 booster
supermarket refrigeration systems operating in warm climates, International Journal of Refrigeration,
64: 61–79, 2016.
12. Gullo P., Cortella G., Theoretical evaluation of supermarket refrigeration systems using R1234ze(E)
as an alternative to high-GWP refrigerants, Science and Technology for the Built Environment, 22(8):
1145–1155, 2016.
13. Orphelin M., Marchio D., Computer aided energy use estimation in supermarkets, In: Proceedings of
the Building Simulation Conference, Prague, Czech Republic, 1997.
14. Arias J., Energy Usage in Supermarkets—Modelling and Field Measurements, Division of Applied
Thermodynamics and Refrigeration, Department of Energy Technology, Royal Institute of Technology
PhD thesis, 2005.
15. Baxter V., Advanced Supermarket Refrigeration/Heat Recovery Systems, Vol. 2, Country Reports,
I.H.P. Programme, Oak Ridge, USA, 2003.
16. Anon., Energy Star, Putting Energy into Profits, Guide for Small Business, Washington, D.C., 2003.
17. European Standard EN ISO 23953–1:2015 Refrigerated display cabinets; Part 1: Vocabulary.
18. European Standard EN ISO 23953–2:2015 Refrigerated display cabinets; Part 2: Classification, require-
ments and test conditions.
19. ASHRAE Standard 72–2014—Method of Testing Open and Closed Commercial Refrigerators and
Freezers.
20. Faramarzi R.T., Woodworth-Szieper M.L., Effects of low-emissivity shields on the performance and
power use of refrigerated display case, ASHRAE Transactions, 105(1): 533–540, 1999.
21. Smale N.J., Moureh J., Cortella G., A review of numerical models of airflow in refrigerated food appli-
cations, International Journal of Refrigeration, 29: 911–930, 2006.
22. Faramarzi R.T., Efficient display case refrigeration, ASHRAE Journal, 41: 46–54, 1999.
23. Bobbo S., Cortella G., Manzan M., The temperature of frozen food in open display freezer cabinets:
Simulation and testing, In: Proceedings of the 19th International Congress of Refrigeration, IIR/IIF,
The Hague, Netherlands, 697–704, 1995.
24. Comini G., Cortella G., Saro O., Finite element analysis of coupled conduction and convection in refrig-
erated transport, International Journal of Refrigeration, 18: 123–131, 1995.
25. Cortella G., CFD-aided retail cabinets design, Computers and Electronics in Agriculture, 34: 43–66,
2002.
26. Madireddi S., Agarwal R.K., Computation of three-dimensional flow field and heat transfer inside an
open refrigerated display case with an air curtain, In: Proceedings of IIR International Conference
Commercial Refrigeration, Vicenza, I, 2005.
27. Morillon C., Penot F., La modélisation: Une aide à la conception thermoaéraulique des meubles frigo-
rifiques de vente, Revue Générale du Froid, 968: 48–53, 1996.
28. Stribling D., Tassou S.A., Marriott D., A two-dimensional computational fluid dynamic model of a
refrigerated display case, ASHRAE Transactions, 103(1): 88–94, 1997.
29. D’Agaro P., Cortella G., Croce G., Two- and three dimensional CFD applied to vertical display cabinets
simulation, International Journal of Refrigeration, 29: 178–190, 2006.
30. Axell M., Fahlen P., Climatic influence on display cabinet performance, In: Proceedings of IIR
International Conference New Technologies in Commercial Refrigeration, Urbana, USA, Hrnjak P.S.,
Ed., 181–190, 2002.
31. Yu K., Ding G., Chen T., A correlation model of thermal entrainment factor for air curtain in a vertical
open display cabinet, Applied Thermal Engineering, 29: 2904–2913, 2009.
32. Navaz H.K., Henderson B.S., Faramarzi R., Pourmovahed A., Taugwalder F., Jet entrainment rate in air
curtain of open refrigerated display cases, International Journal of Refrigeration, 28: 267–275, 2005.
33. Cortella G., Manzan M., Comini G., CFD simulation of refrigerated display cabinets, International
Journal of Refrigeration, 24: 250–260, 2001.
34. Ge Y.T., Tassou S.A., Simulation of the performance of single jet air curtains for vertical refrigerated
display cabinets, Applied Thermal Engineering, 21: 201–219, 2001.
35. Van Oort H., Van Gerwen R.J.M., Air flow optimisation in refrigerated display cabinets, In: Proceedings
of the 19th International Congress of Refrigeration, IIR/IIF, The Hague, Netherlands, 446–453, 1995.
36. Baléo J.N., Guyonnaud L., Solliec C., Numerical simulation of air flow distribution in a refrigerated
display case air curtain, In: Proceedings of the 19th International Congress of Refrigeration, IIR/IIF,
The Hague, Netherlands, 681–688, 1995.
37. Laguerre O., Moureh J., Srour S., Derens E., Commere B., Predictive modelling for refrigerated display
cabinets, In: Proceedings of IIR International Conference Advances in the Refrigeration Systems, Food
Technologies and Cold Chain, Sofia, BG, 480–487, 1998.
38. Cortella G., Manzan M., Comini G., Computation of air velocity and temperature distributions in
open display cabinets, In: Proceedings of IIR International Conference Advances in the Refrigeration
Systems, Food Technologies and Cold Chain, Sofia, BG, 617–625, 1998.
39. Wu Y., Xie G., Chen Z., Niu L., Sun D.-W., An investigation on flowing patterns of the airflow and its
characteristics of heat and mass transfer in an island open display cabinet with goods, Applied Thermal
Engineering, 24: 1945–1957, 2004.
40. Cortella G., D’Agaro P., Air curtains design in a vertical open display cabinet, In: Proceedings of IIR
International Conference New Technologies in Commercial Refrigeration, Urbana, USA, Hrnjak P.S.,
Ed., 55–63, 2002.
41. Cui J., Wang S., Application of CFD in evaluation and energy-efficient design of air curtains for hori-
zontal refrigerated display cases, International Journal of Thermal Sciences, 43: 993–1002, 2004.
42. Gaspar P.D., Carrilho Gonçalves L.C., Pitarma R.A., Experimental analysis of the thermal entrainment
factor of air curtains in vertical open display cabinets for different ambient air conditions, Applied
Thermal Engineering, 31: 961–969, 2011.
43. Wu X., Chang Z., Yuan P., Lu Y., Maa O., Yin X., The optimization and effect of back panel structure
on the performance of refrigerated display cabinet, Food Control, 40: 278–285, 2014.
44. Foster A.M., Madge M., Evans J.A., The use of CFD to improve the performance of a chilled multi-deck
retail display cabinet, International Journal of Refrigeration, 28: 698–705, 2005.
45. Sun J., Tsamos K., Tassou S.A., CFD comparison of open-type refrigerated display cabinets with/
without air guiding strips, Energy Procedia, 123: 54–61, 2017.
46. Marinetti S., Cavazzini G., Fedele L., De Zan F., Schiesaro P., Air velocity distribution analysis in the
air duct of a display cabinet by PIV technique, International Journal of Refrigeration, 35: 2321–2331,
2012.
47. Rossetti A., Minetto S., Marinetti S., A simplified thermal CFD approach to fins and tube heat
exchanger: Application to maldistributed airflow on an open display cabinet, International Journal of
Refrigeration, 57: 208–215, 2015.
48. Marinetti S., Cavazzini G., Lauri I., Testa S., Minetto S., Numerical and experimental analysis of the
air flow distribution in the cooling duct of a display cabinet, International Journal of Refrigeration, 42:
8–13, 2014.
49. Rossetti A., Minetto S., Marinetti S., Numerical modelling and validation of the air flow maldistribution
in the cooling duct of a horizontal display cabinet, Applied Thermal Engineering, 87: 24–33, 2015.
50. Ge Y.T., Tassou S.A., Hadawey A., Simulation of multi-deck medium temperature display cabinets with
the integration of CFD and cooling coil models, Applied Energy, 87: 3178–3188, 2010.
51. Ge Y.T., Tassou S.A., The impact of geometric structure and flow arrangement on the performance
of CO2 evaporators in multi-deck medium temperature display cabinets, International Journal of
Refrigeration, 35: 142–149, 2012.
52. Laguerre O., Hoang M.H., Flick D., Heat transfer modelling in a refrigerated display cabinet: The influ-
ence of operating conditions, Journal of Food Engineering, 108: 353–364, 2012.
53. Foster A.M., Quarini G.L., Using advanced modelling techniques to reduce the cold spillage from
retail display cabinets into supermarket stores to maintain customer comfort, In: Proceedings of the
Institution of Mechanical Engineers, Part E - Journal of Process Mechanical Engineering, 215: 29–38,
2001.
54. D’Agaro P., Croce G. Cortella G., Numerical simulation of glass doors fogging and defogging in refrig-
erated display cabinets, Applied Thermal Engineering, 26: 1927–1934, 2006.
55. Brown T., Corry J.E.L., James S.J., Humidification of chilled fruit and vegetables on retail display
using an ultrasonic fogging system with water/air ozonation, International Journal of Refrigeration, 27:
862–868, 2004.
56. Hu Z., Sun D.-W., CFD simulation of heat and moisture transfer for predicting cooling rate and weight
loss of cooked ham during air-blast chilling process, Journal of Food Engineering, 46: 189–197, 2000.
57. Xia B., Sun D.-W., Application of computational fluid dynamics in the food industry: A review, Computer
and Electronics in Agriculture, 34: 5–24, 2002.
58. Moureh J., Flick D., Airflow characteristics within a slot-ventilated enclosure, International Journal of
Heat and Fluid Flow, 26: 12–24, 2005.
59. Laguerre O., Hoang M.H., Osswald V., Flick D., Experimental study of heat transfer and air flow in a
refrigerated display cabinet, Journal of Food Engineering, 113: 310–321, 2012.
60. Axell M., Fahlen P.O., Tuovinen H., Influence of air distribution and load arrangements in display cabi-
nets, In: Proceedings of the 20th International Congress of Refrigeration, IIR/IIF, Sydney, AU, paper
152, 1999.
2 CFD Optimization of
Perturbed Air Curtains for
Refrigerated Display Cabinets
Jean Moureh
CONTENTS
2.1 Introduction......................................................................................................................... 24
2.2 Energy Consumption and Food Safety Aspects of RDC.................................................... 24
2.3 Numerical Modeling................................................................................................................24
2.4 Airflow Design and Research Methodology of RDC..............................................................25
2.5 Aims and Objectives................................................................................................................26
2.6 Experimental Setup and Procedure......................................................................................... 27
2.6.1 Air Curtain Facility......................................................................................................... 27
2.6.2 Non-Intrusive Experimental Facilities...................................................................... 29
2.6.2.1 Laser Doppler Velocimetry (LDV)............................................................... 29
2.6.2.2 Particle Image Velocimetry (PIV)............................................................. 30
2.6.3 General Features of the Flow Field.................................................................................... 32
2.7 Numerical Modeling Approaches............................................................................................ 33
2.7.1 CFD Numerical Modeling........................................................................................... 34
2.7.1.1 Governing Equations and Hypothesis...................................................... 34
2.7.1.2 Boundary Conditions........................................................................................... 36
2.7.2 Global Modeling Approach................................................................................................ 37
2.7.2.1 Governing Equations................................................................................................37
2.7.2.2 Air Curtain Performance................................................................................. 40
2.8 Results and Discussion.................................................................................................................................41
2.8.1 Without External Perturbation..................................................................................... 42
2.8.1.1 Jet Characteristics and Airflow Patterns....................................................... 42
2.8.1.2 Global Exchanges Through Air Curtain.......................................................44
2.8.2 With External Lateral Flow...................................................................................45
2.8.2.1 Effect of ELF on Airflow Patterns................................................................ 45
2.8.2.2 Turbulence Modeling Performance............................................................ 47
2.8.2.3 Effect of the ELF on the Jet Deflection.....................................................49
2.8.2.4 Effect of ELF on the Jet Decay..................................................................... 50
2.8.2.5 Effect of ELF on the Velocity Profiles.......................................................... 51
2.8.2.6 Effect of ELF on the Global Exchanges Through Air Curtain................... 53
2.9 Conclusions.............................................................................................................................. 55
Nomenclature.................................................................................................................................... 56
References......................................................................................................................................... 58
23
2.1 INTRODUCTION
In many industrial applications, air curtains formed by plane air jets are used to provide a dynamic
barrier for reducing and controlling the heat and mass transfer between two adjoining areas with
different levels of temperature, humidity and pollution. Air curtains are used to create refrigerated
spaces [1,2], open minienvironments for sensitive high-quality products [3], or to reduce the spread-
ing of fire smokes in underground tunnels [4,5]. One of the relevant applications is the vertical
open Refrigerated Display Cabinet (RDC) widely used in supermarkets. Within this context, the
air curtain plays a key role in keeping food at prescribed regular temperatures, while allowing an
energy-saved control and open access for customers.
As air curtains play a very important role in the cold preservation of display cabinets, much
research has been made in this field, including experimental studies [6–8], computational fluid
dynamics (CFD) simulations [9–16] and global modeling approaches [17–19].
The majority of these studies take into consideration only an idealized scenario in which no
perturbation affects the dynamic behavior of the air curtains. However, these devices are generally
used in open spaces like supermarkets, and are very sensitive to external perturbations generated by
human activities, such as pressure difference due to door opening, parasitic air draughts, action of
air conditioning system, etc. Such perturbations may strongly affect the stability, the air tightness,
the transfer mechanisms and therefore the efficiency of the air curtain due to a lack of confinement
and an increase of energy consumption [16]. D’Agaro et al. [9] indicated that the incidence of exter-
nal air currents on air entrainment needs further investigation. Therefore, the flow characteristics,
efficiency and performance are yet to be understood, especially with consideration of external per-
turbations, which one would encounter in practical industrial applications.
2.2 ENERGY CONSUMPTION AND FOOD SAFETY ASPECTS OF RDC

In 2002 it was estimated that there were 322,000 supermarkets and 18,000 hypermarkets world-
wide and that the refrigeration equipment in these supermarkets used on average 35–50% of the
total energy consumed in these supermarkets [20]. In supermarkets, refrigerated display cabinets
are the most common method of keeping chilled food at the required temperature and allowing the
customer almost unrestricted access to the food.
In a typical RDC, the air jet flows from nozzle inlet located at the top front to a nozzle exit
located at the bottom front of the case, acting as a thermal barrier between the warm ambient air
and the chilled compartment. Due to their design, RDC are very sensitive to ambient conditions and
are considered as the weakest link of the cold chain. It has been shown that mean food temperatures
among chilled cabinets can range from –1°C to +16°C [21].
These higher values could be explained by the fact air curtains are easily disturbed by the out-
side ambient air, which in addition results in higher temperature rise and more power consumption.
These disadvantages are difficult to overcome since the European testing standard, EN441, was
established for steady ambient conditions where the surrounding velocity is below 0.15 m/s.
2.3 NUMERICAL MODELING
Many authors [11–15] have shown computational fluid dynamics (CFD) to be a valuable tool to rap-
idly provide design options to improve airflow within display cabinets. In their studies, the authors
use the CFD models to optimize RDC design or to minimize energy losses through the air curtain
by testing the influence of the main factors, which include the dimensions of the nozzle inlet, the
nozzle exit, the length of air curtains, the initial velocity, turbulent intensity and temperature of air
curtains and the temperature and velocity of the ambient environment.
Smale et al. [14] offer a review on the importance of the complementary role of the CFD
approach and its ability to handle the complex configurations of refrigerating facilities, including
CFD Optimization of Perturbed Air Curtains for Refrigerated Display Cabinets 25
RDC. The authors stress the importance of validating CFD models, which necessitates rigorous
comparisons with experimental data.
D’Agaro et al. [9] perform 2D and 3D CFD simulations to investigate the effects of the cabinet
length, the air curtain and longitudinal ambient air movement on air flow pattern and temperature
distribution in a frozen food vertical display cabinet. The authors pointed out the importance of the
3D effects. The computed refrigerating power shows that even low room air velocity of 0.2 m/s,
due to its interaction with the end-wall vortices, has a significant impact on cabinet performance.
The introduction of the ambient air movement affects the return air temperature and consequently
the refrigerating power, which has been estimated to increase about 30%. A similar result was
found by Gaspar et al. [6], who conducted an experiment to study the heat transfer rate and thermal
entrainment factor of air curtains in a RDC in different ambient air conditions. In this study, the
high values of thermal entrainment observed at the sidewall locations can be attributed to sidewall
effects, as the air curtain is unable to restrict the free entry of external air at the extremities of the
RDC. The results of Gaspar et al. [6] also show that the increase of ambient air velocity magnitude
from 0.2 to 0.4 m/s, even parallel to the plane of the equipment’s frontal opening, promotes thermal
interaction between the conservation zone and ambient air masses by disturbance of aerothermody-
namics barrier provided by air curtain. The total heat transfer rate increases 53% due to increase of
air infiltration load across the air curtain.
However, it’s worth noting that the validation of the majority of CFD models reported in the liter-
ature [7,11,12,16] are obtained by comparisons with test results limited to temperature and humidity
(no velocity measurements were performed). Even when velocity measurements have been per-
formed, they are limited to some velocity profiles [1], which is not enough to validate complex 3D
airflow patterns [2]. Thus, there is a need to use more advanced non-intrusive techniques like PIV
and LDV in order to obtain high enough resolution to characterize air flow patterns and velocity
profiles and turbulence levels. The obtaining of such results allows a better understanding of airflow
characteristics and also improves the quality of CFD validation.
Two important physical parameters appear in the bibliography analysis and have been studied
in general by all others: (1) thermal entrainment, reflecting the energy loss, and (2) the air cur-
tain tightness, which is responsible for the temperature or pollution infiltration to the confined
cavity. According to Howell and Adams [22], 75% of the refrigeration load related to an RDC is
induced by the air curtain entrainment, which could be viewed as a global air exchange with the
ambient. The results obtained by Navaz et al. [17] indicate that lowering the Reynolds number of
the air curtain reduces the entrainment rate. However, sufficiently high momentum should still
exist to enforce the integrity of the air curtain structure. A similar result was found by Moureh
et al. [18] and Flick et al. [19], who developed a global analytical model to predict jet velocities
and temperature profiles along a turbulent jet used to seal/confine a heated cavity. In a forced
convection regime, the thermal entrainment (i.e., energy loss) becomes proportional to the flow
rate of the jet.
Amin et al. [8] proposed an original approach combining experimental technique and a global
analytical model to quantify the steady state infiltration rate of RDC. The infiltrating rate was
defined as the part of entrained flow drawn into the nozzle exit. This technique is based upon the
concentration measurements of a tracer gas at three locations: nozzle inlet, nozzle exit and the
ambient.
2.4 AIRFLOW DESIGN AND RESEARCH METHODOLOGY OF RDC

In RDC, there are several factors affecting the complexity of refrigerated air curtains flow, due
to the nature of the regime flow, thermal and geometrical aspects. According to Field and Loth
[23], with typical RDC dimensions and the corresponding jet flow rates, the Reynolds number is
about 5,000 and the typical air curtain lengths are about ten jet widths. Therefore, the air curtain
flows reside in the transitional flow regime, which is not well understood, and little information
was available concerning the velocity profiles. In addition, the transitional flow is more complex to
predict, since classical turbulence model RANS are more adapted to fully developed turbulence.
Unlike free turbulent jets, refrigerated air curtains experience aerodynamic and thermal asym-
metric design; one side has a cold temperature and no stagnant air within a confined space, while
the ambient side has a high temperature and lower velocities. Another complexity concerning the
thermal aspect is the fact that refrigerated air curtains are negatively buoyant jets, which combine
the effects of forced convection and natural convection due to density difference. In addition, the
entrainment of warm air from the ambient causes the air curtain to become warmer as it falls down
the cabinet. Thus the predominance of natural convection decreases progressively along the cabi-
net. The measurements of [Field and Loth, 2006] indicated negatively-buoyant acceleration follow-
ing the jet exhaust, followed by a more linear curtain growth characteristic of isothermal wall jets.
This increases the edge instabilities, which increase the mixing and entrainment into the air curtain.
Finally, the geometry employed is complex and specific to a particular display case. The free
turbulent shear flow developed by the jet is strongly affected by aerodynamic interactions with
shelves, loading arrangements of products, back flow panel and secondary recirculation developed
as semi-confined flow within the cabinet. Given this complexity, the global studies cited above per-
formed on specific cases, which behave as black-boxes, could be considered empirical approaches
since the main geometric and air flow parameters are fixed. Consequently, it was difficult to develop
a fundamental understanding of the underlying fluid dynamics of air curtains in general and thus to
derive universal recommendations aiming to improve their design.
In order to perform more fundamental studies related to fluid dynamics of the air curtain while
eliminating case-specific flow features particular to each manufacturer, some authors perform
investigations on more simplified air curtain geometries, such as isothermal wall jet [23,24] or plane
jet confining rectangular empty cavities [18,19,25].
2.5 AIMS AND OBJECTIVES

Previous studies only assessed the efficiency of air curtain devices in quiescent ambient conditions.
Some fundamental studies have also been extended by a number of researchers to jets discharged
in non-quiescent surroundings with “co-flow” [26,27] and “cross-flow” [28,29]. However, to the best
of our knowledge, the flow configuration of a perturbed air curtains resulting from an aerodynamic
interaction with an external lateral flow has not yet been studied in detail either experimentally or
numerically.
The aim of this research is to experimentally and numerically characterize the dynamic behav-
ior and the confinement effectiveness of an air curtain used to confine a rectangular cavity and
subjected to external perturbation. This study is performed on a reduced-scale model representing
an idealized configuration of a RDC, which is formed by an isothermal simple jet plane, and inves-
tigates the influence of the main parameters related to the jet, cavity and the lateral perturbation.
This allows a more fundamental study of the physics of an air curtain, while avoiding the more case-
specific geometries used by many authors [7,11,13,30]. The study focuses on the near field region
(x/e < 10), which includes the transition zone where strong interactions are expected between the jet
core, cavity and external lateral flow, due to this region’s strong relevance to RDC.
Velocity measurements are performed by means of LDV and PIV techniques. The numerical
approach was performed using computational fluid dynamics (CFD) Fluent code with the standard
k-ɛ model, and the more advanced RSM. The main purpose is to acquire reliable data on the physics
of the air curtain development and transfer mechanisms with and without external perturbation. It
also allows a critical evaluation of the performance of these models and therefore validation of the
numerical model by comparisons with LDV and PIV measurements.
The analysis of experimental and numerical data obtained with and without external perturba-
tion make it possible to quantify the effect of the perturbation on the main jet characteristics related
to maximum velocity decay, growth rate, self-similarity, as well those related to air flow patterns,
jet stability and confinement effectiveness. The numerical developed model was then validated with
PIV measurements and with the available results in terms of the thermal entrainment and infiltra-
tion rate in the literature. The validated CFD model was then combined with an original modeling
approach based on entrainment spill mechanism of the jet. This enables the construction of a global
model used to predict heat or mass fluxes exchanged between cavity, air curtain and the ambient
with and without external perturbation. In addition to the RDC case, studying transitional disturbed
plane jets can provide further understanding of the fundamental nature of the flow physics, lead-
ing to improvements in various practical applications where air curtains evolve in non-quiescent
ambient.
2.6 EXPERIMENTAL SETUP AND PROCEDURE

2.6.1 Air Curtain Facility
The experimental setup shown in Figure 2.1 is a scaled-down (1:5) model of a cavity representing
a generic configuration of an open display cabinet with respect to Reynolds dimensionless number,
defined as Re = e U/v. It represents an air curtain confining cavity subjected to an external lateral
flow (ELF) under isothermal conditions. Using a scaled -down model as a generic configuration of
a RDC is very instructive to characterize the fundamental nature of flow field while eliminating
case-specific issues particular to each manufacturer.
The key benefit and originality of this facility derives from its ability to obtain simultaneously
the control of the air curtain flow and the ELF. The cavity is of 0.4 m height and 0.5 m width with
an adjustable back wall enabling different depths ranging from 0.15 m (which corresponds to an air
curtain confining a cavity) to 0 m, representing a plane wall jet that corresponds to a fully stocked
RDC [31]. In this paper, only the configuration with a cavity depth of 0.15 m is considered.
The experimental device is composed of two airflow separate circuits related to the air curtain
and to the ELF. Airflows for both air curtain and ELF are provided by two centrifugal fans, which
can deliver a nominal flow rate of 0.5 m3s−1 with pressure loss of ΔP = 1469 Pa for the air curtain
and 1.25 m3s−1 for the lateral flow with ΔP = 1959 Pa. Both airflow circuits are put into two closed
FIGURE 2.1 Experimental scaled-down test facility of a display cabinet: (a) settling chamber, (b) smooth
contraction nozzle, (c) rectangular channel, (d) cavity wall, and (e) exit of the recycled jet. Blue vertical arrow:
Air curtain; red horizontal arrow: External Lateral Flow (ELF).
recycled air loops. This configuration also reduces the amount of injected tracer particles needed to
achieve better measurement.
The main dimensions and the axis system are presented in Figure 2.2. Cartesian coordinates are
used, x axis for streamwise direction related to the air curtain flow, y axis for transverse direction
and the third z axis for lateral or span-wise evolution where the ELF flows. The origin is fixed in
the middle of the nozzle inlet and in the (x-y) mid-plane of the air curtain (Figure 2.2b). In the air
curtain airflow circuit, the air jet blows from top to bottom (Figure 2.2a). The rectangular settling
chamber contains two honey-comb flow straighteners and four perforated screens of progressively
finer mesh sizes of 4 × 10 −3 m, 3 × 10 −3 m, 2 × 10 −3 m and 1 × 10 −3 m are put in order to ensure the
uniformity of the mean velocity with a low turbulence level.
A rectangular channel (of section 0.04 m × 0.5 m) leading to the jet nozzle inlet is connected
to the settling chamber with a smooth contraction ratio of 1/5. The aspect ratio (l/e) of the channel
is 12.5. The air curtain length extended to 10 jet widths (H/e = 10) downstream to better focus
the initial stages of the plane jet development in the near field region. Data were acquired in
FIGURE 2.2 Sketch of the scale model: (a) front view, (b) coordinate system, and (c) top view.
TABLE 2.1
Geometrical Characteristics of the Jet and the Cavity
e (nozzle inlet width) 0.04 m
D (cavity depth) 0.15 m
H (cavity height) 0.4 m
L (cavity length) 0.5 m
the turbulent regime at a Reynolds number based on the slot exit (e) of 0.04 m thickness, giving
Re ≈ 8000.
In the ELF circuit, the air is injected in a rectangular horizontal channel of dimensions 0.4 m
height, 0.5 m width and 2.4 m length. To control the uniformity of the ELF flow, 0.5 m of the
channel entrance has been devoted to a settling chamber composed of a honeycomb and a set of
perforated plates (Figure 2.2c). The distance between the last perforated plate and the lateral side
of the air curtain, which corresponds to the beginning of the interaction with the ELF, is 0.7 m.
At z = –0.25 m, located downstream of the interaction with the air curtain, the air flow rate of the
ELF is extracted in a channel in order to form a closed loop. The scaled-down experimental model
walls are made of Plexiglas of higher optical quality material, which is advantageous for LDV and
PIV image recording since the laser sheet can be directed through the wall cavity surface, allowing
the flow field to be free of optical obstructions. The geometrical dimensions characteristics of the
scaled-down model are summarized in Table 2.1.
In this study three external lateral flow velocities (Ulf) have been chosen: 0 ms−1 (unperturbed air
curtain), 0.5 ms−1 and 1 ms−1, which correspond to the ratios Ulf /U0 of 0 (unperturbed air curtain),
0.16 and 0.33, respectively.
To investigate the three-dimensional effect of the ELF on the air curtain flow in the lateral
direction, four positions were selected for PIV measurements: distances of 0.05 m, 0.1 m, 0.25 m
and 0.4 m from the lateral side where the ELF comes in, which correspond to z/L = 0.4, z/L = 0.3,
z/L = 0 (mid-plane) and z/L = –0.3, respectively.
2.6.2 Non-Intrusive Experimental Facilities

Two experimental techniques, Laser Doppler Velocimetry (“LDV”) and Particle Image
Velocimetry (“PIV”), have been used to characterize the mean velocity field and its fluctuating
components. For both techniques, the flow was seeded by white atomized oil particles of 4 × 10 −6 m
that scatter light as the flow carries them through the measurement volume of LDV or the laser sheet
of PIV. The atomizer was set in a box placed in the air curtain jet loop before entering the settling
chamber.
2.6.2.1 Laser Doppler Velocimetry (LDV)

One-dimensional Laser Doppler Velocimetry (Laservec) produced by TSI (Shoreview, MN, USA)
manufacturer was used. The LDV consists of an LDP-100 probe and the raw data are processed
using an IFA 600 signal processor. This measurement technique does not interfere with the flow
and is able to correctly resolve the sign, as well as the magnitude of velocity, and determine mean
velocity and its fluctuation. It comprised a 50 × 10 −3 W laser diode emitting a visible red beam at a
690 × 10 −9 m wavelength, a beam splitter, a Bragg (acousto-optic) cell, a focusing and receiving lens
to collect scattered light from the measurement point and a photomultiplier. The accuracy is consid-
ered to be below 1%. For the present experiments, a maximum of 10,000 samples was specified for
each point measured, with a maximum sampling time of 2 min. The data acquisition was stopped
depending on which of these two events occurred earlier. The data rate varied between 100 and
1000 Hz for all measurements. The probe was carried on an automatic three-dimensional displace-
ment system that provides a position resolution of ±0.5 × 10 −3 m in three directions.
2.6.2.2 Particle Image Velocimetry (PIV)

The 2D-PIV was used to provide instantaneous velocity field measurements in vertical cross-
sections of the flow downstream of the air curtain. This technique is non-intrusive and measures
the velocities of micron-sized particles following the flow. The PIV system used for the determina-
tion of the velocity fields was essentially composed of a CCD camera of 1375 × 1086 pixel resolu-
tion (LaVision sense M2/E 12 bits) with an objective Nikon (60 mm), a pulsed Nd:Yag laser with
the energy of 15 mJ/pulse (Double cavity ) and a ‘‘LaVision” correlator. The laser light is passed
through suitable optics to form a light sheet with approximately 1.5 mm thickness. The whole sys-
tem is driven by the software ‘‘Davis V7.2.” The laser light scattered from the seeded particles is
imaged on a camera and a 64 × 64 pixel region is used on the first pass and a 32 × 32 pixel region
with 50% overlap is used on the second pass. The time delay between two laser pulses was 3.5 ms
and the time delay between the capture of two pairs of pictures was 200 ms. For all measurement
cases 1,000 pairs of images are used for each velocity measurement to allow good accuracy in
statistical calculations of the mean flow and turbulence fluctuation.
To validate the 2D behavior of the air curtain device, Figure 2.3 shows the quasi-uniformity
of inlet velocities and turbulence intensities at (x/e = 1; y/e = –0.5) along the jet nozzle (–0.5 <
z/L < 0.5). Furthermore, Figure 2.4 confirms the top-hat character of the jet obtained in the
mid-plane (z/L = 0) within the jet nozzle (–1 < y/e < 0). To assess the uniformity of the ELF, air
velocity measurements were performed within the duct, upstream of the interaction with the
air curtain maintained at 0m/s along the x direction at z = 0.44 m and y = 0.15 m (Figure 2.5).
The obtained results show the quasi-uniformity of the velocities (Ulf = –1.05 ms –1 ± 2.5%) and
turbulence intensities (4.4% ± 0.1%).
FIGURE 2.3 Inlet velocities and turbulence intensities at (x/e = 1; y/e = –0.5) along the jet nozzle inlet
(–0.5 < z/L < 0.5).
FIGURE 2.4 Profile of inlet velocities (U0) in the mid-plane (z/L = 0) within the jet nozzle (–1 < y/e < 0) at
x/e = 1.
FIGURE 2.5 Velocity and turbulence intensity measurements in the ELF at z = 0.44 m and y = 0.15 m.
2.6.3 General Features of the Flow Field

On the basis of the present investigation and the results reported by many authors [18,19,32], a
recirculating air curtain developed in a relatively quiescent ambient behaves as a free jet. The cor-
responding flow field, schematically represented in Figure 2.6, consists of two, linearly growing
mixing layers separated by a diminishing potential core.
As can be seen, the two main regions identified in this figure may be defined as follows: the
first region is referred to as the potential core region, in which the axial component of velocity
is essentially a constant, (Ucl = U 0; where Ucl is the mean axial velocity along the centerline of
the jet and U 0 is the mean velocity at the nozzle exit), and the zone of fully established flow,
which originates at about the location where the two mixing layers meet. In the fully turbu-
lent region, velocity profiles at various distances from the nozzle inlet are similar. According
to Reichardt [33], the self-similar profiles could be approximated by the following Gaussian/
exponential function:
   
2
U/ U m = exp  − ln(2).  y
 y1/ 2  
 (2.1)


The normalised velocity half-widths are given by
y0.5  x 
= K y  x + 0  (2.2)
e e e
where Ky is the jet spreading rate and x0 is the virtual origin.

y0.5 represents the distance at which the velocity is one-half the value of the maximum velocity
Um:
U (x) = 1 Um( x )
2
Figure 2.7 displays instantaneous flow visualization via PIV tomography of the shear layer
within the cavity. It shows the onset and development of initial eddies of Kelvin–Helmholtz
(K–H) instabilities that dominate the early stages of air jet transition process and play a relevant
FIGURE 2.6 Schematic view of the plane turbulent jet regions.

x /e = 3
x /e = 4
FIGURE 2.7 PIV visualization of Kelvin–Helmholtz instability for unperturbed jet (Ulf = 0 ms–1) at the
mid-plane (z/L = 0).
role in enhancing mixing and entrainment between the jet and the ambient. The roll-up and pair-
ing process of the K–H instability that leads to the formation of larger structures represent the
principal mechanisms governing the growth of the shear mixing layer. The dominance of large-
scale structures diminish as they convect downstream due to the generation of a broader range of
smaller eddies.
Downstream of the jet nozzle inlet, the jet spreads out, which causes a decrease in air jet velocity
and an increase of the mass flow rate of the jet by entrainment mechanisms from both sides: ambient
and cavity. This process continues until the jet flows approach the nozzle exit located at the bottom
of the cavity at x/e = 10. At this level, the global flow rate exceeds the recirculated part. Therefore,
the jet splits into three different air flow components. The main or central part of the jet, which cor-
responds to the mass flow rate of the jet (m 0), is drawn into the nozzle exit to continue the steady
flow operation of the jet. The outer part of the air curtain is spilled back to the ambient surroundings
in order to balance the quantity of air that was entrained by the air curtain on the outside. The inner
part of the air curtain is spilled into the cavity and entrains the formation of a global recirculation
within the cavity, as observed by many authors [17,19,25].
2.7 NUMERICAL MODELING APPROACHES

In this study, two complementary modeling approaches were performed. The first is based on the
CFD numerical predictions, which enable better understanding of the local air flow characteristics,
while the second deals on the global fluxes exchanged between the air curtain, the ambient and the
cavity.
2.7.1 CFD Numerical Modeling

2.7.1.1 Governing Equations and Hypothesis
The description of airflow development is based on the conservative law of mass and momentum.
The solved equations can be written as follows:
∂U j
Mass conservation: = 0 (2.3)
∂x j
∂U jUi 1 ∂P ∂  ∂Ui 
Momentum conservation: =− +  v ∂x − ui u j  (2.4)
∂x j ρ ∂xi ∂x j j
Passive transport scalar equation:

The general equation form of the convection-diffusion of a passive scalar tracer in steady state
without chemical reaction is written as:
∂U jC ∂  v ∂C 
=  Sc ∂x − u j c  + S (2.5)
∂x j ∂x j  j 

where C is the mean mass fraction of the passive tracer, S is the source term and ui u j and u j c
are, respectively, the unknown Reynolds stresses and mass fluxes. The obtaining of these quanti-
ties depends on the turbulence closure. For the case investigated, two levels of turbulence modeling
closure have been employed:
• on one hand, with the standard two-equation k-ɛ model described by Launder and Spalding
[34];
• on the other hand, with a second-moment closure with the Reynolds stress model (RSM).
a) The standard k-ε model

Using the Boussinesq hypothesis, the Reynolds stresses can be described as follows:
 ∂U ∂U j  2
− u i u j = vt  i + − kδ ij (2.6)
 ∂x j ∂xi  3

The eddy (turbulent) viscosity νt is obtained from:
k2
vt = Cµ fµ (2.7)
ε
The turbulence kinetic energy k and the dissipation rate ε are determined using the fol-
lowing transport equations respectively:
∂k ∂  vt  ∂k 
Ui =  v +  + Pk − ε (2.8)
∂xi ∂xi  σ k  ∂xi 

∂ε ∂  vt  ∂ε  ε
Ui =  v +  + (C1 Pk − C2 ε ) (2.9)
∂xi ∂xi  σ ε  ∂xi  k

Pk represents the shear production term:
 ∂U ∂U j  ∂Ui
Pk = vt  i +  (2.10)
 ∂x j ∂xi  ∂x j

The model coefficients in the standard k-ε model are:
(Cµ , C1 , C2 , σ k , σ ε ) = (0.09, 1.44, 1.92, 1.0, 1.3) (2.11)

b) The RSM
The transport equations for Reynolds turbulent stress are obtained by subtracting the
product of the mean velocities by the time-averaged Navier–Stokes equations from the
product of instantaneous velocities by instantaneous Navier–Stokes equations. This gives
rise to:
∂ui u j ∂  p ∂(ui u j )  p  ∂u ∂u  ∂u ∂u j
Uk =− ui u j uk + (δ kj ui + δ ik u j ) − v  + Pij −  i + j  − 2v i (2.12)
∂x k ∂x k  ρ ∂x k  ρ  ∂xj ∂xi  ∂x k ∂x k
∂U j ∂Ui
where Pij = −ui uk − u j uk represents the production term.
∂x k ∂x k
The diffusive transport term was represented by a simplified form of the generalized gradient
diffusion hypothesis as:
∂  p ∂ (ui u j )  ∂  νt ∂ 
− ui u j uk + (δ kj ui + δ ik u j ) − ν =  (ui u j ) (2.13)
∂x k  ρ ∂ x k  ∂x k  σ k ∂x k 

p  ∂ ui ∂ u j 
 +  represents the pressure-strain term leading to a redistribution process
ρ  ∂ xj ∂ xi 
within the Reynolds-stresses towards an isotropic state. It is modeled by Wilcox formula-
tion [35], which is recommended to shear and swirling flows.
The dissipation term was assumed isotropic, and was approximated by:
∂ui ∂u j 2
2ν = δ ij ε (2.14)
∂x k ∂x k 3
The dissipation rate was computed via the ε transport equation as described by the
standard k-ε model [34]. It implies the isotropy of the dissipation process at smallest length
scales.
2.7.1.2 Boundary Conditions
The computational domain may be surrounded by inflow, outflow boundaries and solid walls. At
the two inlet sections concerning the jet and the external flow, uniform distribution is assumed for
velocity components, transport variables, kinetic energy of turbulence (k0), the energy dissipation
rate and the dissipation rate (ε 0). The turbulence quantities are imposed by means of two param-
eters: the turbulence intensity as obtained from experiments, and the hydraulic diameter, assum-
ing a fully developed duct flow upstream. For the RSM, turbulence is assumed to be isotropic:
2
(u i u j = k0δ ij ). For the air curtain, the magnitude of the inlet velocity was tacked constant for all
3
performed simulations at 3 ms–1 and 6 ms–1 for the nozzle inlet width of e = 4 10 –2 m and 6 ms–1 for
2 10 –2 m. This allows the finding of:
• Two different Reynolds numbers: Re = 8000 and Re = 6000 for e = 4 10 –2 m, and Re = 8000
for e = 2 10 –2 m.
• Two different jet aspect ratios: H/e = 10 for e = 4 10 –2 m and H/e = 20 for e = 2 10 –2 m.
For the external lateral perturbation, different ratios of Ulf /U0 were simulated, ranging from 0 to
0.33 for H/e = 20 and from 0 to 0.66 for H/e = 10.
At the outlet boundaries related to the air curtain and to external flow, the air flow rate is imposed
with respect to the mass balance, while the velocity profiles are unknown. Based on the present
model, the non-diffusion outlet boundary conditions for velocity and concentration are assumed.
These have been extensively analysed and proven to be a correct and reasonable assumption for air
flow within ducts.
The near wall region was modeled using the standard logarithmic wall function. The boundary
conditions are summarized in Table 2.1 and indices (τ, η, λ) represent the local coordinate system
of the wall, where τ is the tangential coordinate, η is the normal coordinate, and λ is the binormal
coordinate. Then the local Reynolds stresses at the wall-adjacent cells were computed from:
uτ′2 u′2 u′ 2 u′ u′
= 1.098, η = 0.247, λ = 0.655, and τ η = 0.255. (2.15)
k k k k
The values of k used at the boundaries are obtained from a transport equation of turbulence
kinetic energy similar to the one in the standard k-ε model. For reasons of computational conve-
nience, this equation is solved globally (on the domain), even though the values of k thus computed
are needed only near the wall. The boundary condition for k imposed at the wall is:
∂k
= 0 (2.16)
∂n
where n is the local coordinate normal to the wall.

The production of kinetic energy and its dissipation rate ε at the wall-adjacent cells, which are
the source terms in the k equation, are computed on the basis of the local equilibrium hypothesis.
The different boundary conditions for different simulations were summarized in the Table 2.2.
The governing equations are solved using the finite-volume ANSYS Fluent® CFD code [36] in a
staggered grid system and the SIMPLER algorithm for pressure–velocity coupling. For the spatial
discretization, a second-order accurate upwind differencing scheme is used for the momentum, all
convective-diffusively transported variables and Reynolds stress components.
A three-dimensional, non-uniform hexahedral grid was used in this study, with high-density
mesh in regions near the inlet, outlet, walls and jet boundaries where high gradients are expected.
The growth ratio between two adjacent cells does not exceed 20%.
TABLE 2.2
Boundary Conditions
Inlet (jet) U0 = 3 and 6 (m/s)
I0x = 5%, Dh = 0.074 (m)
k0 = 3/2 (U0I0x)2
( ) 2
ε 0 = Cµ0.75 k01.5 / 0.07 DH ui u j =
3
k0δ ij
Inlet (ELF) Ulf = 0, 0.2, 0.5, 0.7, 1.0, 1.2, 1.5, 1.8 and 2 m/s
I = 5%, Dh = 0.44
Outflow ∂u/∂n = 0 (1/s)
Wall (no slip condition) Ui = 0
Wall treatment
1 u p Cµ1/ 4 k 1p/ 2 ρ Cµ1/ 4 k 1p/ 2 y p
u* = ln( E y* ), u* = , y* =
κ τ w /ρ µ
(a) (b)
cavity
FIGURE 2.8 Mesh topology in (a) xy and (b) yz cross sections.
The numerical results are obtained with 1.5 million cells. It was observed that increasing the grid
to larger than this value does not affect the numerical results. A view of the used mesh in two cross-
sections showing the jet and the cavity domain is given in the Figure 2.8. The computations were
run on the PC station, with total CPUs of 8, and 10 Go of RAM. The convergence was established
after 10,000 iterations, which “take[s] approximately 2 days.”
2.7.2 Global Modeling Approach

Due to the strong 3D effect exerted by the ELF of the jet behaviour described in the paragraph 2.8.2,
the air curtain flow loses its initially two-dimensional character and thus experiences substantial vari-
ations along the cavity in the lateral direction. Therefore, a global modeling approach appears more
appropriate to characterize the fluxes exchanged between the whole cavity, ambient and the air curtain.
2.7.2.1 Governing Equations
To evaluate exchanges through the air curtain with and without lateral flow, a global modeling
approach based on the entrainment-spill mechanism was built. This model considers:
• Four compartments supposed to be homogeneous in temperature or concentration: cavity

(c), ambient (a), jet nozzle inlet (0) and jet nozzle exit (s) (Figure 2.9). Each compartment
is represented by an indice.
• Seven fluxes exchanged between the compartments (Figure 2.9). Each flux involves two
compartments and is represented by their respective indices. For example, the ac flux rep-
resents the flux flowing from ambient (a) to the cavity (c).
Ambient
T a
ca oa Jet Exit
as
Jet inlet Uo To os T s
cs
ac oc
Tc
Cavity
FIGURE 2.9 Schematic view of the seven air fluxes exchanged between the four compartments.
In a turbulent regime (Sct = Prt), heat and mass transfer are driven by forced convection and
turbulent diffusion, while molecular diffusion is neglected. As a consequence, all the fluxes defined
above are proportional to the Reynolds number of the jet. This finding was observed by many
authors [18,19,32,37]. For simplicity, we consider dimensionless fluxes by dividing them by the air
mass flow rate of the jet m 0 .
Assuming heat and mass transfer analogy, non-dimensional temperature and concentration are
similar. This also concerns heat and mass fluxes exchanged between the four compartments. For
example: m 0 ca represents the mass flux of air molecules flowing from cavity c to the ambient a.
The heat and mass fluxes transferred from the cavity to the ambient could be written as: m 0 C p ca Tc
and m 0 ca Cc, respectively (where T = 0°C represents the enthalpique reference).
This representation reflects the physics of a jet. At the nozzle exit level, the mass rate of flow
that is entrained by the jet is spilled out into the cavity (ac + oc) and into the ambient (oa + ca). In
addition, the recirculated part flowing into the nozzle exit, which represents the mass rate of the
jet (m 0), contains three components: os, as and cs, which come from the three compartments: jet
nozzle inlet, ambient and cavity.
Among these air fluxes, two are particularly important:
• The air flux flowing from the jet nozzle inlet to the jet nozzle exit: os represents the capacity
of the air curtain to seal the cavity. Higher values of os close to 1 indicate lower exchanges
between the air curtain and its surroundings.
• The air flux flowing from the ambient into the cavity by the spilled air: ac characterizes the
air tightness efficiency of the air curtain. It also represents the level of the cavity protection
against ambient contaminants, such as seeds, dust, smoke, etc. For example, if Ca repre-
sents the mass concentration of contaminant in the ambient, the quantity of contaminants
carried into the cavity could be expressed as: Ca m 0 ac.
• In the case of a retail display cabinet, (ca Tc + oa T0) represents the spillage of cold air from
the case to the store.
The conservation of (total mass flow rate of the “air + tracer gas” mixture) mass at the nozzle
inlet, nozzle exit and cavity, allowsfor three relations:
oc + os + oa = 1 → as + os + cs = 1 → oc + ac + cs + ca (2.17)
To identify the seven unknown air fluxes involved in this model, a numerical approach based
on the transportation of tracer gas within the studied system is developed by using the CFD model
built with the RSM. Two numerical but independent simulations were performed with and without a
source term for the gas tracer within the cavity. The momentum equations for velocity and pressure
were solved first, and the transport equation for the tracer was then solved based on the converged
velocity results.
Obtaining concentrations of the tracer within the compartments makes it possible to write four
additional relations:
• The first simulation consists of injecting a constant concentration of the tracer gas in the
jet discharge (C01) and computing, by the CFD model, the corresponding concentrations of
the cavity (Cc1), ambient (Ca1) and jet exit (Cs1).
The mass conservation of tracer gas at the nozzle exit and the cavity result in:
os.C01 + cs. Cc1 + as.Ca = (as + os + cs).Cs1 (2.18)

oc.C01 + ac.Ca + = (oc + ac).Cc1 (2.19)

• The second simulation consists of injecting a constant concentration of the tracer gas in the
jet (C02), and applying a uniform source term of gas tracer S (kg/m3.s) within the cavity
domain in such a way to obtain:
C02 = Cc 2 (2.20)

Several trials were required to obtain the appropriate value of S . As in the first simulation, the
mass conservation of tracer gas at the nozzle exit and the cavity result in:
os.C02 + cs. Cc 2 + as.Ca 2 = (as + os + cs).Cs 2 (2.21)

oc.C02 + ac.Ca + S .Vc / m 0 = (oc + ac).Cc 2 (2.22)

In these equations, the mass concentration of gas tracer at the nozzle exit is computed as:
m gas tracer
Cs 2 = (2.23)
m gas tracer + m air

In the cavity, the mass concentration of gas tracer was calculated as the numerical average value:
Cc =
∫ C dv (2.24)
Vc
Where Vc represents the volume of the cavity and is equal to 0.022 m3.
By combining (2.20) and (2.22), it is possible to determine the air flux ac as:

SV
ac = c
(2.25)
m 0 (Cc 2 − Ca )
Solving of the algebraic Equations (2.17), (2.18), (2.19), (2.21) and (2.22) with MATLAB soft-
ware allow for the finding of the seven unknown mass-air fluxes depicted in Figure 2.9. This meth-
odology is applied under different external velocity perturbations.
In the particular case of no external lateral flow and assuming a symmetrical jet hypothesis, we
can write:
ac = ca and oc = oa = as = cs = (1 − os) / 2 (2.26)

2.7.2.2 Air Curtain Performance

Prior to the results presentation, a few concepts and definitions should be introduced, in view of the
analysis and the assessment of the sealing performance of the air curtain. The objective is to build
dimensionless parameters related to heat or mass exchanges through air curtains as a function of
characteristic fluxes defined above (Figure 2.9).
By replacing concentrations by temperatures in the studied system (Figure 2.9), the heat balance
of the cavity domain and the jet can be expressed as:
oc(To − Tc ) = ac(Tc − Ta ) (2.27)

Q j / m 0 C p = (To − Ts ) = cs(Tc − Ta ) + as(To − Ta ) (2.28)

By combining Equations (2.27) and (2.28) and after manipulating, one can find:
To − Ts (Q j ) (m 0 C p ) ac.cs ac.as

= = + + as (2.29)
Tc − Ta Tc − Ta oc oc
This quantity corresponds to a dimensionless form of the energy loss of the jet required to
provide a given level of thermal sealing, i.e., the temperature difference between the cavity and
the ambient. This quantity is inversely proportional to the thermal confinement efficiency (TCE),
defined as the ratio between the thermal confinement level induced by the air curtain and the cor-
responding heat leakage:
T − Ta T − Ta
TCE =  c = c (2.30)
(Q j ) (m 0 C p ) T0 − Ts

To − Ts
This means that higher values of reflect lower TCE levels of the air curtain.
Tc − Ta
In the particular case of a symmetrical jet (Equation (2.26)), the Equation (2.29) can be expressed:
To − Ts (Q j ) (m 0 C p ) 1 − os
= = 2 ac + (2.31)
Tc − Ta Tc − Ta 2
The right term of Equation (2.31) shows the relative importance of the two characteristic fluxes:
ac and (1-os)/2 in TCE.
By defining a heat transfer coefficient h and a Nusselt number Nu related to the air curtain as
follows:
Q j = hHe(Tc − Ta ) (2.32)

hH Q j
Nu = = (2.33)
λ λ (Tc − Ta )
the relation (31) could also be written as:
To − Ts (Q j ) (m 0 C p ) Nu
= = (2.34)
Tc − Ta Tc − Ta Re Pr

ρU o e Cpµ
where Re = and Pr = .
µ λ
In a forced convection regime, many authors [18,19,32,37,38] concluded that (Nu/Re Pr) depends
only on the ratio (H/e).
Another important quantity that was characterized by many researchers [8,17,23,25,39] is the
infiltration rate or thermal entrainment defined by the ratio:
To − Ts
TE = (2.35)
To − Ta
TE varies between 0 and 1 and can be interpreted as the efficiency of an air curtain in infiltrating
the outside air. The higher values of TE represent higher infiltrations; those are undesirable.
In lower limit, TE → 0 occurs when Cs → C0. In our model, this corresponds to os → 1 and
as → 0. This means that the infiltrated part entrained from the ambient goes to zero and the
cavity is completely sealed.
In contrast, TE → 1 occurs when Cs → Ca. Therefore, at the nozzle exit the air is entirely
entrained from the ambient and similarly no portion of the air curtain reaches the nozzle exit. In our
model, this corresponds to as → 1 and os → 0. This means the air curtain ceases to seal the cavity,
due to its breaking by external forces, for example. Evidently, such a situation is highly undesirable
in regards to the cavity sealing purposes.
In our model, TE could be written as:
 T − To 
TE = as + cs  c (2.36)
 Ta − To 
The first term “as” represents the direct infiltration flowing directly from the ambient to the
nozzle exit. The second term represents the portion of “cs” originating from the ambient. It could
be viewed as an indirect infiltration flowing from the ambient to the nozzle exit but passing through
 Tc − To 
the cavity. The ratio  is induced by exchanges between the cavity and ambient. Note that
 Ta − To 
the simplified model elaborated by Amin et al. [8] neglects exchanges between the cavity and the
ambient. As a consequence, Tc = To and TE = as.
2.8 RESULTS AND DISCUSSION

In this part, experimental and numerical results obtained with the two modeling approaches,
CFD and the global model, are compared and contrasted. The comparisons concern jet charac-
teristics, air flow pattern and global exchange through the air curtain with and without external
lateral flow.
2.8.1 Without External Perturbation

2.8.1.1 Jet Characteristics and Airflow Patterns
Figure 2.10 illustrates a typical cross-section in the mid plane (z/L = 0) showing numerical and
experimental PIV results of velocity vectors related to the 2D flow induced by the air curtain
surrounded by the cavity and the ambient.
A qualitatively reasonable agreement is observed between predicted and experimental results
concerning the main airflow related to the air curtain and the global recirculation within the cavity,
which could be viewed as a secondary flow. The presence of the circulation cell in the cavity affects
the air jet trajectory, which is slightly deviated towards the ambient. In comparison with experimen-
tal data, the RSM predicts a more stretched recirculation. The high rate of mixing induced by this
recirculation could potentially entrain the homogenisation of temperature or concentration in case
of heat or mass transfer with the ambient.
In Figure 2.11, related to the centreline jet velocity decay in the mid-plane (z/L = 0), experimental
results exhibit a potential core that extends up to x/e = 3. This finding accords with previous work
FIGURE 2.10 Velocity vectors for unperturbed jet (Ulf = 0 ms–1) at the mid-plane (z/L = 0): comparison
between numerical (RSM) and experimental (PIV) results.
FIGURE 2.11 Decay of the dimensionless centerline jet velocity at the mid-plane (z/L = 0): comparisons
between experimental and numerical results.
concerning free jet [40] or air curtains [41]. Comparisons between turbulence models and experi-
mental data clearly show that k-ɛ model significantly overpredicts the length of the potential core,
and therefore fails to predict the jet centerline decay in the transition zone. On the contrary, a good
agreement is observed with the RSM predictions concerning the jet centerline decay in the initial
and the transition zones. A critical comparison between the two turbulence models is provided in
the paragraph 2.8.2.2.
The unnormalized and normalized streamwise velocity profiles at selected downstream loca-
tions are shown in Figures 2.12a and 2.12b. Figure 2.12a compares numerical and experimental
results, while Figure 2.12b compares experimental data with the theoretical profile obtained by
Gaussian-exponential function (Equation 2.1).
In Figure 2.12a, numerical and experimental results related to the position x/e = 5, which cor-
responds to the half-length of the air curtain, confirm the jet deviation outwards. However, local dif-
ferences are observed between numerical and experimental values concerning velocity profiles near
the nozzle exit at x/e = 9, even if the trends of these two profiles are similar. In fact, the RSM profile
exhibits unphysical deviation of the jet outwards at x/e = 9, which could be attributed to the effect
of the recirculation into the cavity on the jet airflow. We also noted a reasonable agreement between
experimental and numerical data concerning velocity profiles. Figure 2.12b shows that downstream
of the potential core (x/e > 3), the mean velocity profiles are Gaussian and become approximately self-
similar, especially in the free part of the air curtain flowing into the ambient. Due to the confinement
effect and the interaction with the recirculation into the cavity, differences are observed between
FIGURE 2.12 (a) Unnormalized and (b) normalized streamwise velocity profiles at the mid-plane (z/L = 0)
for unperturbed air curtains.
FIGURE 2.13 Half-width growth of the jet at the mid-plane (z/L = 0).
experimental and theoretical values related to the inner part of the air curtain developed into the
cavity. In addition, the good agreement with the theoretical formulation proposed by Reichardt [33]
justifies the assumption made by many authors to assimilate the air curtain to a plane free jet. All
the self-similar profiles conform closely to a Gaussian relation (Equation 2.1). Likewise, the stream-
wise variations (Figure 2.13) of the normalized velocity half-widths, y0.5/e, conform to the far-field
theoretical relationship (Equation 2.2). A quasi-linear regression approximation of this experimental
growth rate curve gives a value of the slope Ky = dy0.5/dx = 0.13, which is slightly higher than values
generally quoted in the literature [40]. This could be easily explained by the fact that our investiga-
tions concern the near field region, while the majority of studies in the literature concern the far field
region. We also notice that the RSM slightly underestimates the jet expansion.
2.8.1.2 Global Exchanges Through Air Curtain

Assuming heat and mass transfer analogy makes it possible to validate the global model by com-
paring numerical results with experimental data obtained from literature related to heat transfer
through air curtains. Figure 2.14a and 2.14b present comparisons between our model and experimental
T − Ts T − Ts Nu 1
or theoretical data related to TE = o and o = = respectively. Qualitatively,
T0 − Ta Tc − Ta Re Pr TCE
a good agreement between numerical and experimental data is observed for H/e = 10 and 20. The
differences were on the same order of magnitude as those related to experimental data provided
by the different authors. The results confirm the quasi-proportionality of exchanges through the
air curtain in regard to the H/e ratio. An increase of H/e from 10 to 20 entrains an increase of
 To − Ts   To − Ts 
 T − T  and  T − T  from 0.16 and 0.32 to 0.32 and 0.34 respectively. Two times higher values
0 a c a
for these quantities mean that the jet is two times less efficient in terms of energy loss and thermal
sealing because of the higher amount of infiltrated ambient air.
In addition to these quantities, the global model developed in this study permits quantification of
the various dimensionless fluxes exchanged between all compartments for H/e = 10 and 20 (Table
2.3). Overall, predicted values confirm the quasi-symmetry of the air fluxes exchanged from both
sides of the air curtain. (oc ≈ oa ≈ cs ≈ as). However, this symmetry is less more pronounced for
e = 0.02 m (H/e = 20) than for e = 0.04 m (H/e = 10).
As indicated above, these results confirm the importance of the H/e parameter on air fluxes across the
air curtain. For the studied configuration, the air flux exchanged between the ambient and the cavity “ac”
is twice as important for H/e = 20 than for H/e = 10, where it increases from 0.16 to 0.315. Conversely,
increasing H/e causes a reduction of the “os” flux by 50% because of the infiltrated ambient air.
To − Ts T − Ts
FIGURE 2.14 Evolution of (a) and (b) o vs H/e: Comparisons between our model and experi-
T0 − Ta Tc − Ta
mental or theoretical data quoted in literature.
TABLE 2.3
Air Fluxes Obtained with the Global Model for Unperturbed Jet
(e; U0; H/e; Re) ac oc os cs as ca oa
(0.02; 3; 20; 8.103) 0.0315 0.25 0.45 0.287 0.263 0 0.3
(0.04; 3; 10; 16.103) 0.016 0.16 0.68 0.16 0.16 0.016 0.16
2.8.2 With External Lateral Flow

2.8.2.1 Effect of ELF on Airflow Patterns
In an attempt to investigate the three-dimensional effects of the ELF on the air curtain, Figure
2.15 shows mean flow velocity vector fields for perturbed air curtain at Ulf = 1 m/s in four planes:
0.05 m, 0.1 m, 0.25 m and 0.4 m from the lateral side where the ELF comes in, which correspond
to z/L = 0.4, z/L = 0.3, z/L = 0 and z/L = –0.3 respectively. The results related to PIV, RSM and
k-ɛ are depicted in three columns (a, b and c, respectively) and ordered from top to bottom in the
FIGURE 2.15 Airflow patterns in x-y plane at different z-locations, for perturbed air curtain (U0 = 3 m/s and
Ulf = 1 m/s): comparison between (a) PIV, (b) RSM, and (c) k-ɛ model.
ELF/lateral direction. This representation enables direct comparisons between numerical results
obtained with RSM and k-ɛ with the PIV data for the different locations.
The results clearly show a strong 3D effect along the axis of the ELF. In comparison with Figure
2.10, which could be considered as a reference case (without perturbation), the results of Figure 2.15
suggest that the highest effect occurred at the lateral side of the air curtain, which corresponds to
the beginning of the interaction between the air curtain and the lateral flow (z/L = 0.5). Among the
presented data, Figure 2.15a (z/L = 0.4) exhibits the strongest effect of the external perturbation on
the air curtain. As can be seen, the jet is completely deflected towards the outside and the cavity
is locally unsealed. In this plane, the jet deviates immediately downstream of the exit nozzle and
breaks at x/e = 0.6 approximately, which in turn prevents the formation of the recirculation within
the cavity.
With increased distance downstream of the lateral edge of the jet, the effect of the perturbation
on the air curtain is progressively reduced. At z/L = 0.3 (Figure 2.15a), the jet conserves its vertical
trajectory until x/e = 2, when it begins to deviate. In the lower part (2 < x/e < 10), the jet exhibits a
strong curvature. The interaction with the lateral flow enhances the lateral spread of the jet, espe-
cially at the bottom level (x/e = 10). Therefore, a great part of the air curtain is deflected towards
the outside and the cavity is partially sealed. By increasing the jet deflection, the air spilling into
the cavity is reduced and the corresponding global recirculation is less active and entrained to the
bottom. For actual RDC, this behavior for the air curtain is highly undesirable because it amplifies
the ‘cold feet’ effect; a much larger mass of air from the cold air curtain overspills the bottom of
the cabinet [9].
At the mid plane (z/L = 0) no noticeable effects due to the ELF were observed on the jet behavior,
which conserves its stability (Figure 2.15a). As a consequence, the airflow related to the air curtain
and to the recirculation within the cavity is very close to the baseline. Further downstream (z/L =
–0.3) the jet conserves its stability but it is slightly deviated inside the cavity.
Due to the ELF effect, experimental (Figure 2.15a) and RSM results (Figure 2.15b) confirm the
presence of an external vortex loop (EVL), developing in the outer shear layer of the jet in the ambi-
ent side. This EVL, which could be viewed as an additional secondary stress induced ELF-flow with
swirl effects, results in a complex shear flow instead of simple free shear layer when compared with
an unperturbed jet. This EVL strongly interacts with the air curtain, implying a substantial increase
of streamline curvature associated with a strong 3D effect along the air curtain in the lateral direc-
tion. As can be seen, this EVL located initially near the jet inlet nozzle at z/L = 0.4 is convected
downstream along the ELF direction and thus it moves progressively to the bottom area of the jet
at z/L = –0.3. For actual RDC configurations, this EVL induced by the ELF would enhance mix-
ing processes between refrigerated air curtain and warm ambient, implying an increase of energy
consumption and air temperature in the conservation zone above the higher limit for the proper
conservation of food products.
2.8.2.2 Turbulence Modeling Performance

Despite the complexity of airflows, the results obtained with the RSM (Figure 2.15b) show a rea-
sonably good agreement between numerical and experimental data. This concerns jet deviation and
stability, airflow patterns and their evolution under the ELF. The same figure also shows the poor
predictions given by the k-ɛ model (Figure 2.15c) and underlines its inability to predict the general
behavior of air motion related to the air curtain sheared laterally. At z/L = 0.4, when experimental
and RSM results show a jet deflected outside the cavity, the k-ɛ predicts a jet deflected inwards.
Conversely, when experimental and RSM results show a slight deflection of the jet inside the cavity
at z/L = –0.3, the k-ɛ predicts a slight jet deflection outwards. In addition, the k-ɛ overestimates the
stability of the jet at z/L = 0.3 in comparison with PIV results.
The failure of the k-ɛ model predictions could be explained by the complexity of the airflow
resulting from the interaction between the air curtain and the ELF, implying a complex shear layer
instead of a simple free shear layer including a high streamline curvature effect. For these complex
flows, different authors [35,42,43] agree on the inadequacy of k-ε model to predict airflow patterns
and underline its limitation by comparison with experimental data. Gibson and Rodi [44] showed
that the standard k-ε model lacks sensitivity to curvature in contrast to the full Reynolds stress
model. Leschziner and Rodi [45] derived after some simplifications the expression of the Cµ coef-
ficient in the Reynolds stress transport equations in curvilinear coordinates. Cµ is not a constant, as
in the standard k-ε models, but depends strongly on streamline curvature. In a comparative review,
Nallasamy [43] concludes that the use of RSM models to account for curvature effects and second-
ary flows would improve the confidence in turbulence closure models. Leschziner [46] pointed out
that k-ɛ model does not respond correctly to curvature strain, normal straining and rotation and is
more appropriate for flows in which a single shear stress is the dominant dynamic link between
turbulence and the mean flow. A similar observation was formulated by Bsara and Jakirlic [47],
who consider that even if the performance of k-ɛ model could be regarded as acceptable in the flows
dominated by the mean shear, it fails to capture important flow features in most of the flows depart-
ing from the thin shear layer approximation: flows induced by streamline curvature, swirling flows,
flows with stagnant regions, etc.
The poor results given by the k-ɛ model could also be explained by the fact that it’s more appro-
priate for fully developed isotropic turbulence, which is not the case with anisotropic K–H vortices
due to their similar topologies and alignment of their rotational axis in the inner field region (Figure
2.4). The additional rotating, shear and swirling effects due to streamline curvature and the EVL
impose a strong anisotropy on the eddy viscosity field. Eddy viscosity has no longer an isotro-
pic, scalar nature as enforced by the k-ɛ formulation, but a tensorial one. To better illustrate the
anisotropic effect in the inner field region, Figure 2.16 shows the Urms and Wrms (root mean square
velocity fluctuation in x and z directions: streamwise and lateral directions) profiles obtained with
LDV in the mid plane (z/L = 0) at x/e = 2 without and with ELF (Ulf = 1 m/s). Without perturbation,
a strong anisotropy is observed between streamwise and lateral components. This concerns outer
shear layers in the ambient and the cavity sides as well as the peak values, where Urms values are
almost 60% higher than Wrms values. Under the ELF effect, the Urms and Wrms values become closer,
especially in the outer shear layer in the ambient side, due to the direct influence of the ELF. The
ratio between the peak values decreased to 30% for the outer shear layer in ambient side and to
53% in the cavity side. Obviously, the ELF enhances the isotopy of turbulence but with vanishing
FIGURE 2.16 Effect of the ELF on RMS velocity fluctuations (LDV) in the mid plane (z/L = 0) at x/e = 2:
comparison between (a) unperturbed and (b) perturbed air curtain (Ulf = 1 m/s).
effect from the ambient side where the ELF flows to the cavity side. These strong modifications
affecting the behavior of normal stresses, including attenuation and amplification of the turbulence
anisotropy, reflect the importance of the pressure-strain term (Equation (2.10)), which controls the
redistribution of turbulence energy among the normal stresses in the RSM. Conversely, the k-ɛ does
not resolve normal-stress anisotropy.
In spite of comprising two principal directions (X: related to air curtain and Z: related to ELF),
this configuration displays all features of a three-dimensional flow, necessitating solution of the
transport equation for all three velocity components and all six Reynolds stress components.
Note that the orientation of the jet deflection occurs in the Y direction, where no dominant mean
flow is present. As a consequence, the wrong predictions given by the k-ɛ model concerning the jet
deflection in the Y direction only result in the selective influence of the anisotropic Reynolds-stress
field on the mean flow. This point highlights the superiority of the RSM, as it only enables resolving
Reynolds-stress anisotropy.
All these aspects justify the use of the second moment closure instead of the two equation tur-
bulence models in this near field region (x/e <10) where the effect of anisotropy of turbulence and
streamline curvature are pronounced. For all these reasons, the remaining simulations are only
performed with the RSM.
2.8.2.3 Effect of the ELF on the Jet Deflection

To better illustrate and quantify the effect of the ELF on the jet deflection, Figure 2.17 shows the
locations of the maximum streamwise velocity (Um) along the cabinet in the lateral direction at three
downstream positions: x/e = 3, 5 and 8 for Ulf = 1 m/s obtained with LDV.
Due to the ELF effect, the air curtain loses its 2D character and experiences a 3D distortion in
the lateral direction. At the beginning of the interaction with the ELF (z/L= 0.4), the jet undergoes a
strong deviation outwards. In Figure 2.10, the missed values of Um positions related to higher down-
stream x/e locations indicate that the air curtain is completely broken on its bottom side near the
nozzle exit. At the end of the interaction with the external perturbation (z/L = –0.4), the air curtain
is deviated inwards.
As can be seen, the higher jet deviations are located on the cabinet sidewalls. This result is in
accordance with the findings of [6,0]. These strong jet deviations reflect a complex 3D flow within
the cavity and also between the cavity and the ambient. To better illustrate these induced flows in
the lateral direction, Figure 2.18a and 2.18b, show numerical results of velocity vectors and their
FIGURE 2.17 Locations of the maximum streamwise velocity (Um) along the cavity in the lateral direction
at x/e = 3, 5, and 8.
FIGURE 2.18 (a) Velocity vectors and (b) dimensionless temperature (T*) levels obtained with RSM for
perturbed air curtain (Ulf = 1 ms–1) at x/e = 8.
T − T0 *
corresponding dimensionless temperature (T * = , T0 = 0 and Ta* = 1) fields, respectively, in a
Ta − T0
horizontal plane at x/e = 8. It thus appears that the break-up of the air curtain on the lateral cabinet
side where the ELF comes in entrains air from the ambient inside the cavity, implying high tempera-
ture levels in this area. For actual RDC, these intensive convective exchanges induced by the ELF
from the warm ambient into the cavity would increase the energy consumption and food tempera-
tures within the chilled compartment.
2.8.2.4 Effect of ELF on the Jet Decay

Figure 2.19 compares the decay of the maximum streamwise velocity (Um /U0)2 in the mid plane
(z/L = 0) with and without external perturbation. Despite some scatter related to perturbed jet
curves, a consistent trend emerges. In fact, Figure 2.19 revealed that downstream of the potential
2
U   x 
core, the well-known relationship  m  = K u  x + 0  related to free jets appears to be also
 0
U e e
valid for perturbed jets.
FIGURE 2.19 Decay of the maximum streamwise velocity (Um /U0)2 in the mid-plane (z/L = 0) for perturbed
and unperturbed air curtains.
The values of Ku related to Ulf /U0 = 0, 0.16 and 0.33 were obtained by a least squares fit to the
data as 0.15, 0.24, and 0.43, respectively.
Figure 2.19 also shows that external perturbation does not affect the length of the potential core.
However, even given the fact that the air curtain conserves its local stability (z/L = 0), a strong effect
is observed in the transient region located downstream of the potential core, where high values of Ulf
entrain higher rates of decay of the maximum streamwise velocity. This could be partially explained
by the effect of the extra shear induced by the ELF on the formation of discrete K–H vortices that dom-
inate the early stage of wall jet transition process and play a relevant role in the jet entrainment inwards.
2.8.2.5 Effect of ELF on the Velocity Profiles

As for the unperturbed jet, Figure 2.20a and 2.20b depict the unnormalized and normalized stream-
wise velocity profiles obtained in the mid-plane (z/L = 0) with Ulf = 1 m/s at selected downstream
locations. Figure 2.20a compares numerical and experimental results, while Figure 2.20b compares
experimental data with the theoretical profile obtained by Gaussian-exponential function (Equation
(2.1)). For Ulf = 0.5 m/s, comparisons between normalized experimental data and the theoretical
profile are shown in Figure 2.21.
For Ulf = 0.5 m/s, the air curtain conserves the self-similarity profiles and despite some deviation,
the Gaussian trend given by a free jet for velocity profiles is also conserved.
For Ulf = 1 m/s, the mean velocity profiles exhibit asymmetrical behavior. In the inner part (cavity
side), the jet conserves the self-similarity profiles given by a free jet. On the contrary, in the outer part
(ambient side), the self-similarity profiles were partially conserved and limited to the core of the jet
(–1 < y/y0.5 < 1). To some extent, the shear effect exerted by the external flow is limited to jet boundar-
ies dominated by K–H instabilities and does not affect the jet core where higher velocities are present.
Figure 2.20b also shows that the biggest deviation from the theoretical profile given by a free jet
is observed at x/e = 5, which corresponds to an external circulation (Figure 2.15a) induced by the
ELF. However, this local perturbation does not affect the capacity of the jet to recover the theoreti-
cal profile further downstream, as can be seen at x/e = 7.
Concerning velocity profiles (Figure 2.20a), we could notice that higher differences between
numerical and experimental data are located at the outer edges of the jet dominated by second-
ary flows (recirculation within the cavity and vortices induced by shearing flow in the outer side).
Such flows are difficult to predict accurately and cause noticeable differences in velocity profiles.
FIGURE 2.20 (a) Unnormalized and (b) normalized streamwise velocity profiles at the mid-plane (z/L = 0)
for perturbed air curtains (Ulf = 1 m/s).
FIGURE 2.21 Normalized streamwise velocity profiles at the mid-plane (z/L = 0) for perturbed air curtains
(Ulf = 0.5 m/s).
However, differences are still within 10% and could be also explained by the complexity of the
three-dimensional effect induced by the external flow on the air curtain.
2.8.2.6 Effect of ELF on the Global Exchanges Through Air Curtain

In order to illustrate the effect of the ELF on the air curtain, Figures 2.22 and 2.23 present the
To − Ts Nu To − Ts
evolution of air fluxes (“ac” and “os”) and thermal parameters ( = , TE = )
Tc − Ta Re Pr T0 − Ta
related to air curtain efficiency as a function of the ratio Ulf /U0. The predicted results of these quan-
tities exhibit a consistent dependence on two dimensionless parameters: H/e and Ulf /U0. The first
is related to the air curtain aspect ratio and the second is related to the external perturbation and
expresses the ratio of the destabilizing to the stabilizing velocity related to the jet.
Overall, predicted results of Figure 2.22 clearly show the great influence exerted by external
perturbations on air fluxes exchanged across the air curtain. However, air fluxes evolve in different
manner. In Figure 2.22a, predicted values of ac underline the extreme sensitivity of this air flux on
H/e especially for small external perturbations ratio. As can be seen, for H/e = 10, ac experiences a
plateau until Ulf /U0 < 0.3. Conversely, no plateau was found for H/e = 20 for the same Ulf /U0 values.
This could be partially explained by the shorter potential core of H/e = 20, for which the effect of
the external-flow becomes more pronounced than for H/e = 10. Consequently, ac increases more
rapidly with Ulf /U0 for H/e = 20 than for H/e = 10. Concerning os air flux (Figure 2.22b), predicted
FIGURE 2.22 Evolution of air fluxes: (a) ac and (b) os for perturbed air curtains.
To − Ts Nu T − Ts
FIGURE 2.23 Evolution of (a) = and (b) TE = o for perturbed air curtains.
Tc − Ta Re Pr T0 − Ta
values corresponding to H/e = 20 are 30 % lower than those obtained with H/e = 10. However, the
rate of decreasing of os values as a function of external perturbations appears to be almost indepen-
dent on H/e values.
T − Ts Nu
The results of o = (Figure 2.23a) show a similar trend as that observed for ac curves
Tc − Ta Re Pr
related to H/e = 10 and 20 and could be explained by the same mechanisms. However, the slope for
H/e = 20 is 2.6 times higher than that related to H/e = 20. In Figure 2.23b, the two TE curves related
to H/e = 10 and 20 exhibit linear increase versus Ulf /U0. According to predicted results, increasing
Ulf /U0 from 0 (unperturbed air curtain) to 0.3 entrains:
• An increase of TE by a factor of 68% and 65% for H/e = 10 and 20 respectively.

T − Ts Nu
• An increase of the energy loss o = by a factor of 45% and 245% for H/e = 10
Tc − Ta Re Pr
and 20 respectively. These same factors could also be used to quantify the decrease of TCE
of the air curtain under the effect of the ELF.
TABLE 2.4
Air Fluxes Obtained with the Global Model for H/e = 10 as a Function of Ulf
Ulf (m/s) ac oc os cs as ca oa
0.5 0.017 0.19 0.62 0.19 0.19 0.017 0.19
0.7 0.018 0.22 0.56 0.22 0.22 0.018 0.22
1 0.019 0.24 0.49 0.23 0.27 0.015 0.26
1.2 0.04 0.25 0.40 0.26 0.34 0.02 0.36
1.5 0.07 0.22 0.35 0.27 0.38 0.017 0.43
1.8 0.1 0.24 0.28 0.32 0.39 0.018 0.47
2 0.11 0.24 0.28 0.32 0.40 0.020 0.49
TABLE 2.5
Air Fluxes Obtained with the Global Model for H/e = 20 as a Function of Ulf
Ulf (m/s) ac oc os cs as ca oa
0.5 0.047 0.315 0.40 0.320 0.278 0.04 0.29
0.7 0.07 0.345 0.284 0.412 0.305 0.003 0.37
1 0.12 0.48 0.270 0.38 0.35 0.22 0.25
1.2 0.15 0.51 0.28 0.37 0.34 0.19 0.20
1.5 0.16 0.50 0.25 0.44 0.31 0.22 0.25
1.8 0.22 0.62 0.21 0.53 0.25 0.3 0.167
2 0.24 0.66 0.18 0.56 0.25 0.32 0.153
As in the unperturbed air curtain, the global model was used to quantify the various dimension-
less fluxes exchanged between all compartments for H/e = 10 (Table 2.4) and H/e = 20 (Table 2.5).
We can notice that the symmetry of fluxes is highly dependent on the H/e ratio. For H/e = 20, this
symmetry is well conserved until Ulf <= 0.7 m/s and vanishes progressively beyond this value. For
H/e = 10, the symmetry of fluxes is less pronounced until Ulf <= 0.5 m/s and really breaks beyond
this value.
2.9 CONCLUSIONS
Experiments and CFD simulations were performed on a reduced-scale model of a RDC (1:5) under
isothermal conditions to investigate the aerodynamic behavior of an air curtain (Re = 8000) sheared
laterally by an external perturbing flow induced at a uniform velocity (Ulf). This study focuses on
the near field region (x/e < 10) where strong interactions are expected between the jet core, K–H
instabilities and external lateral stream (ELS). Three configurations: Ulf = 0 ms–1 (Unperturbed air
curtain), Ulf = 0.5 ms–1 and Ulf = 1 ms–1 were investigated and compared. LDV and PIV measure-
ments were used to investigate the air flow characteristics.
Due to the ELF effect, the air curtain loses its 2D character and experiences a 3D distortion in
the lateral direction. The experimental results obtained with PIV clearly show that the highest effect
occurred at the lateral side of the air curtain, which corresponds to the beginning of the interaction
between the air curtain and the lateral flow. At z/L = 0.4, the jet is completely deflected towards the
outside and the cavity is locally unsealed. With increased distance downstream the lateral edge of
the jet, the effect of the perturbation on the air curtain is reduced. At the mid plane (z/L = 0), no
noticeable effects due to the external stream were observed on the jet behavior, which conserves
its stability. Further downstream (z/L = –0.3), the jet conserves its stability but is slightly deviated
inside the cavity.
Two complementary modeling approaches were performed. The classical CFD modeling approach
was built with the k-ε model and the RSM. Comparisons between CFD numerical results and experi-
mental data clearly show that k-ε model significantly overpredicts the length of the potential core, and
therefore fails to predict the jet centerline decay in the transition zone. On the contrary, a good agree-
ment is observed with the RSM predictions concerning jet decay, jet expansion, velocity profiles,
airflow patterns and their evolution under the external lateral perturbation. Then, the CFD model/
RSM was used to build an original global model that enables the prediction of fluxes exchanged
between the air curtain, the ambient and the cavity. Assuming heat and mass transfer analogy, the
global model was validated by comparing numerical results with experimental data obtained from
the literature related to heat transfer through air curtains operating in no perturbed ambient.
To analyze the efficiency of the air curtain with and without external perturbation, two dimen-
T − Ts
sionless parameters were derived in the global model. The first: TE = o represents thermal
T0 − Ta
To − Ts Nu
entrainment and the second: = represents the ratio between the energy leakage and
Tc − Ta Re Pr
the corresponding thermal confinement level. The behavior of these two parameters exhibits a con-
sistent dependence on H/e and Ulf /U0. The results confirm the quasi-proportionality of exchanges
through the air curtain in regard to the H/e ratio. An increase of H/e from 10 to 20 entrains an
T − Ts T − Ts
increase of o and o from 0.16 and 0.32 to 0.32 and 0.34 respectively. Two times higher
T0 − Ta Tc − Ta
values for these quantities mean that the jet is two times less efficient in terms of energy loss and
thermal sealing because of the higher amount of infiltrated ambient air. For perturbed air curtains,
the results show that increasing Ulf /U0 from 0 (unperturbed air curtain) to 0.3 entrains:
• An increase of TE by a factor of 68% and 65% for H/e = 10 and 20 respectively.

T − Ts Nu
• An increase of the energy loss o = by a factor of 45% and 245% for H/e = 10
and 20 respectively. Tc − Ta Re Pr
The global model was also used to quantify the various dimensionless fluxes exchanged between
all compartments with and without external perturbation. The results confirm the importance of the
H/e parameter and Ulf /U0 on air fluxes across the air curtain. For unperturbed air curtains, the air
flux exchanged between the ambient and the cavity ac representing the airtightness of the air curtain
is twice as important for H/e = 20 than for H/e = 10, where it increases from 0.16 to 0.315. Similarly,
increasing H/e causes a reduction of the os flux by 50% because of the infiltrated ambient air. For
perturbed air curtains, the global model clearly shows the great influence exerted by external per-
turbations on air fluxes exchanged across the air curtain. The air flux ac increases more rapidly with
Ulf /U0 for H/e = 20 than for H/e = 10. Concerning os air flux, predicted values corresponding to
H/e = 20 are 30% lower than those obtained with H/e = 10.
NOMENCLATURE
C Dimensionless mass fraction (-)
Cp Specific heat of air (J kg–1 K–1)
DH Hydraulic diameter of inlet (m)
e Nozzle inlet width (m)
H Height of the cavity (m)
ui2
I Turbulence intensity ( Ii = ) (%)
Ui
k Kinetic energy of turbulence (m2 s–2)

Ky Jet-spreading rate (-)
L Nozzle inlet length (m)
m 0 Mass flow rate of the jet (kg s–1)
Nu Nusselt number, Nu = hd/λ
p Static pressure (Pa)
Cp µ
Pr Prandtl number, Pr = (-)
λ
Re Reynolds number, Re = ρU e/µ (-)
T Temperature (K)
T0, Ts and Ta Temperatures of the jet inlet, jet outlet and the ambient (K)
T − T0
T* Dimensionless temperature = T * = (-)
Ta − T0
ui u j Reynolds stresses component (m2 s–2)
Ui, ui Mean and fluctuating velocity component in xi direction (m s–1)
Urms, Wrms Streamwise and lateral RMS velocity: ux2 and ux2 (m s–1)
Vc Volume of the cavity (m3)
y1/2 Jet half-width, calculated at the y-location at which U(x) = ½ Um (x)
x,y,z Streamwise (x), transverse (y) and lateral (z) directions in Cartesian
coordinates (m)
Greek symbols
ε Turbulent dissipation rate (m2 s–3)
δij Kronecker symbol
μ Laminar dynamic viscosity (Pa s)
μt Turbulent viscosity (Pa s)
λ Thermal conductivity (W m–1 K–1)
v Kinematic viscosity (m2 s–1)
ρ Density (kg m–3)
Subscript
⊥ Normal
0 Relative to inlet boundary condition
a Ambient
c Cavity
cl Centerline
lf Lateral flow
m Maximun
i, j, k Relative to coordinate system
x, y, z Relative to coordinate system
Acronyms
ELF External lateral flow
EVL External vortex loop
K–H Kelvin–Helmholtz
LDV Laser doppler velocimetry
PIV Particle image velocimetry
RDCs Refrigerated display cabinet

RMS Root mean square
RSM Reynolds stress model
REFERENCES
1. Foster A.M., Barrett R., James S.J., Swain M.J., Measurement and prediction of air movement through
doorways in refrigerated rooms, International Journal of Refrigeration, 25: 1102–1109, 2002.
2. Foster A.M., Swain M.J., Barrett R., D’Agaro P., Ketteringham L.P., James S.J., Three-dimensional
effects of an air curtain used to restrict cold room infiltration, Applied Mathematical Modelling, 31(6):
1109–1123, 2007.
3. Hu S.C., Chuah Y.K., Yen M.C., Design and evaluation of a minienvironment for semiconductor manu-
facture processes, Building and Environment, 37: 201–208, 2002.
4. Rivera J., Elicer-Cortés J.C., Pavageau M., Turbulent heat and mass transfer through air curtains devices
for the confinement of heat inside tunnels, International Communications in Heat and Mass Transfer,
38: 688–695, 2011.
5. Guyonnaud L., Solliec C., Dufresne de Virel M., Rey C., Design of air curtains used for area confine-
ment in tunnels, Experiments in Fluids, 28(4): 377–384, 2000.
6. Gaspar P.D., Carrilho Gonçalves L.C., Pitarma R.A., Experimental analysis of the thermal entrainment
factor of air curtains in vertical open display cabinets for different ambient air conditions, Applied
7. Laguerre O., Hoang M.H., Osswald V., Flick D., Experimental study of heat transfer and air flow in a
refrigerated display cabinet, Journal of Food Engineering, 113: 310–321, 2012.
8. Amin M., Dabiri D., Navaz H.K., Tracer gas technique: A new approach for steady state infiltration rate
measurement of open refrigerated display cases, Journal of Food Engineering, 92(2): 172–181, 2009.
9. D’Agaro P., Cortella G., Croce G., Two- and three-dimensional CFD applied to vertical display cabinets
simulation, International Journal of Refrigeration, 29(2): 178–190, 2006.
10. Yu K.Z., Ding G.L., Chen T.J., Simulation of air curtains for vertical display cases with a two-fluid
model, Applied Thermal Engineering, 27: 2583–2591, 2007.
11. Cortella G., CFD-aided retail cabinets design, Computers and Electronics in Agriculture, 34: 43–66,
2002.
12. Cortella G., Manzan M., Comini G., CFD simulation of refrigerated display cabinets, International
Journal of Refrigeration, 24(3): 250–260, 2001.
13. Foster A.M., Madge M., Evans J.A., The use of CFD to improve the performance of a chilled multi-deck
retail display cabinet, International Journal of Refrigeration, 28(5): 698–705, 2005.
14. Smale N.J., Moureh J., Cortella G., Modelling of airflow in refrigerated food applications, International
Journal of Refrigeration, 29: 911–930, 2006.
15. Moureh J., Letang G., Palvadeau B., Boisson H., Numerical and experimental investigations on the
use of mist flow process in refrigerated display cabinets, International Journal of Refrigeration, 32:
203–219, 2009.
16. Ge Y.T., Tassou S.A., Hadawey A., Simulation of multi-deck medium temperature display cabinets with
the integration of CFD and cooling coil models, Applied Energy, 87: 3178–3188, 2010.
17. Navaz H.K., Henderson B.S., Faramarzi R., Pourmovahed A., Taugwalder F., Jet entrainment rate in air
curtain of open refrigerated display cases, International Journal of Refrigeration, 28(2): 267–275, 2005.
18. Moureh J., Ding Y., Flick D., Caractérisation expérimentale et numérique des transferts de chaleur à
travers un jet d’air plan horizontal, Revue Générale de Thermique, 35: 469–474, 1996.
19. Flick D., Moureh J., Ding Y., Confinement thermique d’une enceinte par jet d’air en convection force,
Revue Générale de Thermique, 36: 443–452, 1997.
20. United Nations Environment Programme (UNEP), Report of the Refrigeration, Conditioning and Heat
Pumps Technical Options Committee, 197, 2002.
21. James S.J., James C., The food cold-chain and climate change, Food Research International, 43: 1944–
1956, 2010.
22. Howell R.H., Adams P.A., Effects of indoor space conditions on refrigerated display case performance,
ASHRAE Research Project 596-RP, Department of Mechanical Engineering, University of South
Florida, Tampa, FL, USA, 1991.
23. Field B.S., Loth E., Entrainment of refrigerated air curtains down a wall, Experimental Thermal and
Fluid Science, 30: 175–184, 2006.
24. Bhattacharjee P., Loth E., Simulations of laminar and transitional cold wall jets, International Journal
of Heat and Fluid Flow, 25: 32–43, 2004.
25. Chen Y.G., Yuan X., Simulation of a cavity by a vertical single band cold air curtain, Energy Conversion
and Management, 1745–1756, 2005.
26. Everitt K.W., Robins A.G., The development and structure of turbulent plane jets, Journal of Fluid
Mechanics, 88: 563–583, 1978.
27. Youssef J., Étude expérimentale d’un jet plan turbulent se développant dans un flux uniforme en cocou-
rant, Thèse ISAE-ENSMA Ecole Nationale Supérieure de Mécanique et d’Aérotechique-Poitiers, 2012.
28. Kelso R.M., Lim T.T., Perry A.E., An experimental study of round jets in cross flow, Journal of Fluid
Mechanics, 306: 111–144, 1996.
29. Pathak M., Dewan A., Dass A.K., Computational prediction of a slightly heated turbulent rectangular
jet discharged into a narrow channel crossflow using two different turbulence models, International
Journal of Heat and Mass Transfer, 49: 3914–3928, 2006.
30. Cao Z., Gu B., Han H., Mills G., Application of an effective strategy for optimizing the design of air
curtains for open vertical refrigerated display cases, International Journal of Thermal Sciences, 49:
976–983, 2010.
31. Kaffel A., Moureh J., Harion J.L., Russeil S., Experimental investigation of a plane wall jet subjected to
an external lateral flow, Experiments in Fluids, 56: 1–19, 2015.
32. Howell R.H., Shibata M., Optimum heat transfer through turbulent recirculated plane air curtains,
ASHRAE Transactions, 2567, 1980.
33. Reichardt H., Zeitschrift für Angewandte Mathematik und Mechanik, 21, 1941.
34. Launder B.E., Spalding D.B., The numerical computation of turbulent flows, Computer Methods in
Applied Mechanics and Energy, 3: 269–289, 1974.
35. Wilcox D.C., Turbulence Modeling for CFD. DCW Industries, Inc., La Cañada Flintridge, CA, USA,
1994.
36. ANSYS Fluent 14.5 Theory Guide, ANSYS Inc., Canonsburg, PA, USA, 2012.
37. Hetsroni C., Hall C.W., Dhanak A.M., Heat transfer properties of an air curtain, Transactions of the
ASRAE, 328–334, 1963.
38. Costa J.J., Oliveira L.A., Silva M.C.G., Energy savings by aerodynamic sealing with a downward-blowing
plane air curtain—A numerical approach, Energy and Building, 38: 1182–1193, 2006.
39. Ding Y., Etude du confinement thermique d’une enceinte chaude par un jet d’air horizontal. These de
doctorat, Engref, Paris, France, 1994.
40. Deo R.C., Mi J., Nathan G.J., The influence of nozzle-exit geometric profile on statistical properties of
a turbulent plane jet, Experimental Thermal and Fluid Science, 32: 545–559, 2007.
41. Felis F., Pavageau M., Elicer-Cortés J.C., Dassonville T., Simultaneous measurements of temperature
and velocity fluctuations in a double stream-twin jet air curtain for heat confinement in case of tunnel
fire, International Communications in Heat and Mass Transfer, 37: 1191–1196, 2010.
42. Menter F.R., Eddy viscosity transport equations and their relation to the k-e model, ASME Journal of
Fluids Engineering, 119: 876–884, 1997.
43. Nallasamy M., Turbulence models and their applications to the prediction of internal flows: A review,
Computers and Fluids, 151–194, 1987.
44. Gibson M.M., Rodi W., A Reynolds-stress closure model of turbulence applied to the calculation of a
highly curved mixing layer, Journal of Fluid Mechanics, 103: 161–182, 1981.
45. Leschziner M.A., Rodi W., Computation of strongly swirling axisymmetric free jets, AIAA Journal,
22: 1742–1747, 1984.
46. Leschziner M.A., Modelling turbulent separated flow in the context of aerodynamic applications, Fluid
Dynamics Research, 38: 174–210, 2006.
47. Bsara B., Jakirlic S., A new hybrid turbulence modelling strategy for industrial CFD, International
Journal for Numerical Methods in Fluids, 42: 89–116, 2003.
3 CFD Modeling to Improve
the Performance of Industrial
Cooling of Large Beef Carcasses
Mulugeta A. Delele, Kumsa D. Kuffi,
Bart Nicolai, and Pieter Verboven
CONTENTS
3.1 Introduction..............................................................................................................................................62
3.2 Postmortem Metabolism and Meat Quality............................................................................ 62
3.3 Beef Carcass Cooling Methods................................................................................................... 63
3.3.1 Cold Air Cooling..................................................................................................63
3.3.2 Spray Cooling.......................................................................................................... 64
3.3.3 Vascular Perfusion Cooling..................................................................................... 65
3.3.4 Cooling Using High Thermal Conductivity Inserts...................................................... 65
3.3.5 Cooling Enhanced by Hot Boning...............................................................................66
3.4 Heat and Mass Transfer Mechanisms During Cold Air Cooling of Beef Carcass.................. 67
3.4.1 Surface Heat and Mass Transfer..................................................................................... 67
3.4.2 Heat and Mass Transfer Inside the Carcass..........................................................................67
3.5 CFD Modeling of Beef Carcass Cooling................................................................................ 68
3.5.1 Governing Equations....................................................................................................... 68
3.5.1.1 Cooling Air Side................................................................................................... 68
3.5.1.2 Carcass Side...................................................................................................... 68
3.5.1.3 Air–Carcass Interface.......................................................................................... 69
3.5.2 Model Parameters................................................................................................................ 70
3.5.3 Carcass Geometry....................................................................................................... 71
3.5.4 Simulation Procedure.......................................................................................................71
3.6 Cooling Air Flow Characteristics............................................................................................ 72
3.7 Cooling Characteristics of a Large Beef Carcass.................................................................... 73
3.7.1 Heat and Mass Transfer Coefficients........................................................................... 73
3.7.2 Carcass Temperature Distribution...................................................................................74
3.7.3 Carcass Glucose Level................................................................................................. 75
3.8 CFD Model Application in Assessing the Beef Carcass Cooling Process.............................. 76
3.8.1 Cooling Air Velocity Direction and Magnitude............................................................... 77
3.8.2 Cooling Air Temperature............................................................................................. 79
3.8.3 Evaporative vs. Convective Heat Flux............................................................................ 80
3.8.4 Cooling Room Design......................................................................................................81
3.9 Conclusions....................................................................................................................................81
Acknowledgments............................................................................................................................. 83
References......................................................................................................................................... 83
61
3.1 INTRODUCTION
For commercial purposes, beef is usually cooled at a carcass side stage before being processed to
the required size and shape of consumer meat portions. The meat industry thereby aims to produce
meat products that are hygienically safe, of good quality and visually attractive to consumers. Most
often the strategies used to achieve quality standards are associated with cooling [1].
During air blast cooling, carcasses are exposed to cooling air at a certain temperature, velocity
and relative humidity, all of which determine the cooling rate of the carcasses. The cooling rate in
turn affects weight loss, shelf life, rate of biochemical reactions taking place in the pre-rigor muscle
and final meat eating quality [2]. Rapid carcass cooling reduces weight loss, as the surface tempera-
ture quickly approaches the cooling air, which consequently reduces driving forces for evaporation.
However, a rapid chilling rate may accelerate sarcomere shortening in muscles and, thus, adversely
affect the eating quality of the meat. The rate at which carcasses are cooled also determines the rate
of biochemical reactions taking place in the muscle, which in turn affects the net flow of hydrogen
ions into the muscle, the variable on which the final meat quality mostly depends [3].
As both very fast and very slow cooling rates have side effects on the final quality of meat, mod-
ern slaughterhouses employ controlled cooling rate profiles to produce meat with required quality
traits in the shortest possible cooling time. Often, fast chilling methods during the initial hours of
the chilling process are followed by slow chilling methods. The exact advantages and appropriate
operational conditions of this approach are unclear.
This chapter will present the mechanisms of beef carcass cooling processes and the associated
challenges in terms of cooling efficiency and beef carcass quality. The use of computational fluid
dynamic (CFD) modeling techniques to understand the heat and mass transfer phenomena and to
improve the carcass cooling process at industrial scale will be discussed.
3.2 POSTMORTEM METABOLISM AND MEAT QUALITY

When the animal is slaughtered, it is subjected to a state of shock that acts upon the muscles as a
complex of nervous stimuli that triggers production and utilization of adenosine triphosphate (ATP)
during the death struggle. During muscle transition from rest to work (movement due to death
struggle), ATP turnover may exceed the normal production [4]. This requires the muscle to generate
new ATP. The major source of ATP production in postmortem is thus the glycolysis process, where
glycogen is used as primary substrate. Muscle glycogen is broken down and metabolized through
glycolysis to pyruvate.
ATP hydrolysis is believed to be the major source of proton load during the anaerobic muscle
metabolism [5], as shown in the following reaction:
ATP + H 2 O → ADP + Pi + H +

where Pi represents a phosphate ion. According to [6] the hydrolysis of ATP releases an enthalpy
of ΔH ATP = –134 kJ mol–1 Pi, which is the source of heat that can raise the temperature of beef
by 2.1–2.7°C above body temperature after slaughter [7]. This means that, apart from being the
main source of hydrogen ions, the hydrolysis of ATP also adds heat to the muscle, the overall
magnitude of which depends on the rate at which ATPase converts ATP into ADP and H+ in the
entire meat. Thus, two processes occur—namely, the production of heat and the net production of
hydrogen ions—during the postmortem cooling of carcasses. The two processes are also coupled
via temperature- and pH-dependent enzymatic reaction rates [2]. High muscle temperature favors
a fast rate of pH decline due to the effect of temperature on enzyme activity and subsequently the
rate of glycolysis [8].
Not only does postmortem metabolism lead to heat production and pH decline, high tempera-
ture and decreased pH can potentially change meat tenderness, water-holding capacity and color,
CFD Modeling to Improve the Performance of Industrial Cooling of Large Beef Carcasses 63
7.5
7 Cold
shortening
6.5 window
Muscle pH
6
Heat
5.5 shortening
window
5
40 35 30 25 20 15 10 5
Muscle temperature °C
FIGURE 3.1 The pH(t)—T(t) diagram used by Meat Standard Australia (MSA) to define the optimal decline
in pH relative to the temperature of the muscle. (From J. Thompson, “Managing meat tenderness,” Meat Sci.,
vol. 62, no. 3, pp. 295–308, Nov. 2002.) The solid line represents an optimal rate of decline of pH with tem-
perature that avoids the shortening conditions due to too slow (heat shortening) or too fast (cold shortening)
cooling. Typical suboptimal curves are shown by dotted curves.
primarily due to protein denaturation [2]. Often, one uses a T(t)—pH(t) diagram presentation to
describe the dynamic changes in the meat. Each point on a curve plotted in this diagram represents
a time point during cooling of the carcass at a specific position in the muscle (Figure 3.1), starting
from high T and pH. There are two windows in the T(t)—pH(t) diagram that need to be avoided
during the cooling process in order to produce a palatable meat [9]. These regions are identified as
heat and cold shortening windows. In Figure 3.1 they are shown in the bottom left and top right of
the figure, respectively.
3.3 BEEF CARCASS COOLING METHODS

Chilling of freshly harvested beef carcasses is essential not only for food safety but also to maintain
the required meat quality. The most commonly used methods for chilling of beef carcasses are:
3.3.1 Cold Air Cooling

Cold air cooling is the most commonly used method of beef carcass cooling (Figure 3.2). In cold
air cooling systems, the cooling rate and uniformity are determined by airflow rate, cooling air
temperature, relative humidity, carcass characteristics and a combination of these parameters. In
conventional cooling methods, i.e., when the cooling air temperature is maintained just above 0°C
and the velocity is kept around 0.5–0.6 m/s, cooling time usually exceeds two days to reduce the
temperature of the slowest cooling point to 7°C, which is the regulatory temperature for transport
of meat products in the EU. For many reasons, including weight loss, meat quality and shelf life
issues, but also logistics, a cooling time of two days is not recommended [10]. Therefore, systems
that accelerate the cooling rate of beef carcasses have been suggested.
It has been reported that precooling of the beef carcass using a rapid cooling system improves the
cooling rate and minimizes the quality deterioration. Such precooling systems decrease the carcass sur-
face temperature more quickly. This way, part of the sensible heat of the carcass can be removed before
moving the carcass into a conventional cooler and the surface of the carcass is dried to control as soon
as possible potential microbial issues. A low surface temperature creates a larger temperature gradient
inside the carcass that increases conduction heat transfer rates. Rapid cooling is an energy-consuming
process leading to high moisture loss rates and, in the case of cooling using negative air temperature,
freezing risks. Therefore, rapid cooling is usually used to extract only part of the initial heat load from
the hot carcasses before the carcasses are transported to a conventional chiller room [2,10,11].
FIGURE 3.2 Typical cold air carcass cooing system; the arrow shows the direction of cold airflow for hori-
zontal configuration.
Blast and deep cooling systems are types of rapid cooling systems that apply relatively high and
low cooling air velocity and temperature, respectively. In blast cooling condition, the air velocity
can reach up to 10 m/s in the pre-chilling stage [12], which can accelerate weight loss and surface
drying unless the air temperature is reduced proportionally. During deep cooling at low tempera-
ture, the cooling air temperature can be as low as –30°C or less [13]. At this low temperature,
surface freezing of the muscle can occur but surface freezing has positive consequence on meat
tenderness, as the hard surface acts as a restraint against the contraction process [10].
The CFD model has been used by different researchers to study cold air cooling mechanisms of
carcasses [12,14–17]. The models have been applied to study the effects of carcass geometry, cool-
ing air flowrate and temperature and the orientation of the carcass to the airflow direction on heat
and mass transfer characteristics.
3.3.2 Spray Cooling
Spray cooling is the process that involves automated intermittent spraying of cold water onto car-
casses during the first three to eight hours post-slaughter. The principal purpose of spray cooling is
to reduce carcass weight loss due to water evaporation from the surface [10,18]. The larger evapora-
tion rate also creates fast surface cooling, resulting in surface temperatures approaching the wet
bulb temperature, thus creating similar effects as rapid cooling while minimizing water loss. Based
on their experimental observation, Jones and Robertson [19] reported that spray cooling reduced
the temperature of round and loin sections of beef carcass by 1–2°C as compared to conventional
cooling after 24 h. Similarly, Brown [20] reported small difference in cooling time of spray cooled
legs and loins of lamb carcasses as compared to those conventionally chilled.
The effect of spray duration on weight loss was investigated by Jones and Robertson [19]. After
observing the decrease in weight loss of 0.48–0.89% when the spray duration was increased from 4
to 12 h, the authors reported that carcass shrinkage at 24 h was proportionally related to the length
of spray cooling treatments. It was also reported that spray cooling of carcasses has no effect on
the growth of microorganisms as compared to conventional air cooling [21]. With respect to color
of lean muscle, two different observations have been reported. According to Brown [22], the lean
muscle on the surface of spray-cooled carcasses was significantly more yellow than that of the con-
ventionally air cooled carcasses. However, Greer and Jones [21] reported a contrary observation that
spray cooling has no effect on color of lean meat. Both studies, however, agree on the effect of spray
cooling on fat color as it turns significantly more yellow after spray application.
Kuitche et al. [23] used a simplified mathematical analytical model to investigate the effect of
evaporative cooling on the temperature and weight loss kinetics of meat by assuming the carcass as
an infinite cylinder. Later, Issa [24] applied a two phase CFD model to study the effect of droplet
diameter, air to liquid loading, nozzle to surface distance and relative humidity during spray cooling
of food products, such as beef carcasses.
3.3.3 Vascular Perfusion Cooling

This less conventional method uses the vascular system as a conduit into the deep tissue so that
pre-cooled water or a mixture of pre-cooled water with a low concentration of salt can be deliv-
ered following carcass exsanguination [25]. The feasibility of vascular perfusion as a method for
carcass cooling was tested by Brown et al. [26] on a lamb carcass (Figure 3.3). The lamb carcass
was equipped with catheter and hose systems by which a coolant solution (in this case a solution
containing a very fine ice slurry, sodium chloride and water at –1.3°C) is circulated. According to
the study on lamb, a carcass cooled with vascular perfusion cooled faster during the initial hours
as compared to conventionally air cooled carcass. According to the study, vascular perfusion had
no effect on microbial count as compared to the conventional air cooling method when tested after
24 h. However, carcasses cooled with this technique gained weight and became lighter in appear-
ance when compared to conventionally air cooled carcasses. This technology is not yet tested in
the beef industry and CFD studies of such cooling process of beef carcasses are not available yet.
3.3.4 Cooling Using High Thermal Conductivity Inserts

The cooling of animal carcasses is limited by the rate at which heat can flow from the deep muscles
to the surface. High thermal conductivity inserts have been employed to improve the heat flow from
inside the carcasses to the surface [27]. These devices are inserted deep into the carcass, with the
free end protruding into cold air, to allow rapid heat removal from the slowest cooling point in the
carcass, which as a result accelerates the cooling rate. Heat tubes are sealed metal tubes contain-
ing a refrigerant, which transfer heat to the cooler end of the tube if there is a small temperature
difference between its ends. When the tube is inserted in a warm carcass, heat is extracted from
the muscle to evaporate the refrigerant and transferred to the cooler end of the tube that remains in
the cooled air. The refrigerant condenses and flows back in a wick to the other end and the cycle
FIGURE 3.3 Potential perfusion system for lamb. (From T. Brown, R. Ian Richardson, C. A. Wilkin, and
J. A. Evans, “Vascular perfusion chilling of red meat carcasses—A feasibility study,” Meat Sci., vol. 83, no. 4,
pp. 666–671, 2009.)
FIGURE 3.4 Application of heat pipe technology on beef carcass sides (right) and the corresponding con-
trol side (left). (From R. H. Jacob, D. T. Beatty, and R. D. Warner, “A preliminary study into the use of ‘heat
pipes’ to prevent high rigor temperature in beef carcasses by increasing cooling rate,” Anim. Prod. Sci.,
vol. 54, no. 4, pp. 504–509, 2014.)
continues until the required temperature is achieved. The high surface area used on the tube part
exposed to the cooling environment is just to increase the rate of heat transfer process.
Heat tubes were tested on beef carcasses to improve the cooling rate [27]. According to this
study, the time taken for the leg temperature to reach 35°C, measured at a depth of 100 mm, reduced
from 150 to 76 min by using heat pipes containing methanol as a refrigerant (Figure 3.4). Besides
improving the cooling rate, high thermal conductivity inserts are relatively easy to use, robust,
silent in operation and can last for many years without needing repairs or replacement. However,
concerns relate to ensuring the inserts are free from microorganisms before inserting them into the
carcasses during commercial production. Also, care should be taken to avoid physical damage to the
muscle due to the destructive nature of penetrating the carcass. CFD studies on the cooling of beef
carcasses using high thermal conductivity inserts are not available yet.
3.3.5 Cooling Enhanced by Hot Boning

As the name implies, hot boning is the removal of muscle from the skeleton while the carcass is still hot.
After removal, the muscle is vacuum packed and then stored at a relatively high temperature for some
hours followed by cooling in low temperature air until the required temperature is achieved [28]. As hot
boning is a well-established process, several studies have been conducted to figure out the advantages
and disadvantages as compared to conventional cooling systems [29,30]. Hot boning has the advantage
of reducing cooling time, weight loss and enhancing water holding capacity when compared to the
conventional cooling system. The authors also reported that hot boned muscle can be subjected to cold
shortening if delayed cooling or electrical stimulation practices are not exercised properly.
Removal of the muscle from the carcass during the pre-rigor phase allows more muscle contrac-
tion than muscles that conventionally cooled while attached to the bone [30], so additional facilities
for restricting excessive muscle contraction may be required if the practices of delayed cooling are
not exercised. There may also be the practical constraint of having the processing facilities close
to the abattoir, as transport of hot carcasses is not allowed. Recently Kuffi et al. [31] developed an
empirical model and reported that hot boning accelerates cooling and slows down the metabolism,
resulting in a decreased rate of pH decline. Detailed CFD-based studies of the cooling process of
beef carcasses linked to hot boning have not yet been presented.
3.4 HEAT AND MASS TRANSFER MECHANISMS DURING

COLD AIR COOLING OF BEEF CARCASS
During cold air cooling of beef carcass, heat is transferred from the carcass to the cooling air
by convection, evaporation and radiation. The next subsections discuss the heat and mass transfer
mechanisms at the surface and inside the beef carcass.
3.4.1 Surface Heat and Mass Transfer

The carcass surface is exposed to a cold air flow over the surface, which generates a boundary layer
that is the driving force for heat transfer by convection. The temperature reduction of the surface
in turn causes thermal diffusion from the inner muscle to the surface. Thus, surface temperature
is determined by the balance between heat taken away from the surface by cold air and that which
diffuses to the surface from inside the carcass.
At the start of the cooling process, the surface of the carcass is wet and the surface tempera-
ture is greater than the air temperature; hence, evaporation occurs. This cools the surface by
latent heat removal while moisture and heat transfer from the carcass to the cooling air takes
place by convection, the rate of which is proportional to the air moisture and temperature gradi-
ent between the surface and the air. As surface temperature and water activity drop, evaporation
minimizes and heat transfer becomes dominated by convection only. The convective transfer
coefficients for surface heat and mass transfer depend on the boundary layer around the car-
casses, which in turn depends on the carcass shape and dimension, the air flow rate, the orien-
tation of the carcass towards the approach flow, the turbulence characteristics of the approach
flow, the surface roughness of the product and the specific configuration (i.e., carcass arrange-
ment and placing in the chiller).
In industrial chillers, Kondjoyan and Daudin [32] observed the highest air velocity near the exit
of the circulation fan, and this high air velocity region is associated with a high level of turbulence
intensity. At other positions the air could be relatively stagnant (i.e., the velocity is small both in
magnitude and fluctuation) and the turbulence intensity is low. This heterogeneity in flow conditions
in cooling rooms can create variations in air velocity in the range of 0.5 to 5 m/s, while turbulence
intensity can range from 6 to 40%. Carcasses located at different positions in the chiller room
can experience different cooling rates. Transfer coefficients are affected by turbulence intensity
and velocity. The study also confirmed that the mean transfer coefficients are affected by turbu-
lence intensity and velocity differently on a pork hindquarter and on cylinders, and thus seem very
dependent on body shape. For complex shapes like beef carcasses, both mass and heat transfer
coefficients vary locally along the surface during cooling. This variation creates local variations in
values of surface water activity as well [16]. If these local variations are not considered during chill-
ing process design, potential areas for microorganism growth may be masked. For proper design of
carcass cooling process, therefore, accurate determination of local transfer coefficients is important.
However, owing to the large size and complex shape of beef carcasses, experimental determination
of local transfer coefficients could be difficult and, if feasible, only determined for specific condi-
tions. For this purpose, fluid flow simulation tools could be a means to predict those parameters
controlling the cooling process at local level.
As the surface temperature of the carcasses is higher than the wall temperature of cooler room
during most of the cooling time, consideration of radiation heat transfer is expected to be relevant
as well.
3.4.2 Heat and Mass Transfer Inside the Carcass

The rate of conductive heat transfer within the carcass is proportional to the thermal conductivity
and the temperature gradient. The temperature gradient is the driving force that determines the
direction and magnitude of heat flux, while thermal conductivity determines the rate at which the
temperature changes within the carcass. During beef carcass chilling, the temperature gradient is
initiated by the drop of temperature of the carcass surface, which drives the diffusion of heat from
within the carcass to the surface. In a similar way to the heat transfer, the moisture evaporated
from the surface causes diffusion of moisture from the inner muscle to the surface [1]. The balance
between water evaporated from the surface and that which diffuses to the surface from within the
carcass determines the surface water activity [16].
Since carcasses have a low thermal conductivity, the rate of heat conduction in the carcass is very
low. The carcass temperature during cooling thus varies both with time and position. Apart from
thermal conductivity, the heat transfer within the carcass depends on other thermophysical proper-
ties like the density and specific heat. The different carcass components (muscle, fat and bone) have
significantly different thermophysical properties. Fat, muscle and bones occupy distinct regions
inside the carcass that could influence the heat transfer process.
As the moisture content of beef carcass muscles is high and the diffusivity of the moisture in the
carcass is low, a moisture concentration gradient is expected to occur only near the surface [33]. In
most studies so far, however, internal mass transfer has been neglected [16].
3.5 CFD MODELING OF BEEF CARCASS COOLING

During the last two decades, researchers have proposed different models to predict heat and mass
transfer during the carcass cooling process [14–16,23,34]. The modeling techniques that are com-
monly used are empirical, analytical and numerical. Experimentally validated numerical models
such as CFD are reported as a method of choice to get a relatively accurate prediction of complex
heat and mass transfer processes of beef carcass cooling [35,36]. The models were developed based
on the equations for conservation of momentum, energy and mass.
3.5.1 Governing Equations
3.5.1.1 Cooling Air Side
On the air side, commonly the Reynolds-averaged Navier–Stokes (RANS) equations are solved.
The equations are based on the general conservation laws for continuity, momentum, energy and
species transport combined with an appropriate turbulence closure model [14,16,36–38].
3.5.1.2 Carcass Side
The governing equation of heat conduction inside the carcass is given by the Fourier equation with
heat production:
dTc
ρc c pc = ∇ ⋅ (λc ∇Tc ) + S (3.1)
dt
where ρc (kg m–3), cpc (J kg–1 K–1) and λc (W m–1 K–1) are the density, heat capacity and thermal conduc-
tivity of the carcass, respectively. Tc is the temperature of the carcass (K), and S (W m–3) is the volumet-
ric heat production rate due to the postmortem energy production in the muscles. Until recently, in most
published studies the source term S in Eq. (3.1) was either ignored or given a constant value. Recently,
Hamoen et al. [3] modeled the source term as a function of ATP production rate for slaughtered calves
and showed how the source term is dependent on the temperature profile within the carcasss. model-
ing Kuffi et al. [14] simplified this model and applied it in the modeling of large beef carcass cooling
processes. The rate of heat production per metric cube of muscle can be estimated as [6,39]:
S = −1.25ν G ∆H ATP (3.2)

where vG is the rate of glucose conversion (mol m–3 s–1) and ΔHATP (J mol–1) is the enthalpy change
due to heat production, which corresponds to ΔHATP = −134 kJ per mole of ATP hydrolyzed [6].
E  1 1 
Vmax [G ] ⋅ exp  a  − 
d [G ]  R  Tref T  
= νG = (3.3)
dt  [ H + ]3 
( K m + [G])  1 + K 
a
where [G] (mol m−3) is the unconverted glucose concentration, Vmax,ref (mol m−3 s−1) is the maximum
conversion rate of the reaction at reference temperature, Tref (K) is equal to 37°C, Km (mol m−3) is
the Michaelis–Menten constant, Ka (mol m−3) is the acid dissociation constant, Ea is the activa-
tion energy (kJ mol–1), R is the universal gas constant (J mol−1 K−1) and [H+] is the hydrogen ion
concentration.
The relation of pH fall to lactate (La) formation during postmortem is conventionally
expressed as the buffering capacity (BC), which is the slope of the curve relating the two param-
eters [3,14,40]:
d [ La ]
BC = ; pH = − log10 [ H + ] (3.4)
dpH
The production rate of hydrogen [H+] is related to the rate of glucose conversion, and by applying
stoichiometry rules between glucose and lactate:
dpH −ν La 2ν G 2 d [G ]
= = =− (3.5)
dt CB CB BC dt
3.5.1.3 Air–Carcass Interface
The effect of radiation, convection and moisture evaporation at the carcass surface during cooling
are modeled as boundary conditions. The conservation of energy at the carcass surface [14–16]:
λc ∇Tc ⋅ n = qconv + qevap + qrad (3.6)

where qconv, qevap, and qrad (W m–2) are the heat fluxes due to convection, evaporation, and radiation,
respectively. n is the unit normal. The convective heat transfer at the surface of the carcass:
qconv = hh (Ts − Ta ) (3.7)

where hh is the surface heat transfer coefficient (W m–2 K–1), Ts is carcass surface temperature (K)
and Ta is cooling air temperature (K).
The moisture transfer at the surface of the carcass is determined using the mass balance equation:
Dc ∇Cc ⋅ n = hm (Cs − Ca ) (3.8)

where Dc is the moisture diffusivity in the carcass (m2 s–1), hm is surface mass trasfer coefficient
(m s–1) and Cc, Cs, and Ca are moisture concentration in the carcass, on the carcass surface and air
around the surface (kg m–3), respectively. As moisture transfer in the carcass is often neglected, the
evaporative flux is calculated from the amount of moisture that is convected from the surface mul-
tiplied with the latent heat of evaporation (Lv) of water:
qevap = hm (Cs − Ca ) Lv (3.9)

Cs and Ca can be determined from an ideal gas law equation by assuming the water vapor as an ideal
p
gas [14,38], as: Cs ( a ) = vs ( a ) where Cs(a) is the moisture concentration at the carcass surface or air,
Rv T
Pvs(a) is the water vapor pressure (Pa) at the carcass surface or air, T is the temperature (K) and Rv
is the specific gas constant for water vapor (J kg−1K−1). The water vapor pressure at the surface of
the carcass and the air can be defined by assigning a water activity of the carcass multiplied with
the saturated pressure at a given temperature. The mass transfer coefficient hm is related to the heat
transfer coefficient hh using the Lewis relationship, where the airflow is assumed tubulent and the
Lewis number (Le) is close to 1:
hh
hm = 2
(3.10)
paC pa (Le) 3

λa
where Le is the Lewis number, Le = , ρa (kg m–3), cpa (J kg–1 K–1) and λa (W m–1 K–1) are the
paC pa Da
density, heat capacity, thermal conductivity and moisture diffusivity of air, respectively.
Radiation heat transfer can be calculated using:

( )
qrad = Fs→ wσε Ts4 − Tw4 (3.11)
where Tw is the wall temperature of the cooling room (K), σ is the Stefan–Boltzmann constant
(5.67 × 10 −8 W m−2 K−4) and ε is the radiation emissivity of the carcass surface. Fs→w is a view
factor (the proportion of radiation that leaves the carcass surface and reaches the wall, floor and
roof). When only a single carcass is placed in cooling room, the view factor will be 1. In a loaded
chiller room the view factor can vary between 0 < Fs→w < 1, depending on the arrangement and
the number of carcasses in the chiller room.
3.5.2 Model Parameters
Most prediction studies assumed the carcass as homogenous material and average thermophysical
properties are calculated from the values and mass fraction of the individual components [14–16].
Such crude assumption could decrease the accuracy of the models. For instance, when fat, which
has more insulating properties than muscle, is present as a layer on the surface of the carcass, the
homogeneous approach may lead to inaccuracies in the temperature prediction close to the surface.
For instance, Kuffi et al. [14] assumed 75.2% lean meat, 12.3% fat and 12.5% bone, on a weight
base for their cooling model of beef carcass, and the corresponding values of carcass density,
heat capacity and thermal conductivity were 1080 kg m−3, 3320 J kg−1 K−1 and 0.41 W m−1 K−1,
respectively.
Kuffi [31] measured the pH change of the beef carcass during cooling and fitted the data to
the kinetic models of the glycolysis (Eqs. 3.3–3.5) and calculated the relevant kinetic model
parameters. For instance, the calculated Vmax, K m, pKa and BC values are 0.0026 mol m−3 s−1,
0.6 mol m−3, 5.7 mol m−3, 55.1 mol m−3 pH unit−1, respectively.
3.5.3 Carcass Geometry
The cooling characteristics of the carcass during the cold air cooling process are affected by the
shape of the carcass [14,16,41]. To maximize the accuracy of CFD-based carcass cooling predic-
tions, it is important to have the correct geometrical shape of the carcass. During early days of
CFD modeling of beef carcass cooling, Davey and Pham [41] combined simple shapes of cylinders
and slabs to construct 2D geometry of a carcass. Later Trujillo and Pham [16] and Pham et al.
[15] used a method that was developed by Davey and Pham [42] to construct the 3D geometry of a
beef carcass in their CFD model. This method related the shapes and dimensions of the carcass to
the carcass mass and fat cover index by means of regression equations, and the carcass geometry
was formed from these two characteristics. These methods gave an approximation of the surface
structure of the carcass. Recently, Kuffi et al. [14] used a high-resolution 3D laser scan system to
generate a relatively accurate surface structure of the beef carcass. The laser scan generates 3D data
points on the surface of the carcass (3D point cloud), which were processed by reverse engineering
to obtain a solid surface formed from the 3D surface points (STL geometry file). This information
was used to generate a 3D CAD model for subsequent generation of the computational model, as
shown in Figure 3.5.
3.5.4 Simulation Procedure
The cooling of the carcass starts on the slaughter floor as soon as the skin is removed. It is important
to understand the initial cooling behavior inside the slaughter room. On the slaughter floor, car-
casses are exposed to natural convective flow, evaporation, long-wave radiation exchange and heat
production due to the glycolysis. Kuffi et al. [14] reported a 2 h stay in the slaughter room before
the carcass was placed in the cooler. They simulated this transient cooling process with an initial
body temperature of 40°C, which is slightly above the normal body temperature taking into account
FIGURE 3.5 Beef carcass geometry that is generated from a laser scan.
FIGURE 3.6 3D computational domain and boundary conditions for carcass cooling room simulations.
the rise in temperature due to muscle activity during death struggle. The water activity of the wet
carcass in the slaughter floor was assumed as 1.
After the slaughter room, the carcass is placed in a forced air cooling system. For an accurate
prediction it is important to know the cooling air flow rate, turbulence, temperature, relative humid-
ity, fan characteristics and room design (dimension and insulation). The boundary conditions that
were applied in the model of Kuffi et al. [14] are given in Figure 3.6 for the cooling of a single side in
horizontally approaching airflow. The beef carcass is taken as a solid domain. To accurately predict
the boundary layer of the surrounding cooling air, it is important to discretize its surface with the
appropriate mesh size. The accuracy of surface heat and mass transfer coefficient is highly depen-
dent on the size of the surface mesh. For instance, Kuffi et al. [14] used a relatively small surface
mesh size. After conducting a mesh sensitivity study, smallest mesh size of 0.625 mm on the carcass
surface was selected and then the size grew towards a larger mesh size away from the carcass. At the
carcass surface five layers of prismatic element were used. The model was solved using Shear Stress
Transport (SST) turbulence model. This turbulence model is recommended for a more accurate near
wall treatment [43–46].
3.6 COOLING AIR FLOW CHARACTERISTICS

The predicted airflow profile for a horizontal and vertical flow arrangements through multiple
carcasses is shown in Figure 3.7. In the modeled system, the velocity of the cooling air within
the stack of the carcass was lower than the surrounding free zone. In the horizontal arrangement,
FIGURE 3.7 Cooling air velocity distribution around a stack of beef carcasses in a cool store for horizontal
(left) and vertical airflow direction (right).
the first carcass blocked a large portion of the coming cooling air and the amount of cooling air
that reached the rest of the carcasses was relatively low. This shows that the design of cooling
room, stacking pattern of the carcasses and air flow direction affect the uniformity of the cooling
air velocity.
3.7 COOLING CHARACTERISTICS OF A LARGE BEEF CARCASS

3.7.1 Heat and Mass Transfer Coefficients
The local heat transfer coefficient during air cooling of beef carcass is highly variable; it varies from
one cross-section to another and even within a cross-section of the carcass [14,15]. In the model of
Pham et al. [15], the beef carcass was divided into twelve cross-sections and up to six subsections
and the local transfer coefficients were predicted. The local heat and mass transfer coefficients
were calculated using a regression equation in each section and subsection as a function of veloc-
ity and turbulence intensity. The equations were expressed in the form of dimensionless numbers
(Re, Pr, Nu and Sh); this could allow the equation to be applied over a wide range of temperature
and carcass sizes, for a vertical downward airflow. The predicted local heat transfer coefficient is
shown in Figure 3.8 for a horizontal approaching airflow. In this model a relatively higher cooling
air velocity and turbulence intensity was used (0.58 m s–1 and 45%, respectively). The variation of
local heat transfer coefficient across the carcass surface is large, namely by almost a factor 10, with
maximum values in regions that are exposed to the flow and minimum values in the body cavity
and regions that are less exposed to the normal flow direction. The variability and value of the heat
transfer coefficient was higher than that of the results of Pham et al. [15], it could be due to the dif-
ference in air flow velocity and turbulence intensity. The relatively accurate surface structure that
was used in this study (Figure 3.5) contributed to the relatively higher resolution of the local heat
transfer coefficient.
FIGURE 3.8 Convective heat transfer coefficient distribution on the carcass surface in the cooling room (air
velocity = 0.58 m s–1, TI = 45%). (a) outside of the body; (b) body cavity; red (15 W m–2 °C–1), blue (0 W m–2 °C–1).
3.7.2 Carcass Temperature Distribution

Experimentally validated carcass cooling models have been used to predict the spatial temperature
distribution inside the carcass [14–16,42,47,48]. A typical temperature distribution inside the carcass
during cooling is presented in Figure 3.9. Some of the models were based on wind tunnel cooling
studies [15,16,48], whereas others have explored conditions in real industrial beef carcass coolers
[14,15]. Pham et al. [15] found a relatively higher prediction error in the model of an industrial car-
cass cooler compared to the wind tunnel study. Uncertainties in the position of the thermocouples,
uncertainties in air velocity and turbulence and interference of the neighboring beef carcass were
reported as possible causes of the error.
For an accurate prediction of the carcass temperature distribution during cooling, the initial tem-
perature distribution of the carcass that comes from the slaughter floor should be known. Kuffi et al.
[14] developed and validated a model of the cooling in the slaughter floor and observed a difference
in temperature between the different sections and depth of the carcass. The temperature at 2 cm
from the surface of the carcass decreases soon after slaughter while the temperature at deeper loca-
tions (18 cm) inside the carcass continues to increase. These differences in temperature distribution
within the carcass are induced by heat generation due to glycolysis and conductive resistance, while
heat is removed by convection, evaporation and radiation at the surface. Deep inside the carcass, the
heat generation exceeds the amount that is diffused towards the surface.
Kuffi et al. [14] used the prediction of the slaughter floor temperature as an initial condition
Comparison of calculated and measured deep meat temperature after 10 h for wind tunnel tests [15].
Kuffi et al. [28] used the prediction of the slaughter floor temperature as an initial condition in the
model of the industrial carcass cooler. A comparison of measured hindquarter temperature against
predicted values at different depths is presented for the entire cooling process (Figure 3.10). There
was a difference in predicted cooling rate when a surface-averaged heat transfer coefficient and a
local, spatially-varying heat transfer coefficient distribution over the carcass surface was used. It
is clear that the use of local heat transfer coefficient values gives better predictions than using an
average heat transfer coefficient, which is expected since large variations in local heat transfer coef-
ficients have been observed. Similar comparison was made for the loin section as well.
In the hindquarter area, the predicted values generally lie below the experimental values, while
in the loin area they lie above the experimental values. This may be linked mainly to the variation
in carcass composition (e.g., bones, muscles) [14]. The bone proportion, which has relatively higher
thermal conductivity, is higher in the loin area than in the hindquarter area. Thereby, the predicted
values in the loin area lie slightly above the measurements, while the simulated ones in the hind-
quarter area lie a bit below the measurements. The hindquarter temperatures at deeper locations
continue to increase for about an hour after the carcass has entered into the chiller room. This
FIGURE 3.9 Typical temperature distribution inside a carcass during the cooling process (air velocity =
0.58 m s–1, TI = 45%, temperature = 2.3°C), after 3 h (left), 12 h (middle) and 50 h (right) of cooling.
FIGURE 3.10 Predicted vs measured hindquarter temperature during forced air cooling of a carcass (air
velocity = 0.58 m s–1, TI = 45%, temperature = 2.3°C), at the depth of 2 cm and 18 cm.
indicates the enzymes are still active during the initial hours of cooling in the chiller and producing
heat that is larger than the amount transferred to the surface by conduction. Close to the surface of
the hindquarter area and at all locations in the loin area, such a temperature increase is not present.
3.7.3 Carcass Glucose Level

During the first 20 h of cooling, there is still a relatively high level of glucose but the heat production
rate decreased to a very low level and the pH dropped below 5.5 at the depth of 18 cm (Figure 3.11).
Similar to the pH decrease, the predicted glucose concentration level in the carcass decreases very
fast in the initial hours of the cooling. It starts in the slaughter floor and continues through the first
FIGURE 3.11 Predicted change in carcass pH level during cooling (air velocity = 0.58 m s–1, TI = 45%,
temperature = 2.3°C) inside the hindquarter at different depth from the surface of the carcass, at the depth of
2 cm (broken line) and 18 cm (solid line).
hours in chiller room [14]. This high conversion rate during the initial hours is related to the high
temperature of the carcass after exsanguination, by which the activity of glycolytic enzymes is sus-
tained. The influence of meat temperature on the conversion rate is clearly seen from the difference
in concentration of glucose at different positions. The corresponding simulated volumetric heat
production rate was significant during the initial hours of cooling and also location-dependent. The
results of the temperature and pH distribution show that deeper muscles are more susceptible to heat
shortening than the outer muscles.
3.8 CFD MODEL APPLICATION IN ASSESSING

THE BEEF CARCASS COOLING PROCESS
For beef processing management, the shortest possible time between slaughter and distribution is
desirable [34,49]. For this, in principle, rapid carcass chilling systems are required. However, rapid
carcass chilling can make the cooling rate so fast that the muscle can be subjected to cold shortening
of the sarcomeres [3], which has a deleterious effect on tenderness. At the same time, rapid chilling
may require additional energy and affect other parameters, such as weight loss. Optimal conditions
should therefore exist, depending on the aim of the operators. A common way of managing the
chilling rate is by controlling airflow variables.
The cooling kinetics of carcasses and design and operation aspects of cooler rooms have been
studied using experiments and CFD simulations [12,17,50]. Mirade and Picgirard [51] performed
assessments of airflow patterns inside six industrial beef carcass coolers (three batch and three con-
tinuous) by conducting air velocity measurements and then by CFD simulation for the continuous
coolers to identify poorly ventilated areas. The authors reported that the airflow pattern in a batch
type industrial cooler is heterogeneous with the mean air velocity in the range of 0.5-1 m s–1. Mirade
et al. [12] used CFD to calculate the 3D velocity distribution in a pork carcass cooler. They obtained
good agreement between calculated 3D velocity distribution and measured velocity values and used
CFD for improving process efficiency of cooler rooms as it enables to identify poorly ventilated
areas. From the 3D velocity distribution, they determined surface-averaged heat and mass transfer
coefficients around the pork carcass to study the cooling kinetics of pork carcasses. Mirade and
Picgirard [17] used CFD to predict air velocity in industrial coolers and to design low-cost techni-
cal solutions to improve ventilation homogeneity around the carcasses. Mirade and Picgirard [17]
used CFD to predict air velocity in industrial coolers and to design low-cost technical solutions to
improve ventilation homogeneity around the carcasses. The solutions were based on adding deflec-
tors at different positions in the cooler room in order to force the air in a certain direction.
It is reported that one way of minimizing the mass loss of the carcass during cooling is reducing
the surface temperature of the carcass as fast as possible. Rapid cooling of beef carcasses with an
air temperature of –20°C and air velocity of 2.32 m s–1 for 5 h decreased weight loss by approxi-
mately 0.8 g kg–1 h–1, whereas rapid cooling with an air temperature of –35 °C at 2.32 m s–1 for 5 h
decreased the weight loss by 1.6 g kg–1 h–1 [51]. To achieve such a rapid fall in surface temperature,
modern slaughterhouse cooling systems include a pre-cooling stage, where surface temperature is
quickly reduced to decrease the amount of evaporative weight loss [25] and part of the initial heat
load is absorbed from the carcasses [22]. As a standard procedure, carcasses stay in the pre-cooling
stage for a short duration followed by the conventional cooling stage [11,52], where the carcasses
stay until the required deep muscle temperature is reached. However, retention of good meat qual-
ity is one of the practical problems observed with rapid cooling, due to extreme shortening of sar-
comeres when the pre-rigor muscle temperature is slightly above the freezing point. Interestingly,
when the carcass surface starts to freeze, it becomes hard and thus minimizes the cold shortening
phenomenon, as the hard surface resists the muscle contraction process [10,53]. If the rate of cooling
of the muscle tissue is high, the rate of biochemical changes becomes low, and thus the pH of the
FIGURE 3.12 A 3D view of a typical conventional beef carcass cooler.
muscle falls more slowly than is recommended [9]. On the other hand, when the rate of cooling is
low, the rate of biochemical changes is high thus the pH of the muscle can fall very quickly, result-
ing in hot shortening.
Rapid cooling systems usually demand high air velocity and low air temperature combinations,
which inevitably increase power consumption by the fans and the compressor of the refrigeration
system. The power required by fans to move the air increases with the cube of the airflow rate [54].
Furthermore, the power input to the compressor of the refrigeration system increases when the
temperature lift (between evaporator and condenser) increases. Thus, depending on the variable/
parameter that the slaughterhouse operator wants to minimize, the operation conditions of slaugh-
terhouse can vary. Again, optimal conditions depend on the aim of the operators.
In conventional carcass cooling systems, carcasses are directly railed into the cooler room. Here they
are cooled by means of a forced cold air flow that is typically circulated through the room using fans on
an evaporator unit attached to ceiling on one end of the room, as shown in Figure 3.12. Typically, the
entire width of the room is covered by multiple units and fans. The airflow in these cooling rooms exits
from the fan, flows along the ceiling, and depending on the length and airflow resistance, descends and
returns to the fan unit along the carcasses near the floor of the room. The carcasses are thus not directly
exposed to the outlet flow of the cooling units where typically higher air velocity (>3 m s–1) exists than
in the rest of the room (<1 m s–1). Alternatively, rapid (pre)chilling may be achieved by exposing the
carcasses more directly to high velocity (>5 m s–1) air flow from fans positioned above the carcasses
blowing cool or cold air downwards directly onto the carcass surfaces (Figure 3.13).
3.8.1 Cooling Air Velocity Direction and Magnitude

Cooling air flow direction and magnitude change the value of the surface heat transfer coefficient
(Figure 3.14) [39]. The horizontal flow direction resulted in larger variation of the heat transfer coef-
ficient than the vertical flow direction, which appears more uniform. On the outer surface the values
are higher for the horizontal flow. In the body cavity, on the other hand, it is evident that the values
are not only more uniform, but also larger for the vertical flow direction than for the horizontal flow
direction. In the horizontal flow direction, the orientation of the carcass to the inlet velocity may
increase the chance for air separation from the body, which as a result produces low local transfer
coefficient values in this region. The orientation of the carcass to the inlet velocity for the vertical
flow direction minimizes such separation.
FIGURE 3.13 A 3D view of a typical beef carcass pre-chiller.
FIGURE 3.14 Convective heat transfer coefficient contour on the carcass surface for horizontal (top) and
vertical (bottom) flow directions (inlet air velocity = 5 m s–1, TI = 45%, air temperature 2.3°C); outside of the
body (left); at the body cavity (right). (From K. D. Kuffi, “CFD Modeling of postmortem glycolysis and heat
transfer during beef carcass chilling,” PhD Thesis. KU Leuven, Belgium, 2016.)
FIGURE 3.15 Average surface heat transfer coefficient (CHTC) at different air velocity for vertical and
horizontal flow directions.
As the air velocity increases, the predicted surface averaged heat transfer coefficient values
increase more or less linearly for both the vertical and horizontal flow directions; however, the
slopes differ (Figure 3.15). Vertical flow appears to be beneficial for increasing the average surface
heat transfer coefficient; the values are almost double that of horizontal flow. The distribution pat-
tern of the surface heat transfer coefficient was not affected by the magnitude of the cooling air
velocity.
The effect of cooling air velocity magnitude on cooling rate of the carcass is not that significant.
For instance, the time needed to decrease the hindquarter temperature to 10°C at 18 cm depth
using 0°C cooling air was 39.7 and 39 h for air velocity from 0.58 m s–1 and 10 m s–1, respectively.
 Rc 
 λc 
The relatively high Biot number  Bi = = 2.7 shows that the internal heat flow resistance is
 1 
hc
greater than the external resistance at this velocity and the cooling was limited by the internal heat
transfer resistance, where Rc (m) is the radius of the hindquarter at the thickest region. As the value
of heat transfer coefficient increases with air velocity, Bi increases further; thus, the cooling process
is limited by internal resistance. Increasing the heat transfer coefficient by increasing the velocity is
thus expected not to have a direct effect on the cooling rate.
3.8.2 Cooling Air Temperature

During the carcass chilling process the variation in air temperature is more important than the
variation in air velocity for improving the rate of chilling (Figure 3.16). Air temperature cre-
ates a gradient that drives the heat flux, while variation in air velocity changes the values of the
convective transfer coefficient. For processes where the internal heat flow resistance is large,
like beef carcass cooling, lowering the air temperature enhances the cooling rate more than
increasing air velocity.
FIGURE 3.16 Predicted temperature of the beef hindquarter at 2 cm depths for different air velocities (left)
and different temperature (right) conditions and similar pre-chilling times.
3.8.3 Evaporative vs. Convective Heat Flux

Evaporation heat loss is the dominant heat transfer mechanism on the slaughter floor as well as during
the initial hours in the forced convection system, as shown in Figure 3.17, for different air velocities
in the pre-chilling stage. A distinct increase in evaporation heat loss at onset of pre-cooling is mainly
linked to the increase in convective mass transfer coefficient as a result of the increased air velocity.
Unlike evaporation and convection heat losses, the increase in velocity does not affect the radiation
heat loss. However, as the radiation heat loss depends on the temperature, the low surface temperature
of the carcass in the pre-chiller reduces the magnitude of radiation to a relatively low level.
FIGURE 3.17 Predicted evaporation and convective heat flux during beef carcass cooling at air velocity of
5 m s–1 and –15°C in the pre-chiller.
1.80 Outlines of the porous media representing the

volumes exactly filled by the carcasses
1.50
1.20
0.90
0.60
0.30
Y
Air velocities higher than 1.8 m.s−1 in the these areas
Z X
0.00
FIGURE 3.18 Air velocity (m s−1) fields calculated on a longitudinal section located at half-width (top) and
on a horizontal section located at a height of 3.1 m from the floor in the original batch-type carcass cooler.
The two arrows represent the air inlet and outlet sections. (From P. S. Mirade and L. Picgirard, “Improvement
of ventilation homogeneity in an industrial batch-type carcass chiller by CFD investigation,” Food Res. Int.,
vol. 39, no. 8, pp. 871–881, 2006.)
3.8.4 Cooling Room Design

It has been reported that the flow uniformity of the cooling air in chillers is affected by room design.
Mirade and Picgirard [17] developed a validated CFD model of an industrial batch-type carcass
cooler and analyzed the airflow distribution profile of the original cooler, which was stacked with
nine rows of carcasses. The air velocity was far from uniform; relatively high velocity was observed
in the upper half (above 1.8 m s–1) and relatively low velocity was observed on the lower half
(0.3 m s–1 or less) (Figure 3.18).
Mirade and Picgirard [17] applied the model to identify efficient and low-cost technical solutions
that significantly improve cooling airflow homogeneity. The solutions were focused on directing
the cooling air into the carcass by adding air deflectors and walls. The presence of air deflectors
and walls improved the uniformity of the cooling airflow between the upper half and lower half
of the cooler (Figure 3.19). The large area with air velocities higher than 1.8 m s–1 that is visible
in Figure 3.18 was reduced significantly in Figure 3.19. The model was also applied to optimize the
number, dimensions, location and orientation of these deflectors.
3.9 CONCLUSIONS
Cooling of beef carcasses is performed immediately after slaughter and cleaning. To cope with the
increasing demand and the consumer preference for good quality and safe meat, faster and more
uniform chilling practices need to be implemented at an affordable cost. A clear understanding of
the physical phenomena involved during the cooling process is essential, and a tool that contains
this knowledge is required. Today the tool commonly employed for such purposes is computational
fluid dynamics (CFD).
The biochemical reactions taking place within the pre-rigor muscle produce heat and hydrogen
ions. For accurate predictions of the cooling time and pH evolution, the rate of this heat generation
needs to be known and incorporated as a source term in the energy equation. Hereto a model of the
glycolysis process was developed.
During industrial beef carcass chilling, heat transfer from the surface to the cooling air takes
place by convection, evaporation and thermal radiation. The rate of heat and mass transfer depends
1.80
Plastic walls added around the row 9
1.50 Jet deflectors added in front of the blowing area
1.20
0.90
0.60
0.30
Y Outlines of the porous media representing
the volumes exactly filled by the carcasses
Z X
0.00 Air velocities higher than 1.8 m.s–1 in these areas
Air velocities lower than 0.3 m.s–1 in these areas
FIGURE 3.19 Air velocity (m s–1) fields calculated on a longitudinal section located at half-width, in which
the set of deflectors and three plastic walls were added in front of the cold air blower outlet and behind row 9,
respectively. (From P. S. Mirade and L. Picgirard, “Improvement of ventilation homogeneity in an industrial
batch-type carcass chiller by CFD investigation,” Food Res. Int., vol. 39, no. 8, pp. 871–881, 2006.)
on the airflow conditions. The temperature difference between the surface and the cooling air drives
the convection heat transfer, while the moisture concentration difference between the air near the
meat surface and the cooling air drives the evaporative heat transfer.
On the beef carcass surface, the local value of the surface heat transfer coefficient (hence, the
surface mass transfer coefficient through the Lewis relation) can vary about tenfold during indus-
trial chilling of a beef carcass. Determination of such a large difference in local surface heat transfer
coefficient values along the surface is only possible with CFD, as it resolves the air flow distribution
around the complex shape of beef carcass at arbitrary spatial resolution. Experimental measurement
of the surface heat transfer coefficient at such local positions along the surface is difficult. However,
these measurements are important for wet and nutritious food like meat for microbial safety rea-
sons, as heat or moist spots along the surface can be masked when a global surface heat transfer
coefficient is used for calculations. On the slaughter floor, evaporative heat transfer is the dominant
heat transfer mechanism compared to convection and thermal radiation, while in the chiller room
convection heat transfer is more dominant than the other two heat transfer mechanisms.
The mass transfer coefficient is related to the heat transfer coefficient by the Lewis relationship;
hence, the factors that affect the heat transfer coefficient also affect the mass transfer coefficient.
Accurate prediction of the convective mass transfer coefficient is important as it determines the
value of the surface water activity, which is an important factor for meat safety through its effect on
microbial growth. For this, besides the use of an appropriate turbulence model and boundary condi-
tions, the use of realistic carcass geometry plays a great role.
A CFD model coupled with the heat generation model was used to simulate the cooling process
of beef carcass in industrial conditions, and then the predicted temperature at different depths of
hindquarter and loin sections were compared against experimental temperature values conducted
at a commercial slaughterhouse. A good agreement between model predictions and measurements
was achieved. To investigate the effect of operation and design parameters of the beef carcass cooler
on airflow uniformity and cooling time, a model-based assessment was conducted using the CFD
model. Based on these assessments, the installation of air deflectors and guiding walls can improve
uniformity of cooling airflow. The variations in air temperature have more effect on cooling time
than variations in air velocity.
ACKNOWLEDGMENTS
The authors thank the Flemish Agency of Innovation and Entrepreneurship (VLAIO), Flanders’
Food and the Belgian Slaughterhouses for financial support (project ChillMeat).
REFERENCES
1. Cepeda J.F., Weller C.L., Negahban M., Subbiah J., Thippareddi H., Heat and mass transfer modeling
for microbial food safety applications in the meat industry: A review, Food Engineering Reviews,
5(2): 57–76, 2013.
2. Jacob R.H., Hopkins D.L., Techniques to reduce the temperature of beef muscle early in the post mortem
period—A review, Animal Production Science, 54: 482–493, 2014.
3. Hamoen J.R., Vollebregt H.M., Van Der Sman R.G.M., Prediction of the time evolution of pH in meat,
Food Chemistry, 141(3): 2363–2372, 2013.
4. Hofbauer P., Smulders F.J.M., The muscle biological background of meat quality including that of game
species, In: Game Meat Hygiene in Focus: Microbiology, Epidemiology, Risk Analysis and Quality
Assurance, Paulsen P., Bauer A., Vodnansky M., Winkelmayer R., Smulders F.J.M., Eds., Wageningen
Academic Publishers, Wageningen, Netherlands, 273–295, 2011.
5. Scheffer T.L., Gerrard D.E., Mechanisms controlling pork quality development: The biochemistry con-
trolling postmortem energy metabolism, Meat Science, 77(1): 7–16, 2007.
6. de Meis L., Role of the sarcoplasmic reticulum Ca2+-ATPase on heat production and thermogenesis,
Bioscience Reports, 21(2): 113–137, 2001.
7. Morley M.J., Measurment of heat production in beef muscle during rigor mortis, Journal of Food
Technology, 9: 149–156, 1974.
8. Newbold R.P., Scopes R.K., Postmortem glycolysis in ox skeletal muscle: Effect of temperature on
the concentrations of glycolvtic intermediates and cofactors, Biochemical Journal, 105(1): 127–136,
1967.
9. Thompson J., Managing meat tenderness, Meat Science, 62(3): 295–308, 2002.
10. Savell J.W., Mueller S.L., Baird B.E., The chilling of carcasses, Meat Science, 70(3): 449–459, 2005.
11. Liu Y., Mao Y., Zhang Y., Liang R., Wang R., Zhu L., Meng X., Luo X., Pre-rigor temperature control
of Chinese yellow cattle carcasses to 12–18°C during chilling improves beef tenderness, Meat Science,
100: 139–144, 2015.
12. Mirade P.S., Kondjoyan A., Daudin J.D., Three-dimensional CFD calculations for designing large food
chillers, Computers and Electronics in Agriculture, 34(1–3): 67–88, 2002.
13. van der Wal P.G., Engel B., van Beek G., Veerkamp C.H., Chilling pig carcasses: Effects on tempera-
ture, weight loss and ultimate meat quality, Meat Science, 40(2): 193–202, 1995.
14. Kuffi K.D., Defraeye T., Nicolai B.M., De Smet S., Geeraerd A., Verboven P., CFD modeling of
industrial cooling of large beef carcasses, International Journal of Refrigeration, 69: 324–339,
2016.
15. Pham Q.T., Trujillo F.J., McPhail N., Finite element model for beef chilling using CFD-generated heat
transfer coeffcients, International Journal of Refrigeration, 32(1): 102–113, 2009.
16. Trujillo F.J., Pham Q.T., A computational fluid dynamic model of the heat and moisture transfer during
beef chilling, International Journal of Refrigeration, 29(6): 998–1009, 2006.
17. Mirade P.S., Picgirard L., Improvement of ventilation homogeneity in an industrial batch-type carcass
chiller by CFD investigation, Food Research International, 39(8): 871–881, 2006.
18. Kinsella K.J., Sheridan J.J., Rowe T.A., Butler F., Delgado A., Quispe-Ramirez A., Blair I.S., McDowell
D.A., Impact of a novel spray-chilling system on surface microfora, water activity and weight loss dur-
ing beef carcass chilling, Food Microbiology, 23(5): 483–490, 2006.
19. Jones S.D.M., Murray A.C., Robertson W.M., The effects of spray chilling pork carcasses on the shrink-
age and quality of pork, Canadian Institute of Food Science and Technology Journal, 21(1): 102–105,
1988.
20. Brown T., Chourouzidis K.N., Gigiel A.J., Spray chilling of lamb carcasses, Meat Science, 34(3): 311–
325, 1993.
21. Greer G.G., Jones S.D.M., Quality and bacteriological consequences of beef carcass spray-chilling:
Effects of spray duration and boxed beef storage temperature, Meat Science, 45(1): 61–73, 1997.
22. Brown T., James S.J., Process design data for pork chilling, International Journal of Refrigeration, 15:
281–289, 1992.
23. Kuitche A., Daudin J.D., Letang G., Modeling of temperature and weight loss kinetics during meat
chilling for time-variable conditions using an analytical-based Method-I. The model and its sensitivity
to certain parameters, Journal of Food Engineering, 28: 55–84, 1996.
24. Issa R.J., Numerical investigation of the chilling of food products by air-mist spray, World Academy of
Science and Engineering Technology, 2(11): 364–373, 2008.
25. Savell J.W., Beef carcass chilling: Current understanding, future challenges, available on http://www.beef
research.org/CMDocs/BeefResearch/BeefCarcassChilling%20White%20Paperfinal.pdf, 1–12, 2012.
26. Brown T., Ian Richardson R., Wilkin C.A., Evans J.A., Vascular perfusion chilling of red meat
carcasses—A feasibility study, Meat Science, 83(4): 666–671, 2009.
27. Jacob R.H., Beatty D.T., Warner R.D., A preliminary study into the use of ‘heat pipes’ to prevent high
rigor temperature in beef carcasses by increasing cooling rate, Animal Production Science, 54(4): 504–
509, 2014.
28. Pastsart U., Lescouhier S., D.B.M., Claeys E., D.S.S., Effect of hot boning on colour stability and anti-
oxidant enzyme activities in beef inner and outer biceps femoris, In: 57th International Congress of
Meat Science and Technology, (3): 1–4, 7–12 August, 2011.
29. Pinto Neto M., Beraquet N.J., Cardoso S., Effects of chilling methods and hot-boning on quality
parameters of M. longissimus lumborum from Bos indicus Nelore steer, Meat Science, 93(2): 201–
206, 2013.
30. Taylor J., Toohey E.S., van de Ven R., Hopkins D.L., SmartStretchTMtechnology. IV. The impact on the
meat quality of hot-boned beef rostbiff (m. gluteus medius), Meat Science, 91(4): 527–532, 2012.
31. Kuffi K.D., Lescouhier S., Nicolai B.M., de Smet S., Geeraerd A., Verboven P., Modeling postmortem
evolution of pH in beef M. biceps femoris under two different cooling regimes, Journal of Food Science
and Technology, 55(1): 233–243, 2018.
32. Kondjoyan A., Daudin J., Optimisation of air-flow conditions during the chilling and storage of carcasses
and meat products, Journal of Food Engineering, 34(3): 243–258, 1997.
33. Van der Sman R.G.M., Simple model for estimating heat and mass transfer in regular-shaped foods,
Journal of Food Engineering, 60(4): 383–390, 2003.
34. Mallikarjunan P., Mittal G.S., Optimum conditions for beef carcass chilling, Meat Science, 39: 215–
223, 1995.
35. Christian James S.P., Ketteringham L., James S., Prediction of heat transfer during food chilling,
freezing, thawing, and distribution, In: Predictive Modeling and Risk Assessment, Costa K.K.R., Ed.,
Springer-Verlag, New York, USA, 2009.
36. Trujillo Q.T., Pham F.J., CFD modeling of simultaneous heat and mass transfer in beef chilling, In:
Computational Fluid Dynamics in Food Processing, Sun D.-W., Ed., CRC Press, Boca Raton, FL, USA,
2007.
37. Versteeg H.K., Malalasekera W., An Introduction to Computational Fluid Dynamics: The Finite Volume
Method, 2nd ed. Pearson Education, Harlow, UK, 2007.
38. Delele M.A., Schenk A., Tijskens E., Ramon H., Nicolaï B.M., Verboven P., Optimization of the humidi-
fication of cold stores by pressurized water atomizers based on a multiscale CFD model, Journal of
Food Engineering, 91(2): 228–239, 2009.
39. Kuffi K.D., CFD modeling of postmortem glycolysis and heat transfer during beef carcass chilling, PhD
Thesis. KU Leuven, Belgium, 2016.
40. Bendall J.R., Relations between muscle pH and important biochemical parameters during the post
mortem changes in mammalian muscles, Meat Science, 3(2): 143–157, 1979.
41. Davey L.M., Pham Q.T., Predicting the dynamic product heat load and weight loss during beef chilling
using a multi-region finite difference approach, International Journal of Refrigeration, 20(7): 470–482,
1997.
42. Davey L.M., Pham Q.T., A multi-layered two-dimensional finite element model to calculate dynamic
product heat load and weight loss during beef chilling, International Journal of Refrigeration, 23: 444–
456, 2000.
43. Defraeye T., Herremans E., Verboven P., Carmeliet J., Nicolai B., Convective heat and mass exchange
at surfaces of horticultural products: A microscale CFD modeling approach, Agricultural and Forest
Meteorology, 162–163: 71–84, 2012.
44. Defraeye T., Blocken B., Carmeliet J., CFD simulation of heat transfer at surfaces of bluff bodies in
turbulent boundary layers: Evaluation of a forced-convective temperature wall function for mixed con-
vection, Journal of Wind Engineering and Industrial Aerodynamics, 104–106: 439–446, 2012.
45. Dixon A.G., Ertan Taskin M., Nijemeisland M., Stitt H.H., Systematic mesh development for 3D CFD
simulation of fixed beds: Single sphere study, Computers and Chemical Engineering, 35(7): 1171–1185,
2011.
46. Menter F.R., Kuntz M., Langtry R., Ten years of industrial experience with the SST turbulence model
turbulence heat and mass transfer, In: Proceedings of the 4th International Symposium on Turbulence,
Heat and Mass Transfer, Begell House Inc., West Redding, 4: 625–632, 2003.
47. Kuffi K., Defraeye T., Nicolai B., Geeraerd A., Verboven P., Koninckx E., Lescouhier S., De Smet S., CFD
modeling of air cooling of multiple beef carcasses using 3D geometrical models, Acta Horticulturae,
1008: 159–164, 2013.
48. Davey L.M., Pham Q.T., A multi-layered two-dimensional finite element model to calculate dynamic
product heat load and weight loss during beef chilling, International Journal of Heat and Mass Transfer,
23: 444–456, 2000.
49. Mallikarjunan P., Mittal G.S., Selection criteria for beef carcass chilling, Food Research International,
29(I): 1996.
50. Mirade P.S., Picgirard L., Assessment of air flow patterns inside six industrial beef carcass chillers,
International Journal of Food Science and Technology, 36: 463–475, 2001.
51. Aalhus J.L., Robertson W.M., Dugan M.E.R., Best D.R., Very fast chilling of beef carcasses, Canadian
Journal of Animal Science, 82(964): 59–67, 2002.
52. Jacob R.H., Hopkins D.L., Techniques to reduce the temperature of beef muscle early in the post mor-
tem period—A review, Animal Production Science, 54: 482–493, 2014.
53. Joseph R.L., Very fast chilling of beef and tenderness-a report from an EU concerted action, Meat
Science, 43: 217–227, 1996.
54. James S.J., Bailey C., Process design data for beef chilling, International Journal of Refrigeration,
12(1): 42–49, 1989.
4 CFD Modeling of Heat
and Mass Transfer in a
Hydrofluidization System
During Food Chilling and
Freezing
Juan Manuel Peralta and Susana E. Zorrilla
CONTENTS
4.1 Introduction......................................................................................................................... 87
4.2 CFD in the Hydrofluidization Process..................................................................................... 88
4.3 Mathematical Modeling of the Transport Phenomena................................................................. 90
4.3.1 General Approach........................................................................................................90
4.3.2 Heat, Mass and Momentum Transfer in the Liquid Medium..........................................91
4.3.2.1 Flow Field...............................................................................................................91
4.3.2.2 Particles.........................................................................................................92
4.3.3 Heat and Mass Transfer in the Solid Medium.............................................................97
4.4 Conclusions............................................................................................................................ 100
Nomenclature......................................................................................................................... 101
References....................................................................................................................................... 103
4.1 INTRODUCTION
Hydrofluidization (HF) is a freezing and chilling technique that consists of an unfreezable liquid
that is pumped through orifices or nozzles into a vessel to create agitating submerged liquid jets.
As a consequence, extremely high surface heat transfer coefficients and high freezing rates can be
expected when products are immersed in this highly turbulent liquid [1–3]. A schematic diagram of
an HF system is shown in Figure 4.1. Usually, very low temperatures can be achieved using aqueous
solutions of solutes such as NaCl, CaCl2, KCl, ethanol, glucose, etc. [1–4].
Some advantages when using HF as a freezing technique are: it is a fast freezing method, it saves
considerable amounts of energy, it is environmentally friendly, and new delicious products can
easily be formulated [3]. This last feature is related to the mass transfer associated with the direct
contact between the refrigerant and the product; that is, a transport of appropriate antioxidants,
flavorings and macronutrients to extend the shelf-life of the products can be expected and managed
to improve the nutritional value and sensory properties of products [3]. However, an uncontrolled
mass transfer can lead to degradation of the immersion solution and undesirable levels of solutes in
the foodstuffs [5].
HF was developed to combine the advantages and overcome the drawbacks of two different indi-
vidual quick freezing (IQF) technologies: immersion and air fluidization [2]. In the first one, food is
87
Hydrofluidized
food samples
Freezing
vessel
Fluid jets
Orifices plate
Refrigerant
FIGURE 4.1 Laboratory-scaled HF system.
frozen by immersing it in a vessel containing liquid refrigerant, and in the second one, freezing is
achieved by fluidizing food pieces using an upwards flow of low temperature air. Originally called
“water fluidization” (WF), the HF technology was introduced by a Bulgarian patent in 1985 by Prof.
A. G. Fikiin [6]. Then, subsequent studies were published describing some basic aspects of experi-
mental HF rigs and values of the heat transfer coefficients obtained for freezing and chilling of fish,
fruits and vegetables (see for example: Fikiin [1]; Fikiin and Fikiin [2]; Scheerlinck et al. [7]; Fikiin
[8]; Verboven et al. [9]; Fikiin [3]; Peralta et al. [4]; Alvarez-Cardenas et al. [10]). Although those
studies greatly contributed to the understanding of the underlying transport phenomena involved in
HF, a mathematical model to address some of the most important aspects related to heat and mass
transfer was not proposed there. Moreover, the solute uptake by food and its relation to the operative
variables was not studied.
4.2 CFD IN THE HYDROFLUIDIZATION PROCESS

The complex nature of an HF system (i.e., high turbulence in liquid and stochastic movement of
food) makes it difficult to propose a general mathematical model to describe the transport phenom-
ena involved. Usually, the study of complex systems is carried out by dividing it into simpler parts,
where few aspects of the general system are studied and validated. Then, the whole system can be
modeled. Bear in mind, a set of studies were published using computational fluid dynamics (CFD)
to programmatically describe the transport phenomena in an HF system in increasing detail
[11–18].
As pointed out by Peralta [11], the ratio of the time scale of the momentum transfer to the
time scale of heat and mass transfer in an HF system can be very different from 1. That is, the
fluid field can achieve a steady state condition much faster than the heat and mass transfer,
mainly due to the high levels of turbulence found in HF. Then, the mathematical modeling of
the transfer in the fluid and in the food domain can be treated as decoupled. The decoupling
of the transfer phenomena leads to a more economic and simpler computational approach to
solve the heat and mass balances. First, the fluid domain and heat transfer is modeled until a
steady state is achieved. Second, this information is used to calculate surface parameters of
heat and mass transfer that can be used as a boundary condition on the food surface. Finally,
these boundary conditions are used to solve the mathematical model of the heat and mass trans-
fer in the food domain.
Following the above procedure, Peralta et al. [12] proposed a mathematical model for the trans-
port phenomena that take place in the fluid domain of an HF system using a simplified configura-
tion, consisting of a single stationary sphere impinged by a single submerged liquid jet. The fluid
field and heat transfer in the fluid domain were simulated by CFD. The proposed model was vali-
dated using experimental data from literature of pressure coefficient, local and stagnation Nusselt
CFD Modeling of Heat and Mass Transfer in a HF System During Food Chilling and Freezing 89
coefficient, turbulence and local velocity near the surface sphere. Different values of velocity of
the liquid at the orifice exit, temperature of the liquid refrigerant, and distance between the orifice
and the stagnation point of the sphere were tested. The refrigerant medium was considered to be an
aqueous solution of NaCl.
The information on the heat and mass transfer on the surface of the spheres obtained by Peralta
et al. [12] was condensed into local profiles of surface transfer coefficients. These profiles were
used as boundary conditions to describe the heat and mass transfer in food spheres [13]. Briefly,
the transport phenomena in the solid domain were solved by a multi-phase (i.e., solid matrix,
occluded liquid and ice) mathematical model using a control-volume approach adapted from the
one proposed by Zorrilla and Rubiolo [19,20]. The food samples were assumed to be potato
spheres. An experimental validation for the central temperature and the average solute uptake
of the potato spheres was carried out. Errors between simulated and experimental values were
considered acceptable. An important feature of this model is its ability to estimate local ice frac-
tions in the food.
Belis et al. [16] extended the study of Peralta et al. [12,13] to multiple jets (i.e., multiple orifices)
and multiple stationary spheres. In this case, the quantities used to study the transport phenomena
were: local surface heat transfer coefficient, average heat transfer coefficient, average stagnation-
point heat transfer coefficient, pressure coefficient, area-averaged turbulence intensity for different
heights in the domain, volume-averaged turbulence intensity, central temperature profile, and aver-
age solute uptake of the spheres.
So far, the described studies of the HF system were conducted using stationary spheres. This
was a good idea to gain basic insight of the general phenomena without dealing with the stochastic
nature of fluidization. As a first step, to take into account the movement of the food samples in an
HF system and minimize computational efforts, a combination of CFD with methods like Discrete
Phase Model (DPM) and Discrete Element Method (DEM) was used [17,18].
One of the goals is to model particle movements as a result of simplified versions of particle-
fluid, particle-particle and particle-wall interactions. In this manner, DPM considers a system com-
posed by dispersed (particles) and continuum (fluid) phases. Navier–Stokes equations are solved for
the fluid phase while particles are tracked using the calculated flow field. The particle-particle and
particle-wall interactions are estimated by DEM using the so-called “soft sphere” method approach
[21,22]. In this approach, the influence of the particle of interest is assumed to be limited only to
neighboring particles (in direct contact) by allowing a slight overlap [23,24]. Then, all the interac-
tions can be estimated as a dynamic process where contact forces accumulate or dissipate over
time [24].
As a proof of concept, Oroná et al. [17] presented a mathematical model to study the transport
phenomena in an HF system by using CFD+DPM+DEM to describe the momentum and heat trans-
fer in the refrigerant domain. The heat and mass transfer in the food was estimated using the model
proposed by Zorilla and Rubiolo [19,20], and modified by Peralta et al. [13]. This is performed by
taking advantage of the decoupling procedure described earlier. The proposed model was validated
using heat and momentum data from bibliography and previous works. Also, experimental values
of restitution coefficients for the particle-particle and particle-wall interactions were determined.
The simulations consisted of the freezing of 13 potato spheres immersed in an aqueous solution of
NaCl. Results have shown more realistic values, compared to previous studies, for representative
parameters of heat and mass transfer.
Finally, a sensitivity analysis of the model proposed by Oroná et al. [17] was carried out to com-
plete the presentation [18]. Here, some of the studied parameters were the velocity contours and
streamlines, the behavior of spheres in the fluid domain, the volume-averaged velocity of spheres,
the volume-averaged relative fluid velocity, the turbulence levels, the dispersion level of the spheres,
the surface heat transfer coefficient, and the mass and heat transfer in the food domain. The opera-
tive variables were the average velocity of the refrigerant fluid at the orifices, the refrigerant tem-
perature, the distance between orifices, and the number of spheres.
FIGURE 4.2 Schematic diagram that summarizes the evolution of the modeling approach for an HF system
using CFD. (a) Central temperature of the food sample, and (b) solute uptake. (Adapted from Oroná, J.D.,
Zorrilla, S.E., Peralta, J.M., 2017. Computational fluid dynamics combined with discrete element method and
discrete phase model for studying a food hydrofluidization system. Food and Bioproducts Processing 102,
278–288. doi:10.1016/j.fbp.2017.01.005.)
A schematic diagram that summarizes the evolution of the modeling approach presented in this
work, to study the heat and mass transfer of an HF system using CFD, is presented in Figure 4.2.
As far as the authors know, the mathematical model proposed by Oroná et al. [17,18] is currently
the most realistic, versatile, and usable approach to describe the transport phenomena (i.e., phase
change, heat, mass, and momentum transfer) in an HF system. Therefore, the following sections will
be dedicated to present this mathematical model as succinctly as possible.
4.3 MATHEMATICAL MODELING OF THE TRANSPORT PHENOMENA

4.3.1 General Approach
Taking into account that the momentum transfer is much faster than in analog counterparts, a
decoupling approach is used. That is, transport phenomena (and phase change) in the food domain
are solved after the balances in the fluid domain are computed. Then, this presentation is split into
sections dealing with transport phenomena in the liquid and in the solid media.
The mathematical modeling of the transport phenomena proposed by Oroná et al. [17,18]
solves for each time step: (a) the heat, mass and momentum transfer in the liquid refrigerating
medium by combining CFD with DPM and DEM, and (b) the heat and mass transfer inside the
food sample using a validated mathematical model. The information of the first step is used to
estimate the velocity (and position) of the particles through DPM and DEM (i.e., soft sphere
model). An average surface heat transfer coefficient for each sphere is obtained and used to
transfer information from the liquid to the solid phase. The second step consists of solving a
heat and mass transfer model proposed in previous studies [12,13,19,20], assuming that trans-
port phenomena mainly occur in the radial direction.
4.3.2 Heat, Mass and Momentum Transfer in the Liquid Medium

4.3.2.1 Flow Field
The momentum (Navier–Stokes) and mass (continuity) balances in the liquid refrigerating domain
are solved by CFD. Usually, the aqueous solutions used in HF can be considered as Newtonian [11].
These balances are averaged to take into account the turbulence phenomenon. Then, the flow of a
fluid can be described by the following equations [25,26]:
∂ρ
+ ∇ ⋅ ( ρ v) = 0 (4.1)
∂t
∂ρ v
∂t
( )
+ ∇ ⋅ ρ vv = −∇P + ∇ ⋅ τ − ∇ ⋅ ρ v′v′ + ρ g + S M (4.2)

∂ρ Htot ∂p
∂t
−
∂t
( ) ( )
+ ∇ ⋅ ( ρ vHtot ) = ∇ ⋅ (α ∇ρ H ) − ∇ ρ v′H + ∇ ⋅ v ⋅ τ + SE (4.3)

1 2
Htot = H + v + κ (4.4)
2
1 2
κ= v ′ (4.5)
2
where Htot and κ are the total enthalpy and the turbulence kinetic energy, respectively. The
quantities − ρ v ′v ′ and − ρ v ′H are the Reynolds fluxes for momentum and energy, respectively.
These quantities are related to turbulence and can be estimated using an eddy-viscosity turbu-
lence model such as the κ - ω Shear Stress Transport (SST) [27], through [26]:
2
− ρ v ′v ′ = µt ( ∇v ) + ( ∇v )  − δ ( ρκ + µt ∇ ⋅ v ) (4.6)
+
µt
− ρ v ′H = ∇H (4.7)
Prt
where μt and Prt are the turbulence viscosity and turbulence Prandtl number, respectively.
As demonstrated in previous works for HF systems [12,16–18], standard velocity contours and
2D-3D streamlines are useful to gain insight into the effect of operative variables on the flow field
(Figure 4.3). The 3-D model in CFD allows for effective study of the liquid flow characteristics.
FIGURE 4.3 Representative outputs for a domain, mesh, velocity contours and streamlines.
However, it is very important to recall that special care should be taken when representing the
domain and boundary conditions, selecting the mesh (particularly close to the jet region), and test-
ing the mesh independence.
4.3.2.2 Particles
The basic information needed to characterize the transport in and around a particle is its position,
velocity, heat, and mass transfer. Using DPM, the velocity of the particle p ( v p ) is estimated as
[22,23,28]:
dv p ( ρ − ρ)
= FD ( v f − v p ) + p g + FDEM (4.8)
dt ρp

where v f is the local fluid velocity, ρp is the density of the particle p, ρ is the fluid density, g is the
gravity acceleration vector, FDEM is the force per unit particle mass due to the interactions between
particles (and between particles and solid surfaces) and FD is the drag force per unit particle mass
estimated by Gidaspow [29]:
18 µ C D Re D
FD = (4.9)
ρ p D 2 24

where μ is the fluid viscosity, D is the diameter of the particle p, CD is the drag coefficient and ReD
is the Reynolds number based on the diameter of the particle p and on the magnitude of the relative
(
particle local velocity to the fluid local velocity ρ Dvslip µ .)
The model used to estimate the drag coefficient depends on the operational range of ReD. For HF
systems and for simplicity, the Wen-Yu model can be used. This model was suggested for systems
where the solid phase volume is less than the 20% of the total volume. Then, CD is estimated as
[22,29]:
CD =
24
Re D
( )
1 + 0.15 Re 0D.687 (4.10)

The force FDEM can be estimated using DEM. In this approach, a Kelvin–Voigt model is used to
estimate the forces from the particle collisions as the overlap between pairs of particles and between
particles and a boundary. FDEM can be decomposed in:
FDEM = FN + FT (4.11)

On one hand, the simplest way to estimate the normal component (FN ) is [23]:

(
FN =  Kδ + γ v p− p′ ⋅ ep− p′  ep− p′ ) (4.12)
where K is the spring constant accounting for the elastic effect of the interactions between the par-
ticles p and pʹ, δ is the overlap between the particles p and pʹ,γ is the damping coefficient, v p− p′ is the
relative velocity between the particles p and pʹ and ep− p′ is the unit vector between particles p and
pʹ. Briefly, these parameters can be determined by Crowe et al. [23]:
ep− p′ =
(r p − rp′ ) (4.13)
rp − rp′

(
δ = rp − rp′ − Rp + Rp′ ) (4.14)
v p− p′ = v p − v p′ (4.15)

m p− p′ ln(η)
γ = −2 (4.16)
tcoll
m p m p′
m p− p′ = (4.17)
m p + m p′

m p− p′
tcoll = floss (4.18)
K
floss = π 2 + ln 2 (η) (4.19)

where ri is the position vector of particle i, Ri is the radius of particle i, mp−pʹ is the reduced mass,
mi is the mass of particle i, tcoll is the collision time scale, f loss is a loss factor and η is a coefficient
of restitution. It is important to mention that 0 < η ≤ 1 and that the damping coefficient is γ ≥ 0 s–1
[22,23]. Moreover, the Hertzian contact theory (i.e., δ3/2) [23] should be used to estimate FN when
more general and accurate results are needed.
On the other hand, the tangential component ( FT ) is assumed to be proportional to FN :
FT = µ friction FN (4.20)

where µ friction is the friction coefficient between particles p and pʹ. The direction of FT is opposite to
the relative tangential motion. Here, μfriction is considered as a function of the tangential component
(vp−pʹ, T) of the relative velocity vector between particles ( v p− p′ ) . This parameter can be determined
experimentally [23]. A simple way to estimate μfriction is by [22]:
v  v 
µ friction = µstick + ( µstick − µ glide )  p− p′ ,T − 2..0   p− p ',T  ( v p− p ',T ≤ vglide ) (4.21)
 vglide   vglide 

µ friction = µ glide ( vglide < v p− p ',T ≤ vlimit ) (4.22)

slimit + ( v p− p′ − vlimit )
µ friction = ( v p− p′ ,T > vlimit ) (4.23)
slimit + ( µ glide µlimit )( v p− p′ − vlimit )

where μstick is the sticking friction coefficient, μglide is the gliding friction coefficient, μlimit is the high
velocity limit friction coefficient, vglide is the gliding velocity, vlimit is the limit velocity, and slimit is a
parameter determining how fast μfriction approaches μlimit.
The parameters K, η, μstick, μglide, μlimit, vglide, vlimit, and slimit are necessary to calculate the earlier
expressions. Sensible testing values are: μstick = 0.5, μglide = 0.2, μlimit = 0.1, vglide = 1 m s–1, vlimit =
10 m s–1 and slimit = 100 s m–1 [17,18]. In the case of the spring constant, K = 10000 s–2 can be assumed
for potato. This value can be determined by trial and error following a practical guide of Fluent [22],
where it was suggested that K values should not be too low because particles could overlap or too
high because the interaction would be too fast and the simulation time would be too small. Oroná
et al. [17] presented a simple and economic method to estimate values of η. For example, values of
η for potato were found to be ηw = 0.52 and ηp = 0.77.
So far, local and individual velocities of particles and fluid can be calculated (i.e., v p and v f ).
These parameters can be used to estimate averaged quantities that are useful to characterize the
behavior of a group of particles [16,17]. Also, integrated quantities can facilitate the validation
of the studied mathematical model due to their capacity to be directly compared to macroscopic
experimental parameters.
A set of useful integrated quantities are: average velocity of particles v p , average relative fluid ( )
( )
velocity vslip , and volume-averaged turbulence intensity ( Tu ). They are defined as [16–18]:
1
vp =
VT ∫v
VT
p dV (4.24)

1 1
vslip =
VT ∫v slip dV =
VT ∫v p − v f dV (4.25)
VT VT
1
Tu =
VT ∫ Tu dV (4.26)
VT

where v p is the velocity of particle p, v p is the magnitude of v p, v f is the fluid velocity, vslip is the
magnitude of the particle velocity relative to the fluid velocity, Tu is the local turbulence intensity
(calculated from CFD), V T is the total domain volume and V is the fluid volume.
Usually, the use of DPM implies that heat transfer calculations are simplified by the use of
validated correlations based on fluid flow parameters previously estimated. Then, the problem in
estimating the heat transfer is using the most accurate dimensionless number correlation for the
simulated system. For HF systems using spheres, Oroná et al. [17,18] tested successfully the follow-
ing expression [22,23]:
( )  µ 4
1 2
NuD = 2 + 0.4 Re D + 0.06 Re D Pr
2 3 0.4
 µ  (4.27)
s
where NuD is the Nusselt number based on the particle diameter (hcD/k), ReD is the Reynold num-
ber (ρDvslip/μ), Pr is the Prandtl number (Cpμ/k), μ is the fluid viscosity at the bulk refrigerating
medium temperature, μs is the fluid viscosity at the particle surface temperature, hc is the surface
heat transfer coefficient, k is the fluid thermal conductivity, and Cp is the fluid heat capacity. Based
on Perry and Green [30], Equation (4.27) is valid for 0.7 < Pr < 380, 3.5 < Re D < 8 × 10 4 and 1 < μ/
μs < 3.2. The viscosity μs is calculated at the average temperature of the refrigerating medium and
initial particle surface temperature.
Heat transfer coefficients can be used to estimate how fast solutes enter or leave a particle. Oroná
et al. [17,18] showed that the Chilton–Colburn analogy (CCA) is a useful tool to quickly calculate
the mass transfer coefficients (kc) in an HF system. This parameter is used as a boundary contour
for the mass transfer balances that model what occurs inside the solid particles. CCA is used due to
its simplicity and successful history [13].
Similar to the momentum parameters, useful averaged heat transfer quantities can be defined for
HF systems [12,16–18]: the volume-averaged surface heat transfer coefficient hc and the time- ( )
averaged surface heat transfer coefficient ( )
hc . They are defined as:
1
hc =
VT ∫ h dV
VT
c (4.28)

1
hc =
tRe s ∫
tRe s
hc dt (4.29)

where hc is the surface heat transfer coefficient of each particle, t is time, and tRes is the total resi-
dence time of the particles in the domain.
One of the most important features of HF is that, when correctly controlled, the frozen products
tend to have an attractive appearance and not to stick together due to the fluidization [3]. As stated
by Oroná et al. [17], the fluidization (or dispersion level of particles) of a large number of particles
is usually correlated to the pressure drop through the bed [31]. This is not the case for systems
composed of few particles (i.e., low volume fraction), where there are not enough particles to form a
distinctive bed and the pressure drop would not be correctly determined. However, for these systems
the position of all particles is relatively easy to compute and the dispersion level can be calculated
directly. Oroná et al. [17] proposed a way to compute a dispersion level of particles using a dimen-
(
sionless average minimum distance d t [32–34]: )
n
d min
d t=
d min
t
=
1
d min ∑ min{d }
i =1
i≠ j
n
ij j =1 (4.30)
0 0
where min{dij}nj = 1 is the minimum value of dij in the Euclidean distances of the geometrical centers
i≠ j
{di1, di3, … , din} excluding the distance di = j.
As a summary, Figure 4.4 shows typical profiles obtained by Oroná et al. [18] of v p , vslip , and
hc as a function of time for different operative variables (orifice separation, number of particles,
average jet velocity at orifice, and temperature). These outputs emphasize the prediction capabil-
ity of the approach proposed. There is a broad spectrum of variables that can be analyzed to get a
deep understanding of the process with relatively small computational effort and time. Moreover,
the outputs can be conveniently selected to specifically represent the characteristics of each system
under study.
Dispersion coefficient profiles showing their corresponding arrangements for different num-
ber of spheres and selected times are shown in Figure 4.5. Such information is very relevant
FIGURE 4.4 Particle velocity, slip velocity and heat transfer coefficient profiles for different orifice sepa-
rations (S), number of spheres (E), average fluid velocity at orifices (Vo) and refrigerant temperature (Tsn).
(Adapted from Oroná, J.D., Zorrilla, S.E., Peralta, J.M., 2018. Sensitivity analysis using a model based on
computational fluid dynamics, discrete element method and discrete phase model to study a food hydrofluidi-
zation system. Journal of Food Engineering 237, 183–193. doi: 10.1016/j.jfoodeng.2018.05.19.)
FIGURE 4.5 Dispersion coefficient profiles showing their corresponding arrangements for different num-
ber of spheres and selected times. (Adapted from Oroná, J.D., Zorrilla, S.E., Peralta, J.M., 2018. Sensitivity
analysis using a model based on computational fluid dynamics, discrete element method and discrete phase
model to study a food hydrofluidization system. Journal of Food Engineering 237, 183–193. doi: 10.1016/j
.jfoodeng.2018.05.019.)
when modeling dispersed-phase flows in fluidized beds (i.e., to avoid uneven spatial distribu-
tion, which may result in undesirable frozen food quality). It is worth mentioning that this
approach for determining the particle dispersion is relatively simple and can be used to study
similar fluidization phenomena during food processing, which is a rather incipient modeling
matter.
4.3.3 Heat and Mass Transfer in the Solid Medium

The heat and mass transfer in an HF system can be estimated using the mathematical model pro-
posed by Zorrilla and Rubiolo [19] for ICF processes and adapted to HF systems by Peralta et al. [13].
This model assumes the food domain as composed of three phases: (1) ice (phase I), (2) occluded
liquid (phase OL), and (3) solid matrix (phase S). After averaging the heat and mass balances (pro-
posed for each phase) over an arbitrary volume, a pseudo-continuum domain is obtained. Thus, the
heat and mass transfer in food can be estimated by:

ρ
∂ H
∂t
(
= ∇ ⋅ keff ∇ T ) (4.31)

∂
∂t
(
ε OL ρ1 OL )+ m = 0 (4.32)
  ρ2 
∂
∂t
(ε OL ρ2 OL ) = ∇ ⋅  ρ OL OL Deff ∇  OL

 ρOL OL  
(4.33)


∂
∂t
(ε I ρI I )= m (4.34)
where keff is the effective thermal conductivity, H is the average mass enthalpy, T is the average
temperature, ρ is the spatial average in the domain of the density, m is the mass rate of water
solidification per unit volume, εOL is the volume fraction of phase OL, ρ1 OL is the average concen-
tration of component 1 in phase OL, ρ2 OL is the average concentration of component 2 in phase
OL, ρOL is the average concentration of all species in phase OL, Deff is the effective diffusion
OL
coefficient of the solute, εI is the volume fraction of phase I, and ρI is the average concentration
I
of all species in phase I.
Equations (4.32)–(4.33) were presented assuming the simplest configuration. That is, the occluded
liquid is composed of water (i.e., component 1) and one solute (i.e., component 2). For multicompo-
nent occluded solutions, individual mass balances for each species must be solved.
For an HF system, Peralta et al. [13] proposed for Equations (4.31)–(4.34) the following boundary
conditions (t > 0):
∂ T
keff
∂s
(
= hc Tsn − T (4.35) )

∂ ρ2
Deff OL
= kc ( ρ2 )sn − ρ2  (4.36)

∂s OL
and initial conditions (t = 0):
T = T 0 (4.37)
ρ1 = ρ1 (4.38)
OL OL ,0
ρ2 = ρ2 (4.39)
OL OL ,0
ε OL = ε OL ,0 (4.40)

ε I = ε I ,0 (4.41)

where s is a coordinate perpendicular to the food surface, Tsn is the fluid temperature near the
food surface, ( ρ2 )sn is the concentration of component 2 in the refrigerant, T 0 is the initial
average temperature, ρ1 OL,0 is the initial average concentration of component 1 in phase OL,
ρ2 OL,0 is the initial average concentration of component 2 in phase OL, ε OL,0 is the initial
volume fraction of phase OL, and ε I,0 is the initial volume fraction of phase I. Again, for
multicomponent occluded solutions, individual boundary conditions must be stated for each
species.
To complete the mathematical model, a set of thermophysical properties must be estimated.
Peralta et al. [13] used the following expressions:
(
H = T − Tref

) Cp +
e1 ∆H 0 T0 − T f ( ) 
 (4.42)

f
(T 0 − Tref )(T
0 − T )
e1 ∆H 0 (T0 − T f )
Cpeff = Cp f + (4.43)
(T )
2
0 − T

keff = k f + ( k u − k f )
(T 0 − Tf ) (4.44)

(T 0 − T )
where Tref is a reference temperature (–40°C), T0 is the freezing point of pure water, Tf is the initial
freezing point of the food, Cpf is the specific heat of the frozen food, e1 is the total initial freezable
water fraction, ΔH0 is the latent heat of fusion of ice, kf is the thermal conductivity of the frozen
food, and ku is the thermal conductivity of the unfrozen food.
The phase equilibrium that takes place between the solvent in the occluded liquid and the ice
can be estimated using the excess Gibbs energy model proposed for multicomponent solutions by
Peralta et al. [35] and validated by Tello Alonso et al. [36].
Finally, for systems where the transport of only one component is important and independent
from the rest of the species, the effective diffusion coefficient can be estimated by [13]:
ε OL 
Deff = (4.45)
τ
where Đ is the Fick binary diffusion coefficient and τ is a factor called tortuosity. For multicom-
ponent systems, where the transport of one component is affected by the movement of the other
species, the model proposed by Tello Alonso et al. [37] to estimate diffusion coefficients should be
considered.
Peralta et al. [13] showed in detail a simple numerical implementation for solving Equations
(4.31)–(4.41) applied to a spherical domain. Complementary examples of application to other geom-
etries can be found elsewhere [19,20].
Useful local and averaged heat and mass transfer related quantities can be calculated using the
approach presented in this work. Local profiles of temperature and solute concentration are used
most commonly to analyze the effect of the freezing phenomena on the heat and mass transfer.
Local temperature profiles in the geometric center of the particles and the global solute uptake
(i.e., volume-averaged solute concentration) are good quantities to compare with experimental data
[13]. Also, one of the most important quality features of frozen food products is the time and/or the
velocity that the food freezes. These parameters can be calculated by the evolution of local volume
FIGURE 4.6 Central temperature and solute uptake profiles along their respective solute, temperature and
water fraction fields for different times. (Adapted from Peralta, J.M., Rubiolo, A.C., Zorrilla, S.E., 2012.
Mathematical modeling of the heat and mass transfer in a stationary potato sphere impinged by a single
round liquid jet in a hydrofluidization system. Journal of Food Engineering 109, 501–512. doi:10.1016/j
.jfoodeng.2011.10.032.)
fractions of the ice and the occluded liquid [13]. Some aforementioned profiles are shown in Figure
4.6. The modeling approach proposed is able to predict, as realistically as possible, global variables
related to solute uptake and temperature of the slowest cooling point, which is highly valuable for
prototype and industrial applications.
4.4 CONCLUSIONS
A brief review of the main contributions to the mathematical modeling of the heat and mass
transfer in an HF system using CFD was presented. These works were part of a general strategy
to programmatically increase the level of description of the main transport phenomena in an
HF system. As a consequence, a novel and state-of-the-art mathematical model for the transport
phenomena in an HF system was developed and summarized in this presentation. This model can
deal with three features that are very important in a food freezing process using HF. First, it can
simulate conveniently and reasonably well the fluidization phenomenon through a combination of
CFD, DPM, and DEM. Second, the freezing process in the food can be rigorously described in
detail. Finally, the heat and mass transfer is known in both domains (i.e., food and liquid refrig-
erant). Moreover, multi-component mass transfer can be computed with minimal computational
efforts.
The mathematical approach involves a rigorous and convenient set of momentum, heat, and
mass balances that are solved for each domain. The highly turbulent nature of the fluid domain
allows decoupling of the momentum transfer in the refrigerant from the heat and mass balances in
both domains. This is true for most practical conditions. Therefore, transport phenomena can be
solved sequentially (momentum transfer in refrigerant → heat transfer in refrigerant → heat and
mass transfer coefficient in particle surface → heat and mass transfer in food particles). Different
validation steps, related to its constituent parts, were carried out using the main heat and mass trans-
fer parameters. Local and integrated parameters related to the momentum, heat, and mass transfer
can be easily calculated.
Due its simplicity, the mathematical approach presented here can be used as a tool to help in the
designing, control, and optimization of an HF system using minimal computational resources. Also,
easy in-line decision-making calculations can be performed.
Despite the described practical benefits, many aspects of the mathematical model need to be bet-
ter addressed to gain incremental understanding of the process (for example, solving the boundary
layer near the particles, describing the fluid-solid interaction, taking into account solid deformation,
and calculating the stresses in the solid matrix).
NOMENCLATURE
CD Drag coefficient (-)
Cp Specific heat capacity (J kg–1 K–1)
D Diameter of the spheres (cm)
Đ Fick binary diffusion coefficient (m2 s–1)
dij Euclidean distance of geometric centers (m)
d t Dimensionless average minimum distance (-)
d min t Average minimum distance at t (m)
d min 0 Value of d min t at t = 0 s (m)
E Number of particles (-)
e Unit vector between two particles (-)
e1 Total initial freezable water fraction (-)
FD Drag force per unit particles mass (m s–2)
FDEM Force per unit particle mass due to the interactions between particles estimated by
DEM (m s–2)
FN Normal component of FDEM (m s–2)
FT Tangential component of FDEM (m s–2)
f loss Loss factor (-)
g Gravity acceleration vector (m s–2)
H Mass enthalpy (J kg–1)
hc Surface heat transfer coefficient (W m–2 K–1)
hc Time-averaged hc (W m–2 K–1)
K Spring constant per unit mass (s–2)
k Thermal conductivity (W m–1 K–1)
kc Mass transfer coefficient (m s–1)
m Mass (kg)
m Mass rate of water solidification per unit volume (kg m–3 s–1)
NuD Nusselt number based on D (hcD/k) (-)
P Pressure (Pa)
Pr Prandtl number (CPμ/k) (-)
R Radius (m)
ReD Reynolds number based on D and vslip (ρDvslip/μ) (-)
R Radial position (m)
r Position vector (m)
S Minimum distance between the geometric centers of the orifices (cm)
SE Heat source or link (W m–3)
SM Momentum source or sink (N m–3)
s Coordinate perpendicular to the food surface (m)
slimit Parameter determining how fast μ approaches μlimit (s m–1)

T Temperature (°C)
Tc Temperature of the geometric center of the spheres (°C)
T0 Freezing point of pure water (°C)
Tf Initial freezing point (°C)
t Time (s)
tcoll Collision time scale (s)
tRes Total residence time of the particles in the domain (s)
Tu Local turbulence intensity ( 2 κ
3 )
v f (-)
V Volume (m ) 3
Vo Area averaged fluid velocity at the orifices (m s–1)

V T Volume of the fluid domain (m3)
v Fluid velocity vector (m s–1)
vglide Gliding velocity (m s–1)
vslip Magnitude of v p − v f (m s–1)
vlimit Limit velocity (m s–1)
y Axial position in the fluid domain (m)
Greek symbols
∝ Thermal diffusivity (m2 s–1)
γ Damping coefficient (s–1)
ΔH0 Latent heat of fusion of ice (J kg–1)
δ Overlap between the particles p and pʹ (-)
δ Kronecker delta tensor (-)
ε Volume fraction (-)
η Coefficient of restitution (-)
κ Turbulence kinetic energy (m2 s–2)
μ Viscosity (Pa s)
μfriction Friction coefficient (-)
μglide Gliding friction coefficient (-)
μlimit High velocity limit friction coefficient (-)
μt Turbulence viscosity (Pa s)
μstick Sticking friction coefficient (-)
ρ Density or concentration (kg m–3)
τ Tortuosity (-)
τ Viscous stress tensor (-)
ω Specific turbulence dissipation rate (s–1)
Subscripts
0 Initial
1 Component 1 (i.e., water)
2 Component 2 (i.e., solute)
eff Effective
f Fluid or frozen food
I Ice phase
OL Occluded liquid phase
p Particle p or particle-particle interactions
p − pʹ Relative between particles p and pʹ
ref Reference state
sn In the refrigerant bulk

T Tangential
u Unfrozen food
w Particle-wall interactions
Special symbols
〈•〉 Volume averaged quantity
Acronyms
CFD Computational fluid dynamics
DEM Discrete element method
DPM Discrete phase method
HF Hydrofluidization
ICF Immersion chilling and freezing
REFERENCES
1. Fikiin A.G., New method and fluidized water system for intensive chilling and freezing of fish, Food
Control, 3: 153–160, 1992. doi:10.1016/0956-7135(92)90100-O.
2. Fikiin K., Fikiin A., Individual quick freezing of foods by hydrofluidisation and pumpable ice slur-
ries, In: IIR Proceedings Series Refrigeration Science and Technology, Presented at the International
Conference on Advances in the Refrigeration Systems, Food Technologies and Cold, Fikiin K., Ed.,
Sofia, Bulgaria, 319–326, 1998. doi:10.13140/RG.2.1.3901.1283/1.
3. Fikiin K., Emerging and novel freezing processes, In: Frozen Food Science and Technology, Evans J.A.,
Ed., Blackwell Publishing Ltd., Oxford, UK, 101–123, 2008.
4. Peralta J.M., Rubiolo A.C., Zorrilla S.E., Design and construction of a hydrofluidization system, Study of
the heat transfer on a stationary sphere, Journal of Food Engineering, 90: 358–364, 2009. doi:10.1016/j
.jfoodeng.2008.07.004.
5. Lucas T., Raoult-Wack A.L., Immersion chilling and freezing in aqueous refrigerating media: Review and
future trends, International Journal of Refrigeration, 21: 419–429, 1998. doi:10.1016/S0140-7007(98)00014-0.
6. Fikiin A.G., Method and system for water cooling of fish, Bulgarian Patent No. 40164, 1985.
7. Scheerlinck N., Jancsók P., Verboven P., Nicolaï B.M., Influence of shape on the fast freezing of small
fruits by means of hydrofluidisation, In: AgEng, Budapest, Hungary, 8, 2002.
8. Fikiin K., Novelties of food freezing research in Europe and beyond fair-flow europe synthetic brochure
for SMEs, No. 10, In: Institut National de la Recherche Agronomique, Paris, France, 2003.
9. Verboven P., Scheerlinck N., Nicolaï B.M., Surface heat transfer coefficients to stationary spherical par-
ticles in an experimental unit for hydrofluidisation freezing of individual foods, International Journal
of Refrigeration, 26, 328–336, 2003. doi:10.1016/S0140-7007(02)00110-X.
10. Alvarez-Cardenas A., Hernandez-Viveros J., Vazquez-Avila T.L., Quintanar-Guerrero D., Zambrano-
Zaragoza M.L., Precooling by nanocapsule coating application and hydrofluidization on fresh-cut fruit,
In: Advances in Science, Biotechnology and Safety of Foods, Nevarez-Morillón G.V., Garcia Galindo
H.S., Garcia S., Eds., Asociacion Mexicana de Ciencia de los Alimentos, San Nicolas de los Garza,
Mexico, 269–276, 2015.
11. Peralta J.M., Food freezing by hydrofluidization (in spanish) (Ph.D.), Universidad Nacional del Litoral,
Santa Fe, Argentina, 2009.
12. Peralta J.M., Rubiolo A.C., Zorrilla S.E., Mathematical modeling of the heat transfer and flow field of
liquid refrigerants in a hydrofluidization system with a stationary sphere, Journal of Food Engineering,
99: 303–313, 2010. doi:10.1016/j.jfoodeng.2010.03.003.
13. Peralta J.M., Rubiolo A.C., Zorrilla S.E., Mathematical modeling of the heat and mass transfer in a
stationary potato sphere impinged by a single round liquid jet in a hydrofluidization system, Journal of
Food Engineering, 109: 501–512, 2012. doi:10.1016/j.jfoodeng.2011.10.032.
14. Oroná J.D., Zorrilla S.E., Peralta J.M., Study of a hydrofluidization system using computational fluid
dynamics and a discrete element method I, Flow field and velocity profiles, In: Anais Do XX Congresso
Brasileiro de Engenharia Quimica, Florianopolis, Brazil, 2014a.
15. Oroná J.D., Zorrilla S.E., Peralta J.M., Study of a hydrofluidization system using computational fluid
dynamics and a discrete element method II, Heat transfer and turbulence intensity, In: Anais Do XX
Congresso Brasileiro de Engenharia Quimica, Florianopolis, Brazil, 2014b.
16. Belis E.E., Zorrilla S.E., Peralta J.M., Effect of the number of orifices and operative variables on the
heat and mass transfer in a hydrofluidization system with static spheres, Journal of Food Engineering,
153: 96–107, 2015. doi:10.1016/j.jfoodeng.2014.12.016.
17. Oroná J.D., Zorrilla S.E., Peralta J.M., Computational fluid dynamics combined with discrete element
method and discrete phase model for studying a food hydrofluidization system, Food and Bioproducts
Processing, 102: 278–288, 2017. doi:10.1016/j.fbp.2017.01.005.
18. Oroná J.D., Zorrilla S.E., Peralta J.M., Sensitivity analysis using a model based on computational fluid
dynamics, discrete element method and discrete phase model to study a food hydrofluidization system,
Journal of Food Engineering, 237: 183–193, 2018. doi: 10.1016/j.jfoodeng.2018.05.019.
19. Zorrilla S.E., Rubiolo A.C., Mathematical modeling for immersion chilling and freezing of
foods: Part I: Model development, Journal of Food Engineering, 66: 329–338, 2005a. doi:10.1016/j
.jfoodeng.2004.03.026.
20. Zorrilla S.E., Rubiolo A.C., Mathematical modeling for immersion chilling and freezing of foods: Part II:
Model solution, Journal of Food Engineering, 66: 339–351, 2005b. doi:10.1016/j.jfoodeng.2004.03.027.
21. Cundall P.A., Strack O.D.L., A discrete numerical model for granular assemblies, Géotechnique, 29:
47–65, 1979. doi:10.1680/geot.1979.29.1.47.
22. Fluent, ANSYS Fluent Theory Guide, Release 14.0, ANSYS, Inc., Canonsburg, PA, USA, 2011.
23. Crowe C.T., Schwarzkopf J.D., Sommerfeld M., Tsuji Y., Multiphase Flows with Droplets and Particles,
2nd ed. CRC Press, Boca Raton, FL, USA, 2012.
24. Wu C.-Y., Discrete Element Modelling of Particulate Media, Special Publication. Royal Society of
Chemistry, Cambridge, UK, 2012.
25. Bird R.B., Stewart W.E., Lightfoot E.N., Transport Phenomena, Rev. 2 ed., John Wiley & Sons,
Hoboken, NJ, USA, 2007.
26. Versteeg H.K., Malalasekera W., An Introduction to Computational Fluid Dynamics: The Finite Volume
Method, 2nd ed. Pearson Education Ltd, Harlow, UK, 2007.
27. Wilcox D.C., Turbulence Modeling for CFD, 3rd ed. DCW Industries, La Cañada Flintridge, CA, USA,
2006.
28. Schulze S., Schmidt R., Nikrityuk P.A., Modeling of moving particles. Review of basic concepts and
models, In: Gasification Processes, Nikrityuk P.A., Meyer B., Eds., Wiley-VCH, Weinheim, Germany,
43–72, 2014.
29. Gidaspow D., Multiphase Flow and Fluidization: Continuum and Kinetic Theory Descriptions,
Academic Press, Cambridge, MA, USA, 1994.
30. Perry R.H., Green D.W., Perry’s Chemical Engineers’ Handbook, 8th ed. McGraw-Hill, New York,
USA, 2008.
31. Smith P.G., Applications of Fluidization in Food Processing, 1st ed. Blackwell Publishing Ltd., Oxford,
UK, 2007.
32. Clark P.J., Evans F.C., Distance to nearest neighbor as a measure of spatial relationships in populations.
Ecology, 35: 445–453, 1954. doi:10.2307/1931034.
33. Illian J., Penttinen A., Stoyan H., Stoyan D., Statistical Analysis and Modelling of Spatial Point Patterns,
Statistics in Practice, John Wiley & Sons, Hoboken, NJ, USA, 2008.
34. O’Sullivan D., Unwin D., Geographic Information Analysis, 2nd ed. John Wiley & Sons, Hoboken, NJ,
USA, 2010.
35. Peralta J.M., Rubiolo A.C., Zorrilla S.E., Prediction of heat capacity, density and freezing point of liquid
refrigerant solutions using an excess Gibbs energy model, Journal of Food Engineering, 82: 548–558,
2007. doi:10.1016/j.jfoodeng.2007.03.010.
36. Tello Alonso H.A., Peralta J.M., Rubiolo A.C., Zorrilla S.E., Prediction of the freezing point of multi-
component liquid refrigerant solutions, Journal of Food Engineering, 104: 143–148, 2011. doi:10.1016/j
.jfoodeng.2010.12.006.
37. Tello Alonso H.A., Rubiolo A.C., Zorrilla S.E., Prediction of the diffusion coefficients in multicom-
ponent liquid refrigerant solutions, Journal of Food Engineering, 109: 490–495, 2012. doi:10.1016/j
.jfoodeng.2011.10.034.
5 Improving the Performance
of a Partially Loaded
Cold Store by CFD
Fumihiko Tanaka and Fumina Tanaka
CONTENTS
5.1 Introduction...................................................................................................................................... 105
5.2 Modeling Heat and Mass Transfer in a Cold Store............................................................... 106
5.2.1 Governing Equations..................................................................................................... 106
5.2.2 Modeling Heat and Mass Transfer within Loaded Boxes Assuming Porous Media...... 107
5.2.3 Boundary Conditions................................................................................................. 108
5.2.3.1 Inlet, Outlet, and Fan..............................................................................................109
5.2.3.2 Wall............................................................................................................. 109
5.3 Simulation of Heat and Mass Transfer in a Partially Loaded Cold Store.................................111
5.3.1 Governing Equations......................................................................................................111
5.3.2 Boundary Conditions................................................................................................. 113
5.3.3 Effect of Loading Patterns on the Cooling Process of Stacked Products....................113
5.3.4 Effect of Radiation on the Cooling Process of Stacked Products.................................116
5.4 Simulation of Heat and Mass Transfer for Ethylene Control in a Partially Loaded
Shipping Container......................................................................................................................................117
5.5 Conclusions..................................................................................................................................118
References....................................................................................................................................... 119
5.1 INTRODUCTION
The temperature field, which is closely related to the airflow field, affects the quality of stacked
agricultural products in cold storage rooms. The most important factor that affects quality loss is
the temperature related to respiration, ripening and microbial activity; therefore, adequate cool-
ing is required to avoid quality loss and to extend the shelf-life of agricultural products. Cooling
systems designed to provide environmental temperature control need to be carefully evaluated to
ensure that the most efficient cooling operation and uniform temperature distribution is achieved.
Proper methods of cooling may ensure greater temperature uniformity of products stacked in a cold
store and improve the heat transfer efficiency between the products and air. A number of studies
have demonstrated the effectiveness of various CFD (computational fluid dynamics) approaches
in modeling airflow patterns, temperature, moisture and chemical species distributions with the
goal of improving the cooling performance of cold stores [1–6]. Most of the previous works have
discussed the operating parameters of the hardware system. Delele et al. [7,8] used a multiscale
CFD model to investigate the cooling process of model food stacked in cold stores. Chourasia and
Goswami [2,9,10] studied the heat and mass transport phenomena during natural convection cool-
ing of bagged potatoes in cold storage, investigating the moisture loss from potatoes in the stack
during cooling for different values of bulk medium porosity. Refrigerated agricultural products can
either be stored in large bulks or in smaller units or packages. A wide range of packages is used in
105
practice, constructed from a variety of materials, with many different designs and ranging in size
[11]. Ambaw et al. [6] investigated the effect of the material used to construct boxes of apples on the
distribution of 1-MCP gas during cold storage, and used a geometrical model to predict the distribu-
tion of 1-MCP gas in such boxes.
Stacked products substantially affect the flow field in cold stores. Therefore, when optimizing
the cooling process and storage system of agricultural products, it is important to investigate the
effect of the loading pattern on the cooling rate and temperature, humidity and chemical species
distributions [12]. In practice, however, it is difficult to identify the optimum loading pattern based
on empirical data because of limitations imposed by, for example, the cost of experiments and tech-
nical problems with data collection. The CFD can be applied to investigate heat and mass transfer
phenomena, phase changes and chemical reactions while cooling products in a store with various
loading patterns, and is thus a useful method to gain marked benefits at low cost.
The objective of this chapter is to develop a CFD model to investigate the cooling performance
of a partially loaded cold store during the cooling process and storage of agricultural produce. We
demonstrate the effectiveness of a CFD model that predicts storage room air velocity and tempera-
ture distribution in improving cooling efficiency. We present a method of modeling heat and mass
transfer within stacked agricultural produce, assuming porous media. Since some fruits are highly
sensitive to ethylene during storage, it must be removed using absorbents or chemical reactors. In
this case, the chemical species transport model, incorporating phase change or chemical decompo-
sition reaction models, is useful to predict and control ethylene gas. The models introduced in here
will help to understand and improve the cooling performance of refrigerated rooms for agricultural
produce during the cooling process and storage.
5.2 MODELING HEAT AND MASS TRANSFER IN A COLD STORE

In our model, we consider incompressible flows. We employ the system of conservation equations
for mass, momentum and energy, taking into account density changes that result from tempera-
ture variations. The continuous turbulent airflow is solved using the Reynolds-averaged fluid flow
equations:
∂ ∂
ρ+ ( ρui ) = 0 (5.1)
∂t ∂xi
In addition, the gravity term is added in the body force in Eq. (5.2):
∂ ∂ ∂   ∂u j ∂ui   ∂p ∂
( ρu j ) + ( ρui u j ) = µ + − − ( ρui′u′j ) − ( ρ − ρref ) g + Su (5.2)
∂t ∂xi ∂xi   ∂xi ∂x j   ∂x j ∂xi

∂ ∂ ∂  ∂T  ∂
( ρC pf T ) + ( ρui C pf T ) =  λ  − ( ρC pf ui′T ′) + Se (5.3)
∂t ∂xi ∂xi  ∂xi  ∂xi

where u, u’, T, T’, ρ, ρref, μ, p, λ, and Cpf are the mean velocity, fluctuating velocity, temperature,
fluctuating temperature, density, reference density, dynamic viscosity, pressure, thermal conductiv-
ity and specific heat of fluid, respectively. x is the Cartesian coordinates and t is time. The terms
ui′u′j and ui′T ′ are the specific Reynolds stress and specific Reynolds flux, respectively. In order to
solve turbulent flows with the Reynolds-averaged Navier–Stokes (RANS) equations, it is necessary
Improving the Performance of a Partially Loaded Cold Store by CFD 107
to develop turbulence models to estimate the Reynolds stresses and the scalar transport terms. The
mixing length and k-ε models are the most widely used and validated. The two-layer k-ε, Spalart–
Allmaras, Wilcox k-ω, and Menter SST k-ω models were also used in an advanced model.
5.2.2 Modeling Heat and Mass Transfer within Loaded Boxes Assuming Porous Media
We consider airflow in a porous media and two way coupling with the discrete phase. In this case,
the Reynolds-averaged fluid flow equations based on the averaged internal fluid velocity are given by
Equations (5.2) and (5.3). Most studies modeling airflow, heat and mass transfer in storage of stacked
products have used the porous medium approach, and treated turbulence using the conventional tur-
bulence models that are employed in the Reynolds-average approach [7,8]. By assuming thermal equi-
librium between the air and the porous media, the energy Equation (5.3) becomes:
∂ ∂ ∂  ∂T  ∂
φρC pf T + (1 − φ ) ρ pC ppT  + ( ρuiC pf T ) =  λeff −φ ( ρaC pf ui′T ′) + Se (5.4)
∂t ∂xi ∂xi  ∂xi  ∂xi

where ρp and Cpp are the density and specific heat of solid, respectively. ϕ and λ eff are the porosity
and effective thermal conductivity of porous media, respectively. Se is a heat source, representing
the heat of respiration of bulk agricultural produce and evaporation of water from the produce. The
heat exchange with the cooler is also included as an energy source.
The momentum source term Su in Equation (5.2) contains flow resistance due to porous media.
The airflow resistance of porous media Sup was estimated using the Darcy-Forchheimer-Brinkman
equation:
µ 1 ∂2 ui
Sup = −φ ui − φ 2 βij ρui u j u j + µeff (5.5)
K ij 2 ∂x 2j

where Kij and βij are the Darcy permeability and Forchheimer drag coefficient, respectively. The
first and second terms on the right hand side represent the Darcy term and Forchheimer term,
respectively. The Darcy and Forchheimer terms account for viscous drag and internal drag, respec-
tively. The last term on the right hand side represents the Brinkman term, which is important to
resolve the small boundary layer at solid or porous media interfaces. The equation has been used
to describe airflow through vented horticultural packages, and it can give good predictions of the
total pressure drop over the vented porous media load. The effective viscosity (μeff) is expressed as
a function of porosity (μeff = μ/ϕ).
The transport equation for chemical species is:
φ
∂( ρYk )
+φ
∂( ρuiYk )
=φ
∂  ∂Yk  ( )
∂ ρui′Yk′
∂t ∂xi ∂xi  ρ D ∂x  − φ ∂x + Rm + Sm (5.6)
i i
where D, Y and Yʹ are the diffusivity, volume fraction and fluctuating volume fraction of species,
respectively. The terms Rm and Sm take into account the net production of chemical species due to
chemical reaction and creation from source, respectively.
A scheme of forward-only reactions is represented by the stoichiometric equation:
n n
∑ ν ′ M →∑ ν ′′ M
k =1
kj k
k =1
kj k j = 1, 2, 3,… , m (5.7)

ν kj′ and ν kj′′ are stoichiometric coefficients of reactant and product species Mk in reaction j,
respectively. n is the total number of species involved, and m is the number of reactions in the
scheme. The reaction rate r (kmol m–3 s–1) for the above reaction (5.7) is given by:
dc product n
ν′
r= = k f Π cmkkj (5.8)
dt k =1
The net progress reaction is given by:
n
• ν′
q kj = (ν k′′ − ν k′ ) k f Π cmkkj (5.9)
k =1
where cmk represents the molar concentration of the species mk in kmol m–3. The proportionality
constant kf is called the specific reaction rate constant and is expressed as:
 E 
k f = AT α exp  − a  (5.10)
 RT 

A, α, Ea and R are the pre-exponential constant, temperature exponent, activation energy and
universal molar gas constant (=8.314 kJ kmol–1 K–1), respectively. Although the unit of factor T α
depends on the value of α, in most cases α = 0.
The general form of a reversible reaction is given by:
•  n ν′ n
ν′ 
q kj = (ν k′′ − ν k′ )  k f Π cmkkj − kb Π cmkkj  (5.11)
 k =1 k =1 
where kf and kb are the forward and backward reaction rate constants for reaction j. The total rate of
production is given by:

•
qk = ∑q•
kj
(5.12)
The source term used in species transport equations can be written as:
•
Rm = φ MWk q k (5.13)

where MW is the molecular weight.

Finally, the chemical energy equation is considered. Chemical energy is released as heat during
the chemical reaction, and the resulting enthalpy is obtained. The net rate of increase/decrease of
enthalpy due to pressure work must be considered when solving the energy equation.
5.2.3 Boundary Conditions
All CFD problems are defined in terms of initial and boundary conditions, and it is important to
specify these correctly. The initial values of heat and mass flow variables must be specified at all
solution points in the computational domain; fortunately, this is a trivial problem. Therefore, we will
only discuss the boundary conditions required to solve the governing equations.
5.2.3.1 Inlet, Outlet, and Fan

When modeling heat and mass transfer in a cold store, we need to specify the distribution of all
flow variables at inlet boundaries. These variables include inlet air velocity, mass flow, pressure,
turbulent intensity and temperature. In cold store modeling, the air outlet of a refrigerator is some-
times defined as an inlet boundary. The inlet flow distribution and temperature profiles are usually
specified on the basis of experimental data as shown in 5.3.2. Outlet boundary conditions are usu-
ally used in conjunction with inlet boundary conditions. In Section 5.3, we employed inlet and outlet
boundary conditions to realize airflow supplied by a refrigerator fan. The discontinuous pressure
rise across the fan is specified as a function of the velocity. Equations for predicting the swirl veloc-
ity [13] generated by an axial flow fan are given by:
∑
M
Uθ = fm r m ; −1 ≤ M ≤ 6 (5.14)
m =−1
∑
M
Ur = gm r m ; −1 ≤ M ≤ 6 (5.15)
m =−1
where Uθ and Uγ are, respectively, the tangential and radial velocities on the fan surface, f m and gm
are the tangential and radial velocity polynomial coefficients, and r is the distance to the fan center.
f m and gm are determined on the basis of experimental data. In Section 5.4, we applied this boundary
condition to describe the pressure jump at a reactor unit fan and refrigerator.
5.2.3.2 Wall
The wall is the most common boundary considered when solving heat and mass flow problems. The
non-slip condition (ui = 0) is the appropriate condition for velocity components at solid surfaces. In
turbulent flows, the effect of wall boundaries on fluid flow is evaluated by wall function formulae.
For boundary conditions in heat transfer, it is necessary to specify the rate of heat transfer through
the walls. In general, the heat flux q is defined by:
q = K (Tout − Tin )rs → or q = constant (5.16)

where Tout and Tin are the air temperatures surrounding and inside storage, respectively. The overall
heat transfer coefficient K depends on the thermal conductivity and thickness of the conductive solid
material, and on properties of the fluid (density, heat capacity, etc.) and properties of the flow (geom-
etry, turbulence, etc.). The surrounding temperature Tout depends on variables including meteoro-
logical conditions, season, and location, and is sometimes given as a periodic boundary condition.
In long distance transportation, the surrounding temperature is a time-dependent variable. Iwamoto
[14] defined radiant heat coefficient rs (1.0 and 1.2 for night and day, respectively) and added it to
Equation (5.16).
In order to model thermal radiation precisely, the Monte Carlo ray-tracing technique [15] was
useful to simulate the behavior of photons, which carry energy between physical surfaces. This
technique can be described as follows:
The surfaces in the geometry of the walls and stacked products were subdivided into sets of N
non-overlapping primitive surfaces. The source location of photons on the primitive surface and the
initial direction of travel were determined stochastically with a pseudo-random number generator
(0 ≤ ξi ≤ 1). The random sampling of the source location on a primitive surface with local curvilinear
coordinates (rr1 ≤ rr ≤ rr2, sr1 ≤ sr ≤ sr2, tr1 ≤ tr ≤ tr2) consisted of three instances of the pseudo-random
variable 0 ≤ ξi ≤ 1 with i = 1–3.
rr = rr1 + ξ1 (rr 2 − rr1 )

sr = sr1 + ξ2 (sr 2 − sr1 ) (5.17)
tr = tr1 + ξ3 (tr 2 − tr1 )

Each simulated photon required not only a location but also an initial direction of travel. We
generated direction cosines from:
rr′ = ξ4 cos φ
(5.18)
sr′ = ξ4 sin φ

where:
tr′ = 1 − ξ4
(5.19)
φ = π (2ξ5 − 1)

and where ξ4 and ξ5 are two instances of the pseudo-random variable 0 ≤ ξi ≤ 1 with i = 4 and 5. The
initial travelling direction can be expressed as the following unit vector:
e = rr′e x + sr′ e y + tr′e z (5.20)

A photon travelling across the geometry was determined to be absorbed or reflected at a physical
surface according to the following criteria:
ξ6 ≤ ε → absorption (5.21)
ξ6 > ε → reflection
where ξ6 is the pseudo-random variable 0 ≤ ξ6 ≤ 1. The absorbed ratio of photons is simply pro-
portional to the emissivity of the sample (ε), assuming a gray body. In general, the emissivity of
objects can be defined as the ratio of the radiation emitted by the surface to the radiation emitted
by a blackbody at the same temperature. When reflection was determined to occur at a diffuse
surface, a new direction of travel was determined according to the Equations (5.18)–(5.20). Thus,
the directional distribution governing photons at a real surface can be simulated. Tracking photons
across the geometry of a cold store is the most computationally intensive task in our modeling, and
the outcome provides the number of photons absorbed by each surface.
The resulting irradiation on surface i is:
N
n ij
Gi =
1
Si ∑
j =1
nj
S j ε j σ Tj4 (5.22)

where n ij being the number of photons emitted by surface j and incident on surface i. Each primitive
surface Sj was treated as a separate uniform photon source with emissivity εj and temperature Tj.
The Boltzmann constant is indicated by σ. The number of photons nj emitted by surface j is:
S j ε j σ Tj4
nj = N p = N
(5.23)
∑ S ε σT
x =1
x x x
4
Thus, the following equation was derived:
∑n i
j N N
Gi =
1
Si
j =1
Np ∑
x =1
S x ε x σ Tx4 =
1 ni
Si N p ∑S ε σT
x =1
x x x
4
(5.24)

It therefore suffices to add all incident photons on surface i to the photon current ni and multiply
the relative photon current with the available emissive power in the system to obtain the irradiation.
The net radiation flux qrad,i leaving surface i is equal to:
qrad ,i = ε i σ Ti 4 − ε i Gi (5.25)

In order to obtain good estimates of this quantity of interest, it was necessary to generate many
photon tracks.
5.3 SIMULATION OF HEAT AND MASS TRANSFER

IN A PARTIALLY LOADED COLD STORE
In this simulation we investigated the performance of a partially loaded cold store during the cool-
ing process [15]. The cold store used in this simulation is a working prototype of a 12 ft. (3.66 m)
shipping container equipped with a humidity and temperature control system (Mayekawa Co. Ltd.,
Tokyo, Japan). The total dimensions of usable storage are 2.80 m × 2.32 m × 2.18 m. Forty-eight cor-
rugated fiberboard containers (270 mm × 225 mm × 400 mm) filled with PET bottled water were
loaded into the cold store in six different configuration models (model 1–6), as shown in Figure 5.1.
The thermal properties of the corrugated fiberboard containers filled with PET bottled water are
listed in Table 5.1. The flat loading configurations with (model 1) and without air gaps (model 2) and
the high-stacked configuration (model 3) models were used to validate the CFD simulation. The
models were validated successfully in cooling experiments with a mean error of 0.36 m s–1 and
1.4°C for the air velocity in a refrigerated room and the temperature of the product, respectively
[16]. Based on the model, we studied the effect of different loading patterns on the effectiveness of
cooling corrugated fiberboard containers filled with products in a cold store.
The governing equations for the CFD model are introduced in 5.2.1. In order to compute turbulent
flows using RANS equations, it is necessary to develop turbulence models to estimate Reynolds
stresses and scalar transport terms. Though many widely applicable turbulence models that are use-
ful in general-purpose CFD code were considered, the mixing length and k-ε models are the most
widely used and validated. Thus, our simulation accounted for turbulence by means of the standard
Cold air inlet
Moisture inlet
Stacked Products Outlet
FIGURE 5.1 Model geometry of cold store used in heat and mass transfer simulation. (From Tanaka, F., Verboven,
P., Scheerlinck, N., Morita, K., Iwasaki, K., Nicolaï, B.M., 2007. Investigation of far infrared radiation heating as
an alternative technique for surface decontamination of strawberry, Journal of Food Engineering, 79(2), 445–452.)
TABLE 5.1
Boundary Conditions and Thermal Properties of Product Used in
Heat and Mass Transfer Simulation in Partially Loaded Cold Store
Initial Conditions Boundary Conditions
All domains Cold air inlet Moisture inlet Outlets
Temperature 30°C 2.1°C Mean temp. of Free
air in the store
Air velocity 0 m s–1 Measured value Measured value Free
Loading Product
Apparent Apparent specific Apparent
density heat thermal
conductivity
545 kg m–3 3.95 kJ kg–1 K–1 0.43 W m–1 K–1
k-ε model. The model consists of the turbulent kinetic energy Equation (5.26) and its dissipation
rate Equation (5.27):
∂( ρa k ) ∂( ρa ui k ) ∂  µt  ∂k 
+ =  µ +  + Pk − ρa ε (5.26)
∂t ∂xi ∂xi  σ k  ∂xi 

∂( ρa ε ) ∂( ρa ui ε ) ∂  µt  ∂ε  ε ε
+ =  µ +  ∂x  + k C1 Pk − k C2 ρa ε (5.27)
∂t ∂xi ∂xi   σ ε i 


where k, ε, μt, σk (=1.0), σε (=1.3), Pk, are the turbulence kinetic energy, turbulence energy dissipation
rate, turbulent viscosity, turbulent Prandtl number for k, turbulent Prandtl number for ε, and turbu-
lent generation rate, respectively. C1 (=1.44) and C2 (=1.92) are constants in the transport equation
for ε. μt and Pk are given by:
k2 k2
µt = C µ ρa = 0.09 ρa (5.28)
ε ε
∂ui  ∂ui ∂u j  2 ∂uk  ∂ui 

Pk = µt + −  3µt ∂x + ρa k  + Pkb (5.29)
∂x j  ∂x j ∂xi  3 ∂x k i

where Pkb is the turbulent generation rate.
5.3.2 Boundary Conditions
All physical surfaces were specified as non-slip walls. At the interface between fluid and corru-
gated fiberboard containers, a conservative heat flux interface condition was applied. Heat flux
from outside to inside was applied on the floor of the room and the walls that were exposed to the
atmosphere. The cold air and moisture inlets and outlet were arranged as shown in Figure 5.1. Based
on experimental results, the air inlet velocity condition is given by:
 2π xi   2π x j 
ui = −1.55 sin  + 1.55 sin  (5.30)
 0.36   0.36 

 2π xi   πxj 
u j = −3.82 sin  × sin  (5.31)
 0.36   0.36 

 2π xi   πxj 
uk = 7.1 sin   × sin  (5.32)
 0.36   0.36 

Moisture inlet velocity is given by:
ui = −5.0 m s −1 (5.33)

u j = 0 m s −1 (5.34)

uk = 8.7 m s−1 (5.35)

In this study, the humidifying device was used only as a blower without humidification.
Simulations were performed for the conditions outlined in Table 5.1.
5.3.3 Effect of Loading Patterns on the Cooling Process of Stacked Products

Six consignation models with 48 loaded corrugated fiberboard containers were used to investigate
the cooling performance in the store, as shown in Figure 5.2. The mean, maximum and mini-
mum temperatures in the corrugated fiberboard containers were calculated for all loading patterns.
Models 1 and 2 were the flat loaded configurations, models 4 and 5 were the mid-height configu-
rations, and models 3 and 5 were the high-stacked configurations. The volume of the stacked pile
was 1.17 m3. The specific surface areas per unit volume were 10.6, 6.9, 5.8, 5.9, 6.1, and 6.1 m–1 for
models 1, 2, 3, 4, 5, and 6, respectively. These values were almost identical for model 3–6.
FIGURE 5.2 Six Loading Patterns of containers stacked in cold store. (From Tanaka, F., Verboven, P.,
Scheerlinck, N., Morita, K., Iwasaki, K., Nicolaï, B.M., 2007. Investigation of far infrared radiation heating
as an alternative technique for surface decontamination of strawberry, Journal of Food Engineering, 79(2),
445–452.)
The predicted airflow vector field in the empty room is given in Figure 5.3A. The main stream formed
by merging two flows from the cold air and moisture inlets was observed near the ceiling. Subsequently,
it flowed toward the floor and spread widely, before reaching the outlet positioned in the side wall. Figure
5.3B shows the predicted airflow vector field in the flat loaded configuration with air gaps. Although the
main stream was qualitatively similar to the empty model, an upward flow was observed near the center
of the cold store. Figure 5.4 shows the airflow patterns in the central cross section for the six different
FIGURE 5.3 Predicted airflow vector fields in the (A) empty and (B) partially loaded stores. (From Tanaka,
F., Verboven, P., Scheerlinck, N., Morita, K., Iwasaki, K., Nicolaï, B.M., 2007. Investigation of far infrared
radiation heating as an alternative technique for surface decontamination of strawberry, Journal of Food
Engineering, 79(2), 445–452.)
Air inlet Air inlet Air inlet
Air outlet Air outlet Air outlet
Model 1 Model 2 Model 3
Air inlet Air inlet Air inlet
Air outlet Air outlet Air outlet

Model 4 Model 5 Model 6
FIGURE 5.4 Airflow distributions in the loaded store for six different configuration models.
configuration models. In the prediction of airflow in a partially loaded store, cool air flowed out of the
cooler and spiraled around the corrugated fiberboard containers. The results showed a complex airflow
pattern in the store. When the height of the product was increased, the circumfluence over the product
was segmented by the product pile. The results of the simulation indicated that the loading pattern of
product strongly influences the airflow and temperature fields inside the cold store. Figure 5.5 shows the
mean and maximum temperature difference of the product after 27 h cooling from an initial temperature
of 30°C. The final mean temperature of the product decreased with increasing specific surface area;
however, the maximum temperature differences were almost the same for model 1 and 2 despite the dif-
ference in specific surface areas. As a result, the flat loaded pattern with air gaps (model 1) was identified
as the optimal configuration to achieve high uniformity and rapid cooling. Uniform cooling was also
observed for the flat loaded pattern without air gaps (model 2). In both cases, the flat loaded patterns
had an advantage in uniform cooling in the store. On the other hand, the mid-height and high-stacked
configurations had disadvantages in uniformity and rapid cooling. The findings of this study can help to
understand and improve the cooling performance of refrigerated rooms for agricultural products.
FIGURE 5.5 The mean temperature (shaded bar) and maximum temperature difference (unshaded bar)
of the loaded product for all models after 27 h cooling with initial temperature of 30°C. (From Tanaka, F.,
Konishi, Y., Kuroki, Y., Hamanaka, D., Uchino, T., 2012. The use of CFD to improve the performance of a
partially loaded store. Journal of Food Process Engineering, 35(6), 874–880.)
5.3.4 Effect of Radiation on the Cooling Process of Stacked Products

In order to investigate the effect of radiation between the storage walls and the loaded container, we
employed the Monte Carlo ray-tracing technique described in 5.2.3.2. As shown in Figure 5.6, 256
corrugated fiberboard containers separated into 4 blocks with a 5 cm gap between the blocks were
loaded into the cold store described in 5.3. The volume of the stacked pile was 6.22 m3. The specific
surface area per unit volume was 5.46 m–1 for the model used in this study.
The emissivity of a real surface is defined as the ratio of the energy radiated by a real surface to the
energy radiated by a blackbody at the same temperature. The emissivity of the walls and the loaded
containers depend not only on the material but also on the nature of the surface. In this simulation, the
emissivity of the wall was set to 0, 0.22, 0.6 and 0.8 to investigate the effect of thermal radiation on the
cooling process. The emissivity of the corrugated fiberboard containers filled with PET bottled water
was set to 0.84. The Monte Carlo method was employed as a radiation model with 100,000 photon
histories generated in order to reduce error introduced by random events. Figure 5.7 shows the surface
Cold air inlet Moisture inlet
Outlet
Stacked Products
FIGURE 5.6 Model geometry of cold store used in radiation heat transfer simulation.
Wall emissivity = 0.8 0.6

30℃
25℃
20℃
15℃
10℃ 0.22 0
5℃
0℃
FIGURE 5.7 Effect of the emissivity of wall on the surface temperature distribution of stacked container.
TABLE 5.2
The Effect of Wall Emissivity on the Mean and Standard Deviation
of Loaded Product Temperature after 27 h Cooling
Emissivity of wall 0 0.22 0.6 0.8
Mean temperature 19.3°C 18.9°C 18.7°C 18.6°C
Standard deviation 5.90°C 5.98°C 6.09°C 6.12°C
temperature distributions of the containers for four different wall emissivity values after 27 h cooling.
A slight difference in surface temperature of the containers was obtained from radiation heat transfer
simulations. Table 5.2 shows the mean and standard deviation of the temperature of the containers after
27 h cooling. The mean temperature decreased with increasing wall emissivity, whereas the standard
deviation increased slightly. The results of this Monte Carlo simulation or radiation between the storage
walls and loaded containers suggest that radiation influences the cooling performance of a cold store.
5.4 SIMULATION OF HEAT AND MASS TRANSFER FOR ETHYLENE

CONTROL IN A PARTIALLY LOADED SHIPPING CONTAINER
Ethylene is a natural plant growth hormone that induces fruit ripening and accelerates fruit softening
and senescence, and so has a detrimental impact on the product quality and shelf-life of agricultural
products. Since some fruits and vegetables are highly sensitive to ethylene, it is necessary to remove
it using absorbents or chemical decomposition system during storage and distribution. This is of par-
ticular concern when mixed storage of fruits and vegetables cannot be avoided. For example, persim-
mons are highly sensitive to ethylene, and the rapid softening of fruit flesh tissues becomes a serious
problem during storage or transportation with products such as apples that produce a lot of ethylene.
The chemical species transport model with chemical decomposition reaction described in 5.2.2 can be
used to predict and control the ethylene gas concentration in a cold store. In this section, we predicted
the profile of ethylene gas concentration in a partially loaded shipping container [17].
The cold store used in this simulation is a 20 ft. (6.1 m) shipping container equipped with a tem-
perature and ethylene control system as shown in Figure 5.8. The total dimensions of storage with
refrigerator zone are 5.96 m × 2.33 m × 2.29 m. Pallets filled with persimmons (B) or apples (C)
were loaded into the container. The loaded pallets were modeled as porous media. The pressure drop
within the container was determined experimentally. Two decomposition reactor units for ethylene
equipped with fans (gas flow rate of 6.6 × 10 –2 m3 s–1) were placed over the apple pallet side of the
container. In this reactor, the ethylene is decomposed using a pulse ionizer. The decomposition reac-
tion takes place in a reactor unit of volume 7 × 10 –2 m3 and we assume a volumetric reaction. In this
study, the decomposition reaction of ethylene was defined simply as follows: C2H4 + 3O2→ 2CO2 +
2H2O. Based on our experimental results, the forward rate constant kf was set to 4.30 × 10 –8. The
ethylene production rate from the stacked apples (ethylene source) was set to 2.2 × 10 –8 m3 s–1. Two
models, with and without a separation board of height 1.93 m between the apple and persimmon
loads, were employed to assess the efficacy of a separation board. Figure 5.9 shows the distributions
of ethylene mass fraction and air velocity in a central cross section of the shipping container after
0.5 h, with and without a separation board. Although the circumfluence patterns in free space were
similar, the two simulated cases showed different ethylene diffusion patterns. Without the separa-
tion board, ethylene moved directly from the apple to the persimmon. By separating the two stacked
loads using the board, the direct ethylene transport from apples to persimmon piles was suppressed
and the ethylene mass fraction decreased in the persimmon pile. We conclude that a partition should
be placed between loads to prevent the exposure of persimmons to ethylene. It is expected that the
proposed model will be useful to optimize ethylene control systems.
FIGURE 5.8 Model geometry of shipping container used in ethylene transport simulation.
Mass fraction of ethylene ( ×10 -6) Mass fraction of ethylene ( ×10 -6)
1.5 4.0
3.3
1.4
2.6
1.3
1.9
1.2
1.2
1.1
0.5
1.0 0.0
Velocity (m s -1) Velocity(m s -1)

7.00 7.00
5.25 5.25
3.50 3.50
1.75 1.75
0 0
FIGURE 5.9 Predicted ethylene (top) and velocity vector (bottom) distributions on a central cross section of
shipping container: without (left)/with (right) partition board at 0.5 h.
5.5 CONCLUSIONS
This chapter presented a CFD model for investigating the cooling performance of a partially
loaded cold store during the cooling process and storage of agricultural produce. A turbulent
airflow model was employed to predict the heat and mass transfer in the cold store and ship-
ping container. The governing equations were solved numerically using the Reynolds-averaged
fluid flow model. Using this model, we investigated the effect of different loading patterns
of corrugated fiberboard containers filled with products on the effectiveness of cooling. The
results provided the optimal loading configuration to achieve high uniformity and rapid cool-
ing. Radiant heat transfer between the cold store walls and the stacked agricultural products
was also investigated, and the effect of radiation was discussed. In order to control ethylene
produced by fruit and vegetables, a chemical species transport model incorporating a model for
chemical reaction was employed, and the mass fraction of ethylene was assessed in a shipping
container equipped with ethylene decomposition reactors. In this model, the loaded products
were considered porous media and the distribution of ethylene mass fraction was predicted in
a computational domain including porous zones. It is concluded that our models will help to
understand and improve the cooling performance of refrigerated rooms for agricultural pro-
duce during the cooling process and storage.
REFERENCES
1. Nahor H.B., Hoang M.L., Verboven P., Baelmans M., Nicolaï B.M., CFD model of airflow, heat and
mass transfer in cooling stores, International Journal of Refrigeration, 28: 368–380, 2005.
2. Chourasia M.K., Goswami T.K., CFD simulation of effects of operating parameters and product on heat
transfer and mass loss in the stack of bagged potatoes, Journal of Food Engineering, 80(3): 947–960, 2007.
3. Konishi Y., Tanaka F., Uchino T., Hamanaka D., CFD prediction of the temperature distribution within
a refrigerated truck filled with fruit and vegetables during transport, Transactions of the Japanese
Society of Refrigerating and Air Conditioning Engineers, 26(2): 159–165, 2009.
4. Delele M.A., Vorstermans B., Creemers P., Tsige A.A., Tijskens E., Schenk A., Opara U.L., Nicolaï
B.M., Verboven P., CFD model development and validation of a thermonebulisation fungicide fogging
system for postharvest storage of fruit, Journal of Food Engineering, 108(1): 59–68, 2012a.
B.M., Verboven P., Investigating the performance of thermonebulisation fungicide fogging system for
loaded fruit storage room using CFD model, Journal of Food Engineering, 109(1): 87–97, 2012b.
6. Ambaw A., Verboven P., Defraeye T., Tijskens E., Schenk A., Opara U.L., Nicolaï B.M., Effect of
box materials on the distribution of 1-MCP gas during cold storage: A CFD study, Journal of Food
Engineering, 119(1): 150–158, 2013.
7. Delele M.A., Schenk A., Tijskens E., Ramon H., Nicolaï B.M., Verboven P., Optimization of the humidi-
fication of cold stores by pressurized water atomizers based on a multiscale CFD model, Journal of
Food Engineering, 91(2): 228–239, 2009a.
8. Delele M.A., Schenk A., Tijskens E., Ramon H., Nicolaï B.M., Verboven P., Evaluation of a chicory root
cold store humidification system using computational fluid dynamics, Journal of Food Engineering,
94(1): 110–121, 2009b.
9. Chourasia M.K., Goswami T.K., Simulation of transport phenomena during natural convection cooling
of bagged potatoes in cold storage, part I: Fluid flow and heat transfer, Biosystems Engineering, 94(1):
33–45, 2006a.
10. Chourasia M.K., Goswami T.K., Simulation of transport phenomena during natural convection cooling of
bagged potatoes in cold storage, part II: Mass transfer, Biosystems Engineering, 94(2): 207–219, 2006b.
11. Verboven P., Flick D., Nicolaï B.M., Alvarez G., Modeling transport phenomena in refrigerated food bulks,
packaged and stacks: Basics and advances, International Journal of Refrigeration, 29: 985–997, 2006.
12. Xie J., Qu X.-H., Shi J.-Y., Sun D.-W., Effect of design parameters on flow and temperature fields of a
cold store by CFD simulation, Journal of Food Engineering, 77: 355–363, 2006.
13. ANSYS Inc., Fluent Release 14 User’s Guide, Canonsburg, PA, USA, 2011.
14. Iwamoto S., A numerical prediction of indoor thermal environment: With floor panel heating in peri-
odical steady state. Summaries of technical papers of Annual Meeting Architectural Institute of Japan,
Environmental Engineering, 801–802, 1994.
15. Tanaka F., Verboven P., Scheerlinck N., Morita K., Iwasaki K., Nicolaï B.M., Investigation of far infra-
red radiation heating as an alternative technique for surface decontamination of strawberry, Journal of
16. Tanaka F., Konishi Y., Kuroki Y., Hamanaka D., Uchino T., The use of CFD to improve the performance
of a partially loaded store, Journal of Food Process Engineering, 35(6): 874–880, 2012.
17. Umeno Y., CFD applications to advances preservation technique for agri-food products, Ph.D. disserta-
tion at the Graduate School of Bioresource and Bioenvironmental Sciences, Kyushu University, 2016.
6 CFD Investigation of Fresh
Produce Cooling Processes
and Effects of Package Stacking
Umezuruike Linus Opara, Alemayehu Ambaw, and Tarl Berry
CONTENTS
6.1 Introduction....................................................................................................................... 122
6.2 Types of Fresh Produce Cooling Processes........................................................................... 122
6.2.1 Precooling.................................................................................................................. 122
6.2.2 Cold Storage Room.................................................................................................... 123
6.2.3 Transport (Shipping) Cooling................................................................................. 124
6.2.4 Refrigerated Display Cabinet........................................................................................ 124
6.3 Experimental Approaches to Investigating Fresh Produce Cooling....................................... 125
6.3.1 Airflow Measurement.................................................................................................... 125
6.3.2 Temperature and Relative Humidity (RH) Measurements........................................ 125
6.3.3 Energy Consumption................................................................................................. 128
6.4 CFD Modeling of Fresh Produce Cooling Processes............................................................ 128
6.4.1 Mathematical Model.................................................................................................. 129
6.4.2 Geometry and Flow Domain..................................................................................... 129
6.4.3 Discretization of the Model Geometry...................................................................... 130
6.4.4 Modeling Mass Transfer............................................................................................ 132
6.4.5 Calculating the Refrigeration Heat Load................................................................... 133
6.4.6 Modeling Produce Quality.............................................................................................135
6.4.7 Model Validation................................................................................................................ 136
6.5 Modeling Fresh Fruit Cooling Inside Ventilated Packaging....................................................137
6.5.1 Apples..............................................................................................................................137
6.5.2 Citrus.................................................................................................................................... 139
6.5.3 Pomegranate....................................................................................................................139
6.5.4 Strawberries........................................................................................................... 139
6.5.5 Table Grape................................................................................................................ 141
6.6 Effects of Package Stacking Pattern...................................................................................... 141
6.7 Conclusions............................................................................................................................ 143
Nomenclature.........................................................................................................................144
References....................................................................................................................................... 145
121
6.1 INTRODUCTION
With the ever-increasing cost of energy and attention to environmental problems, understanding
energy use and exploring energy saving options is becoming a priority for the economic sector,
including food and other bioprocesses. Computational fluid dynamic (CFD) has been applied to
predict the airflow, heat and mass transfers in postharvest handling of horticultural products along
the cold chain. Fundamentals of the CFD modeling steps are presented in this chapter together with
their application in postharvest cooling processes. This chapter consolidates and relates various
CFD techniques used to study the postharvest handling of produce, starting from the precooling to
retail display. The modeling approaches are discussed following the normal path of the postharvest
life of fruits: from precooling to cold storage, refrigerated transport, and retail (display) cooling.
The main objective of the modeling task, basic procedure of the modeling approach and execution
of the mathematical models are presented.
6.2 TYPES OF FRESH PRODUCE COOLING PROCESSES

Food losses have been in the order of one third of the total produced for human consumption across
the globe [1,2]. A significant amount of these losses can be avoided by employing temperature
management in the cold chain. This method, when effectively executed, has been shown to be
highly cost effective compared to continually increasing production. Temperature management
includes:
1. Precooling of freshly harvested fruits and vegetables

2. Short- or long-term storage of perishables
3. Cold and controlled conditions transport
4. Marketing displays (cold handling of produce during storage and displays at markets)
6.2.1 Precooling
Precooling is a process of rapidly removing the field heat from freshly picked perishables down
to their optimum storage temperature. It is applied before fruits and vegetables are placed in
long-term refrigerated storage or before long distance transportation. Different methods of
precooling are available, including: room cooling, forced-air cooling (FAC), vacuum cooling,
hydrocooling or spray cooling and package icing. FAC is the most preferred cooling technique
[3,4]. It involves forcing chilly air through stacked fruit along an induced pressure gradient.
Different fan positions and produce stacking arrangements have proven successful for FAC. The
tunnel horizontal airflow system (Figure 6.1) is the most common FAC arrangement [5]. In this
arrangement two parallel rows of produce are positioned approximately 0.5 to 1 m apart, form-
ing a tunnel. The top and back of the tunnel are covered by cloth or plastic sheet (Figure 6.1).
At the front end of the tunnel fans are mounted to pull chilly air through the space between the
rows and out through the fans. Refrigerated air travels horizontally through the stacked produce
(Figure 6.1b).
The box design (vent area, shape, number, position, etc.), the box arrangement, the thermophysi-
cal properties of the produce, the fruit-stacking pattern within the package, and the ambient con-
ditions determine the rate and uniformity of cooling. This process can be made more efficient by
improving box design, box stacking configurations or orientation. Carton design must allow a mini-
mum of 5% side-wall venting [5]. The analysis and quantification of the interaction of the numerous
factors is a complex problem. Mathematical models, especially computational fluid dynamics, have
been used to assist studies of this operation [6,7].
CFD Investigation of Fresh Produce Cooling Processes and Effects of Package Stacking 123
FIGURE 6.1 The tunnel horizontal precooling arrangement (a) zoomed view of the major airflow path
through the stacked produce during a horizontal precooling process. Color on flow lines corresponds to veloc-
ity magnitudes.
6.2.2 Cold Storage Room

Cold storage room is a system of preserving the quality of agricultural materials and preventing
deterioration for a specific period beyond their normal shelf life. In cold storage room, a bin or
carton of produce is simply placed inside a room fitted with a refrigeration unit (Figure 6.2). Cold
storage rooms can also be used for additional treatments like gassing and fungicide applications
[8,9]. The produce can either be handled in bulk or placed inside containers or packages of unique
design. As in the precooling process, the effectiveness of the cooling depends on the flow resistance
that is induced by the container and the product (Figure 6.2b). Non-uniform flow of air inside the
cool store could cause uneven cooling and non-uniform distribution of fumigants, leading to loss
of product quality.
FIGURE 6.2 Schematics showing produce stacking, cooling unit, and airflow directions inside cold stor-
age room (a) and closer look of the airflow profile inside of a bin of fruit (b). Color corresponds to airflow
magnitude.
FIGURE 6.3 Schematic showing bottom-air delivery T-bar reefer (a), closer view of the T-bar floor
structure (b) and the section view showing the dimensions of the T-bar (c).
6.2.3 Transport (Shipping) Cooling

Fruits and vegetables for distant markets should be kept cool and at the optimum condition during
transit. This is accomplished by using cargo that has its own, self-powered system to keep itself
cool. Reefers are the most common means of transporting fresh produce to distant markets [10,11].
Reefers are designed to distribute chilled air from the floor via specific T-shaped decking (Figure
6.3). The air delivery system should be powerful enough to ensure a sufficient and uniform flow of
air through the stacked produce inside the shipment.
In addition to the design of the reefer container, the package design (size of boxes and vent-hole
design) and arrangement (stacking patterns) significantly influence the performance of the cool-
ing and the space utilization [10–12]. Packaging boxes, if not properly designed and appropriately
stacked, can limit the uniformity and sufficiency of the cold air distribution. It is crucial that pack-
age design and arrangement should take the vertical airflow inside a reefer into account.
6.2.4 Refrigerated Display Cabinet

Fruits and vegetables should be kept at optimum temperature, and radiation and other sources of
heat should be minimized during sales in stores and supermarkets. Retail (display) cooling systems
accomplish this by employing a specialized refrigeration technique. While keeping the produce
cool, display coolers should allow good visibility and ensure free access to stored food for shop
costumers. This is accomplished by a virtual insulation barrier called the air curtain developed by
recirculating air from the top to the bottom of the case (Figure 6.4). The air curtain is a non-physical
barrier between chilly air inside the case and the warm shop environment. Airflow from the top of
the case provides a curtain between customer and the cold produce. This air passes over all the food
products, resulting in heat transfer from the food to the air, which keeps the produce at a predefined
temperature. Simultaneously, heat is transferred from the shop environment to the air curtain. This
causes the temperature of the air curtain to increase and reduces the effectiveness of the air curtain
in the lower compartments of the display case. It is important to quantitatively determine the amount
of the curtain reaching the bottom of the case to be re-chilled and the amount that escapes to the
shop environment. This is important to optimize the energy usage and reduce the discomfort to the
FIGURE 6.4 (a) Components of a vertical multideck display cabinet and (b) CFD model of the airflow path
inside the display cabinet. (Taken from Roşca, R., Ţenu, I. and Cârlescu, P., 2017. Food Chilling Methods and
CFD Analysis of a Refrigeration Cabinet as a Case Study. In: Refrigeration. London: IntechOpen Ltd.)
consumer [13]. Hence, the study of air curtains is required because these can be easily disrupted
by air circulation in front of the cabinet or by the consumers taking food from the shelves [13–15].
6.3 EXPERIMENTAL APPROACHES TO INVESTIGATING

FRESH PRODUCE COOLING
6.3.1 Airflow Measurement
Airflow is one of the key components that determine the performance of cooling process. Though
CFD gives a spatially more resolved visualization and quantification of the airflow, it should first be
validated using experimental data. A wide variety of devices are commercially available for mea-
suring air velocities. Flowmeters measure different flow quantities with different technologies and
it is difficult to categorize them by one parameter. Typically, airflow meters are referenced by: the
technology employed, the configuration of the instrumentation, the physical quantity measured, and
flow quantity converted. Table 6.1 summarizes the commonly used airflow measuring techniques.
Wind tunnels are frequently used to measure the airflow resistance of stacked produce [10,11]. In
this setup the stacked produce is placed under a known pressure gradient and the corresponding airflow
velocity measured using one of the instruments discussed above. Alternatively, a dedicated FAC system
can be used to measure the airflow characteristics as well as the cooling characteristics [16,17].
6.3.2 Temperature and Relative Humidity (RH) Measurements

Temperature and relative humidity (RH) are crucial factors that control respiratory activity, physi-
ological disorders and growth of microbial pathogens during storage of fruits. Hence, it is crucial to
control the temperature and humidity of the cooling environment to reduce waste.
Temperature measuring sensors come in a wide variety and have one thing in common: they
all measure temperature by sensing some change in a physical characteristic. The basic types of
temperature measuring sensors are: Thermistors, thermocouples, RTDs (resistive temperature
detectors), digital thermometer ICs and analog thermometer ICs. Thermistors are temperature
TABLE 6.1
Commonly Used Airflow Measuring Techniques
Technology Measure Pros Cons
Differential Elbow Pressure • Low to medium initial set • Medium to high pressure
pressure Nozzle up cost drop
Orifice • Can be used in wide • Limited sensitivity
Pitot Tube ranges of fluid phases and
Venturi flow conditions
Wedge • Simple and sturdy
structures
Thermal CTA Heat transfer • Medium initial set up cost • Fragile, high repair cost
CCA • Low pressure drops • For (clean) gas only
• Require regular calibration
Vane anemometer Rotational • Simplicity • Unsuitable for use within
speed small void spaces
Ultrasonic wind Doppler Acoustic • No moving parts • Higher initial set up cost
sensors waves • Minimal maintenance cost
• High precision
Tracer gas CO2 Concentration • Can simultaneously • Not suitable for use in a
methods CO of the tracer generate spatial system where measurements
gas distribution of the airflow are required over an
extended time
Laser Doppler Photodetector/ Scattered • Insensitive to fluid • Only a single point can be
Anemometry laser light temperature, density, or sampled at any given time
(LDA) composition
Particle Image Optical/laser Location of • Provides instantaneous • Use hazardous lasers and
Velocimetry tracer velocity 3D vector require high-resolution,
(PIV) Particles in measurements in an entire high-speed cameras, which
time cross-section of the fluid bring cost and safety
flow constraints
transducers based on a thermally sensitive resistor that exhibits a large, predictable, and precise
change in resistance correlated to variations in temperature. A thermocouple consists of two wires
of different metals connected at two points. The varying voltage between these two points reflects
proportional changes in temperature. RTDs (resistive temperature detectors) are a lot like thermis-
tors in the sense that they are resistive temperature transducers. An RTD consists of a film or, for
greater accuracy, a wire wrapped around a ceramic or glass core. The most accurate RTDs are made
using platinum but lower cost RTDs can be made from nickel or copper. They provide high accuracy
over a range of roughly −200 to 500°C. Digital thermometer ICs are semiconductor-based tempera-
ture sensors placed on integrated circuits (ICs). These sensors are effectively two identical diodes
with temperature-sensitive voltage vs. current characteristics that can be used to monitor changes in
temperature. Table 6.2 gives a summary of the commonly used temperature sensors.
By far the most common way of temperature measurements in the postharvest cold handling of pro-
duce is with a thermocouple. Thermocouples can be used to measure the air temperature in the cooling
environment and to monitor the time history of the produce core temperature during cooling [16,17].
Spatiotemporal temperature measurement data were processed to estimate cooling coefficients,
cooling times and cooling uniformities to perform comparative assessment of various configurations
and analyze air temperature distribution in cold stores [18]. Also, direct measurement data was used to
create contour maps of the profile of temperature, relative humidity and air velocity in cold stores [19].
Humidity is the measure of the amount of moisture in the air. There are many techniques that
can be used to measure humidity [20]. Table 6.3 summarizes humidity measurement techniques.
TABLE 6.2
Comparison of the Commonly Used Temperature Sensors
Typical Temperature
Sensor Type Range (°C) Pros Cons
Thermistor −40 to 125 High sensitivity Limited range
Fast Fragile
Low cost Current source required
Durable Self-heating
Small size
Thermocouple −200 to 1450 Self-powered Least stable
Simple Least sensitive
Inexpensive Reference and calibration is required
Wide variety
Wide range
RTD −260 to 850 Most stable Expensive
Most accurate High thermal initiative (slow)
Linear output Fragile
Accurate
Analog IC −40 to 125 Easy to use More expensive than thermistores
Limited temperature range
Digital IC −55 to 125 Simple to use with Requires a microcontroller
microcontrollers Expensive than thermistores
Accurate
TABLE 6.3
Guide to Humidity Measurement
Hygrometer Type Sensing Principle
Absorption spectrometer Infrared light absorption by water vapor
Acoustic Humidity-dependent acoustical transmission or resonance
Adiabatic expansion “Cloud” formation in chamber on expansion cooling of sample gas
Cavity ring-down Decay time of absorbed, multiply-reflected infrared light
spectrometer
Color change Crystals or inks using cobalt chloride or other chemicals changing color with hydration
Condensation Temperature of formation of water or ice, on cooling humid air
Electrical impedance Relative humidity sensor – electrical change on absorption of water vapor into
(typically) polymer film
“Dew-point” probe – electrical change on absorption of water vapor into (typically)
porous metal oxide film.
Electrolytic (phosphorous Electric current proportional to dissociation of water into hydrogen and oxygen
pentoxide)
Gravimetric By weighing – mass of water gained or lost by humid air sample
Mechanical Dimensional change of humidity sensitive material (hair, polymer, paper …)
Optical fibre Change in reflected or transmitted light, using hygroscopic coating, or optical grating
Quartz crystal resonator Change in resonant frequency due to mass of surface-adsorbed water
Saturated lithium chloride Conductivity of hygroscopic salt
Wet- and dry-bulb Humidity-dependent evaporative cooling of wetted surface
(psychrometer)
Zirconia Humidity derived from oxygen content of gas
Source: Taken from Bell, S., 2012. A beginner’s guide to humidity measurement. National Physical Laboratory, 2.
FIGURE 6.5 The main energy consuming components of the cold storage system (a) and a typical measure-
ment of the energy demand taken within 72 h (b).
6.3.3 Energy Consumption
The energy usage of a cooling facility depends greatly on the type and configuration of the cool-
ing system (efficiency of individual components). Storage at chilled (above zero) or frozen (below
zero) temperatures may be needed for a proper preservation. There are reports showing cold stores
consume over eight times more energy than their efficient counterpart [21,22]. Implementations
of energy saving strategies require the precise measurement of energy consumption. Stand-alone
or portable meters are available to measure power consumption. Normally, the electric usage of a
cooling process is obtained through an energy audit. To measure the energy consumption, the elec-
tricity usage of the system elements such as the compressor, condenser and fan of the evaporator
(Figure 6.5a) must be measured within a specific period (Figure 6.5b). This can be done using elec-
tric power measurement instruments. Modern watt meters easily provide volt, current and power
factors to calculate the electrical power in watts.
6.4 CFD MODELING OF FRESH PRODUCE COOLING PROCESSES

The main objectives of modeling produce cold handling processes are to visualize and quantify
the airflow, cooling rate, and cooling uniformities and to calculate refrigeration heat loads. Many
studies implemented CFD and successfully solved the 3D airflow velocity field, the spatiotemporal
temperature distributions, cooling rate, cooling uniformities and refrigeration heat loads of stacked
produce. Such models are used to predict precooling time, identify hot spots and cold spots in the
cooling environment and calculate energy utilization to optimize the operation of the cooling pro-
cess. The effect of package designs, stack arrangements and operational parameters on the cooling
process were investigated using CFD.
6.4.1 Mathematical Model
CFD simulation is based on the mathematical statements of the conservation laws (conservation
of mass, momentum, and energy) that govern fluid flow. In this approach fluid is assumed to be a
continuum and its properties, like pressure, flow velocity, density and temperature, to be differen-
tiable. Further, by taking time-averaged motion of the fluid flow, the well-known Reynolds-averaged
Navier–Stokes equations (or RANS equations) are obtained (Equations 6.1 to 6.3) to model fluid
flow based on the time-averaged flow velocity.
∇ ⋅ U = 0 (6.1)

∂U   µ + µt   1
+ ∇ ⋅ (U ⊗ U ) − ∇ ⋅    ∇U  − SU + ∇p = 0 (6.2)
∂t  ρa   ρa

 ∂T 
ρaC pa  a + U ⋅ ∇Ta  − ∇ ⋅ ( ( ka + kt )∇Ta ) − Q = 0 (6.3)
 ∂t 
where U is the vector of the velocity (m s−1), t is time(s), μ is the dynamic viscosity of air (kg m−1s−1),
μt is the turbulent eddy viscosity (kg m−1s−1), p is pressure (Pa) causing the fluid flow and SU (m/s2)
is any momentum source inside the fluid domain. SU accounts for any momentum source in the flow
domain, Cpa (J kg−1K−1) is the heat capacity of air, ρa (kg m−3) is the density of air Ta (K) is the air
temperature, ka (W m−1K−1) is the thermal conductivity of air, kt (W m−1K−1) is the turbulent thermal
conductivity. The turbulent thermal conductivity is a function of the turbulent eddy viscosity, the
heat capacity and Prandtl number of air.
The turbulent eddy viscosity is calculated by the turbulence model. Turbulence modeling is
a key subject in CFD simulations. There are several different formulations for solving turbulent
flow problems. The most widely used are the k-ε and the k-ω eddy viscosity models. Another very
popular turbulence model which has proven to be very successful in different applications is the
Shear Stress Transport k-ω (SST k–ω) model, which combines the best aspects of the k-ε and k-ω
models. Most frequently, the SST k–ω turbulence model is used in modeling flow in postharvest
cold handling systems [23,24].
6.4.2 Geometry and Flow Domain

When analyzing cooling of stacked perishables using CFD approaches, the geometry of the system
to be analyzed is first modeled. This system can be an individual fruit, a single container/box filled
with produce, a group of packages arranged in a layer or a column, an entire pallet or a fully loaded
cool room (Figure 6.6). The complexity and challenge of acquiring the digitized model geometry
increases when going from an individual fruit to the loaded cool room. In many cases, due to limited
computational facility and time, the actual geometry is simplified. Assumptions include geometry
and flow symmetry, or nearly identical flow condition between parts—for instance, through rows or
columns of a stack. The geometry model fixes the boundaries of the computational domain. Proper
definition of interfaces between solids (produce and packing materials) and fluid (air) is crucial for
realistic and accurate capturing of the airflow and cooling process. A properly generated geometry
considerably reduces the subsequent modeling process.
Detailed CFD analysis of large postharvest systems, like reefers and cool store rooms, remains
a challenge due to high computational demand. Fruit inside cold storage room and reefer containers
is packed in bulk in boxes and stacked on pallets forming complex geometry. For such a system,
model discretization is very expensive due to the high size ratio of geometries that make up the
FIGURE 6.6 (a) Individual fruits, (b) fruit loaded box, (c) a layer of stack to study a horizontal precooling
process, (d) palletized stack of produce, and (e) schematics of half cool store room used as model geometry
based on symmetry principle.
domain [8,10,11]. To solve this problem, the actual geometry is usually simplified by assuming
porous media. In this approach, stacked fruit is simplified to a porous medium consisting of a fluid
phase (air) and a solid phase (fruit and packaging materials) with constant volume fraction in space
and time. This approach is beneficial because it manages the computational power requirement and
solves the airflow and heat transfer in a reasonably short computational time.
6.4.3 Discretization of the Model Geometry

This step undertakes the partitioning of the model geometry into several nonoverlapping subdomains
called a computational grid. There are several cell shapes available to do the discretization (Table 6.4).
Cells are arranged in the geometry into a grid. Depending on the shape and complexity of the
geometry, cells are arranged into one of the three grid types: structured grids, unstructured grids
and hybrid grids (Figure 6.7). Structured grids are identified by regular connectivity. The possible
element choice in 3D is hexahedron. This grid is highly space efficient and is characterized by a
better convergence and higher resolution than an unstructured grid. Cell elements can be irregularly
connected to form an unstructured grid. This allows for any possible element that a solver might
be able to use. These grids typically employ tetrahedra in 3D. A hybrid grid contains a mixture
of structured portions and unstructured portions. This can help in optimizing grids in complex
geometries. Parts of the geometry with regular structure can have structured grids and irregular
structures can have unstructured grids. Due to the relatively complex geometry of packed fruits,
unstructured or hybrid grids are commonly used in CFD models of cold handling systems in the
postharvest (see Table 6.5).
TABLE 6.4
The Basic 3-Dimensional Cells or Elements Used in 3D CFD Discretization
Cell Shapes Description Remark
Tetrahedron Has 4 vertices, 6 Typically employed for
edges, and is bounded structurally complex
by 4 triangular faces geometry
Pyramid Has 5 vertices, 8 Transition element,

edges, bounded by 4 between square and
triangular and 1 triangular faced elements
quadrilateral face and other in hybrid grids
Triangular prism Has 6 vertices, 9 Resolves boundary layer

edges, bounded by 2 efficiently
triangular and 3
quadrilateral faces
Hexahedron Has 8 vertices, 12 For the same cell

edges, bounded by 6 amount, the accuracy of
quadrilateral faces solutions in hexahedral
meshes is the highest.
FIGURE 6.7 Grid arrangement: structured grid (a), unstructured tetrahedral grid (b), and hybrid grid (c).
TABLE 6.5
CFD Simulation Techniques in the Postharvest
Numerical
Reference Fruit Model Geometry Technique Meshing Grid Sensitivity Study
[25] Apple Single fruit FVM Tetrahedral, Against “wall Y+” and
unstructured product temperature
[26,27] Apple Single package FVM Tetrahedral, Skewness and the y+
unstructured parameters
[28,29] kiwifruit Single layer taken FVM Tetrahedral, Using the Richardson
out of a pallet unstructured extrapolation
[30,31] Citrus Single layer taken FVM Hybrid grid Richardson extrapolation
out of a pallet (hexahedral and
tetrahedral cells)
[32] Strawberries Single layer taken FEM Tetrahedral elements Product temperature
out of a pallet
[33,34] Spherical Single package/ FVM Tetrahedral hybrid “wall y+” and product
plastic balls half a pallet mesh temperature
[35] Apple Single package FVM Hybrid grid Richardson extrapolation
(tetrahedral and method
hexagonal cells)
[8,36] Apple Cool store room Porous Hybrid grid Richardson extrapolation
medium (tetrahedral and
FVM hexagonal cells)
[16] Pomegranate Single layer/ FVM Tetrahedral elements Richardson extrapolation
pallet method
The assessment of the accuracy, sensitivity and robustness of the generated grid is a very impor-
tant stage of the CFD modeling process. The grid quality should be assessed through a dedicated
grid sensitivity study, by measuring the sensitivity of the model prediction with respect to mesh size.
Frequently, the Richardson extrapolation method [8,10,11] is used to perform grid sensitivity analy-
sis. This is a method for obtaining a higher-order estimate of the flow fields from a series of lower-
order discrete values. In this technique, extrapolation is made from the results of at least two different
grid solutions. Grid quality measurement can be supported by various mesh quality metrics gener-
ated by the meshing software: the orthogonality (especially at the boundaries), relative grid spacing
(15% to 20% stretching is considered a maximum value), and grid skewness are a few. Additionally,
the spacing between cells should be in accordance with the desired resolution of key features.
Simulation or solving the flow problem requires numerical techniques that transform the par-
tial differential equations representing conservation laws into discrete algebraic equations over the
elements or cells. This is accomplished using one of the three most popular techniques: the finite
difference method (FDM), the finite volume method (FVM), or the finite element method (FEM).
A common opinion is that the FDM is the easiest to implement and the FEM the most difficult. The
finite element method requires sophisticated mathematics for its formulation and large memory
space. Due to its robustness, ease of coding and flexibility, the FVM is the most used. Table 6.5
summarizes the commonly employed CFD techniques in the postharvest.
6.4.4 Modeling Mass Transfer

In addition to the airflow, the distribution of moisture and other gaseous substances are important
during cold handling of produce. Gaseous substances like 1-methylcyclopropen (1-MCP) and eth-
ylene are frequently used to control respiratory activity and physiological disorders during cold
handling of produce. Controlled atmosphere (CA) and ultra-low oxygen (ULO) storage methods
are based on controlled oxygen and carbon dioxide concentrations to minimize the respiration of
the products and preserve quality for several months of storage. Hence, uniformity of distribution
of these substances in the cool store atmosphere is crucial. The CFD method can also be used to
model the transport and distribution of moisture and other gaseous substances. However, since CFD
is basically for fluid flow, incorporating the transfer of moisture and gaseous substances requires
additional effort. Basically, these processes are transient in nature and are governed by additional
parameters like: permeation, diffusion, adsorption, desorption and respiration kinetics.
For instance, incorporating the moisture transport in the CFD model requires a modification to
the basic heat transfer model discussed previously. It should incorporate the respiration and tran-
spiration processes of the produce and the heat loss/generation due to evaporation/condensation of
water at the surface of the produce as given by Equations (6.4) and (6.5):
 ∂T 
( ρaC pa )  a + U ⋅ ∇Ta  = ∇ ⋅ ( ( ka + kt )∇Ta ) + h pa (Tp − Ta ) (6.4)
 ∂t 

∂Tp
( ρ pC pp ) = ∇ ⋅ ( k p ∇Tp ) + h pa (Ta − Tp ) + Qr − Qv (6.5)
∂t
where hpa(Tp – Ta) is the heat exchange across the interface between the produce and the cool store
atmosphere, Qr is the respiration heat generation and Qv is the heat loss due to evaporation of water
from the surface of the produce. Modeling the moisture distribution needs Equation (6.6) to be
coupled to the basic fluid flow equations (Navier–Stokes equation).
∂G
ρa + ∇ ⋅ ( GU − ( Da + Dt )∇G ) = m (6.6)
∂t
where G is the moisture concentration (relative humidity), Da is the diffusivity of moisture in air,
Dt is the turbulent diffusion coefficient and m is the rate of evaporation of water from the surface
of produce into the cool store atmosphere, which is governed by the equilibrium between the room
moisture content and water activity of the produce. The turbulent diffusion coefficient, through the
empirical turbulent Schmidt number, is a function of the turbulent viscosity (Dt = μt/Sct).
6.4.5 Calculating the Refrigeration Heat Load

The result of the CFD simulation is frequently postprocessed to generate the required information
to perform analysis of the refrigeration heat load. The refrigeration heat load of a produce cold han-
dling process depends on the configuration and operation of the process. Generally, it includes: the
field heat to be removed from the produce (for a precooling stage), the heat load from the evaporator
fan-motors, the heat load from other fan-motor (for instance, fans that drive air through the stack
during a precooling process) and the conduction heat load from the outside environment. Heat loads
due to lighting and other activities are frequently not considered during calculation. Depending
on the location of the cold handling facility, the solar heat load may be significant and should be
incorporated.
The field heat is calculated based on the average produce initial temperature and the cooling
air temperature. The heat of respiration and the effect of moisture transport during a precool-
ing process are frequently neglected. However, in cold storage room, the respiration heat load
is significant [36]. The heat gain from the outside environment through (walls, ceiling, floor and
the door) is calculated based on the outside ambient condition. The type, thickness and thermal
properties of the insulation of the wall material should be clearly known to calculate the transmis-
sion heat load (Equation 6.7).
Qw = uw Aw (T∞ − Tw ) (6.7)

where Qw is the transmission heat load (W), uw (W m−2 °C−1) is the wall heat transfer coefficient, Aw
(m2) is the total surface area of the walls, ceiling and flour of the cool store room, Te is the outside
air temperature (°C), and Ta is the inside temperature of the cool store. Heat loads due to illumina-
tion and other activities are usually assumed negligible in the calculation. The wall heat transfer
coefficient, Uw was estimated using Equation (6.8).
1 1 1
uw = + + (6.8)
1 x 1
hi k h0
where x is the wall insulation thickness (m), k is the thermal conductivity of wall material (W m−1
°C−1), hi is the convection heat transfer coefficient for air corresponding to the inside wall of the cool
store (W m−2 K−1), h 0 is the convection heat transfer coefficient for air corresponding to the outside
wall of the cool store (W m−2 K−1).
FIGURE 6.8 Components of the refrigeration loads of cool room with a constant set point temperature (full
curve) and with cyclic set point temperature (dot curve) and with on-off cycling (16 h off/8 h on) of the cool-
ing unit (dash curve). Heat loads: due to respiration of the produce (a), due to conduction through boundaries
(b), fan heat (c) and cumulated refrigeration load (d). (Taken from Ambaw, A., Bessemans, N., Gruyters, W.,
Gwanpua, S.G., Schenk, A., De Roeck, A., Delele, M.A., Verboven, P. and Nicolai, B.M., 2016. Analysis of
the spatiotemporal temperature fluctuations inside an apple cool store in response to energy use concerns.
International Journal of Refrigeration, 66, 156–168.)
FIGURE 6.9 Refrigeration heat load of a pomegranate precooling system.
Figure 6.8 depicts a typical refrigeration load obtained from a CFD simulation of apple fruit loaded
cool store room [8], showing the refrigeration load of the cool store under different temperature con-
trol strategies. Figure 6.9 shows the refrigeration heat load of a pomegranate precooling system.
6.4.6 Modeling Produce Quality

During storage of fruit, quality attributes such as taste, firmness, color and flavor are measured
to track quality degradation. Including produce quality in a CFD model of cold handling process
is interesting. Through simulation it is possible to evaluate the effect of storage conditions on the
quality of the final products. In this way it is possible to identify critical points during postharvest
storage and to adjust or improve the decision-making. However, there are few published studies
that incorporate produce quality in a CFD model. Wu, W. and Defraeye [37] incorporated a generic
quality model (Equation 6.9) to the basic CFD model of airflow and heat transfer inside ventilated
cartons for different cold chain scenarios and modeled the quality evolution of individual fruit.
dA
− = γ An (6.9)
dt
where t is the time (s), γ is the rate constant [s−1], n is the order of the reaction. Experimentally
obtained time history data of quality variables such as firmness and peel color can be modeled
using Equation (6.9) to determine the value of γ and n. γ is normally temperature-dependent, which
is often described by an Arrhenius relationship, Equation (6.10).
− Ea
RTp
γ (Tp ) = γ 0 e (6.10)

where γ0 is a constant (d−1), Ea is the activation energy (J mol−1), R is the ideal gas constant (8.314 J mol−1
K−1), T is the absolute temperature (K). The constants γ0 and Ea can be inferred from quality decay data.
6.4.7 Model Validation
CFD models must be validated before being used to assess and compare scenarios or in any decision-
making design steps. The main objective is to prove the accuracy of the developed CFD model so
that it is used with acceptable levels of uncertainty and error. The level of accuracy required from a
CFD analysis depends on the desired use of the results. For qualitative information, like the profile of
the flow field for understanding on a qualitative level the behavior of the flow field, accuracy require-
ments are low. On the other hand, absolute quantities like local magnitude of flow velocity, tempera-
ture and other transport variables require the highest level of accuracy. Validation of model-predicted
absolute quantities require quantification of the prediction errors. This can be accomplished by com-
paring the CFD solution with experimental data or against highly accurate numerical solutions.
In the postharvest, airflow prediction capabilities of CFD models are compared against velocity and
temperature measurement data taken from different spatial locations. Getahun [10,11] measured the local
temperature and airflow velocity on sampling points, as shown in Figure 6.10. In addition to the properties
of the cool store atmosphere, the time history of fruit core temperatures is used to validate predicted cool-
ing rate and produce temperature distribution during precooling of pomegranate fruit (Figure 6.11) [16].
FIGURE 6.10 Schematics of the position of pulp temperature sensors (red circle) and air velocity sensors (blue
circle) in a fully packed reefer container. (a) Showing the positions of the temperature and air velocity sensors as
seen from top, notice that air velocity sensors were situated in the air gap between pallet row 1 and row 2 and tem-
perature sensors on the side of each pallet, (b) position of temperature and air velocity sensors as seen from side of
a pallet in row 1, (c) position of airflow sensors in the overhead space (between pallets and the reefer ceiling), and
(d) a snapshot (door side) of a fully packed reefer wired with the sensors. (Taken from Getahun, S., Ambaw, A.,
Delele, M., Meyer, C.J. and Opara, U.L., 2017. Analysis of airflow and heat transfer inside fruit packed refrigerated
shipping container: Part I–Model development and validation. Journal of Food Engineering, 203, 58–68.)
FIGURE 6.11 Schematics showing the package arrangement and sampling location to monitor the fruit core
temperature during precooling of pomegranate fruit. (Taken from Ambaw, A., Mukama, M. and Opara, U.L.,
2017. Analysis of the effects of package design on the rate and uniformity of cooling of stacked pomegranates:
Numerical and experimental studies. Computers and Electronics in Agriculture, 136, 13–24.)
6.5 MODELING FRESH FRUIT COOLING INSIDE VENTILATED PACKAGING

6.5.1 Apples
The precooling, cold storage and cold transport of apple fruit have been addressed using CFD by
many researchers [38,39]. Also, the application of gaseous fumigants and disease control fungicides
during cold storage of apple fruit has been reported [8,40].
Han et al. [25] studied how air velocities affect the cooling rate and temperature distribution
(inside fruit) during forced-convection cooling of individual apples. This work modeled airflow and
heat transfer simultaneously. The authors showed the relationship between cooling rate and airflow
velocity. It was demonstrated that a reasonable increase in cooling rate is achieved with an increase
in airflow velocity up to 2.5 m s−1; further increase in airflow velocity resulted in a low increase in
the cooling rate. Hence, increasing the air velocity above the optimum value is a waste of energy.
This work also assessed the importance of produce heat generation on the accuracy of the tem-
perature prediction. To do this, the authors compared CFD models with and without produce heat
generation. The simulated temperatures of the two models differed, at maximum, by only 0.033 K.
This showed the negligible effect of the produce heat generation on the precooling process.
Hoang et al. [41] compared two CFD modeling approaches: the porous medium approach and the
solid blocks approach (Figure 6.12). The two models showed a good agreement with the experimen-
tal data of air velocity and produce temperature. The authors reported a better prediction of product
temperature evolution by the solid block approach than the porous medium approach. In the porous
medium model of this work, individual pallets were assumed to be single porous medium blocks.
However, it would be more appropriate to assume individual fruit-loaded boxes as porous media, so
that a pallet would be collection of blocks of porous medium.
FIGURE 6.12 Simulated temperature distribution inside apple fruit loaded cool store room. (a) Porous
medium approach and (b) solid block approach. (Taken from Hoang, H.M., Duret, S., Flick, D. and Laguerre,
O., 2015. Preliminary study of airflow and heat transfer in a cold room filled with apple pallets: Comparison
between two modeling approaches and experimental results. Applied Thermal Engineering, 76, 367–381.)
Ambaw et al. [36] used a porous medium CFD model of airflow and temperature dynamics
inside a fully loaded cool store room of apple fruit. Using this model, the authors analyzed several
load-shifting scenarios by cycling the temperature set point between 1.2°C and 0.6°C following a
day/night regime to reduce energy cost. The air circulation fan was shown to be the major source
of heat load. Discontinuous use of the cooling operation, including 12 h on/12 h off; 10 h on/14 h
off; and 8 h on/16 h off, was investigated. Using the CFD model the authors were able to assess the
produce temperature history under various scenarios to identify the hot and cold zones inside the
cool store (Figure 6.13).
Berry et al. [35] studied the effect of carton designs on airflow resistance, cooling rate, cooling
uniformity and energy usage of precooling of apple fruit. Four carton designs with different vent
hole areas were compared. The authors used an experimentally validated CFD model to investi-
gate the airflow resistance, cooling rate, cooling uniformity and package related energy consump-
tion of the four designs. Experiments were used to quantify box compression strength and study
the effect of vent hole proportion on the mechanical strength of the carton. The study showed an
energy use reduction of 58% with a slight change on the vent-hole design compared to the standard
vent design.
FIGURE 6.13 CFD simulated produce temperature contours in time after fan off and cooling off inside
apple fruit cool store loaded with 240 bins each holding ≈ 380 kg of Jonagold apple fruit and kept under con-
trolled atmosphere (CA) conditions ([O2] = 1% and [CO2] = 2.5%). (Taken from Ambaw, A., Bessemans, N.,
Gruyters, W., Gwanpua, S.G., Schenk, A., De Roeck, A., Delele, M.A., Verboven, P. and Nicolai, B.M., 2016.
Analysis of the spatiotemporal temperature fluctuations inside an apple cool store in response to energy use
concerns. International Journal of Refrigeration, 66, 156–168.)
6.5.2 Citrus
Citrus fruit have a longer post-harvest life than some other fruits and vegetables, although there
are considerable differences between species and cultivar. Citrus is generally sensitive to low tem-
peratures and should normally be stored between 4 and 7°C to avoid chilling injury (CI). Rapidly
cooling freshly harvested citrus fruit to the optimum storage temperature has a beneficial effect
in preserving quality during cold store. Citrus fruits are the highest value fruit crop in terms of
international trade. Improving the cold handling processes of this fruit has a huge economic and
environmental value. The goal of achieving better precooling and cold transport of citrus has been
addressed experimentally and by means of CFD models.
Defraeye et al. [30,31] used a CFD model to investigate the cooling performance of an existing pack-
age design and of two new ones (Supervent and Ecopack). The CFD model helped the researchers obtain
a detailed and closer quantification and visualization of the airflow path to identify airflow short circuits
of alternative designs. Through redesigning of the vent holes, airflow short circuits were removed and
the rate and uniformity of the cooling process enhanced. The energy usage, fruit quality preservation
and throughput of the process were also significantly improved by the new design. In a subsequent study,
Defraeye et al. [42] further investigated alternatives to forced-air precooling. In this work the authors
used an experimentally validated CFD model to investigate warm loading of citrus fruit into refrigerated
containers for cooling during marine transport for logistic and economic savings. The authors under-
lined the importance of box design and proper stacking to reduce airflow short-circuits between pallets.
6.5.3 Pomegranate
The control of temperature, humidity and decay in the cool store room are crucial in postharvest
handling of pomegranate fruit. Pomegranate fruit is susceptible to chilling injury (CI) and should
ideally be stored at temperatures ≥ 7.5°C when stored for more than two months. CI in pomegran-
ates is manifested as brown discoloration and development of necrotic areas (pitting) on the peel,
and browning of the white fleshy mesocarp tissue and segments surrounding the arils. Low humid-
ity levels can dry out the skin readily and the rind becomes dark and hard. Sometimes, prolifera-
tion of decay causing microorganisms becomes the primary limiting factor for long-term storage.
However, maintaining optimal temperature and relative humidity throughout postharvest handling
of pomegranate fruit are prerequisite. Right after harvest, the produce is normally at ambient tem-
perature and contains a large amount of field heat, exasperating the moisture loss and creating favor-
able conditions for microbial growth. To minimize these problems, removal of the field heat is very
important. While research on their health benefits and improved cold storage handling methods for
pomegranates have been reported, little is known about their precooling. An understanding of the
impact of internal packages on the cooling airflow, the cooling rate and the postharvest quality is
important in the design and choice of appropriate packaging.
The cooling characteristics of two different package designs (with or without plastic liner) used
for postharvest handling of pomegranate fruit were analyzed experimentally and by using computa-
tional fluid dynamics (CFD) [16]. Figure 6.14 demonstrated a detailed analysis of the effect of pack-
age design on the pressure distribution (to characterize the airflow resistance) during the precooling
process. The authors demonstrated that liners had a prominent effect on cooling rates and delayed
cooling time by factors of three compared to stacks with no liners.
6.5.4 Strawberries
Cooling of strawberries in vented packages is very important since the shelf life of this product is
relatively short; the fruit is a high value crop and is one of the most perishable fruit. Strawberries are
susceptible to mechanical damage, microbial decay and water loss. If strawberries are left without
cooling for a few hours, their quality will be reduced to an extent that a proportion of marketable
FIGURE 6.14 CFD simulated contour of pressure distribution on vertical plane sectioning the CT1 (a) and
CT2 (b) stacks. The pressure gradient along horizontal line passing through the CT1 (c) and CT2 (d) stacks.
Simulations were at airflow rate of 0.5 L kg−1 s−1. (Taken from Ambaw, A., Mukama, M. and Opara, U.L.,
2017. Analysis of the effects of package design on the rate and uniformity of cooling of stacked pomegranates:
Numerical and experimental studies. Computers and Electronics in Agriculture, 136, 13–24.)
fruit is lost. Hence, strawberries need stricter temperature management than many other fruits.
Through properly executed temperature control, the microbial activity and the respiration process
can be reduced. Wilting and shriveling due to moisture loss can also be reduced.
Ferrua and Singh [43–45] undertook CFD analysis of the FAC process of retail packages of
strawberries. The authors clearly demonstrated that the design of the packaging system significantly
influenced the uniformity of the cooling process. Here, airflow path through clamshells were visu-
alized. The evaluation of the flow path obtained from the CFD simulation helped the researchers
easily identify bypass flows.
Nalbandi et al. [32] used experimental and modeling approaches to design a new package and
cooling system for precooling of strawberries. The CFD model of the airflow through the traditional
system helped the researchers perform an in-depth visualization of the airflow path and distribution.
This led to modification of the original design by positioning a small air restriction plate (ARP)
(Figure 6.15). The innovative design resulted in faster and more uniform cooling compared with the
traditional system.
FIGURE 6.15 Modification of traditional strawberry package by incorporating an air restriction plate (ARP).
(Taken from Nalbandi, H., Seiiedlou, S., Ghasemzadeh, H.R. and Rangbar, F., 2016. Innovative Parallel
Airflow System for forced-air cooling of strawberries. Food and Bioproducts Processing, 100, 440–449.)
6.5.5 Table Grape
Table grapes are packed in vented cardboard boxes with multiple inner packaging materials that
include a plastic liner, SO2 pad, moisture absorber and bunch carry bag. These packaging compo-
nents help maintain the quality of the grapes by providing a shield against mechanical injuries, min-
imizing product moisture loss and retarding microbial decay. Low temperature handling is essential
to control the rate of respiration, moisture loss and microbial growth. Normally, table grapes are
cooled to their optimum storage temperature using forced-air cooling techniques as soon as pos-
sible after harvest and packing. Due to the high water-activity of the produce, moisture loss during
forced-air cooling of this commodity is a severe problem. Understanding the relationship between
the airflow, temperature and relative humidity is important for proper handling. Understanding the
effect of the different components of the packaging material on the rate and uniformity of the cool-
ing process and on the moisture distribution is required for properly managing the cold handling.
Delele et al. [46] used a CFD model of the table grape cooling process to predict the cooling air
velocity, temperature, RH and product moisture loss. The model considered the detailed geometries
of the packaging components (box, liner and pads). The authors validated the CFD model using
experimental measurement data. Using the validated model, the authors performed a detailed analy-
sis of alternative grape packaging and cooling procedures. First, bulk grape bunches were placed
inside the vented box and cooled to storage temperature. As soon as the storage temperature was
attained, the grape bunches were placed inside the carry bag and plastic liner. With this method, the
seven-eighth cooling time was reduced by 78.11%, showing that this approach could help to reduce
the total energy consumption of system.
6.6 EFFECTS OF PACKAGE STACKING PATTERN

Stacking is the placement of several cartons (boxes, packages, containers) on a pallet to facilitate
bulk handling, storage, transport and distribution. Given the tremendous large volume of fresh pro-
duce grown and traded globally, stacking is used in cold stores, refrigerated trucks and shipping
containers as part of the cold chain to deliver fresh produce from the packhouse to markets. With
produce such as citrus, bananas and pomes, ventilated cartons are stacked in layers on a pallet con-
taining up to 20 cartons per pallet. One of the challenges of bulk handling of stacks of ventilated
packages is ensuring adequate and uniform cooling of produce. Historically, individual packages
were designed to protect the produce against mechanical damage (due to compression, impact and
vibration) and to facilitate efficient cooling. However, research has shown that during stacking, the
misalignment of vent holes (openings to facilitate airflow into the carton) is major and common
problem [7,47] that limits the effective delivery of the cooling medium (cold air) to fruit inside the
package.
A survey of ventilated packages used to handle pome fruit in South Africa [48] showed that over
eleven different carton designs were used, categorized as either ‘Display’ or ‘Telescopic’ packages.
Furthermore, four stacking configurations commonly used to palletize the cartons were identified.
Total ventilated area (TVA) per carton varied between 1.92 and 8.81%, with an average of 3.80 and
4.44% for display and telescopic designs, respectively. After stacking cartons on pallet, improper
vent positioning on some packages resulted in non-alignment of ventilation, which in turn altered
airflow patterns during FAC.
Due to the different stacking configurations, these cartons will not have complete ventilation
alignment (Figure 6.16) in both length and width directions. The lack of ventilation alignment
will result in large pressure drops (Vigneault, Markarian, & Goyette, 2004) and poor airflow
rates during forced-air cooling (FAC) and will therefore contribute to ineffective precooling of
fruit (Kader, 2002). A combination of experimental and CFD modeling of airflow, heat and mass
transfer inside ventilated multi-scale packaging of spherical objects showed the stacking pat-
terns have considerable influence on airflow patterns and fruit cooling rates (Delele et al., 2008).
FIGURE 6.16 Pallet stacking arrangements showing (a) 5, (b) 7, (c) 8, and (d) 10 cartons per layer.
Earlier experiment studies on precooling of grape packages by Nelson [49] observed significant
effects of the number and arrangement of packages on the pallet. Due to the low penetration of the
cooling air to the central region of the stack, the larger the pallet, the more difficult to pre-cool.
Recently, experimentally validated CFD-based studies at the Postharvest Technology Research
Laboratory of the South African Research Chair in Postharvest Technology showed that the effect
of the package stacking pattern on fresh produce cooling processes and energy consumption is sig-
nificant. The orientation and stacking pattern of packages over the pallet significantly affected the
airflow pattern, cooling rate and moisture transfer [47,50]. Analysis of the airflow pattern and tem-
perature distribution within and around the stack of grape packages (Figure 6.17) showed that only
some portion of the cooling air appeared to penetrate the stack through the gaps between the boxes
(Figure 6.17a). Furthermore, the hottest region was located near the central region of the stack, and
the region that was located near the entry of the cooling air had a faster cooling rate than the interior
region (Figure 6.17b,c). Correspondingly, the one-half and seven-eighth cooling time at the center of
the stack was 23.19 and 63.92 h, respectively.
v (m/s)
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
(a)
0.0
(b) (c)
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
Temperature (°C)
FIGURE 6.17 Predicted airflow and temperature profile within and around stack of grape packages, after
6 h of cooling from the initial temperature of 21°C (294.15 K); (a) airflow vectors; (b) temperature along the
length of the stack; (c) temperature along the width of the stack.
FIGURE 6.18 Table grape ventilated package stacked on a pallet.
Compared to the corresponding cooling times of two other packages reported by Ngcobo [51],
these cooling times were 193.42 and 204.09% higher. In order to attain the maximum relative at
the hottest spot in the stack, a cooling time of 88 h was needed, which represented an increase of
213% relative to the two packages studied by Ngcobo et al. [51]. Figure 6.18 shows a typical table
grape package stacked over a standard pallet. The airflow patterns and temperature profiles were
highly dependent on the orientation of the stack to the incoming cooling air. For this stacking pat-
tern, forcing the cooling air from direction 2 gave better cooling rate than forcing from direction 1
(Figure 6.19b). These results show the applicability of CFD modeling in optimizing packaging
systems. Overall, these findings demonstrated the considerable influence of package stacking pat-
tern and the applicability of CFD-based modeling as a valuable tool in analyzing the cooling per-
formance of refrigerated cold chains.
6.7 CONCLUSIONS
Cooling of fresh produce such as fruit and vegetables is a critical step in postharvest handling to
maintain quality, extend shelf life and minimize the incidence of food safety hazards. Cooling is
an energy intensive preservation technique, and in the fresh produce industry, the use of cold air is
the most common method due to its applicability to a wide range of commodities. CFD has been
successfully applied to design and investigate the performance of fresh produce cooling process of
commodities such as apples, strawberries, table grapes and pomegranates. In addition to quantita-
tive prediction of cooling rates and energy consumption, CFD-based modeling provides the addi-
tional benefit of visualizing flow patterns, thereby facilitating more comprehensive understanding
of the cooling process, spatially and temporally. Combining CFD modeling and simulation further
allows for scenario analysis (especially at extreme industrial conditions) and process optimization,
which are not feasible to evaluate using experimental approaches alone. However, it should be noted
that despite these advantages, experimental measurements are still required to validate the results
of CFD modeling and simulation of cooling and other bioprocesses.
Stacking ventilated packages during cooling introduces another scale of complexity in CFD
modeling of fresh produce cooling, especially due to misalignment of vent holes and increased
requirement for better meshing of the domain. Nevertheless, considerable progress has been
made during the past decade in applying CFD to characterize the effects of package stacking
on cooling. By including vent hole alignment as an additional objective function, novel package
designs have been proposed and experimentally validated. Increasing use of internal packages
Airflow direction
(a)
0.0 0.4 0.9 1.3 1.7 2.2 2.6 3.0 3.5 3.9
(b)
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0
Temperature (°C)
(c)
60.0 64.0 68.0 72.0 76.0 80.0 84.0 88.0 92.0 98.0
RH (%)
FIGURE 6.19 Predicted airflow, temperature and RH profiles around and inside the package: (a) air velocity
vector when air was forced along direction 1 (left), air velocity vector when air was forced along direction
2 (right); (b) temperature distribution after a cooling time of 4 h from an initial temperature of 21°C; (c) RH
distribution after a cooling time of 4 h from an initial temperature of 21°C.
such as polyliners and carry-bags also presents new opportunities to develop the next genera-
tion of ventilated packages for optimal cooling of fresh produce. This is an important area of
research with considerable potential for industry applications, especially given the widespread
use of these multi-scale packaging in combination with stacking on pallets for handling fresh
produce.
NOMENCLATURE
Ta Air temperature inside cool store (K)
Tw Cool store wall temperature (K)
Qr Heat of respiration (W m−3)
Qv Heat loss due to evaporation (W m−3)
Cpa Heat capacity of air (J kg−1 K−1)
Cpp Heat capacity of the produce (J kg−1 K−1)
hpa Interfacial heat transfer coefficient (W m−2 K−1)
Su Momentum source term (kg m−2 s−2)
T∞ Outside air temperature (K)

p Pressure (Pa)
Tp Produce temperature (K)
ka Thermal conductivity of air (W m−1 K−1)
kp Thermal conductivity of produce (W m−1 K−1)
t Time (s)
Sct Schmidt number
kt Turbulent thermal conductivity (W m−1 K−1)
D t Turbulent diffusion coefficient (m s−1)
U Vector of velocity, ux,y,z (m s−1)
Greek symbols
ρa Density of dry air (kg m−3)
ρp Density of proudce fruit (kg m−3)
μ Dynamic viscosity (kg m−1 s−1)
μ t Turbulent eddy viscosity (kg m−1 s−1)
Sub and super-scripts

a Air
p Solid
t Turbulent
Abbreviations
FAC Forced-air cooling
AFR Airflow resistance
RH Relative humidity (%)
SECT Seven eights cooling time (h)
CI Chilling injury
REFERENCES
1. Lipinski B., Hanson C., Lomax J., Kitinoja L., Waite R., Searchinger T., Reducing food loss and waste,
World Resources Institute Working Paper, June, 2013.
2. Gustavsson J., Cederberg C., Sonesson U., van Otterdijk R. Meybeck A., Global food losses and food
waste: Extent, causes and prevention, FAO, Rome, 2011.
3. Brosnan T., Sun D.-W., Precooling techniques and applications for horticultural products—A review,
International Journal of Refrigeration, 24: 154–170, 2001.
4. Dehghannya J., Ngadi M., Vigneault C., Mathematical modelling procedures for airflow, heat and mass
transfer during forced convection cooling of produce: A review, Food Engineering Reviews, 2: 227–243,
2010.
5. Aswaney M., Forced-air precooling of fruits and vegetables, Air Conditioning and Refrigeration
Journal, 57–62, January–March, 2007.
6. Opara L.U., Zou Q., Sensitivity analysis of a CFD modelling system for airflow and heat transfer of fresh
food packaging: Inlet air flow velocity and inside – package configurations, International Journal of
Food Engineering, 3(5): 1556–3758, 2007. DOI: 10.2202/1556-3758.1263.
7. Opara U.L., From hand holes to vent holes: What’s next in innovative horticultural packaging? South
Africa: Stellenbosch University, Inaugural Lecture, 24, 2011.
8. Ambaw A., Verboven P., Delele M.A., Defraeye T., Tijskens E., Schenk A., Verlinden B.E., Opara
U.L., Nicolai B.M., CFD-based analysis of 1-MCP distribution in commercial cool store rooms: Porous
medium model application, Food and Bioprocess Technology, 7(7): 1903–1916, 2014.
loaded fruit storage room using CFD model, Journal of Food Engineering, 109(1): 87–97, 2012.
10. Getahun S., Ambaw A., Delele M., Meyer C.J., Opara U.L., Analysis of airflow and heat transfer inside
fruit packed refrigerated shipping container: Part I–Model development and validation, Journal of Food
Engineering, 203: 58–68, 2017.
11. Getahun S., Ambaw A., Delele M., Meyer C.J., Opara U.L., Analysis of airflow and heat transfer inside
fruit packed refrigerated shipping container: Part II–Evaluation of apple packaging design and vertical
flow resistance, Journal of Food Engineering, 203: 83–94, 2017.
12. Vigneault C., Thompson J., Wu S., Hui K.P.C., LeBlanc D.L., Transportation of fresh horticultural
produce, In: Postharvest Technologies for Horticultural Crops, Benkeblia N., Ed., Research Signpost,
Kerale, India, 2: 1–24, 2009.
13. Roşca R., Ţenu I., Cârlescu P., Food chilling methods and CFD analysis of a refrigeration cabinet as a
case study, In: Refrigeration, IntechOpen Ltd., London, UK, 2017.
14. Ge Y.T., Tassou S.A., Simulation of the performance of single jet air curtains for vertical refrigerated
display cabinets, Applied Thermal Engineering, 21(2): 201–219, 2001.
15. Chaomuang N., Flick D., Laguerre O., Experimental and numerical investigation of the performance of
retail refrigerated display cabinets, Trends in Food Science & Technology, 70: 95–104, 2017.
16. Ambaw A., Mukama M., Opara U.L., Analysis of the effects of package design on the rate and uni-
formity of cooling of stacked pomegranates: Numerical and experimental studies, Computers and
Electronics in Agriculture, 136: 13–24, 2017.
17. Mukama M., Ambaw A., Berry T.M., Opara U.L., Energy usage of forced air precooling of pomegranate
fruit inside ventilated cartons, Journal of Food Engineering, 215: 126–133, 2017.
18. Gil R., Bojaca C.R., Schrevens E., Suay R., Analysis of air temperature distribution inside a cold store by
means of geostatistical methods, 4th International Conference on Postharvest Unlimited, Leavenworth,
WA. Acta Horticulturae, 945: 29–37, 2012.
19. Akdemir S., Arin S., Spatial variability of ambient temperature, relative humidity and air velocity in a
cold store, Journal of Central European Agriculture, 7(1): 101–110, 2006.
20. Bell S., A beginner’s guide to humidity measurement, National Physical Laboratory, 2, 2012.
21. Roşca R., Ţenu I. Cârlescu P., Food chilling methods and CFD analysis of a refrigeration cabinet as a
case study, In: Refrigeration, IntechOpen Ltd., London, UK, 2017.
22. Evans J.A., Gigiel A.J., Reducing the energy consumption in cold stores, In: The 22nd IIR International
Congress of Refrigeration, 21–26, August, 2007.
23. Evans J., Cold storage of food: Review of available information on energy consumption and energy sav-
ing options, Technical report, Food Refrigeration and Process Engineering Research Centre (FRPERC),
University of Bristol, UK, 2007.
24. Delele M.A., Tijskens E., Atalay Y.T., Ho Q.T., Ramon H., Nicolaï B.M. et al., Combined discrete ele-
ment and CFD modelling of airflow through random stacking of horticultural products in vented boxes,
25. Delele M.A., Ngcobo M.E., Opara U.L., Meyer C.J., Investigating the effects of table grape package
components and stacking on airflow, heat and mass transfer using 3-D CFD modelling, Food and
Bioprocess Technology, 6(9): 2571–2585, 2013.
26. Han J.W., BadÍa-Melis R., Yang X.T., Ruiz-Garcia L., Qian J.P., Zhao C.J., CFD simulation of airflow
and heat transfer during forced-air precooling of apples, Journal of Food Process Engineering, 40(2):
e12390, 2017.
27. Han J.W., Qian J.P., Zhao C.J., Yang X.T., Fan B.L., Mathematical modelling of cooling efficiency
of ventilated packaging: Integral performance evaluation, International Journal of Heat and Mass
Transfer, 111: 386–397, 2017.
28. Han J.W., Zhao C.J., Yang X.T., Qian J.P., Fan B.L., Computational modeling of airflow and heat
transfer in a vented box during cooling: Optimal package design, Applied Thermal Engineering, 91:
883–893, 2015.
29. O’Sullivan J.L., Ferrua M.J., Love R., Verboven P., Nicolaï B., East, A., Forced-air cooling of poly-
lined horticultural produce: Optimal cooling conditions and package design, Postharvest Biology and
Technology, 126: 67–75, 2017.
30. O’Sullivan J.L., Ferrua M.J., Love R., Verboven P., Nicolaï B., East A., Modelling the forced-air cooling
mechanisms and performance of polylined horticultural produce, Postharvest Biology and Technology,
120: 23–35, 2016.
31. Defraeye T., Lambrecht R., Delele M.A., Tsige A.A., Opara U.L., Cronjé P., Verboven P., Nicolai B.,
Forced-convective cooling of citrus fruit: Cooling conditions and energy consumption in relation to
package design, Journal of Food Engineering, 121: 118–127, 2014.
32. Defraeye T., Lambrecht R., Tsige A.A., Delele M.A., Opara U.L., Cronjé P., Verboven P., Nicolai B.,
Forced-convective cooling of citrus fruit: Package design, Journal of Food Engineering, 118(1): 8–18,
2013.
33. Nalbandi H., Seiiedlou S., Ghasemzadeh H.R., Rangbar F., Innovative parallel airflow system for forced-
air cooling of strawberries, Food and Bioproducts Processing, 100: 440–449, 2016.
34. Delele M.A., Ngcobo M.E.K., Getahun S.T., Chen L., Mellmann J., Opara U.L., Studying airflow and
heat transfer characteristics of a horticultural produce packaging system using a 3-D CFD model, Part I:
Model development and validation, Postharvest Biology and Technology, 86: 536–545, 2013.
35. Delele M.A., Ngcobo M.E.K., Getahun S.T., Chen L., Mellmann J., Opara U.L., Studying airflow and
heat transfer characteristics of a horticultural produce packaging system using a 3-D CFD model, Part
II: Effect of package design, Postharvest Biology and Technology, 86: 546–555, 2013.
36. Berry T.M., Fadiji T.S., Defraeye T., Opara U.L., The role of horticultural carton vent hole design on
cooling efficiency and compression strength: A multi-parameter approach, Postharvest Biology and
Technology, 124: 62–74, 2017.
37. Ambaw A., Bessemans N., Gruyters W., Gwanpua S.G., Schenk A., De Roeck A., Delele M.A., Verboven
P., Nicolai B.M., Analysis of the spatiotemporal temperature fluctuations inside an apple cool store in
response to energy use concerns, International Journal of Refrigeration, 66: 156–168, 2016.
38. Wu W., Defraeye T., Identifying heterogeneities in cooling and quality evolution for a pallet of packed
fresh fruit by using virtual cold chains, Applied Thermal Engineering, 133: 407–417, 2017. doi: https://
doi.org/10.1016/j.applthermaleng.2017.11.049.
39. Zou Q., Opara L.U., McKibbin R., A CFD modeling system for airflow and heat transfer in ventilated
packaging for fresh foods I: Initial analysis and development of mathematical models, Journal of Food
Engineering, 77(4): 1037–1047, 2006a.
40. Zou Q., Opara L.U., McKibbin R., A CFD modeling system for airflow and heat transfer in ventilated
packaging for fresh foods II: Computational solution, software development and model testing, Journal
of Food Engineering, 77(4): 1048–1058, 2006b.
loaded fruit storage room using CFD model, Journal of Food Engineering, 109(1): 87–97, 2012.
42. Hoang H.M., Duret S., Flick D., Laguerre O., Preliminary study of airflow and heat transfer in a cold
room filled with apple pallets: Comparison between two modelling approaches and experimental results,
Applied Thermal Engineering, 76: 367–381, 2015.
43. Defraeye T., Verboven P., Opara U.L., Nicolai B., Cronjé P., Feasibility of ambient loading of citrus fruit
into refrigerated containers for cooling during marine transport, Biosystems Engineering, 134: 20–30,
2015.
44. Ferrua M.J., Singh R.P., Modeling the forced-air cooling process of fresh strawberry packages, Part I:
Numerical model, International Journal of Refrigeration, 32(2): 335–348, 2009.
45. Ferrua M.J., Singh R.P., Modeling the forced-air cooling process of fresh strawberry packages, Part II:
Experimental validation of the flow model, International Journal of Refrigeration, 32(2): 349–358, 2009.
46. Ferrua M.J., Singh R.P., Modeling the forced-air cooling process of fresh strawberry packages, Part III:
Experimental validation of the energy model, International Journal of Refrigeration, 32(2): 359–368, 2009.
47. Berry T.M., Delele M.A., Griessel H., Opara U.L., Geometric design characterisation of ventilated
multi-scale packaging used in the South African pome fruit industry, Agricultural Mechanization in
Asia, Africa, and Latin America, 46(3): 1–19, 2015.
48. Delele M.A., Ngcobo M.E., Opara U.L., Meyer C.J., Investigating the effects of table grape package
components and stacking on airflow, heat and mass transfer using 3-D CFD modelling, Food and
49. Nelson K.E., Pre-cooling—Its signifcance to the market quality of table grapes, International Journal
of Refrigeration, 1(4): 207–215, 1978.
50. Delele M.A., Ngcobo M.E., Opara U.L., Pathare P.B., CFD modelling to study the effects of table grape
packaging and stacking on cooling and moisture loss, Acta Horticulturae, 1008: 105–112, 2012.
51. Ngcobo M.E.K., Opara U.L., Thiart G.D., Effects of packaging liners on cooling rate and quality
attributes of table grape (cv. Regal Seedless), Packaging Technology and Science, 25(2): 73–84, 2011.
doi:10.1002/pts.961.
7 Optimization of Ventilation
Ports of Packaging for Fresh
Produce Using CFD
Hiroaki Kitazawa
CONTENTS
7.1 Introduction....................................................................................................................... 149
7.2 Basic Flow for Optimizing Ventilation Efficiency.............................................................. 149
7.3 Validation of Simulation.................................................................................................... 150
7.4 Airflow Visualization of Packaging with Different Sizes of Ventilation Ports..................... 152
7.4.1 Simulation Conditions for CFD.............................................................................. 152
7.4.2 Relationship between Different Ventilation Ports and Airflow Inside the Box........ 152
7.5 Optimizing the Layouts of Ventilation Ports in Packaging................................................. 157
7.5.1 Sequence of Analysis................................................................................................. 157
7.5.2 Simulation Models for CFD................................................................................... 157
7.5.3 Estimation of Ventilation Ability.............................................................................. 158
7.5.3.1 Estimation with Various Port Layouts (Simulation 1)................................ 158
7.5.3.2 Optimized Port Layouts (Simulation 2)......................................................161
7.5.3.3 Estimation of Ventilation Ability with Optimized Port Layouts (Simulation 3)....162
7.6 Conclusions................................................................................................................................. 164
Nomenclature......................................................................................................................... 166
References....................................................................................................................................... 166
7.1 INTRODUCTION
Packaging for fresh produce sometimes requires features that ensure uniform and rapid airflow to
enable cooling and gas exchange. Therefore, ventilation ports are employed. However, manufactur-
ing various types of packages in order to obtain a package with maximum cooling and ventilation
efficiency requires substantial time and effort. Moreover, it is difficult to determine the airflow in
all regions of a closed and narrow space. Consequently, the size and layout of ventilation ports in the
packaging of fresh produce are not based on any rigorous or tested rationale.
Several studies have employed computational fluid dynamics (CFD), whereby the changes in the
airflow and/or temperature in packaging containers have been analyzed for certain fruits such as
apples [1]. This shows that a CFD simulation can be used to optimize the ventilation efficiency of
fresh produce packaging.
In this chapter, two examples of a CFD simulation are introduced for strawberry packaging, one
of which helps in clarifying the effects of air velocity and ventilation port diameter on the airflow
in a carton box. The other example helps in analyzing the effect of different ventilation port layouts
on the airflow distribution in the carton box.
7.2 BASIC FLOW FOR OPTIMIZING VENTILATION EFFICIENCY

The author would like to introduce research on optimizing ventilation efficiency of packaging for
strawberries by Ferrua and Singh [2–6].
149
FIGURE 7.1 Scheme for optimizing ventilation efficiency.
The author believes that the scheme of their research (Figure 7.1) is a good example for the opti-
mization of ventilation efficiency of fresh produce packaging.
First, using CFD analysis, they simulated the airflow with forced cooling inside a carton and
around strawberries packed into plastic trays. They then demonstrated the accuracy of numerical
models in terms of the thermal energy by comparing the measured and calculated values from the
CFD analysis. In addition, they predicted the points to optimize the ventilation efficiency on the
carton by CFD simulation. Based on the simulation results, they proposed an optimized packaging
carton.
7.3 VALIDATION OF SIMULATION
In this section, an example is introduced for validating the CFD simulation using an actual package.
The packaging used for this study comprises two ventilation ports with diameters (d) in the range
of 15–16 mm, located on the two longer sides of the carton box (Figure 7.2).
In this type of packaging, two trays filled with strawberries are placed inside the box such that
the edges of the trays touch the inner walls of the box; accordingly, the interior space is divided
horizontally. Air flows into the box from the inlet port, which is one of the ventilation ports, for
18 min. The temperature and velocity (U) of the air were 21.5°C and 1.2 m s−1, respectively. The
change in the temperature in the upper layer of the box was then measured. The initial temperature
of the upper layer was 15.7°C. A shape similar to that of the interior space of the box was created
using a pre-processor (GAMBIT 2.4.6, ANSYS, Inc., PA, USA). The airflow distribution and the
change in temperature of the interior space model were then determined using a CFD analysis soft-
ware (FLUENT 6.3.26, PA, USA).
The results show that the average temperature difference between the calculated and measured
values is 0.24°C (Figure 7.3), thus demonstrating the high accuracy of the CFD simulation for the
airflow analysis inside the packaging carton, though the airflow distributions were not measured
directly.
Optimization of Ventilation Ports of Packaging for Fresh Produce Using CFD 151
FIGURE 7.2 Outer (top) and inner (bottom) images of a strawberry package for CFD analysis. Inner dimen-
sions: 210 × 290 × 45 mm. Diameter of ventilation port: 15 mm.
FIGURE 7.3 Differences between measured and calculated values of the temperature change in the upper
layer of strawberry packaging (Figure 7.2).
7.4 AIRFLOW VISUALIZATION OF PACKAGING

WITH DIFFERENT SIZES OF VENTILATION PORTS
7.4.1 Simulation Conditions for CFD
In this section, the airflow in the packaging is visualized to determine the points for optimizing the
ventilation ports. Air was made to flow into the carton box through the ventilation port on one side,
and the airflow in the box was reproduced using the CFD analysis software. The size and layout of
the box are the same as those of the box shown in Figure 7.2. The box was divided into two halves
owing to its symmetrical shape, thus reducing the calculation load and calculation time. U was set as
0.3, 0.6, 0.9, 1.2, and 1.5 m s−1. For the physical properties of the air, the values at 15°C were adopted
(viscosity (μ): 1.7894 × 10 −5 Pa s, density (ρ): 1.225 kg m−3).
The Reynolds number (NRe) of the flow field in the box can be obtained using the following equation:
N Re = U d ( µ /ρ)−1 (7.1)

The Reynolds number at the inlet port determined using Equation 7.1 is maximum, i.e., approximately
2150, when U is 1.5 m−1 and d is 21 mm. The airflow through the inlet port is expected to form a turbu-
lent field in a very limited region. However, a laminar flow, which is simpler for analysis, was assumed
because the main objective was to investigate the overall ventilation state in the box. In the simulation, a
steady state calculation was performed, and when the residual of the numerical calculation concerning
the air velocity was less than 10−3, the calculation was assumed to be converged. For calculation, a per-
sonal computer (OPTIPLEX 740, Dell Japan, Kawasaki, Japan, CPU: AMD Athlon 64 X2 4000+ Dual
core processor, RAM: 1.93 GB) was connected to the Calculation System for Science and Technology
of Agriculture, Forestry and Fisheries Research Information Technology Center of MAFF, Japan.
The effects of U and d on the ventilation of the box were analyzed in terms of the air velocity dis-
tribution and flow direction, and the air velocity distributions along Lines 1–9, shown in Figure 7.4,
were calculated and categorised. In addition, a dimensionless number R, which is the ratio of the air
velocity at each position (U′) to the inlet air velocity (U), was calculated using the following equation.
R = U ′/U (7.2)
The volumetric flow rate Q of the inlet air flowing into a packaging box with ventilation can be
obtained using the following equation.
Q = A U , (7.3)

where A is the cross-sectional area of the ventilation port. The relationship with the average R along
Lines 1–9 was calculated to investigate the volumetric flow rate (Q)-based ventilation efficiency
using Equation 7.3. Furthermore, to investigate the lateral diffusion of the airflow through the ven-
tilation port in the packaging box, U′ near the center of Line 1 (102.5 mm from the ventilation
port) was divided into components in the X and Y directions, and the X velocity-to-Y velocity ratio
representing the diffusion in this region was calculated.
7.4.2 Relationship between Different Ventilation Ports and Airflow Inside the Box
With respect to the vertical airflow along the symmetrical boundary in the analysis model, air
flows in through the inlet port and gradually diffuses in the vertical direction in the box under all
conditions (Figure 7.5). There is a region with no airflow near the wall on the inlet side. Figure 7.6
shows the effects of U and d on the overall flow in the entire box. When U = 0.3 m s−1, airflow is
limited near the symmetric axis for all values of d. When U = 0.6 m s−1, the flow slightly strengthens
FIGURE 7.4 Positions for calculating air velocity distribution [7]. Figure shows the upper layer (left)
and lower layer (right) of the inner regions of the box. Lines 1 and 4 are located on the symmetry
boundary.
FIGURE 7.5 Contour and vector of air velocity on the symmetry surface. (From H. Kitazawa, T. Funaki,
M. Nakao, Y. Ohshiro, M. Hiruta, and Y. Ishikawa. Air flow visualization for fresh produce packaging by
CFD analysis. Food Science and Technology Research 18(4): 525–534. 2012.)
FIGURE 7.6 Vector of air velocity distribution within tested packaging. (From H. Kitazawa, T. Funaki,
M. Nakao, Y. Ohshiro, M. Hiruta, and Y. Ishikawa. Air flow visualization for fresh produce packaging by
CFD analysis. Food Science and Technology Research 18(4): 525–534. 2012.) Left: upper layer, Right: bottom
layer.
FIGURE 7.7 Effects of the differences of inlet velocity (U) and diameter of ventilation port (d) on the local
velocity (U′). (From H. Kitazawa, T. Funaki, M. Nakao, Y. Ohshiro, M. Hiruta, and Y. Ishikawa. Air flow
visualization for fresh produce packaging by CFD analysis. Food Science and Technology Research 18(4):
525–534. 2012.)
close to the axis; however, the tendency is similar to that when U = 0.3 m s−1. When U = 0.9, 1.2,
and 1.5 m s−1, air flows in through the ventilation port, reaches near the wall at the outlet port side,
and partially releases. However, the unreleased air generates a flow circulating along the wall and
merges into the flow near the center. This tendency was particularly noticeable with further increase
in d. When U = 0.9 m s−1, a marked swirling flow is generated for a ventilation port with a diameter
of 21 mm or greater in the upper space, and with a diameter of 17 mm or greater in the lower space.
Here, U = 1.5 m−1 and d = 1 mm.
When U = 1.2 m s−1, d is 15 mm or greater, and when U = 1.5 m s−1, the flow swirls in both the
upper and lower spaces of the box for all values of d.
However, regions exist where the flow generation is less, particularly near the end and center of
the walls at the ventilation and outlet port sides in the upper space under all analytical conditions.
The air velocity distribution (U′) along Lines 1–9 increases with the increase in U for most positions
(Figure 7.7); however, the flow decreases with the increase in U at approximately 150 mm from the
inlet port side along Line 2. This region corresponds to the position with less flow generation, as
described above (Figure 7.6). Hence, the flow near the region is believed to be obstructed by the air-
flow along the wall; however, this needs to be further investigated. With respect to the relationship
between R and U, a tendency similar to that in Figure 7.7 was noted in most regions (Figure 7.8);
however, when U = 0.3 m s−1, the reduction in R was marked at a distance of 50 mm or more from
the inlet port side along Lines 1, 4, 5, 7, 8, and 9.
The effect of U’ along these lines is significant depending on U. Therefore, to develop a pack-
aging box, attention should be paid to the ventilation layout to avoid decreasing the R value
along these lines, regardless of whether U is low. With respect to the air velocity distribution, the
relationship between the average values of R and Q was analyzed at various positions in the box
FIGURE 7.8 Effects of the differences of inlet velocity (U) and diameter of ventilation port (d) on the
velocity-to-inlet velocity ratio (R). (From H. Kitazawa, T. Funaki, M. Nakao, Y. Ohshiro, M. Hiruta,
and Y. Ishikawa. Air flow visualization for fresh produce packaging by CFD analysis. Food Science and
Technology Research 18(4): 525–534. 2012.)
using Equation 7.2. The effect of Q on the average R value is significant along Lines 3, 5, and 7–9
(Figure 7.9), followed by those along Lines 1, 2, and 4. Along Lines 1 and 4, the dependence of the
average R value on d was marked, in addition to that on Q. The X velocity-to-Y velocity ratio near
the center of Line 1 ranges from 0.02 to 0.08, indicating that most of the inlet air widely diffuses
in the lateral direction before reaching the outlet port side (Figure 7.10). The lateral diffusion
did not decrease even with the increase in U. In contrast, when d is in the range of 15–21 mm,
FIGURE 7.9 Relationship between volumetric flow rate (Q) and velocity-to-inlet velocity ratio (R) at each posi-
tion. (From H. Kitazawa, T. Funaki, M. Nakao, Y. Ohshiro, M. Hiruta, and Y. Ishikawa. Air flow visualization for
fresh produce packaging by CFD analysis. Food Science and Technology Research 18(4): 525–534. 2012.)
FIGURE 7.10 Relationship between inlet velocity (U) and X velocity-to-Y velocity ratio near the center of Line 1.
(From H. Kitazawa, T. Funaki, M. Nakao, Y. Ohshiro, M. Hiruta, and Y. Ishikawa. Air flow visualization for fresh
produce packaging by CFD analysis. Food Science and Technology Research 18(4): 525–534. 2012.)
the ratio decreases with the increase in U, suggesting that the lateral diffusion of the air decreases
with the increase in U. This may be the reason for the above marked dependence of the average
R value on d along Lines 1 and 4. Along Line 6, the effect of Q on the average R was weaker
than those in the other regions, indicating that the ventilation condition cannot be improved by
increasing Q in this region.
The ventilation was heterogeneous in the analyzed packaging, and it is unlikely that homogenous
ventilation can be achieved by increasing Q through the ventilation port. The elimination of regions
with less flow generation identified using this analysis is a much better alternative to optimize the
ventilation of the box.
7.5 OPTIMIZING THE LAYOUTS OF VENTILATION PORTS IN PACKAGING

In this section, the effect of different ventilation port layouts on the airflow distributions in the
carton box is discussed.
7.5.1 Sequence of Analysis
Figure 7.11 shows the analysis sequence of the present study. The objective of this study was to use
CFD analysis to identify issues related to airflow inside the packaging with the normal port layout,
and subsequently, to propose an optimized layout.
7.5.2 Simulation Models for CFD

The model for this analysis is a carton box, generally used for the single-layer packaging of straw-
berries (Figure 7.12), though the number of ventilation ports is different. A three-dimensional shape
resembling such a box was created using a fluid analysis pre-processor. The number of meshes in
each shape was approximately 130,000. The number of ports and their corresponding diameters and
layouts are discussed in Section 7.5.3. The airflow in the box when air was allowed to flow from one
FIGURE 7.11 Simulation scheme.

FIGURE 7.12 Basic port layouts for CFD analysis.
side was reproduced using CFD analysis software (ANSYS Fluent® 14.5, ANSYS Inc., PA, USA). In
the analysis given in Section 7.4, the range of U was 0.3–1.5 m s−1. In the present study, to examine
the ventilation capability when the airflow is negligible, U was set to 0.2 m s−1. The physical proper-
ties of air at 15°C were used (μ = 1.7894 × 10 −5 Pa s and ρ = 1.225 kg m−3).
From Equation 7.1, if the port diameter is set to a typical length, i.e., 19 mm, NRe for the inlet
portion in this analysis will be approximately 260. Consequently, a simple laminar flow state is
assumed, and the numerical computations were conducted accordingly. The calculation was per-
formed in the simulation analysis, wherein the numerical calculation is said to have converged when
the residual numerical calculations for the inlet velocity is lower than 10 −3. A personal computer
(Vostro 260S, Dell Japan CPU: Intel® CoreTM i5-2400 3.1GHz, 4GB RAM) was connected to the
Calculation System for Science and Technology of Agriculture, Forestry and Fisheries Research
Information Technology Center of MAFF, Japan for the calculation.
7.5.3 Estimation of Ventilation Ability

7.5.3.1 Estimation with Various Port Layouts (Simulation 1)
The diameter was set to 19 mm, which is close to the limiting size, regardless of whether the ports
are placed on the upper or lower surfaces of the box. Two ports were made on each of the longer
sides of the box; a pair of ports served as the inlet and the other pair as the outlet. Furthermore,
considering the various conditions under which packages might actually be stored, the center of
gravity of the box was considered the point of symmetry, and the ports were located at distances of
0, 24, 48, 72, 96, and 120 mm from the center of the longer side such that the ports are symmetrical
with respect to both the inlet and the outlet sides. Accordingly, twenty port layout patterns were
analyzed in this study, as shown in Figure 7.12.
In this section, the port layout for which the airflow in the box is most uniform is deter-
mined. First, the effect of different port layouts on the average velocity distribution along Lines
1–8 (Figure 7.13) was studied. Next, the average and standard deviation along Lines 1–5 in the
upper layer and along Lines 6–8 in the lower layer of the box were calculated. The port layout
for which the variation in the average velocity is negligible and for which the average velocity
is high in the upper and lower layers was selected. The results shown in Section 7.4 suggested
that it was possible to adjust the magnitude of the velocity to a certain extent by changing the
size of the ports. However, it was also suggested that controlling the airflow distribution by
changing either the size of the ports or the inlet velocity of the air through the ports is dif-
ficult. Therefore, it was decided that in the port layout, a smaller variation in the average flow
velocity would be prioritised over the maximum velocity, and this reasoning is followed in the
subsequent analyzes.
Table 7.1 lists the simulation results of the airflow distributions in the interior space for several
port layouts. For a port located at the center of the longer side of the box (Cases 1–5), the variation
in the velocity for both the lower and upper layers was considerable (hereafter, this variation is
referred to as the standard deviation (SD)). Moreover, in the port layouts, although the airflow in
some regions is considerable, regions exist wherein the airflow is limited (Figure 7.14).
FIGURE 7.13 Positions for calculating air velocity distribution. (From H. Kitazawa, and N. Hasegawa.
Improving the layout of ventilation ports in packaging for fresh produce using computational fluid dynamics.
Journal of Food, Agriculture and Environment 12(3–4): 46–50. 2014.)
160
TABLE 7.1
Airflow Distribution in Packaging with Various Port Layouts
Upper Layer (U’ [m s−1]) Lower Layer (U’ [m s−1])
Analysis Pattern Line 1 Line 2 Line 3 Line 4 Line 5 Average SDa Line 6 Line 7 Line 8 Average SD
Case 1 0.020 0.020 0.079 0.012 0.010 0.028 0.026 0.043 0.108 0.024 0.058 0.036
Case 2 0.015 0.034 0.077 0.016 0.004 0.029 0.026 0.025 0.114 0.022 0.054 0.042
Case 3 0.011 0.085 0.079 0.021 0.004 0.040 0.035 0.024 0.119 0.019 0.054 0.046
Case 4 0.009 0.016 0.085 0.010 0.004 0.025 0.030 0.020 0.116 0.016 0.050 0.046
Case 5 0.006 0.012 0.099 0.008 0.003 0.026 0.037 0.011 0.115 0.011 0.046 0.049
Case 6 0.008 0.013 0.013 0.021 0.022 0.015 0.005 0.033 0.043 0.044 0.040 0.005
Case 7 0.003 0.016 0.014 0.031 0.015 0.016 0.009 0.029 0.051 0.027 0.036 0.011
Case 8 0.004 0.018 0.013 0.098 0.010 0.029 0.035 0.025 0.052 0.026 0.034 0.012
Case 9 0.005 0.009 0.021 0.014 0.008 0.011 0.005 0.022 0.054 0.022 0.032 0.015
Case 10 0.006 0.011 0.032 0.011 0.006 0.013 0.010 0.014 0.069 0.014 0.033 0.026
Case 11 0.008 0.032 0.014 0.020 0.020 0.019 0.008 0.039 0.038 0.046 0.041 0.004
Case 12 0.006 0.035 0.015 0.032 0.014 0.020 0.011 0.034 0.047 0.030 0.037 0.007
Case 13 0.007 0.034 0.016 0.109 0.010 0.035 0.038 0.031 0.049 0.030 0.036 0.009
Case 14 0.007 0.014 0.016 0.015 0.007 0.012 0.004 0.027 0.052 0.028 0.036 0.011
Case 15 0.011 0.106 0.011 0.026 0.022 0.035 0.036 0.041 0.036 0.045 0.041 0.004
Case 16 0.009 0.109 0.011 0.034 0.014 0.035 0.038 0.036 0.044 0.034 0.038 0.005
Case 17 0.010 0.110 0.013 0.112 0.010 0.051 0.049 0.032 0.047 0.033 0.037 0.007
Case 18 0.015 0.011 0.008 0.011 0.028 0.015 0.007 0.042 0.033 0.050 0.042 0.007
Case 19 0.010 0.016 0.008 0.015 0.011 0.012 0.003 0.035 0.040 0.036 0.037 0.002
Case 20 0.039 0.010 0.006 0.010 0.039 0.021 0.015 0.053 0.025 0.052 0.043 0.013
Source: H. Kitazawa, and N. Hasegawa. Improving the layout of ventilation ports in packaging for fresh produce using computational fluid dynamics. Journal of Food,
Agriculture and Environment 12(3–4): 46–50. 2014.
a Standard deviation of average value.
Computational Fluid Dynamics in Food Processing
FIGURE 7.14 Simulated airflow in Case 3. (From H. Kitazawa, and N. Hasegawa. Improving the layout
of ventilation ports in packaging for fresh produce using computational fluid dynamics. Journal of Food,
Agriculture and Environment 12(3–4): 46–50. 2014.) Arrows indicate the positions of inlet ports.
Based on Cases 6–10, wherein the ports are located 24 mm from the center of the longer
side, despite the trend towards a higher SD for the average airflow in the upper layer in Case 8
and in the lower layer in Case 10, the SDs for the average velocity are lower in both the lower
and upper layers in other cases. In Cases 11–14, wherein the ports are located 48 mm from the
center of the longer side, both the average velocity and its SD in the lower layer appear to be
largely unaffected by the port layout. However, a trend towards a higher average velocity was
observed in Case 13 in the upper layer, and a trend towards a lower SD is seen in Case 14. In
Cases 15–17, wherein the ports are located 72 mm from the center of the longer side, despite
the higher average velocity in the upper layer, a tendency towards a higher SD for the velocity is
observed. Furthermore, a similar trend in both the velocity and its SD is observed in the lower
layer in Cases 11–14. In Cases 18–19, wherein the ports are located 96 mm from the center of
the longer side, both the average velocity in the upper layer and the SD for the average veloc-
ity in the lower layer are low; this trend is similar to that observed in Cases 11–17. In Case 20,
wherein all the ports are located 120 mm from the center of the longer side, the SD for the
average velocity in the upper layer is considerable.
In summary, the results show that when the ports are located at a distance in the range of
24–120 mm from the center of the longer side of the box and are positioned such that the inlet
and outlet ports do not overlap, the SDs for the velocity decrease in most cases (even though the
SD for the average velocity in the upper layer may be assumed to be greater in some portions of
the layout). Moreover, the velocity distribution in the interior space can be considered to approach
uniformity.
7.5.3.2 Optimized Port Layouts (Simulation 2)

Ten patterns of port layouts (Figure 7.15) were selected to simulate the effect of port layouts on the
velocity distribution in the interior space from the results of the analysis in Simulation 1. From the
results of Simulation 1, several port layouts wherein the airflow in the upper and lower layers is
considerable and mostly uniform were selected, based on which the airflow in the interior of the box
FIGURE 7.15 Optimized port layouts for CFD analysis. (From H. Kitazawa, and N. Hasegawa. Improving
the layout of ventilation ports in packaging for fresh produce using computational fluid dynamics. Journal of
Food, Agriculture and Environment 12(3–4): 46–50. 2014.)
was simulated. Therefore, ports were made in the upper and lower surfaces. Thus, four ports were
made on the inlet and outlet ends of the longer sides of the box. The port diameter and the analysis
method from Simulation 1 were applied to this simulation.
Among Cases 21–30, in Cases 22, 25, 28, and 29, a tendency towards greater average velocity is
observed in the upper layer (Table 7.2).
However, the SD for the average velocity was considerable, and therefore, it was expected that
a bias may occur in favor of the accompanying port layouts in areas with high velocity. The other
instances of high average velocity in the upper layer occurred in Cases 26 and 27. The SD for the
upper layer in Case 27 was the lowest among the cases analyzed in this section. For the lower layer,
the highest average velocity and the lowest SD occurred in Cases 27 and 29. Because the SD for the
average velocity in the upper layer was considerable in Case 29, Case 27 was deemed to present the
best port layout in this section.
7.5.3.3 Estimation of Ventilation Ability with Optimized Port Layouts (Simulation 3)

Ten patterns with different diameters (Figure 7.16) were simulated on the basis of the port layout
of Case 27 from the results of the analysis in Simulation 1. Minimizing the port size is essential
for maintaining the strength of the box. In each port layout discussed in Simulation 1, the air enter-
ing the ports is divided along two directions with respect to the vertical tray placed inside the box.
Consequently, to ensure that the As values located in both the upper and lower portions were half
the size of the port with the 19 mm diameter, the ports were replaced with smaller ports with a
diameter of 13.44 mm.
These ports were then increased by factors of 1.05, 1.10, 1.20, 1.30, and 1.40, correspond-
ing to diameters of 14.10, 14.78, 16.12, 17.46, and 18.80 mm, respectively, and upon placing the
enlarged ports at the positions proposed in Simulation 2, the airflow in the interior was simu-
lated. The analyzes related to different port diameters were conducted separately for the lower
and upper layers.
In Cases 31–35 wherein the effect of different diameters of the ports on the lower layer on the
average velocity in this layer was analyzed, the results show that the average velocity increases
with an increase in the port diameter. This finding is clear from comparing the 1.4 fold increase in
TABLE 7.2
Airflow Distribution in Packaging with Optimized Port Layouts
Upper Layer (U’ [m s−1]) Lower Layer (U’ [m s−1])
Analysis Pattern Line 1 Line 2 Line 3 Line 4 Line 5 Average SD a Line 6 Line 7 Line 8 Average SD
Case 21 0.003 0.006 0.015 0.008 0.005 0.008 0.004 0.025 0.057 0.025 0.036 0.015
Case 22 0.004 0.019 0.010 0.112 0.007 0.031 0.041 0.028 0.056 0.028 0.037 0.013
Case 23 0.004 0.016 0.010 0.025 0.012 0.013 0.007 0.030 0.054 0.030 0.038 0.011
Case 24 0.005 0.010 0.011 0.012 0.026 0.013 0.007 0.043 0.047 0.033 0.041 0.006
Case 25 0.006 0.030 0.011 0.118 0.007 0.034 0.043 0.032 0.049 0.032 0.038 0.008
Case 26 0.004 0.032 0.011 0.028 0.011 0.017 0.011 0.033 0.047 0.033 0.038 0.006
Case 27 0.006 0.026 0.012 0.017 0.027 0.018 0.008 0.046 0.042 0.037 0.042 0.004
Case 28 0.010 0.030 0.008 0.119 0.006 0.035 0.043 0.036 0.046 0.036 0.039 0.005
Case 29 0.024 0.021 0.009 0.124 0.009 0.037 0.044 0.039 0.039 0.047 0.042 0.004
Case 30 0.026 0.007 0.006 0.008 0.013 0.012 0.008 0.040 0.037 0.047 0.041 0.004
Source: H. Kitazawa, and N. Hasegawa. Improving the layout of ventilation ports in packaging for fresh produce using computational fluid dynamics. Journal of Food, Agriculture and Environment
12(3–4): 46–50. 2014.
a Refer to Table 7.1.
Optimization of Ventilation Ports of Packaging for Fresh Produce Using CFD
163
FIGURE 7.16 Various port diameters with optimized port layouts for airflow simulation. (From H.
Kitazawa, and N. Hasegawa. Improving the layout of ventilation ports in packaging for fresh produce
using computational fluid dynamics. Journal of Food, Agriculture and Environment 12(3–4): 46–50.
2014.)
the port diameter with the 1.05 fold increase, wherein the average velocity increased by 1.78 times,
as listed in Table 7.3. In comparison to Case 1 (analyzed in Simulation 1, wherein the average
velocity in this space is highest), when the port diameter is increased by 1.4 times, the average
velocity was found to increase by 1.43 times. In contrast, with the increase in the port size, the SD
for the average velocity in this space tends to decrease. In Cases 36–40 wherein the upper layer
is analyzed, although the average velocity increases with the increase in the port diameter, the
maximum increase was only 1.35 times. Therefore, compared to the lower layer, the increase in
the average velocity in the upper layer is negligible, even with the increase in the port diameter.
Furthermore, when compared to Cases 1–20 in Simulation 1 (Table 7.1), the average velocity in
most cases was found to be lower than the corresponding values in that space. Furthermore, it was
assumed that, in this space, the SD for the average velocity would increase with the increase in the
port size. However, in this space, neither an increase nor a uniformity in the average velocity could
be observed via increasing the port size.
7.6 CONCLUSIONS
When evaluating the ventilation or cooling efficiency of packaging optimized using the simulation
results, the effect of such an improvement on the quality of the packaged fresh produce must be
simultaneously considered. However, only a few studies have clarified the relationship between the
ventilation or cooling efficiency and the quality of fresh produce.
The cooling delay for strawberries after the harvest leads to a decrease in the content of ascor-
bic acid in the fruit, as shown in Figure 7.17. Such a finding would contribute to setting the goal of
optimization of ventilation efficiency in packaging. Nevertheless, the most important part of the
optimization is to maintain the quality of fresh produce.
TABLE 7.3
Effect of Differences in Port Diameter on the Airflow Distribution in Packaging with Improved Port Layouts
d (mm) Upper Layer (U’ [m s−1]) Lower Layer (U’ [m s−1])
Analysis Pattern Upper Layer Lower Layer Line 1 Line 2 Line 3 Line 4 Line 5 Average SDa Line 6 Line 7 Line 8 Average SD
Case 31 13.44 14.10 –b – – – – – – 0.052 0.048 0.041 0.047 0.005
Case 32 13.44 14.78 – – – – – – – 0.057 0.052 0.045 0.051 0.005
Case 33 13.44 16.12 – – – – – – – 0.067 0.062 0.054 0.061 0.005
Case 34 13.44 17.46 – – – – – – – 0.079 0.073 0.064 0.072 0.006
Case 35 13.44 18.80 – – – – – – – 0.091 0.084 0.075 0.083 0.007
Case 36 14.10 13.44 0.007 0.032 0.012 0.019 0.027 0.019 0.009 – – – – –
Case 37 14.78 13.44 0.008 0.035 0.012 0.020 0.027 0.020 0.010 – – – – –
Case 38 16.12 13.44 0.009 0.041 0.012 0.022 0.027 0.022 0.012 – – – – –
Case 39 17.46 13.44 0.010 0.048 0.012 0.025 0.025 0.024 0.013 – – – – –
Case 40 18.80 13.44 0.011 0.053 0.012 0.028 0.024 0.026 0.015 – – – – –
Source: H. Kitazawa, and N. Hasegawa. Improving the layout of ventilation ports in packaging for fresh produce using computational fluid dynamics. Journal of Food, Agriculture
and Environment 12(3–4): 46–50. 2014.
a Refer to Table 7.1.
b Not simulated.
Optimization of Ventilation Ports of Packaging for Fresh Produce Using CFD
165
FIGURE 7.17 Effect of post-harvest cooling delay on ascorbic acid content of strawberry fruit after three
days of storage at 5°C. (From H. Kitazawa, T. Sato, N. Nakamura, and S. Motoki. Effects of post-harvest
cooling delay on weight loss, soluble solid and ascorbic acid contents of strawberry fruit. Journal of Food,
Agriculture and Environment 11(3–4): 372–376. 2013.) Asterisks indicate treatments that differed signifi-
cantly (Dunnett’s test, p ≤ 0.05; n = 4) from the fresh samples, which are sampled immediately after the
harvest.
NOMENCLATURE
A Cross-sectional area of the ventilation port (m2)
d Diameter of the ventilation port (m)
μ Viscosity (Pa s)
ρ Density (kg m−3)
NRe Reynolds number
Q Volumetric flow rate (m3 s−1)
R Velocity-to-inlet velocity ratio
U Inlet velocity (m s−1)
U′ Local velocity (m s−1)
REFERENCES
1. Zou Q., Opara L.U., Sensitivity analysis of a CFD modeling system for airflow and heat transfer of
fresh food packaging: Inlet air flow velocity and inside-package configurations, International Journal
of Food Engineering, 3(5): article No. 16, 2007.
2. Ferrua M.J., Singh R.P., Modeling the forced-air cooling process of fresh strawberry packages, Part I:
Numerical model, International Journal of Refrigeration, 32(2): 335–348, 2009.
3. Ferrua M.J., Singh R.P., Modeling the forced-air cooling process of fresh strawberry packages, Part II:
Experimental validation of the flow model, International Journal of Refrigeration, 32(2): 349–358,
2009.
4. Ferrua M.J., Singh R.P., Modeling the forced-air cooling process of fresh strawberry packages, Part III:
Experimental validation of the energy model, International Journal of Refrigeration, 32(2): 359–368,
2009.
5. Ferrua M.J., Singh R.P., Design guidelines for the forced-air cooling process of strawberries,
International Journal of Refrigeration, 32(8): 1932–1943, 2009.
6. Ferrua M.J., Singh R.P., Improved airflow method and packaging system for forced-air cooling of straw-
berries, International Journal of Refrigeration, 34(4): 1162–1173, 2011.
7. Kitazawa H., Funaki T., Nakao M., Ohshiro Y., Hiruta M., Ishikawa Y., Air flow visualization for fresh
produce packaging by CFD analysis, Food Science and Technology Research, 18(4): 525–534, 2012.
8. Kitazawa H., Hasegawa N., Improving the layout of ventilation ports in packaging for fresh produce
using computational fluid dynamics, Journal of Food, Agriculture and Environment, 12(3–4): 46–50,
2014.
9. Kitazawa H., Sato T., Nakamura N., Motoki S., Effects of post-harvest cooling delay on weight loss, sol-
uble solid and ascorbic acid contents of strawberry fruit, Journal of Food, Agriculture and Environment,
11(3–4): 372–376, 2013.
8 Optimization of Horticultural
Carton Vent Hole Design
by CFD
Umezuruike Linus Opara, Tarl Berry, and Alemayehu Ambaw
CONTENTS
8.1 Introduction....................................................................................................................... 169
8.2 Ventilated Cartons Used in the Horticultural Industry................................................... 170
8.2.1 Types of Ventilated Cartons................................................................................... 170
8.2.2 Internal Packaging..................................................................................................... 171
8.3 Recommended Carton Vent Hole Geometrical Characteristics............................................ 172
8.3.1 Experimental Studies................................................................................................. 172
8.3.2 Modeling Studies....................................................................................................... 172
8.4 CFD Modeling of Vent Hole Design on Cartons................................................................ 173
8.4.1 Multi-Scale Modeling Approach.......................................................................... 173
8.4.2 Modeling Airflow, Heat, and Mass Transfer............................................................175
8.4.3 Quantifying Cooling Performance of Ventilated Cartons......................................... 176
8.5 CFD-Based Optimization of Vent Hole Design for Citrus.................................................... 178
8.6 CFD-Based Optimization of Vent Hole Design for Pome Fruit............................................ 181
8.7 Effects of Internal (Liner) Packaging................................................................................. 184
8.8 Conclusions....................................................................................................................... 188
References....................................................................................................................................... 189
8.1 INTRODUCTION
Ventilated horticultural cartons are the preferred packaging for export of fresh produce around the
globe. Cartons are designed with vent holes along the top, bottom and sides to facilitate cooling
airflow distribution between the fresh produce, towards the maintenance of an unbroken cold chain.
This is critical for the preservation of fruit quality, as respiration rates are directly linked to the tem-
perature at which produce are stored. Vent holes are particularly important during forced-air cool-
ing (FAC), which is used to rapidly reduce field heat and produce temperature to an optimal storage
temperature. FAC is typically applied to palletized carton, packed with warm fruit. The excess
heat in produce after harvest is mainly attributed to field heat (which implies that produce was not
fully cooled before packaging) or as a result of fruit increasing in temperature during postharvest
handling (such as packaging and palletization) under unrefrigerated conditions. FAC is illustrated
in Figure 8.1 and entails the use of fans to create reduced pressure conditions behind one or more
pallet stacks, which drives cold air horizontally through the stacks and accelerates cooling through
convective heat transfer.
Ventilated cartons are manufactured using corrugated paperboard (CP), which is preferred over
other materials due to several unique advantages. For instance, CP is lightweight and has a high
rigidity for its density; can be safely and sustainably manufactured from cellulose-based products;
is completely recyclable and biodegradable; and can dampen mechanical impacts and vibration,
which are sources of damage on fruit. Given the competitive nature of the fresh produce export
169
FIGURE 8.1 Schematic of pallet stacks being cooled using a tunnel-type forced-air cooler.
market, CP cartons are a predominant packaging mode, as they are often more cost-effective than
other materials. However, a challenge of using CP is ensuring cartons maintain their mechanical
strength under cold chain conditions and over extended durations of handling and storage. Factors
that can negatively influence carton integrity include high humidity environments, large compres-
sion forces (due to pallet stacking) and the configuration of vent holes [1]. Packaging designers thus
need to maintain a strict balance between minimizing cost (materials) and packaging performance
(mechanical strength and cooling).
8.2 VENTILATED CARTONS USED IN THE HORTICULTURAL INDUSTRY

8.2.1 Types of Ventilated Cartons
Ventilated horticultural cartons can generally be categorized into two broad design types, namely,
display (also called open top; Figure 8.2a) and telescopic (Figure 8.2b). A survey of the geometric
design of cartons used in the South African horticultural industry showed that display cartons are
used to handle approximately 25% of pome fruit exports and have a dominant market-related func-
tion [2]. Display cartons are designed to be placed directly on the shelf at point of retail, and thus
help to streamline the handling chain. The design thus prioritizes the aesthetic presentation of the
fruit over some of the other design features, such as packing density. In contrast, the telescopic
design emphasizes high-density packing towards bulk handling, whereby fruit are repacked near
the final destination retail market.
The fruit export industry utilizes a wide range of carton designs within the two main carton
design categories. These sub-category designs have been developed to accommodate the many
FIGURE 8.2 Example of the (a) display and (b) telescopic ventilated horticultural carton design.
Optimization of Horticultural Carton Vent Hole Design by CFD 171
FIGURE 8.3 Schematic of stacking patterns used on the standard (a–c) and euro (d) pallets.
diverse requirements of unique fresh produce cold chain and end user. Thus, a carton design might
be used to meet the requirements of different fruit types and sizes, importer demands and various
handling equipment and infrastructure standards. A plethora of different carton designs are thus
currently in use worldwide. Nevertheless, most cartons have a footprint (width and length) that
facilitates stacking on either a 1.2 × 1.0 m pallet (ISO 2; often referred to as standard pallet) or less
commonly on a 1.2 × 0.8 m pallet (ISO 1; Euro). Figure 8.3 shows some of the more common stack-
ing patterns using these pallet bases [2]. No specific carton footprint dominates the export industry;
instead, certain footprints are preferred for certain fruit types. For instance, pome fruit are often
exported using 0.5 × 0.3 m (telescopic) or 0.6 × 0.4 m (display) cartons; citrus is commonly shipped
in 0.4 × 0.3 m (telescopic) cartons; and table grapes are frequently shipped using 0.4 × 0.3 m (dis-
play) or 0.6 × 0.4 m (display) cartons [2].
8.2.2 Internal Packaging
In addition to the many different ventilated carton designs used in postharvest handling, fruit in
each individual carton can also be packed using several different internal packaging types [2].
Internal packaging can perform both functional and aesthetic roles in the postharvest handling of
fruit. As telescopic cartons are typically used for bulk packing (high density), they are often packed
in combination with trays. This provides a consistent and rigid packing structure in the carton
and also absorbs impact and vibration forces, thus minimizing the incidence of mechanical dam-
age. Trays can be made from both pulp paper or polyethylene. The choice of packaging materials
depends on type and economic value of produce, the handling system and potential for mechanical
damage of both the package and produce.
Liner bags are another common internal packaging type and are typically manufactured from
polyethylene (20–60 μm thick), although other material types can be used. Liner bags are placed
around the fruit along the inside of the carton walls and are applied to modify the environmental
conditions around the fruit and to exclude pests and impurities, which compromise phytosanitary
standards in international trade. In many cases, such as in the handling of pomegranates, liner bags
are added mainly to maintain high relative humidity for the purpose of reducing fruit moisture loss.
However, liner bags can also be used to achieve modified atmospheres, so as to influence CO2 and
O2 concentrations towards controlling fruit metabolism and improving overall quality preservation.
Liner bags also form a physical barrier around the fruit, which can reduce exposure to pathogens
and thus decrease the potential for decay.
Other important internal packaging types are carry bags and clam shells, which are used to pack
several smaller retail portions of fruit (such as table grapes) inside an individual carton. Carry bags
and clam shells are thus frequently packed into retail display cartons, enabling the consumer to pur-
chase pre-packaged fruit directly out of the ventilated carton. Carry bags have similar advantages to
liner bags with respect to modifying conditions around the fruit. However, carry bags are generally
packed loose into a carton (fruit are more susceptible to mechanical damage), whereas liner bags
enclose fruit packed on trays, and this packing system reduces the potential for mechanical damage.
Clam shells can also be used to modify the environmental condition around fruit, but unlike carry
bags, they are manufactured from rigid moulded plastic to form transparent miniature containers.
8.3 RECOMMENDED CARTON VENT HOLE GEOMETRICAL CHARACTERISTICS

8.3.1 Experimental Studies
Various attempts to optimize carton vent holes for improved fresh produce quality preservation
during export have been documented for over 50 years [3,5]. Initial investigations were primarily
approached at industry levels based on expert experience; however, with increasing globalization
and the wider distribution of packaged fruit from site of production to distant markets thou-
sands of kilometres away, a lot of research has been pursued at academic and research institu-
tions during the past couple of decades to optimize the design of ventilated cartons for improved
cold chain performance. Early academic work was performed experimentally with the objective
of developing standard guidelines. These investigations typically made use of wind tunnels to
explore the effect of vent hole configuration and total ventilation area percentage (TVA = ratio
between vent area and total carton face area) on airflow distribution, resistance to airflow and
cooling properties.
Past work has shown that cartons with TVAs larger than about 25% do not significantly reduce
pressure loss or improve cooling performance during forced convective cooling [4]. However, vent
holes of this size are impractical in CP cartons, as this would compromise mechanical strength.
Instead, application of the 25% vent hole area recommendation is commonly reserved for packages
made from stronger materials such as plastic. Conversely, a well-recognized compromise between
carton mechanical strength and cooling performance is to use TVAs between 5–6% [5]. In practice,
actual vent holes sizes in industry are often smaller, ranging between 2–6% [2]. Given the lim-
ited vent hole area available, it is critical to maximize vent hole effectiveness by utilizing optimal
vent hole configuration in terms of vent shape, orientation, position on the carton, and number of
openings.
An effective method of improving cooling performance during FAC is to uniformly distribute
cooling airflow through the cartons and between the fruit. This can be achieved by positioning
vent holes evenly across the carton walls [6]. However, a potential drawback of this approach is
that smaller vent holes can reduce cooling performance under room (or static) cooling conditions
(i.e., no fans) [7]. The internal contents of a package can also have a significant effect on how the
airflow is distributed inside the carton. For loose packed produce, it is recommended that vent holes
be placed nearer the center of the carton than at the corners and sides. Apart from the research
reported by Ngcobo et al. [29] on table grapes, which focused on characterizing resistance to air-
flow and cooling performance under multi-scale packaging, relatively little experimental work has
been performed on optimizing vent hole designs for cartons packed with liner bags, carry bags and
clam shells. However, indications for designs used in industry typically suggest vent holes should
be placed along the top and bottom edges [8]. Identifying the exact mechanisms for this general
recommendation can be difficult using experimental methods. The use of modeling techniques is
thus expected to provide important insight into how vent holes can be optimized in cartons packed
with more complex geometries.
8.3.2 Modeling Studies
In recent years, computational resources have increased considerably, making it possible to numeri-
cally investigate the effect of carton design on cooling performance at higher resolutions than are
possible experimentally. One of the earliest examples was by Zou and Opara [9–11], who examined
apple fruit packing on trays in a ventilated carton. Computational fluid dynamics (CFD) has since
become the main numerical approach to evaluating carton vent hole designs.
Past studies have shown significant correlation between improved airflow distribution and vent
hole areas; however, this relationship is only relevant up to about 10% TVA [12,13]. Similarly,
prior work has shown little improvement in cooling rates beyond TVAs of about 7% [14,15]. With
respect to vent hole positioning and number, modeling studies have shown improved cooling per-
formance with openings placed nearer the carton center and with increasing number of vent holes.
Conversely, vent holes placed at the top and bottom of the carton showed comparatively poor cool-
ing performance.
Trays form impermeable layers that compartmentalize the volume of a carton of fruits into dis-
crete regions. Airflow between the respective tray layers is thus restricted to the small openings
between the tray edges and carton faces. In this scenario, vent holes should, therefore, be deliber-
ately configured to deliver airflow evenly to each fruit layer. Typically, this is achieved using elon-
gated oblong vent holes, which have also been recommended for increased compression strength
[16]. Studies investigating vent hole designs that distribute cooling airflow to only a single tray have
noted large cooling heterogeneity and reduced cooling rates [17]. Similar to cartons packed with
loose produce, it is also preferable to position vent holes near the carton center compared with plac-
ing them along the top and bottom edges [18]. With respect to vent hole size in cartons packed using
trays, past studies have recommended TVAs between 4 and 8% [19], which is similar to recommen-
dations for loose packed cartons.
Cartons packed with liner bags have unique vent hole requirements, as the liner bags can obstruct
vent holes while also thermally insulating the produce from cooling air streams. In the pomegran-
ate industry, cartons packed using liner bags in commercial settings frequently position vent holes
along the top or bottom of the carton sides [20]. This enables air to flow along the faces of the car-
ton and liner bag. Using this approach can significantly reduce pressure drop while simultaneously
increasing cooling rates, and is thus recommended. The optimization of vent hole area on cartons
packed with liner bags has been a relatively unexplored area and thus represents an important future
research objective given the increased use of liner bags in commercial postharvest handling of fresh
fruit and other types of produce.
Evidence discussed in this section has shown that the internal geometry of a carton can signifi-
cantly influence airflow distribution inside a carton. Vent hole evaluations, therefore, need to be
designed according to what is being packed—the type of produce and cold chain. Additionally,
recommendations applicable to individual cartons are often not applicable during pallet stacking of
cartons because in the latter, vent holes often become obstructed by adjacent cartons, which results
in ineffective ventilation. The principle of ensuring vent hole alignment during stacking is, thus, a
critical design factor that should be considered at all stages of design and development of ventilated
packaging for handling fresh produce.
8.4 CFD MODELING OF VENT HOLE DESIGN ON CARTONS

8.4.1 Multi-Scale Modeling Approach
The multi-scale approach is a valuable technique used to problem solve systems comprised of
multiple interconnected levels of complexity [21]. The approach is thus useful in agricultural pack-
aging systems [29], which are often impractical to model as a whole. The multi-scale approach
incorporates elements from the small scale, through an averaging procedure, into large scales. In
the case of packaging systems, multi-scale modeling has primarily been used to include the pres-
ence of packaging systems (e.g., bins, pallet stacks) into large refrigerated cold chain structures,
such refrigerated containers and cold rooms. This is typically achieved using a porous media
approach, thus enabling a CFD model to simulate a relatively complex system without having to
model each individual fruit or package [21]. It should be noted that many other small-scale proper-
ties, such as the physiological processes in the fruit, can also be incorporated into a larger model.
However, implementation of multi-scale modeling requires clear understanding of the physics
behind the small-scale features of interest, which are often not yet available due to limited knowl-
edge in this area.
In order to apply the multi-scale approach to fresh produce packaging systems, it is helpful to
use a multi-scale packaging perspective, which describes and outlines the various interconnected
structures making up packaging systems in the cold chain. Figure 8.4 illustrates a multi-scale pack-
aging system as follows: (i) an individual fruit, (ii) a collection of fruit either in bulk (loose) or in/on
internal packaging (e.g., trays, carry bags, clam shells), (iii) a carton packed loose or with packaging
fruit, (iv) a pallet stack made of stacked cartons on a pallet base, and (v) a refrigerated holding area
(e.g., refrigerated container or cold room). Accurate understanding of the relevant heat and mass
transfer processes and impacts on produce cooling at each scale is therefore important in providing
a comprehensive multi-scale modeling of the packaging and cooling system.
Multi-scale modeling is, therefore, very useful in the design and performance evaluation of any
packaging system across different and potential cold chain conditions, especially given huge cost
of experiments and the practical difficulties experimental assessment of such complex situations.
At the individual carton scale, vent holes should be designed to facilitate the alignment of openings
during both horizontal (FAC) and vertical (refrigerated container) cooling [18]. Design recommen-
dations should be interpreted in the context that higher-scale aspects may influence the performance
of lower-scale systems [22]. For example, vent hole alignment, air velocity airflow direction, pal-
let slats blocking bottom vent holes and even non-uniform airflow distribution in the holding area
should all be considered at the initial design and development phases.
Similarly, the multi-scale packaging perspective should also be applied to other packaging design
considerations including mechanical strength and manufacturing cost. The design vent hole on a
FIGURE 8.4 Diagram illustrating a multi-scale packaging perspective, which emphasises the various hier-
archical scales of a packaging system. Superscript numbers indicate level of scale.
carton will not be optimized without taking into account the carton strength and cost of packaging
to business. The use of a multi-scale packaging perspective thus enables a carton to be designed and
evaluated in a much more holistic way, resulting in much more practical outcomes [23].
8.4.2 Modeling Airflow, Heat, and Mass Transfer

During the past couple of decades, computational fluid dynamics modeling has become recognised
as a reliable and robust method to evaluate convective cooling processes in packaged produce.
Most fresh produce package simulations using CFD have observed good agreement between CFD
predictions and experimental results using very similar model setup approach. Chapter 6 in this
book addresses the specific nuances and details of these modeling approaches. However, the section
below briefly outlines a typical numerical setup used to model fresh produce package during forced
convective cooling. This was notably used in the simulations discussed in Sections 8.5 and 8.6.
Figure 8.5 represents a typical computational model setup and is representative of most CFD
evaluations. In the case of apples and oranges, individual fruit are typically modeled discretely as
spheres, which significantly reduces both computational time and cost compared to modeling actual
fruit geometries. Defraeye et al. [24] performed a sensitivity analysis to examine the potential of
using spheres in place of spherical fruit and showed relatively similar airflow predictions between
the two geometries.
The upstream and downstream domains are included to limit the effect of the inlet and outlet
boundaries on flow and cooling conditions. Depending on the research or design objective, inlets
can either be set as pressure or velocity inlets, whereas outlets are typically set as pressure outlets.
Superficial air velocities (velocity at inlet) during FAC are dependent on the type of fan in use and
the number of pallet stacks in a tunnel (Figure 8.1). However, a general estimate is between 1 and
3 L s−1 kg−1 [25,26], although much lower values are also feasible in industry. Refrigerated contain-
ers have considerably lower flow rates that range between 0.02 and 0.06 L s−1 kg−1 [27].
The top, bottom and sides domains are set as symmetry planes. This is a realistic assumption, as
the individual carton or carton stack layer is part of a row of pallet stacks. For instance, it is expected
that the stacked cartons will have similar setup present along both sides (Figure 8.1). Additionally,
the carton stack layer is comprised of many layers and, therefore, it is expected that similar carton
stack layers will be positioned below and above the simulation domain.
FIGURE 8.5 An example of a standard approach to model packaging under cooling conditions. Diagram
depicts the typical geometry setup and boundary conditions used.
In most cases, a hybrid grid (tetrahedral and hexagonal cells) is built within the computational
domain. Grid size is determined from a grid sensitivity analysis, usually in the form of a Richardson
extrapolation method [28]. This can be used to determine the spatial discretization error of various
factors of interest such as mass flow, heat transfer rates, shear stress or product cooling rates, and
the range of acceptable errors should ideally be <5%. The construction of a good quality grid is
critical for the prediction of accurate results; however, the knowledge, experience and time needed
to generate a high-quality grid are very often not mentioned in iterature.
8.4.3 Quantifying Cooling Performance of Ventilated Cartons

Packaging should facilitate rapid and uniform cooling to quickly remove undesirable heat from
fresh produce after harvest or after a break in the cold chain. Quantifying the performance of this
cooling process is particularly relevant to ensuring cold chain efficiency and maintaining produce
quality. The cooling performance influences fresh produce quality and shelf life. It also determines
how long the cooling facility and equipment must operate, which, in turn, has significant impact
on running costs (e.g., power usage). A major benefit of CFD modeling is the high-resolution
data it produces, which can be a valuable tool for visualization and evaluation of package perfor-
mance. However, appropriate evaluation methods and approaches need to be implemented if the
predictions are to be interpreted correctly. Some of the main approaches used to quantify cooling
performance are briefly reviewed below and can be helpful when identifying trade-offs between
design features.
Fruit cooling rates can be quantified using the seven-eighths cooling time (SECT) and half cool-
ing time (HCT), which are defined as the time taken to cool produce to seven-eighths or half of the
temperature difference between the set air temperature and initial fruit temperature, respectively.
The SECT is commonly used in evaluations, as it more closely represents the point at which pre-
cooling would be considered sufficient under commercial settings.
The temperature-time curve can best be expressed using the dimensionless temperature (Y) value
and is calculated using Eq. 8.1. As shown in Figure 8.6, the SECT (Y = 0.125) and HCT (Y = 0.5)
can be determined directly from both the original or dimensionless temperature curves; however, the
dimensionless temperature is helpful when comparing multiple cooling curves with different initial
FIGURE 8.6 Typical volume-averaged temperature-time cooling profile. The numbers on the left indicate
the rescaled dimensionless temperature, whereas numbers on the right show the original actual temperature in
°C. Dashed lines intersect with the cooling curve and indicate the respective HCT and SECT values (x-axis).
temperatures as all data is represented using the same scale (i.e., initial temperature at 1 and a set tem-
perature to 0). Both SECT and HCT can also be calculated from the parameters ( j and C) using Eq. 8.2.
T f − Ta
Y= = je − Ct (8.1)
Ti − Ta
ln(8 j) ln(2 j)
t7 /8 = ; t1/ 2 = (8.2)
C C
where t is time, Ti (°C) is the initial product temperature, Ta (°C) is the set cooling air temperature,
j is the lag factor and C is the cooling coefficient.
Experimentally, Tf is often determined from a thermocouple sensor and thus represents a single
point near the fruit center. A similar value can also be determined from CFD predictions; however,
a volume-averaged temperature across the whole fruit domain is also possible and provides a more
representative description of the overall fresh produce temperature state. The use of different Tf
sampling types and positioning should thus be considered carefully, as this will likely influence the
cooling curve data and thus also affect the values of the SECT and HCT determined.
The convective heat transfer factor (CHTF; W m−2 K−1) at the surface of the produce is a help-
ful parameter to calculate or visualize cooling uniformity. In this context, the CHTF (Eq. 8.3) is a
defined parameter, which relates the convective heat flux normal to the surface (q; W m−2), at the
air-product interface, to the difference between the surface temperature (Tw; °C) and a reference
temperature (Tref ; °C). Typically, the set air temperature (Ta) is selected as the Tref and product tem-
perature is set as a constant value in the simulation, which holds Tw at the initial fruit temperature
across all fruit surfaces. The advantage of this approach is that the CHTF value can be calculated
without any transient simulations and is relatively independent of the heat transfer rate magnitude
as it incorporates the temperature difference (Ti − Ta).
q
CHTF = (8.3)
Tw − Tref

Cooling uniformity (or homogeneity) among the packaged fruit is necessary to ensure uniform
fruit quality preservation throughout the cold chain by avoiding “hot and cold spots” that devi-
ate from the optimal produce temperature. Cooling uniformity can be calculated using multiple
approaches, including using (a) CHTF values at the start of a simulation, (b) SECT/HCT after
cooling is complete, and samples of temperature at a specific moment (e.g., SECT or HCT) or over
time. These respective values are sampled across the packaged fruit spatial domain and expressed
as a measurement of variation. A useful sampling approach is to use the relative standard devia-
tion (RSD), which relates the standard deviation to the mean and is expressed as a percentage. For
example, Figure 8.7 shows the RSD over time using temperature samples across the surface of pack-
aged fruit. An important factor to note is that the final RSD values at SECT are relatively similar,
whereas the values differed significantly halfway through the cooling process.
Cooling rates are usually closely correlated to air velocity; however, the fan-related power needed
to achieve the respective air velocity is dependent on the resistance to airflow of the packaging system.
The resistance to airflow of a system can be determined by relating the pressure difference (Δp) between
the inlet and outlet to the flow rate at the inlet (G; m3 s−1). This relationship can be characterised using the
Darcy–Forchheimer equation, which has two coefficients, namely the Darcy term and the Forchheimer
term (ξ). The Darcy term is generally only significant at very low air velocities (~0.0001 m s−1) and is thus
not applicable under forced-air convective cooling processes [23]. Resistance to airflow can therefore be
characterized using the Forchheimer term as determined in Eq. 8.4.
FIGURE 8.7 An example of the relative standard deviation development over time for different carton
designs. In this case, multiple temperature samples across all fruit surfaces were used to calculate the RSD.
Black crosses indicate the SECT and the moment at which cooling can be ceased.
∆p = ξG 2 (8.4)

The fan-related power consumption needed to cool the packaged fruit to a set temperature is
another important performance parameter, and it influences the cooling operating cost. Fan-related
power consumption is a function of the operational time (e.g., SECT), the airflow rate and resistance
to airflow. Both fan and motor efficiency can generally be ignored because they vary between facili-
ties. However, it should be noted that a packaging system should ideally be optimized with respect
to the equipment being used. The fan-related power (w; watts) needed to maintain a particular flow
rate through a packaging system can be calculated using Eq. 8.5.
w = ∆PG = ξG 3 (8.5)

The total fan-related energy consumption (E; Joules) required to cool fruit to a desired tem-
perature can thus be calculated using Eq. 8.6. This determines the theoretical energy consumption
needed by a facility to precool produce and can be extrapolated to calculate precooling throughput.
E = wt = ∆PGt (8.6)
8.5 CFD-BASED OPTIMIZATION OF VENT HOLE DESIGN FOR CITRUS

Improving cooling performance of citrus carton design in the South African horticultural industry
was one of the earliest examples of vent hole optimization using CFD methods. For citrus export,
fruit is typically packed loose inside a telescopic carton design (0.40 × 0.30 × 0.27 m; Figure 8.8a).
The citrus package was first evaluated with respect to FAC at an individual carton scale and then
in later studies at the pallet scale. More recently, the carton has been evaluated towards improving
refrigerated freight container cooling (vertical airflow).
With respect to individual carton simulations, a CFD model was first developed and then vali-
dated [14]. Predictions showed good agreement with experimental results. A range of vent hole
FIGURE 8.8 Illustration of the (a) original standard vent hole and (b) new vent hole designs investigated
using CFD modeling. (From S. Getahun, A. Ambaw, M.A. Delele, C.J. Meyer, and U.L. Opara. Analysis of
airflow and heat transfer inside fruit packed refrigerated shipping container: Part II – Evaluation of apple
packaging design and vertical flow resistance. Journal of Food Engineering 203: 83–94, 2017.)
sizes, shapes and positions were then evaluated with respect to cooling rate, cooling uniformity and
pressure drop (Figure 8.8b). The study showed that vent hole size was the most significant geometric
design factor in determining fruit cooling rate. However, it was observed that increasing vent hole
size beyond 7% was less effective in improving cooling rates and cooling uniformity. It was also
noted that vent hole shape did not have a significant effect on cooling performance, whereas vent
hole positioning and number were found to be relatively important. Specifically, as with earlier
experimental studies, it was shown that placing many vent holes near the center improved cooling
performance [15]. Curiously, as shown in Figure 8.9, the study noted reduced cooling performance
and energy consumption when vent holes were positioned along the top and bottom of the carton.
Despite these findings, it was noted in industry level that placing vent holes along the top and
bottom of the carton could significantly reduce airflow resistance during refrigerated freight con-
tainer shipping, which utilizes vertical airflow delivery. The next study thus compared the standard,
FIGURE 8.9 Temperature distribution after 2 h of FAC (30 Pa pressure difference between inlet and
outlet) at a temperature of −0.5°C from an initial temperature of 21°C. Carton vent area was equal to 7%
in all cases, using (a) 2 round vents; (b) 9 round vents; (c) top and bottom vents and (d) a rectangular vent
configurations. (From S. Getahun, A. Ambaw, M.A. Delele, C.J. Meyer, and U.L. Opara. Analysis of airflow
and heat transfer inside fruit packed refrigerated shipping container: Part II – Evaluation of apple packaging
design and vertical flow resistance. Journal of Food Engineering 203: 83–94, 2017.)
FIGURE 8.10 The “Supervent” carton design with vent holes along the top and bottom edges, showing
(a) dimensions of individual carton and (b) a layer of cartons.
centrally positioned vent hole design (Figure 8.8a) to the Supervent design, which has openings
placed along the top and bottom edges of the cartons (Figure 8.10). An important priority in these
studies was to improve both horizontal (FAC) and vertical (refrigerated container) cooling perfor-
mance for the purposes of achieving efficient cold sterilization. For citrus fruit, cold sterilization
treatments required sub-optimal temperatures for extended durations, which made fruit susceptible
to chilling injury. A vent hole design that enabled uniform and controlled fruit cooling was therefore
desired to apply treatments as efficiently and rapidly as possible.
A comparison of the “standard” and “Supervent” vent hole designs highlighted the importance of
considering a multi-scale perspective. It was noted that the standard vent hole design resulted in consider-
able vent hole misalignment during stacking. In contrast, the Supervent carton design provided good vent
alignment in stacks of cartons. As a result, the Supervent enabled higher flow rates for the same fan-related
energy consumption as in the standard designs. Heat transfer rates were thus higher and more evenly dis-
tributed across the packaged fruit in the Supervent compared to the standard design (Figure 8.11), and
both FAC and cold sterilization treatments were thus significantly more energy efficient.
Later work by the same research group investigated vertical cooling using the Supervent design
(Figure 8.10). Results indicated a relatively even airflow distribution through the cartons, although
cooling rates were significantly reduced compared to FAC as a result of the slower flow rates and
greater stacking depth. Consequently, air temperature rapidly increased near the inlet before reach-
ing fruit in the remaining cartons, and a large vertical temperature gradient developed across the
stack, which can only be reduced through the use of higher airflow rates. This allows cold air to
reach the final cartons before it is heated by fruit in the upwind carton. The Supervent carton design
has undergone several other assessments at both academic research and industry levels [23], and has
FIGURE 8.11 Contour plot of the convective heat transfer coefficient across citrus fruit surfaces in a layer of
(a) “Standard” and (b) “Supervent” carton designs. Values indicate an adjusted form of the heat transfer rate and
emphasis regions with high or low cooling rates. (From T. Defraeye, R. Lambrecht, M.A. Delele, A. Ambaw,
U.L. Opara, P. Cronjé, P. Verboven, and B.M. Nicolaï. Forced-convective cooling of citrus fruit: Cooling condi-
tions and energy consumption in relation to package design. Journal of Food Engineering 121(1): 118–127, 2014.)
since been adopted for the export of citrus fruit. The Supervent design is thus an excellent example
of how CFD can be used to accelerate package design and testing.
8.6 CFD-BASED OPTIMIZATION OF VENT HOLE DESIGN FOR POME FRUIT

Ventilated cartons used for handling pome fruit (such as apples and pears) are another type of
packaging that have been investigated using CFD methods. In contrast to citrus fruit, pome fruit are
typically packed on trays (internal packaging), which introduces an additional level of complexity to
the packaging system. The vent hole recommendations used for loose packed cartons are therefore
inappropriate in these instances. Recent studies concentrated on a “Mark 4” carton, which is simi-
lar in design to the “Bushel” carton in other citrus industries in Europe and elsewhere. The carton
makes use of a telescopic design (0.50 × 0.33 × 0.27 m; Figure 8.12a and b).
A major challenge facing the use of standard vent hole design in the pome fruit indutry is that air-
flow is only delivered to the middle fruit layers. Cooling rates could therefore be potentially improved
if vent holes were positioned or shaped to deliver airflow to each respective layer. The Altvent and
Multivent hole designs (Figure 8.12c) were therefore proposed as novel approaches to improve FAC
performance [18]. In accordance with the lessons learned in the citrus ventilated carton optimiza-
tion studies, the proposed vent holes were configured to align during stacking. A comparison of the
Altvent and Multivent to the standard vent hole design showed ~50% reduction in fan-related energy
needed to cool the produce to a set temperature. As depicted in Figure 8.13c and d, the Altvent and
Multivent also reduced cooling heterogeneity as a result of improved airflow distribution.
FIGURE 8.12 “Mark 4” carton vent hole design used in the fruit industry showing (a) an individual carton,
(b) alignment of cartons in a stack layer, and (c) dimensions of vent holes. Diagram also shows the (c) vent hole designs
that were investigated using CFD. (From T.M. Berry, T. Defraeye, B.M. Nicolaï and U.L. Opara. Multiparameter
analysis of cooling efficiency of ventilated fruit cartons using CFD: Impact of vent hole design and internal packag-
ing. Food and Bioprocess Technology 9(9): 1481–1493, 2016.) Dashed line indicates positions of the trays.
FIGURE 8.13 Contour plot of the convective heat transfer coefficient across the apple fruit surfaces at
1 L s−1 kg−1. Values indicate an adjusted form of the heat transfer rate and emphasis regions with high or
low cooling rates. (From T.M. Berry, T. Defraeye, B.M. Nicolaï and U.L. Opara. Multiparameter analy-
sis of cooling efficiency of ventilated fruit cartons using CFD: Impact of vent hole design and internal
packaging. Food and Bioprocess Technology 9(9): 1481–1493, 2016; T.M. Berry, T.S. Fadiji, T. Defraeye,
and U.L. Opara. The role of horticultural carton vent hole design on cooling efficiency and compression
strength: A multi-parameter approach. Postharvest Biology and Technology 124: 62–74, 2017.) Figure
shows the (a) Standard Vent, (b) Edgevent, (c) Altvent and (d) Multivent carton designs.
In contrast to the Altvent and Multivent, the Edgevent design (Figure 8.12c) decreased cool-
ing rates compared to the standard design (Figure 8.13b) and this was attributed to airflow being
directed along the top and bottom fruit layers only. The presence of trays thus restricted vertical air-
flow, resulting in larger temperature gradients during FAC. Although the Edgevent design reduced
cooling rates, it also decreased resistance to airflow due to the low fruit-packing density along the
sides of the cartons where fruit do not interlock with each other. All things considered, the use of
Multivent design, which performed slightly better than the Altvent in cartons packed with trays, was
thus recommended for improved FAC efficiency. However, similar to the citrus case, the low airflow
resistance in the Edgevents may be more valuable under stacking conditions where fan power is
often limited.
As a next step in the optimization process, the Multivent design was applied to various carton
shapes. Two new carton shapes were thus proposed and evaluated under both FAC (horizontal
airflow) and refrigerated container (vertical airflow) conditions. Results obtained showed uniform
airflow distribution between all fruit layers and across all fruit surfaces. Table 8.1 shows the relative
standard deviation (RSD) percentage of heat transfer across the fruit surfaces at the start of FAC and
RFC. Overall, the RSD values were all relatively low, with similar values between carton types. The
cooling heterogeneity that was present could be attributed to the air temperature gradient between
the inlet and outlet, as it exchanged heat with the fruit.
TABLE 8.1
Percentage Relative Standard Deviation (Heterogeneity) of Convective Heat
Transfer across the Fruit Surfaces with Respect to Air Velocity During FAC
Air Velocity (m s–1) Mark 4 Hex Tes
0.2 83 96 94
0.4 72 81 79
0.6 69 78 75
FIGURE 8.14 Distribution of convective heat transfer coefficient values over surfaces of apple fruit for the
three packaging designs at 0.4 m s−1 (velocity at inlet) during FAC conditions.
Figure 8.14 shows the heat transfer rates across the packaged fruit surfaces, with relatively uni-
form heat transfer rates between the different pallet stacks. Similarly, with respect to cooling rate,
the Hex design facilitated relatively similar cooling rates compared to the Mark 4 carton design.
Conversely, the Tes design improved cooling energy efficiency compared to the Mark 4 carton. A
detailed analysis of the simulated results (Figure 8.14) showed that this improvement can be attrib-
uted to the Tes carton design having a low airflow resistance, which consequently improved the
overall cooling efficiency.
With respect to refrigerated container cooling, CFD simulations showed significant differences
between the different carton designs (Figure 8.15). Generally high cooling heterogeneities were
observed in all three carton shapes (Table 8.2). As with the FAC cases, this was primarily attrib-
uted to a gradient in air temperature, which increased significantly in temperature (Figure 8.16)
between the inlet and outlet. Cooling heterogeneity thus decreased with increasing airflow rates.
The Mark 4 carton shape facilitated the most efficient FAC; however, this was credited to the use
of a centrally positioned vent hole, which forced airflow to travel from the carton side walls (gap
between trays and carton) to the carton’s mid-point (Figure 8.14). For the Hex type, this distance
was significantly longer than the Mark 4, whereas the Tes type had a squarer vent hole that was
more easily obstructed by nearby fruit. The high resolution of these CFD simulations thus showed
that improved vent hole designs could circumvent these factors and thus considerably improve both
cooling uniformity and cooling efficiency.
The results from this CFD case study demonstrate the considerable advantages of applying
CFD methods to ventilated carton design. CFD predicts details flow patterns at high resolu-
tion, which offers a much higher level of insight that would have been possible from experi-
ments alone. As a result of the detailed predictions from the CFD simulations, it is possible to
investigate the underlying causes of observed thermodynamic phenomena, thereby allowing
the design engineer to make informed decisions and more rapidly prototype optimal packaging
designs.
FIGURE 8.15 Distribution of convective heat transfer coefficient values over surfaces for the three packag-
ing designs at 0.04 m s−1 (velocity at inlet) during refrigerated container cooling conditions.
TABLE 8.2
Percentage Relative Standard Deviation (Heterogeneity)
of Convective Heat Transfer across the Fruit Surfaces with
Respect to Air Velocity During Refrigerated Container Cooling
Air Velocity (m s−1) STD Hex Tes
0.02 377 343 365
0.04 273 254 267
0.06 225 214 224
8.7 EFFECTS OF INTERNAL (LINER) PACKAGING

Liner bags covering fresh produce inside cartons perform an important role in preserving quality
over extended cold storage and transport periods. However, the use of liners bags also presents
several challenges with regard to insulating the produce from cooling air streams and blocking vent
hole. Airflow is thus constricted between the vent holes and liner bags, resulting in large pressure
drops, reduced cooling rates and higher refrigeration energy costs.
Figure 8.17a shows a computed tomography (CT) image of plastic spheres packed inside carry bags
and a liner bag. The image illustrates very different geometries between the two setups and suggests a
much higher porosity in the case of carry bags versus liner bags. A similar observation was reported
by Ngcobo et al. [29], who experimentally demonstrated considerably lower pressure drops in cartons
packed with table grapes in carry bags compared to liner bags. Currently, little or no published work
FIGURE 8.16 Example of temperature gradient in a column of cartons under refrigerated freight cooling
conditions. Image illustrates how air rapidly increases in temperature before reaching fruit higher up.
FIGURE 8.17 CT image of a (a) box packed with spheres in carry bags and (b) a carton packed with plastic
spheres inside a liner bag.
is available that explores modeling of cartons packed with carry bags. A potentially promising strat-
egy is to treat the individual carry bags as large fruit-air bodies. Vent hole recommendations that are
applicable to fruit packed loosely may also, therefore, be applicable to cartons packed with carry bags.
However, it should be noted that vent holes could be easily blocked by the loose bundles of fruit.
The geometry of a liner bag inside a carton is somewhat more challenging to model than loose
fruit, in large part because all the mass (air) flow takes place in the small air domain between the
carton wall and the liner bag. For display cartons, this region is quite narrow along the bottom
and sides, whereas the gap above the liner bag top is substantially larger [30]. On the other hand,
telescopic cartons (Figure 8.17b) have a consistently narrow gap (<5 mm) across all faces of the
liner bag. In addition to accurately capturing the liner bag’s geometry, the narrow spacing will also
require a relatively high-quality grid in these narrow regions if accurate CFD-based predictions are
to be generated. The computational cost is thus expected to increase significantly when modeling
liner bags, which is evident from the limited amount of research reported on this topic.
Figure 8.18 shows the temperature profiles at the start of FAC in telescopic cartons packed with
apples in liner bags and on trays. In this case, the four vent hole configurations in Figure 8.12 were
investigated. Figure 8.18 contrasts the different cooling processes using each internal packaging
type. Specifically, cartons packed with trays enable the fruit surfaces to be cooled convectively,
resulting in relatively rapid cooling rates. However, when using liner bags, heat transport must take
place conductively from the fruit to the inner air domain and then finally to the liner bag, where
it can be convectively transferred to the outer airflow. Fruit packed in liner bags thus took over
seven times longer to cool (Table 8.3), with relatively small differences in cooling uniformity (<3%)
between the four configurations.
With respect resistance to airflow, liner bags increased pressure drops by about 2.8 times.
Cartons with vent holes along the top and bottom carton (Edgevent) showed over 50% reduction
in pressure drop compared to other carton designs packed with liner bags, which resulted in lower
power consumption. These findings indicate that improved cooling performance can be achieved
FIGURE 8.18 Temperature contour planes of cartons packed with trays and liner bags through the air and
fruit domains inside the carton designs, when packed with trays and liner bags, at the start of FAC.
TABLE 8.3
Pressure Drop and Cooling Rate for the Four Vent Hole
Designs When Packed with Trays or Liner Bags at 1 L s−1 kg−1
Seven-Eighths Cooling
Vent Hole Design Pressure Drop (Pa) Time (hours)
Standard vent (Trays) 177 2.83
Standard vent (Liner) 772 27.46
Altvent (Trays) 149 2.67
Altvent (Liner) 594 29.05
Edgevent (Trays) 101 5.05
Edgevent (Liner) 268 28.70
Multivent (Trays) 114 2.59
Multivent (Liner) 424 27.84
by placing vent holes along the top and bottom of carton walls. Similar observations were made in
the case of display cartons, where the industry has extensively implemented top/bottom ventilation.
Studies examining display cartons packed with pomegranates (Figure 8.19), table grapes (Figure 8.20)
and kiwifruit (Figure 8.21) in liner bags have also shown good cooling performance using vent holes
along the cartons top edge [30–33]. This performance was attributed to the vents distributing air-
flow through the large gap above the liner bag, which enables low airflow resistance between indi-
vidual cartons in a stack, and thereby improves stack cooling uniformity and cooling of fruit inside
individual cartons. Designing cartons that ensure vent hole alignment is, therefore, critical, espe-
cially where the use of liners is necessary to meet market requirements.
FIGURE 8.19 Temperature profile along fruit surfaces and streamlines showing airflow distribution through
a stack of cartons packed with pomegranates in liner bags. (From A. Ambaw, M. Mukama, and U.L. Opara.
Analysis of the effects of package design on the rate and uniformity of cooling of stacked pomegranates:
Numerical and experimental studies. Computers and Electronics in Agriculture 136: 13–24, 2017.)
FIGURE 8.20 Temperature profile along the (a) length and (b) width of a pallet stack made up of table grapes
in liner bags. (From M.A. Delele, M.E.K. Ngcobo, U.L. Opara, and C.J. Meyer. Investigating the effects of
table grape package components and stacking on airflow, heat and mass transfer using 3-D CFD modeling.
Food and Bioprocess Technology 6(9): 2571–2585, 2012.)
FIGURE 8.21 Air velocity through stacked cartons packed with kiwi fruit in liner bag. Images compare
(a) an original and (b) modified vent hole configuration. (From J.L. O’Sullivan, M.J. Ferrua, R. Love, P.
Verboven, B.M. Nicolaï, and A.R. East. Forced-air cooling of polylined horticultural produce: Optimal cool-
ing conditions and package design. Postharvest Biology and Technology 126: 67–75, 2017.)
The effect of liner bags and carry bags on table grape cooling and moisture loss under cold chain
conditions has also been investigated using CFD [34]. In this case, a porous media approach was
implemented to describe the fruit, liner and carry bag domains. Predictions using this approach
showed good agreement with experimental results, which further demonstrated the capability of
CFD modeling as a valuable tool to investigate cooling performance of ventilated packaging used
in postharvest handling and storage of horticultural commodities.
8.8 CONCLUSIONS
Ventilated packaging has become the mainstay for postharvest handling, storage and distribution
of fresh horticultural commodities. Over the years, the geometric configuration and purpose of
vents has evolved from “hand-holes” to facilitate the lifting of individual cartons to more sophisti-
cated design of vents, resulting in multiple combinations of vent size, shape, number and position
on the carton.
Earlier studies based on experimental investigations focused on assessing the resistance to air-
flow and cooling performance of individual carton designs and thus offered limited scope towards
optimization of vent to maintain the cold chain and produce quality. With rapid advancements in
computational fluid dynamics (CFD) and associated lower computing costs, CFD applications in a
wide range of industries followed, including horticultural packaging. This provided the opportunity
for simultaneous heat and mass transfer modeling, while the visualization of fluid flow patterns
enabled better understanding and appreciation of the simulation results. With increasing public
interest in sustainability goals and resource conservation, recent applications of CFD in packaging
design and performance assessment have incorporated the predication of energy amount used in
cooling fruit inside ventilated cartons.
With the plethora of cartons designs emerging to meet market demand to reduce produce weight
loss and exclude pests at export destination, the use of multi-scale packaging has continued to grow.
In particular, the use of internal packaging such as polyliners, individual fruit wrapping, punnets
and carry bags provide additional challenges in the application of CFD to optimize ventilated car-
tons. It is expected that future advancements in this area will progress quickly in the next decade
with improvements in meshing techniques and computational speed.
REFERENCES
1. Frank B., Corrugated box compression—A literature survey, Packaging Technology and Science, 27(2):
105–128, 2014.
2. Berry T.M., Delele M.A., Griessel H., Opara U.L., Geometric design characterisation of ventilated
multi-scale packaging used in the South African pome fruit industry, Agricultural Mechanization in
Asia, Africa, and Latin America, 46(3): 1–19, 2015.
3. Maltenfort G.G., Performance and Evaluation of Shipping Containers: A Collection of Seminal Articles
and Technical Papers in the Corrugated Container Industry, Jelmar Publishing Company, Plainview,
New York, USA, 1989.
4. Vigneault C., Goyette B., Design of plastic container opening to optimize forced-air precooling of fruits
and vegetables, Applied Engineering in Agriculture, 18(1): 73–76, 2002.
5. Pathare P.B., Opara U.L., Vigneault C., Delele M.A., Al-Said F.A.J., Design of packaging vents for cool-
ing fresh horticultural produce, Food and Bioprocess Technology, 5(6): 2031–2045, 2012.
6. de Castro L.R., Vigneault C., Cortez L.A.B., Cooling performance of horticultural produce in contain-
ers with peripheral openings, Postharvest Biology and Technology, 38(3): 254–261, 2005.
7. Mitchell F.G., Parsons R.A., Mayer G., Cooling trials with plastic tray pack nectarines in various con-
tainers, California Agriculture, 25(9): 13–15, 1971.
8. Anderson B.A., Sarkar A., Thompson J.F., Singh R.P., Commercial-scale forced-air cooling of packaged
strawberries, Transactions of the ASAE, 47(1): 183–190, 2004.
9. Opara U.L., Zou Q., Sensitivity analysis of a CFD modeling system for airflow and heat transfer of
fresh food packaging: Inlet air flow velocity and inside-package configurations, International Journal
of Food Engineering, 3(5): 1–13, 2007.
10. Zou Q., Opara U.L., McKibbin R., A CFD modeling system for airflow and heat transfer in ventilated
packaging for fresh foods: I. Initial analysis and development of mathematical models, Journal of Food
Engineering, 77(4): 1037–1047, 2006.
11. Zou Q., Opara U.L., McKibbin R., A CFD modeling system for airflow and heat transfer in venti-
lated packaging for fresh foods: II. Computational solution, software development, and model testing,
12. Tutar M., Erdogdu F., Toka B., Computational modeling of airflow patterns and heat transfer prediction
through stacked layers’ products in a vented box during cooling, International Journal of Refrigeration,
32(2): 295–306, 2009.
13. Dehghannya J., Ngadi M., Vigneault C., Simultaneous aerodynamic and thermal analysis during cool-
ing of stacked spheres inside ventilated packages, Chemical Engineering Technology, 31(11): 1651–1659,
2008.
14. Delele M.A., Ngcobo M.E.K., Getahun S., Chen L., Mellmann J., Opara U.L., Studying airflow and
heat transfer characteristics of a horticultural produce packaging system using a 3-D CFD model. Part I:
Model development and validation, Postharvest Biology and Technology, 86: 536–545, 2013.
15. Delele M.A., Ngcobo M.E.K., Getahun S., Chen L., Mellmann J., Opara U.L., Studying airflow and heat
transfer characteristics of a horticultural produce packaging system using a 3-D CFD model. Part II:
Effect of package design, Postharvest Biology and Technology, 86: 546–555, 2013.
16. Han J., Park J.M., Finite element analysis of vent/hand hole designs for corrugated fibreboard boxes,
Packaging Technology and Science, 20(1): 39–47, 2007.
17. Han J.-W., Zhao C.-J., Yang X.-T., Qian J.-P., Fan B.-L., Computational modeling of airflow and heat
transfer in a vented box during cooling: Optimal package design, Applied Thermal Engineering, 91:
883–893, 2015.
18. Berry T.M., Defraeye T., Nicolaï B.M., Opara U.L., Multiparameter analysis of cooling efficiency
of ventilated fruit cartons using CFD: Impact of vent hole design and internal packaging, Food and
19. Berry T.M., Fadiji T.S., Defraeye T., Opara U.L., The role of horticultural carton vent hole design on
cooling efficiency and compression strength: A multi-parameter approach, Postharvest Biology and
Technology, 124: 62–74, 2017.
20. Mukama M., Ambaw A., Berry T.M., Opara U.L., Energy usage of forced air precooling of pomegranate
fruit inside ventilated cartons, Journal of Food Engineering, 215: 126–133, 2017.
21. Ho Q.T., Carmeliet J., Datta A.K., Defraeye T., Delele M.A., Herremans E., Opara U.L., Ramon H.,
Tijskens E., van der Sman R.G.M., Van Liedekerke P., Verboven P., Nicolaï B.M., Multiscale modeling
in food engineering, Journal of Food Engineering, 114(3): 279–291, 2013.
22. Defraeye T., Lambrecht R., Ambaw A., Delele M.A., Opara U.L., Cronjé P., Verboven P., Nicolaï B.M.,
Forced-convective cooling of citrus fruit: Package design, Journal of Food Engineering, 118(1): 8–18,
2013.
23. Defraeye T., Cronjé P., Berry T.M., Opara U.L., East A.R., Hertog M.L.A.T.M., Verboven P., Nicolaï
B.M., Towards integrated performance evaluation of future packaging for fresh produce in the cold
chain, Trends in Food Science & Technology, 44(2): 201–225, 2015.
24. Defraeye T., Herremans E., Verboven P., Carmeliet J., Nicolaï B.M., Convective heat and mass exchange
at surfaces of horticultural products: A microscale {CFD} modeling approach, Agricultural and Forest
Meteorology, 162–163: 71–84, 2012.
25. Brosnan T., Sun D.-W., Precooling techniques and applications for horticultural products—A review,
International Journal of Refrigeration, 24(2): 154–170, 2001.
26. Thompson J.F., Mitchell F.G., Rumsey T.R., Kasmire R.F., Crisosto C.H., Commercial Cooling of Fruits,
Vegetables, and Flowers, University of California Department of Agriculture and Natural Resources,
Oakland, CA, USA, 2008.
27. Defraeye T., Cronjé P., Verboven P., Opara U.L., Nicolaï B.M., Exploring ambient loading of citrus
fruit into reefer containers for cooling during marine transport using computational fluid dynamics,
Postharvest Biology and Technology, 108: 91–101, 2015.
28. Celik I.B., Ghia U., Roache P.J., Freitas C.F., Coleman H., Raad P.E., Procedure for estimation and
reporting of uncertainty due to discretization in CFD applications, Journal of Fluids Engineering,
130(7): 1–4, 2008.
29. Ngcobo M.E.K., Delele M.A., Opara U.L., Zietsman C.J., Meyer C.J., Resistance to airflow and cooling
patterns through multi-scale packaging of table grapes, International Journal of Refrigeration, 35(2):
445–452, 2012.
30. Ambaw A., Mukama M., Opara U.L., Analysis of the effects of package design on the rate and uni-
formity of cooling of stacked pomegranates: Numerical and experimental studies, Computers and
Electronics in Agriculture, 136: 13–24, 2017.
31. O’Sullivan J.L., Ferrua M.J., Love R., Verboven P., Nicolaï B.M., East A.R., Modeling the forced-
air cooling mechanisms and performance of polylined horticultural produce, Postharvest Biology and
Technology, 120: 23–35, 2016.
32. O’Sullivan J.L., Ferrua M.J., Love R., Verboven P., Nicolaï B.M., East A.R., Forced-air cooling of poly-
lined horticultural produce: Optimal cooling conditions and package design, Postharvest Biology and
Technology, 126: 67–75, 2017.
33. Delele M.A., Ngcobo M.E.K., Opara U.L., Pathare P.B., CFD modeling to study the effects of table
grape packaging and stacking on fruit cooling and moisture loss, Acta Horticulturae, 1008: 105–112,
2013.
34. Delele M.A., Ngcobo M.E.K., Opara U.L., Meyer C.J., Investigating the effects of table grape pack-
age components and stacking on airflow, heat and mass transfer using 3-D CFD modeling, Food and
35. Getahun S., Ambaw A., Delele M.A., Meyer C.J., Opara U.L., Analysis of airflow and heat transfer
inside fruit packed refrigerated shipping container: Part II – Evaluation of apple packaging design and
vertical flow resistance, Journal of Food Engineering, 203: 83–94, 2017.
36. Defraeye T., Lambrecht R., Delele M.A., Ambaw A., Opara U.L., Cronjé P., Verboven P., Nicolaï B.M.,
Forced-convective cooling of citrus fruit: Cooling conditions and energy consumption in relation to
package design, Journal of Food Engineering, 121(1): 118–127, 2014.
Section II
CFD Applications in Thermal
Processing and Heat Exchangers
9 Three-Dimensional CFD
Modeling of Continuous
Industrial Baking Process
Weibiao Zhou and Nantawan Therdthai
CONTENTS
9.1 Introduction....................................................................................................................................194
9.2 Baking Process......................................................................................................................... 194
9.2.1 Heat and Mass Transfer Mechanisms........................................................................ 196
9.2.1.1 Heat Transfer Mechanism........................................................................... 196
9.2.1.2 Mass Transfer Mechanism.......................................................................... 197
9.2.2 Changes During Baking Process............................................................................... 198
9.2.2.1 Volume Expansion........................................................................................ 198
9.2.2.2 Solidification............................................................................................... 198
9.2.2.3 Color Development......................................................................................... 199
9.2.2.4 Flavor Development......................................................................................200
9.3 CFD Modeling.......................................................................................................................200
9.3.1 Continuous Industrial Baking Oven..........................................................................200
9.3.2 Oven Monitoring System in the Continuous Industrial Baking Oven.......................200
9.3.3 Thermal Properties of Dough During Baking............................................................200
9.3.3.1 Modeling......................................................................................................... 203
9.3.3.2 Model Assumption......................................................................................204
9.3.3.3 Volume Condition Settings.......................................................................... 205
9.3.3.4 Boundary and Initial Condition Settings........................................................205
9.3.3.5 Solving the Model.......................................................................................207
9.3.3.6 Model Validation............................................................................................ 207
9.4 Applications of CFD Model........................................................................................................... 209
9.4.1 Predicting Temperature Profiles for Quality Estimation of Baked Bread.................209
9.4.2 Simulation of Oven Operation Under Increasing Oven Load.....................................211
9.4.3 Simulation of Baking Index....................................................................................... 212
9.4.4 Design of Operating Condition to Achieve the Optimum Tin Temperature
Profiles.............................................................................................................................215
9.4.5 Design of Process Controller for a Continuous Bread Baking Process.................... 216
9.4.6 Design of Indirect Gas-Fired Radiant Burner for a Tunnel Baking Oven................... 217
9.4.7 Design of Operational Condition for Radiant Tube of a Burner...................................219
9.4.8 Design of an Air Impingement Baking Oven........................................................... 220
9.5 Conclusions..................................................................................................................................221
Nomenclature.................................................................................................................................. 221
References....................................................................................................................................... 222
193
9.1 INTRODUCTION
Baking is the process by which a raw dough piece is transformed into a light, porous, readily digest-
ible and flavorful product under the influence of heat. With the requisite quality attributes, the pro-
duction of bakery products such as bread presumes a carefully controlled baking process. Factors
that have vital influence on the final product quality include the rate and amount of heat applica-
tion, the humidity level in the baking chamber and baking time. During baking, the most apparent
interactions are volume expansion, crust formation, inactivation of yeast and enzymatic activities,
protein coagulation, partial gelatinization of starch in dough [1] and moisture loss.
Heat distribution in an oven depends on a number of parameters, including heat source, airflow
pattern, flow rate, oven load, baking time, etc. To manipulate the oven conditions at the optimum
temperature profiles, the relationship among these important parameters needs to be established.
Computational fluid dynamics (CFD) modeling may be the only method to effectively solve such
a complicated problem. This chapter presents the results of three-dimensional CFD models. The
effect of heat source, flow rate through convection fans, baking time and oven load on baking tem-
perature profiles will be investigated. The results of simulation can be used for manipulating an
oven control to maintain its baking temperature profiles at the optimum conditions.
9.2 BAKING PROCESS
During baking, heat is applied and the final leavening occurs before yeast is destroyed at around
60°C. As a result, a rapid expansion of water vapor and carbon dioxide produces oven spring. The top
crust is pushed up. The optimal temperature of baking depends on the size and richness of the prod-
uct. For example, small products should be baked at 204 to 238°C, whereas richer products should
be baked at lower temperatures. Baking is the key step in bread making. It can be divided into three
stages [2]. The first stage starts at around 204°C and is composed of one-fourth of a total baking time
of 26 minutes. The temperature of the outer crumb increases at an average rate of 4.7°C per minute
to 60°C. An increase in temperature enhances enzymatic activity and yeast growth resulting in an
oven rise (a perceptible increase in loaf volume). When temperature reaches 50–60°C, most enzymes
are inactivated and yeast is killed. Consequently, carbon dioxide is released from the solution, result-
ing in rapid loaf expansion called an oven spring. The volume increases by one-third of the original.
Furthermore, surface skin loses elasticity, thickens and begins to appear brown color. In the second
stage, oven temperature is maintained at 238°C for 13 minutes. Crumb temperature increases at a
rate of 5.4°C per minute to 98.4–98.9°C before keeping constant. At this temperature, all reactions are
maximized, including evaporation, starch gelatinization and protein denaturation. Dough transforms
into crumb structure from outer to inner portions by penetrating heat. A typical brown crust can be
observed when crust temperature reaches 150–205°C. Finally, the volatilization of organic substances
is designated as the bake-out-loss. This period also takes one-fourth of the total baking time.
Operation of an oven is crucial for producing high-quality products. Even though a well-adjusted
oven cannot fully compensate for the errors committed earlier in the processing sequence, it can
bring about the potential of a well-processed dough piece.
A travelling tray oven is efficient in space utilization because it can be made for long horizontal
runs and does not require high ceilings. The trays can be conveyed from the front to the back of the
oven and then returned to the front by a lower track. In addition, a double lap oven can be designed
resulting in twice capacity in the same space. Ducts below and above a tray conveyer are arranged to
regulate temperatures in each part of the oven by forming control zones. Each control zone contains
its individual air and gas supply, temperature controller and groups of burners.
As stated previously, temperature is the dominating factor for final product quality during bak-
ing. This is because it affects enzymatic reaction, volume expansion, gelatinization, protein denatur-
ation, non-enzymatic browning reaction and water migration [3]. The temperature gradient provides
a pressure gradient in the product. Then the pressure gradient causes a gas cell lattice to dilate in
Three-Dimensional CFD Modeling of Continuous Industrial Baking Process 195
one dimension, from the centercenter of bread toward the surface. If such an expanded foam lat-
tice is transformed to be a pore system, the pore structure will show the same dilation. As a result,
such bread is easily fractured along the inside of the crust. Therefore, to avoid the dilation effect,
the heat gradient in an oven has to be adapted to the strength of gluten film forming the interface
of gas cells [4]. The optimum temperature also needs to be maintained at the right time. Otherwise,
product quality can be reduced. For example, supplying too much heat at the early stage of baking
might cause an early crust formation, shrunken bread loaf and too dark of a crust. Too much heat
at the bottom may cause holes towards the bottom of a loaf, and then a triangular shape. Moreover,
cavities may be found at the bottom [5].
A baking air temperature profile can be divided into three stages for sandwich bread [2]. However,
the correlation between air temperature and product quality is poor because there are some other
parameters affecting the product quality at the same time. Instead of air temperature, product sur-
face temperatures were used to study the consequent quality of hi-top bread. It was noticed that the
temperature on the bottom surface increased slightly faster than at the side surface. This was due to
the heat conduction through the plate. At temperatures below 80°C, the top temperature was about
0–10°C lower than the bottom temperature [6]. Then it became higher than both the bottom and
side temperatures, where the bread contacted the tin. An alternative way to measure the combined
influence of the baking temperature and other baking parameters is the heat flux measurement. Heat
flux is defined as the heat transfer rate per unit area required for baking from the oven chamber to
the product [7]. The heat flux measurement was claimed to be more useful than the air temperature
measurement for controlling the quality of bakery products [7–9].
In addition to temperature, airflow velocity affects baking quality. Lack of product uniformity is
possibly due to the non-uniformity of airflow around the product during baking. It was confirmed
that heat flux to a product could be increased by increasing airflow velocity [9,10]. According to [11],
increasing airflow velocity results in a higher weight loss, lower softness and darker surface. Therefore,
either baking time or baking temperature should be reduced with respect to an increase in heat transfer
rate. However, baking requires a minimum temperature to produce an adequate color. When bread
is baked at a very low temperature, very high airflow velocity is required to increase the drying rate
at the surface. Nonetheless, drying rate alone is not enough to produce an acceptable crust color [12].
Similarly, in the case of a microwave baking oven, the product is baked at a low environment tempera-
ture and short time. Not only is the lack of baking color and flavor a problem with microwave-baked
products, but also condensation at the product surface [13]. To overcome the problem, in one study,
airflow velocity was increased. As a result, heat and mass transfer coefficients in the air were increased.
However, the mass transfer coefficient at the surface was still not high enough to rapidly remove the
significant moisture accumulated. Therefore, higher forced airflow velocity should be applied at the
optimal area to increase the surface mass transfer coefficients [14]. When mass transfer coefficients in
the air and at the surface are high enough to significantly reduce the water content on the surface, color
and flavor development could be enhanced at the same time as texture improvement.
Kinetic reactions, including starch gelatinization and browning reactions, depend on not only
temperature but also baking time. Therefore, to reduce the baking time by increasing either air-
flow velocity or baking temperature, gelatinization and browning reactions must be completed.
Otherwise product quality can be degraded. Even though the starch gelatinization and browning
reactions are completed, quality of the product baked by short and long baking times can still be
quite different. Longer baking time can produce caving on the loaf side as well as less softness [5].
With the same oven temperature and airflow velocity, increasing air humidity by either injecting
water vapor into the oven chamber or water vapor migration from the product increases heat flux.
According to a CFD model, the average temperature of an oven composed of 100% water vapor can be
5°C higher than that of an oven containing only dry air [10]. However, water vapor could limit the crust
formation [15]. Therefore, water vapor is normally applied to an oven only at the beginning of the bak-
ing process for bread products. On the other hand, an improper water vapor control and damper con-
trol could render the humidity too low in an oven chamber. As a result, baking loss might increase [2].
9.2.1 Heat and Mass Transfer Mechanisms

9.2.1.1 Heat Transfer Mechanism
During baking, heat is transferred through the combination of three well-known mechanisms: con-
duction, convection and radiation. However, the actual form of combination and proportions are
very different for heat transfer within dough pieces and heat transfer within an oven chamber.
Heat transfer within a dough piece occurs via a combination of conduction from band or tins to the
dough, conduction in the continuous liquid/solid phase of the dough and evaporation-condensation
in the gas phase of the dough. In one study [16], the evaporation-front temperature was assumed to
be 100°C, the point at which a massive amount of unbound water evaporates in the process of boil-
ing. There were four steps proposed for the heat transport inside the dough. Firstly, water evaporates
at the warmer side of a gas cell that absorbs latent heat of vaporization. Then water vapor immi-
grates though the gas phase. When it meets the cooler side of the gas cell, it condenses and becomes
water. Finally heat and water are transported by conduction and diffusion through the gluten gel to
the warmer side of the next cell. The water diffusion mechanism becomes more important when the
dough is a poor conductor that limits the heat transfer via conduction [17].
In an oven chamber, molecules of air, water vapor, or combustion gases circulate throughout the
oven and transfer heat by convection until they contact solid surfaces such as tin, band, conveyor and
so on. Then the heat transfer mode changes to conduction. Meanwhile, radiant energy coming from
burner flames and all hot metal parts in the oven travels in a straight line. Much of it never reaches the
product because it is intercepted by some substances not transparent to radiation. Radiation has two
characteristics different from the other modes of heat transfer. Firstly, it is subject to shadowing or
blocking by intervening layers that are opaque to the radiation. Secondly, it is responsive to changes in
the absorptive capacity of the dough. For example, color changes influence the progression of baking
by increasing the absorption of infrared rays. An increase in the absorptive capacity for infrared rays,
although not apparent visually, is concurrent with the visible change. As a result, there is a tendency
for color changes to accelerate after the first browning appears. Such a tendency might be either good
or bad depending on the desired characteristics of the final product. Therefore, radiation tends to cause
localized temperature differentials of an exposed surface, particularly the darkened area, whereas con-
vection tends to even out temperature gradients [18]. Radiation was confirmed to be the most important
heat transfer mode to bake sandwich bread [9,16], whereas conduction was found to be the most impor-
tant heat transfer mode in the oven chamber to bake an Indian flat bread [19]. According to several
studies [10,20–22], increasing airflow velocity in an oven chamber would increase heat flux to the
product. A change in the chamber gas composition during baking also affects the heat flux. When the
oven chamber was filled with radiation absorbing gases (e.g., water vapor and carbon dioxide), the aver-
age temperature was estimated to be increased by 5°C [10]. So far, the effect of the migration of water
and water vapor has not been integrated into the account of the heat transfer in an oven chamber yet.
9.2.1.1.1 Radiation
The quantity of heat transmitted between two bodies by radiation is directly proportional to the
difference between the fourth powers of the absolute temperatures, which can be described by the
following equation:

( )
q = σε Th4 − Tb4 (9.1)
where q is heat flux (W/m2), σ is Steffan-Boltzman constant (5.67 × 10 –8 W/m2 · K4), ε is emissivity
of tin, Th is heat source temperature (K) and Tb is heat absorbing body temperature (K).
9.2.1.1.2 Convection
Convection is the transfer of heat between a body surface and a fluid, such as air, that is in contact
with the surface. As air heats up, its density decreases. As a result, heated air moves upwards.
In contrast, cooling air with a higher density moves downward. This mechanism is known as a
natural convection. In the case of a baking oven, forced convection is applied, wherein air is gener-
ally moved by fans. The heating rate by forced convection is faster. Convective heat transfer can be
described by the following equation:

( )
q = h Ta − Tb (9.2)
where q is heat flux (W/m2), h is heat transfer coefficient (W/m2 · °C ), Ta is air temperature (°C ) and
Tb is body surface temperature (°C).
9.2.1.1.3 Conduction
Conduction is the transmittal of heat from one part to another part of the same body, or from one
body to another that is in physical contact with it, without macro movement of molecules through
space. The mechanism can be described by an equation known as Fourier’s Law (Equation 9.3).
k
q = − (T1 − T2 ) (9.3)
x
where q is heat flux (W/m2), k is thermal conductivity (W/m · °C), x is distance (m), T1 and T2 are
temperature of location 1 and location 2 (°C), respectively.
9.2.1.2 Mass Transfer Mechanism

Diffusion, evaporation and condensation have been assumed to be the mass transfer mechanisms inside
dough [16,23–25]. The transport of water is driven by the gradients in water content. At the centercenter
of a loaf, the measured water content decreased until the center temperature was at 70 + 5°C because
of the volume expansion. However, the total water content of the loaf should be constant because dough
does not have a continuous pore system. When the temperature reached 70°C, some structural changes
commenced; as a result, the discrete pores became continuous and then allowed water to move freely
[6]. To reduce the partial water vapor pressure due to the temperature gradient, water moves towards
the loaf centercenter and the surface by condensation and evaporation. As a result, crumb temperature
increasing is accelerated. At the surface exposed to oven air, where the partial water vapor pressure is
far from saturation, the water vapor diffuses into the air; as a result, the surface starts to dry out [4]. At
this stage, a differentiation in bread structure is observed. Crumb is a wet core that contains as much
moisture as the dough. Crust is a dried portion; the longer the baking, the higher the thickness [24]. In
a study [25], the moisture content in the center of crumb was measured at 45.7%, compared to 37.2% at
the edge. The moisture movement in crumb and crust can be described by Fick’s Law [25,26].
Almost all moisture loss in bakery products occurs during the baking process because of evapo-
ration. Variations in moisture loss are caused by dough characteristics and baking conditions [2].
As stated earlier, due to the effect of heat during baking, dough structure changes into a continuous
pore structure to allow water to move. Some water vapor is lost through the crust layer by evapora-
tion while some water vapor condenses at dough center. Consequently, moisture content and water
activity change and moisture gradients are formed [27]. Moisture differential has a consequence on
the mechanisms of staling. It may affect the activity of heat stable anti-staling amylolytic enzymes,
because amylolytic activity increases with an increased hydration in dough [28].
In one study, crumb moisture at the loaf center was predicted by the following equation [28]:
Moisture content(%) = 25.3% + 0.3333 ( dough water addition(%) ) (9.4)

According to the models based on water evaporation and diffusion, the water content of crumb
after baking remained the same as the initial dough water content, whereas the crust water content
was close to zero [29]. Therefore, the total moisture loss was basically from the loss at the surface.
A total moisture loss of 53 g per loaf could be from the top crust (29 g), the layer below the top
crust (12 g) and the side crust (12 g) [6]. In the case of a frozen bun, during re-heating, the moisture
loss from bread buns without crust was three times greater than the moisture loss from bread buns
with crust. This is because crust acts as a barrier to mass transfer. Heat supplied to the surface is
used for heating the surface and then conducted through the crust. Without a crust, heat supplied to
the surface is used for evaporating the water from buns. As a result, the center temperature of buns
without crust levels off at a lower level than that of buns with crust [30].
9.2.2 Changes During Baking Process

Baking is a process of gradually turning dough pieces into light, porous and flavourful bakery prod-
ucts, e.g., bread. During this process, yeast fermentation continually produces carbon dioxide at the
beginning and then is killed once dough temperature reaches about 55°C. Subsequently, the semi-
solid dough is solidified into bread due to starch gelatinization and protein coagulation/denaturation.
Meanwhile, brown crust is produced by a non-enzymatic browning reaction. Besides the brown
crust, flavor is also developed at this baking stage [31].
9.2.2.1 Volume Expansion
Carbon dioxide gas is mainly produced by the yeast fermentation process. The stoichiometry of the
reaction is as follows:
C6H12O6 → 2C2H 5OH + 2CO 2 + 2 ATP

The production of carbon dioxide gas by yeast continues at an increased rate during the first stage
of baking until yeast is destroyed at a temperature of about 55°C. According to the Gay-Lussac Law,
occluded gas expands when the temperature increases from 25 to 70°C [32]. At temperatures below
55°C, volume expansion is slightly influenced by temperature. However, temperature shows a sig-
nificant effect on volume expansion after dough temperature reaches 60°C [33]. When temperature
increases during baking, the solubility of carbon dioxide in a liquid dough phase decreases. Then
dissolved carbon dioxide vaporizes. At the same time, saturated vapor pressure of water increases
rapidly; as a result, gas cells expand. In the presence of a constant pressure, the volume of the
occluded gas increases by a factor of 1.15 [32]. However, the pressure remains constant until dough
temperature reaches 50°C because dough viscosity does not affect oven rise up to this tempera-
ture. After that the increased viscous resistance causes an increase in pressure [33]. If the pressure
increases, the expansion will accordingly decrease. The expansion is ceased by the exhaustion of
baking powder, if used, by the inactivation of yeast at higher temperatures, and by the resistance
of the dough to extension. The resistance can be due to either the viscous resistance of the bulk to
deformation or a crust formation at the surface [32].
9.2.2.2 Solidification
The metamorphosis from dough to bread involves crucial steps induced by starch gelatinization and
transition from gel to coagel. The most striking changes are the opening of foam-type gas cells to
produce pores and the solidification of aqueous bulk medium. The solidification is caused partly
by gelatinization and partly by the loss of cohesiveness during transition from gluten gel to coagel.
Bread is ready when the entire crumb reaches 100°C [4].
9.2.2.2.1 Starch Gelatinization
Starch granule is composed of amylose and amylopectin, which can form crystalline structures
together. The space between the crystalline structures is called an amorphous region. In the pres-
ence of water, when molecules are heated to the point that water has enough energy to break the
bond in the amorphous region, gelatinization starts [34]. According to the nuclear magnetic reso-
nance (NMR) baking [35], gelatinization starts at 55°C and finishes at 85°C when evaporation at the
surface dominates. A series of processes at the molecular scale include swelling, melting, disruption
of starch granules and exudation of amylose [32]. As a result of starch gelatinization, the partially
swollen granules can be stretched into elongated forms to allow gas cells to expand. Therefore, the
texture and structure of the product is dependent on starch gelatinization [34]. The extent of starch
gelatinization can be used as a baking index. Besides the baking conditions, addition of some ingre-
dients would also influence the starch gelatinization. Emulsifiers, for example, delay the gelatiniza-
tion; as a result, the period of time for volume expansion is prolonged or the increase in viscosity is
delayed. Therefore, adding emulsifiers is expected to improve the baking performance [4].
9.2.2.2.2 Protein Denaturation
In wheat flour-based products (bread, cracker, cookies, cake, etc.), dough is prepared through the
formation of a wheat gluten network. When dough is subject to high temperature during baking,
changes in its viscoelastic properties are found, depending on the physicochemical characteristics
of the wheat gluten [36]. Heating at temperatures above 60°C leads to an increase in the storage
modulus that characterizes elastic properties. This effect can be explained by the polymerization
of glutenins as a result of a thiol-disulphide interchange reaction. The thermal effect induces the
change from gluten gel to coagel [32]. A previous study [37] hypothesized that changes in the solu-
bility of wheat gluten during baking were dependent on its gliadin fraction. Otherwise the changes
might depend on the high temperature that allowed the activation of thermosetting reactions pro-
ducing the intra- and intermolecular covalent bonds of a protein network. The change in the gluten
phase also enhances the effects of starch gelatinization, such as the transformation from viscous
dough into an elastic material [32]. After the protein is denatured during baking, water adsorbed in
the gluten is released. Then starch uses this water for gelatinization [34,38]. Consequently, dough
becomes semi-rigid bread. Before baking, water in the dough is estimated to be combined with
starch (46%), protein (31%) and pentosan (23%) [39]. Immediately after baking, gelatinized starch
granules are in an amorphous state. When bread is cooled down, water is redistributed as some
starch turns to a crystalline state. More water from gluten is released to be incorporated into the
crystalline structure of starch; as a result, staling is developed [40]. At this time, it is estimated that
no water is associated with proteins. However, some water is still combined with pentosans due to
their high hydration capacity [39].
9.2.2.3 Color Development
Color is one of the important characteristics of baked products contributing to consumer preference.
In bread crust, the higher temperature and lower water content activate non-enzymatic browning
reactions, including the Maillard reaction (sugar-amine) and caramelization [41]. At the beginning
of a Maillard reaction, furosine amino acid is formed. However, the furosine decreases after a high
intensity is reached. In contrast, hydroxymethylfurfural, an intermediate product in the Maillard
reaction and sugar degradation, keeps increasing with higher temperature and baking time [42]. The
resulted brown polymeric compound is called melanoidine. The browning reaction rate depends on
water activity and temperature. The water activity that produces the maximum browning reaction
rate is in the range of 0.4 to 0.6, depending on the type of food substance [41]. At temperatures below
60°C, the browning reaction performs as a zero-order reaction. When temperature is higher, reac-
tion curve follows the first-order reaction. Due to the short period of sub-60°C surface temperature
in baking processes, the overall color reaction is assumed to follow first-order kinetics [43–44]. Its
kinetic constant increases with increased temperature [45] and decreased moisture content [44]
following Arrhenius equation. In bread baking, the crust browning reaction occurs at temperatures
greater than 110°C [30].
Crust color intensity can be measured by an Infrared Engineering (Colorex) sensor or Hunterlab
[46], a monochrome or color image [47] and a Food Analyzer [30]. However, one study [42] claimed
that color intensity measurement was less sensible than measuring the intermediate compound,
hydroxymethylfurfural. Therefore, hydroxymethylfurfural was suggested to be used as a browning
indicator for sliced bread toasting.
9.2.2.4 Flavor Development
Flavor is another quality attribute developed during the baking process in the form of n-heterocycles
via the Maillard reaction. 2-Acetyl-1-pyroline and 2-acetyltetrahydro pyridine are the major flavor
compounds formed in wheat bread crust. Based on the relationship between flavor development and
progress of baking, a flavor sensor was developed to be used as a process indicator for baking and
toasting processes [48]. During baking, the flavor compounds formed are absorbed by pore cur-
vatures. Crust structure also provides a barrier against the loss of flavors [4]. Differences in flavor
development during baking are mainly due to the variation in radiative and convective heat flux
incident on the top of bread [49] and the variation in humidity in the baking oven [50].
9.3 CFD MODELING
9.3.1 Continuous Industrial Baking Oven
In a continuous baking process, dough pieces placed inside trays are delivered continuously from
a prover to an oven. As the dough moves into the oven, it is gradually baked until exiting the oven.
After a designated baking period, the bread is depanned and cooled down. Baking ovens that can
be used for continuous baking can either be tunnel ovens or travelling-tray ovens. The tunnel oven
has its entrance and exit at different ends. In contrast, a travelling-tray oven brings the trays in and
discharges them both at the front end. It requires less space than the tunnel oven [18]. In this chapter,
an indirect-heating travelling-tray oven is studied.
In the indirect-heating system for the oven, hot air is generated from two burners located in the
bottom part of the oven. The area of the burners is separated from the baking chamber by metal
plates. The hot air is then supplied to ducts in the oven chamber through tubes. As shown in Figure
9.1, the oven chamber can be systematically divided into four zones. Each zone contains its own
ducts to transfer heat to travelling trays. Because there are no solid walls between the zones, heat
supplies to the baking zones interfere with each other. Therefore, it is a difficult task to operate the
oven such that a set temperature in each zone can be achieved. In addition, zones 3 and 4, which are
below zones 1 and 2, have two convective fans in the middle area of the heating ducts, respectively.
The convective fans force the hot air to move horizontally towards the front and back of the oven.
In the oven chamber, there are 26 travelling rows; each row consists of 12 trays and each tray is
composed of four tins (Figure 9.2).
9.3.2 Oven Monitoring System in the Continuous Industrial Baking Oven

The oven monitoring system included 15 type-K thermocouples, an in-line anemometer [51]
and a Bakelog (BRI Australia Ltd). Stationary type-K thermocouples were installed in the oven
chamber to monitor temperatures at different locations, as shown in Figure 9.3. To monitor the air
temperature and velocity near the travelling trays, travelling sensors including thermocouples and
an in-line anemometer were also attached to a travelling tin, as shown in Figure 9.4.
9.3.3 Thermal Properties of Dough During Baking

As mentioned earlier, baking is a process of transforming dough pieces, which are viscous, into
bakery products, which are rigid. Thus, some physical properties of dough (e.g., moisture content,
specific volume, porosity) experience significant changes during baking. Consequently, the thermo-
physical properties of dough can vary significantly throughout the baking process. By assuming
FIGURE 9.1 Schematic diagram of the industrial bread baking oven. (Reprinted from Therdthai, N., Zhou,
W., and Adamczak, T., J. Food Eng., 60, 211, 2003. With permission from Elsevier.)
FIGURE 9.2 Illustration of the travelling trays and tins in the oven.
FIGURE 9.3 Diagram of the placement of the stationary thermocouples in the oven.
FIGURE 9.4 Diagram of the placement of the travelling sensors on the tin.
constant thermo-physical properties in the CFD simulation, some serious errors can be yielded. The
sensitivity of CFD simulation to variation in the thermo-physical properties of dough was investi-
gated by Wong et al. [31] in order to identify those properties whose accuracy is more critical than
others.
For a dough piece with density of 245 kg/m3, initially (at temperature of 28°C) its heat capacity
and thermal conductivity were at 2.80 × 103 J/kg K and 0.282 W/m.K, respectively. When the dough
was heated to 60°C, its heat capacity and thermal conductivity were decreased to 2.32 × 103 J/kg
K and 0.126 W/m.K, respectively. After baking for a while, the temperature of the dough increased
to the evaporation point of water, resulting in moisture loss. Protein denaturation and starch gela-
tinization occurred, resulting in crumb formation. The temperature of the crumb or inner layer
kept increasing and reached 120°C. At this point, the heat capacity and thermal conductivity were
decreased by half to around 1.13 × 103 J/kg K and 0.0560 W/m.K, respectively. For the outer layer,
the temperature can be much higher than the crumb to form crust. However, the thermo-physical
properties did not change much. At 227°C, the heat capacity and thermal conductivity were around
1.01 × 103 J/kg K and 0.0528 W/m.K, respectively [31].
The density of dough varies depending on the formulation of dough. With a higher density of
327 kg/m3, its heat capacity and thermal conductivity are around 1.26 × 103–3.50 × 103 J/kg K and
0.0660–0.353 W/m.K, respectively, depending on dough temperature. In contrast, the heat capacity
and thermal conductivity can be decreased to around 0.757 × 103–2.10 × 103 J/kg K and 0.0396–
0.212 W/m.K, respectively, when the density is lowered to 164 kg/m3 [31].
To test the sensitivity of each property, an effect analysis should be carried out. The effect analy-
sis can estimate the relative strength of each property and their interaction, which subsequently
explains how the variation of each property affects the prediction [52]. In the case of a full baking
load, density and heat capacity showed the highest effect on the temperature profiles, including side,
top, bottom, and center crumb temperatures. In addition, their interactions also had a significant
effect. Dough with lower density and heat capacity would have higher temperature profiles than one
with higher density and heat capacity. Regarding thermal conductivity, it should theoretically have
a positive effect on the crumb center temperature but a negative effect on the crust temperature.
However, in the case of a baking oven that could supply enough energy to maintain conductive
heat towards the loaf center and high temperature on the surface, the effect of thermal conductivity
might not be significant compared with density and heat capacity.
As baking progressed, the effects of density and heat capacity on the surface temperature profiles
continually increased. However, their effects on crumb center temperature remained constant for
the whole baking process. The crumb temperature was more affected by thermal conductivity than
by density and heat capacity. The effect of thermal conductivity increased as baking time increased.
The relationship between thermo-physical properties and dough/bread temperature was estab-
lished by Wong et al. [31], as shown in Equation (9.5).
∆Tgi = ∑3j =1 agij ∆x j + ∑3j ,k =1, j ≠ k agijk ∆x j ∆x k + agi123 ∆x1∆x 2 ∆x3 + ∑ 2j =1 bgij ∆x 2j (9.5)

where ∆Tgi (g = 1,2,3,4; i = 1,2,3,4) are temperature changes: ∆T1i (i = 1,. . . ,4) are the changes in
the average top temperature in the four zones, respectively; ∆T2i (i = 1,. . . ,4) are the changes in the
average side temperature in the four zones, respectively; ∆T3i (i = 1,. . . ,4) are the changes in the
average bottom temperature in the four zones, respectively; ∆T4i (i = 1,. . . ,4) are the changes in
the average dough temperature in the four zones, respectively.
xj (j = 1,2,3) are the physical properties: x1 the density (Factor A), x2 the heat capacity
(Factor B) and x3 the thermal conductivity (Factor C).
∆xj∆xk (j,k = 1,2,3; j ≠ k) are the two-factor interactions (A * B, A * C, B * C),
∆x1∆x2∆ x3 is the three-factor interaction (A * B * C).
agij, agijk, agi123, and bgij are model parameters.
By using Equation (9.5) to calculate changes in the bread crust temperature profiles, the error
of prediction was very small, ranging from 0.0148 % to 0.359%. The prediction of crumb center
temperature yielded larger errors, ranging from 0.142 to 0.695%. However, the maximum error
was equivalent to ±0.7°C. This was similar to the error of the instrumental measurement, which is
normally around ±0.5°C [31].
9.3.3.1 Modeling
9.3.3.1.1 Simplification of the Baking Oven Geometry
Due to the complexity of the oven geometry, it was necessary to simplify it in the model to reduce
processing time during CFD simulation. In the actual industrial baking process, lids are preheated
before covering the tins. Thus, the initial tin temperature on the top is always higher than those on
the bottom and side, which are approximately the same as the dough temperature after proving. To
simulate this situation, a preheat block was created to heat the top surface of dough to the actual
initial top tin temperature before the dough moved into the oven (Figure 9.5). Using this set up,
when dough entered the oven, its initial bottom and side temperatures were the same as the dough
FIGURE 9.5 Illustration of the simplified baking oven
temperature, whereas its initial top temperature was higher and equivalent to the actual heated lid
temperature in the industrial operation. One travelling tray (0.55 m × 0.12 m × 0.28 m) actually
consists of four tins (0.12 m × 0.12 m × 0.28 m) with air gaps between them. To predict the side
temperature, the tins were individually simulated with air gaps between them. In addition, two con-
vection fans in zones 3 and 4 were simplified as blocks in the middle of the duct panels to simulate
the flow generated by the convection fan to the left and right sides of the oven. To reduce the number
of cells, non-uniform mesh was used. A coarser mesh was applied inside the oven chamber, whereas
a finer mesh was applied to the travelling tins. Consequently, the number of cells and faces became
93,654 and 323,786, respectively.
9.3.3.2 Model Assumption
9.3.3.2.1 Three-Dimensional Flow
To simulate the moving of trays into the oven as in the continuous baking process, a three-dimensional
cross-section of the oven was taken as the calculation domain (Figure 9.5). At the cross-section
(280 mm thickness, which is equivalent to the width of a baking tray), there were 52 trays of dough
continuously moving via both the top and bottom tracks in the oven.
9.3.3.2.2 Turbulent Flow
Due to the convection fans generating flow above the bottom ducts, air in the oven can be influenced
by both forced convection and natural convection. Flow can be observed to have three character-
istics: laminar, transitional and turbulent. The Reynolds number (NRe) is used to describe the flow
characteristics. It provides insight into energy dissipation caused by viscous effects. When viscous
forces dominate the effect on energy dissipation, the Reynolds number is small. In the oven that
contained 3.65 × 0.90 m2 of opening area and the typical 0.3 m/s of airflow velocity, the Reynolds
number was approximately 2.7255 × 104. Thus, the flow characteristics could be assumed to be
turbulent.
9.3.3.2.3 Transient State Computation

The objective of this chapter is to simulate the continuous baking process with a constant mass
flow rate through the oven. Both temperature and airflow velocity depend on time and location.
Therefore, a transient state was set up for the simulation.
9.3.3.3 Volume Condition Settings

9.3.3.3.1 Dough Properties
The physical properties of dough and bread, including density, specific heat and thermal conductiv-
ity, were set up as variables depending on temperature. The physical properties were also dependant
on moisture content. However, in the case of bread, the difference between the moisture content of
crumb and that of dough was not significant. Furthermore, bread is composed of dominantly crumb
rather than crust. Therefore, in the model, dough properties were set as piece-wise linear functions
temperature only, as shown in Table 9.1.
9.3.3.3.2 Flow Source
Due to the convection fans located at the bottom ducts, small blocks of convection fans were added
in the middle of the bottom duct panels, as shown in Figure 9.5. Therefore, flow sources were set up
in the small blocks. In addition, the sides of the blocks were set up as flow inlets with fixed velocity
on the x-axis to generate flow to the oven chamber. At the top of the blocks, a boundary condition
was set as a wall to prevent air from flowing up. In the real oven, there were plates on top of the fans
to force air to move towards the front and the back of the oven rather than the top area. In addition
to the sides and top of the blocks, a boundary condition at the bottom of the blocks was set up as
flow outlet to suck air out of the oven chamber.
9.3.3.3.3 Heat Source
The U-turn movement in the actual oven (Figure 9.1) could not be modeled due to the limitation
of grid deformation in the software (CFD-ACE+). Therefore, the system was simplified by having
only the top shelf moving towards the back of the oven, and then moving out. After that, hot dough,
equivalent to the dough that had been baked for 50% baking time, moved in via the bottom shelf
towards the front end of the oven. To heat up the dough entering the bottom shelf to be equivalent
to the last dough of the top shelf, heat blocks with a total length covering 50% baking time were set
up as isothermal heat sources.
9.3.3.4 Boundary and Initial Condition Settings

9.3.3.4.1 Convection Heat Transfer from Duct Surfaces
The burners were to supply hot air to the ducts. Then the hot air heated up and maintained the duct
surfaces at certain temperatures. Therefore, in the model the duct surfaces were set up as isothermal
TABLE 9.1
Thermal Properties of Dough and Bread
Specific Heat Thermal Conductivity
Temperature (°C) Density (kg/m3) (J/kg · °C) (W/m · °C)
28 420 2883 0.20
120 380 1470 0.07
227 340 1470 0.07
TABLE 9.2
Duct Temperatures
Duct Temperature (°C) Zone 1 Zone 2 Zone 3 Zone 4
At supply air duct 336 336 397 370
At return air duct 311 300 300 311
walls. Finally, the temperature gradient due to convective heat transfer from the isothermal duct
surfaces to air in the oven chamber was computed by the CFD-ACE+ codes (Equation 6). The heat
transfer coefficient (ha) was calculated according to flow status inside the oven chamber.
qa = ha (Tduct − Tair _ inside _ oven ) (9.6)

The initial condition was Tair_inside_oven = Ta0 at t = 0. Duct temperatures at different locations are
shown in Table 9.2.
9.3.3.4.2 Radiation Heat Transfer from All Metal Surfaces Inside Oven Chamber
In addition to the convective heat transfer, radiant heat came from all hot metal parts in the oven. It
travelled straight through the space and caused localized temperature differentials. In the model, it
was calculated as follows:
qb = σε (TA + 273)4 − (TB + 273)4  (9.7)

where σ is Stefan–Boltzmann constant (5.669 × 10 −8 W/m2 °C4), ε is emissivity (it was assumed to
be at 0.95 [43], TA is temperature of heat source (°C) and TB is temperature of heat absorb (°C). The
initial condition was TB = TB0 = 40°C at t = 0.
9.3.3.4.3 Heat Loss Through Oven Walls

Due to the insulation, heat loss though the oven walls was very small. The overall heat transfer coef-
ficient from the combined conduction and convection was approximately 0.3 W/m2°C. Heat loss can
be calculated by Equation (9.8):
qc = hc (Toven _ inner _ wall − Tair _ outside _ oven ) (9.8)

The initial condition was Toven_inner_wall = Tc0 at t = 0.
9.3.3.4.4 Travelling Trays

The boundaries of the travelling trays and the oven chamber above and below the travelling trays
were set as arbitrary interfaces. This allowed fluid to flow between the travelling trays and the static
oven chamber, when the trays were travelling inside the oven. There was no fluid flow between the
oven chamber and the waiting trays outside the oven. This was close to the actual baking scenario.
Dough started baking after entering the oven and finished when exiting the oven. Conductive heat
transfer inside the dough was calculated by Equation (9.9).
qd = kd (Tdough _ surface − Tdough _ center ) /thicknessdough (9.9)

The initial condition was Tdough_center = Td0 at t = 0.

Therefore, heat transfer at the dough surface can be estimated by Equation (9.10).
qd A = mCdTdough /dt (9.10)

The initial condition was Tdough = T0 at t = 0.
9.3.3.4.5 Exhaust Box
Normally an exhaust is installed to suck cold air from the front door out of the oven chamber. To
simulate the exhaust, one section of the bottom floor was set up as a flow outlet with fixed flow
velocity. This boundary allowed cold air to be removed from the oven with a constant flow rate.
9.3.3.5 Solving the Model

Together with the initial and boundary conditions, the continuity, Navir-Stokes, and energy con-
servation equations were solved by a CFDACE+ (V2002) code. The program discretized the dif-
ferential equations to produce a numerical solution. CFD-ACE+ code uses an iterative, segregated
solution procedure. Two nonlinear equation solvers, the conjugate gradient squared (CGS) plus pre-
conditioning solver and the algebraic multigrid (AMG) solver, were applied. The first-order upwind
spatial differencing scheme was used. The relevant set of linear finite difference equations was
solved sequentially and repeatedly until a converged solution was obtained.
Generally, a radiating boundary can be treated as a direct boundary problem. Recently, a bound-
ary radiating temperature profile (at oven walls) was estimated by formulating an inverse boundary
problem. Initially, the objective function was optimized for one time-step at a time. The first-order
Tikhonov regularization technique and the pre-filtering technique were used to deal with the noise
in the input data. The first-order Tikhonov regularization technique required a realistic set-up of ini-
tial conditions to perform reasonably well. In contrast, the pre-filtering technique performed better
with no requirement of realistic set-up initially. The solving time of the inverse boundary problem
was about 50% longer than that of the direct method [54].
9.3.3.6 Model Validation
Despite the oven geometry simplification in the model, the prediction of the top temperature
profile was reasonably accurate. However, over-prediction in the bottom and side temperature
profiles appeared in zones 3 and 4 (Figure 9.6). The model performance could be described
by the values of correlation coefficient (R) and mean square error (MSE) between the model
predicted values and the values measured during industrial production, as shown in Table 9.3.
For the velocity profile, as shown in Figure 9.7, a few undesired peaks and valleys appeared. The
most severe peaks were around the U-turn. This could be due to the simplification of the model
configuration at the U-turn area. As mentioned earlier, the present model could not simulate the real
U-turn movement due to the limitation in the software’s capability. As a result, some errors could be
produced from the simplification of the tray moving direction.
According to Table 9.4, in comparison of the predicted velocity with the measured velocity, a
relative error of 37.31% was observed for the whole oven. The error was mainly from the U-turn
area in zones 2 and 3. Due to the simplification of tray movement at the U-turn at the end of zone 2
and the beginning of zone 3, a large relative error was found in a high proportion to the measured
velocity. A previous study [55] observed an error of 26–28.5% in using a CFD model to simulate the
airflow pattern in a cold store with palloxes. In addition, an error of 40% when predicting airflow
pattern in a chiller with objects was reported in [56].
By comparing the difference between measured velocity and predicted velocity relative to the
velocity at the flow source, the error was 4.29%. Due to a much lower flow rate in zones 1 and 2,
the error relative to the velocity at the flow source in these two zones was significant lower than
that in zones 3 and 4. In one of the earlier studies, an error of 5% in simulating airflow pattern in a
FIGURE 9.6 Top, side and bottom surface temperature profiles. (Reprinted from Therdthai, N., Zhou, W.,
and Adamczak, T., J. Food Eng., 65, 599, 2004. With permission from Elsevier.)
TABLE 9.3
Model Performance
Top Bottom Side
Performance Temperature Temperature Temperature Velocity
R 0.9132 0.9065 0.9096 0.6019
MSE 141.1939 276.7656 281.9919 0.0336
FIGURE 9.7 Velocity profile. (Reprinted from Therdthai, N., Zhou, W., and Adamczak, T., J. Food Eng., 65,
599, 2004. With permission from Elsevier.)
TABLE 9.4
Relative Error Compared to Measured Velocity and Flow Source
Compared to Measured Compared to Velocity at Flow Source
Relative Error Velocity (3 m/s)
Zone 1 37.69 % 2.35 %
Zone 2 39.69 % 2.85 %
Zone 3 39.16 % 5.84 %
Zone 4 32.29 % 5.26 %
Whole oven 37.31 % 4.29 %
mechanically ventilated livestock building (the bulk jet velocity was 6.1 m/s) without objects was
also observed [57].
Overall, the model presented an acceptable performance in predicting airflow velocity compared
to other related works [55–57], although its performance was not as good as that in predicting tem-
perature profiles.
9.4 APPLICATIONS OF CFD MODEL

9.4.1 Predicting Temperature Profiles for Quality Estimation of Baked Bread
CFD has been successfully used for the prediction of temperature and airflow distributions in bak-
ing ovens. The CFD predicted temperature profiles could be used to further predict key quality
attributes of bread, including color, texture and baking loss. In order to obtain essential data for
establishing a relationship between the temperature profiles and the quality attributes, bread was
sampled from an industrial production process to determine variations in crust color by a colorimeter
(Minolta CR-310) and in weight loss by a scale with accuracy up to 0.1 g. The color was expressed
as lightness (L*), redness (a*) and yellowness (b*). To measure weight loss, 24 dough pieces of white
sandwich bread were sampled from a conveyor before entering a prover in the industrial produc-
tion line and marked on the surface. Their initial weight (W0) was recorded before returning to the
prover. After baking was completed, the bread samples came out of the oven and were depanned
automatically. Then the weight of all samples were checked and recorded as W1. The weight loss
was the difference between W0 and W1.
With good performance in predicting the top temperature profile and reasonable performance
in predicting the bottom and side temperatures, the CFD model could be used to predict the cor-
responding bread quality attributes by utilizing the mathematical models developed by [58].
xi = WTi*Ti + WSi*Si + WBi* Bi (9.11)
where xi (i = 1…4) are the weighted temperatures in the four zones and W Ti, WSi and W Bi are
weighting factors whose values are given in Table 9.5.
Using Equation (9.11) and Table 9.5, the average weighted temperature profiles were 116,130,172
and 170°C for zones 1,2,3 and 4, respectively. The corresponding quality attributes were then pre-
dicted by using Equation (9.12) [58].
5 5 5
yi = bi 0 + ∑
j =1
bij x j + ∑
j ,k =1, j ≠ k
bijk x j x k + ∑b x
j =1
ijj
2
j (9.12)

where yi (i = 1…6) are the quality attributes: % weight loss (y1), internal temperature (y2), side crust
color (y3), top crust color (y4), bottom crust color (y5), and average crust color (y6). xi (i = 1…4) are
the tin temperatures in the four zones, respectively: x1, x2, x3, and x4. x5 is the baking time. bi0, bij,
bijk and bijj (i = 0,1…5; j = 1…5; k = 1…5; j≠k) are model parameters.
As shown in Table 9.6, by using the temperature profiles obtained from the CFD model [59]
and the mathematical models [58], the predicted quality attributes were in good agreement with
the measured values.
TABLE 9.5
Weighting Factors for the Top, Sides, and Bottom Tin Temperatures
Weight Factor Zone 1 Zone 2 Zone 3 Zone 4
WTi 0.7500 1.4000 1.0811 1.7889
WSi 0.4453 0.4219 0.4516 0.3518
WBi −0.1953 −0.8219 −0.5328 −1.1406
TABLE 9.6
Comparison of the Measured Bread Quality Attributes with Those Predicted from the CFD
Model and Mathematical Models
Internal Top Crust Side Crust Bottom Average Crust
Quality Weight Loss Temperature Color Color Crust Color Color
Attributes (%) (°C) (L* value) (L* value) (L* value) (L* value)
Measured quality 9.10 98.41 49.93 70.47 50.57 56.99
Predicted quality 9.36 97.86 48.81 71.15 49.80 56.82
FIGURE 9.8 Effect of thermo-physical properties of dough on bread quality. (Reprinted from Wong, S.Y.,
Zhou, W., and Hua, J., J. Food Eng., 77, 784–791, 2006. With permission from Elsevier.)
It is confirmed that baking temperature plays the most important role in determining the quality
attributes of bread. Weight loss in baking process can be from moisture diffusion and evaporation.
However, the majority of weight loss is due to the evaporation of water, particularly at dough surface.
Differences in weight loss are also due to variations in dough density and heat capacity. Dough of lower
density and heat capacity requires less energy to reach the evaporation point. Thus, for the same baking
time, the surface temperature of the dough increases faster and crust is formed earlier. As a result, its
weight loss is relatively higher, compared with dough that has higher density and heat capacity.
In fact, early crust formation also causes a darker crust. Crust color is due to the browning reac-
tion, which is enhanced by temperature. Normally, crust color turns to brown when the temperature
reaches 150–200°C. From an effect analysis, dough density and heat capacity have a positive effect
on the side, top and bottom crust colors (Figure 9.8).
9.4.2 Simulation of Oven Operation Under Increasing Oven Load

During a continuous baking process, baking trays were continually delivered into the oven. Each
step of the tray movement increased the oven load unless the oven load already reached its maxi-
mum (i.e., was full). The change in oven load affected the temperature gradient; as a result, the
convective flow simultaneously changed. This led to a change in overall airflow pattern. At the same
time, the tray movement directly forced the airflow to change its pattern. The variation of airflow
pattern could be simulated throughout the whole baking time using the three-dimensinal CFD
model. Each travelling tray experienced different airflow and thereby different temperature profiles.
From the simulation results of the CFD model [59], the average temperature profiles of the earlier
tins were slightly higher than those of the later tins (Figure 9.9).
With the mechanistic models developed earlier in [58], the simulated temperature profiles could
be used to estimate the product quality attributes influenced by the change in oven load. The esti-
mated weight loss of bread in the moving tray Nos. 4, 8, 12, and 16 was 9.71, 9.56, 9.49 and 9.45%,
respectively (Figure 9.10, Part A). At the beginning (i.e., when fewer trays were inside the oven),
the variation in weight loss between trays was significant, due to the significant difference in the
corresponding crust temperature. However, the variation was reduced and became nonsignificant
when the number of trays increased. Similarly, the significant decrease of temperature profile at
the beginning made the crust color lighter (i.e., higher L* value) as the number of trays in the oven
FIGURE 9.9 Effect of the oven load on temperature profiles.
increased (Figure 9.10, Part C). This was due to the slowing down of browning reactions on the sur-
face. Although the variation in the crust temperature was significant, there was no significant dif-
ference in the internal temperature among the different trays (Figure 9.10, Part B). These simulated
phenomena were consistent with the observations of the actual baking in industry during a start-up
period as well as when there was a production gap in the oven chamber.
9.4.3 Simulation of Baking Index

The baking index is an indicator of the degree of baking that is related to the gelatinization rate of
starch. By integrating a kinetic model of starch gelatinization in [60] with the CFD model in [59],
the baking index for bread baked in the travelling -tray oven could be simulated [61]. Normally an
industrial baking tray is composed of four tins. Tin No. 2 had two tins on its right and one tin on its
FIGURE 9.10 Effect of the oven load on the quality attributes of the final bread. A: weight loss, B: internal
temperature and C: crust color. (Reprinted from Therdthai, N., Zhou, W., and Adamczak, T., J. Food Eng., 65,
left. To study the extent of starch gelatinization within a loaf, bread baked in Tin No. 2 was exam-
ined. As shown in Figure 9.11, at positions 40, 35, 25, 20, 10 and 0 mm from the center to the left and
right sides, there was significant variation in the starch gelatinization rate. In the outer crumb layers
(40 mm from the center) on the left and right sides of the loaf, the gelatinization process started in
zone 1 and reached the maximum extent (i.e., baking index = 1) within 10 minutes. In zone 1, only
the outer crumb layer (40 mm from the center) started the gelatinization process. When baking time
was increased, heat penetrated to the inner crumb layer. Other layers towards the center started and
finished the gelatinization process sequentially later. Due to the position of loaf No. 2, which had
two tins on its right and one tin on its left, the gelatinization rate of the left side layer was faster as
a consequence of the temperature gradients within the tray. The difference in the gelatinization rate
between the left and right sides became more obvious in zone 3, where the airflow pattern changed
due to the U-turn movement. In zone 4, the variation in the gelatinization rate was not significant,
because the gelatinization process nearly reached the maximum extent.
Considering the variation in the gelatinization rate from the top to the bottom of the loaf (0–40 mm
away from the loaf center), the bottom crumb layer tended to have a faster gelatinization rate.
FIGURE 9.11 Variation of starch gelatinization within a loaf: (A) top-half; (B) bottom-half; (C) left-half;
and (D) right-half. (Reprinted from Therdthai, N., Zhou, W., and Adamczak, T., J. Food Eng., 65, 543, 2004.
With permission from Elsevier.)
This was due to more radiation heat from the heating ducts to the bottom of the loaf. The temperature
gradient made the loaf temperature on the bottom higher and thereby increased the gelatinization
rate. However, the variation in the gelatinization extent decreased in the area near the loaf center
(0–20 mm away from center vertically and horizontally), possibly due to the condensation of water
vapor. Condensation maintained the temperature at the loaf center region below 100°C [16]. In addi-
tion, the moisture content of crumb at the loaf center region was at the same level as that of dough.
Therefore, the starch gelatinization, a thermal reaction, could be nearly uniform at the loaf center.
In addition to the variation among the travelling tray, variation within the same travelling tray was
also found (Figure 9.12). When a tray moved from left to right along the top travelling track (Figures 9.1
and 9.2), Tin No. 4 was generally the hottest of the four tins in the same tray. After the moving direction
of the tray changed to left-to-right in zones 3 and 4 (i.e., along the bottom travelling track), Tin No. 1
became the hottest. In zones 1 and 2, the gelatinization was just activated; therefore the correspond-
ing variation was not significant. Significant variation in the gelatinization extent was found in zone 3.
In zone 4, variation became nonsignificant again, as the gelatinization already reached the maximum
extent. Although the starch gelatinization extent was taken as the baking index, the effect of gelatiniza-
tion rate and profile during baking on the final product quality attributes has not been studied yet.
FIGURE 9.12 Variation in the gelatinization extent among tins in the same tray. (Reprinted from Therdthai,
N., Zhou, W., and Adamczak, T., J. Food Eng., 65, 543, 2004. With permission from Elsevier.)
9.4.4 Design of Operating Condition to Achieve the Optimum Tin Temperature Profiles

According to the three-dimensional CFD simulation results [59], the current oven operating condi-
tion resulted in a temperature profile of 116, 130, 171, and 170°C for zones 1, 2, 3 and 4, respec-
tively. Using the mechanistic models in [58], the consequent weight loss was 9.35% after baking for
25 minutes. To reduce the weight loss without changing the baking time, the duct temperatures and/
or airflow volume needed to be adjusted. The optimum average weighted temperature profile for
a total baking time of 25 minutes was found to be 106, 130, 166, and 178°C for zones 1, 2, 3 and
4, respectively. Compared to the current profile, there was some significant difference. Under the
current operating condition, the average weighted tin temperatures in zone 1 and 3 were too high,
whereas the average weighted tin temperature in zone 4 was too low. According to the simulation
results from a two-dimensional CFD model [62], the most effective way to manipulate the baking
condition was to adjust the airflow pattern. Increasing the volume of airflow tended to increase the
baking temperature profiles. Therefore, heat supply to the oven could be reduced. To obtain the
optimal temperature profile that reduced weight loss, the three-dimensional CFD model was used to
simulate the oven operating condition with various airflow volumes and heat source temperatures.
Due to the temperature profiles in zones 1 and 3 being too high, the duct temperatures in zones
1, 3 and 4 were reduced by approximately 10°C in order to decrease the tin temperatures in zones 1
and 3. By only reducing the duct temperatures in some zones, the temperature profiles in zones 2
and 4 might become too low. Therefore, at the same time as decreasing the duct temperatures, the
flow volume applied to the convection fan in zone 3 should be doubled in order to maintain the tin
temperatures in zones 2 and 4. By increasing the flow volume supplied to zone 3, the difference
between the top, side and bottom tin temperatures in zone 4 also became less. From the simula-
tion results, by integrating the CFD model with the mechanistic models in [58], the proposed new
operating condition would produce a weight loss of 7.95% with the lightness values of the crust
color on the top, bottom and side of the loaf being around 50.68, 55.34, and 72.34, respectively.
Although heat supply was reduced, the obtained bread still had a completed baking inside the loaf,
with its internal temperature reaching 98.69°C. Based on the simulation results utilizing the kinetic
model of starch gelatinization, the gelatinization rate of dough baked under the proposed operating
FIGURE 9.13 Comparison of the starch gelatinization profiles during baking under current and modified
conditions. (Reprinted from Therdthai, N., Zhou, W., and Adamczak, T., J. Food Eng., 65, 543, 2004. With
permission from Elsevier.)
condition was slower than that baked under the current operating condition, especially in zone 3.
This was due to the reduction in energy supply in zone 3, which was the major factor causing high
weight loss. However, the gelatinization was speeded up in zone 4 and finally reached the maximum
gelatinization extent at the end of the proposed baking process, as shown in Figure 9.13.
Therefore, by adjusting the oven operating condition as proposed, the energy supply could be
reduced, while the yield could be increased and the quality attributes retained within an acceptable
range. The proposed oven operating condition is highly recommended for industrial process operations.
9.4.5 Design of Process Controller for a Continuous Bread Baking Process

Despite the fact that the baking condition might be optimized through various approaches, to imple-
ment a set of optimum conditions into the actual baking process is not an easy task. In fact, keeping
the oven operation at the optimum condition for the entire baking process is even more difficult. To
overcome this problem for industrial continuous bread baking, applying CFD models to the design
of process control system was explored in [63]. The CFD models were incorporated into a feedback
control system with multiple decoupled PI controllers to control the temperatures during baking.
The controllers assisted by the CFD models performed satisfactorily in response to changes in set
points and disturbances. The immediate effect of a controller action on the baking process could
be evaluated continuously throughout the system, rather than only at specific points where sensors
were located. CFD software can be combined with process simulation software (e.g., a commercial
CFD package such as Fluent can be integrated with a general-purpose advanced process simula-
tor such as gPROMS) to exchange information and thereby simultaneously solve the problems of
process behavior prediction and controller action determination. CFD could be used as a tool to
generate time-series data that can be subsequently used to develop a reduced-order transfer function
matrix (TFM) model. Based on the TEM, the controller can be designed.
FIGURE 9.14 Integration system of CFD model and feedback controller. (Reprinted from Wong, S.Y., Zhou,
W., and Hua, J., J. Food Eng., 81, 523–534, 2007. With permission.)
In the integration system, the CFD model developed through software (Fluent 6.2.16) was used
to calculate heat transport, natural convection and turbulent flow in the baking oven [63]. At the
same time, the feedback controllers were introduced through user defined functions. The structure
of the combined system is shown in Figure 9.14.
CFD model was used to predict two temperatures: T1 and T2. T1 was given by a sensor located
in the front part of the oven in zone 1 while T2 was given by a sensor located in the back part
of the oven in zone 2. However, the measured temperatures always had some oscillations due to
the movement of dough/bread. The oscillation period was equivalent to the time required for one
dough/bread tray to pass one of the temperature sensors. It is counter-productive to feed oscillated
temperature measurements to a feedback controller. Therefore, a filter of 30 s moving average was
applied to smooth the temperature measurement data before feeding them to the controller.
In the control system, the measured temperatures T1 and T2 were compared with their set points
(Ts1, Ts2). The differences between them were defined as error signals (Er1, Er2). When the error
signals were detected, the first controller adjusted MV1 to bring T1 back to its set point. However,
adjusting MV1 had some impact on Er2. Furthermore, Er2 was used to adjust MV2 by the controller
in the second loop, in order to bring T2 back to its set point. Then MV1 and MV2 were sent to the
CFD part to manipulate the output temperature of the burners below zone 4 (the bottom-front part
of the oven) and zone 3 (the bottom-back part of the oven), respectively. Consequently, the duct and
tube temperatures in zones 4 and 3 would be changed. Therefore, the boundary conditions of the
CFD model were changed accordingly. The CFD solver ran subsequently to produce a converged
solution. To avoid the numerical load for the CFD model from becoming too high, changes in the
boundary conditions were implemented at an interval of 30 s.
As a common industrial practice, prior to the baking of dough, the empty oven is preheated to a high
temperature. Thus, both Ts1 and Ts2 initially had to be set at that high temperature and then reduced at
a later stage. Under the designed controller, similar surface temperature profiles of dough/bread across
different zones were achieved with and without the initial preheating. Elimination of the preheating
stage will reduce the start-up period and save energy, while producing the same quality of bread.
Compared to the traditional controller design, the integrated process control system with the CFD
model could better provide full fluid flow information during baking, not limited to only some moni-
toring points or under specific operational parameters. With the better information and understand-
ing, a dual-mode controller was designed to efficiently deal with nonlinearity of the baking process.
Both modes performed satisfactorily in response to changes in disturbances and set points.
9.4.6 Design of Indirect Gas-Fired Radiant Burner for a Tunnel Baking Oven

A tunnel oven is one type of continuous baking oven. It is normally heated up indirectly, using
a burner or burners to generate radiant heat to the oven chamber. To increase energy efficiency,
the burner should be properly designed. CFD was used to predict radiation heat fluxes in the oven to
help design a gas-fired burner in [64]. In addition, the effect of humidity and surface emissivity on
the radiation heat distribution in the oven was studied.
Three CFD models were developed using CFX 11.0 (ANSYS Inc., Waterloo, Canada). When the
root mean square residual values for momentum and energy balances were decreased to less than
10 −4, convergence was considered achieved. The first model was developed to simplify multiple
radiant tubes into a single curve radiating surface. To account for natural convective heat transfer,
inflated boundary layers were used while other surface boundaries were set as symmetric planes.
The number of nodes was 45, 825. Time to achieve convergence was around 30−0 min after 200
iterations. Based on the CFD simulation, the effects of moisture content in atmosphere and radiant
tube surface emissivity on heat flux to bread and conveyers were investigated.
In the second CFD model, the multiple radiant tubes were included to be closer to the actual
configuration of the oven. The measured wall temperatures of the radiant tubes were used as model
inputs for simulating radiative heat distribution. Due to the increased complexity and the increased
number of nodes to 59, 770, the numerical calculation became more time-consuming. Thus, vertical
symmetric planes through the center of burner were assumed to reduce the calculation time to a
range of 50−60 min. After 100 iterations, convergence could be obtained.
The third CFD model was established to design the locations and optimum numbers of radia-
tive tubes. The number of nodes was dramatically increased to 388, 495 and the calculation time
increased to 5−6 h. After 170 iterations, convergence could be obtained.
To validate the model, measurement data were compared with model predicted data. Only a
10% error of prediction was observed, which confirmed the predictability of the CFD model to be
further used for burner design. Based on the CFD simulation results, a relationship between the
radiative heat flux and the radiant tube emissivity was established, which was clearly linear. Thus,
variation in the emissivity of radiant tubes could seriously affect the radiant heat incident on the
product. The emissivity of typical stainless steel was around 0.3. After being used for a period of
time, it could be tarnished by heat and thereby its emissivity might go up to 0.8. This means that
the total radiative heat flux varyied although the oven setting remained unchanged. Baking loss and
quality attributes of bread could vary from time to time because the radiant tubes were getting more
tarnished. Therefore, the radiant tubes for a prototype burner were pre-treated to have a uniform
emissivity of 0.85.
In the prototype oven, there were four radiant tubes. The set temperatures were 760, 670, 600 and
542°C for tubes 1, 2, 3, and 4, respectively. With this design, the difference between the maximum
and minimum temperatures was huge. The ratio of maximum to minimum temperatures was 3.15.
To reduce the temperature difference, the burner was lifted. The clearance between tube 4 and the
product was increased from 110 to 210 mm. As a result, the ratio of maximum to minimum tempera-
tures was reduced from 3.15 to 2.20. With this design, further slight improvement of temperature
uniformity could be achieved when the temperature of tube 1 was increased from 760 to 856°C. The
ratio of maximum to minimum temperatures was thus reduced to 2.11. However, reducing the num-
ber of tubes by removing tube 4 did not improve the uniformity of temperature. In fact, the ratio of
maximum to minimum temperatures was increased to 2.21. Thus, the burner design kept four tubes.
Based on the CFD simulation results, the best design of the burner is shown in Figure 9.15.
The radiant tube temperatures were 856, 620, 550, and 500°C for tubes 1, 2, 3, and 4, respectively.
Plenum shield emissivity was increased from 0.3 to 0.85. Plenum shield temperature was adiabatic
and set at 500°C. A convection air jet shield was installed to be an additional source of radiation.
This design was shown to be able to dramatically improve the uniformity of heat flux incident on
the product. It yielded the smallest ratio of maximum to minimum temperatures, which was 1.18.
The tunnel oven in the traditional design was limited by its maximum radiation intensity, due
to the issue of non-uniform heat transfer. The distance between the bread and the burner could
significantly impact bread quality. The new design of the burner would allow higher radiative heat
flux intensity per unit of oven length. Therefore, a shorter oven length could be designed to save
FIGURE 9.15 Schematic diagram of an indirect gas-fired burner for a tunnel baking oven, which was
designed through CFD modeling and simulation. (Modified from Williamson, E. and Wilson, DI., J. Food
Eng., 91, 64–71, 2009.)
capital cost, reduce space requirements and reduce heat loss through the oven walls. Based on the
new burner design, radiant heat was transferred more uniformly to the top of travelling trays across
the width and the length of the tunnel oven.
9.4.7 Design of Operational Condition for Radiant Tube of a Burner

A three-dimensional CFD model was developed in [65] to study the heat transfer and combustion
phenomena in a radiant tube and thereby provide key information for designing a U-shape radiant
tube for a gas-fired burner (22 kW). The U-shape tube had a length of 460 cm, an inner diameter
of 7 cm and a distance of 12 cm between the two legs. Inside the heating tube, air and fuel were
mixed. Outside the heater, there was a reflector placed above it to reflect the radiant heat. During
combustion, when the temperature of the exhaust gas was higher than that of the inlet air and fuel,
the exhaust gas expanded rapidly. The velocity inside the heating tube then increased. The area
around 30 cm from the flame provided the highest radiant heat.
The effect of air factors (AF) such as the air/fuel ratio on the efficiency of heat transfer was stud-
ied. According to CFD simulation results, decreasing excess air could increase heating efficiency.
However, the effect of excess air became insignificant when the preheating temperature was high
[65]. AF is defined in Equation (9.13).
Air factor ( AF ) = 1 + [excess air present /100] (9.13)
Operating at lower AF (around 1.1) led to a higher heat flux, due to the lower amount of excess
air as carry gas in the heating tube. The magnitude of velocity was thus not high. The travelling
distance of the hot gas to the inner tube’s wall from the entrance could be very long. Therefore, the
highest heat flux was observed at a long distance (120 cm) from the burner.
In order to design the optimum operational condition, the efficiency of heating should be deter-
mined with Equation (9.14).
Efficiency = heat radiated from tube/heat released from combustion (9.14)
From the calculation results, the heater with a lower AF would have a higher efficiency than that
with a higher AF (around 1.5). However, the pre-heating temperature could be another factor in the
efficiency. An increase in the air inlet temperature from 27 to 377°C could improve the efficiency
from around 60 to above 70, in the case of AF 1.1. However, the effect of AF was reduced when the
inlet air was pre-heated. In fact, by pre-heating the inlet air to 377°C, the efficiency of heating at AF
1.1 was not significantly different from that at AF1.5.
9.4.8 Design of an Air Impingement Baking Oven

Previous studies [66,67] reported that energy waste during baking was as high as 46% of the heat
supply. This large amount of energy waste motivates the baking industry to improve their oven
design. The commercial baking ovens can be categorized into two major types: indirect fired ovens
(radiation) and direct fired ovens (forced convection) ovens. Earlier in this chapter, investigations on
some indirect fired baking ovens have been described. Configurations of burners and their radiative
tubes and process controllers were re-designed to increase the uniformity of heat distribution and
the heating efficiency of the oven. In this section, a different design of baking oven proposed by
Khatir et al. [68] is described. Instead of supplying energy by radiative heating, the proposed oven
design heats up mainly by forced convection, a process known as “air impingement.” CFD model-
ing and simulation were used to estimate its energy efficiency in accordance with the oven design.
The air impingement baking oven used air as a medium to transfer heat to the product. The sys-
tem was designed to increase airflow velocity and rapidly generate forced convective heat. The con-
vective heat transfer coefficient was expected to be high. Oven configuration (Figure 9.16), including
nozzle dimension and nozzle-to-surface distance, was investigated through the CFD approach.
Based on the CFD simulation results, only a small temperature variation of 0.04°C was detected.
This allows the bread to be baked efficiently. With uniform heat distribution in the air impingement
FIGURE 9.16 Schematic diagram of an air impingement baking oven. (Modified from Khatir, Z., Paton, J.,
Thompson, H., Kapur, N., and Toropov, V., Applied Energy, 112, 918–927, 2013.)
oven, a 10% energy reduction could be achieved. The total heat flux in the oven was composed of
93% convection and 7% radiation coming from oven walls. The convective heat flux varied, depend-
ing on the distance from the nozzles. In the area underneath the nozzle outlets, the convective heat
flux was at its maximum. The minimum convective heat flux was observed at the midpoint between
two nozzles. In contrast, the radiative heat flux remained relatively constant for all the distances
from the nozzles. Therefore, the total heat flux underneath the nozzles was up to three times larger
than the area away from the nozzles. Due to the high ratio of convection in the total heat flux, the
distance of bread surface to the nozzle outlets should be reduced to increase the heating rate at the
dough center. Thus, the baking time could be shortened.
In addition, the oven should be designed to provide the optimum heat transfer coefficient of
35 W/m2K. Then, another 5% energy reduction could potentially be achieved. With a heat transfer
coefficient of 35 W/m2K, the baking time (time when the core temperature reaches 94°C) of typical
commercial bread would be around 24 min. The baking time could only be slightly further reduced
(still more than 20 min) when the heat transfer coefficient was notably increased to 100 W/m2K.
Moreover, when the heat transfer coefficient increased to above 35 W/m2K, the requirement of
the specific electrical energy for the convective fan was dramatically increased. Although it could
slightly save more energy due to reduced heat loss to ambient, the total energy requirement was
extremely high, at about 400 kJ/kg when the heat transfer coefficient was 100 W/m2K [68].
9.5 CONCLUSIONS
Baking is a complex process due to the many thermal reactions involved, including starch gelatiniza-
tion and non-enzymatic browning reactions, which are coupled with heat and mass transportation.
To obtain bakery products of highest quality, temperature profiles should be optimized in accor-
dance with the required quality attributes. However, it is challenging to operate an industrial con-
tinuous baking oven to achieve the optimum temperature profile, because of the complexity of the
oven. Both heat supply to and airflow in the oven have to be adjusted. Adjusting the oven operation
through an online trial-and-error process would not be practical, as it is very time-consuming and
costly. This chapter has presented the capability of CFD modeling and simulation for optimizing an
industrial continuous travelling-tray bread-baking oven. A three-dimensional CFD model with mov-
ing grids was established to specifically simulate the tin temperature profiles and airflow patterns
during the continuous baking process. The CFD model demonstrated reasonably good performance
via model validation, where the predicted temperature and airflow profiles were compared to the
measured ones during industrial operations. The model was subsequently used to estimate the effect
of oven load on airflow pattern, temperature profiles, and the corresponding bread quality attributes.
The CFD simulation results demonstrated a dynamic change in the tin temperature profiles at the
beginning of the baking process. However, the change became non-significant when the oven was
filled with more trays of dough. By integrating the CFD model with other mathematical models
established for quality attributes, variations in the weight loss and crust color of bread among dif-
ferent trays and among different tins within the same tray could be estimated. The CFD model was
also used to study how the oven operating condition should be adjusted to achieve the optimum tin
temperature profile. Based on the prediction, the weight loss and energy consumption could both be
reduced, while the product quality attributes were maintained in an acceptable range. Information
obtained from CFD modeling and simulation was further demonstrated to be very useful in modify-
ing the oven configuration and designing the control system for industrial continuous baking ovens.
NOMENCLATURE
C Specific heat (J kg–1 °C–1)
T Temperature
h Convective heat transfer coefficient (W m–2 °C–1)
k Thermal conductivity (W m–2 °C–1)

q Heat flux (W m–2)
x Distance (m)
α Degree of gelatinization
σ Stefan–Boltzmann constant
ε Emissivity
REFERENCES
1. Pyler E.J., Baking Science & Technology, Vol. II, 3rd ed., Sosland Publishing Company, USA, 1346,
1988.
2. Swortfiguer M.J., Dough absorption and moisture retention in bread, The Bakers Digest, 42(4): 42–44,
1968.
3. Anonymous, Heat flow, BRI Technical Bulletin, March 2002.
4. Eliasson A.-C., Larsson K., Cereals in Breadmaking: A Molecular Colloidal Approach, Marcel Dekker,
5. Cauvain S., Young L., Bread, In: Baking Problems Solved, Cauvain S., Young L., Eds., Woodhead
Publishing, Sawston, Cambridgeshire, UK, 81–109, 2001.
6. Thorvaldsson K., Skjoldebrand C., Water diffusion in bread during baking, Lebensmittel-Wissenschaft
und-Technologie, 31(7–8): 658–663, 1998.
7. Van Son M., Flux future, Asia-Pacific Baker, 16–22, 2001.
8. Fahloul D., Trystram G., McFarlane I., Duquenoy A., Measurement and predictive modeling of heat
fluxes in continuous baking ovens, Journal of Food Engineering, 26(4): 469–479, 1995.
9. Carvalho M.G., Nogueira M., Improvement of energy efficiency in glass-melting furnaces, cement kilns
and baking ovens, Applied Thermal Engineering, 17(8–10): 921–933, 1997.
10. Velthuis H., Dalhuijsen A., De Vries U., Baking ovens and product, Food Technology International
Europe, 61–66, 1993.
11. Sato H., Matsumura T., Shibukawa S., Apparent heat transfer in a forced convection oven and properties
of baked food, Journal of Food Science, 52(1): 185–188, 193, 1987.
12. Wahlby U., Food quality in a low temperature convection oven—Experiments and modilisation during
development of a new household oven, In: Proceedings of ACoFoP IV, Goteborg, Sweden, 479–484, 1998.
13. Sumnu G., A review on microwave baking of foods, International Journal of Food Science &
Technology, 36(2): 117–127, 2001.
14. Verboven P., Datta A.K., Anh N.T., Scheerlinck N., Nicolai B.M., Computation of airflow effects on heat
and mass transfer in a microwave oven, Journal of Food Engineering, 59(2–3): 181–190, 2003.
15. Chevallier S., Della G., Colonna P., Broyart B., Trystram G., Structural and chemical modifications of
short dough during baking, Journal of Cereal Science, 35(1): 1–10, 2002.
16. Zanoni B., Peri C., Pierucci S., A study of the bread-baking process I: A phenomenological model,
17. De Vries U., Sluimer P., Bloksma A.H., A quantitative model for heat transport in dough and crumb dur-
ing baking, In: Cereal Science and Technology in Sweden, Asp N.-G., Ed., STU Lund University, Lund,
174–188, 1989.
18. Matz S., Equipment for Bakers, Elsevier, Amsterdam, Netherlands, 475, 1989.
19. Gupta T.R., Individual heat transfer modes during contact baking of Indian unleavened fat bread
(chapati) in a continuous oven, Journal of Food Engineering, 47(4): 313–319, 2001.
20. Carvalho M.G., Mertins N., Mathematical modeling of heat and mass transfer phenomena in bak-
ing ovens, In: Computational Methods and Experiment Measurements V, Sousa A., Brebbia C.A.,
Carlomagno G.M., Eds., Computational Mechanics Publications, 359–370, 1991.
21. De Vries U., Velthuis H., Koster K., Baking ovens and product quality—A computer model, Food
Science and Technology Today, 9(4): 232–234, 1995.
22. Noel J.Y., Ovenden N.A., Pochini I., Prediction of flow and temperature distribution in a domestic forced
convection electric oven, In: Proceedings of ACoFoP IV, Goteborg, Sweden, 491–496, 1998.
23. Tong C.H., Lund D.B., Microwave heating of baked dough products with simultaneous heat and mois-
ture transfer, Journal of Food Engineering, 19(4): 319–339, 1993.
24. Zanoni B., Pierucci S., Peri C., Study of bread baking process-II. Mathematical modeling, Journal of
25. Thorvaldsson K., Janestad H., A model for simultaneous heat, water and vapour diffusion, Journal of
26. Sablani S.S., Marcotte M., Baik O.D., Castaigne F., Modeling of simultaneous heat and water transport
in the baking process, Lebensmittel-Wissenschaft und-Technologie, 31(3): 201–209, 1998.
27. Czuchajowska Z., Pomeranz Y., Jeffers H.C., Water activity and moisture content of dough and bread,
Cereal Chemistry, 66(2): 128–132, 1989.
28. Larsen N.G., Greenwood D.R., Water addition and the physical properties of mechanical dough develop-
ment doughs and breads, Journal of Cereal Science, 13(2): 195–205, 1991.
29. Hall J.E., Bravington V., Otto S.R., Simulation of the transition from dough to bread, In: Proceedings of
ACoFoP IV, Goteborg, Sweden, 503–509, 1998.
30. Wahlby U., Skjoldebrand C., Reheating characteristics of crust formed on buns, and crust formation,
31. Wong S.Y., Zhou W., Hua J., Robustness analysis of a CFD model to the uncertainties in its physical
properties for a bread baking process, Journal of Food Engineering, 77: 784–791, 2006.
32. Bloksma A.H., Rheological aspects of structural changes during baking, In: Chemistry and Physics of
Baking, Blanshard J.M., Frazier P.J., Galliard T., Eds., The Royal Society of Chemistry, Great Britain,
170–178, 1986.
33. Fan T., Mitchell J.R., Blanshard J.M.V., A model for the oven rise of dough during baking, Journal of
34. Medcalf D.G., Wheat starch properties and their effect on bread baking quality, The Bakers Digest,
42(4): 48–52, 1968.
35. Engelsen S.B., Jensen M.K., Pedersen H.T., Norgaard L., Munck L., NMR-baking and multivariate
prediction of instrumental texture parameters in bread, Journal of Cereal Science, 33(1): 59–69,
2001.
36. Hansen L.P., Johnston P.H., Ferrel R.E., Heat-moisture effects on wheat four, I. Physical-chemical
changes of four proteins resulting from thermal processing, Cereal Chemistry, 52(4): 459–472, 1975.
37. Cuq B., Boutrot F., Redl A., Lullien-Pellerin V., Study of the temperature effect on the formation of
wheat gluten network: Influence on mechanical properties and protein solubility, Journal of Agricultural
and Food Chemistry, 48(7): 2954–2959, 2000.
38. Marston P.E., Wannan T.L., Bread baking-the transformation from dough to bread, The Bakers Digest,
50(4): 24–28, 49, 1976.
39. Bushuk W., Distribution of water in dough and bread, The Bakers Digest, 40: 38–40, 1966.
40. Gray J.A., Bemiller J.N., Bread staling: Molecular basis and control, Comprehensive Reviews in Food
Science and Food Safety, 2: 1–21, 2003.
41. Gogus F., Duzdemir C., Eren S., Effects of some hydrocolloids and water activity on nonenzymatic
browning of concentrated orange juice, Nahrung, 44(6): 438–442, 2000.
42. Ramirez-Jimenez A., Guerra-Hernandez E., Garcia-Villanova B., Browning indicators in bread, Journal
of Agricultural and Food Chemistry, 48(9): 4176–4181, 2000.
43. Zanoni B., Peri C., Bruno D., Modeling of browning kinetics of bread crust during baking, Lebensmittel-
Wissenschaft und-Technologie, 28(6): 604–609, 1995.
44. Moyano P.C., Rioseco V.K., Gonzalez P.A., Kinetics of crust colour changes during deep-fat frying of
impregnated french fries, Journal of Food Engineering, 54(3): 249–255, 2002.
45. Ibarz A., Pagan J., Garza S., Kinetic models of non-enzymatic browning in apple puree, Journal of the
Science of Food and Agriculture, 80(8): 1162–1168, 2000.
46. Trystram G., Computerized process control for the bakery/cereal industry, In: Computerized Control
Systems in the Food Industry, Mittal G.S., Ed., Marcel Dekker, New York, USA, 491–512, 1997.
47. Kim S., Cho S.I., Neural network modeling and fuzzy control simulation for bread-baking process,
Transactions of the ASAE, 40(3): 671–676, 1997.
48. Kohl D., Heinert L., Bock J., Hofmann Th., Schieberle P., Gas sensors for food aroma during baking
and roasting processes based on selective odorant measurements by an array (HRGC/SOMMSA), Thin
Solid Films, 391(2): 303–307, 2001.
49. Fahloul D., Trystram G., Duquenoy A., Kinetics of biscuit colour in continuous baking ovens and cor-
relations with heat fluxes, In: Proceedings of the IUFOST 2006, Nantes, 2006.
50. Xue J., Walker C.E., Humidity change and its effect on baking in an electrically heated air jet impinge-
ment oven, Food Research International, 36: 561–569, 2003.
51. Therdthai N., Zhou W., Adamczak T., The development of an anemometer for industrial bread baking,
52. Gardiner W.P., Statistical Analysis Methods for Chemists—A Software-Based Approach, 1st ed.
Cambridge, UK, 263–280, 1997.
53. Singh R.P., Heldman D.R., Introduction to Food Engineering, 3rd ed., Food Science and Technology,
Academic Press, Cambridge, MA, USA, 331, 2001.
54. Bouleta M., Marcosa B., Moresoli C., Dostie M., Sequential inverse method implemented into CFD
software for the estimation of a radiation boundary, International Journal of Thermal Sciences, 51:
7–15, 2012.
55. Hoang M.L., Verboven P., De Baerdemaeker J., Nicolai B.M., Analysis of the air flow in a cold store by
means of computational fluid dynamics, International Journal of Refrigeration, 23(2): 127–140, 2000.
56. Mirade P.S., Daudin J.D., Numerical simulation and validation of the air velocity field in a meat chiller,
International Journal of Applied Science and Computations, 5(1): 11–24, 1998.
57. Harral B.B., Boon C.R., Comparison of predicted and measured air flow patterns in a mechanically
ventilated livestock building without animal, Journal of Agricultural Engineering Research, 66(3):
221–228, 1997.
58. Therdthai N., Zhou W., Adamczak T., Optimisation of temperature profile in bread baking, Journal of
59. Therdthai N., Zhou W., Adamczak T., Three-dimensional CFD modeling and simulation of the tem-
perature profiles and airflow patterns during a continuous industrial baking process, Journal of Food
Engineering, 65(4): 599–608, 2004.
60. Zanoni B., Peri C., Bruno D., Modeling of starch gelatinisation kinetics of bread crumb during baking,
Lebensmittel-Wissenschaft und-Technologie, 28(3): 314–318, 1995.
61. Therdthai N., Zhou W., Adamczak T., Simulation of starch gelatinisation during baking in a travelling-tray
oven by integrating a three-dimensional CFD model with a kinetic model, Journal of Food Engineering,
65(4): 543–550, 2004.
62. Therdthai N., Zhou W., Adamczak T., Two-dimensional CFD modeling and simulation of an industrial
continuous bread baking oven, Journal of Food Engineering, 60(2): 211–217, 2003.
63. Wong S.Y., Zhou W., Hua J., Designing process controller for a continuous bread baking process based
on CFD modeling, Journal of Food Engineering, 81: 523–534, 2007.
64. Williamson M.E., Wilson D.I., Development of an improved heating system for industrial tunnel baking
ovens, Journal of Food Engineering, 91: 64–71, 2009.
65. Ahanj M.D., Rahimi M., Alsairaf A.A., CFD modeling of a radiant tube heater, International
Communications in Heat and Mass Transfer, 39: 432–438, 2012.
66. Probert D., Newborough M., Designs, thermal performances and other factors concerning cooking
equipment and associated facilities, Applied Energy, 21: 81–222, 1985.
67. Paton J.B., Khatir Z., Thompson H.M., Kapur N., Toropov V.V., Thermal energy management in the
bread-baking industry using a system modeling approach, Applied Thermal Engineering, 53(2): 340–
347, 2013.
68. Khatir Z., Paton J., Thompson H., Kapur N., Toropov V., Optimisation of the energy efficiency of bread-
baking ovens using a combined experimental and computational approach, Applied Energy, 112: 918–
927, 2013.
10 Improving Bread-Baking
Process Under Different
Oven Load Conditions
by CFD Modeling
N. Chhanwal, J. A. Moses, and C. Anandharamakrishnan
CONTENTS
10.1 Introduction..................................................................................................................... 225
10.2 Overview of Bread Making Process.................................................................................... 226
10.3 CFD Modeling of Baking Process..................................................................................... 229
10.4 Radiation Models................................................................................................................. 230
10.4.1 Discrete Transfer Radiation Model (DTRM)...................................................... 230
10.4.2 Surface to Surface (S2S)........................................................................................... 231
10.4.3 Discrete Ordinate (DO).......................................................................................... 231
10.5 Velocity and Temperature Profile of Oven............................................................................ 231
10.6 Evaporation Condensation Mechanism................................................................................. 232
10.7 Modeling of Starch Gelatinization and Bread Softness...................................................235
10.8 Modeling Browning of Bread Surface................................................................................. 238
10.9 Conclusions..................................................................................................................... 239
References....................................................................................................................................... 239
10.1 INTRODUCTION
Bread is a basic dietary item for a vast portion of the global population. Though there are differ-
ent variants, the baking process is common to all. The process of baking breads dates back to the
Neolithic age and it is believed that the science was first mastered by Egyptians. Advancement
in technology and changing lifestyles demand energy-efficient baking process. The success and
growth of any industry depend on good quality products and energy-efficient processes. Baking is
one of the most heat sensitive process in which heating has shown significant influence on quality
(chemical, physical and sensory) and cost of the final product [1,2]. The baking process involves
several operations like mixing, proving, baking, cooling, etc. These processes have a specific role
in the final product quality [3]. Operating conditions of the baking process have an influence on
key quality parameters of bread (texture, moisture content, surface color and geometry of bread).
Product load in the oven also influences the heat transfer rate and eventually affects the product
quality [2,4].
Ovens are the integral part of the baking process and serve as the energy source that leads to heat
and mass transfer in the product. Baking ovens play an important role in deciding the final product
quality of any baking process [4,5]. The success or failure of any baking venture is determined by
the oven and its operating conditions. Ovens may be classified based on scale (size and capacity),
product being baked (bread or biscuit, etc.), physical arrangements (batch or continuous), heating
source (electrical or hot air heating) and mode of heating (conductive or convective). Batch baking
225
ovens are mainly used in the small-scale industry or pilot plants and at domestic level. In batch-
baking process, electrical heating ovens are widely used because of their adaptability for different
bakery products. The most commonly used continuous bread-baking ovens are travelling-tray ovens
and rotary rack ovens. Continuous ovens are more energy-efficient and cost-effective compared to
batch-baking ovens [2,4].
Heat distribution inside the oven chambers depends on the location of the heating source, air
flow, product load, vent position, placement of the bread and baking time. The product quality var-
ies during the initial period of baking process due to the high temperature gradient between the hot
oven and product. The temperature rise due to the initial rate of heat absorption initiates a number
of physical mechanisms. Early crust formation, darkening of the crust, more water loss and less
volume expansion may be caused by very high initial temperature [2,4].
Bread baking is a complex process where various physiochemical and biological transformations
take place simultaneously, such as evaporation of water, gelatinization of starch, volume expansion,
crust formation, denaturation of proteins and browning reactions etc. [5–9]. Producing high qual-
ity bread is a great challenge in the baking process; as substandard product has to be discarded as
baking is a non-reversible process, and this could be economically unfavorable. Water content and
temperature are responsible for physiochemical and biological processes. Variation in color and
moisture content occur due to the non-uniform temperature distribution inside the oven. These
variations can be minimized by the proper design of the oven as well as maintaining proper process-
ing conditions, such as air temperature, heating power, baking time and bread size [4,5,9,10]. The
placement of the bread is also important, as it affects flow pattern of heat and thus affects various
transformation processes of bread baking. Moreover, electrical heating pilot-scale ovens are nor-
mally operated in a batch mode. In the continuous process, bread is exposed to different heating
cabinets due to U-turn movement, etc.; hence it gets heated uniformly. In electrical heating ovens,
bread is almost stationary throughout the baking process. Therefore, the placement of bread plays a
vital role in deciding the final product quality [4,10,11].
10.2 OVERVIEW OF BREAD MAKING PROCESS

In the text to avoid confusion, term ‘bread making’ refers to the entire bread-baking process (mix-
ing, fermentation, moulding, proofing, baking and cooling), whereas baking refers to the only bak-
ing operation carried out in the oven. To understand the effect of heat distribution during bread
baking process, it is important to first understand the entire bread making process and physico-
chemical transformation associated with it. The bread-making process is presented in Figure 10.1.
Bread making starts with the mixing process, in which all the ingredients (flour, water, yeast, fat
and salt) are mixed in the defined proportions. Precise choice of raw materials cannot be compro-
mised, as it has direct implications on baked bread quality. The proportion of different ingredients
varies based on the type of bread. Mixing is one of the critical stages, where incorporation of
gas and energy (rotational) transferred to the ingredients while dough formation. If mixing is not
done properly, the final product quality will be inferior or the product will have significant defects.
Mixing is carried out mechanically in an automated mixer with blades at variable mixing speed for
the optimized time. The transferred energy is utilized for gluten network formation, which is key
for the bread structure [12].
Fermentation is crucial for the generation of CO2 and dough expansion with the application of
yeast (Saccharomyces cerevisiae). The dough is allowed to ferment by yeast for 90 min at 30°C
and 75% relative humidity (RH). Gas generated during fermentation by yeast remains entrapped
inside the dough due to the gluten network and results in dough expansion. With time, the bread
is leavened with associated production of CO2. An important aspect of the leavening process is
that it is proper only if the dough has the ability to trap gas molecules within the system. This
implies that the baking process is a complex technology involving engineering of food structures.
Significant physical and chemical changes occur in the process of dough becoming bread. These
Improving Bread-Baking Process Under Different Oven Load Conditions by CFD Modeling 227
FIGURE 10.1 Outline of bread making process.
include changes in starch and protein matrices to effect visually prominent volumetric expansion
and development of the porous structure. Several researchers have attempted to understand these
series of changes and such findings are well documented in the literature. Shortening agents such as
ghee or margarine are often used to improve the machinability of the dough. In commercial bread-
baking processes, emulsifiers are added to (1) improve dough strength, (2) improve hydration behav-
ior, (3) improve slicing properties, (4) improve the structure of the developing crust, (5) improve
gas holding capacity and, (6) increase the shelf-life of the baked bread. Next the dough is punched,
rounded, weighed, moulded and placed in the pan. Proofing is carried out for 45–55 min at 30°C and
85% RH. Meanwhile, the oven has been preheated to the desired temperature (180–220°C). Later,
proofed dough in the pan is placed inside the preheated oven, and the baking process is carried out.
The major proceedings during the bread-baking process are summarized in Figure 10.2. Cooling
is the final step of the bread-making process, and it is important to avoid water vapor condensation
over the bread surface during this stage [3,4,12].
During the baking process, heat is transferred to the bread surface by all three modes (i.e., con-
duction, convection, and radiation). The heat arrived at the bread surface by different heating mode
is transported to the bread core by conduction and evaporation-condensation mechanism. The bread
surface hardens and crust forms due to very high bread surface temperature, and rapid evaporation
FIGURE 10.2 Major proceedings during the bread-baking process. (From Chhanwal, N., Tank, A.,
Raghavarao, K., Anandharamakrishnan, C., 2012. Computational fluid dynamics (CFD) modeling for bread
baking process—A review. Food and Bioprocess Technology, 5, 1157–1172.)
of water dehydrates the surface. Early crust formation during baking helps to achieve higher vol-
ume rise, and it can be achieved by employing high heat density during the initial baking period.
During the first stage of baking, due to the rapid increase of bread core temperature, oven spring is
observed. Oven spring occurs as gas produced due to the evaporation of water and carbon dioxide
release, pushes the crust. Later, crumb develops due to starch gelatinization and finally gradual
color development or surface browning occurs. Low moisture content and very high temperature
(>120°C) induces a non-enzymatic browning reaction, which results in the crust formation at the
surface of the bread [5,11,13,14]. Structural change also occurs during the bread-baking process, and
it comprises solidification and expansion. The network like the structure of bread crumb is the result
of protein denaturation and starch gelatinization. During gelatinization, starch granules hydrate
and swell. Starch gelatinization is typically used as the minimum baking index criteria. Under the
influence of temperature, gluten gel converts to cogel due to protein denaturation [5,15,16]. Moisture
loss is typical for any baked product due to evaporation of water, but it varies with the properties
of dough and baking conditions. Maximum moisture loss occurs at the surface of the bread while
moisture at the core region of bread remains unchanged due to evaporation and condensation of
water. The completion of the baking results in three different zones each with a unique texture,
crumb, crust and evaporation front (Figure 10.3) [2].
Crumb: Wet inner zone, where the temperature does not exceed 100°C and dehydration
a.
does not occur.
Crust: Dry outer zone, where temperature increases above 100°C and dehydration take place.
b.
Evaporation front: Embedded between the crumb and crust, where the temperature is
c.
100°C and water evaporates.
Apart from the effect of product variables, a range of process parameters also affects the baking
quality of bread. Importantly, the design of the oven, modes of heat transfer and process variables
during the baking process are critical deciders of product quality. They also affect overall produc-
tivity and process economics.
FIGURE 10.3 Schematic of bread structure.

10.3 CFD MODELING OF BAKING PROCESS

Apart from its applications in the baking process, CFD has been used in food industry mainly for
spray drying, refrigeration, cooling, sterilization and pasteurization. Due to the advancement in
computational resources and successful application to other food processing operations, CFD has
been applied intially for the design and development of baking ovens. Later, it was also applied
to understand the fluid flow and heat transfer during baking process [4,18–33]. Apart from the
computational power, increase in trained technical people and availability of user-friendly and
robust software such as Fluent, CFX, and COMSOL Multiphysics has boosted the CFD applica-
tions in the baking industry [4,18,33]. CFD is mainly used for the improvement of oven design and
optimizing the baking conditions. CFD techniques for the efficiency enhancement of ovens were
reviewed by Wong et al. [22]. Later, Chhanwal et al. [4] emphasized the application of CFD mod-
eling in the bread-baking process, including baking ovens and physico-chemical transformation
of bread. They also highlighted the strengths and limitations of CFD modeling concerning bread
baking process [1,4].
Figure 10.4 depicts the steps involved in the development of a CFD model for the bread-
baking process. A system (bread) was selected and simplified geometry (avoiding geometrical
complexities) of the desired system created using Gambit. Geometry was discretized into the
number of meshes using mesh elements. A selection of mesh elements for discretizing the
system was crucial to saving computational time and obtaining precise model predictions.
Meshed geometry was exported to the solver, desired models (e.g., radiation, turbulent flow)
were selected, thermo-physical properties and boundary conditions were applied and simula-
tion was carried out in transient or stationary mode. Simulation predictions were experimen-
tally validated for parameters such as temperature and velocity to define the accuracy of the
model. On model validation, desired parameters were extracted and plotted graphically for the
entire system [1,11,17].
Figure 10.5 depicts the pilot scale oven and one of the chambers used for the CFD modeling
study of the baking process.
Let us now discuss the radiative heat transfer equation and different radiation models used for
the bread-baking process.
FIGURE 10.4 Steps involved in CFD modeling of the bread-baking process.

FIGURE 10.5 Pilot scale electric heating oven (one chamber shown among eight) along with meshed oven
geometry and fully loaded oven.
10.4 RADIATION MODELS
The radiative transfer equation is given by [34],
 
dI ( r , s )
4π
  σT 4 σ    
+ (α + σ s ) I ( r , s ) = α n 2
ds π
+
4π ∫ I (r , s′ ) Φ ( s ⋅ s′ ) d Ω′ (10.1)
0

where I is radiation intensity (kJ/m2 rad), α is gas absorption coefficient (1/m), r is position vector,
 
s is direction vector, s′ is scattering direction vector, s is path length (m), n is refractive index, σS
is scattering coefficient, Φ is phase function, Ω’ is solid angle and σ is Stefan–Boltzmann constant
(W/m2 K4).
Although there are five radiation models available, but mainly three radiations models (i.e.,
DTRM, S2S and DO) were applicable for the baking oven and hence discussed below.
10.4.1 Discrete Transfer Radiation Model (DTRM)

The DTRM model assumes that the radiation leaving the surface element in a particular range of a
solid angle can be approximated by a single ray. The energy source in the fluid due to the radiation
is then computed by summing the change in intensity along the path of each ray that traced through
the control fluid volume. The equation for a change of radiant intensity can written as
dI ασ T 4
+αI = (10.2)
ds π
where, I is radiation intensity (kJ/m2 rad), α is gas absorption coefficient (1/m), s is path length (m),
and σ is Stefan–Boltzmann constant (W/m2 K4).
The DTRM ray tracing technique predicts the radiative heat transfer between surfaces without
calculating view factors. The accuracy of prediction of DTRM model mainly depends on the num-
ber of computational domains and rays traced.
10.4.2 Surface to Surface (S2S)

The S2S radiation model considers only the surface-to-surface radiation and neglects all absorption,
emission or scattering of radiations. It is mostly used for the radiative heat transfer in gray-diffusive
surfaces. In gray-diffusive surfaces, emissivity and absorptivity are independent of the radiation
wavelength. View factors account for radiative heat transfer. The energy flux leaving a surface is
qout . = εσ T 4 + ρqin. (10.3)

where qout. is energy flux leaving the surface (W/m2), qin. is energy flux incident on the surface from
surrounding (W/m2), ε is emissivity of surface and ρ is density (kg/m3).
10.4.3 Discrete Ordinate (DO)

The DO radiation model solves the radiative transfer equation for a finite number of discrete solid
angles, each linked with a vector direction fixed in the global Cartesian system. The DO model
solves radiative transfer equations in all directions. The DO model has the following equation
     σT 4
∇.[ I (r , s )s ] + (α + α s ) I (r , s ) = α n 2 (10.4)
π
  
where r is the position vector, s is direction vector, s′ is scattering direction vector, n is the refrac-
tive index and αs is absorption coefficient of surface.
10.5 VELOCITY AND TEMPERATURE PROFILE OF OVEN

Tank et al. [24] studied the velocity and temperature profile of a baking oven with fully loaded
(eight trays) and partially loaded configuration (one or two trays), depicted in Figures 10.6
and 10.7, respectively. Figure 10.6(a–c) shows that non-uniform velocity and swirl flows were
observed in the case of the partially loaded oven, whereas more uniform velocity was obtained
for the fully loaded oven. The temperature gradient creates natural convection, which changes
the velocity of the air flow and consequently affects heat distribution inside the oven. The non-
uniform velocity of air in the fully loaded oven was due to more number of trays placed. In
the partially loaded oven, less velocity was observed near the trays, heating source and buns,
whereas in the case of the fully loaded oven, it was only observed near the trays and heating
source [24].
The temperature profile of the partially loaded oven (Figure 10.7a) shows a higher tempera-
ture in the top portion of the oven compared to the bottom due to natural convection and the
presence of a vent at the bottom. Similar observations were reported by Anishaparvin et al.
[25] earlier.
Non-uniform heat distribution was observed inside the oven; higher temperatures were noted
at the extreme left near the oven wall and at the right central portion compared to the remaining
portion of the oven. The arrangement of the heating source (heating coils) covering an uneven area
within the oven cavity and the location of vent may be the main causes. The temperature profile of
the partially (diagonally loaded bun tray) loaded oven was shown in Figure 10.7(b), which shows
a different temperature profile than Figure 10.7(a) due to diagonal placement of the product tray.
Moreover, the temperature gradient was less between the top and bottom portion of oven with
diagonal placement of the product tray compared to the oven with a single tray. Increasing the oven
load lowered the temperature inside the fully loaded oven compared to the partially loaded oven
(Figure 10.7). Thus, loading of more buns affects heat distribution inside the oven, which eventually
increases the baking time [24].
FIGURE 10.6 Velocity profile of oven (a) partially (one tray) loaded, (b) partially (two trays) loaded and
(c) fully loaded at 300 s of baking process. (From Tank, A., Chhanwal, N., Indrani, D., Anandharamakrishnan,
C., 2014. Computational fluid dynamics modeling of bun baking process under different oven load conditions.
Journal of Food Science and Technology, 51, 2030–2037.)
10.6 EVAPORATION CONDENSATION MECHANISM

The bread-baking process was categorized as simultaneous heat and mass transfer (SHMT) pro-
cesses. Earlier CFD modeling studies [10,23,25] show a discrepancy in bread center temperature
predicted by the CFD model and experimentally measured. However, the bread surface temperature
predictions were closer to experimental observations, while center temperature predictions showed
large discrepancies. Moreover, the predicted bread center temperature crosses 100°C in the above
studies while in the real bread-baking process it never crosses 98–99°C due to the evaporation-
condensation mechanism. This was the major drawback of the earlier CFD modeling studies; they
fail to predict the correct bread core temperature.
Variation in the thermo-physical properties of foods is the major hurdle during computational
model development for any food processing operation. This holds true for baking; it is a tedious
task to develop a numerical model. Zanoni et al. [35–37] first used variable thermo-physical prop-
erties for dough during the bread-baking process and later Purlis et al. [6–8] and Purlis [38,39]
incorporated them into evaporation-condensation mechanism. Chhanwal et al. [18] used the
evaporation-condensation mechanism to predict actual bread crumb temperature by including in
FIGURE 10.7 Temperature profile of oven at 300 s: with (a) partially (one tray) loaded, (b) partially
(two trays) loaded and (c) fully loaded oven of baking process. (From Tank, A., Chhanwal, N., Indrani, D.,
Anandharamakrishnan, C., 2014. Computational fluid dynamics modeling of bun baking process under dif-
ferent oven load conditions. Journal of Food Science and Technology, 51, 2030–2037.)
the CFD model. In the evaporation-condensation mechanism, the specific heat of bread defined as a
function of temperature and enthalpy jump at the phase change is added. Also, the thermal conduc-
tivity of bread defined as a function of temperature [18]. Recently, Khatir et al. [26] overpredicted
bread core temperature due to the exclusion of the evaporation-condensation mechanism. Lucas
et al. [40] used evaporation-condensation-diffusion mechanism to model the bread-baking process.
The evaporation-condensation mechanism governs simultaneous heat and mass transfer (SHMT)
during the bread-baking process, as depicted in Figure 10.8. Rapid heat transfer from the surface to
the core of the bread occurs due to the evaporation-condensation mechanism.
Steps involved in the evaporation-condensation mechanism are as follows:
1. On the warmer side of the gas cell, (1) water evaporates by absorbing the latent heat of
vaporization.
2. Due to the vapor concentration gradient inside the cell, water vapor migrates through the
gas phase inside the gas cell (1).
FIGURE 10.8 Simultaneous heat and mass transfer mechanism during the bread-baking process. (From
Chhanwal, N., Tank, A., Raghavarao, K., Anandharamakrishnan, C., 2012. Computational fluid dynamics
(CFD) modeling for bread baking process—A review. Food and Bioprocess Technology, 5, 1157–1172.)
3. Water vapor condenses, setting free its latent heat on the colder side of the gas cell (1).
4. Water diffuses through a dough membrane and heat is transferred by conduction to the
warmer region of the next gas cell (2). Processes reiterate in the gas cell (2).
During evaporation-condensation, when the gas phase becomes continuous, the fourth stage no
longer exists. Evaporation-condensation mechanism takes place until temperature gradient exists
and the temperature of the whole crumb reaches 100°C [4,6–10,39,41–43].
During CFD modeling of the bread-baking process, the bread was assumed solid with constant
density and moisture content [44,45]. All heat transfer modes, i.e., conduction, convection, and radi-
ation, were enabled for the bread. The evaporation-condensation mechanism was employed through
the thermo-physical properties specific heat and thermal conductivity as a function of temperature
and phase change. The specific heat of bread was given as a function of temperature including the
enthalpy jump at the phase change [7,35,36].
C p = (0.013T 2 − 0.217T + 4179)0.39 + (5T + 1390)(0.61) + λ (T − TV , ∇T ) (10.5)

where T is temperature (°C), λ is heat of phase change (J/m3), T V is phase change temperature and ∇T
is the temperature difference (100 ± 0.5°C) in which phase change occurs and it considered as 0.5°C.
Thermal conductivity increased until the bread temperature reached 100°C and dropped to a
constant, i.e., 0.2 W/m2 K, for the crust after the phase change [46]. The low thermal conductivity
of the crust region was due to low moisture content.
0.9
k (T ) = + 0.2 if T ≤ 100°C
1 + exp(–0.1(T-353.16)) (10.6)
k(T) = 0.2 if T ≥ 100°C

In this study [18] application of the evaporation-condensation mechanism was able to predict the
crumb temperature profile in a way that matched experimental observations. The difference in bread
FIGURE 10.9 CFD model predictions and experimental observations for bread center temperature profile
during the bread-baking process with and without the evaporation-condensation (E-C) mechanism. (From
Chhanwal, N., Indrani, D., Raghavarao, K.S.M.S., Anandharamakrishnan, C., 2011. Computational fluid
dynamics modeling of bread baking process. Food Research International, 44, 978–983.)
center temperature predictions with and without incorporation of the evaporation-condensation mech-
anism is shown in Figure 10.9. Moreover, this accurate temperature prediction helps to predict other
temperature-dependent properties like starch gelatinization and bread softness. This model is very
useful for predicting the bread temperature more accurately than the earlier CFD models published.
10.7 MODELING OF STARCH GELATINIZATION AND BREAD SOFTNESS

Gelatinization of starch causes disruption of the intermolecular hydrogen bonds, which maintain the
structural integrity of the granules, and an exudation of amylose. Differential scanning calorimetric
(DSC) and X-ray diffraction methods are used to measure the degree of starch gelatinization during
the bread-baking process. Starch gelatinization is a function of time/temperature and follows first order
kinetics [36,37,47]. Starch gelatinization of bread during the bread-baking process can be predicted by
integrating the kinetic model for starch gelatinization into the CFD model for the bread-baking process
[23,25,48]. Therdthai et al. [48] reported that the starch gelatinization progressed faster at the top and
bottom of bread compared to the left and right halves of the bread. Chhanwal et al. and Anishaparvin
et al. [23,25] also predicted the degree of starch gelatinization based on the kinetic model.
The degree of starch gelatinization is considered a minimum baking index to decide the bread
baking time during the industrial baking process. Starch gelatinization is a function of time/
temperature and follows first order kinetics [36,37,47]. Zanoni et al. [36,37] studied starch gela-
tinization using the differential scanning calorimetric (DSC) method and calculated the model
parameters, such as k0 and Ea, as shown in the following equations.
1− α g = exp(− kt ) (10.7)

where αg is the degree of starch gelatinization, k is the reaction rate constant and t is the time in
seconds. The reaction rate constant (k) in turn can be calculated using Arrhenius type equation,
 E 
k = ko exp  − a  (10.8)
 RT 

where k0 = 2.8 × 1018 (1/s) and Ea = 138 kJ/mol [37].

Similar to starch gelatinization, bread softness is also one of the essential quality features for
bread. Bread softness is defined as a function of glass transition temperature and starch gelatiniza-
tion [49]. However, starch gelatinization should be 0.3 for obtaining soft product and maximum
softness is achieved with complete starch gelatinization. Glass transition temperature (Tg) is influ-
enced by the water, starch and sugar content of the product [50]. It was determined based on the
non-linear regression formula given by Hadiyanto et al. [49].
Tg = 457.1 − 396.3(S/Z) − 853.2 ⋅ W + 716.7 ⋅ (S/Z) ⋅ W + 430.3 ⋅ (S/Z)2

(10.9)
+ 778.4 ⋅ W 2 − 1424.7(S/Z)2 ⋅ W 2

where (S/Z) is the ratio of sugar and starch and W is the water content.
Bread softness can be determined using the following equations [49]:
Softness = Softness (dTg ) × Softness (α ) (10.10)

0 If dTg < 0

Softness(dTg ) =  0.01 × dTg If 0 < dTg < 100 (10.11)

 1 If dTg > 100



0 If α (t ) < 0.3
Softness(α ) =  3 10 (10.12)
 − + × α (t ) If 0.3 < α (t ) ≤ 1
 7 7
where dTg is the difference between room temperature and glass transition temperature, and α (t) is
the degree of starch gelatinization at time t.
The temperature profile of bread at six different points (shown in the inset) was obtained and pre-
sented in Figure 10.10. Most of the points (1, 4, 5 and 6) were located in the transition zones where
FIGURE 10.10 The temperature profile of bread crumb at different locations during the baking process.
FIGURE 10.11 Starch gelatinization profile of bread during the baking process at various locations.
crumb and crust meets, and points 3 and 2 were located near bread center. These points were selected
to define the starch gelatinization and bread softness based on temperature. The temperatures of
points 1, 4, 5 and 6 show a rapid increase and increase beyond 100°C. However, the temperatures
of points 2 and 3 do not cross 100°C as they are located in the crumb region. The large temperature
gradient (40–50°C) between point 3 (center region) and point 6 (bottom region) observed during the
initial stage of the bread-baking process underlines the transition of heat from bread surface to core.
Figure 10.11 depicts the progression of starch gelatinization for points 1–6 (derived from the
temperature depicted in Figure 10.10). Figure 10.11 indicates that complete gelatinization for point 3
occurs at 780 s and for point 4 at 300 s of the baking process. The softness of the bread is a function
of the degree of starch gelatinization and glass transition temperature. Initiation of softness develop-
ment occurs once the degree of starch gelatinization reaches 0.3. Figure 10.12 shows the initiation
of bread softness started at 180 s for point 4 and reached its peak value (0.73) at 360 s. Similarly, for
point 3, it starts at 600 s and completes at 840 s. In the present study, bread softness was predicted
based on temperature, assuming constant moisture content (0.39) throughout the bread. Therefore,
FIGURE 10.12 Bread softness profile during the bread-baking process at various locations.
the softness prediction holds true for the crumb portion, where moisture content remains 0.39 for
the entire bread-baking process. Therefore, a computational model including moisture content may
predict softness profile more precisely.
10.8 MODELING BROWNING OF BREAD SURFACE

Surface browning is an important quality parameter of bread which occurs due to low water activity
(0.4–0.8) and high surface temperature (<120°C). Excess browning is undesirable, as it affects flavor
and appearance of bread as well as the formation of acrylamide, which is a carcinogenic substance
[51,52]. Therefore, evaluation of the browning index of the bread surface is essential for optimizing
the baking process. Many researchers have proposed various models based on first-order kinetics
for the determination of browning of the bread surface [8,36,38,53]. Zhang and Datta [51] predicted
surface browning during the baking process. Chhanwal et al. [18] predicted surface browning by
integrating a kinetic model developed by Zanoni et al. [36] in a CFD model. In this study, the model
developed by Zanoni et al. [37] was adopted as this model based on bread surface temperature and
baking time. The model equation is as follows:

∆E∞ − ∆Et = (∆E∞ − ∆Et − ∆t ) exp(− kb ∆t ) (10.13)
From Arrhenius equation, the reaction rate constant (kb) was determined.
 E 
kb = kb 0 exp  − b  (10.14)
 RT 

where, kb0 = 42,000 s−1, Eb = 64,151 J/mol and ΔE∞– = 52.

The surface temperature of the bread at various locations (shown in inset) during the baking pro-
cess was obtained and presented in Figure 10.13. During the initial stage of bread baking, a sudden
temperature rise (90–100°C) was observed at the bread surface. Later, a linear increase during the
rest of the baking period was followed. Bread surfaces crossed 100°C within the first two minutes of
the baking process, which was responsible for the crust formation. At the end of the baking process,
the bread surface attained oven temperature (≈220°C). However, points located at the right corners
(points 1, 2, 3 and 4) of the bread indicate high temperature compared to the left corners (points 5,
6, 7 and 8). Non-uniform temperature distribution around the bread was responsible for this.
FIGURE 10.13 The temperature profile of the bread surface at various locations during the baking process.
FIGURE 10.14 Browning index of the bread surface at various locations during the baking process.
The browning index across the bread surface was obtained by integrating a kinetic model for
browning temperature predictions at various points and presented in Figure 10.14. As the surface
temperature of the bread increases, browning also increases. The browning index reached the satu-
ration value (i.e., constant value) towards the end of the baking process. The degree of browning at
points 2, 3 and 4 was rapid compared to the other points, which were proportional to the tempera-
tures at the corresponding points (Figure 10.13). Zanoni et al. [36] reported the maximum browning
index (dark brown charred bread) and minimum browning indexes were 52 and 30, respectively.
The browning indexes of all points fall within this range during the complete baking process.
10.9 CONCLUSIONS
To understand, predict and control the quality attributes of bread during the baking process, a com-
putational fluid dynamics (CFD) model was developed by incorporating evaporation-condensation
mechanism. The temperature profile for the oven and bread, and other quality parameters such as
starch gelatinization, softness and browning of bread surface were predicted based on bread tem-
perature. On the concluding of this chapter, CFD models help to understand the baking process
thoroughly in terms of various physico-chemical changes and heat distribution inside the oven.
REFERENCES
1. Anandharamakrishnan C., Computational Fluid Dynamics Applications in Food Processing, Springer-
Verlag, New York, USA, 2013.
2. Chhanwal N., Anandharamakrishnan C., Temperature and moisture based modeling for prediction of
starch gelatinization and crumb softness during bread-baking process, Journal of Texture Studies, 45:
462–476, 2014.
3. Cauvain S.P., Bread Making—Improving Quality, CRC Press, Boca Raton, FL, USA, 2003.
4. Chhanwal N., Tank A., Raghavarao K., Anandharamakrishnan C., Computational fluid dynamics (CFD)
modeling for bread baking process—A review, Food and Bioprocess Technology, 5: 1157–1172, 2012.
5. Zhou W., Therdthai N., Three-dimensional modeling of a continuous industrial baking process, In:
Computational Fluid Dynamics in Food Processing, Sun D.-W., Ed., CRC Press, Boca Raton, FL, USA,
287–312, 2007.
6. Purlis E., Salvadori V.O., Bread baking as a moving boundary problem. Part 1: Mathematical modeling,
Journal of Food Engineering, 91: 428–433, 2009a.
7. Purlis E., Salvadori V.O., Bread baking as a moving boundary problem. Part 2: Model validation and
numerical simulation, Journal of Food Engineering, 91: 434–442, 2009b.
8. Purlis E., Salvadori V.O., Modelling the browning of bread during baking, Food Research International,
42: 865–870, 2009c.
9. Therdthai N., Zhou W., Recent advances in the studies of bread baking process and their impact on the
bread baking technology, Food Science and Technology Research, 9: 219–226, 2003.
10. Wong S.Y., Zhou W., Hua J., CFD modeling of an industrial continuous bread-baking process involving
U-movement, Journal of Food Engineering, 78: 888–896, 2007a.
11. Zhou W., Hui Y.H., Bakery Products Science and Technology, John Wiley & Sons, Hoboken, NJ, USA,
2014.
12. Chhanwal N., Ezhilarasi P.N., Indrani D., Anandharamakrishnan C., Influence of electrical and hybrid
heating on bread quality during baking, Journal of Food Science and Technology, 52: 4467–4474, 2015.
13. Wagner M., Quellec S., Trystram G., Lucas T., MRI evaluation of local expansion in bread crumb during
baking, Journal of Cereal Science, 48: 213–223, 2008.
14. Vanin F.M., Michon C., Trystram G., Lucas T., Simulating the formation of bread crust in a DMTA
rheometer, Journal of Cereal Science, 51: 277–283, 2010.
15. Scanlon M.G., Zghal M.C., Bread properties and crumb structure, Food Research International, 34:
841–864, 2001.
16. Lucas T., Baking, In: Bakery Products Science and Technology, Zhou, W., Hui Y.H., Eds., John
Wiley & Sons, Hoboken, NJ, USA, 335–354, 2014.
17. Chhanwal N., Anandharamakrishnan C., Design and optimization of food processing equipments
using computational fluid dynamics modeling, In: Engineering Applications of Computational Fluid
Dynamics, Sadiq Al-Baghdadi M.A.R., Ed., International Energy & Environment Foundation, Najaf,
Iraq, 243–268, 2012.
18. Chhanwal N., Indrani D., Raghavarao K.S.M.S., Anandharamakrishnan C., Computational fluid
dynamics modeling of bread baking process, Food Research International, 44: 978–983, 2011.
19. De Vries U., Velthuis H., Koster K., Baking oven and product quality—A computer model, Food Science
and Technology Today, 9: 232–234, 1995.
20. Verboven P., Scheerlinck N., de Baerdemaeker J., Nicolai B.M., Computational fluid dynamics model-
ling and validation of the isothermal airflow in a forced convection oven, Journal of Food Engineering,
43: 41–53, 2000.
21. Mirade P.S., Daudin J.D., Ducept F., Trystram G., Clément J., Characterization and CFD modeling
of air temperature and velocity profiles in an industrial biscuit baking tunnel oven, Food Research
International, 37: 1031–1039, 2004.
22. Wong S.Y., Zhou W., Hua J., Improving the efficiency of food processing ovens by CFD techniques,
Food Manufacturing Efficiency, 1: 35–44, 2006b.
23. Chhanwal N., Anishaparvin A., Indrani D., Raghavarao K., Anandharamakrishnan C., Computational
fluid dynamics (CFD) modeling of an electrical heating oven for bread-baking process, Journal of Food
Engineering, 100: 452–460, 2010.
24. Tank A., Chhanwal N., Indrani D., Anandharamakrishnan C., Computational fluid dynamics modeling
of bun baking process under different oven load conditions, Journal of Food Science and Technology,
51: 2030–2037, 2014.
25. Anishaparvin A., Chhanwal N., Indrani D., Raghavarao K.S.M.S., Anandharamakrishnan C., An
investigation of bread baking process in a pilot-scale electrical heating oven using computational fluid
dynamics, Journal of Food Science, 75: E605–E611, 2010.
26. Khatir Z., Paton J., Thompson H., Kapur N., Toropov V., Lawes M., Kirk D., Computational fluid
dynamics (CFD) investigation of air flow and temperature distribution in a small scale bread-baking
oven, Applied Energy, 89: 89–96, 2012.
27. Khatir Z., Paton J., Thompson H., Kapur N., Toropov V., Optimisation of the energy efficiency of bread-
baking ovens using a combined experimental and computational approach, Applied Energy, 112: 918–
927, 2013a.
28. Khatir Z., Thompson H., Kapur N., Toropov V., Paton J., Multi-objective computational fluid dynamics
(CFD) design optimisation in commercial bread-baking, Applied Thermal Engineering, 60: 480–486,
2013b.
29. Paton J., Khatir Z., Thompson H., Kapur N., Toropov V., Thermal energy management in the bread bak-
ing industry using a system modelling approach, Applied Thermal Engineering, 53: 340–347, 2013.
30. Smolka J., Nowak A.J., Rybarz D., Improved 3-D temperature uniformity in a laboratory drying oven
based on experimentally validated CFD computations, Journal of Food Engineering, 97: 373–383,
2010.
31. Smolka J., Bulinski Z., Nowak A.J., The experimental validation of a CFD model for a heating oven
with natural air circulation, Applied Thermal Engineering, 54: 387–398, 2013.
32. Smolka J., Genetic algorithm shape optimisation of a natural air circulation heating oven based on an
experimentally validated 3-D CFD model, International Journal of Thermal Sciences, 71: 128–139,
2013.
33. Williamson M.E., Wilson D.I., Development of an improved heating system for industrial tunnel baking
ovens, Journal of Food Engineering, 91: 64–71, 2009.
34. Fluent 6.3., Fluent user’s guide, ANSYS Inc., Canonsburg, PA, USA, 2006.
35. Zanoni B., Petronio M., Effect of moisture and temperature on the specific heat of bread, International
Journal of Food Science, 3: 239–242, 1991.
36. Zanoni B., Peri C., Gianotti R., Determination of the thermal diffusivity of bread as a function of poros-
ity, Journal of Food Engineering, 26: 497–510, 1995a.
37. Zanoni B., Peri C., Bruno D., Modeling of starch gelatinization kinetics of bread crumb during baking,
LWT-Food Science and Technology, 28: 314–318, 1995b.
38. Purlis E., Browning development in bakery products—A review, Journal of Food Engineering, 99:
239–249, 2010.
39. Purlis E., Bread baking: Technological considerations based on process modeling and simulation,
Journal of Food Engineering, 103: 92–102, 2011.
40. Lucas T., Doursat C., Grenier D., Wagner M., Trystram G., Flick D., Modeling of bread baking with a
new, multi-scale formulation of evaporation-condensation-diffusion and evidence of compression in the
outskirts of the crumb, Journal of Food Engineering, 149: 24–37, 2015.
41. Thorvaldsson K., Janestad H., A model for simultaneous heat, water and vapour diffusion, Journal of
Food Engineering, 40: 167–172, 1999.
42. Wagner M.J., Lucas T., Le Ray D., Trystram G., Water transport in bread during baking, Journal of
43. Purlis E., Baking process design, In: Handbook of Food Process Design, Rahman A., Ed., Wiley-
Blackwell, Hoboken, NJ, USA, 743–768, 2012b.
44. Hamdami N., Monteau J.Y., Le Bail A., Heat and mass transfer in par-baked bread during freezing,
Food Research International, 37: 477–488, 2004a.
45. Hamdami N., Monteau J.Y., Le Bail A., Thermophysical properties evolution of French partly baked
bread during freezing, Food Research International, 37: 703–713, 2004b.
46. Rask C., Thermal properties of dough and bakery products: A review of published data, Journal of Food
Engineering, 9: 167–193, 1989.
47. Lund D.B., Influence of time, moisture, ingredients, and processing conditions on starch gelatinization,
Critical Reviews in Food Science and Nutrition, 20: 249–273, 1984.
48. Therdthai N., Zhou W., Adamczak T., Simulation of starch gelatinization during baking in a traveling—
Tray oven by integrating a three dimensional CFD model with a kinetic model, Journal of Food
Engineering, 65: 543–550, 2004a.
49. Hadiyanto, Asselman A., van Straten G., Boom R.M., Esveld D.C., van Boxtel A.J.B., Quality predic-
tion of bakery products in the initial phase of process design, Innovative Food Science and Emerging
Technologies, 8: 285–298, 2007.
50. Farhat I.A., Blanshard J.M.V., Descamps M., Mitchell J.R., Effect of sugars on retrogradation of waxy
maize starch-sugar extrudates, Cereal Chemistry, 77: 202–208, 2000.
51. Zhang J., Datta A.K., Mathematical modeling of bread baking process, Journal of Food Engineering,
75: 78–89, 2006.
52. Vanin F.M., Lucas T., Trystram G., Crust formation and its role during bread baking, Trends in Food
Science and Technology, 20: 333–343, 2009.
53. Purlis E., Salvadori V.O., Bread browning kinetics during baking, Journal of Food Engineering, 80:
1107–1115, 2007.
11 CFD Applications of Food
Packaging Sterilization and
Filling
Giuseppe Vignali and Filippo Ferrari
CONTENTS
11.1 Introduction............................................................................................................................244
11.2 Filling Technologies................................................................................................................ 244
11.2.1 Filling at Room Temperature and Post-Packaging Treatment................................. 244
11.2.2 Hot Filling............................................................................................................... 245
11.2.3 Aseptic Filling.. ............................................................................................245
11.2.4 Ultra Clean Filling...................................................................................................... 246
11.3 Food Packaging Sterilization............................................................................................... 247
11.3.1 Heat Sterilization...........................................................................................................247
11.3.1.1 Saturated Vapor......................................................................................... 247
11.3.1.2 Overheated Steam and Hot Air............................................................ 247
11.3.1.3 Heating due to Extrusion or Blowing........................................................ 248
11.3.2 Decontamination with Radiation............................................................................248
11.3.2.1 UV Rays...................................................................................................... 248
11.3.2.2 UV Rays Combined with Hydrogen Peroxide........................................248
11.3.2.3 Infrared Rays...........................................................................................248
11.3.2.4 Ionizing Radiation...................................................................................248
11.3.2.5 Pulsed Light................................................................................................ 249
11.3.3 Sterilization with Chemical Agents........................................................................ 249
11.3.3.1 Hydrogen Peroxide.................................................................................... 249
11.3.3.2 Peracetic Acid.......................................................................................... 250
11.3.3.3 Ozone...................................................................................................... 250
11.3.3.4 Ethanol.................................................................................................... 250
11.4 CFD Applications in Food Packaging Sterilization and Filling........................................... 250
11.4.1 Simulation and Analysis of Fluid Dynamic Behavior of Foods During
Filling Processes..................................................................................................... 250
11.4.2 Flow Investigation of the Product Fill Valve of Filling Machine for Packaging
Liquid Products....................................................................................................... 251
11.4.3 Simulation of the Filling of Polyethylene-Terephthalate Bottles (PET) with a
Volumetric Swirl Chamber Valve (VODM 40355) on the Basis of Calculation
Models and Experiments............................................................................................ 252
11.4.4 Improvement of BIB Packaging Product Filling Valve CIP Performance
and Efficiency................................................................................................ 254
11.4.5 CFD Analysis of Coffee Packaging in Capsules Using Gas Flushing Modified
Atmosphere Packaging......................................................................................... 255
11.4.6 Decontamination of Primary Packaging by Means of Atmospheric Plasmas....... 256
11.4.7 Modeling and Multi-Objective Optimization of the VHP Pouch Packaging
Sterilization Process............................................................................................ 257
11.5 Conclusions.......................................................................................................................... 258
References....................................................................................................................................... 259
243
11.1 INTRODUCTION
Computational Fluid Dynamics (CFD) methodology is used to predict what will happen in circum-
stances when it is not clear how fluids will flow. This is often coupled with the complications of
simultaneous flow of heat, mass transfer, phase changes (melting, freezing, boiling), chemical reac-
tions (combustion, rusting), mechanical movement of pistons, fans, etc. and stresses in and displace-
ment of immersed or surrounding solids. CFD uses a computer to solve the relevant science-based
mathematical equations, using information about the domain and boundaries in question.
The simulation process is usually performed in three stages: pre-processing, processing and
post-processing. In the pre-processing step, all the tasks that take place prior to the numerical
solution process are carried out. This includes problem solving, meshing and the generation of a
computational model. The processing phase involves the use of a computer to solve mathematical
fluid flow equations. This is a very intensive process and usually requires several thousands of
equations to be solved. In each case, boundary conditions are applied to the integrated equations. In
the post-processing step, an evaluation of the data generated by CFD analysis has to be made. When
the model is solved, the results can be analyzed both numerically and graphically [1].
In the food industry, CFD applications may allow a better understanding of complex physical
mechanisms. Since the CFD technique can be of great benefit to the food processing industry, accel-
erated development has taken place in the past few years. CFD, as a research tool for enhancing the
design process and understanding the basic physical nature of fluid dynamics, can provide benefits
for the food processing industry in many areas, such as drying, mixing, refrigeration, sterilization,
filling and other production phases.
In this chapter, articles that discuss packaging filling and sterilization will be presented.
The Scopus database was used to find articles published from 2007 to 2017 on these two themes.
Five articles were found on the filling process and two on the packaging sterilization process.
Literature on the latter topic was more limited as most of the articles consider sterilization of the
product inside the packaging as a secondary issue within the analysis, and thus do not take it into
account.
Before describing the application of CFD in this field of interest, it seemed appropriate to provide
a brief description of the two processes, highlighting the most important areas of application.
11.2 FILLING TECHNOLOGIES
11.2.1 Filling at Room Temperature and Post-Packaging Treatment
In this case, filling takes place at room temperature without special hygiene precautions, and
the product and container are then sterilized (or pasteurized) with a retort process inside a pres-
surized autoclave that circulates hot water or an alternative heating fluid. This is also known as
Appertification after Nicolas Appert, who in 1795 began to produce foods packed in sealed glass
jars and treated for six hours in boiling water.
This is the most effective process in terms of microbiological safety, but it also causes the great-
est heat damage to food. It is not possible to heat the product uniformly once it is inside the packag-
ing and also ensure that the treatment temperature is reached even at the most disadvantaged point,
which is usually at the center of the packaging. This involves long thermal transients both in the
heating and cooling stages and consequently involves a significant baking effect.
The parameters that influence this process are: the initial temperature of the product (usually the
higher the temperature, the shorter the transitional period), the viscosity of the product, the weight
of the product (especially in the case of liquid in the liquid state), the size of the container and the
temperature of the heating fluid. One particular disadvantage of this process is the lack of direct
control over the temperature of the product, unlike that of the heating fluid.
CFD Applications of Food Packaging Sterilization and Filling 245
At present, this type of conditioning is rarely used for beverages but is mainly applied to solid
and semi-solid products (such as vegetable preserves and animal products packaged in metal pack-
aging) or to products partially made up of liquid control.
This type of treatment provides products with high microbiological quality and a significant
shelf life of between 6–12 months.
11.2.2 Hot Filling
The hot filling process provides aseptic packaging for highly acidic products. It essentially involves
rapidly heating the food up to the treatment temperature and maintaining this temperature for the
time it takes to get the desired number of microorganisms’ decimals reduction. The heated food is
then transferred by means of an appropriate filling system into the containers, which may have been
preheated. At this point the packages are sealed and capped to ensure that the product comes into
contact with all the interior walls. In this case, the treatment of the packaging and the environment
is ensured by the residual heat delivered by the product during and after filling.
For processes of this kind, containers that can withstand high temperatures must obviously be
used. With regard to plastic containers, only those with a glass transition temperature significantly
higher than the filling temperature of the product, which is never lower than 65°C, can be used.
This type of process, although conceptually more complicated than previously illustrated, is
easier to carry out because it allows direct control over the process’s main variable, i.e., the prod-
uct’s treatment temperature.
Thermal damage, albeit less relevant than in the retort process, always results; although it is pos-
sible to guarantee rapid warming of the product (which is treated in bulk), the cooling phase is less
rapid as it takes place after the package has been sealed. Nevertheless, some steps can be taken to
accelerate this phase as much as possible.
As described above, this process can only be successfully used to sterilize and package liquid
(or semi-liquid) and particularly acidic products. It is used mainly for bottling or packaging acidic
beverages (such as nectars and fruit juices), sauces and vegetable preserves (such as tomato sauces)
and for solid products in acid control liquids. In order to successfully use it for packaging low-acid
products, the filling environment would have to be pressurized.
Products packaged with this system are highly stable and can have a shelf life comparable to that
achieved with post-packaging sterilization or pasteurization.
11.2.3 Aseptic Filling
Aseptic packaging is a technology that involves mass-product recovery, cooling of the product and,
lastly, filling at room temperature in pre-sterilized containers, within an aseptic zone called the
white chamber.
In this case, the product is heated to reach the treatment temperature, maintained at this tem-
perature and cooled prior to packaging. This allows higher temperatures to be applied (130 to 150°C
can also be used for low acid drinks) and this results in shorter treatment times (typically 5–6 s
depending on the product) and thermal transients of lesser duration, both in heating and cooling.
This makes it possible to achieve a considerable sterilizing effect and at the same time minimize
the effects on the product, to develop highly stable microbiological foods with a high organoleptic
and nutritional quality. This system can only be used for liquid or liquid products with small pieces/
fibers.
The most commonly used technology for refining foods in aseptic filling environments is UHT
(Ultra-High Temperature) processing, which allows the product to be heated up to the treatment tem-
perature and then cooled at the same rate. UHT treatments can be divided into direct and indirect.
In indirect treatments, the heating of the product takes place through a heat exchanger (which can
be either plate or tubular, depending on the product to be treated) and therefore there is no direct
contact between the food and the heating fluid.
In direct treatments, however, heating is achieved by injecting high temperature steam into the
product (injection), or by spraying the product inside a chamber where high temperature steam
(infusion) is also inserted.
In both cases cooling occurs by passing the food into a heat exchanger in which a coolant fluid
circulates (usually water at room temperature). If the treatment was direct, the product is circulated
in a vacuum chamber to separate the previously injected steam (now condensed) prior to cooling.
In the aseptic process, containers are not sanitized together with the food or by the residual heat
delivered at the time of filling, and must therefore be treated before reaching the filling station since
the packaging must always undergo a microbiologically equivalent treatment to that of the product.
The most commonly used solution for the treatment of containers and closure systems is generally
chemical sterilization by means of H2O2 solutions or peracetic acid, followed by sterilized water/air
rinsing to remove cleansing agent residues.
At this point the filling must take place in a zone of the machine where aseptic conditions and the
white chamber are guaranteed. Creating a white chamber generally isolates some of the machinery
from the rest by means of siphon seals or air-flow curtains, and then creates a slight overpressure
inside. The air circulating inside a white chamber is filtered through High Efficiency Particulate Air
(HEPA) filters to retain 99.97% of the airborne particles with a diameter of less than 0.3 μm. The
pipes carrying the product after it has been cooled to the dosing station should also be aseptic in
order to avoid recontamination.
The advantages of the aseptic system over the other filling technologies illustrated so far are:
• Better organoleptic and nutritional products;

• Lower operating costs due to the high energy savings made with the short duration of
thermal treatments and transients and the fact that the packaging does not require heating;
• Possibility of using containers with lower mechanical and thermal properties and therefore
less expensive.
The main disadvantages are:
• Very high initial investment costs;

• Possibility of long plant shutdowns: if a failure occurs, in addition to restoring the opera-
tion, it is necessary to restore aseptic conditions in the white room;
• It may be problematic to convey the enclosures and closure systems in an aseptic manner
from the area where they are treated to the white chamber.
Aseptic filling can be successfully used to pack all types of beverages, both for flat and add-ons
of CO2, and low- and high-acid products.
Products obtained with aseptic technology are also known as UHT products due to the type of
treatment employed and have a room temperature shelf-life ranging from three to six months.
11.2.4 Ultra Clean Filling

From the point of view of application, ultra clean technology is very similar to aseptic filling. The main
differences lie in the stabilization treatment of the product and the sterilization of containers. This
technology packages products under less microbiological severe conditions rather than the aseptic one.
In this case, the product treatment is usually High Temperature Short Time (HTST) pasteuriza-
tion, possibly preceded by microfiltration, and is thus an even less effective treatment of the prod-
uct’s organoleptic and nutritional characteristics than that used in aseptic filling.
Containers are frequently sanitized in the same way (i.e., chemical treatment), and the main dif-
ference lies in the degree of sterilization achieved. While in aseptic filling it is necessary to have at
least six decimals for the packs, in ultra-clean technology it is enough to have a maximum of three
to four decimals.
This type of filling does not generally achieve commercial sterility, and products packaged in
this way must be kept in the cold chain, from storage to the point of sale, in order to prevent the
germination of spores and the proliferation of vegetative cells that may have survived the treatment.
This technology is usually used for bottling beverages for which the consumer requires some
degree of “freshness” but also a slightly extended shelf life, though not comparable to those obtain-
able with aseptic treatments. Products filled in this way are usually fresh pasteurized milk, milk-
based drinks, fresh fruit juices and smoothies, etc.
Ultra clean products are called ESL (Extended Shelf-Life) and, if properly maintained, the cold
chain can reach one to two months.
11.3 FOOD PACKAGING STERILIZATION

As explained in the previous paragraph, sterilization of packaging plays an increasingly important
role, and in particular with the development of aseptic filling and ultra-clean filling, the need for
new methodologies for reclaiming containers has increased. It is not possible to sterilize products
and packaging together using these packaging technologies, or to take advantage of the residual heat
of the food for packaging remediation. In this section we therefore intend to give an overview of the
various packaging sanitization options developed as a result of these changes.
11.3.1 Heat Sterilization
Inactivation of microorganisms in foods is usually achieved through thermal treatments; these can
be successfully used for sterilization of the packaging, as well as for the product inside them. Before
deciding whether to use heat as a sanitizing agent, the packaging being used must be taken into
account: Plastic or poly-coupled packaging cannot be effectively sterilized in this way, while there
are no problems for cans and other metal packaging.
Microorganisms, and spores in particular, showed a higher resistance to dry heat than wet heat.
The former causes microbial death as a result of dehydration and subsequent oxidation of proteins,
while the latter kills by denaturing and coagulating essential proteins for the cell.
11.3.1.1 Saturated Vapor
Saturated steam is a form of wet heat. In order to create a short-term sterilization phase compatible
with high productivity, the temperature of the surfaces of the packaging must reach at least 135°C
and consequently problems of a technological nature arise concerning how to obtain these tempera-
tures without softening the packaging materials. One solution is to cool the external surfaces, i.e.,
those that do not come into contact with the food. For example, in some processes, polystyrene cups
are subjected to saturated steam at 165°C and 6 bar immediately after removal from the mold, and
the outer surface of the container is cooled at the same time. Polypropylene packaging, which has
a higher thermostability, is treated at lower temperatures (140–147°C) for a slightly longer time. In
any case, the sterilization phase never exceeds 5–6 s.
11.3.1.2 Overheated Steam and Hot Air
Overheated steam and hot air are dry heat forms and therefore less effective than saturated vapors at
killing microbial cells. Longer treatment times are thus required for the same temperature.
When using paper-based packaging material (such as laminated paper and aluminum sheets), hot
sterilization is preferred to saturated or overheated steam because it does not alter the packaging
properties. In some processes, sheets of paper and aluminum are sterilized in warm air at 145°C for
3 min. Hot air sterilization units are available for juice and other beverage bottling systems.
Since dry heat recovery is slow and occurs at very high temperatures, it cannot be used for many
plastic containers.
11.3.1.3 Heating due to Extrusion or Blowing

Temperatures between 180°C and 230°C can easily be reached during the extrusion process of plas-
tic granules to obtain preforms, which will then be blown at high temperatures in molds to obtain
bottles. However, only if the bottles are generated starting from the granule and then kept in aseptic
condition, can the sterility of the container be ensured in this way. If the process is not completely
aseptic, the extruded and then blown containers should be sterilized with H2O2 or peracetic acid to
ensure commercial sterility.
11.3.2 Decontamination with Radiation

When neither heat nor chemical means can be used to sterilize a certain type of packaging, radia-
tion is considered a valid option. The dosage and the type of radiation to be used depend on the type
and volume of microorganisms to be removed from the container. Moreover, the radiation does not
leave any residue on sterilized surfaces and therefore could potentially provide the perfect solution.
However, it can indirectly alter the chemical properties of a material and thus lead to unexpected
interactions with food. The International Atomic Energy Agency (IAEA) states that the following
packaging types can be sterilized by radiation: polyethylene, polyester, polypropylene, nylon, PVC,
etc.
11.3.2.1 UV Rays
The microbicidal effect of UV rays is particularly high for radiation with a wavelength range of
200 to 315 nm, which are called UV-C rays. A 253.7 nm wavelength, for example, has a very high
germicidal potential against yeasts, molds, bacteria, viruses and algae. Mold spores are, however,
especially resistant to UV rays. The lethal dose of radiation is defined as the product of intensity
 mWs 
and time and is expressed as milliwatts second per square centimeter 
 cm 2 
. Despite what the
literature says about UV rays being able to be safely used as a packaging sterilization method, it is
very difficult to get reductions of more than 4 to 5 logs.
11.3.2.2 UV Rays Combined with Hydrogen Peroxide

The bactericidal action of peroxide is enhanced by the simultaneous application of the rays, the
overall effect being greater than that of the two sanitizing agents alone. The effect is excellent at
relatively low H2O2 concentrations (even below 5%) and excessively high concentrations have a
protective effect.
In the case of polychrome packaging, the decontamination of preformed containers can be much
more problematic than the film.
11.3.2.3 Infrared Rays
Infrared beams with a wavelength of 0.8 to 15 × 10–6 m are easily generated. On an absorbent sur-
face they are converted into sensitive heat, causing a surface temperature increase. This is how their
microbicidal effect is generated. This method was used for sterilizing the inner surface of aluminum
lids with a plastic coating on the outside. Since an excessive temperature increase can cause plastic
softening, the maximum treatment temperature and exposure time are limited.
11.3.2.4 Ionizing Radiation
Ionizing radiation is an excellent sterilizing agent that penetrates deep into the objects. Primary
interest ionizing radiation types are: α rays, β beams, X-rays, γ rays and electron beams. Ionizing
radiation kills microorganisms in a very similar way to heat, with comparable sterilization effects.
Sterilization of food packaging with electron beams is a very advanced and effective technol-
ogy that is able to meet both legislative and customer requirements. Unfavorable effects of electron
beams on packaging have not yet been observed. As they are less penetrative than γ rays, they
require less caution in use.
11.3.2.5 Pulsed Light
The technology of using short pulsations of light is an attractive alternative to sterilizing packaging
and equipment in aseptic packaging lines. The spectrum of light used for sanitation ranges from
wavelengths in the ultraviolet region to wavelengths in the near infrared region.
The sterilization material must be exposed to at least one light pulse having a surface energy den-
J
sity of 0.01 to 50 and the duration of the pulse should be between 1 ms and 0.1 s. Packaging
cm 2
materials typically come from 1 to 20 high-intensity and short-duration light pulses. Several flashes
applied at a speed of 1 to 10 per second can generate a high microbicidal action, making this system
applicable in continuous packaging lines. This system has been tested for some polymeric materials
such as LDPE, HDPE, Nylon and PP.
11.3.3 Sterilization with Chemical Agents

Chemical sterilization is the most commonly used system in which containers are stored prior to
filling and is probably also the most energy-efficient because no heat or radiation is generated. In
this case, sanitization must always be followed by a rinsing step with sterile water and/or blowing
with sterile air to remove sterilization agent residues from the containers. The levels of residues
permitted in products are legally regulated.
11.3.3.1 Hydrogen Peroxide
Hydrogen peroxide is one of the most widely used chemical sterilization agents for food packaging.
At present, many aseptic packaging plants use hydrogen peroxide at a concentration of 30–35%
(the concentrations available on the market) for reclaimed packaging, followed by hot air blowing
(60–125°C) to increase the sterilizing effect.
Sterilization performance increases as the concentration of peroxide and the treatment tempera-
ture increases. At room temperature the activity of hydrogen peroxide against microbial forms
is relatively slow. Consequently, if effective sporicidal action is required, the application of H2O2
should occur at a higher temperature than the ambient temperature or should be followed by hot air
blowing. This aspect is very important when the container is used as packaging for low-acid foods.
Another alternative is to use a hydrogen peroxide solution and peracetic acid. The presence of these
elements helps to bring the maximum efficiency temperature of H2O2 closer to room temperature.
In aseptic packaging systems that start from coil (and therefore not from preformed containers),
sterilization usually takes place by immersing the packaging material in a hydrogen peroxide bath,
heating it to increase the bactericidal effect and then vaporizing it to eliminate the residuals of the
chemical agent.
However, when sanitizing pre-formed containers, a nozzle sprinkles or atomizes the solution
within the same. In this case, it is important to make sure that a uniform film of sterilizing agent
covers all the inner walls of the package. Rinsing or drying with sterile water/air to remove the resi-
dues must then follow, since after filling the amount of hydrogen peroxide allowed in the containers
is 0.05 ppm at most according to the FDA rules. This process is beginning to be replaced by the use
of hot air mixtures and vaporized hydrogen peroxide, which is able to guarantee good bactericidal
action and at the same time reduce the amount of peroxide used and hence that absorbed and then
released by packaging material.
In the case of preformed containers with complex geometry, such as plastic or glass bottles and
some types of metal packaging, the treatment consists of an internal rinse with a hydrogen peroxide
solution (30 to 33%) and peracetic acid. The bottles are then rinsed with sterile water, allowed to
drain and dried thoroughly by means of sterile hot air blow.
11.3.3.2 Peracetic Acid
Peracetic acid is a peroxide of acetic acid that when decomposed forms acetic acid and water. Its
low pH and its oxidizing power make it an excellent sporicidal agent. It is typically used as a liquid
sterilizer and is particularly effective against aerobic and anaerobic bacteria spores at lower tem-
peratures than hydrogen peroxide. In practice, as explained above, it is usually employed in solution
with H2O2 (Oxonia). This solution is effective against bacterial spores at 20°C but the application
temperature is typically 35–40°C as this greatly reduces the time of action. This is the most com-
monly used system for the sterilization of glass bottles, plastic material and other preformed con-
tainers of complex geometry.
Peracetic acid vapors are very dangerous and irritant, so if this sterilization method is used, a
system must be implemented to avoid dispersion in the environment. Moreover, the presence of
these vapors is also undesirable in the packaging head space as they can cause off-flavors in the
product.
11.3.3.3 Ozone
Ozone is a powerful oxidizing and disinfecting agent and has recently been introduced in the treat-
ment of food and packaging materials. The aqueous and gaseous forms of this substance are highly
effective against a wide range of microorganisms, including bacteria, spores, fungi and viruses.
It is most frequently used in the form of liquid for the sterilization of packaging, and is obtained
by bubbling gaseous ozone in demineralized sterile water.
At significantly lower concentrations, ozone has proven more effective than hydrogen peroxide
and peracetic acid in eliminating bacterial spores. It has also proven to be effective at room tempera-
ture and with shorter contact times.
Bearing these considerations in mind, although ozone is not currently used at industrial level, its
use could become widely established in the future.
11.3.3.4 Ethanol
Ethanol is the only alcohol used at industrial level with significant bactericidal and fungicidal
action, although it is ineffective against spores. It is typically used at very high concentrations of up
to 70–80%, but in very few applications. As it is not sporicidal, it cannot be used for sterilization
where aseptic packaging is required, but only for ultra-clean packaging or other technologies that
extend product shelf-life.
11.4 CFD APPLICATIONS IN FOOD PACKAGING STERILIZATION AND FILLING

The following studies are the result of research into papers firstly describing the filling process and
then the sterilization process. They are ordered chronologically beginning with the oldest papers
and ending with the most recent. A brief summary was then made explaining what was tested and
how CFD was able to optimize the process.
11.4.1 Simulation and Analysis of Fluid Dynamic Behavior

of Foods During Filling Processes
In this study [2], the authors focus on modeling the filling process of a non-Newtonian shear-
thinning fluid, namely a low-fat yogurt. This choice was motivated by the fact that recently
significant areas of evolution of filling systems have centered on the advent of new products,
and especially on items that do not come under traditional dairy categories, of which yogurts
and fruit juices are important examples.
Once the simulation model had been validated, the subsequent analyses focused on examining
the dynamic behavior of the fluid when passing through the filling valve. In particular, the fluid flow
in a core-annular section of the valve (30 mm above the valve outlet) was investigated, due to its
impact on the performance of the filling process. Specifically, homogeneous fluid velocity within
the core annular section allows better distribution of the fluid at the valve exit, resulting in improved
aseptic bottle filling.
The resulting velocity fields at ΔPFV = 3,000 Pa (a) and PFV = 5,000 Pa (b) are within the range
of the pressure drops applied in industrial applications.
With PFV = 3,000 Pa, the flow is mainly located on the internal side of the filling valve, causing a
possible internal deflection of the fluid during the filling process. Conversely, a uniform distribution
of the fluid within this filling system is reached under a 5,000 Pa pressure drop between valve inlet
and outlet. Moreover, operating at PFV = 5,000 Pa involves increased productivity of the bottling
plants. However, such a pressure drop may cause transient flows during the filling process; the flow
of fluids through the valve should instead be kept laminar and stable.
On the basis of this premise, further analyses were aimed at investigating whether and under
which pressure drops transitions from laminar to turbulent flow occur, and, in this case, which val-
ues of the “critical” generalized Metzner-Reed Reynolds numbers ReMR involve such transitions.
ρVD
Re MR = n −1
(11.1)
 8V 
m′ 
 D 
After several steps and calculations, it can be concluded from the computation that the laminar
flow at the inlet of the valve subsists when the generalized Metzner-Reed Reynolds number is less
than 444 for the filling system examined.
11.4.2 Flow Investigation of the Product Fill Valve of Filling

Machine for Packaging Liquid Products
A full-scale computational fluid dynamics (CFD) model was developed to investigate the complex
flows in a product fill valve that is used for packaging liquid products [3]. Detailed investigations of
pressure, velocity and wall shear stress distribution were made to examine the effect of the current
valve design on CIP (cleaning in place), smooth operation and the product-filling rate.
The case under analysis is an application of Bag-in-Box (BIB) packaging, where the product is
packed in a flexible plastic bag that is in turn packed inside a cardboard box. In this study a product
valve used to fill bags in wineries was analyzed to determine if recent developments in materials,
processes, modeling and manufacturing could produce an assembly with a significantly improved
flow rate, while retaining all of the benefits of the existing device.
In re-designing this piece of equipment, emphasis was placed on improving the CIP ability,
reducing cavitation and increasing filling speed. In addition, the sound level generated by the
valve was investigated. CIP provides several significant benefits for the equipment. These include
improved product quality, longer shelf life, lower labor cost and increased production capacity.
The velocity of fluid flowing through the inlet plane shows a uniform velocity distribution and is
suited to the core-annular nature of the pipe flows. The velocity varies from zero at the valve wall
to a maximum of 9 m/s in the center. However, the contours of velocity at the outlet are of interest
due to their non-symmetrical nature.
Unlike the usual straight tube flow, the position of the highest velocity shifted from the center.
The observed maximum velocity at the outlet was 17 m/s. This unexpected distribution is a result
of the internal shape of the valve.
To overcome this problem several alternative configurations (the inlet stem of the valve can be
lifted further than its usual position, the shape of the main valve body can be made more spheri-
cal or the entry angle of the inlet tube can be altered to avoid the current sharp 60° bend) can be
considered.
In this paper, the geometry of the valve was modified by raising the valve stem 10 mm further
than its usual opening position to investigate the effect of shape changes on the valve’s performance.
The inlet and outlet boundary conditions along with other fluid flow parameters (such as viscosity,
density, etc.) were kept similar to the previous study for comparison purposes.
The results of this analysis indicate significant improvement in velocity distribution.
For the original valve stem position the mid-plane velocity plot shows velocity contours through
the valve assembly in the fully open position. The flow is fully developed with a mass flow of approx-
imately 5.1 kg/s, which is very close to the manufacturer’s rated product-filling rate of 5.0 kg/s,
and provides the validation of this model. However, an analysis of this result reveals some concerns
that the valve’s efficiency is reduced. For example, a rapid increase in velocity was observed where
the flow turns through 60°. The high velocity in this area causes a local pressure drop that is severe
enough to cause cavitation of the fluid. This cavitation is responsible for the reverse flow, developed
at the outlet, resulting in a loss in filling efficiency.
Furthermore, very low velocity was observed in most of the near wall region of the valve, associ-
ated with low turbulence intensity. This certainly reduces the CIP performance of the valve. A case
study with a 10 mm valve lift of current configuration shows significant improvement in velocity
distribution and filling rate. With this change in configuration, the filling rate increased by 7.3%.
The plot of wall shear stresses is of interest as the CIP performance of the valve depends on
adequate wall shear stresses to remove soils.
Wall shear stress is velocity-dependent, so theoretically almost any shape could be cleaned if
adequate pressure and therefore velocity are available. In practice it is preferable to keep CIP pres-
sures as low as feasible for a number of practical reasons, including piping and fitting pressure rat-
ings, pump and holding vessel sizes, equipment size and cost and operator safety.
Much of the internal surface of the valve was subjected to very low wall shear stress (<90 kPa),
which may not be sufficient for cleaning, leading to poor CIP performance. However, even at low
shear stress (<100 Pa), effective cleanability can be achieved if sufficient fluctuation in shear stress
is present.
Therefore, investigation into shear stress fluctuations in lower shear stress regions is vital, in
evaluating the cleanability of the valve.
The use of alternative designs, as suggested in the previous sections, may result in better fluid
circulation and hence greater turbulence intensity in the valve body, resulting in improved CIP
performance.
According to the CFD analysis, alternative designs of this valve can be achieved with no cavita-
tion and better CIP performance to generate considerable benefits, including gentler product treat-
ment, improved filling rates and lower labor costs.
11.4.3 Simulation of the Filling of Polyethylene-Terephthalate Bottles (PET)

with a Volumetric Swirl Chamber Valve (VODM 40355) on
the Basis of Calculation Models and Experiments
This article [4] is about the simulation and validation of the product flow when PET-bottles are
filled, taking into consideration physical and mathematical aspects using a CFD-modeling system.
At first the pure fluid streaming through the inlet pipe system, turn area, bellow and swirl insert
valve is simulated in order to monitor the distribution of the dynamic pressure and velocity mag-
nitude along the paths and cross sections. The boundary conditions were defined to be 1013 hPa
(working pressure), 288 K (working temperature) and 1.22 m/s (inlet velocity).
The simulation revealed that the pressure and velocity rates have significant differences along the
inlet elements. Furthermore the contours of velocity magnitude within some specific cross sections
of the swirl chamber valve were examined in more detail. Despite identical guide blades in this area
the mass flow rates vary from 0.0403 kg/s (135%) to 0.0299 kg/s (100%). These differences will
probably result in different radial velocities of the fluid streams mixing together after leaving the
valve and therefore making a significant impact on the filling quality.
When heated PET pre-forms are placed under pressure from inside and formed in two-piece
molds, undercuts at the transition area between the cylindrical part of the bottle and its shoulder are
likely to occur as a result of the manufacturing process, so the absolute dimensions of the transition
geometry vary from bottle to bottle.
The influence parameters for filling examined in this study are:
• Undercuts at the transition area

The simulation was carried out on the basis of measured data from real bottles and
afterwards extended by the variation of specific geometric parameters. The series of
simulations comprises a bottle geometry without an undercut, a narrow undercut of
3 mm, a wider one with 8 mm and an undercut directed at the opposite side of the
contour. The filling failure influences identified by CFD-simulation correspond to a high
degree with the outcome of practical experiments. As a result, it can be stated that either
wide or narrow undercuts lead to inappropriate fluid appearances at filling. On the other
hand, if an ideal inlet streaming is provided (also simulated), the results are completely
satisfactory in theory. This again shows that valve flow conditions significantly influence
filling quality.
• Transition radius
The series of simulations considered several transition radii measuring 2, 4 and 6 mm.
In conclusion, a bigger radius leads to better results even if a radius of 2 mm produced
fewer disturbances than expected and also allowed fairly successful filling. A transition
radius greater than 4 mm is generally recommended.
• Assembly seams
The series of simulations considered several assembly seams, two of which are pre-
sented with a height measuring 0.2 mm in both cases and widths that are 0.5 mm and
0.3 mm. It was concluded that bottles with distinctive assembly seams are subject to fluid
interruptions, increasing with the amount of discontinuities.
• Transition angle
The transition angle represents the angle of tangents at a theoretical intersection point
of the bottle contours.
Smaller angles (respectively slopes of the contour) lead to a more rapid change in diam-
eter requiring a higher radial velocity component of the fluid to ensure it is pressed against
the inner bottle wall. The series of simulations considered several transition angles: 115°,
105° and 90°. Bottles with angles of 115° and 105° provide near-perfect filling results,
whereas an angle of 90° is subject to fluid interruptions. A transition angle larger than 90°
is therefore preferable.
• Machine rotation
The series of simulations considered realistic centrifugal accelerations of 0.5 m/s2,
1.2 m/s2, 2.0 m/s2 and unusually 4.0 m/s2. The bottle volumes vary between 0.5 liters (small
diameter) and 2.0 liters (bigger diameter). It can be stated that bottles with small diameters
exposed to realistic centrifugal accelerations can be filled without difficulty as well as big-
ger bottle volumes in combination with small accelerations, both showing slight displace-
ments on the fluid front. Only bigger bottles (2.0 liters) exposed to 4.0 m/s2 are subject to
fluid interruptions. Therefore, if standard filling parameters are provided, the influence of
the machine rotation on filling quality is considered marginal.
• Product properties
A further influence parameter for filling quality is considered to be the surface tension
of a fluid and contact angle between fluid and counterpart, which could be PET material
or glass. The surface tension, which varied between 0.073 N/m (water), 0.050 N/m (milk)
and 0.023 N/m (ethanol), exerts minor influences on the filling front of the fluid; only water
tends to remain continuous.
When bottles with a transition angle of 90° are filled, both water (0.073 N/m) and a
theoretical fluid (0.0073 N/m) tend to exhibit volume fractions. The influence of the contact
angle, is varying from 90° (water/PET) to 25° (water/glass). It reveals that small contact
angles lead to appropriate filling results even if unfavorable bottle transition angles (90°)
are used.
• Fluid flow velocity
A fast-filling velocity leads to appropriate filling characteristics, although not at the
beginning of the process when the front of the medium is disturbed due to inertia of the
medium.
A slow velocity is thus advisable at the beginning followed by fast filling at a later stage.
This process has not yet been CFD-simulated but is adequately controllable and has been
verified by practical experiments.
11.4.4 Improvement of BIB Packaging Product Filling

Valve CIP Performance and Efficiency
This paper [5] presents the work carried out to improve the cleaning-in-place (CIP) performance of
a product fill valve used extensively for the bag in box (BIB) packaging of wine. Computational fluid
dynamics (CFD) models were developed in an attempt to improve CIP, remove areas of cavitation
and increase the flow through the valve.
Detailed investigations were undertaken on velocity, pressure distribution, wall shear stress and
turbulence intensity with a view to improving valve performance. The original valve is of cylindri-
cal type and initial attempts therefore maintained the cylindrical shape but raised the valve stem by
10 mm, changing the angle of the inlet tube from 30° to 37° and lengthening the outlet tube by 7 mm.
In the original valve, the inlet plane velocity distribution shows a uniform velocity distribution
that is suited to the core-annular nature of the pipe flow. However, the velocity contours at the outlet
are non-symmetrical and the velocity shifts from the center. This unexpected distribution is due to
the internal shape of the valve.
One of the objectives of changing the valve configurations is to improve this velocity contour
at the outlet. The velocity contours at the mid-plane through the valve assembly at the fully open
valve position have been assessed. The flow is found to be fully developed with the mass flow rate
of approximately 5.07 kg/s, which is very close to the manufacturer’s rated product filling rate of
5.0 kg/s. This provides a basic validation of the model. The model also predicted a local pressure
drop in areas where the flow turns 60°, which is severe enough to cause cavitation.
The CFD investigation into fluid path modifications of the filling valve has confirmed the effec-
tiveness of the method. It appears feasible to increase the filling rate and reduce or eliminate cavita-
tion through the current device by simple changes to the valve configuration, such as raising the valve
stem 10 mm further than the standard position, increasing the angle of the inlet tube and increasing
the entry height of the inlet tube. Thus, the product filling rate can be increased while at the same
time reducing the pressure drops that may be injurious to product quality. If the cylindrical valve
body shape is retained, however, these changes have a negative impact on the CIP performance.
Thereafter, a semi-spherical valve body is designed with a 2 mm increase in outlet diameter

and a 5° decrease in the angle of the inlet tube that joins the valve body. Unlike other cylindrical
designs, the semi-spherical shape showed an increase in wall shear stress and turbulent intensi-
ties. Therefore, cleanability with CIP methods using a semi-spherical shape has been significantly
improved. Performance with this semispherical valve body also showed an increased mass flow of
47% over the original design.
11.4.5 CFD Analysis of Coffee Packaging in Capsules Using

Gas Flushing Modified Atmosphere Packaging
The aim of this paper [6] is to analyze the gas flow in a packaging machine used to fill polymeric
capsules with coffee by means of CFD. The final goal is to optimize the geometric shape of some of
the mechanical components in the machine’s sealing station in order to reduce inert gas consump-
tion (N2 in this case) and at the same time achieve an O2 residual that is constantly equal to or below
1% by mass in the center of the capsule.
The CFD analysis focused only on the sealing station of the packaging machine because this is
the area where the N2 flow rate is highest.
The main components present in this station are: a stainless steel plate, within which a groove
for the distribution of gaseous N2 is formed by milling, four diffuser rings, which introduce the N2
into the capsules, four small pipes (internal diameter measuring 8 mm) that direct N2 flow inside the
groove in the plate, the closing lid and, finally, four electric heaters welding the lid onto the capsule
edge.
The capsule considered is the “Blue” Lavazza model, which has a truncated-cone shape, with
a maximum diameter of 40 mm, a minimum diameter of 32 mm and an overall height of about
26 mm. It is important to specify the capsule’s model because the kind of diffuser ring that has to
be mounted is dependent on the model and the capsules were left empty.
The experimental procedure used is as follows:
1. Execution of the packaging cycle;

2. Removal of the capsules coming out of the packaging line;
3. Placing a small piece of adhesive rubber on the capsule lids, so as to prevent external air
entering the container during perforation;
4. Introduction of the sensor needle;
5. Test execution;
6. Noting the value of O2 residual shown on the sensor monitor.
After analyzing the original case with CFD, its weaknesses were evaluated and modified by the
authors, who then went on to analyze the new case. The main variation is in the shape of the groove
created within the plate:
The changes introduced resulted in an increase in the number of N2 inlet points (from 4 to 8).
The results obtained for O2 residual with the new plate are shown in Table 11.1.
The changes made achieved the target goal of an O2 residual lower than 1% by mass. In particu-
lar, the new configuration shows a mean decrease in O2 residual of −72.25 %, achieving a mean O2
residual of 0.66% (max residual 0.78 %).
A more homogeneous pressure distribution was noted, with values closer to atmospheric pres-
sure and higher velocity values corresponding to some of the injection holes (maximum velocity of
about 16 m/s).
TABLE 11.1
O2 Residual in Capsule’s Center for New Configuration
O2 Residual in Capsule’s Center
N2 Increase in N2 Decrease in O2
Capsule Concentration O2 Residual Conc. Residual
Capsule 1 96.44% 0.75% 6.4% −61.7%
Capsule 2 97.34% 0.56% 16.6% −83.8%
Capsule 3 97.42% 0.55% 16.5% −84.1%
Capsule 4 96.25% 0.78% 5.6% −59.4%
Mean O2 Residual 0.66%
Mean Increase in N2 Conc. 11.28%
Mean Decrease in O2 Residual −72.25%
Source: Adapted from Spanu S., Mosna D., Vignali G., 2016. CFD analysis of coffee pack-
aging in capsules using gas flushing modified atmosphere packaging, International
Journal of Food Engineering 12(9):875–887.
11.4.6 Decontamination of Primary Packaging by Means of Atmospheric Plasmas

The aim of this study [7] was to analyze the decontamination of PET bottles by means of microwave
plasma.
An ionized gas distribution analysis was carried out using CFD to avoid hot spots during the
treatment of the bottle. This sterilization process is optimal due to the absence of chemicals that
may migrate into the packaging. The main disadvantage is the duration of the process, which in
some cases can be too lengthy for the food industry sector.
The material’s resistance to high temperatures is the critical constraint in decontamination pro-
cesses and this also applies to this case. The retraction of the material and its deformation occur
under three conditions:
• Temperatures above 60°C;

• Extended treatment periods;
• Non-homogeneous action of the treatment.
It is therefore important to know the number of cycles and the power of the microwaves at
a certain speed of the plasma. Through the simulations, the upper limit of these parameters
was found, which if exceeded will result in deformation of the bottle. The values found are as
follows:
• Maximum number of cycles 3;

• Cycle time of 300 ms;
• Power of 1000 Watts.
From simulations, it has been found that for a number of higher cycles, the temperature inside the
bottle exceeds 60°C, causing bottle deformation.
The experiments carried out experimentally confirm the analyses made. If a higher number of
cycles or power are used, keeping the other conditions constant, there is a substantial deformation
of the bottle.
A microbiological study was subsequently carried out to see if the treatment was able to reach
the conditions necessary to create a suitable packaging in this regard.
The analysis showed that the results achieved did not meet the requirements for using this tech-
nology in the food sector. In addition, process times remain high.
The article concludes by stating that a more in-depth study of these factors needs to be made.
This point can be improved since it can be seen that the bottlenecks are only slightly stressed, so
further CFD analysis is needed to see how to optimize the positioning of the lance, in order to treat
the inside the bottle more uniformly.
11.4.7 Modeling and Multi-Objective Optimization of the VHP

Pouch Packaging Sterilization Process
This paper [8] deals with the pouch packaging sterilization process used in aseptic technology. A
mixture composed of vaporized hydrogen peroxide and hot sterile air is injected into the package
through a sterilization nozzle. Starting from an initial position of the nozzle inside the pouch, this
study aims to model and simulate the process, firstly varying the flow rate and the position of the
nozzle (in order to ensure contact between Vaporized Hydrogen Peroxide (VHP) and all areas of the
packaging wall) and then varying the process time and hydrogen peroxide concentration in order
to reach the sterilization target of 5-log microbial reduction (according to what was suggested by
Moruzzi et al. [9] to reach at least 4 log for packaging decontamination in aseptic processing), with
minimal processing costs.
The paper is divided into two parts: the first concerns the modeling and fluid dynamics simula-
tion of the sterilization process on the pouch with subsequent experimental validation. A CFD
multicomponent model was created and validated in order to simulate the real process. The second
deals with the multi-objective process optimization, with the aim of finding the most suitable system
configuration. Based on this model, the paper aims to minimize hydrogen peroxide consumption
and optimize the sterilization treatment of the packaging. This issue was approached using multi-
objective optimization software applied to CFD multicomponent simulations. The most suitable
nozzle position inside the pouch was established, as well as the optimal treatment time, concentra-
tion and flow rate of the sterilizing mixture.
Three-dimensional, multicomponent and two-fluid simulations were developed to investigate the
VHP pouch packaging sterilization process. The aim of the first series of simulations is to deter-
mine the most suitable position of the nozzle inside the packaging during the sterilization phase.
Time-dependent simulations were carried out to obtain the best fluid dynamic configuration. In
particular, the speed profile of the sterilizing mixture inside the pouch was evaluated in order to
select the configuration with the most homogenous flow. The total time of the simulation is 18.5 s.
Four nozzle positions were set, namely at 15, 45, 75 and 105 mm from the bottom of the pouch.
The initial concentration of hydrogen peroxide was set at 5,000 ppm.
The results of each simulation are reported in terms of mixture velocity (air and hydrogen perox-
ide) and hydrogen peroxide concentration (ppm). The average value of hydrogen peroxide concen-
tration (ppm) inside the pouch was then determined, exporting the results from the CFD software
to Microsoft Excel.
The sterilization phase was achieved by injecting a mixture of hot air and VHP inside the con-
tainer. The VHP was obtained by vaporizing a 30% hydrogen peroxide commercial solution with a
plate maintained at a temperature of 200°C.
The vapor was then mixed with hot (55°C) sterile air, which sent the VHP towards the nozzle.
Air-flow rate was adjustable between 0.1 and 0.5 L/s.
The nozzle then injected the sterilizing mixture inside the pouch to achieve decontamination.
The treatment time could be adjusted manually between 15 and 21 s. The sterilization phase was
followed by a purging phase, where any residual sterilizing mixture can be removed by a quick hot
air injection (not evaluated in this study).
The simulations generated speed streamlines within a range of between 0 and 1 m/s at regime
(after 6 s of test), and Table 11.2 shows the volume percentages where the speed of the sterilizing
TABLE 11.2
Percentage of Pouch Volume with Velocity Lower Than 0.05 m/s,
in a Range of 0.05–0.20 m/s and Greater Than 0.20 m/s
Nozzle Position Percentage of Pouch Volume with Velocity
(mm) <0.1 m/s 0.1–0.2 m/s >0.2 m/s Total
15 0.6 8.4 90.9 100.0
45 0.3 4.7 94.9 100.0
75 0.7 6.9 92.4 100.0
105 0.3 3.5 96.2 100.0
Source: Adapted from Spanu S. and Vignali G., 2016. Modelling and Multi-objective
Optimisation of the VHP Pouch Packaging Sterilisation Process, International
Journal of Food Engineering 12(8):739–752.
mixture at that moment is lower than 0.05 m/s, between 0.05 and 0.2 m/s and greater than 0.2 m/s.
All these simulations were performed using a fixed concentration of 5,000 ppm.
The simulation where the distance between the nozzle and the pouch bottom was 105 mm shows
the best results with the lowest percentage of volume with low velocities.
Only about 0.3% of the volume shows velocities lower than 0.05 m/s, and more than the 96% of
the volume has a speed higher than 0.20 m/s.
The setting of the optimization model is one of the most important phases of the process, given
that it is necessary to define the optimization strategy.
ModeFrontier™ software was adopted in this study.
This software is able to connect with and act on existing models in order to optimize one or more
objectives using specific optimization algorithms.
After the first series of simulations, the ModeFrontier™ software analyses the results and, on the
basis of chosen criteria, automatically generates new simulation trials.
The objectives of the simulations were connected in the ModeFrontier™ workflow to the output
variables. The two main objectives of the study to be achieved were as follows:
1. To reach the efficacy target (5-log reductions) of the sterilization treatment;

2. To minimize the cost of hydrogen peroxide consumption.
Starting from the best geometrical shape, 144 process configurations were simulated using the
ModeFrontier™ software.
The two conflicting objectives determined a Pareto frontier composed of 19 non-dominated solu-
tions. Finally, a multiple criteria decision making (MCDM) software tool ranked all the optimal
designs and identified design 127 as the optimal configuration. The optimal configuration shows
a sterilization target of 2.96 × 10–01, 15.2 s as process time and a low treatment cost of 2.61 ×
10–04 €/bottle.
11.5 CONCLUSIONS
As can be seen from the work described in this chapter, CFD analysis is important in the food sec-
tor. It can provide an accurate picture of behavior and distribution, and thus generate better results in
the processes studied by improving these aspects. The filling and packaging sterilization phases in
particular present some specific characteristics, which frequently necessitate the use of multiphase
modeling and transient analysis with a specific focus on heat transfer.
In the light of what has been seen in the filling process, CFD analysis has been extremely useful
for analyzing fluid behavior during the process. This analysis aims to investigate the transition from
laminar flow for shear thinning fluids under different pressure conditions, to investigate the complex
flows in a product fill valve, to simulate the behavior of two immiscible fluids with different density
properties, to investigate velocity, pressure distribution, wall shear stress and turbulence intensity
with a view to increasing valve performance and to optimize the geometric shape of some of the
mechanical components.
The steady state model was used in the CFD analysis of these studies, with the exception of the
uses of the time dependent model and both methods of analysis in two studies.
For all these studies a 3D model multi-component simulation was set up, as the behavior and the
interaction of two or more fluids were studied.
In the two cases concerning the sterilization of packaging, CDF analysis was used to establish
the temperature inside the bottles in order to avoid deformation and to understand the velocity and
the concentration of the chemical solution during the sterilization process. In the last case in par-
ticular, a three-dimensional process, multicomponent and two fluid simulations were developed to
investigate the process.
Before modeling the process it is useful to study the original process to be replaced so that the
CFD model can be validated through experimental testing.
In conclusion, there are several ways customers can benefit from incorporating CFD into their
operations: firstly, by using validated virtual processes, fluid dynamic simulations differ only mini-
mally from experimental data. This means we can be confident that the processes will always gen-
erate the desired results. Secondly, making a simulation takes far less time than building several
prototypes and performing consecutive tests. Furthermore, it is faster and cheaper to vary the process
parameters in computer simulations than it would be to stop on-going production and run physical
tests. Thirdly, computational simulations provide us with detailed information that contributes to
our understanding the processes. Important process parameters are identified and can be improved.
Food manufacturers and consumers insist on a perfect product every time. To achieve this, it is
important to stay in complete control of every aspect of food production from beginning to end.
REFERENCES
1. Kaushal P., Sharma H.K., Concept of computational fluid dynamics (CFD) and its applications in food
processing equipment design, kaushal and sharma, Journal of Food Processing Technology, 3: 138,
2012.
2. Bottani E., Rizzo R., Vignali G., Simulation and analysis of fluid dynamic behaviour of foods during
filling processes, Chemical Product and Process Modeling, 2(3): Article 18, 2007.
3. Rahaman M.F., Bari S., Veale D., Flow investigation of the product fill valve of filling machine for
packaging liquid products, Journal of Food Engineering, 85: 252–258, 2008.
4. Hain K., Wels H., Muhr M., Simulation of the filling of polyethylene-terephthalate bottles (PET) with a
volumetric swirl chamber valve (VODM 40355) on the basis of calculation models and experiments, In:
Seventh International Conference on CFD in the Minerals and Process Industries CSIRO, Melbourne,
Australia, December 2009, 2009.
5. Bari S., Veale D., Improvement of BIB packaging product filling valve CIP performance and efficiency,
Food and Bioproducts Processing, 90(4): 849–857, 2012.
6. Spanu S., Mosna D., Vignali G., CFD analysis of coffee packaging in capsules using gas flushing modi-
fied atmosphere packaging, International Journal of Food Engineering, 12(9): 875–887, 2016.
7. Ehlbeck J., Andrasch M., Schnabel U., Rackow K., Neumann T., Krohmann U., Weltmann K.-D.,
Großjohann B., Braun D., Klein M., Schultz G., Dekontamination von Primärverpackungen mittels
Atmosphärendruckplasmen, Vakuum in Forschung und Praxis, 22(5): 10–16, 2010.
8. Spanu S., Vignali G., Modelling and multi-objective optimisation of the VHP pouch packaging sterilisa-
tion process, International Journal of Food Engineering, 12(8): 739–752, 2016.
9. Moruzzi G., Garthright W.E., Floros J.D., Aseptic packaging machine pre-sterilisation and package
sterilisation: Statistical aspects of microbiological validation, Food Control, 11: 57–66, 2000.
12 CFD Analysis of Food
Pasteurization Processes
Padma Ishwarya and C. Anandharamakrishnan
CONTENTS
12.1 Introduction..................................................................................................................... 261
12.2 Food Pasteurization Process................................................................................................ 262
12.2.1 Thermal Pasteurization........................................................................................... 262
12.2.2 Pulsed Electric Field (PEF) Pasteurization......................................................... 265
12.3 CFD Modeling of Pasteurization Process....................................................................... 266
12.3.1 CFD Modeling........................................................................................................266
12.3.1.1 Conservation of Mass Equation..............................................................266
12.3.1.2 Conservation of Momentum Equation....................................................266
12.3.1.3 Conservation of Energy Equation...........................................................266
12.3.2 CFD Modeling of Thermal Pasteurization of Food Products...................................267
12.3.2.1 Canned Milk............................................................................................... 267
12.3.2.2 Bottled Beer.................................................................................................271
12.3.2.3 Egg in Stationary and Rotation Modes................................................... 274
12.3.3 CFD Modeling of PEF Pasteurization.................................................................... 279
12.3.3.1 Governing Equation for Electrostatics.................................................... 279
12.3.3.2 Design and Optimization of PEF Treatment Chamber...........................280
12.4 Conclusions.......................................................................................................................... 285
References....................................................................................................................................... 285
12.1 INTRODUCTION
The term “Pasteurization” was coined in 1857 by Louis Pasteur, a French microbiologist and chem-
ist. Pasteur established that spoilage of milk can be delayed by heating the milk to temperatures
ranging between 50–61°C. Initially, pasteurization was defined as a mild thermal preservation pro-
cess applied on a wide range of food products, mainly liquid foods, to inactivate vegetative cells of
pathogenic and spoilage microorganisms and enzymes that cause deterioration of food quality. The
approach was intended to extend the shelf life of foods with minimal chemical, physical, and organ-
oleptic changes in a product. Later, pasteurization was redefined by the United States Department
of Agriculture (USDA) as “any process, treatment, or combination thereof, that is applied to food
to reduce the most resistant microorganism(s) of public health significance to a level that is not
likely to present a public health risk under normal conditions of distribution and storage” [1].
This definition also considers application of non-thermal processes. such as high pressure (HP) and
high intensity pulsed electric fields (PEF), for pasteurization of foods. Dairy products (butter, milk,
cheese, cream, yoghurt, and ice cream), fruit and vegetable-based products (fruit juices, purees,
jams, jellies, sauerkraut, and pickles), beverages (carbonated drinks, beer, and wine), egg products
(egg white, egg yolk, whole eggs, and dried eggs), honey, vinegar, and meat and fish products (lob-
ster meat, crab meat, cured, cooked ham, hot-smoked fish) are pasteurized.
261
In the thermal pasteurization process, precise choice of temperature-time combinations is vital to

achieve microbial inactivation without affecting product quality. A comprehensive understanding of
the pasteurization process can be obtained with an analysis of temperature distribution in addition
to the changes in thermophysical properties involved. With the aid of computational fluid dynamics
(CFD) simulations, it is now possible to determine the temperature distribution and velocity profiles
during the thermal pasteurization of food products. A common application of CFD simulation with
respect to the thermal processing of foods is the prediction of temperature distribution profiles dur-
ing the process [2]. This is significant as pasteurization relies on the exchange of heat to raise the
temperature at a product’s slowest heating zone (SHZ) to a pre-specified temperature. SHZ is the
region in a food pack/container that receives minimum heating. CFD simulations enable monitor-
ing of SHZ from the temperature distribution predictions. Thus, optimization of the pasteurization
process can be done with the help of CFD modeling. Further, in the case of PEF pasteurization, the
treatment chamber in which the food is exposed to the pulsed electric field, the electric field strength
and the temperature distribution in the treatment chamber have been identified as the key process-
ing parameters that determine the efficacy of treatment. These aspects also have implications on
the sensory quality of the food. Recently, there has been an increased usage of CFD modeling to
understand the application of PEF technology for microbial inactivation. CFD models are useful in
describing the stationary flow pattern, the electric field, and the temperature distributions in a PEF
treatment chamber [3]. This chapter presents the methodology and applications of CFD modeling
pertaining to thermal and non-thermal pasteurization of different food products.
12.2 FOOD PASTEURIZATION PROCESS

12.2.1 Thermal Pasteurization
Thermal pasteurization comprises of heating a food to a specific temperature below its boiling
point over a specific duration followed by an immediate phase of cooling. Pasteurization is a key
critical control point (CCP) in the HACCP plan of various processing operations in food indus-
tries. The purpose and extent of heat treatment during thermal pasteurization vary with the pH of
the food product. While in low acid foods (pH > 4.5), the heat treatment is intended to inactivate
the pathogenic (disease-causing) bacteria, in high acid foods (pH < 4.5), pasteurization aims to
destroy spoilage microorganisms and inactivate enzymes. The actual pasteurization temperature is
determined by the heat resistance of the most heat-resistant microorganism or enzyme in the food
product. The D-value, also termed the decimal reduction value, is a measure of the heat resistance
of a microorganism. It is the time (in minutes) at a given temperature required to destroy 1 log
cycle or 90% of the target microorganism. For instance, in the case of liquid milk pasteuriza-
tion, a pathogenic organism of relevance is the Mycobacterium tuberculosis. The temperature for
milk pasteurization is decided based on the D-value of this bacillus, which is 0.2–0.3 minute at
65.6°C (denoted as D65.6°C) [4]. Similarly, conditions for Coxiella burnetti are also well established.
Alkaline phosphatase, a naturally occurring enzyme in raw milk, has a similar D-value to that of
most heat-resistant pathogens in milk. Thus, phosphatase activity is commonly used as the marker
to determine the effectiveness of the milk pasteurization process. Presence of phosphatase activ-
ity implies inadequate pasteurization of milk. Similar to alkaline phosphatase in milk, certain
enzymes are indicators in heat treatment of fruit and vegetable products. Knowledge of the time-
temperature relationships of these enzymes can enable optimization of the thermal pasteurization
processes for commercial applications.
The striking difference between pasteurization and sterilization is that the former is a less drastic
treatment that is not intended to kill all the microorganisms in the food. Nevertheless, pasteuriza-
tion inactivates the viable disease-causing pathogens and reduces their numbers to a “safe” level,
restricting possibility of the incidence of any disease. Pasteurization conditions do not inactivate
heat-resistant spores.
CFD Analysis of Food Pasteurization Processes 263
Pasteurization can be classified based on two criteria: (i) the temperature-time combination used
and (ii) the mode of operation. Based on the temperature-time combination, pasteurization can be
further classified into several types, as described below:
i. High-temperature-short-time (HTST) pasteurization: In this process, products are

exposed to a higher amount of heat over short durations to inactivate pathogenic bacteria.
HTST pasteurization is the most commonly employed technique for the heat treatment of
fruit juices and milk. The temperature-time combination for HTST pasteurization of liquid
milk is 72°C for 15 s.
ii. Low-temperature-long-time (LTLT) pasteurization: This process uses a relatively lower
amount of heat but at extended duration to achieve the intended degree of microbial inac-
tivation. Milk is subjected to LTLT pasteurization at 63°C for 30 minutes.
iii. Flash pasteurization: This is a form of HTST pasteurization in which the liquid product
is rapidly heated to the processing temperature and held for a shorter time duration, usually
for 15-30 s, followed by rapid aseptic cooling. Such products stay longer and do not need
refrigeration during the storage period.
iv. Tunnel pasteurization: Tunnel pasteurization is an alternative to flash pasteurization.
However, this is a longer process than flash pasteurization and operates at a lower process-
ing temperature of 60°C. The arrangement of this process enables the pasteurization to be
carried out after the product is filled in bottles and sealed (i.e., in-bottle pasteurization).
Consequently, tunnel pasteurization is the predominantly employed technique for beer
packed in cans or bottles. During the process, beer bottles pass through a tunnel that is
divided into various temperature zones. The temperature within each zone is controlled by
the temperature of the water that is sprayed down onto the packages and the bottles remain
in each temperature zone for a specific residence time.
The schematic representation of a tunnel pasteurizer comprising of seven zones is
depicted in Figure 12.1. The first two zones are the preheating zones where the tempera-
ture of the beer is gradually raised from 2°C to around 20°C. This is done to avoid the risk
of breakage of bottles due to pressure build-up. Zone 3 is the superheating zone, where the
water spray temperature is maintained at 65°C for 14 minutes and the temperature of the
beer rises up to 60°C. This zone is the most critical section as any lapse in temperature
control here will lead to overpasteurization and flavor deterioration in beer. Zone 4 is the
holding section, in which temperature of both the water spray and beer are maintained at
60°C for 6 minutes. Pasteurization of beer takes place in this holding section. The next
FIGURE 12.1 Schematic diagram of tunnel pasteurizer. (From E. Bhuvaneswari and C. Anandharamakrishnan.
Heat transfer analysis of pasteurization of bottled beer in a tunnel pasteurizer using computational fluid
dynamics. Innovative Food Science and Emerging Technologies, 23:156–163, 2014.)
three zones are the cooling sections where the temperature of the beer is reduced to 28°C.
The bottles or cans move on a conveyor belt through the tunnel pasteurizer at a controlled
speed. The tunnel pasteurization process is measured in “Pasteurization Units” (PU), in
which 1 PU denotes treatment time of 1 minute at 60°C. Thermal processing of beer is
carried out in levels from 5 to 15 PUs [5,6].
v. Ultra-pasteurization: This involves heating of the product at or above 138°C for at least
2 s to produce a product that has an extended shelf life under refrigerated conditions [8].
Based on the mode of operation, pasteurization can be classified into two types, elaborated as
follows:
i. Batch pasteurization: In the batch process, a large quantity of the liquid product is held
in a jacketed stainless steel vessel through which hot water or steam is supplied. An agi-
tator is mounted onto the vessel to facilitate uniform heat transfer and avoid localized
overheating of the product. The product is heated in the jacketed vessel under specific
time-temperature conditions followed by rapid cooling. The LTLT pasteurization process
is best conducted in batch mode. However, higher heating and cooling time requirements
are the major disadvantage of this system. In addition, the thermal efficiency of a batch
pasteurization process is low as heat regeneration is not possible owing to a considerable
loss of latent heat through the heated product.
ii. Continuous flow pasteurization: Generally, HTST processes are continuous. In the con-
tinuous flow pasteurization process, the liquid food is typically treated in a heat exchanger
(HE), which can be plate-type (PHE), tubular, shell and tube or spiral. In the case of highly
viscous products that are very difficult to pump through plate heat exchangers, scraped sur-
face heat exchangers are used. In PHE, the gap between the plates is adjusted to handle
high viscosity food products. The flow velocity is also increased to overcome the drag force
created by the viscous food product. Further, pasteurization temperatures are maintained at
relatively higher values for viscous products, so as to overcome difficulties in heat transfer
through the viscous mass. The product is forced between the metal plates or through pipes
that are indirectly heated from outside by a heating medium. Hot water or steam is com-
monly utilized as the heating medium. The holding tube is an important component of this
pasteurization system. The lethality accumulated in the holding section is considered impor-
tant by the Food and Drug Administration (FDA) [9]. Thus, the design of the holding tube is
vital to attain a uniform and effective thermal process. While flowing under pressure within
the holding tube, the product is maintained at 72°C for at least 15 s. The flow is controlled
such that the residence time is just above the heating time required to achieve lethality. The
required product flow rate in the holding section is maintained by a metering pump (posi-
tive displacement type) that is located upstream. A flow diversion valve (FDV) is installed
downstream to divert the product back to the heating section if the product is not heated to
the set temperature. This mechanism of the FDV is aided by a temperature sensor located at
the exit of holding tube. Before being chilled back to 4°C in a heat exchanger with cold water
as the cooling medium, the liquid food can be allowed to flow through a heat exchanger to
pre-heat the product that is just entering the system. The latent heat of the outgoing food
product at high temperature is utilized to heat the incoming cold product. This concept is
known as “regeneration” and the corresponding section of the heat exchanger is known as
the “preheating or regeneration section.” Accordingly, regeneration improves the thermal
efficiency of the heat exchangers. The cold pasteurized food product is hermetically sealed
and packed before refrigeration. This is followed by low temperature storage and transporta-
tion prior to retailing. Alternatively, pasteurized food products can be hot-filled and sealed,
followed by cooling with water at ambient temperature. Hot-filling is adopted for viscous
products because, at higher temperatures, viscous products behave as free-flowing liquids.
Continuous pasteurization is appropriate for the bulk heat treatment of pumpable products
such as liquids, puree, semi-liquids, and suspensions with small particles.
12.2.2 Pulsed Electric Field (PEF) Pasteurization

Pulsed electric field (PEF) pasteurization involves subjecting the liquid product to high voltage
electrical pulses at a maximum intensity in the order of 10-80 kV/cm for a very short duration, in
the range of microseconds [10]. The PEF pasteurizer comprises of two major components: a high
voltage pulse generator and a treatment chamber. The treatment chamber includes a fluid handling
system and the required monitoring and controlling devices (Figure 12.2). The food product to be
treated is exposed to an electric field between a pair of electrodes within the treatment chamber. A
non-conductive material connects both the electrodes to avoid electrical flow between them. High
voltage electrical pulses applied to the electrodes conduct the high-intensity electrical pulse to the
product placed between the two electrodes. The food product experiences a force per unit charge,
termed the electric field, which is responsible for the irreversible cell membrane breakdown in
microorganisms [10,11]. This results in microbial inactivation during PEF pasteurization. Electric
field strength and temperature distribution in the PEF treatment chamber influence the efficiency of
the pasteurization process and the sensorial characteristics of PEF-treated food products [3].
The advantages of pulsed electric field (PEF) pasteurization over thermal pasteurization are the
better retention of flavor, textural and nutritional properties and the extended shelf life of the food
products, in addition to microbiological safety [12,13]. The above merits render PEF suitable for the
pasteurization of heat-sensitive foods. A range of liquid and semisolid food products have been sub-
jected to PEF pasteurization, including milk, yoghurt, apple juice, grape juice, orange juice, sugar
cane juice, soups, brine solutions and liquid eggs [14,15].
The efficacy of the PEF pasteurization process is measured by a number of process variables,
including electric field distribution, temperature profile, flow velocity characterization, treatment
time, and properties of the treated liquid product. Experimental measurement of these parameters
FIGURE 12.2 A PEF pasteurizer system. (From M.E.A. Mohamed and A.H.A. Eisa. Pulsed Electric Fields
for Food Processing Technology (Chapter 11). In: Structure and Function of Food Engineering. London:
IntechOpen Ltd., 2012, http://www.intechopen.com/books/structure-and-function-of-foodengineering.)
is not possible, owing to extremely short treatment times. CFD aids in the prediction of temperature
distribution in the PEF treatment chamber and thus finds application in the optimization of process
and equipment design of PEF pasteurization systems.
12.3 CFD MODELING OF PASTEURIZATION PROCESS

12.3.1 CFD Modeling
CFD is a simulation tool that uses powerful computers in combination with applied mathematics
to model fluid flow situations and aid in the optimal design of industrial processes. In recent years,
rapid developments in the application of CFD in food processing operations have occurred [16,17].
CFD modeling is governed by the equations of fluid dynamics, primarily derived from the laws of
conservation. These comprise of the equations of continuity (Equation 12.1), momentum (Equation
12.3) and energy (Equation 12.4). The above equations are generally expressed as partial differ-
ential transport equations, collectively known as the Navier–Stokes equations, which describe the
conservation of mass, momentum, and energy and state equations [2,17]. CFD modeling seeks the
solution of these equations governing the flow of fluids inside defined flow geometry. These equa-
tions are derived by considering the mass, momentum and energy balances in an element of fluid,
resulting in a set of partial differential equations. They are completed by adding other algebraic
equations from thermodynamics, such as the equation of state for density and a constitutive equa-
tion to describe rheological properties [18,19].
12.3.1.1 Conservation of Mass Equation

The continuity equation describes the rate of change of density at a fixed point resulting from the
divergence in the mass velocity vector ρv. Equation (12.1) is the unsteady, three-dimensional, mass
conservation or continuity equation for the simplified case of an incompressible or constant density
fluid.
∇⋅v = 0 (12.1)

where ∇ has the dimension of reciprocal length.
∂ ∂ ∂
∇= i+ j+ k (12.2)
∂x ∂y ∂z
12.3.1.2 Conservation of Momentum Equation

The principles of the conservation of momentum are an application of Newton’s Second Law of
Motion to an element of fluid and state that a small volume of an element moving with the fluid is
accelerated because of the force acting upon it.
Dv
ρg = −∇p + ∇ ⋅ τ + ρg g (12.3)
Dt
In Equation (12.3), the convection terms are on the left side and terms on the right-hand side
include the pressure gradient (p), source terms of gravitational force (g) and stress tensor (τ ), which
is responsible for the diffusion of momentum.
12.3.1.3 Conservation of Energy Equation

The energy equation is based on the first law of thermodynamics, which states that the time rate of
change of internal energy plus kinetic energy is equal to the rate of heat transfer minus the rate of
work done by the system. A CFD solver computes the solution for the energy equation in the fol-
lowing form (Equation 12.4):
∂
∂t
( ρ E ) + ∇ ⋅  v ( ρ E + p)  = ∇ ⋅  keff ∇T −
 ∑ h J + (τ .v)
j j (12.4)
 j 
where E is the internal (thermal) energy, keff is the effective conductivity (kta+ kt, where kta is thermal
conductivity and kt is turbulent thermal conductivity), T is the temperature, τ is the stress tensor
and J j is the diffusion flux of species j and hj is the enthalpy of species j. The three terms on the
right-hand side of the equation represent energy transfer due to conduction, species diffusion and
viscous dissipation, respectively [19,20]. CFD involves an understanding of the physical meaning
of the above equations.
In the context of thermal pasteurization, CFD is mainly utilized as a tool to determine the tem-
perature distribution pattern inside the product and to determine the position of the slowest heating
zone (SHZ) at various time stages during the heat treatment [21–23].
12.3.2 CFD Modeling of Thermal Pasteurization of Food Products

The industrial pasteurization process aims to extend shelf life without deteriorating quality attri-
butes of the food product. These objectives are attained when the product receives the adequate
degree of heat treatment. Heat transfer mechanisms during the thermal pasteurization include [24]:
• Conduction (in solid foods),

• Natural convection (in low viscosity liquid foods), and
• A combination of convection and conduction (in liquid foods containing solid particles or
viscosity modifiers such as starch).
The study of the temperature profile inside the product during pasteurization is an unsteady state
problem, as the temperature within the package depends on both position and time. An experimen-
tal investigation for this case requires thermocouples being placed at different positions to record the
temperature in a container during heating. This disturbs the flow patterns as the physical presence of
thermocouple wires may restrict the free/actual movement of the liquid. This leads to errors in tempera-
ture measurements [25]. Consequently, studying the thermal behavior of foods under actual conditions
is laborious as it involves such complexities as variation in initial temperature, non-linear and non-
isotropic thermal properties, irregular-shaped geometries and time-dependent boundary conditions
[26]. Further, because of the complex nature of heat transfer in natural convection heating, it is a dif-
ficult task to locate the slowest heating zone (SHZ) [27]. Also, it is difficult to measure the temperature
at the SHZ because this region continuously shifts its position during the heating progress [28]. In the
above context, CFD modeling assists in predicting the temperature profiles within the package by find-
ing numerical solutions for the partial differential equations that govern the heat transfer process. The
CFD models developed to study the thermal pasteurization of different food products are elaborated
upon in the subsequent sections. This section focuses on the development and interpretation of CFD
models for thermal pasteurization of food products and on the deliverables obtained from the model.
12.3.2.1 Canned Milk
Pasteurization of milk is an important thermal preservation approach to increase the shelf life of
milk. The processing temperature and time for the thermal pasteurization of canned milk should be
selected after considering the heat sensitivity of various nutrients, such as thiamine, α-lactoalbumin,
and β-lactoglobulin, present in milk [29]. Uniform heat distribution inside the can is critical. The cho-
sen processing time is thus also based on the time required to achieve uniform temperature inside
the can. However, determining the temperature profile within canned milk is challenging during
an industrial-scale continuous-type operation. A CFD model was developed for the pasteurization
process of canned milk to study the effect of thermal processing time on temperature distribution
inside the can and to validate the same with experimental measurements [30]. In addition, this study
also investigated the effect of can rotation on the temperature distribution, processing time and pas-
teurization unit (PU). The degree of heat treatment received by the product was expressed in terms of
the pasteurization unit (PU) given by the following equation [31,32], similar to the F value equation:
t
∫ 10
Τ − Tref )/ z
PUTref = dt (12.5)
0
where Tref is the reference processing temperature and T is the temperature of the milk. Equation 12.5
implies that a PU value of 1 is equivalent to a process of 1 min at a specified processing temperature.
This value is used to find out the effect of time and temperature on the inactivation of microorganisms.
The CFD model utilized a three-dimensional cylindrical structure of the milk can (Figure
12.3[a]). A hexahedral mesh with 73,584 grid cells of size 0.001 m (Figure 12.3[b]) was used in
order to minimize the discretization error. The finite volume method was used to solve the partial
differential forms of the continuity equations using the SIMPLE (Semi-Implicit Pressure-Linked
Equations) approach for pressure-velocity coupling and a second-order upwind scheme was used
to interpolate the variables on the surface of the control volume due to its stability with respect to
time step size. The thermo-physical properties of milk listed in Table 12.1 were used as the model
parameters. CFD model predictions were validated with experimental measurements of tempera-
ture recorded at three different positions along the height of the can in the radial direction (bottom:
z = 0.01 m, middle: z = 0.0425 m and top: z = 0.07 m) (Figure 12.3[c]).
The governing equations for the model were the transport equations for conservation of mass,
momentum, and energy (Equations 12.1 to 12.4), as explained above. This study considered a com-
bination of conduction and convection heat transfer mechanisms. Heat transfer to milk near the
walls of the can was considered to be governed by conduction, and heat transfer into the milk was
described by natural convection effects. An unsteady state 3D CFD simulation was carried out with
three-side heating without slip conditions. A 0.2 mm wall thickness was assumed with bottom and
side wall set at two different temperatures (Twall): 74°C and 85°C, approximated to be equivalent
to the hot water temperature, owing to the negligible thermal resistance of the can material. The
temperature of the top wall was defined to be 32°C (ambient temperature).
FIGURE 12.3 (a) Can geometry; (b) meshed can; (c) radial planes. (From D. Anand Paul, A. Anishaparvin
and C. Anandharamakrishnan. Computational fluid dynamics studies on pasteurisation of canned milk.
International Journal of Dairy Technology, 64(2):305–313, 2011.)
TABLE 12.1
Thermo-Physical Properties of Milk
Properties Value⁄Equation Reference
Density (kg/m3) 1029 (Boussinesq approximation) [29]
Thermal conductivity k = (326.8 + 1.0412T [33]
(W/m K) – 0.00337*T2)
(0.44 + 0.54Xmwater) 1.73 × 10–3)
Thermal expansion 0.0002 [34]
coefficient (K–1)
Source: D. Anand Paul, A. Anishaparvin and C. Anandharamakrishnan.

Computational fluid dynamics studies on pasteurisation of canned milk.
International Journal of Dairy Technology, 64(2): 305–313, 2011.
Note: Xm: mass fraction; T: temperature (K).
In the stationary position, the simulation predictions were in good agreement with the experimental
results at point-1 and point-2 for both the wall temperatures considered (Figure 12.4[a & b] for Twall = 74°C
and Figure 12.5[a & b] for Twall = 85°C). At 74°C, there was a linear increase in temperature during the
first 180 s of heating, after which it reached a steady state temperature of 69.5°C for point-1 and 69.2°C
for point-2 at 480 s (Figure 12.4[a & b]). A similar trend was observed at 85°C, but the steady state tem-
perature was reached at 540 s which was 79.6°C for point-1 and 79.2°C for point-2 (Figure 12.5[a & b]).
100 100
(a) (b)
Temperature (ºC)
80
Temperature (ºC)
80
60 60
40 40
Experiment Experiment
20 20
CFD simulated CFD simulated
0 0
0 150 300 450 600 750 900 0 150 300 450 600 750 900
Time (s) Time (s)
FIGURE 12.4 Comparison of experimental and CFD simulation results at 74°C (a) Point-1 (from can top to
0.05 m down and 0.01 m from the center); (b) Point-2 (from can top to 0.065 m and 0.02 m from the center).
(From D. Anand Paul, A. Anishaparvin and C. Anandharamakrishnan. Computational fluid dynamics studies
on pasteurisation of canned milk. International Journal of Dairy Technology, 64(2):305–313, 2011.)
100 100
(a) (b)
Temperature (ºC)
Temperature (ºC)
80 80
60 60
40 40
Experiment Experiment
20 CFD Simulated 20 CFD Simulated
0 0
0 150 300 450 600 750 900 0 150 300 450 600 750 900
Time (s) Time (s)
FIGURE 12.5 Comparison of experimental and CFD simulation results at 85°C (a) Point-1 (from can top to
0.05 m down and 0.01 m from the center); (b) Point-2 (from can top to 0.065 m and 0.02 m from the center).
FIGURE 12.6 Velocity profile of milk inside the can during pasteurization (85°C) of milk inside a stationary
can at time intervals [(a) 60 s and (b) 360 s] and can rotated at 5 rpm at time intervals [(c) 60 s and (d) 360 s].
The rapid increase in temperature was attributed to the low viscosity of milk. Differences in experimental
and simulated results were attributed to variation in milk properties used for simulation studies.
With respect to locating SHZ in the can, the model indicated that during the initial stage of heat-
ing (60 s), the zone of minimal heating was formed at the bottom portion of the can due to changes
in the viscosity, density, and re-circulating pattern of the milk (Figure 12.6[a & b]). With time, the
SHZ moved towards the core region, indicated by lower milk temperatures at the central plane
compared to the surrounding regions (Figure 12.7[d]).
In the stationary can, the velocity profiles depicted a gradual increase in velocity at the top central
region of the can (Figure 12.6[a]) due to the re-circulatory flow of milk as a result of the convective
heating. The temperature difference between wall and the interior regions of the can created density
differences in the canned milk. This stimulates a buoyancy force in the fluid and thereby leads to
natural convective flow patterns. When heated at 60 s (Figure 12.6[a]), the velocity of milk at the
core region was relatively lower compared to that at 360 s (Figure 12.6[b]). However, when the can
was rotated at 5 rpm, the velocity distribution was uniform compared to that in the stationary posi-
tion (Figure 12.6[c & d]). Due to radial mixing, the temperature distribution was almost uniform,
restricting the formation of SHZ. The key finding from this CFD study was that with the rotation
FIGURE 12.7 Temperature (°C) profiles during pasteurization (85°C) process of milk in a can held at
stationary position, at (a) 60 s (b) 120 s (c) 240 s and (d) 360 s. (From D. Anand Paul, A. Anishaparvin
and C. Anandharamakrishnan. Computational fluid dynamics studies on pasteurisation of canned milk.
of can, uniform heating of milk was achieved at a reduced processing time, due to the absence of
SHZ. This is significant as reduced processing time leads to lower processing cost and eventually a
reduced final product cost. The study also showed that rotation of the can at higher rpm for a shorter
time (100 rpm for 360 s) or lower rpm for a longer time (5 rpm for 480 s) was ideal for the thermal
pasteurization of milk. This was further confirmed by the effect of can rotation on the PU (Figure
12.8). Can rotation can result in higher PU values, corresponding to improved scope for inactivation
of microorganisms when compared to pasteurization of a stationary can (Figure 12.8). Further, PU
values increased when the can was rotated at a higher speed (100 rpm). Maximum PU (1 unit) can
be achieved only when the product temperature is equal to the process temperature, thus demon-
strating the positive effect of pasteurization conducted in rotation mode on microbial inactivation.
12.3.2.2 Bottled Beer
Beer is one of the most widely consumed alcoholic beverages in the world. Pasteurization is an
important unit operation in beer processing [7]. Heat treatment received by the product during pas-
teurization influences beer quality in terms of flavor, color, bitterness, chill haze, and protein sensi-
tivity [35]. Higher pasteurization temperatures and shorter holding times are considered favorable
towards obtaining an optimal beer quality. Thus, it is imperative to control the process temperature
by regulating the temperature of the water that is sprayed on the beer bottles during their residence
inside the tunnel pasteurizer [36]. For reasons similar to those explained with milk pasteurization, it
is difficult to determine the temperature profile and slowest heating zone (SHZ) inside bottled beer
FIGURE 12.8 Effect of can rotation speed on pasteurization unit (PU) during pasteurization of milk at
85°C in stationary, rotation of 5 rpm, 50 rpm, and 100 rpm. (From D. Anand Paul, A. Anishaparvin and
C. Anandharamakrishnan. Computational fluid dynamics studies on pasteurisation of canned milk.
during industrial-scale tunnel pasteurization. CFD modeling can be used as a tool to determine the
heat distribution inside bottled beer and to examine the position of SHZ [7].
Augusto et al. (2010) [37] carried out CFD simulations to predict the influence of bottle orien-
tation on the “in-package” thermal pasteurization of beer at 60°C for up to 15 PU. The cans were
assumed to be in conventional, inverted, and horizontal positions. Temperature profiles and velocity
of convection currents during the process and the variation of the PUs were evaluated in relation to
time and bottle orientation. Simulation results showed that the package position did not improve the
process. Later, a transient three-dimensional CFD model was developed for the tunnel pasteuriza-
tion of beer [7]. The approach followed in this study was different, as the CFD simulation considered
the movement of different heating zones around the stationary beer bottle. The temperature profile
inside the bottle was predicted by considering the bottles stationary. Thus, the conditions used in the
simulation were similar to those prevalent in industrial-scale tunnel pasteurization, in which beer
bottles move through different heating zones. In addition, the execution of this modeling approach
addressed the limitations associated with the availability of computational resources. The study
also investigated the effect of different temperature zones on the inactivation of Saccharomyces
cerevisiae, the most common spoilage microorganism in beer. The temperature profile in the bot-
tled beer predicted by CFD simulation was validated with experimental measurement.
The 3D geometry of the beer bottle considered in this study (Figure 12.9) consisted of three volumes;
the first represented the outer glass bottle with a height of 28.3 cm and 3 mm thickness, the second
represented the beer in the bottom portion of the bottle with a height of 19.4 cm, and the third volume
represented the headspace with a height of 8.9 cm. The surface and volume mesh were generated and
exported to CFD code for further processing. The total number of tetrahedral cells in the volume mesh
comprising of all three volumes was 128370 (bottle: 34261; beer: 88219; and headspace: 5890). The
authors used coarse meshing, as it is an established approach to reduce the computational time. The study
employed the finite volume approach to solve the partial differential forms of the continuity equations
using the SIMPLE method for pressure-velocity coupling. A first-order implicit upwind scheme was used
to interpolate the variables on the surface of the control volume, owing to its stability with respect to sim-
ulation time step size (set as 1 s) with five iterations/time steps. The equations of conservation for mass,
momentum, and energy were considered the governing equations for the model (Equations 12.1 to 12.4).
The CFD simulation assumed a convective mode of heat transfer from the headspace air to
the beer bottle. The beer inside the bottle was assumed to be at a uniform initial temperature of
27°C considering laminar flow throughout the thermal processing. The hot water temperature was
FIGURE 12.9 3D geometry of a standard long neck beer bottle. (From E. Bhuvaneswari and C.
Anandharamakrishnan. Heat transfer analysis of pasteurization of bottled beer in a tunnel pasteurizer using
computational fluid dynamics. Innovative Food Science and Emerging Technologies, 23:156–163, 2014.)
assumed to remain constant at 60°C along the boundaries of the heating wall during pasteurization.
The water temperature was approximated to be the temperature of the outer surface of the bottle
wall, i.e., Twater = T = Touter wall = 60°C. This study used the actual thermo-physical properties of beer,
headspace, glass bottle, and metal cap (Table 12.2) instead of using the properties of water. This
resulted in closer correlation between the CFD simulation predictions and experimental measure-
ments. A no-slip condition was applied at bottle walls and beer surface.
The CFD prediction of beer temperature during pasteurization was found to be in good agreement
with experimental measurements. A curvilinear increase in temperature was observed in the preheat-
ing zones and the beer temperature maintained a constant value of 60°C after 1200 s of processing
(Figure 12.10[a]). As in the case of milk, the low viscosity of the product was responsible for the
initial rapid increase in temperature. The temperature contours of the beer bottle (Figure 12.10[b])
at different time intervals during pasteurization showed that during the initial stage of heating, SHZ
was formed at the bottom of the bottle due to changes in viscosity and density. The headspace region
heated faster due to the difference in thermophysical properties between the headspace air and beer.
The above observations were further confirmed by the CFD simulated temperature profile of the beer
TABLE 12.2
Thermo-Physical Properties of Beer, Headspace, Glass, and Metal
Properties Value ⁄ Equation Reference
Beer: 2 * 10−5 × − T2 − 2 * 10−3 × T + 4.118 [38]
Specific heat capacity (kJ/kg °C) −8.116 * 10−6 * T2 + 1.9 * 10−3 × T + 0.54611
Thermal conductivity (W/m °C)
Headspace: 1.067 * 10−5 × T2 − 4.4992 * 10−3 × T + 1.29 [39]
Density (kg/m3) 1007
Specific heat capacity (J/kg °C) 7.31 * 10−5 × T + 0.02367793
Thermal conductivity (W/m °C) 4.5 * 10−8 × T + 1.73568 * 10−5
Viscosity (kg/m s)
Glass bottle: 2500 [40,41]
Density (kg/m3) 750
Specific heat capacity (J/kg °C) 1.4
Metal cap: 2702 [40]
Density (kg/m3) 903
Specific heat capacity (J/kg °C) 237
Source: E. Bhuvaneswari and C. Anandharamakrishnan. Heat transfer analysis of pasteurization of

bottled beer in a tunnel pasteurizer using computational fluid dynamics. Innovative Food
Science and Emerging Technologies, 23:156–163, 2014.
bottle extracted at three different points (Figure 12.10[c]). The slowest heating rate was depicted by
point 3, which was located towards the bottom of the bottle (Figure 12.10[c]).
The temperature contours of bottled beer (Figure 12.11[a-m]) provided details of temperature
distribution in the “preheat,” “superheat,” “holding,” and “cooling” zones of the tunnel pasteurizer.
In the first preheat zone, the hot beer began to flow upwards due to the density difference created
by the natural-convection-mediated temperature gradient. Consequently, the colder fluid was also
carried by viscous drag. The fluid flowing upward was deflected by the top cone surface of the
bottle, after which it started moving in a radial direction and then downwards, thus demonstrating
a recirculation phenomenon until uniform heating was achieved. In the second preheat zone and in
the superheat zone, at the processing time of 510 s (Figure 12.11[c]) and 1140 s (Figure 12.11[f]),
respectively, the SHZ was at the bottom of the bottle due to convective heating and the effect of
funnel-shaped bottle neck geometry. In the second cooling zone, the temperature contour of the
bottle at 2700 s (Figure 12.11[k]) was maintained at 32°C, indicating that the bottom portion of the
bottle cooled slowly compared to the top portion. Thus, the position of SHZ remained at the bottom
region, depicted by the low rate of heat transfer during both heating and cooling processes.
Further, the degree of inactivation of the target microorganism S. cerevisiae (yeast) estimated
using Eq. (12.5) was integrated into the CFD simulation to calculate the PU values for all seven
zones of the tunnel pasteurizer. PU values were 19.89, 15.71, and 13.96 for points 1, 2, and 3, respec-
tively. The obtained PU value was within the safe range [36] and was sufficient to achieve the
required microbial inactivation. All three points followed similar trends of inactivation with slight
variations due to natural convection (Figure 12.12).
12.3.2.3 Egg in Stationary and Rotation Modes

Eggs are consumed worldwide for their nutritive value. Pasteurization is an important unit operation
employed for the inactivation of Salmonella enteritidis, a common pathogenic bacterium in eggs
that can cause food-borne illnesses. Few works have been published on the development of CFD
65
60
55
Temperature (ºC)
50
45
Experiment
40
CFD simulation
35
30
25
0 500 1000 1500 2000
Time (s)
6.00e+01
5.83e+01
5.65e+01
5.48e+01
5.30e+01
5.13e+01
4.95e+01
4.78e+01
4.60e+01
4.43e+01
4.25e+01
4.08e+01
3.90e+01
3.73e+01
3.55e+01
3.38e+01
3.20e+01
3.03e+01
2.85e+01
2.68e+01
2.50e+01
Temperature (ºC) a) 30s b) 90s c) 120s d) 360s e) 420s f ) 600s g) 1140s

60
55
50
Temperature (ºC)
45
Point 1 1
40
Point 2
2
35
Point 3
30 3
25
0 200 400 600 800 1000 1200
Time (s)
FIGURE 12.10 (a) Comparison of experimental measurement (center point: 6.95 cm from the bottom)
of temperature and CFD simulation results during pasteurization of bottled beer at 60°C; (b) Temperature
contours (a–g) of bottled beer during the pasteurization (60°C) process at different time intervals; (c) CFD
simulated temperature profiles (at three points) for beer bottle at pasteurization temperature of 60°C (point
1: [top] 18.4 cm from bottom; point 2: center; point 3: [bottom] 2 cm from bottom). (From E. Bhuvaneswari
and C. Anandharamakrishnan. Heat transfer analysis of pasteurization of bottled beer in a tunnel pasteurizer
using computational fluid dynamics. Innovative Food Science and Emerging Technologies, 23:156–163, 2014.)
models for thermal pasteurization of eggs. A CFD approach for determining the surface heat trans-
fer coefficient during thermal processing of eggs was proposed by Denys et al. (2003) [42]. Later, a
combined conductive and convective heat transfer model for egg pasteurization [43] and a first-order
kinetic model for studying the inactivation of S. enteritidis [44] with respect to the position of egg
yolk during thermal pasteurization were developed. While these earlier works have considered a
FIGURE 12.11 Temperature profiles (a–m) of bottled beer in different zones of a tunnel pasteurizer at differ-
ent time intervals. (From E. Bhuvaneswari and C. Anandharamakrishnan. Heat transfer analysis of pasteuri-
zation of bottled beer in a tunnel pasteurizer using computational fluid dynamics. Innovative Food Science
and Emerging Technologies, 23:156–163, 2014.)
FIGURE 12.12 Pasteurization unit (PU) calculated inside the bottled beer for the seven-zone tunnel pasteur-
izer at point 1, point 2, and point 3. (From E. Bhuvaneswari and C. Anandharamakrishnan. Heat transfer anal-
ysis of pasteurization of bottled beer in a tunnel pasteurizer using computational fluid dynamics. Innovative
Food Science and Emerging Technologies, 23:156–163, 2014.)
FIGURE 12.13 Three-dimensional geometry of egg. (From R. Ramachandran, D. Malhotra, A. Anishaparvin

and C. Anandharamakrishnan. Computational fluid dynamics simulation studies on pasteurization of egg in
stationary and rotation modes. Innovative Food Science and Emerging Technologies, 12:38–44, 2011.)
stationary position of the egg during thermal processing, a more recent work compared the effects
of thermal pasteurization of whole egg with yolk at 55.6°C on the pasteurization time in stationary
and rotational modes at two different speeds: 2.5 rpm and 5 rpm [45].
The three-dimensional geometry of the egg used in this study, assuming the yolk to be at the
center of the egg, is shown in Figure 12.13. The ellipsoid shape of the egg was created in two dimen-
sions and its three-dimensional structure was obtained by revolving the contour 360° about its axis
of symmetry. The thickness of the shell wall was included in the geometry in order to account for
conductive heat transfer through the eggshell. The surface and the volume mesh were generated and
exported to CFD code for further processing. The geometry thus created was meshed using 50,000
tetrahedral/hybrid elements (including that for both albumin and yolk).
A transient 3D CFD simulation was carried out using the finite volume method. The partial dif-
ferential forms of the governing conservative equations (mass, momentum, and energy) were solved
using the SIMPLE method. The egg white and yolk were considered non-Newtonian fluids and their
walls were coupled for simulation studies. The thermal and physical properties of egg white and
yolk were considered individually and expressed as a piece-wise linear function of temperature. The
wall temperature (Twall) was set at 56.5°C, approximated to that of hot water, considering the very
low thermal resistance offered by the eggshell. The temperature of hot water was assumed to remain
constant along heating wall boundaries. An all-side heating was adopted without slip conditions. The
CFD model considered conductive heat transfer to compare the stationary and rotational modes of
egg pasteurization. In the rotational mode of thermal pasteurization, two different speeds of rotation
were chosen (2.5 and 5 rpm), based on the fact that a minimum rotation is required to avoid damage
to yolk and mixing of yolk and albumin. During the CFD simulation of the rotating egg, the effects of
coriolis acceleration and the centripetal acceleration were considered with the momentum equation.
Simulated results were extracted for three points in the egg with the first point at 0.025 m from center
to top; the second point at the center and the third point at 0.0195 m from center to right. Further, the
inactivation kinetics of S. enteritidis was studied using the F value equation, which is similar to that
of the PU given in Equation (12.5). Since egg white and yolk have varied effects on the thermal inac-
tivation of S. enteritidis, two different z-values of 4.37°C and 3.29°C were used for yolk and white,
respectively [39]. Twall was used as the Tref in Equation (12.5). Then the Equation (12.5) was solved
in the CFD simulation with respect to time and temperature of the process to quantify the F value.
FIGURE 12.14 CFD-simulated temperature profiles at three points for whole egg heating at 55.6°C (sta-
tionary mode). (From R. Ramachandran, D. Malhotra, A. Anishaparvin and C. Anandharamakrishnan.
Computational fluid dynamics simulation studies on pasteurization of egg in stationary and rotation modes.
Innovative Food Science and Emerging Technologies, 12:38–44, 2011.)
CFD simulation in the stationary mode showed that thermal pasteurization of a whole egg requires
more time to reach a uniform temperature, explicated by the plot of temperature vs. time shown in
Figure 12.14. The egg yolk heated up very slowly compared to the rest of the egg due to the difference
in thermo-physical properties (thermal conductivity and density) between the egg yolk and egg white.
The temperature of the egg white (points 1 and 3) increased exponentially (Figure 12.14) with time,
as it was located near the eggshell. On the contrary, the temperature of the egg yolk (point 2) showed
a linear increase (Figure 12.14), as it remained at room temperature until heat reached the surface of
the yolk. The center of the yolk took 30 minutes to reach the pasteurization temperature of 55.6°C.
Thus, this showed that the SHZ was located in the region of egg yolk slightly below its geometric
center, which was further confirmed by the temperature contour obtained from the CFD simulation
at various time intervals during the process: 30 s, 150 s, and 300 s (Figure 12.15[a]). The SHZ did
not shift towards the lower half of the egg throughout the processing time.
Strikingly contrasting observations were obtained with the thermal pasteurization of the egg
simulated under the rotational mode. The rotation was found to improve the efficiency of the heating
process by ensuring more uniform heating compared to stationary heating. In the rotational mode,
the egg white reached the set pasteurization temperature at a faster rate due to improved heat pen-
etration throughout the egg. Nevertheless, the yolk portion took more time to heat up compared to
the egg white (Figure 12.15[d–f]), but less time than the stationary egg (Figure 12.15[c]). The SHZ
of the rotating egg was concentrated at the center of the yolk throughout the rotating process (Figure
12.15[d,e]), unlike the SHZ in a stationary egg (Figure 12.15[b]). Thus, the rotating egg achieved a
uniform temperature in approximately 300 s (Figure 12.15[f]), while the stationary egg had an SHZ
that remained at 37°C even at the end of the processing time, i.e., 300 s (Figure 12.15[c]).
Further, the rate of heating improved with an increase in the rotational speed from 2.5 to 5 rpm. While
the egg white (point 1, Figure 12.16[a]) reached maximum temperature in 6.5 min at 5 rpm, at 2.5 rpm
the same was attained in 7 min. However, these time durations were significantly shorter compared to the
stationary egg, which required 30 min. The specified point inside the yolk (Figure 12.16[b]) took a longer
time to reach the pasteurization temperature than the point on the egg white (Figure 12.16[a]). However,
the egg yolk demonstrated a similar trend as the egg white with respect to the effect of rotation on the
heating rate. The egg yolk rotated at 5 rpm took 9 min to reach the maximum pasteurization temperature,
while the egg yolk that rotated at 2.5 rpm took about 13 min (Figure 12.16[b]).
The inactivation kinetics of S. enteritidis showed a similar trend of observations as obtained with
the effect of rotation on the temperature profile of the egg during pasteurization. The F value was
Stationary egg
5.50e+01
5.44e+01
5.31e+01
5.19e+01
5.00e+01
4.94e+01
4.82e+01
4.69e+01
4.57e+01
4.44e+01
4.32e+01
4.21e+01 (a) 30 s (b) 150 s (c) 300 s
4.07e+01 Rotating egg
3.95e+01
3.83e+01
3.70e+01
3.58e+01
3.45e+01
3.33e+01
3.21e+01
3.08e+01
Temperature (ºC) (a) 30 s (b) 150 s (c) 300 s
FIGURE 12.15 Comparison of CFD simulated temperature contours of stationary (a–c) and egg rotat-
ing at 5 rpm (d–f) heated at 55.6°C. (From R. Ramachandran, D. Malhotra, A. Anishaparvin and
C. Anandharamakrishnan. Computational fluid dynamics simulation studies on pasteurization of egg in sta-
tionary and rotation modes. Innovative Food Science and Emerging Technologies, 12:38–44, 2011.)
calculated for stationary and rotating modes for every one-minute interval of the thermal process.
The egg in the stationary position attained the maximum F value (higher inactivation of Salmonella)
only after 30 min of heating (Figure 12.17[a]), whereas the rotated egg attained it within 7 min and
13 min of heating at 5 rpm at 2.5 rpm rotation, respectively. Nevertheless, inactivation of S. enter-
itidis in the yolk (Figure 12.17[b]) required more time than in the egg white.
12.3.3 CFD Modeling of PEF Pasteurization

CFD modeling of pulsed electric field (PEF) pasteurization is governed by the equation for the
electric potential. The treatment conditions of PEF pasteurization can be accurately modeled by
considering the electrostatic and electrodynamic effects.
12.3.3.1 Governing Equation for Electrostatics

The governing equation for the electric potential is written based on the charge conservation, as
given below [3]:

∇.J = J .∇. (σ (T ).∇.ψ ) = 0 (12.6)


where J . is the current density and ψ is the electric potential. The relationship between electric field
and electric potential is described by:
E = −∇.ψ (12.7)
FIGURE 12.16 Comparison of CFD simulated temperature profiles of stationary and rotated egg conductively
heated at 55.6°C: (a) point-1 (bottom); (b) point-2 (center). (From R. Ramachandran, D. Malhotra, A. Anishaparvin
and C. Anandharamakrishnan. Computational fluid dynamics simulation studies on pasteurization of egg in sta-
tionary and rotation modes. Innovative Food Science and Emerging Technologies, 12:38–44, 2011.)
In the above equation, E is the electric field strength. The use of high voltage electric fields dur-
ing the PEF treatment results in an increase in temperature in the processed food product due to the
Joule (ohmic) heating effect. The rate of volumetric heat generation (ü) within the product is depen-
dent on the electrical conductivity (σ) of the food material and the applied electric field intensity
(∇V), given by Equation (12.8).
2
u = s ∇V (12.8)
The average heating rate of the food product per unit mass due to the Joule heating effect can be
expressed as Equation 12.9:
dT u
= (12.9)
dt ρCP
where, dT, dt, ρ, and Cp denote the temperature increase, time elapsed during the pulse, density, and
heat capacity of the food material, respectively [46].
12.3.3.2 Design and Optimization of PEF Treatment Chamber

Schroeder et al. (2009) [3] developed and validated a CFD model to describe the stationary flow,
electric field and the temperature distributions inside a pilot-scale PEF treatment chamber. NaCl
was considered the model liquid product to be treated by the PEF pasteurization. The geom-
etry of the PEF treatment chamber considered for the CFD modeling is shown in Figure 12.18.
FIGURE 12.17 F value predictions for stationary and rotated egg conductively heated at 55.6°C: (a) top
point (0.025 m from center) and (b) center point. (From R. Ramachandran, D. Malhotra, A. Anishaparvin
and C. Anandharamakrishnan. Computational fluid dynamics simulation studies on pasteurization of egg in
stationary and rotation modes. Innovative Food Science and Emerging Technologies, 12:38–44, 2011.)
The treatment chamber includes a central high voltage electrode made of stainless steel, enclosed
by two polytetrafluoroethylene (PTFE) isolators and two low voltage (ground) electrodes on both
ends. The unsteady state problem was solved using the Finite Element Method. A refined mesh with
triangular grid elements was adopted for the simulation. The governing equations comprised that of
the electric potential (Equation 12.6), fluid flow based on the momentum equation (Equation 12.3)
and heat transfer given by the equation for heat balance (Equation 12.10) as expressed below.
∂( ρ.CP T ) 
+ ∇( ρ vCP T ) = Q + ∇. ( ( k1 + kt )∇T ) (12.10)
∂t
where k is the thermal conductivity, kT is the turbulent thermal conductivity, CP is the specific heat
capacity and Q denotes an external source of heat, which is expressed as
Q = σ ⋅ E (12.11)
where E is the electric field strength and σ is the media conductivity. For the stationary simulations,
the external heat source was multiplied by a factor accounting for the time-averaged potential. The
pulse factor ϕ is determined by relating the averaged high voltage pulse over time to its maximum
value. For simplified rectangular pulses, the factor can be determined as
φ = τ ⋅ f (12.12)
where τ is the pulse width and f is the pulse frequency [46].
FIGURE 12.18 (a) Co-linear PEF treatment chamber in axis-symmetrical configuration; (b) full three-
dimensional configuration. (From S. Schroeder, R. Buckow and K. Knoerzer. Numerical simulation of pulsed
electric fields (PEF) processing for chamber design and optimisation. Seventh International Conference on
CFD in the Minerals and Process Industries CSIRO, Melbourne, Australia, 9th–11th December 2009. www.cfd
.com.au/cfd_conf09/PDFs/062BUC.pdf.)
With respect to the boundary conditions, symmetry on the center-axis (r = 0) was assumed. The
conductive media application was limited to the liquid part of the model. For the inlet, outlet and
isolator gap, the electric isolation was assumed as given below:

nJ = 0 (12.13)
At the high voltage electrode, the electric potential was defined as V = V0, where V0 is the maxi-
mum potential of the pulse. The other electrodes were considered grounded with zero potential (V = 0).
The thermo-physical properties of the NaCl solution, such as the density, specific heat capacity,
thermal conductivity, and viscosity, were considered as a function of the temperature. The material
properties of the electrode and the isolators were considered as given below:
Electrode : k = 0.24 W/mK; ρ = 2200 kg /m 3; CP = 1050 J/kgK

Isolators : k = 44.5W/mK , ρ = 7850 kg /m 3; CP = 475J/kgK

The CFD simulations showed that the geometry of the isolator had a significant effect on the elec-
tric field distribution inside the PEF treatment chamber (Figure 12.19). The insertion of the isolator
FIGURE 12.19 Predicted distribution of the electric field strength in the treatment chamber for five different
isolator geometries. (From S. Schroeder, R. Buckow and K. Knoerzer. Numerical simulation of pulsed electric
fields (PEF) processing for chamber design and optimisation. Seventh International Conference on CFD in
the Minerals and Process Industries CSIRO, Melbourne, Australia, 9th–11th December 2009. www.cfd.com
.au/cfd_conf09/PDFs/062BUC.pdf.)
in the treatment zone was found to increase the average electric field strength and improve the
uniformity of the PEF treatment. In the standard chamber (Figure 12.19[a]), a non-uniform electric
field was observed as characterized by high field intensity at the edges and low field intensity in the
center of the isolators. However, reducing the inner diameter of the isolator bore (Figure 12.19[b])
improved the uniformity, but depicted peak intensity at the edges of the isolator. Rounded isolator
edges (Figure 12.19[c]) significantly improved the distribution of the electric field. Further improve-
ments in the uniformity of the electric field resulted by modifying the treatment chamber with an
isolator in the elliptical form, as shown in Figure 12.19[e].
Isolator geometry also exerted an influence on the turbulence characteristics. Turbulence dur-
ing PEF processing allows a good mixing of the fluid and reduces the level and non-uniformities
of temperature in the chamber. The CFD simulation predictions of the turbulent kinetic energy of
different treatment chamber geometries are shown in Figure 12.20. The flow velocity peaked on
the axis and decreased towards the wall of the treatment chamber. In all cases, the fluid enters the
chamber at the base with high velocity (5 m/s). The chamber design shown in Figure 12.20(a) caused
the least turbulence compared to other geometries. The isolator did not reduce the pipe diameter
and thus led to increased velocities and Reynolds numbers. The turbulence intensity drastically
increased to 18 m2/s2 when the diameter of the isolator was reduced to 3 mm (Figure 12.20[b]). On
the other hand, non-uniformity of the turbulence occurred due to the sharp edges. Rounded isolator
edges (Figure 12.20[c]) provided better uniformity of the turbulence, but certain zones in the cham-
ber depicted poor mixing of the media due to flow recirculation areas. Elliptical isolator geometry
(Figure 12.20[d]) resulted in a more uniform pattern of the turbulent kinetic energy and avoided the
dead zones of poor media mixing.
The temperature distribution during the PEF treatment depends on the distribution of the elec-
tric field and the turbulence pattern in the chamber. The CFD simulation of the effect of isolator
geometry on the temperature profile (Figure 12.21) showed that heating of the liquid was highest
in the treatment chamber geometry (Figure 12.21[a]), which was largely due to the heterogenic
electric field and the low turbulence. Very high field strengths at the edges of the isolator cause
a drastic heating of the media and increased the conductivity of the media and the energy input
in this spot. This phenomenon is termed the “thermal runaway heating.” The low turbulence in
FIGURE 12.20 Predicted distribution of the turbulent kinetic energy in the treatment chamber for five dif-
ferent isolator geometries. (From S. Schroeder, R. Buckow and K. Knoerzer. Numerical simulation of pulsed
electric fields (PEF) processing for chamber design and optimisation. Seventh International Conference on
CFD in the Minerals and Process Industries CSIRO, Melbourne, Australia, 9th–11th December 2009. www
.cfd.com.au/cfd_conf09/PDFs/062BUC.pdf.)
the chamber caused the fluid near the wall to move with a lower velocity compared to the center
of the pipe. This led to an enhanced heating of the media near the chamber wall. Due to con-
duction heating, the turbulence increased the temperature of the isolators and electrodes. Under
stationary conditions, the total temperature increase with the isolator geometry was 21 K. But the
geometry shown in Figure 12.21(b) caused a total heating of 6 K in the media, attributed to the
high turbulence kinetic energy observed (Figure 12.20[b]). The heating of the media was simi-
lar in the chambers with isolator geometry, as shown in Figure 12.21(b) and (c). The total tem-
perature increase in the cases represented in Figure 12.21(d) and (e) was higher—approximately
12 K and 9 K, respectively. Although the turbulence in these treatment chambers was less pro-
nounced than in Figure 12.21(b) and (c), these isolator geometries resulted in a uniform electric
field distribution.
FIGURE 12.21 Predicted temperature distribution in the treatment chamber for five different isolator geom-
etries (From S. Schroeder, R. Buckow and K. Knoerzer. Numerical simulation of pulsed electric fields (PEF)
processing for chamber design and optimisation. Seventh International Conference on CFD in the Minerals
and Process Industries CSIRO, Melbourne, Australia, 9th–11th December 2009. www.cfd.com.au/cfd_conf09
/PDFs/062BUC.pdf.)
Thus, the CFD simulations showed that minute modifications in the isolator geometry can reduce
the temperature rise in the treated media by up to 70%. This could potentially decrease the detri-
mental effects on the nutritional and sensorial characteristics of the treated food while facilitating
an energy-efficient operation of the PEF pasteurization process of liquid foods.
12.4 CONCLUSIONS
The food pasteurization process, whether thermal or non-thermal, involves complexities with
respect to real-time investigations owing to the difficulty of measuring temperature without hinder-
ing the fluid flow during an industrial-scale operation, extremely short processing time and intricate
geometries. Computational fluid dynamics simulations aid food processing engineers in predicting
the temperature profile and its uniformity of distribution within the product during the pasteuriza-
tion process. In addition, it is possible to locate the slowest heating zone (SHZ) and study the inac-
tivation kinetics of a spoilage/pathogenic microorganism involved by integrating the equation for
F value in the governing equation of the CFD model. This chapter has presented the capability of
CFD modeling and simulation to predict the temperature profile during thermal pasteurization and
optimize the design of the treatment chamber for treatment of liquid foods using pulsed electric field
(PEF) pasteurization. Three-dimensional CFD models with volume mesh have been established
and solved by the finite volume approach to simulate the temperature profiles and locate the SHZ
during the thermal pasteurization process. The CFD simulation predictions were found to dem-
onstrate reasonable agreement with the experimental measurements during model validation. The
models were also used to estimate the effect of the rotation of the product or the product-containing
can on the effectiveness of thermal pasteurization and the corresponding inactivation kinetics of
the spoilage/pathogenic microorganism. The CFD simulations established the positive effects of
rotational mode pasteurization leading to reduced processing time. In the case of PEF, the elec-
trode geometry was found to have a profound influence on the distribution of electric field strength,
turbulence in the chamber and temperature profile. A significant improvement in the uniformity of
electric field and the turbulence pattern of flow was observed upon the reduction of the inner diam-
eter of the isolator and by modifying its edges to a rounded or an elliptical shape. The information
acquired from the CFD simulations would be very useful in optimizing the time-temperature com-
bination of industrial-scale food pasteurization processes to reduce the processing time by avoiding
slow heating zones within the product.
REFERENCES
1. National Advisory Committee on Microbiological Criteria for Foods (NACMF), Requisite scientific
parameters for establishing the equivalence of alternative methods of pasteurization, Journal of Food
Protection, 69(5): 1190–1216, 2006.
2. Scott G., Richardson P., The application of computational fluid dynamics in the food industry, Trends in
Food Science and Technology, 8: 119–124, 1997.
3. Schroeder S., Buckow R., Knoerzer K., Numerical simulation of pulsed electric fields (PEF) processing
for chamber design and optimisation, In: Seventh International Conference on CFD in the Minerals and
Process Industries CSIRO, Melbourne, Australia, 9th–11th December 2009. www.cfd.com.au/cfd
_conf09/PDFs/062BUC.pdf.
4. Stumbo C.R., Thermal resistance of bacteria, In: Thermobacteriology in Food Processing, 2nd ed.,
Academic Press, Cambridge, MA, USA, 93–120, 1973.
5. Buzrul S., A suitable model of microbial survival curves for beer pasteurization, LWT-Food Science and
Technology, 40(8): 1330–1336, 2007.
6. Dilay E., Vargas J.V.C., Amico S.C., Ordonez J.C., Modeling, simulation and optimization of a beer
pasteurization tunnel, Journal of Food Engineering, 77(3): 500–513, 2006.
7. Bhuvaneswari E., Anandharamakrishnan C., Heat transfer analysis of pasteurization of bottled beer
in a tunnel pasteurizer using computational fluid dynamics, Innovative Food Science and Emerging
Technologies, 23: 156–163, 2014.
8. Public Health Service, US Food and Drug Administration, Grade “A” Pasteurized Milk Ordinance,
Publication No. 229, 1999.
9. Dignan D.M., Barry M.R., Pflug I.J., Gardine T.D., Safety considerations in establishing aseptic pro-
cesses for low-acid foods containing particulates, Food Technology, 43(3): 118–121, 1989.
10. Mohamed M.E.A., Eisa A.H.A., Pulsed electric fields for food processing technology (Chapter 11), In:
Structure and Function of Food Engineering, IntechOpen Ltd., London, UK, 2012, http://www.intecho-
pen.com/books/structure-and-function-of-foodengineering.
11. Zimmermann U., Benz R., Dependence of the electrical breakdown voltage on the charging time in
Valonia utricularis, Journal of Membrane Biology, 53(1): 33–43, 1980.
12. Dunn J., Pulsed electric field processing: An overview, In: Pulsed Electric Fields in Food Processing:
Fundamental Aspects and Applications, Barbosa-Canovas G.V., Zhang Q.H., Eds., Technomic Press,
Lancaster, PA, USA, 1–30, 2001.
13. Singh I., Pulsed electric field processing of foods, International Journal of Farm Sciences, 2(1): 1–16,
2012.
14. Yeom H.W., Evrendilek G.A., Jin Z.T., Zhang Q.H., Processing of yoghurt based products with pulse elec-
tric field: Microbial, sensory and physical evaluations, Journal of Food Processing and Preservation,
28(3): 161–178, 2007.
15. Kumar Y., Patel K.K., Kumar K., Pulse electric field processing in food technology, International
Journal of Engineering Studies and Technical Approach, 1(2): 6–17, 2015.
16. Anandharamakrishnan C., Computational fluid dynamics (CFD)—Applications for the food industry,
Indian Food Industry, 22(6): 62–68, 2003.
17. Norton T., Sun D.-W., Computational fluid dynamics (CFD)—An effective and efficient design and
analysis tool for the food industry: A review, Trends in Food Science and Technology, 17(11): 600–620,
2006.
18. Fletcher A.J., Computational Techniques for Fluid Dynamics, Springer-Verlag, New York, USA, 2000.
19. Anandharamakrishnan C., Computational Fluid Dynamics Applications in Food Processing,
Springer-Verlag, New York, USA, 2013.
20. Anandharamakrishnan C., Experimental and computational fluid dynamics studies on spray-freeze-
drying and spray-drying of proteins, PhD thesis, Loughborough University, UK, 2008.
21. Ghani A.G.A., Farid M.M., Chen X.D., Richards P., Numerical simulation of natural convection heating
of canned food by computational fluid dynamics, Journal of Food Engineering, 41(1): 55–64, 1999.
22. Varma M.N., Kannan A., CFD studies on natural convective heating of canned food in conical and
cylindrical containers, Journal of Food Engineering, 77(4): 1024–1036, 2006.
23. Kannan A., Sandaka P.Ch.G., Heat transfer analysis of canned food sterilization in a still-retort, Journal
24. Chen C.R., Ramaswamy H.S., Visual basic computer simulation package for thermal process calcula-
tions, Chemical Engineering and Processing: Process Intensification, 46(7): 603–613, 2007.
25. Stoforos N.G., Merson R.L., Estimating heat transfer coefficients in liquid/particulate canned food using
only liquid temperature data, Journal of Food Science, 55(2): 478–483, 1990.
26. Puri V.M., Anantheswaran R.C., The finite-element method in food processing: A review, Journal of
27. Ghani A.A.G., Farid M.M., Chen X.D., A computational and experimental study of heating and cooling
cycles during thermal sterilization of liquid foods in pouches using CFD, Proceedings of the Institution
of Mechanical Engineers, Part E, Journal of Process Mechanical Engineering, 217(1): 1–9, 2003.
28. Ghani A.A.G., Farid M.M., Chen X.D., Theoretical and experimental investigation of the thermal inac-
tivation of Bacillus stearothermophilus in food pouches, Journal of Food Engineering, 51(3): 221–228,
2002.
29. Walstra P., Wouters J., Geurts T., Dairy Science and Technology, 2nd ed., CRC Press, Boca Raton, FL,
USA, 2006.
30. Anand Paul D., Anishaparvin A., Anandharamakrishnan C., Computational fluid dynamics studies on
pasteurisation of canned milk, International Journal of Dairy Technology, 64(2): 305–313, 2011.
31. Shapton D.A., Lovelock D.W., Laurita-Longo R., The evaluation of sterilization and pasteurization pro-
cesses for temperature measurements in degrees Celsius (°C), Journal of Applied Bacteriology, 34(2):
491–500, 1971.
32. Holdsworth S.D., Simpson R., Sterilization, pasteurization and cooking criteria, In: Thermal Processing
of Packaged Foods, 3rd ed., Holdsworth S.D., Simpson R., Eds., Springer-Verlag, New York, USA,
123–138, 2007.
33. Riedel L., Thermal conductivity measurements of sugar solutions, fruit juice and milk, Chemie-
Ingenieur-Technik, 21(17): 340–341, 1949.
34. Short A.L., The temperature coefficient of expansion of raw milk, Journal of Dairy Research, 22(1):
69–73, 1955.
35. Buzrul S., Alpas H., Bozoglu F., Effect of high hydrostatic pressure on quality parameter of lager beer,
Journal of the Science of Food and Agriculture, 85(10): 1672–1676, 2005.
36. Briggs E.D., Boulton C.A., Brookes P.A., Stevens R., Brewing: Science and Practice, Woodhead
Publishing, Sawston, Cambridgeshire, UK, 2004.
37. Augusto P.E.D., Pinheiro T.F., Cristianini M., Using computational fluid dynamics (CFD) for the evalu-
ation of beer pasteurization: Effect of orientation of cans, Ciência e Tecnologia de Alimentos, 30(4):
980–986, 2010.
38. Romero T., Cabral R.A.F., Telis V.R.N., Gabas A.L., Effect of corrugation on the hydrodynamic and
heat transfer behavior of beer flowing through rectangular corrugated channels, In: International
Conference on Engineering and Food, 9, 2004.
39. Erdogdu F., Tutar M., Velocity and temperature field characteristics of water and air during natural
convection heating in cans, Journal of Food Science, 76(1): E119–E129, 2011.
40. Pinho C., Marcelo C., Three-dimensional mathematical modeling of microbiological destruction of
Bacillus stearothermophilus in conductive baby food packed in glass container, International Journal
of Food Engineering, 1, 2005.
41. Bailey R.T., Elban W.L., Thermal performance of aluminum and glass beer bottles, Heat Transfer
Engineering, 29(7): 643–650, 2008.
42. Denys S., Pieters J.G., Dewettinck K., Combined CFD and experimental approach for determination of
the surface heat transfer coefficient during thermal processing of eggs, Journal of Food Science, 68(3):
943–951, 2003.
43. Denys S., Pieters J.G., Dewettinck K., Computational fluid dynamics analysis of combined conductive
and convective heat transfer in model eggs, Journal of Food Engineering, 63(3): 281–290, 2004.
44. Denys S., Dewettinck K., Pieters J.G., CFD analysis for process impact assessment during thermal pas-
teurization of intact eggs, Journal of Food Protection, 68(2): 366–374, 2005.
45. Ramachandran R., Malhotra D., Anishaparvin A., Anandharamakrishnan C., Computational fluid
dynamics simulation studies on pasteurization of egg in stationary and rotation modes, Innovative
Food Science and Emerging Technologies, 12: 38–44, 2011.
46. Salengke S., Sastry S.K., Zhang H.Q., Pulsed electric field technology: Modeling of electric field
and temperature distributions within continuous flow PEF treatment chamber, International Food
Research Journal, 19(3): 1137–1144, 2012.
13 CFD Determination of
F Values During Thermal
Processing of Still Cans
Won Byong Yoon and Hyeon Woo Park
CONTENTS
13.1 Introduction..................................................................................................................... 289
13.2 Sterilization of Still Cans............................................................................................................. 290
13.2.1 Analysis of Fluid Flow Patterns in Cans During Thermal Sterilization Process...... 290
13.2.2 F Value.................................................................................................................... 290
13.2.3 Reaction Kinetics of Quality Changes.................................................................... 292
13.3 Applications of CFD Model.......................................................................................................293
13.3.1 CFD Modeling................................................................................................................. 293
13.3.1.1 Modeling Thermal Sterilization for Liquid Foods in Cans..................... 293
13.3.1.2 Assumptions Used in Numerical Simulation............................................ 294
13.3.1.3 Governing Equations and Boundary Conditions..................................... 294
13.3.1.4 Results of Simulation of Thermal Sterilization for Liquid Foods
in a Vertical Can...................................................................................... 295
13.3.1.5 Influence of Geometry and Orientation......................................................296
13.3.2 Thermal Sterilization for Solid-Liquid Food Mixtures in Cans............................. 297
13.4 Examples of CFD Application to Determine F Values During Thermal Processing.......... 299
13.4.1 Prediction of Slow Heating Point During Thermal Processing.............................. 299
13.4.2 Determination of F Value Using Subzones............................................................ 299
13.4.3 F Value Based on Degradation of Spores........................................................... 301
13.5 Conclusions..........................................................................................................................302
References.......................................................................................................................................302
13.1 INTRODUCTION
Thermal processing in the food industry is defined as application of heat to assure food safety and
extend shelf-life of processed food [1]. Canning is the most widely used preservation technology
for many kinds of long shelf-life foods, such as fruits, vegetables, milk, meat, and poultry. Despite
significant advances in food preservation techniques during the last several decades, the canning
process is still considered the most effective preservation technology, especially for developing
shelf-stable products. Hot water under certain pressure or steam is usually applied to food in a con-
tainer to provide enough heat to kill microorganisms.
The most challenging aspect of designing a canning process is providing adequate heat treat-
ment to ensure that the slowest heating zone (SHZ) within a container receives critical heat for a
specific period of time to inactivate microorganisms to avoid detrimental results. However, ther-
mal treatment not only destroys microbes, but also degrades food quality and nutritive properties
due to excessive heating [2]. Minimizing loss of quality and nutrients during thermal processing
while providing the desired sterility is one of the most important goals of the canning industry [3].
Hence, it is necessary to predict and quantify heat transfer in cans correctly to develop a process
289
that not only ensures safety, but also minimizes overcooking. Therefore, it is important to under-
stand the heat transfer phenomenon in thermal sterilization of food in cans. To ensure safe thermal
processing, it is also necessary to understand heat transfer mechanisms and proper predictive
methodologies to choose the proper processing conditions [4].
However, it is difficult to develop suitable mathematical models for pasteurization or steriliza-
tion of canned foods. This is mainly due to natural convection, which is the dominant mechanism
for transferring heat inside cans. Because fluid motion is dependent on buoyancy forces, momen-
tum equations must be solved simultaneously with the energy equation to calculate the velocity
profile and temperature profile, both of which are needed to determine the SHZ of the can [5].
Since the cold point will continuously move during natural convection, SHZ has been widely used
as a reference zone to determine the degree of pasteurization or sterility. The F value is generally
used as an indicator representing the degree of sterilization during the thermal process. The F
value of a thermal process must be estimated based on the heat penetration curve at the cold point.
Numerical prediction of the transient temperature and velocity profiles during natural convection
heating of canned liquid foods was initiated by Datta and Teixeira [6]. Kumar et al. [7] and Kumar
and Bhattacharya [8] then simulated thermal processing of viscous liquid foods in a vertical can.
Numerical simulation methods such as computational fluid dynamics (CFD) have been improved to
develop and validate mathematical models for natural convection heating associated with canned
foods [9–17]. This review presents the results of previous studies on CFD determination of F val-
ues during thermal processing of still cans. Estimating F values of a thermal process based on the
cold point accurately determined by using simulations and experiments is an essential step when
designing a suitable thermal process for canned foods to ensure food safety while minimizing loss
of quality and nutrients.
13.2 STERILIZATION OF STILL CANS

13.2.1 Analysis of Fluid Flow Patterns in Cans During Thermal Sterilization Process
During natural convection heating of canned liquid food products, fluid movement will accelerate
the sterilization process by enhancing the heat transfer rate. Hence, heat transfer by convection is
governed by both fluid motion and temperature difference [18]. CFD and numerical simulation are
commonly applied in the prediction of flow patterns and temperatures during thermal processing of
foods [19]. During natural convection heating, the velocity in momentum equations is coupled with
temperature in the energy equation because the movement of fluid is solely due to buoyancy force.
In thermal sterilization of canned liquid foods, the liquid near the wall is initially at rest due
to application of a no-slip boundary condition. When the steam has contact with the outer wall of
the can later, the liquid adjacent to the wall will heat up to the wall temperature by conduction.
However, liquid away from the wall is still at the initial temperature. Due to gravity and variation
in liquid density throughout the can, buoyancy forces are created. Throughout the heating period,
buoyancy forces are opposed by the viscous forces of the liquid. The velocity of the convective cur-
rent is dependent upon the strength of the buoyancy forces and the magnitude of liquid viscosity’s
resistance to flow. When heating progresses, temperature-dependent viscosity decreases, resulting
in increased velocity, which results in faster heating of the product. As the temperature inside the
container becomes uniform later, buoyancy forces will decrease, leading to significant reduction in
velocity and cessation of recirculation [8].
13.2.2 F Value
Thermal destruction of microorganisms takes place logarithmically. Therefore, a sterile product
cannot be produced with certainty regardless of the process time. However, the probability of sur-
vival of a single microorganism can be predicted based on heat resistance of the microorganism,
CFD Determination of F Values During Thermal Processing of Still Cans 291
temperature distribution, and heating time duration. This gives rise to a concept known as commer-
cial sterility. In canning factories, accelerated storage of randomly selected cans of food is needed
to ensure that commercial sterility is maintained before foods are released for retail sale.
The destruction rate of a microorganism is usually taken as a first order. Generally, the reduction
pattern of a microorganism in thermal processes shows a logarithmic reduction. When microbial
populations are presented on semi-logarithmic coordinates as a function of time on standard coor-
dinates, a linear decrease in microbial population with time is observed at a given temperature. This
is known as logarithmic order of death. It can be described by a death rate curve or survivor curve.
A typical relationship for a microbial spore population versus time on semi-logarithmic coordinates
is shown in Figure 13.1.
The time needed to destroy 90% of microorganisms (to reduce their numbers by a factor
of 10) is defined as the decimal reduction time (DT). It is the time required for a one-log cycle
reduction in microbial population. The DT value has been used throughout thermal processing
literature to quantify the effect of temperature on microbial populations. A plot of decimal
reduction time as a function of temperature on semi-log coordinates results in a linear relation-
ship, which is referred to as the thermal death time (TDT) curve for a given microbial popula-
tion (Figure 13.2).
The slope of the TDT curve is termed the Z value, which is defined as the temperature increase
required to cause a one-log cycle reduction in the decimal reduction time. It is also defined as the
temperature difference that causes a ten-fold change in the rate of microbial destruction. The value
of Z can be used to estimate the effect of changing temperature on processing time.
One of the most important parameters in thermal processing is the F value. It is defined as the
time required to destroy a given percentage of microorganisms at a given temperature and Z value.
This parameter states the reduction in the population of microbial pathogens required to ensure
product safety. A sterilization value or Fref value is used to describe processes that operate at a single
reference temperature (Tref), which is based on a microorganism with a Z value of 10°C. It is defined
by the following equation [20]:
t T (t ) − Tref

Fref =
∫ 10
0 z
dt (13.1)
FIGURE 13.1 Death rate curve of a microbial population.

FIGURE 13.2 Thermal death time curve of a microbial population.
where t is the time for thermal treatment, T(t) is the temperature at time t, and Tref is the reference
temperature. To achieve successful sterilization, this value must be greater than or equal to the
required F value. This approach is widely used in the food industry.
13.2.3 Reaction Kinetics of Quality Changes

Thermal sterilization of food causes biochemical changes in food quality, including positive impact
(such as reduction in the microbial population) and negative impact (such as the loss of nutrients),
as discussed earlier. Most reactions occurring in food are represented by simple reaction kinetics.
Thermal destruction of microorganisms, most nutrients, quality factors (texture, color, and flavor),
and enzymes generally follow first-order reaction kinetics. That is, the destruction rate of each of
these components is linearly dependent on the concentration of the component. This relationship
is frequently referred to as a “logarithmic order of inactivation or destruction” since each compo-
nent is not dependent on its initial concentration. The first-order response describing these changes
occurring during food processing can be expressed mathematically as follows:
dC
− = kT C (13.2)
dt
where C is the value of quality attribute, t is the time duration of exposure, and kT is reaction rate constant.
In fact, most changes occurring during thermal processing of food are more complex [21]. However, it
has been accepted in practice that relatively simple expressions with parameters similar to rate constant
(kT) can be used to describe quality changes in food. Integrating Equation 13.2 gives the following:
C
Ln = − kT t (13.3)
C0
where C0 is the initial value of quality attribute. From the values of kT and the initial value of the
quality attribute, the amount of the attribute remaining can be estimated for any time of exposure
to a process.
The relationship between microbial populations and time is very similar to the relationship
used to describe kinetic parameters in first-order chemical reactions described above. kT is used to
describe changes in concentration of a reactant as a function of time. In microbial populations, the

DT value is used to describe the same relationship. It follows the relationship between kT and DT as
follows:
2.303
kT = (13.4)
DT
There are two principal methods to describe the dependence of reaction rate constant on tem-
perature: (1) TDT curves, and (2) the Arrhenius equation. The dependence of reaction rate constant
on temperature can be described by the following Arrhenius equation:
 − Ea 
kT = A exp   (13.5)
 Rg T 

where A is the preexponential factor, Ea is the activation energy, and Rg is the universal gas constant.
Thus, DT and F values are valuable as they provide information on how fast a specific microor-
ganism can be destroyed. They can also be used to model the effect of temperature on the rate of
destruction. Changes in food quality can be modeled with both time and temperature as mentioned
above. However, none of these approaches mentioned above can be directly used to determine the
required sterilization conditions for cans or pouches. This is because there are large variations in
temperature with position in these containers. To achieve food safety and quality of canning prod-
uct, CFD should be used to provide numerical solutions of partial differential equations describing
fluid motion and heat transfer.
13.3 APPLICATIONS OF CFD MODEL

13.3.1 CFD Modeling
13.3.1.1 Modeling Thermal Sterilization for Liquid Foods in Cans
Numerical solutions have been applied to sterilization problem, mostly focusing on solid foods or
very viscous foods such as purees and concentrates usually assumed to be heated by pure conduc-
tion. One of the earliest applications of numerical methods in thermal processing of foods was
reported by Teixeira et al. [22]. Mathematical analysis in sterilization of stationary cans containing
liquid foods is more difficult because heat is transferred inside the can by natural convection in
such cases. Due to fluid motion caused by buoyancy force, the velocity in momentum equations is
coupled with temperature in the energy equation. Therefore, momentum equations must be solved
simultaneously with the energy equation in order to determine the temperature profile, velocity
profile, and slowest heating zone of the can.
Sterilization of liquid food in a can sitting in an upright position and heated at 121.1°C from all
sides was simulated and results were presented. Water was used as a model liquid. Computations
were performed for a can with a radius of 26.6 mm and height of 124.0 mm. Governing equations
for continuity, momentum, and energy conservation were solved using a commercial CFD package
(ANSYS 18.0). Transient flow patterns and temperature profiles within the model liquid (water)
were simulated. Content in a can or system was meshed in three dimensions into 20,991 small
control volumes.
Including headspace in the model would require analysis of air–water interface problem, thus
adding complexities to the given computational problem. However, Erdogdu et al. [10] and Kiziltas
et al. [12] have neglected the effect of headspace (less than 5% in volume) without having any
significant impact on simulation results. Various other studies [23,24] in food engineering literature
did not account for the effect of headspace on heat transfer rate either.
In simulating water as a model liquid for low-viscosity beverages, total heating time was divided
into 1,200 time-steps. Higher iteration numbers are required to improve accuracy due to the low
viscosity value of water and its high temperature dependency [12]. This was especially significant at
initial stages of heating to solve momentum equations for correct velocity profiles. The time interval
was progressively increased with heating. Simulation was conducted with 600 steps to achieve the
first 60 s of heating and 1,200 steps for 300 s of total heating.
13.3.1.2 Assumptions Used in Numerical Simulation

The initial temperature of the system was set at 20°C while the heating medium temperature was set
at 121.1°C. Thermal and physical properties of water were used as a function of temperature with
fitted equations while buoyancy force caused by density variation with temperature was governed
by the Boussinesq approximation (Table 13.1).
To simplify the problem, the following assumptions were made:
1. Heat generation due to viscous dissipation was negligible.

2. Thermal and physical properties of can were constants (Table 13.1).
3. Assumption of no-slip condition at the inside wall of the can was valid.
4. Condensing steam maintained a constant temperature at the outer surface of the can.
13.3.1.3 Governing Equations and Boundary Conditions

Partial differential equations governing natural convection motion in a cylindrical space included
the Navier–Stokes equation in cylindrical coordinates [6]. The following initial and boundary con-
ditions were used:
At t = 0,
T0 = Twall = 20°C, vr = vθ = vz = 0 m /s, at 0 ≤ r ≤ R and 0 ≤ z ≤ H

At t > 0,
Twall = Tm = 121.1°C, vr = vθ = vz = 0 m /s, at 0 ≤ r ≤ R and 0 ≤ z ≤ H

TABLE 13.1
Physical Properties of Water and Aluminum Can Used in the Model
Property Water Aluminum Can
Density (kg/m3) 998.2 2,719
Thermal conductivity (W/m·K)a −0.73 + 7.32 × 10−3T−9.49 × 10−6T2 202.4
Specific heat (J/kg·K)a 3.99 × 10−4–423.18T + 1.88T2−3.72 × 10−3T3 + 2.76 × 10−6T4 871
Viscosity (Pa·s)a 0.0054−0.2 × 10−4T + 0.2 × 10−7T2
Thermal expansion coefficient 0.0002
(1/K)b
Source Denys et al. [25] ANSYS V18.0
a T in K.
b T in °C.
Laminar flow mode was assumed to occur throughout the entire heating process. This resulted
in oscillations during the first few seconds of the simulation. Because oscillations did not continue,
turbulent flow mode was not used. Oscillations that did occur were disregarded.
13.3.1.4 Results of Simulation of Thermal Sterilization for Liquid Foods in a Vertical Can

Figure 13.3 shows temperature profiles and flow patterns of water in a can after 5, 20, 200, and 400 s
of heating. In the beginning of the process, heating of the extremities resulted in a large temperature
and density gradient, leading to fluid movement in random distribution and characteristic of flows.
When the liquid in the can was heated, it expanded and became less dense. Liquid away from
the side wall remained at a much lower temperature. The buoyancy force created by the change in
liquid density due to temperature variation (from the wall to the core) produced an upward flow near
the side wall. The rising hot liquid was deflected by the top wall. It then travelled radially toward
the core. The heavier liquid in the core moved downward first. It then moved toward the wall. Thus,
a recirculating flow was created. Because of these high velocities, the SHZ kept moving during
FIGURE 13.3 (a) Temperature and (b) velocity profiles in a vertical cylindrical can during thermal processing.
heating. As the heating progressed, a more uniform velocity was obtained, reducing the buoyancy
force in the liquid and significantly reducing the velocity.
13.3.1.5 Influence of Geometry and Orientation

The characteristics of liquid flow inside the can during heating are a function of its geometry. Even
very small alterations can result in differences in the process. CFD modeling studies have been
extended to modifications of thermal processing of foods through changes in geometry or orienta-
tion of its packaging.
Movement of SHZ during heating for cylindrical and conical containers with non-Newtonian
fluid has been investigated by Varma and Kannan [26]. They have found that not only geometry
but also orientation of these containers can significantly influence the thermal sterilization pro-
cess. Boz and Erdogdu [24] have simulated temperature profiles and velocity fields in horizontal
can using 2D approach. Augusto et al. [27] have also studied the effect of orientation of beer cans
on Pasteurization Unit. Their results have demonstrated that the orientation of beer cans (such as
conventional, inverted, and horizontal orientation) can significantly affect thermal profiles and flow
patterns in these cans. Thus, the temperature and velocity profile inside the cans solved by momen-
tum and energy equations can be significantly different depending on the geometry and orientation
of the cans.
13.3.1.5.1 Results of Simulation

Figure 13.4 shows temperature profiles and velocity vector of water in a 3-D cylindrical can
(radius = 26.6 mm; height = 124.0 mm) lying horizontally after 30 s of heating. The velocity of
the fluid at 30 s ranged from 3.19 mm/s to 0.03 mm/s, which was about three times lower than that
observed in a vertical can (ranging from 9.03 mm/s to 0.12 mm/s). This was due to lower Grashof
number in the horizontal can since the characteristic dimension of Grashof number in this case was
FIGURE 13.4 (a) Temperature and (b) velocity profiles in a horizontal cylindrical can after 30 s of heating.
FIGURE 13.5 Temperature profiles in a conical can pointing (a) upward and (b) downward during thermal
processing.
equal to the can diameter, which was smaller than its height. The general behavior of fluid motion in
the horizontal can was similar to that in the vertical one (Section 3.1.4). The buoyancy force created
by the change in liquid density due to temperature variation (from the wall to the core) produced an
upward flow near the side wall. The hot liquid going up was deflected at the middle of the top wall.
It then travelled radially toward the core. Figure 13.4a shows temperature profiles of the same can
at a vertical plane and an angular plane. This figure clearly shows the actual shape of the SHZ. The
SHZ settled down at the lower half of the can as natural convection developed.
Temperature profiles of the conical can (radius = 53.2 mm; height = 124.0) at two different
orientations are shown in Figure 13.5. Due to variation in flow patterns caused by natural convec-
tion and geometry, the hot region penetrated the center of the core faster than it did its periphery.
It was evident that the cone pointing upwards was heated faster than the cone pointing downwards.
Figure 13.6 shows the effect of orientations on velocity profiles of the conical can. In a body heated
uniformly at the surface containing initially cold liquid, temperature differences between the wall
and the core in turn resulted in density differences that induced buoyancy forces in the liquid, result-
ing in natural convection flow patterns. Buoyancy forces were initially increased due to widening
temperature differences between the wall and the core. They were subsequently decreased due to
gradual evening out of temperature gradients. For the cone pointing upwards, the position of the
SHZ was near the geometric center initially. It then moved axially downward. It then moved radi-
ally, finally settling at one location. For the cone pointing downwards, there was only axial move-
ment of the position of the SHZ without radial movement. For both cones (pointing upwards and
downwards), the SHZ settled near the bottom at the height around 10% of total axial distance. Since
the bottom of the can was also heated and the hot fluid in contact with this wall was circulating,
further axial movement of the SHZ was restricted.
13.3.2 Thermal Sterilization for Solid-Liquid Food Mixtures in Cans

For solid food with conduction heating, the location of the SHZ can be determined experimentally
[28] since it always lies at the geometric center of the can. On the other hand, determination of the
SHZ for liquid food is a difficult task. It requires numerical solutions of partial differential equa-
tions describing fluid motion and heat transfer as discussed earlier.
FIGURE 13.6 Velocity profiles in a conical can pointing (a) upward and (b) downward during thermal
processing.
Solid in a mixture of solid-liquid food is heated by conduction while heat transfer in liquid
is dominated by natural convection [7]. However, heating time is usually calculated based on
the assumption that solid-liquid heats up by pure conduction. Ignoring the effect of natural convec-
tion on liquid leads to much longer heating times than needed, resulting in over-heated products
[29].
Recently, CFD modeling studies have been extended to sterilization of canned solid-liquid
food mixtures. The presence of solids has been found to influence the velocity profile and the
position of the SHZ inside the can. Unlike liquid foods, solid-liquid food mixtures are assumed
to be heated by both conduction and convection [5,30–32]. In the case of pineapple slices canned
in liquid sucrose solution, liquid has been assumed to be heated by natural convection exhibiting
recirculation phenomenon [33], while solid food (pineapple slices) has been presumed to be heated
by conduction. Padmavati and Anandharamakrishnan [32] have also studied CFD modeling of
canned pineapple slices. Their results have shown that the position of the SHZ of the entire product
is concentrated on pineapple slices rather than at the geometric center of the can where sugar solu-
tion is present. For solid-liquid mixtures, solids are uniformly distributed in the liquid of canned
peas [12]. They have observed slight changes in the velocity profile. This is due to heat exchange
and surface deflections when the flow is slowly moving through the stack of solid particles. For
canned fruit salad filled with only 16.7% sugar syrup, the F value is still greatly influenced by the
natural convection motion of the syrup [30]. The SHZ of canned solid-liquid food varied in com-
parison with that of canned liquid sterilization. Such an effect is expected as the presence of solids
reduces the effect of natural convection in the fluid, causing the SHZ to move slightly upwards. On
the other hand, in the case of asparagus canned in brine, Dimou and Yanniotis [5] found that nei-
ther the shape of the solid nor the available space between the asparagus restricted the flow of the
brine. Thus, the SHZ was found to lie toward the bottom. A similar result was observed by Kiziltas
et al. [12] for canned peas in water. The SHZ of beef fat in water under a high-pressure processing
unit was studied by Ghani and Farid [34] and a forced convection phenomenon was observed in
the liquid component. Solid foods need to be heated more than liquids due to differences in their
compression heating coefficient.
13.4 EXAMPLES OF CFD APPLICATION TO DETERMINE

F VALUES DURING THERMAL PROCESSING
Because heat transfers occurring in liquid foods and solid-liquid food mixtures during thermal
treatments are very complicated due to combined mechanisms of conduction and convection, SHZs
are continuously changing during thermal treatments. Numerical simulations, primarily CFD, have
been effectively used to investigate SHZs during the pasteurization or sterilization of liquid foods
packed in cans [6,26,35–37].
Recently, CFD modeling studies have been extended to estimation of lethality of canned foods
during thermal processing [13,15,16,31,32,38–40]. Lethalities of canned viscous liquid foods based
on SHZ temperature have been found to be significantly lower than those based on volume-averaged
temperatures [13]. Predicting lethality based on volume-averaged temperature in the can leads to
much shorter heating time than needed. This results in under-heated products. To avoid under-
heating, Dimou et al. [31] calculated F values at every point inside the container where temperature
values are available (that is, at every node of the grid). Vatankhah et al. [40] investigated the effect
of mesh quality on determination of the SHZ position. Using coarse meshes may cause higher SHZ
temperature than actual temperature. However, using fine meshes can dramatically increase com-
putational time.
Estimating the lethality of canned foods is difficult due to the fact that SHZ is not a stationary
point or region in the can undergoing convection heating. Its location is not at the geometric center
of the can (as it is in the case of conduction heating). In the following section, several applications
of CFD to estimate lethality of canned foods are briefly described.
13.4.1 Prediction of Slow Heating Point During Thermal Processing

Datta and Teixeira [6] were the first ones who performed numerical predictions of transient temper-
ature and velocity profiles in a still retort during natural convection heating of canned liquid foods.
Water was used to simulate liquid food. It was heated uniformly around the can’s outside surface in
a cylindrical can. The recirculating flow inside the can consisted of liquid rising at the wall, radial
flow and mixing near the top, and uniform core flow downwards near the axis. The liquid inside
the container was found to be stratified inside the container with increasing temperatures toward
the top. The slowest heating region was located away from the centerline in a donut-shaped region
near the bottom 15% of height of container. However, the slowest heating points migrated within
this region without a particular pattern of migration. The simulation model successfully demon-
strated temperature and velocity profiles of canned liquid foods during thermal processing. They
investigated transient values of temperature with an average of the temperatures at the three slowest
heating points. However, its limitation is that this average temperature does not correspond to any
particular physical location.
13.4.2 Determination of F Value Using Subzones

In a study done by Lee and Yoon [41], subzones in a can domain were created by dividing the entire
domain vertically and horizontally while the SHZ was used as a reference zone to determine the
F value (Figure 13.7). SHZs were found at the bottom part when sizes of subzones were one-half
and one-fourth of the entire domain. However, when the size of the subzone was smaller than one-
eighth of the entire domain, the SHZ was not located at the bottom part. Average F values of each
subzone (FA) were then calculated. The SHZ was determined by overwrapping the SHZ with the
lowest FA in vertical and horizontal directions. As shown in Figure 13.8, FA values were decreased
when the size of the subzone became smaller. The FA value of the entire domain showed the high-
est value. When the size of the subzone was decreased, the FA value became close to the F value
of the cold point.
FIGURE 13.7 Determination of slow heating zone by creating subzones in vertical and horizontal directions
Shaded area represents SHZ at each scale of subzone by overwrapping vertical and horizontal subzones hav-
ing the lowest FA. Subzones in the vertical direction were divided by (a) 1, (b) 2, (c) 4, (d) 8, (e) 12, and (f) 16.
All subzones in the horizontal direction were divided by 6. (From Lee, M.G., Yoon, W.B., J. Food Process
Eng., 37, 493–505, 2014.)
FIGURE 13.8 FA values at various subzone sizes and relative difference between FA from the slowest
heating zone and the F value from the cold point. (From Lee, M.G., Yoon, W.B., J. Food Process Eng., 37,
493–505, 2014.)
The SHZ was determined at a relatively early stage of the heating process (45s). It was then main-
tained in the same zone until the process was completed. Note that the SHZ determined from the
subzones remained constant for the entire process while the cold point continuously changed during
the heating process. Such result shows that an SHZ determined from subzones can be used as an
indicator to evaluate the sterilization value of liquid food in convective heating.
13.4.3 F Value Based on Degradation of Spores

Numerical simulation of natural convection during sterilization of canned liquid food has been
conducted by Siriwattanayotin et al. [42]. They also simulated changes of invert sugar concentration
simultaneously. In addition to governing equations of continuity, momentum, and energy equations,
the following equation for concentration of sucrose is introduced:
Mass balance
dC A
= D∇ 2C A + rA (13.6)
dt
where CA is sucrose concentration, D is diffusion coefficient, rA is the rate of sucrose degradation.

The rate of sucrose inversion has been determined by Chawanasporn and Rattanathanalerk [43] as
shown below:
rA = − kT C A0.5 (13.7)

with
kT = A exp(− B × pH ) (13.8)

where A and B are functions of temperature. Invert sugar resulting from sucrose inversion was used
as an indicator of thermal deactivation of microorganisms.
In this study, the F value was determined by two methods. For the first method, temperature and
time history at fixed point in a can are needed for calculation by numerical integration as shown in
Equation 13.9. The second method, Equation 13.10, is based on the reduction of the initial number
of spore (Ni) to the final value (Nf).
t T − Tref

F0 Temp =
∫
0
10
z
dt (13.9)
F0 Spore = DT (log N i − log Nf ) (13.10)

2.303
DT = (13.11)
60 × kT
where kT was determined by Holdsworth [44].

The result showed that a hot region of the fluid at the top zone of a can was incorporated with
high concentration of invert sugar, which was in good agreement with the prediction of invert sugar
changes during heating. In order to calculate F0 of the process, the kinetic rate of sugar inversion
was replaced by that of C. botulinum spore deactivation.
The F0 Spore showed a smaller value than F0 Temp at the surrounding temperature of 100°C (Figure
13.9a). This was due to the effect of rate constant, as shown in Equation 13.11, which changed
with temperature. Lower surrounding temperature resulted in lower value of rate constant (kT),
while the F0 Temp used was a fixed value of z with a reference temperature of 121°C. Contrary to
the result at 100°C, F0 Temp showed a lower value than F0 Spore when temperature effect was included
(Figure 13.9b). While the safer process time could be established at higher surrounding temperature
FIGURE 13.9 FA of a can content based on Eqs. (13.9) and (13.10) at (a) 100°C and (b) 120°C. (From
Siriwattanayotin, S. et al., J. Food Eng., 73, 307–312, 2006.)
by using F0 Spore, less process time would be required to reduce the spore of C. botulinum to reach
the target value.
13.5 CONCLUSIONS
Heat transfers occurring in canned foods during thermal processing are complicated due to com-
bined mechanisms of conduction and convection, leading to continuous changes of cold points and
SHZs during thermal processing. CFD has been effectively used to understand heat transfer phe-
nomenon in cans and investigate SHZs during pasteurization or sterilization of canned foods. This
chapter presented the capability of CFD modeling and simulation to optimize thermal processing
of canned foods. CFD modeling of thermal processing of canned food can be used to ensure food
safety while minimizing food quality and nutrients loss.
REFERENCES
1. Stumbo C.R., Purohit K.S., Ramakrishnan T.V., Thermal process lethality guide for low-acid foods in
metal containers, Journal of Food Science, 40(6): 1316–1323, 1975.
2. Singh A., Singh A.P., Ramaswamy H.S., Computational techniques used in heat transfer studies on
canned liquid-particulate mixtures, Trends in Food Science & Technology, 43(1): 83–103, 2015.
3. Ghani A.A., Farid M.M., Chen X.D., Numerical simulation of transient temperature and velocity pro-
files in a horizontal can during sterilization using computational fluid dynamics, Journal of Food
Engineering, 51(1): 77–83, 2002.
4. Erdogdu F., A review on simultaneous determination of thermal diffusivity and heat transfer coefficient,
5. Dimou A., Yanniotis S., 3D numerical simulation of asparagus sterilization in a still can using compu-
tational fluid dynamics, Journal of Food Engineering, 104(3): 394–403, 2011.
6. Datta A.K., Teixeira A.A., Numerically predicted transient temperature and velocity profiles during
natural convection heating of canned liquid foods, Journal of Food Science, 53(1): 191–195, 1988.
7. Kumar A., Bhattacharya M., Blaylock J., Numerical simulation of natural convection heating of canned
thick viscous liquid food products, Journal of Food Science, 55(5): 1403–1411, 1990.
8. Kumar A., Bhattacharya M., Transient temperature and velocity profiles in a canned non-Newtonian
liquid food during sterilization in a still-cook retort, International Journal of Heat and Mass Transfer,
34(4–5): 1083–1096, 1991.
9. Bhuvaneswari E., Anandharamakrishnan C., Heat transfer analysis of pasteurization of bottled beer
in a tunnel pasteurizer using computational fluid dynamics, Innovative Food Science & Emerging
Technologies, 23: 156–163, 2014.
10. Erdogdu F., Uyar R., Palazoglu T.K., Experimental comparison of natural convection and conduction
heat transfer, Journal of Food Process Engineering, 33(s1): 85–100, 2010.
11. Ghani A.A., Farid M.M., A numerical simulation study on thermal sterilization of food in pouches using
computational fluid dynamics (CFD), Association for Computing Machinery New Zealand Bulletin,
1(2): 1–9, 2005.
12. Kiziltas S., Erdogdu F., Palazoglu T.K., Simulation of heat transfer for solid–liquid food mixtures in
cans and model validation under pasteurization conditions, Journal of Food Engineering, 97(4): 449–
456, 2010.
13. Koribilli N., Aravamudan K., Varadhan M.A., Quantifying enhancement in heat transfer due to natural
convection during canned food thermal sterilization in a still retort, Food and Bioprocess Technology,
4(3): 429–450, 2011.
14. Lespinard A.R., Mascheroni R.H., Influence of the geometry aspect of jars on the heat transfer and flow
pattern during sterilization of liquid foods, Journal of Food Process Engineering, 35(5): 751–762, 2012.
15. Rawajfeh K., Albaali A.G., Saidan, M., Abureden, S., Modeling of natural convection heating and bio-
chemical changes in a viscous liquid canned food using computational fluid dynamics, International
Journal of Food Science and Nutrition Engineering, 3(4): 71–79, 2013.
16. Shafekhani S., Zamindar N., Hojatoleslami M., Toghraie D., Numerical simulation of transient tempera-
ture profiles for canned apple puree in semi-rigid aluminum based packaging during pasteurization,
Journal of Food Science and Technology, 53(6): 2770–2778, 2016.
17. Spanu S., Vignali G., Modelling and multi-objective optimisation of the VHP pouch packaging sterilisa-
tion process, International Journal of Food Engineering, 12(8): 739–752, 2016.
18. Rao M.A., Anantheswaran R.C., Convective heat transfer to fluid foods in cans, Advances in Food
Research, 32: 39–84, 1988.
19. Scott G., Richardson P., The application of computational fluid dynamics in the food industry, Trends in
Food Science & Technology, 8(4): 119–124, 1997.
20. David J.R., Graves R.H., Szemplenski T., Handbook of Aseptic Processing and Packaging (Vol. 4),
CRC Press, Boca Raton, FL, USA, 2012.
21. Heldman D.R., Harter R.W., Principles of Food Processing, Aspen Publication, Gaithersburg, MD,
USA, 17–180, 1998.
22. Teixeira A.A., Dixon J.R., Zahradnik J.W., Zinsmeister G.E., Computer determination of spore survival
distribution in thermally-processed conduction-heated foods, Food Technology, 23: 78–80, 1969.
convection heating in cans, Journal of Food Science, 76(1): E119–E129, 2011.
24. Boz Z., Erdogdu F., Evaluation of two-dimensional approach for computational modelling of heat
and momentum transfer in liquid containing horizontal cans and experimental validation, Food and
Bioproducts Processing, 91(1): 37–45, 2013.
943–951, 2003.
cylindrical containers, Journal of Food Engineering, 77(4): 1024–1036, 2006.
27. Augusto P.E.D., Pinheiro T.F., Cristianini M., Using computational fluid-dynamics (CFD) for the evalu-
ation of beer pasteurization: Effect of orientation of cans, Food Science and Technology (Campinas),
30(4): 980–986, 2010.
28. Pflug I.J., Procedures for Carrying Out a Heat Penetration Test and Analysis of the Resulting Data.
Department of Food Science and Nutrition, University of Minnesota, 1975.
29. Naveh D., Kopelman I.J., Pflug I.J., The finite element method in thermal processing of foods, Journal
of Food Science, 48(4): 1086–1093, 1983.
30. Cordioli M., Rinaldi M., Copelli G., Casoli P., Barbanti D., Computational fluid dynamics (CFD) mod-
elling and experimental validation of thermal processing of canned fruit salad in glass jar, Journal of
31. Dimou A., Panagou E., Stoforos N.G., Yanniotis S., Analysis of thermal processing of table olives using
computational fluid dynamics, Journal of Food Science, 78(11): E1695–E1703, 2013.
32. Padmavati R., Anandharamakrishnan C., Computational fluid dynamics modeling of the thermal pro-
cessing of canned pineapple slices and titbits, Food and Bioprocess Technology, 6(4): 882–895, 2013.
33. Ghani A.A., Farid M.M., Using the computational fluid dynamics to analyze the thermal sterilization
of solid–liquid food mixture in cans, Innovative Food Science & Emerging Technologies, 7(1): 55–61,
2006.
34. Ghani A.A., Farid M.M., Numerical simulation of solid–liquid food mixture in a high pressure process-
ing unit using computational fluid dynamics, Journal of Food Engineering, 80(4): 1031–1042, 2007.
35. Zechman L.G., Pflug I.J., Location of the slowest heating zone for natural-convection-heating fluids in
metal containers, Journal of Food Science, 54(1): 205–209, 1989.
36. Akterian S.G., Fikiin K.A., Numerical simulation of unsteady heat conduction in arbitrary shaped
canned foods during sterilization processes, Journal of Food Engineering, 21(3): 343–354, 1994.
37. Farid M., Ghani A.A., A new computational technique for the estimation of sterilization time in canned
food, Chemical Engineering and Processing: Process Intensification, 43(4): 523–531, 2004.
38. Derossi A., De Pilli T., La Penna M.P., Severini C., Prediction off heating length to obtain a definite F
value during pasteurization of canned food, Journal of Food Process Engineering, 36(2): 211–219, 2013.
39. Anand Paul D., Anishaparvin A., Anandharamakrishnan C., Computational fluid dynamics studies on
pasteurisation of canned milk, International Journal of Dairy Technology, 64(2): 305–313, 2011.
40. Vatankhah H., Zamindar N., Baghekhandan M.S., Heat transfer simulation and retort program adjust-
ment for thermal processing of wheat based Haleem in semi-rigid aluminum containers, Journal of
Food Science and Technology, 52(10), 6798–6803: 2015.
41. Lee M.G., Yoon W.B., Developing an effective method to determine the deviation of F value upon the
location of a still can during convection heating using CFD and subzones, Journal of Food Process
Engineering, 37(5): 493–505, 2014.
42. Siriwattanayotin S., Yoovidhya T., Meepadung T., Ruenglertpanyakul W., Simulation of sterilization
of canned liquid food using sucrose degradation as an indicator, Journal of Food Engineering, 73(4):
307–312, 2006.
43. Chawanasporn W., Rattanathanalerk M., Effect of temperature and pH on degradation of sucrose solu-
tion to monosaccharide, Fundamentals of food engineering practice, Department of Food Engineering,
King Mongkut’s University of Technology, Thonburi, Thailand, 2002.
44. Holdsworth S.D., Thermal Processing of Packaged Foods, Blackie Academic and Processional, London,
UK, 1997.
14 CFD Modeling of Thermal
Processing of Particulate Foods
Massimiliano Rinaldi, Matteo Cordioli, and Davide Barbanti
CONTENTS
14.1 Introduction..........................................................................................................................305
14.2 CFD Modeling of Canned Foods with Particulates............................................................. 305
14.3 CFD Modeling of Particulate Foods Subjected to Continuous Thermal Treatments............ 312
14.4 CFD Modeling of Particulate Foods’ Fluidized Bed.......................................................... 315
14.5 Conclusions.......................................................................................................................... 317
References....................................................................................................................................... 318
14.1 INTRODUCTION
In the food industry, a great number of particulate foods are packaged in cans or jars, filled with
an appropriate sugar syrup or brine, and thermally processed in order to increase their shelf-life
through the inactivation of both spoilage microorganisms and enzymes. The market for this kind of
product has increased in recent years, mainly driven by the consumer demand for both convenience
foods with homemade quality and beverages. Particles complicate the heat treatment of foods; as
a consequence, heat exchangers are often over-dimensioned, resulting in quality and organoleptic
detriments with possible costs to equipment, operations and the environment.
Particles include a considerable number of products, such as fibers (tomato peels, fruit wires)
gelatinous particles, fruit pieces (peach cubes, pineapple pieces, citrus sacs) and soft/hard particles
(carrots, peas, seeds, stone cells, cereal, nuts, grains).
14.2 CFD MODELING OF CANNED FOODS WITH PARTICULATES

As stated previously, mixtures of solid–liquid foods in cans represent a difficult situation. Generally
the processing time for these cases is determined assuming solid–liquid mixture is heated by con-
duction and ignoring natural convection effects, thus leading to heating times much longer than are
required, with over-processed products [1].
The development of advanced modeling techniques such as computational fluid dynamics (CFD)
modeling has reduced the complexity in solving heat transfer for solid–liquid food mixtures in the
design of thermal processes [2].
However, the literature still lacks in complete simulation studies for predicting temperature
changes of solid–liquid mixtures where solids are randomly dispersed in the liquid phase [1].
Among them, Rabiey et al. [3] successfully simulated natural convection heating in a metal can
containing water and large particles (45 mm in diameter) made of alginate. The modeled tempera-
ture profiles confirmed that the hottest zones were located near the wall of the can and that tempera-
ture stratification was influenced by the presence of large particles inside the container. Particularly,
CFD modeling showed that the warm liquid coming from the boundary layers almost stagnated
at the top of can as the area with downward flow is much greater than the area with upward flow.
This fact was confirmed by a higher upward velocity (about 28 mms−1) compared to the downward
one (about 12 mm s−1). As the liquid descends, its temperature decreases because of conduction
305
and mixing with the lower and colder liquid layers and also because of the exchanges with the cold
particles. Reaching the bottom, the liquid moves again towards the lateral wall and begins a new
cycle (Figure 14.1).
Temperature stratification was reported to be relevant and to be considered in designing ther-
mal processes of food intended to inactivate microbes and enzymes. As shown in the following
Figure 14.2, the stratification leads to locate the slowest heating point near the bottom of the can
(data expressed as dimensionless temperature).
The differences in temperature caused also a significant difference in pasteurizing values (PV):
Authors calculated a difference of PV value higher than 25% at the particle centers of the top and
bottom layers. This difference is unacceptable regarding thermal stabilization of packed foods.
Similarly, Kızıltaş et al. [1] modeled the pasteurization of canned spherical bodies (peas) with
a ratio of the water volume to the volume of can of about 35%, from measurements made on
FIGURE 14.1 Fluid flow inside the can with large particles. (From Rabiey, L., Flick, D. and Duquenoy, A.
(2007). 3D simulations of heat transfer and liquid flow during sterilization of large particles in a cylindrical
vertical can. Journal of Food Engineering, 82, 409–417.)
FIGURE 14.2 Temperature stratification inside the can. (From Rabiey, L., Flick, D. and Duquenoy, A.
(2007). 3D simulations of heat transfer and liquid flow during sterilization of large particles in a cylindrical
vertical can. Journal of Food Engineering, 82, 409–417.)
CFD Modeling of Thermal Processing of Particulate Foods 307
commercial samples of canned peas. CFD modeled data showed that, when there are solid imper-
meable particles distributed in the fluid, the velocity profiles change due to the heat exchange and
surface deflections while the flow is slowly through the stack of solid particles. Particularly, the
velocity profiles changed when fluid passed between solid particles as in Figure 14.3.
The downward velocity increased when the passage between peas was reduced, as expected.
Authors reported also the formation of Benard convective cells as expected for confined natu-
ral convection problems. Natural convection effects were confirmed by the thermal stratification
(Figure 14.4).
FIGURE 14.3 Velocity profiles of canned peas during thermal treatment. (From Kızıltaş, S., Erdoğdu, F,
and Palazoğlu, T. K. (2010). Simulation of heat transfer for solid-liquid food mixtures in cans and model vali-
dation under pasteurization conditions. Journal of Food Engineering, 97, 449–456.)
FIGURE 14.4 Thermal stratification (panel a) and Benard convective cells (panel b). (From Kızıltaş, S.,
Erdoğdu, F., and Palazoğlu, T. K. (2010). Simulation of heat transfer for solid-liquid food mixtures in cans and
model validation under pasteurization conditions. Journal of Food Engineering, 97, 449–456.)
The slowest heating point was located at the bottom of the can, confirming the natural convective
heat exchange.
Regarding more difficult shapes, Ghani and Farid [4] successfully modeled the thermal steriliza-
tion process of canned pineapple slices with governing liquid, indicating that natural convection
effects in the liquid played a significant role in the evolution of temperature.
In the same way, Dimou and Yanniotis [5] used the CFD modeling technique for studying the
temperature profile, the velocity profile and the slowest heating zone in a still can filled with aspara-
gus in brine. Regarding fluid flow, authors reported the presence of vortices in the near-metal wall
region of the can during heating as well as during the cooling phase, depending on the number of
asparagus pieces and free spaces in the can (Figure 14.5).
The authors observed a significant increase in velocities from about 15 mm s–1 to 40 mm s–1 when
the number of asparagus pieces increased from 1 to 19 due to the section reduction: CFD modeling
allowed the quantification of this effect as in Figure 14.6.
In addition, further investigations might study the influence of product and process parameters
such as thermal food properties, liquid rheology (including non-Newtonian), particle size, stacking
FIGURE 14.5 Velocity profile for asparagus in brine. (From Dimou, A. and Yanniotis, S. (2011). 3-D numer-
ical simulation of asparagus sterilization using computational fluid dynamics. Journal of Food Engineering,
104, 394–403.)
FIGURE 14.6 Effect of asparagus number on velocity profile. (From Dimou, A. and Yanniotis, S. (2011).
3-D numerical simulation of asparagus sterilization using computational fluid dynamics. Journal of Food
Engineering, 104, 394–403.)
pattern, solid to particle ratio, and initial and external temperatures. In this context, Padmavati and
Anandharamakrishnan [6] investigated, by means of CFD models, the effect of size reduction of
pineapple slices vs. titbits on the effectiveness of heat transfer during thermal processing, underlin-
ing the effect of natural convection moves in spaces between titbits, compared to slices. The models
were drawn by starting from commercial samples of sliced pineapple in syrup with 8 slices each,
having outer and inner diameters of 8.7 and 3.4 cm, respectively, stacked one above the other, in
order to get a total height of 9.2 cm. Pineapple titbits were rearranged in cylindrical can of dimen-
sions equivalent to that used for the pineapple slices with 105 titbits randomly distributed. The
authors reported that, from the predicted radial profiles of velocity, the velocity resulted higher in
the titbits processing compared to the slice one, probably due to the non-blockage of natural convec-
tion. This fact was confirmed by the temperature simulation results (Figure 14.7).
As reported in Figure 14.8, the rate of heat transfer has been found to be almost uniform through-
out the can in the case of pineapple titbits, while, on the other hand, in the pineapple slices, the mid-
dle point (P2) showed a minimum temperature rise since it was located in the SHZ (slowest heating
zone). This great difference in the rate of heat transfer in the solid and liquid components may lead
to the underprocessing of the solids or to the overprocessing of the liquid component. Authors used
the simulated temperature for calculating F values at different can locations.
More recently, Cordioli et al. [7] studied the heat transfer of a fruit salad, composed of five dif-
ferent fruits with different shapes, sizes and thermal properties, submerged in water/sugar syrup
during the pasteurization treatment. As reported in Figure 14.9, the five different fruits were grape,
pear, cherry, peach and pineapple with very different shapes and quite different sizes.
The authors developed a CFD model that was successfully validated against the experimental
ones; the developed model allowed location of the slowest heating point (SHP) and observed how
the F values was influenced by the distance from the jar bottom as function of natural convection
motion of the syrup (Table 14.1).
Authors reported that fruits positioned close to the top of the jar showed a simulated F value
higher than those placed close to the bottom: for example, grape, which was in the highest location,
presented a pasteurization value (9.70 min) that was much higher than pear’s (1.56 min). The same
effect was also observed for the F values achieved by the syrup: the higher the quota inside the jar,
the higher the pasteurizing value (Table 14.1). This paper demonstrated that, despite the convective
motions, temperature stratification is very relevant and causes uneven thermal treatment in a food
product with solids and liquid and that the differences increase with an increased ratio of solids/
liquid.
More recently, Sarghini and Erdogdu [8] applied the CFD three-dimensional technique to study
the effect of an end-over-end process for canned particulate products by using a volume of fluid (VOF)
FIGURE 14.7 Pineapple titbits vs. slices temperature profile. (From Padmavati, R., Anandharamakrishnan,
C. (2013). Computational fluid dynamics modeling of the thermal processing of canned pineapple slices and
titbits. Food Bioprocess Techology, 6, 882–895.)
FIGURE 14.8 Pineapple titbits vs. slices F values. (From Padmavati, R., Anandharamakrishnan, C. (2013).
Computational fluid dynamics modeling of the thermal processing of canned pineapple slices and titbits. Food
Bioprocess Techology, 6, 882–895.)
FIGURE 14.9 Fruits composing the fruit salad. (From Cordioli, M., Rinaldi, M., Copelli, G., Casoli, P., &
Barbanti, D. (2015). Computational fluid dynamics (CFD) modeling and experimental validation of thermal
processing of canned fruit salad in glass jar. Journal of Food Engineering, 150, 62–69.)
multiphase approach with rotating mesh. The authors studied a three-dimensional simulation of a
can (87.3 mm in diameter – 115.9 mm in length) containing 60 spherical particles (∅ 19 mm) homo-
geneously distributed both in a low viscosity Newtonian and high viscosity non-Newtonian liquid
with 8.63% headspace. The mathematical model allowed study of temperature, apparent viscosity
and velocity magnitude of three particles inside the can in three positions (P6, near the base and
lateral wall, P30 near the lateral wall, and P46, the most internal one). Unexpectedly, the domi-
nance of rotational forces annulled the internal fluid movement (Figure 14.10), and the velocity of
the fluid was simply dictated by rigid rotational effects, hence dramatically reducing the shear rate
and increasing the viscosity near the external wall of the can (Figure 14.10). This fact resulted in
TABLE 14.1
RMSE, Experimental, and Simulated F Values for Fruits Composing
the Fruit Salad
12.9
F90 (min)
12.9
Point of Measurement RMSE Experimental Simulated ∆F90
Syrup 17.5 mm 0.68 0.40 0.36 0.04
30.5 mm 0.79 1.19 1.01 0.18
43.5 mm 1.77 2.52 2.55 0.03
56.5 mm 2.65 4.04 4.99 0.95
Average 1.47 0.30
Peach 1.14 5.41 5.30 0.11
Pear 2.31 1.56 1.63 0.07
Pineapple 1.78 5.03 5.13 0.10
Grape 1.61 9.70 9.68 0.02
Cherry 1.31 7.27 7.31 0.04
Average 1.63 0.07
Source: Cordioli, M., Rinaldi, M., Copelli, G., Casoli, P., & Barbanti, D. (2015). Computational
fluid dynamics (CFD) modeling and experimental validation of thermal processing of
canned fruit salad in glass jar. Journal of Food Engineering, 150, 62–69.
further opposition to the mass transfer and generated an almost pure conductive heating with dra-
matic temperature stratification as demonstrated by particles’ temperature profiles (Figure 14.11).
The temperature profiles of particles located near the bottom and the external wall were signifi-
cantly higher compared to those located at the can center, as expected for a purely conductive heat
exchange. This fact will produce great differences in pasteurizing values with more external par-
ticles overtreated. Consequently, the CFD technique could be used for identifying the best rotation
conditions with improved heat exchange efficiency.
14.3 CFD MODELING OF PARTICULATE FOODS SUBJECTED

TO CONTINUOUS THERMAL TREATMENTS
A further step in CFD modeling is represented by CFD modeling of particle foods submitted to con-
tinuous heat treatments. This condition is a great challenge for the food industry as the measurement
of residence time distribution is quite difficult and has an enormous impact on food safety [9]. The
authors simulated several scenarios in a 2-D bend (∅ = 5 cm) by considering two solid-liquid ratios,
buoyant and neutrally buoyant particles with diameter ranging from 0.6 to 1.2 cm and density from
0.998 to 1.002 kgm–3. Reported data from CFD simulations showed that in the case of the neutrally
buoyant particles, the flow is quite regular in the straight section, while more mixing streams take
place due to the acceleration and deceleration of the fluid in the bend. In general, the particles’
diameter does not seem to play a significant role. In the case of the buoyant particles the flow is less
uniform and flow of particles in “plugs” takes place. Moreover, authors observed that the residence
time distribution (RTD) was influenced by the solids fraction: the higher the solids fraction, the nar-
rower the RTD. Moreover, for neutrally buoyant particles, the distribution is much more symmetric
around the mode compared to buoyant ones.
Krishnan and Aravamudan [10] simulated the processing of spherical food particulates in non-
Newtonian fluids by means of Computational Fluid Dynamics (CFD). The authors studied the effect
FIGURE 14.10 Temperature (panel a and b) and velocity profile (panel c and d) in rotating can with par-
ticulates. (From Sarghini, F., & Erdogdu, F. (2016). A computational study on heat transfer characteristics of
particulate canned foods during end-over-end rotational agitation: Effect of rotation rate and viscosity. Food
and Bioproducts Processing, 100, 496–511.)
FIGURE 14.11 Temperature of particulates in three distinct locations. (From Sarghini, F., & Erdogdu, F.
(2016). A computational study on heat transfer characteristics of particulate canned foods during end-over-end
rotational agitation: Effect of rotation rate and viscosity. Food and Bioproducts Processing, 100, 496–511.)
of power law index (n = 0.94 and 0.76) as well as the blockage ratio (BR) on the dimensionless tem-
perature into the fluid and Nu number on the sphere surface. Results from simulations (Figure 14.12)
showed that the thickness of the thermal boundary layer was higher for higher power law index.
Moreover, in the case of n = 0.76 both for BR 0.5 and BR 0.025 the temperature contours were
closed around the sphere; the thermal boundary layer thickness was higher at higher n, especially in
the rear region of the sphere, as shown in Figure 14.12.
The CFD technique also allowed the study of velocity streamline and viscosity of the fluid
around the sphere for different BR (Figure 14.13): for low blockage ratio, a boundary layer separa-
tion was obtained but this effect was suppressed by higher blockage ratio due to the stabilizing
influence of the tube wall. Regarding viscosity values, at the region downstream of the sphere,
FIGURE 14.12 Dimensionless temperatures around the sphere and variation of Nusselt number at the sphere
surface. (From Krishnan, S., & Aravamudan, K. (2013). Simulation of non-Newtonian fluid-food particle heat
transfer in the holding tube used in aseptic processing operations. Food and Bioproducts Processing, 91(2),
129–148.)
FIGURE 14.13 Velocity around the sphere and variation of viscosity at the sphere surface. (From Krishnan,
S., & Aravamudan, K. (2013). Simulation of non-Newtonian fluid-food particle heat transfer in the holding
tube used in aseptic processing operations. Food and Bioproducts Processing, 91(2), 129–148.)
FIGURE 14.14 Dimensionless temperature distributions in the fluid domain and local Nusselt number dis-
tributions on the sphere surface. (From Krishnan, S., & Aravamudan, K. (2013). Simulation of non-Newtonian
fluid-food particle heat transfer in the holding tube used in aseptic processing operations. Food and Bioproducts
Processing, 91(2), 129–148.)
the effect of blockage ratio was considerable: particularly, the viscosity values increased at the point
of boundary layer separation, where the minimum shear rate occurred (Figure 14.13).
Authors also carried out mathematical simulations by varying BR and Reynold numbers with a
constant power law index of 0.76 (Figure 14.14).
The thermal boundary layer, especially around the front stagnation portion of the sphere, waas
compressed in the confined case (panel a), indicating higher Nusselt numbers, especially at Re of
0.1. However, the effect of blockage on the thermal boundary layer is weaker at Re of 40 due to the
inherently thin boundary layers at higher Reynolds numbers.
14.4 CFD MODELING OF PARTICULATE FOODS’ FLUIDIZED BED

Subsequently, Datta et al. [11] used the CFD technique for studying the velocity profiles of grain-
water suspension and cassava starch particles with different solid fractions (5, 10 and 15% w/w) in a
continuous tube flow treatment. The addition of solid particles to turbulent liquid flow modified the
velocity distribution of the suspension and the solid phase velocity profile was generally asymmetri-
cal about the central axis with the particles flowing at higher velocity in the upper part of the tube.
Datta et al. [11] obtained the magnitude of slip velocity between water and particle from CFD analy-
sis using FLUENT: slip velocity represents information that can be experimentally obtained with
great difficulty but it also represents a crucial parameter for heat treatment design. The obtained
profile was asymmetric along the vertical center line and its magnitude was reported to be higher
near the tube wall, as expected. In general, slip velocity close to the center line increased as particle
size increased, and the profile becomes more asymmetric with the increase in particle size.
Another possible application of CFD analysis is hydrofluidization, an interesting technique for
chilling and freezing that uses a fluidized bed of highly turbulent liquid with moving products.
Oroná et al. [12] successfully modeled a food hydrofluidization system composed of multiple fluid
immersed jets and multiple moving food spheres. The modeled system consisted of a cylindrical
vessel with 69 orifices to produce jets at its base with a regularly spaced squared array of 10-mm
diameter spheres (13 spheres) made of potato. Simulations showed that spheres continuously ascend
according to their initial position without perturbing the fluid field. The fastest spheres were those
initially placed at the axis of the domain and the slowest were those ascending near the domain
walls (Figure 14.15).
The developed approach was validated against other mathematical models with an average per-
centage error less than 13%, indicating good agreement. The developed model could be then used
for heat and mass transfer modeling in the food domain. Figure 14.16 shows the approach to repre-
sent the physical scenario of HF through the models developed by Peralta et al. [13], Belis et al. [14]
and Oroná et al. [12].
The method proposed in the paper from Oronà et al. [12] seems to be more realistic, thanks to
the combination of CFD with discrete phase model (DPM) and discrete element method (DEM).
In DPM, the fluid phase is considered a continuum by solving the Navier–Stokes equations,
while the dispersed phase is solved by tracking the particles through the calculated flow field.
FIGURE 14.15 Dimensionless temperature distributions in the fluid domain and local Nusselt number distri-
butions on the sphere surface. (From Oroná, J. D., Zorrilla, S. E., & Peralta, J. M. (2017). Computational fluid
dynamics combined with discrete element method and discrete phase model for studying a food hydrofluidiza-
tion system. Food and Bioproducts Processing, 102, 278–288.)
FIGURE 14.16 Modeling approaches of hydrofluidization system. (From Oroná, J. D., Zorrilla, S. E., &
Peralta, J. M. (2017). Computational fluid dynamics combined with discrete element method and discrete
phase model for studying a food hydrofluidization system. Food and Bioproducts Processing, 102, 278–288.)
DEM completes the information related to particle–particle interactions by the so-called “soft
sphere” approach. Due to the system complexity, however, authors stated that further validation
studies are necessary to reinforce the knowledge of the model capability to estimate the main
variables involved.
14.5 CONCLUSIONS
CFD is a very promising technique for process modeling in the food technology field, and nowadays
it represents a relevant method in food process design and optimization. The CFD technique is also
gaining more popularity in the field of particulate foods, which represents more complex systems
and generally are difficult to be studied by means of experimental trials or laboratory tests.
CFD can overcome all the limits related to this kind of food product by mathematical calculus;
however, as for any mathematical model, the data obtained from models must be fully validated
through experimental tests and all the parameters and food characteristics used as inputs in the
models must be fully weighted before the utilization.
The present chapter reported several examples of the CFD technique with a vast range of appli-
cations in particulate foods processes. The CFD technique is expected to be used more and more
in particulate food process design, including for very complex scenarios, to allow optimization in
terms of energy and overall food quality.
REFERENCES
1. Kızıltaş S., Erdoğdu F., Palazoğlu T.K., Simulation of heat transfer for solid–liquid food mixtures in
cans and model validation under pasteurization conditions, Journal of Food Engineering, 97(4): 449–
456, 2010.
2. Singh A., Singh A.P., Ramaswamy H.S., Computational techniques used in heat transfer studies on
canned liquid-particulate mixtures—A review, Trends in Food Science & Technology, 43: 83–103, 2015.
3. Rabiey L., Flick D., Duquenoy A., 3D simulations of heat transfer and liquid flow during sterilization of
large particles in a cylindrical vertical can, Journal of Food Engineering, 82: 409–417, 2007.
4. Ghani A.G., Farid M.M., Using the computational fluid dynamics to analyse the thermal sterilization of
solid-liquid food mixture in cans, Innovative Food Science and Emerging Technologies, 7: 55–61, 2006.
5. Dimou A., Yanniotis S., 3D numerical simulation of asparagus sterilization in a still can using compu-
tational fluid dynamics, Journal of Food Engineering, 104: 394–403, 2011.
6. Padmavati R., Anandharamakrishnan C., Computational fluid dynamics modeling of the thermal pro-
cessing of canned pineapple slices and titbits, Food Bioprocess Techology, 6: 882–895, 2013.
7. Cordioli M., Rinaldi M., Copelli G., Casoli P., Barbanti D., Computational fluid dynamics (CFD) model-
ing and experimental validation of thermal processing of canned fruit salad in glass jar, Journal of Food
Engineering, 150: 62–69, 2015.
8. Sarghini F., Erdogdu F., A computational study on heat transfer characteristics of particulate canned
foods during end-over-end rotational agitation: Effect of rotation rate and viscosity, Food and
Bioproducts Processing, 100: 496–511, 2016.
9. Duchanoy C., Jongen T.R., Efficient simulation of liquid–solid flows with high solids fraction in com-
plex geometries, Computers & Fluids, 32(10): 1453–1471, 2003.
10. Krishnan S., Aravamudan K., Simulation of non-Newtonian fluid-food particle heat transfer in the hold-
ing tube used in aseptic processing operations, Food and Bioproducts Processing, 91(2): 129–148, 2013.
11. Datta A.K., Bhunia K., Sharma R., Computational fluid dynamics (CFD) modeling of grain-water sus-
pensions in tube, Agricultural Engineering International: CIGR Journal, 18(1): 269–283, 2016.
12. Oroná J.D., Zorrilla S.E., Peralta J.M., Computational fluid dynamics combined with discrete element
method and discrete phase model for studying a food hydrofluidization system, Food and Bioproducts
Processing, 102: 278–288, 2017.
13. Peralta J.M., Rubiolo A.C., Zorrilla S.E., Mathematical modeling of the heat and mass transfer in a
stationary potato sphere impinged by a single round liquid jet in a hydrofluidization system, Journal of
14. Belis E.E., Zorrilla S.E., Peralta J.M., Effect of the number of orifices and operative variables on the
heat and mass transfer in a hydrofluidization system with static spheres, Journal of Food Engineering,
153: 96–107, 2015.
15 CFD Modeling of Natural-
Convection Heating Processes
Ferruh Erdogdu, Huseyin Topcam,
Fabrizio Sarghini, and Francesco Marra
CONTENTS
15.1 Introduction............................................................................................................................ 319
15.2 Natural Convection Heat Transfer......................................................................................... 320
15.2.1 Fundamentals.......................................................................................................... 320
15.2.2 Mathematical Background...................................................................................... 322
15.2.3 Examples from the Studies in the Literature............................................................... 327
15.3 CFD Examples..................................................................................................................... 330
15.3.1 Natural Convection Heating of a Canned Liquid.......................................................330
15.3.2 Natural Convection Heating of Canned Liquid with Solid Particles...................... 332
15.4 Conclusions.......................................................................................................................... 336
References....................................................................................................................................... 336
15.1 INTRODUCTION
Besides the recent innovative food processes using high pressure and pulsed electric field, heat trans-
fer is the dominant transport mechanism in various food processing operations from pasteurization–
sterilization to cooling, from thawing with conventional (air or water-immersion thawing) and
innovative (radio frequency and microwave applications) methods to freezing. In almost all food
processing operations, temperature changes due to heat transfer are observed; thus, determining or
knowing the temperature distribution in the product and/or inside the processing system is required
for further design and optimization purposes.
Heat, as an energy form, transfers by virtue of a temperature difference between two mediums,
and there are three modes involved in this transfer: conduction, convection and radiation. The
similarity in the first two cases is the requirement of a medium for the heat transfer to occur. In
conduction, heat energy is transferred from more energetic particles in a substance to the adjacent
ones due to the temperature difference, and it takes place in solids. Even though conduction might
be assumed in fluids, its occurrence might be dominated by the eventual convection heat transfer,
depending on the fluid motion. This motion starts due to natural convection triggered by density
differences occurring via the change in temperature. Convection heat transfer, compared to con-
duction, requires a fluid medium, where the fluid motion enhances the heat transfer due to mixing
of cold and warmer regions. Radiation, on the other hand, differs from conduction and convection
in that it does not require a medium. For example, heat energy is directly transmitted through a
vacuum from the sun to Earth. Heating the hands by a camp fire in a cold environment is another
example of radiation. In food processing, use of infrared heating for surface pasteurization of agri-
cultural commodities (e.g., herbs and spices) is a perfect case of radiation heating.
Among these three modes, convection heat transfer is rather significant in food processing due to
the changes in temperature, velocity (and pressure) in a fluid that should be considered for designing
the process. The processed fluid might be the target product itself, like in the case of pasteurization
of milk in a double pipe or plate heat exchanger, or it might be the heat transfer medium, like in the
319
case of air cooling–freezing. Temperature distribution inside a cooling or freezing room, for exam-
ple, will be significant to properly design the whole process where the temperature change inside the
product will be an additional concern. Heat transfer by natural convection in domestic refrigerators
is a typical example of this case [1]. In the cold storage or freezing rooms, however, effects of natural
convection in addition to the forced convection (where the fluid movement is forced using a fan, for
example) will need to be quantified and considered for design purposes.
15.2 NATURAL CONVECTION HEAT TRANSFER

15.2.1 Fundamentals
Convection heat transfer can be considered in two cases: forced and natural convection. In the first
case, fluid motion is obtained by an external force like a pump or fan; in the second, it is triggered
naturally due to evolution of buoyancy forces occurring with density changes with temperature.
Because of the fluid motion, convection heat transfer is expected to take place much faster than
conduction heat transfer with respect to the temperature increase. Erdogdu et al. [2] was one of the
first to demonstrate the comparison of natural convection and conduction in a cylindrical cavity
where the experimental studies were carried out in pasteurization conditions and used to validate
the developed CFD models.
In a fluid medium, when there is a change in the density of a fluid, there will be an upward (with
a decrease in the density of the fluid due to increase in its temperature—heating process) or down-
ward forces (with an increase in the density due to the decrease in its temperature—cooling process)
created. All these changes are based on the gravitational forces, and quantifying the gravitational
force over the viscous forces and/or inertial forces is required to understand the dominating force
mechanism on the change of temperature and flow field evolution. Two dimensionless numbers,
Froud (Fr) and Taylor (Ta) numbers, are used for this comparison:
fr −i ω 2 R
Fr = =
fgbf g
f− i ω 2 R4
Ta = = (15.1)
fviscous ν2
 2π f 
 ω = 
60 
where fgbf is the gravitational buoyancy, frotational is the rotational force, f–i is the viscous force, and R
is the characteristic dimension (m). Tutar and Erdogdu [3] present a detailed analysis for the use of
Froude and Taylor numbers to quantify the various effects on the evolution of temperature in a rotat-
ing two-dimensional cylindrical cavity. In the natural convection case, the force formed by natural
means with the effect of gravitational force is called the buoyancy force:
F = ρ f ⋅ g ⋅V (15.2)

where F is the buoyancy force, ρf is the density of the fluid, g is the gravitational acceleration
(9.81 m/s2) and V is the volume of the body immersed in the fluid. Hence, temperature has a sig-
nificant impact on the density difference. Since the density difference is the main cause of natural
convection heat transfer, the volume expansion coefficient, which represents the variation of density
CFD Modeling of Natural-Convection Heating Processes 321
as a function of temperature in fluids at constant pressure, is an additional physical property, to

consider in natural convection:
1  ∂v  1  ∂ρ 
β= ⋅  = − ⋅  (15.3)
v  ∂T  P ρ  ∂T  P
where T is in K. Cengel and Ghajar [4] demonstrate a perfect description of the effects of buoyancy
force, temperature and volume expansion coefficient on natural convection heat transfer. For an
ideal gas, the volumetric expansion coefficient can easily be demonstrated to be equal to the inverse
of the temperature:
P = ρ ⋅ R⋅T
1 (15.4)
β=
T
A large value for β demonstrates a large change in density as a function of temperature; the larger
the temperature difference, the larger the density change, leading to higher buoyancy forces and
higher natural convection heat transfer rate [4].
The dimensionless number to represent the natural convection effects is the Grashof number. Similar
to the definition of the Reynolds (Re) number in forced convection problems (ratio of the inertial forces
to the viscous forces), Gr is defined as the ratio of the buoyancy forces over viscous forces acted:
δ ⋅v
Re =
 µ
 ρ 
(15.5)
g ⋅ β ⋅ ∆T ⋅ δ 3
Gr = 2
 µ
 ρ 

where g (9.81 m/s2) is the gravitational acceleration, v is the velocity (m/s), β is the volumetric expan-
sion coefficient (1/K), ΔT is the temperature difference (K), δ is the characteristic dimension (m), μ is
the dynamic viscosity (Pa-s) and ρ is the density (kg/m3). The ratio of dynamic viscosity to density is
also known as kinematic viscosity (m2/s). The ratio of Gr to the square of Re determines the relative
Gr
effects of natural and forced convection. When is greater than 1, the relative effects of the buoy-
Re 2
ancy forces dominate the convection over the inertial forces; when the inverse case holds, the inertial
forces dominate [4]. Heat transfer studies of natural convection in the literature mostly focus on the
experimental cases to derive correlations to determine the heat transfer coefficient. In these correla-
tions, the common approach is to determine the Nusselt (Nu) number as a function of the Rayleigh
(Ra) number, where the Ra is defined as the product of Gr and Prandtl (Pr) numbers:
Nu = f ( Ra)
g ⋅ β ⋅ ∆T ⋅ δ 3 c p ⋅ µ
Ra = Gr ⋅ Pr = ⋅ (15.6)
 µ
2
k
 ρ 

where cp is the specific heat capacity (J/kg-K) and k is the thermal conductivity (W/m-K) of the
fluid. For this purpose, various correlations defining various heat transfer conditions over surfaces
and inside the cavities have been experimentally developed and validated.
With advances in computational power, however, the solution of the Navier–Stokes equations
coupled with the energy and continuity equations to determine the flow field velocity and pressure
changes with temperature distribution was introduced. Datta and Teixeira [5] were the first in the
food engineering literature to present the numerical solution of the momentum equations coupled
with continuity and energy equations. In this study, temperature and velocity profiles were predicted
in a cylindrical can using an axi-symmetry condition with constant liquid properties (except the
density in the buoyancy term of the momentum equation in the vertical direction).
15.2.2 Mathematical Background
During a natural convection heat transfer problem, fluid flow and heat transfer are governed by the
Navier–Stokes equations: continuity equation, conservation of momentum and conservation of energy.
These equations might be derived as follows in differential equation format [6]:
• continuity equation:
∂U j
=0 (15.7)
∂x j

• conservation of momentum:
∂Ui ∂UiU j 1 ∂P ∂   ∂Ui ∂U j 

+ =− ⋅ + ν ⋅  + − ui′u′j   (15.8)
∂t ∂x j ρ ∂xi ∂x j   ∂x j ∂xi  

• conservation of energy:
∂T ∂T ∂  ∂T 
ρ ⋅ cv ⋅ + ρ ⋅U j ⋅ cp ⋅ =  k⋅ − ρ ⋅ c p ⋅ u ′j T ′  (15.9)
∂t ∂x j ∂x j  ∂xi 

where U is the average velocity (m/s), u′ is turbulent component of velocity (m/s) and ui′u ′j is the
average of the fluctuating component of velocity. T and T′ are temperature (K) and fluctuating com-
ponent of temperature (K), P is pressure (Pa), ρ is density (kg/m3), v is kinematic viscosity of the
fluid (m2/s) and cp and cv are heat capacity (J/kg-K) at constant pressure and volume, respectively.
The differential forms are:
∂ ∂ ∂ ∂
= , ,
∂xi ∂x ∂y ∂z
(15.10)
∂ ∂ ∂ ∂
= + +
∂x j ∂x ∂y ∂z

∂ ∂ ∂ ∂
Uj ⋅ = Ux ⋅ + U y ⋅ + Uz ⋅ (15.11)
∂x j ∂x ∂y ∂z

Due to the fluctuating components, there might be evolving turbulence in the flow, and this
might require additional solution approaches. Regarding the required turbulence quantities, κ – ε,
κ – ω, Reynolds stress models are special turbulence models to solve this issue. Olsson, Ahrne and
Tragardh [7] report detailed information for the comparisons of these models with the experimen-
tal data available in the literature. To include the natural convection effects, a buoyancy term is
included in the momentum equations in the gravitational direction.
One of the famous natural convection problems for food processing is the sterilization–pasteurization
of canned food products. The movement of the cans during processing, e.g., horizontal rotation,
end-over-end rotation or horizontal oscillated movement, might induce turbulence within the cans
due to the viscosity of the canned liquid. This is especially valid for low-viscosity Newtonian liq-
uids. To numerically evaluate the temperature changes during canning processes, the given govern-
ing equations might be simplified as follows in cylindrical coordinates where an axi-symmetry
condition might also be applied in the vertical direction where the buoyancy term is applied [5]:
• continuity equation:
1 ∂ ∂
⋅ (r ⋅ ρ ⋅ v) + ( ρ ⋅ u) (15.12)
r ∂r ∂z
• conservation of momentum in radial and vertical (where the buoyancy effects are present)
directions:
 ∂v ∂v ∂v  ∂P  ∂ 1 ∂  ∂2 v 
ρ ⋅ + v ⋅ + u ⋅  = − + µ ⋅   ⋅ (r ⋅ v) + 2 
 ∂t ∂r ∂z  ∂r  ∂r  r ∂r  ∂z 
(15.13)
 ∂u ∂u ∂u  ∂P  1 ∂  ∂u  ∂2 u 
ρ ⋅ + v⋅ +u⋅  = − + µ ⋅ ⋅ r ⋅  + 2  + ρ ⋅g

 ∂t ∂r ∂z  ∂z  r ∂r  ∂r  ∂z 
• conservation of energy:
∂T ∂T ∂T k  1 ∂  ∂T  ∂2 T 
+ v⋅ +u⋅ =  ⋅  r ⋅ + 
∂t ∂r ∂z ρ ⋅ c p  r ∂r ∂r  ∂z 2  (15.14)

Temperature dependency of density is either used to be a function of temperature or to follow the

Boussinesq approach:
ρ = ρ0 ⋅ 1 − β ⋅ ( T − T0 )  (15.15)

where T0 and ρ0 are temperature and density at the reference conditions. In the numerical solution of
these equations, a no-slip condition over the solid boundaries and a 3rd kind convection boundary
condition (or constant surface temperature in a simplified form) are mostly applied. Due to increase in
computational power and available software, computational fluid dynamics (CFD) software has been
recently preferred for the solutions of these equations. The common approach is just to determine tem-
perature and flow field changes within a cavity representing the product or a food processing system.
Use of these solutions for further design and optimization purposes are still scarce in the literature.
While certain approximations, like axi-symmetric or symmetric approaches, have been preferred to
reduce the computational time, some studies focused on solving the equation for the exact geometry
without any simplification either due to non-existence of symmetry-related approaches or because of

the possible deviations from expected symmetry solutions. Even though symmetry or axi-symmetry
might be observed in a geometrical shape, it should not directly lead to the expectation of a sym-
metry in the flow or temperature field [8]. This is true for complex geometries, which (and especially
for liquid-gas interface problems) cause disturbances in the flow with the formation of vortices and
eddies. In simple geometries, however, it might still be a good idea to use the symmetry approaches
to reduce the computational time. Figure 15.1 is an example of using an exact geometry rather than a
symmetry case in a square cavity to demonstrate the symmetry in temperature (Figure 15.1a) and flow
field velocity (Figure 15.1b) distributions. In this case, a 3rd kind convective boundary condition in
heat transfer and no-slip conditions in momentum transfer were applied to heat up the viscous liquid
in the cavity, including 0.5% CMC (all thermal and physical properties were assumed to be tempera-
ture dependent and obtained from Tuta and Palazoglu [9]). As indicated, square geometry might be
accepted as a simple geometry (considering the presence of a liquid in the cavity), and due to the
higher viscosity of the 0.5% CMC, computational requirements might be achieved readily compared
to lower viscosity cases where higher number of iterations are required for convergence. Compared to
the perfect symmetry in the temperature and flow field velocity (Figure 15.1a,b), Figure 15.1c shows
FIGURE 15.1 A 2-dimensional square cavity filled with liquid (0.5% CMC) exposed to convective heat
transfer boundary condition along all the surfaces to demonstrate (a) temperature distribution symmetry
during natural convection heating (scale in K), (b) Flow field velocity distribution in the 2-dimensional case.
(Continued)
FIGURE 15.1 (CONTINUED) A 2-dimensional square cavity filled with liquid (0.5% CMC) exposed to
convective heat transfer boundary condition along all the surfaces to demonstrate (c) temperature distribution
(scale in K) in a 2-dimensional circular cross-section including liquid (water) and headspace (air), (d) change
in the air–water interface in the 2-dimensional cross-section during natural convection heating,
the temperature distribution in a circular geometry filled with water. In addition, a certain amount of
headspace, assumed to consist of air, was also considered for this simulation. The outer surfaces of
the two-dimensional circular cavity were also assumed to be exposed to the 3rd kind of convective
heat transfer coefficient. The viscosities of these two liquids, water and 0.5% CMC, are significantly
different (0.001 Pa-s for water versus 0.5 Pa-s for 0.5% CMC at 20 C). Presence of the headspace in
the second case also increased the complexity in the system by requiring more computational time. As
observed in Figure 15.1c, temperature distribution did not conform to the symmetry expected from the
geometrical symmetry. This issue might be considered in the natural convection-related problems for
low viscosity cases and complex geometries, considering that geometrical symmetry might not always
resemble the temperature or flow field symmetry. Figure 15.1c was obtained after a certain process
time, while Figure 15.1d demonstrated the change in the headspace volume (red color section is the
air headspace) due to the evolving velocity in the flow field due to the formation of natural convection.
In Figure 15.1a,b, a convective boundary condition was applied over all four surfaces. However,
these figures indicated a significant effect of the natural convection starting from side walls where,
in their vicinities, density change-triggered natural convection currents took place due to the tem-
perature change. This phenomena was observed over the bottom wall with the presence of convec-
tive cell formations, called Benard convection cells, while there was no effect of natural convection
heating through the heat transfer effect on the top wall. This is due to the fact that the heated liquid
would get lighter with the increase in its temperature, and its upward movement is started with the
nature of the natural convection. In the case of the top wall, the heated fluid in the vicinity would
not be able to make an upward movement in the closed cavity, and it could not be deflected to make
a downward movement since there was no cooling effect. Then, a conduction similar heat trans-
fer might take an effect when the only heating effect is applied over the top wall. Figure 15.2a,b
demonstrate these cases where a heat transfer boundary condition is applied only over the side
walls and only over the top wall, respectively. The evolved velocity field might also be observed
in Figure 15.2a with a very similar temperature and velocity distribution as the case shown in
Figure 15.1a, with the absence of the convection cells formed over the bottom wall.
FIGURE 15.2 Temperature distribution (scale in K) in a 2-dimensional square cavity filled with liquid
(0.5% CMC) exposed to convective boundary conditions along (a) the side walls, (b) the top wall.
The random nature of Benard convection cell formations from the bottom wall were also
observed in these cases. These cells are formed by the instability resulting from the heated bottom
and contacting the heated fluid with the colder ones above. This leads to a burst of the convective
cells, as also explained by Abdul Ghani et al. [10]. While this formation was observed right in the
middle of the bottom surface in Figure 15.1a, regular formation was not seen in the circular geom-
etry case (Figure 15.1c). The finer mesh structure, especially at the walls where steep gradients in
velocity and temperature are observed due to the effect of sudden heating, are preferred to keep the
discretization errors as small as possible [11]. Vargas et al. [12] also indicated the requirement to
use a three-dimensional velocity field to make a correct comparison of experimental and numerical
data, especially for the complex geometry and flow systems. The solutions for Figure 15.1a,b and
Figure 15.2a,b were obtained using COMSOL Multiphysics (V. 5.3, COMSOL AB, Stockholm,
Sweden) while the cases for Figure 15.1c,d were obtained using ANSYS Fluent® V. 18 (ANSYS Inc.,
Canonsburg, PA, USA) CFD software.
15.2.3 Examples from the Studies in the Literature

Following the first numerical modeling study on natural convection heating in the food engineering
literature by Datta and Teixeira [5], Kumar et al. [13] and Kumar and Battacharya [14] used canned
liquids to predict the changes in temperature and flow field for viscous and non-Newtonian liquids.
In these studies, use of a commercial finite element software (FIDAP–Fluid Dynamics International,
Inc., Evanston IL) was demonstrated. In the first study [13], heat transfer from the side surfaces was
considered assuming that the bottom and top surfaces were insulated. In the following study, how-
ever, heat transfer from bottom was also included in the solutions, and no-formed eddies or circulat-
ing loops were observed due to the high viscosity liquid. Some of the pioneering studies, including
the experimental works, on natural convection can be listed with the studies by Evans et al. [15];
Torrence and Rocket [16]; Engelman and Sani [17] and Naveh et al. [18]. While the modeling studies
increased their impact with the evolution of computational power and the available software, Ostrach
[19] outlined the complexity of the natural convection phenomena in enclosures and presented the
implications for numerical methods. Inadequacies of the pure numerical solutions, due to the difficul-
ties in composing a proper mesh, and the requirement of the experimental guidance and validation
of the models were also pointed out.
Abdul Ghani et al. [10] simulated the natural convection heating in a can filled with liquid
(water and carboxy-methyl-cellulose—CMC solution). In this study, a commercial CFD package
(PHOENICS) was used, and in both liquids, viscosity was temperature-dependent. For the case of
CMC, shear rate dependency was neglected. Density change was constant except in the buoyancy
term, and temperature dependency of the density was expressed with the Boussinesq approach. It
was noted that, due to the lower viscosity values of water, more iterations (for convergence of the
solvers) were required. Significant effects of viscosity over other thermal-physical properties on
the evolution of the temperature field and its dependence upon temperature were also addressed.
Calderon-Alvarado et al. [20] also focused on viscosity effects by comparing two viscosity models’
effect on the lethality assumption of canned liquids. Abdul Ghani et al. [21] repeated their first study
[10] for the case of a horizontal can filled with carrot-orange soup. While the former study applied an
axi-symmetric condition in the vertical direction, a three-dimensional approach was applied in the lat-
ter case. Abbasnezhad et al. [22] used a three-dimensional approach in numerical modeling of ther-
mal processing of intact egg where natural convection in egg white was modeled using COMSOL
Multiphysics program (V.3.5a).
Symmetry or axi-symmetry approaches along the vertical direction were also applied by Sakai
et al. [23], Varma and Kanna [11,24], Erdogdu et al. [2], Kiziltas et al. [25], Erdogdu and Tutar [26],
Koribilli et al. [27] and Moraga et al. [28] to reduce the computation time [29]. As demonstrated by
Boz and Erdogdu [30], an axi-symmetry condition (along the horizontal direction compared to the
vertical cans) in horizontal cans was not possible to apply due to the effect of gravity in the enclosure.
Figure 15.3a shows the temperature distribution for the case where axi-symmetry approach was
applied over horizontal (3a) directions. As observed, there was a symmetrical mirror image where
the temperature distribution evaluated for the horizontal case was not correct and was physically
impossible to hold. The solutions for these cases were obtained using COMSOL Multiphysics V.5.3.
The axi-symmetric approach for vertical case is explained in the examples below, and the correct
temperature distribution for horizontal case (in the given cross-section of Figure 15.3a) is shown in
Figure 15.3b. This result was obtained using a three-dimensional simulation study.
Most of the CFD studies on natural convection in the literature neglected the presence of an
air gap, or headspace, within the cavity, assuming that its effect on heat transfer and flow field
were negligible due to its lower volume. However, the presence of headspace promoted the mixing,
especially when there was an external force applied to increase the heat transfer rate. As explained
above, thermal processing of canned liquid foods is a perfect example of natural convection heating
in food processing, and the presence of headspace in canned foods is inevitable due to the process
requirements on holding the pressure increase in the can in balance to prevent bulging of cans and
also holding vacuum. This headspace also increases the mixing in horizontal rotation, end-over-end
rotation or horizontal oscillated shaking processes. Erdogdu and Tutar [26] determined the effect
of 10% headspace (located at the top of the liquid) in a #1 can (36.5 mm in radius and 110 mm
FIGURE 15.3 Temperature distribution (scale in K) in a horizontal cylindrical cavity (a) after applying the
axi-symmetry approach over the horizontal directions, (b) temperature distribution in the same cross-section
after applying a 3-dimensional approach.
in height) on the evolution of temperature and flow field using a 2D axi-symmetry approach with
the volume of fluid (VOF) model using Fluent V.6.3.26. All thermal and physical properties were
assumed to be temperature-dependent, and a mesh independency study was also carried out prior
to the experimental validation of the numerical results. In this study, headspace was assumed to be
air, and the interface movement along the liquid–gas interface was demonstrated with the veloc-
ity contours. It was demonstrated that the headspace effect did not contribute any insulation effect
in contrast to some studies previously reported in the literature [10,31–33]. Kumar et al. [13] and
Joseph et al. [34], however, indicated faster heating rates with the presence of headspace. It was also
observed by Erdogdu and Tutar [26] that the organized velocity motion along the air–water interface
resulted in the oscillation motion of the interface, and this affected the temperature increase with
the formation of velocity and vortex formations inside the liquid. Figure 15.4 shows the interface
FIGURE 15.4 Liquid (water)–gas (air headspace) interface movement (a) at 0 and 60 s and velocity vectors
along the interface (b) at processing time of 1 and 60 s during natural convection heating of a vertical can (the
solution was obtained using an axi-symmetry approach). (From Tutar, M. and Erdogdu, F., J. Food Sci., 76,
E119, 2011.)
movement (Figure 15.4a) at 0 and 60 s and velocity vectors along the interface (Figure 15.4b) at pro-
cessing time of 1 and 60 s. These changes for both liquid and gas phases were presented in Figure
15.4 along the interface using a computational geometry similar to the case shown in Figure 15.5a.
ANSYS Fluent V. 12 (ANSYS Inc., Canonsburg, PA, USA) was used to obtain the interface move-
ment and velocity changes along the interface in Figure 15.6. In this comprehensive study by Tutar
and Erdogdu [3], numerical simulation for heat transfer and velocity field characteristics of a two-
phase (air with low and high viscosity liquid) flow system undergoing an axial rotation was carried
out in a two-dimensional cylindrical cavity. The effect of headspace with the rotation rate was dem-
onstrated. In the viscous liquid, counter to the expectation, the increases rotation rates negatively
affected the temperature evolution due to the evolution of higher viscous forces.
The general assumptions on the studies reported in the literature for modeling the natural con-
vection heating in a cavity might be summarized, as also detailed by Das et al. [35]:
• An incompressible fluid assumption was applied,

• Newtonian or non-Newtonian approaches to quantify the effects of temperature and shear
rate on viscosity were used: due to the lower motion of the fluid inside the (because of the
natural convection triggered flow), the shear rate dependency was mostly neglected,
• Depending upon the temperature difference (to start-up) the natural convection, laminar
flow was mostly assumed,
• Two-dimensional symmetry and axi-symmetry approaches were used assuming that the
geometrical symmetry holds for the temperature and flow field symmetry, and
• Thermal-physical properties of the fluid were constant or temperature dependent: con-
stant viscosity approaches were generally avoided, and density was assumed to function
of temperature or to change with the Boussinesq approach in the body force term of the
momentum equation in the gravitational direction.
15.3 CFD EXAMPLES
15.3.1 Natural Convection Heating of a Canned Liquid
As indicated above, natural convection heating is a rather significant heat transfer case in pasteurization–
sterilization of canned liquids. To demonstrate the changes in the flow and velocity field, the Eqs.
15.12 to 15.14 were solved in #1 can (36.5 mm in radius and 110 mm in diameter) using COMSOL
V.5.3 (COMSOL AB, Stockholm, Sweden) Multiphysics program for the two-dimensional axi-
symmetric vertical cylinder case (Figure 15.5). Angular direction effects were not included in
these simulations, and a two-dimensional axisymmetric approach was assumed to hold due to the
same boundary conditions through all the surfaces of the cylindrical cavity. The objective for this
approach was to reduce the computational time. However, when the same example was considered
for the horizontal cans, a three-dimensional approach should have been applied. The initial and
boundary conditions for this case were:
• No-slip conditions for velocity components (u = v = 0) at the walls with a 3rd kind convec-
tive boundary condition with a constant heat transfer coefficient and constant wall tem-
perature were imposed,
• The liquid (water and 0.5% CMC) was at a constant uniform initial temperature (20°C)
initially,
• Thermal and physical properties of the liquid (water and 0.5% CMC) were assumed to
be temperature-dependent (the temperature dependence of thermal conductivity, den-
sity, specific heat, viscosity and volume expansion coefficient for water were obtained
from Cengel and Ghajar [4], while the properties for CMC were obtained from Tuta and
Palazoglu [9]):
FIGURE 15.5 (a) The mesh structure applied in a 2-dimensional axi-symmetric case (where the axi-symmetry
was applied along the vertical direction: r = 0) for a vertical cylinder filled with water 0.5% CMC. (b) Temperature
(scale in K) distribution for water and 0.5% CMC cases during the natural convection heating. (Continued)
• The presence of the headspace was neglected to reduce the computational time and also
decrease the mesh size (liquid–gas interactions require a rather dense mesh structure along
the interface in the cost of computational time). The presence of the headspace in a vertical
cylindrical can included liquid and solid spherical particles and will be demonstrated in
section 3.2.
A high resolution mesh structure was used to reduce the discretization errors. The dense mesh
was specifically applied in the vicinity of the walls where a sudden changes in temperature and
velocity are expected. Figure 15.5a shows the mesh structure applied in the two-dimensional axi-
symmetric case (where the axi-symmetry was applied along the vertical direction: r = 0). Temperature
changes over the time and velocity in the flow field are shown in Figure 15.5b and c. As observed
in Figure 15.5b, the liquid (both water and 0.5% CMC) adjacent to the side walls received a higher
heat transfer compared to the bottom wall, due to the surface area, and this led to a temperature
increase and a lower density. Since a temperature and density difference evolved within the liquid
adjacent to the wall and the liquid away from the wall, there was a buoyancy force created, induc-
ing an upward motion. This motion was deflected by the top surface towards the radial direction,
while the liquid got colder with increased density, leading to a downward motion. This recirculating
flow is observed in the flow field for both water and 0.5% CMC (Figure 15.5c), and it causes the
cold temperature region to occur at the bottom sections of the can instead of the geometrical center,
FIGURE 15.5 (CONTINUED) (c) Velocity field evolved in the system for 0.5% CMC.
as seen in the conduction heat transfer. The range of Rayleigh numbers is significant to generate an
oscillatory and chaotic flow. The resulting turbulence might then need to be considered for selection
of the turbulent models in the solution. When these two cases were compared for the temperature
evolution, the temperature increase for water was greater than for 0.5% CMC due to the rather sig-
nificant difference in their viscosity values, even though other thermal-physical properties (thermal
conductivity, specific heat and density) were similar. The apparent dynamic viscosity value of 0.5%
CMC, as applied in the simulation, changed from 0.5 Pa-s at 20 C to 0.1 at around 70 C.
Heat transfer literature especially focuses on the analysis of the dimensional groups in the
natural convection studies for scale-up possibilities. Erdogdu and Tutar [26,36] demonstrated the
changes in Ra-number distribution through the natural convection heating in a vertical cylindrical
cavity and a two-dimensional circular cavity. In these cases, the headspace was present in the liquid
(water—low viscosity and semi-fluid—high viscosity) filled cavities. A recent review by Das et al.
[35] is suggested for a detailed analysis of Ra-number in various cavities under the effect of natural
convection heating.
15.3.2 Natural Convection Heating of Canned Liquid with Solid Particles

While a certain number of cases focused on natural convection heating of one or two phases,
canned liquids including solid particles and headspace is another very complex example. There
have been some studies in the literature focusing on developing dimensionless correlations to
determine an effective heat transfer coefficient for evaluation of temperature change inside
the cans. Few studies, like Abdul Ghani and Farid [37], Rabiey et al. [38], Kiziltas et al. [25],
FIGURE 15.6 (a) Temperature (scale in K), (b) Flow field velocity distribution in a 2-dimensional circular
cross-section including water and solid–circular particles during a natural convection heating.
Dimou and Yanniotis [39], Dimou et al. [40,41] and Cordioli et al. [42] focused on determining
temperature distribution, mostly neglecting the effect of headspace and sometimes assuming
a 2-D axi-symmetric approach in this natural convection problem. Singh et al. [43] presented
a detailed review of the computational techniques to determine the temperature change in
canned liquid-particulate mixtures. Besides these reported recent computational works, deter-
mining the temperature history at the center of a particle located in the coldest region of the
liquid was the general approach in the literature, especially for the experimental studies.
FIGURE 15.7 (a) A 3-dimensional can containing liquid (2% CMC) and solid spherical particulates with
10% headspace (in volume) located at the top, (b) Computational geometry with the mesh structure. (Modified
from Sarghini, F. and Erdogdu, F. 2016. A computational study on heat transfer characteristics of particu-
late canned foods during end-over-end rotational agitation: Effect of rotation rate and viscosity. Food and
Bioproducts Processing. 100:496–511.)
Boz [44] demonstrated the temperature and flow field velocity distribution in a two-dimensional

circular geometry where circular solid items were located in water. Figure 15.6a shows the for-
mation of the natural convection heating for temperature distribution while the flow field veloc-
ity changes are observed in Figure 15.6b. These results were obtained using ANSYS CFX V.12
(ANSYS Inc., Canonsburg, PA). Sarghini and Erdogdu [45] developed a numerical model for
end-over-end processing of this case while the improvements in the temperature increase were
compared with the vertical case where the natural convection heating dominated the whole pro-
cess. Figure 15.7 shows the given system and the computational geometry, where a can includes
liquid (2% CMC—this high-viscosity, shear-dependent liquid was used to assume that the solid
particulates would be suspended within the volume), solid–spherical particles and 10% head-
space (in volume) located at the top. A constant surface temperature boundary condition (due
to the expected high heat transfer coefficient over the external boundaries) and no-slip condi-
tion over the external walls and spheres were applied in the set of the Navier–Stokes equations.
Figure 15.8a showed the evolution of temperature in the x-y-z cross-sections with the limited
effect of the natural convection due to the high viscosity value of the 2% CMC (Figure 15.8b—
laminar viscosity distribution over the x-y-z cross-sections). ANSYS Fluent V.18 (ANSYS Inc.,
Canonsburg, PA, USA) was used to obtain the temperature and viscosity distributions observed
in Figure 15.8.
FIGURE 15.8 Evolution of temperature (scale in K) in the x-y-z cross-sections of a vertical can contain-
ing liquid (2% CMC)–solid particles and headspace particles, (b) Laminar viscosity distribution over x-y-z
cross-sections. (Modified from Sarghini, F. and Erdogdu, F. 2016. A computational study on heat transfer
characteristics of particulate canned foods during end-over-end rotational agitation: Effect of rotation rate and
viscosity. Food and Bioproducts Processing. 100:496–511.)
15.4 CONCLUSIONS
While heat transfer is the dominating transport mechanism in food processing operations, heat
transfer by natural convection in an enclosure (internal convection) is a significant problem due to its
dependence on the flow–momentum transfer. The flow formation in the given enclosure starts due
to the density changes in the fluid occurring with temperature changes. The changes in temperature
and flow field in the cavity must be known for further design and optimization problems. Canning
makes up a significant portion of thermal processing, and quantifying the natural convection effects
in canned foods containing liquid (low or high viscosity) or liquids with particulates is required to
design a proper process.
This chapter focused on canning models with cylindrical cavities representing a can. Temperature
and flow field changes in a square cavity were also represented to demonstrate the natural convection-
dependent temperature evolution. With the temperature and flow field changes, the solution of the
Navier–Stokes equations with the energy equation ie required for a complete mapping of a natu-
ral convection problem. This makes the natural convection problem a perfect case for CFD use.
For this purpose, various CFD programs (ANSYS, ANSYS-CFX, ANSYS-Fluent and COMSOL
Multiphysics) were used in the simulations demonstrated in this chapter. In addition, the food engi-
neering literature was also summarized from the perspective of natural convection.
REFERENCES
1. Laguerre O., Flick D., Heat transfer by natural convection in domestic refrigerators, Journal of Food
Engineering, 62: 79–88, 2004.
2. Erdogdu F., Uyar R., Palazoglu T.K., Experimental comparison of natural convection and conduction
heat transfer, Journal of Food Process Engineering, 33: 85–100, 2010.
3. Tutar M., Erdogdu F., Numerical simulation for heat transfer and velocity field characteristics of two-
phase flow systems in axially rotating horizontal cans, Journal of Food Engineering, 111: 366–385,
2012.
4. Cengel Y.A., Ghajar A.J., Heat and Mass Transfer—Fundamentals & Applications (Chapter 9), 5th ed.,
McGraw-Hill, New York, USA, 2015.
5. Datta A.K., Teixieira A.A., Numerically predicted transient temperature and velocity profiles during
natural convection heating of canned liquid foods, Journal of Food Science, 53: 191–195, 1988.
6. Erdogdu F., Ferrua M., Singh S.K., Singh R.P., Air-impingement cooling of boiled eggs: Analysis of flow
visualization and heat transfer, Journal of Food Engineering, 79: 920–928, 2007.
7. Olsson E.E.M., Ahrne L.M., Tragardh A.C., Heat transfer from a slot air jet impinging on a circular
cylinder, Journal of Food Engineering, 63: 393–401, 2004.
8. Sarghini F., Discussion on the symmetry issues in heat and momentum transfer problems and turbu-
lence evolution, Personal communication, 2017.
9. Tuta S., Palazoglu T.K., Finite element modeling of continuous-flow microwave heating of fluid foods
and experimental validation, Journal of Food Engineering, 192: 79–92, 2017.
10. Abdul Ghani A.G., Farid M.M., Chen X.D., Richards P., Numerical simulation of natural convection
heating of canned food by computational fluid dynamics, Journal of Food Engineering, 41: 55–64, 1999.
11. Varma M.N., Kannan A., Enhanced food sterilization through inclination of the container walls and
geometry modifications, International Journal of Heat and Mass Transfer, 48: 3753–3762, 2005.
12. Vargas M., Sierram F.Z., Ramos E., Avramenko A.A., Steady natural convection in a cylindrical cavity,
International Communications in Heat and Mass Transfer, 29: 213–221, 2002.
13. Kumar A., Bhattacharya M., Blaylock J., Numerical simulation of natural convection heating of canned
thick viscous liquid food products, Journal of Food Science, 55: 1403–1411, 1420, 1990.
14. Kumar A., Bhattacharya M., Transient temperature and velocity profiles in a canned non-Newtonian
liquid food during sterilization in a still-cook retort, International Journal of Heat & Mass Transfer, 34:
1083–1096, 1991.
15. Evans L.B., Reid R.C., Drake E.M., Transient natural convection in a vertical cylinder, American
Institute of Chemical Engineers Journal, 14: 251–259, 1968.
16. Torrence K.E., Rocket J.A., Numerical study of natural convection in an enclosure with localized heat-
ing from below: Creeping flow to the onset of laminar instability, Journal of Fluid Mechanics, 36:
33–54, 1969.
17. Engelman M., Sani R.L., Finite element simulation of an in-package pasteurization process, Numerical
Heat Transfer, 6: 41–54, 1983.
18. Naveh D., Kopelman I.J., Pflug I.J., Sterilization of food in containers with an end fat against a retort
bottom: Numerical analysis and experimental measurements, Journal of Food Science, 49: 1086–1093,
1984.
19. Ostrach S., Natural convection in enclosures, Journal of Heat Transfer, 110: 1175–1190, 1988.
20. Calderon-Alvarado M.P., Alvarado-Orozco J.M., Herrera-Hernandez E.C., Martinez-Gonzalez G.M.,
Miranda-Lopez R., Jimenez-Islas H., Effect of two viscosity models on lethality estimation in steriliza-
tion of liquid canned foods, Food Science and Technology International, 22: 496–515, 2016.
21. Abdul Ghani A.G., Farid M.M., Chen X.D., Numerical simulation of transient temperature and velocity
profiles in a horizontal can during sterilization using computational fluid dynamics, Journal of Food
Engineering, 51: 77–83, 2002.
22. Abbasnezhad B., Hamdami N., Montenau J.-Y., Vatankhah H., Numerical modeling of heat transfer and
pasteurizing value during thermal processing of intact egg, Food Science & Nutrition, 4: 42–49, 2015.
23. Sakai N., Tang J.W., Liu C.M., Watanebe M., Predicting temperature during the thermal processing of
canned high-viscosity liquid food, Food Science and Technology Research, 10: 79–85, 2004.
cylindrical containers, Journal of Food Engineering, 77: 1024–1036, 2006.
25. Kiziltas S., Erdogdu F., Palazoglu T.K., Simulation of heat transfer for solid-liquid food mixtures in cans
and model validation under pasteurization conditions, Journal of Food Engineering, 97: 449–456, 2010.
convection heating in cans, Journal of Food Science, 76: E119–E129, 2011.
27. Koribilli N., Aravamudan K., Varadhan M.U.S.V.A., Quantifying enhancement in heat transfer due
to natural convection during canned food thermal sterilization in a still retort, Food and Bioprocess
Technology, 4: 429–450, 2011.
28. Moraga N., Torres A., Guarda A., Galotto M.J., Non-Newtonian canned liquid food, unsteady fluid
mechanics and heat transfer prediction for pasteurization and sterilization, Journal of Food Process
Engineering, 34: 2000–2025, 2011.
29. Augusto P.E.D., Cristianini M., Evaluation of geometric symmetry condition in numerical simulations
of thermal process of packed liquid food by computational fluid dynamics (CFD), International Journal
of Food Engineering, 6(5): Article 1, 2010.
30. Boz Z., Erdogdu F., Evaluation of two-dimensional approach for computational modeling of heat
and momentum transfer in liquid containing horizontal cans and experimental validation, Food and
31. Weintraub S.E., Ramaswamy H.S., Tung M.A., Heating rates in flexible packages containing entrapped
air during overpressure processing, Journal of Food Science, 54: 1417–1421, 1989.
32. Ramaswamy S., Grabowski S., Influence of entrapped air on the heating behavior of a model food pack-
aged in semi-rigid plastic containers during thermal processing, LWT–Food Science and Technology,
29: 82–93, 1996.
33. Mohamed I.O., Determination of an effective heat transfer coefficients for can headspace during ther-
mal sterilization process, Journal of Food Engineering, 79: 1166–1171, 2007.
34. Joseph S.J., Speers R.A., Pillay V., Effect of head space variation and heat treatment on the thermal
and rheological properties of non-agitated conduction-heated materials, LWT–Food Science and
Technology, 29: 556–560, 1996.
35. Das D., Roy M., Basak T., Studies on natural convection within enclosures of various (non-square)
shapes—A review, International Journal of Heat and Mass Transfer, 106: 356–406, 2017.
36. Erdogdu F., Tutar M., A computational study for axial rotation effects on heat transfer in rotating cans
containing liquid water, semi-fluid system and headspace, International Journal of Heat and Mass
Transfer, 55: 3774–3488, 2012.
37. Abdul Ghani A.G., Farid M.M., Using the computational fluid dynamics to analyze the thermal ster-
ilization of solid–liquid food mixture in cans, Innovative Food Science and Emerging Technologies,
7: 55–61, 2006.
large particles in a cylindrical vertical can, Journal of Food Engineering, 82: 409–417, 2007.
39. Dimou A., Yanniotis S., 3-D numerical simulation of asparagus sterilization using computational fluid
dynamics, Journal of Food Engineering, 104: 394–403, 2011.
40. Dimou A., Panagou E., Stoforos N.G., Yanniotis S., Analysis of thermal processing of table olives using
computational fluid dynamics, Journal of Food Science, 78: E1695–E1703, 2013.
41. Dimou A., Stoforos N.G., Yanniotis S., Effect of particle orientation during thermal processing of peach
halves: A CFD simulation, Foods, 3: 304–317, 2014.
42. Cordioli M., Rinaldi M., Copelli G., Casoli P., Barbanti D., Computational fluid dynamics (CFD) model-
ing and experimental validation of thermal processing of canned fruit salad in glass jar, Journal of Food
Engineering, 150: 62–69, 2015.
43. Singh A., Singh A.P., Ramaswamy H.S., Computational techniques in heat transfer studies on canned
liquid-particulate mixtures, Trends in Food Science and Technology, 43: 83–103, 2015.
44. Boz Z., Two dimensional simulation of heat transfer in horizontal cans containing liquid and solid-
liquid mixtures, M.Sc. thesis, Mersin University, Mersin, Turkey, 2012.
45. Sarghini F., Erdogdu F., A computational study on heat transfer characteristics of particulate canned
foods during end-over-end rotational agitation: Effect of rotation rate and viscosity, Food and
16 CFD Modeling of Convective
Drying of Cylindrical Fruit
Slices
Alexandros Vouros, Dimitrios Tzempelikos,
Dimitrios Mitrakos, and Andronikos Filios
CONTENTS
16.1 Introduction..................................................................................................................... 339
16.2 General Description of Diffusion-Based Models for Convective Drying............................ 340
16.2.1 Conjugated Drying Models..................................................................................... 341
16.2.2 Non-Conjugated Drying Models............................................................................ 342
16.3 Convective Heat and Mass Transfer Coefficients...............................................................344
16.4 Advantages and Disadvantages of Drying Models..............................................................348
16.5 Computational Modeling of Heat and Mass Transfer in Cylindrical Fruit Slices............... 349
16.5.1 Model Formulation........................................................................................................349
16.5.2 CFD Simulations for the Determination of CTCs.................................................. 351
16.5.3 Thermophysical Properties..................................................................................... 353
16.5.4 Model Validation................................................................................................................ 355
16.6 Conclusions.......................................................................................................................... 359
Nomenclature............................................................................................................................. 360
References....................................................................................................................................... 361
16.1 INTRODUCTION
Drying of agricultural products is a stand-alone process in the food industry or a part of a chain-
process in the perfume and healthcare industries—for example, before the distillation of essential
oils. Wherever it is involved, drying is critical, considering the exploitation of the plant material and
the quality of the dried product. Drying is also important for food preservation and storage.
Convective drying is one of the most popular methods for drying of agricultural products in the
food industry. The main advantages of this method are its simplicity and low cost for the develop-
ment and installation of the related infrastructure. In convective drying, the moisture is removed
from the product—often in the form of slices—by means of a hot air stream, which is supplied
around the slices. Convective heat increases the temperature of the product and promotes a faster
migration of the water content from the interior towards the surface. Water is then carried away by
the air stream. The process is governed by the strong coupling of heat and mass transport in both
the product and the stream of the drying air. The whole process is controlled by many parameters,
but primarily the properties of the product and the drying conditions, including the temperature, the
velocity and the relative humidity of the hot air stream.
The rate at which the moisture content is removed affects the quality and must be adjusted
according to the morphology of the product to provide the desired effect. Simulating with accu-
racy the drying rate is crucial, considering the development of optimized “drying processes” that
fit the requirements of food production systems or other relevant industries. Considering the con-
vective drying of agricultural products and fruits, computational modeling is an important tool;
339
it contributes to a better understanding of heat and mass transport, and, when simply formulated,
it can be used for the improvement of the design and the control of the drying process in the food
industry.
Drying models can be divided in two prime categories, depending on the physical mech-
anisms presumed to drive the moisture migration inside the product [1]. The first category is
based on the theory of capillary, pressure-driven Darcy-like flow of moisture at the interior
of the dried product, which is often considered as a porous media. This approach seems to be
physically consistent and is expected to provide accurate results, provided that the properties of
the product are available [2]. Sophisticated computational platforms and algorithms used for the
representation of the structure of the porous inside the material support the development of such
models; however, the complex geometry, the uncertainties related to the properties of the porous
media, as well as the high computational cost make their development and extensive use quite
challenging.
The second category comprises the so-called diffusion-based models. In this approach, model-
ing is based on the presumption that the migration of the moisture in the product can be described
by Fickian diffusion theory. The main weakness of this approach is linked with the determination of
an effective diffusion coefficient for the transport of the moisture inside the product, which can only
be obtained from experimental correlations. Nevertheless, diffusion-based models can be relatively
simple. Such models are often combined with Computational Fluid Dynamics (CFD) for the simula-
tion of the flow field of the drying air around the product. Although not as physically consistent as
the porous media models, diffusion-based models have gained extensive attention and can be useful
in designing, optimizing and controlling the drying process. Extensive review of the different mod-
els can be found in the works of Datta [1] and Defraeye [3].
This chapter focuses on diffusion-based models and the computational approaches used for the
numerical simulation of the drying process of agricultural products. The reason is twofold: First,
these models compose the vast majority of the models found in the literature and second, they pro-
vide satisfactory agreement with the existing experimental data, by incorporating relatively simple
formulations and computational approaches. The chapter is structured as follows. First an overview
of the different diffusion-based models is presented. Conjugated and non-conjugated computational
approaches are then discussed. Regarding in particular non-conjugated models, a more extensive
description of the parameters is given, involving the Convective Transfer Coefficients (CTCs).
Strengths and weaknesses of the different models are highlighted. Finally, the successive applica-
tion of a recent model for the simulation of the drying of quince slices is reviewed. The details of the
model are described in brief, and some of the most important results are presented in comparison
with experimental data used for the validation of the model.
16.2 GENERAL DESCRIPTION OF DIFFUSION-BASED

MODELS FOR CONVECTIVE DRYING
Diffusion-based models are typically distinguished in conjugated and non-conjugated [3–5]. The
classification is based on the computational approach used for the treatment of the coupled heat and
mass transport in both the product and drying air domains.
In conjugated models, the unsteady transport equations are solved for both the drying air and the
product domains simultaneously. Computationally, these models are often called multi-region mod-
els, since different thermophysical properties are applied to the different regions that represent the
air and the product. These models do not require special treatment at the air-product interface; the
interaction between the different domains results from the solution of the boundary layer between
the air and the product, which can be resolved or modeled by standard wall functions. In some
CFD Modeling of Convective Drying of Cylindrical Fruit Slices 341
cases, more than two domains are considered in order to represent additional regions, such as the
epidermis of the product. The strong coupling of the heat and mass transfer equations and the bet-
ter representation of the geometry and structure of the product make these models highly accurate;
however, the complexity of the non-linear system of equations and the different scaling needed to
resolve the diffusion inside and outside the product make the solution procedure very sensitive and
computationally very demanding.
In non-conjugated models, the heat and mass transport are treated separately at the product and
the air stream domains. The solution for the drying air fluid flow and temperature fields is handled
independently of the moisture and heat transport in the product. Air flow characteristics (velocity,
temperature and humidity fields outside the product) are considered unaffected by the moisture
and heat exchange with the product so that the effect of the water vapor escaping from the product
surface on the fluid flow is neglected. Coupling between the heat and mass transport within the
product is often considered, in terms of the interdependencies between the various thermophysical
properties, including the effective diffusion coefficient, temperature and moisture content, changes
in product geometry and density, etc.
The typical approach of employing the appropriate CTCs to treat the heat and mass transfer at the
air-product interface is applied to non-conjugated models. The Convective Heat Transfer Coefficient
(CHTC) and Convective Mass Transfer Coefficient (CMTC) are used to form the boundary condi-
tions needed for the solution of the transient heat and moisture transport inside the product. The
choice of CTCs depends on the geometry of the product, the flow configuration, the flow character
(laminar or turbulent) and the range of the drying conditions.
16.2.1 Conjugated Drying Models

In conjugated models, the heat and mass transport in the air and the product are solved simultaneously
in a transient way, and the drying air and the product are treated as two interrelated systems of equa-
tions. At the interface of the two domains, continuity of heat and mass fluxes, temperature and mass
fraction in terms of water vapor is required. A phenomenological representation of the geometry and
the variables involved in conjugated models, considering in particular a cylindrical slice, is illustrated
in Figure 16.1.
Drying chamber
Unknown variables for drying air:

uair, pair, Tair, Cair
Unknown variables for product:

X p , Tp
FIGURE 16.1 Phenomenological representation of conjugated approach.

The exact series of computations required for the conjugated models has been reported by [4].
These include the determination of the geometrical characteristics and the thermophysical proper-
ties of the drying product, the definition of the initial conditions and the solution of heat and mass
transfer equations at the different computational domains that account for the drying air and the
product.
Conjugated models have been applied to simulate the convective drying of a wide variety of
fruits and vegetables. Various flow configurations and drying conditions have been examined.
Computational domains can be two- or three-dimensional while the flow, according to the flow
conditions of the drying air, can be considered laminar or turbulent. When the flow is turbulent,
standard wall functions are often used, since the computational requirements for the resolution of
the boundary layer are very demanding and thus high in computational cost.
Table 16.1 summarizes the studies on convective drying of agricultural products using the
conjugated approach. In most cases, the models are validated by experimental data for the
same product, geometrical configuration and drying conditions. Some of the most recent stud-
ies are those of Curcio et al. [6], Caccavale et al. [7], Khan & Straatman [8], Lemus-Mondaca
et al. [9], Cârlescu et al. [10], Defraeye & Radu [11]. Models extended to more than two regions,
in order to take into account the epidermis and the porous structure of the product, can be
found in the recent studies of Defraeye et al. [12], Defraeye & Verboven [13] and Manickathan
et al. [14].
16.2.2 Non-Conjugated Drying Models

In non-conjugated models, the heat and mass transfer within the product are solved, while the flow,
temperature and concentration of water vapor fields are considered unaffected and thus remain
“frozen” during the simulation. As already mentioned, an effective diffusion coefficient is used
to describe the transport of the moisture content inside the product. In contrast to the conjugated
models, appropriate boundary conditions, through using heat and mass CTCs, are necessary to
account for the heat and mass exchange between the product and the drying air at the interface. A
representation of the geometry and the variables involved in non-conjugated models, considering
in particular a cylindrical slice, is illustrated in Figure 16.2. Such boundary conditions are often
mentioned as third kind or Robin conditions and they are considered realistic from a physical con-
sistency point of view [1,2].
The series of computations required for the non-conjugated models has been reported by Lemus-
Mondaca et al. [4]. Computations include the determination of the geometrical characteristics and
the thermophysical properties of the drying product, the selection of appropriate CTCs, the defini-
tion of initial conditions and boundary conditions and the solution of the heat and mass transfer
equations inside the product.
The utilization of appropriate CTCs for non-conjugated modeling is often debated.
Sometimes, the origins of the CTCs are used to characterize furthermore the drying models
as semi-conjugated [3,5], in particular when the spatial distribution produced by separate flow
calculation or the temporal variation of the CTCs that involves the dependency on the moisture
content at the surface are considered. It has to be noted, however, that, no matter the origins
of the CTCs, the CTCs values are necessary in non-conjugated models and their estimation at
separate flow fields is not necessarily associated with the product. In any case, the above state-
ments indicate that non-conjugated models are inevitably related to CTCs and that they can be
discussed only in conjunction with CTCs. A more extended analysis of the CTCs used in the
most recent models of the literature is presented in the following paragraph.
TABLE 16.1
Summary of Studies on Convective Drying of Agricultural Products Using
Conjugated Models
Food Dimension of Fluid
Authors Material Product Geometry HMT Model Flow Drying Air Conditions
Curcio et al. [15] Carrot Cylindrical 2D Turbulent Tair = 35 to 55°C
axisymetric uair = 1 to 5 m/s
RH = 10 to 30%
De Bonis & Ruocco Carrot Rectangular slice 2D Laminar Tair = 50 to 80°C
[16] uair = 0.2, 2 m/s
RH = 45.75%
Lamnatou et al. [17] Apple Rectangular cylinder 2D Laminar Tair = 50°C
uair = 0.33, 0.67 m/s
RH = 9%
Lamnatou et al. [18] Apple Rectangular cylinder/ 2D Laminar Tair = 50°C
and two plates side by U=1
banana side arrangement RH = 9%
Lamnatou et al. [19] Apple Rectangular cylinder/ 2D Laminar Tair = 50°C
two plates in uair = 0.67 m/s
different arrangement RH = 9%
Marra et al. [20] Potato Rectangular disk 2D Turbulent Tair = 30°C
uair = 1.5 m/s
Curcio [21] Potato Cylindrical 2D Turbulent Tair = 50 to 70°C
axisymetric uair = 1 to 5 m/s
RH = 15 to 50%
Defraeye, Herremans, Apple Spherical/arbitrary 3D Turbulent Tair = 10°C
et al. [22] uair = 0.002 to 6.325 m/s
Sabarez [23] Prune Cylindroid 2D Turbulent Tair = 70, 80°C
axisymmetric uair = 1.5 to 7 m/s
RH = 15, 35%
Defraeye, Verboven & – Spherical 3D Turbulent Tair = 10°C
Nicolaï [24] uair = 0.002 to 6.325 m/s
Kurnia et al. [25] Potato Rectangular slice 2D Laminar/ Tair = 42°C
Turbulent uair = 2, 20 m/s
Curcio et al. [6] Potato Arbitrary 3D Laminar Tair = 25 to 90°C
RH = 10 to 45%
Caccavale et al. [7] – Rectangular 2D/3D Turbulent Tair = 100°C
Re = 1 × 104
RH = 0.01%
Khan & Straatman [8] Apple Rectangular slice 3D Laminar Tair = 60°C
uair = 0.64 m/s
RH = 9%
Lemus-Mondaca et al. Papaya Rectangular slice 3D Turbulent Tair = 60°C
[9] uair = 3 m/s
RH = 57%
Cârlescu et al. [10] Apricot Half-circular slice 2D Laminar Tair = 25 to 80°C
uair = 1 m/s
Defraeye & Radu [11] Apple Half-circular slice 2D Turbulent Tair = 20 to 60°C
uair = 0.01 to 10 m/s
RH = 10 to 90%
Abbreviation: HMT, Heat and Mass Transfer.

Drying chamber
Known variables for drying air:

CHTC, Tair, CMTC, Cair
Tp Xp
kp CHTCavg (Tp , s air ) h fg s Deff
n s
n s
Xp
s Deff CMTCavg ( C p , s Cair
n s
BCs at the interface air - product
Unknown variables for product: Xp, Tp
FIGURE 16.2 Representation of non-conjugated approach.
16.3 CONVECTIVE HEAT AND MASS TRANSFER COEFFICIENTS

Several approaches have been proposed in the literature for the derivation of the CTCs. These
involve constant values or empirical correlations for simplified flow configurations such as flow
around plates, cylinders, etc., time-dependent coefficients based on a zero-dimensional heat and
mass balance, optimization and regression analysis of experimental and numerical data and flow
field simulations using CFD. The above approaches have been adopted in multi- or one-dimensional
modeling of drying of various product geometries and flow configurations. An overview of the
recent non-conjugated models in relation to the determination of the CTCs is given in Table 16.2.
The following remarks can be made:
• For rectangular shaped slices, in the one-dimensional model of Agrawal et al. [26] empiri-
cal values have been used for CHTC and CMTC.
• For rectangular shaped fruit slices, in the two-dimensional models of Aversa et al. [27],
Kaya et al. [28], Oztop et al. [29], Villa-Corrales et al. [30] and Esfahani et al. [31] constant,
empirical, time-dependent and CHTC values produced by CFD were used. Constant or
time-dependent CMTC values were used.
• For the same geometry, in the three-dimensional models of Chandra Mohan & Talukdar
[32], Lemus-Mondaca et al. [33], Ateeque et al. [34], Esfahani et al. [35] and Pasban et al.
[36] constant, time-dependent and values produced by CFD were used for the CHTC. The
values of the CMTC were time-dependent or they were produced by the Heat and Mass
Transfer Analogy (HMTA).
• Considering CFD simulations for rectangular fruit slices, the air-drying stream was con-
sidered laminar. In the model of Ateeque et al. [33] the SST k-ω model was used to predict
the turbulent flow field around the product.
• For cylindrical shaped slices, in the one-dimensional models of Barati & Esfahani [37],
Barati & Esfahani [38], Da Silva et al. [39], Da Silva et al. [40], Tzempelikos et al. [41],
Da Silva et al. [42] and Vahidhosseini et al. [43] constant, empirical, CFD produced values
and values from optimization and regression analysis were used for the CHTC. Constant,
empirical values, values based on the HMTA and values dictated by optimization and
regression analysis were used for the CMTC.
TABLE 16.2
Summary of Studies on Convective Drying of Agricultural Products Using Non-Conjugated Approach
Food Product HMT Drying Air Values/Expressions Evaporation
Authors Material Geometry Model Conditions CHTC Values/Expressions for CHTC CMTC for CMTC Term
Kaya et al. Broccoli Cylindrical 2D Tair = 57°C CFD-La/2D 5 to 60 W/m2K HMTA 5 × 10–7 to 4.5 × 10–6 m/s No
[44] slice uair = 0.3 m/s
Aversa et al. Carrot Rectangular 2D Tair = 50, 60°C Empirical [45,46] HMTA [45,46] Yes
[27] slice uair = 0.3 m/s
RH = 45, 75%
Kaya et al. Kiwi Rectangular 2D Tair = 50°C CFD-La/2D 0 to 38 W/m2K HMTA 0 to 0.035 m/s No
[28] slice uair = 0.3, 0.6, 0.9 m/s
RH = 0%
Oztop et al. Apple & Rectangular 2D Tair = 60, 70, 80°C Constant Biapple = 0.17 to 3.62 Constant 0.73 × 10–6 to No
[29] potato slices uair = 1 and 1.5 m/s Bipotato = 0.03 to 70.80 2.21 × 10–6 m/s
0.21 × 10–6 to
39.15 × 10–6 m/s
Janjai et al. Mango Arbitrary slice 2D Tair = 50, 60, 70, 80°C – – Empirical CMTC = (Dair /d) (2.0 + No
[47] uair = 0.2 m/s 0.522 × Re0.5Sc0.33)
RH = 1.5 to 13.3%
Villa-Corrales Mango Rectangular 2D Tair = 50, 55, 60, 65, TDC TDC CMTC = (V/At) ln(MR) Yes
 h fg dM 
CFD Modeling of Convective Drying of Cylindrical Fruit Slices
et al. [30] slice 70°C CHTC =  c p m ∆T −  / (Ts − Tair )

 A dt 
uair = 0.2 m/s
RH = 15%
Chandra Apple Rectangular 3D Tair = 40 to 80°C CFD-La/2D 2.02 to 46.58 W/m2K HMTA – No
Mohan et al. slice uair = 0.1, 0.2, 0.3 m/s
[32]
Barati & Mango Cylindrical 1D Tair = 50, 60°C Empirical Nu/Pr1/3 = 0.664 × Re0.5 Empirical CMTC = (Dair /d) (2.0 + Yes
Esfahani [37] slice uair = 0.3 m/s 0.522 × Re0.5Sc0.33)
RH = 5%
Barati & Carrot Cylindrical 1D Tair = 70°C Empirical Nu/Pr1/3 = 0.664 × Re0.5 HMTA Yes
CHTC
Esfahani [38] slice uair = 2 m/s = ρair c p,air Le 0.66
CMTC
RH = 1, 10, 40%
345
(Continued )
346
TABLE 16.2 (CONTINUED)

Da Silva et al. Banana Cylindrical 1D Tair = 29.9, 68.4°C – – ORA 9.175 × 10–8 to No
[39] slice uair = 0.38, 0.39 m/s 3.492 × 10–7 m/s
RH = 7.3, 35.7%
Lemus- Papaya Rectangular 3D Tair = 40 to 80°C TDC TDC CMTC = (V/At) ln(MR) Yes
 h fg dM 
Mondaca slice uair = 1.5 m/s CHTC =  c p m ∆T −  / (Ts − Tair )
 A dt 
et al. [33] RH = 57%
Da Silva et al. Banana Cylindrical 1D Tair = 47.9, 58.6, ORA 15.2 to 28.3 W/m2K ORA 4.43 × 10–7 to Yes
[40] slice 66.9°C 4.98 × 10–7 m/s
uair = 1.5, 1.7, 1.84 m/s
RH = 6.6, 10.4, 20.3%
Ateeque et al. Potato Rectangular 3D Tair = 40 to 70°C CFD-Tu/3D Nu = 0.8185 × Re0.5427 HMTA [48] Yes
[34] slice uair = 0.5 to 4.0 m/s Nu = 0.1182 × Re0.6026
RH = 4 to 16% Nu = 0.1573 × Re0.6289
Nu = 0.1623 × Re0.6244
Esfahani et al. Apple Rectangular 2D Tair = 50oC CFD-La/2D 4 to 75 W/m2K HMTA 0.8 × 10–6 to No
[31] slice uair = 0.11, 0.22, 15 × 10–6 m/s
0.33m/s
Tzempelikos Quince Cylindrical 1D Tair = 40, 50, 60°C CFD-Tu/2D Nu/Pr1/3 = 0.472 × Re0.657 HMTA 0.0125 to 0.0966 m/s Yes
et al. [41] slice uair = 1.0, 2.0 m/s axisymmetric Nu/Pr1/3 = 0.150 × Re0.710
RH = 3, 9, 17%
Da Silva et al. Banana Cylindrical 1D Tair = 70°C – – ORA 5.03 × 10–7 to No
[42] slice uair = 0.55 m/s 7.75 × 10–7 m/s
RH = 51, 55, 56%
(Continued )
Esfahani et al. Apple Rectangular 3D Tair = 60°C Constant 2.73 to 23.07 W/m2K HMTA 5.3 × 10–7 to No
[35] slice uair = 0.1, 0.2 and 4.51 × 10–6 m/s
0.3 m/s
Vahidhosseini Quince Cylindrical 1D Tair = 40, 50, 60°C Constant 17.05 to 58.75 W/m2K HMTA 0.0125 to 0.0966 m/s Yes
et al. [43] slice uair = 1.0 and 2.0 m/s
RH = 3, 9, 17%
Da Silva Banana Cylindrical 2D Tair = 40 to 70°C – – ORA 4.2634 × 10–7 to No
Júnior et al. slice 3.1740 × 10–5 m/s
[49]
Pasban et al. Apple Rectangular 3D Tair = 60, 70, 80, 90°C TDC 10.372 to 14.982 W/m2K HMTA 0.009347 to Yes
[36] slice uair = 0.1 m/s 0.014128 m/s
RH = 6%
CFD Modeling of Convective Drying of Cylindrical Fruit Slices
Agrawal et al. Pumpkin Rectangular 1D Tair = 50, 60, 70°C Empirical Nu = 0.664 × Re0.5Pr0.3 Empirical Sh = 0.664 × Re0.5Sc0.3 Yes
[26] slice uair = 6.7, 10.15,
13.6 m/s
Defraeye & Apple Half-circular 3D Tair = 20°C CFD-Tu/3D 0 to 90 W/m2K HMTA CMTC = CHTC × 7.03 Yes
Radu [50] slice uair = 0.5 m/s × 10–9
RH = 30%
Abbreviations: CFD-La, CFD-Laminar flow; CFD-Tu, CFD-Turbulent flow; ORA, Optimization and Regression Analysis; TDC, Time-Dependent Coefficient.
347
• For the same geometry, in the two-dimensional models developed by Kaya et al. [43] and Da
Silva Júnior et al. [49], the values of the CHTC were produced by CFD modeling. The CMTC
was based on the HMTA or it was produced by optimization and regression analysis.
• For the same geometry, the three-dimensional model of Defraeye & Radu [50] incorporate
CFD values, for the CHTC. CMTC was produced by the HMTA.
• In CFD simulations of cylindrical slices, the air-drying stream was considered mainly
laminar. In the recent models of Tzempelikos et al. [41] and Defraeye & Radu [50], the
flow was considered turbulent and the SST k-ω was used to estimate the flow field over the
cylindrical slice.
16.4 ADVANTAGES AND DISADVANTAGES OF DRYING MODELS

The two main categories of models analyzed in the previous sections provide the opportunity to
simulate successively the drying process for a wide variety of fruits and drying conditions. In most
cases, the models demonstrate good agreement with the corresponding experimental data, for a
range of temperatures and velocities of the air stream. Nevertheless, both types of models are char-
acterized by significant weaknesses, so they need to be improved further in order to be used exten-
sively and systematically as powerful engineering tools, independently of the drying conditions, the
geometrical configurations and the product characteristics.
Conjugated models have the advantage of the strong coupling between heat and mass transport in
two or more domains, which can be used to represent the morphological characteristics of the prod-
ucts. The main weakness of these models remains the complexity of the formulation and the high com-
putational cost, due mainly to the different scales of convection and diffusion inside and outside the
product, and the exchange of the boundary conditions between the various computational domains.
Non-conjugated models have the advantage of the low computational cost and they considered
flexible, so that they can be easily adapted to a range of geometrical, although simplified, con-
figurations, flow conditions and products. The main weakness of these models is the incorporation
of CTCs, which are necessary for modeling the heat and mass transfer exchange, at the interface
between the drying air and the product. The values or the empirical correlations that are employed
are in most cases derived for simplified configurations, the temporal or spatial variation of the
coefficients variables are not taken into account, and when the heat and mass transfer analogy is
incorporated, it is applied under conditions that are not always met, including the absence of cou-
pling between heat and mass transfer. A comparative summary of the strengths and weaknesses of
conjugated and non-conjugated models is provided in Table 16.3.
TABLE 16.3
Strengths and Weaknesses of Conjugated and Non-Conjugated Drying Models
Strengths Weaknesses
Conjugated models • Strong coupling of heat and mass • Complex formulation
transfer between the drying air • “Case-specific”-not easily adapted
and the product to different conditions and products
• Detailed representation of product • High computational cost
geometry and morphology
Non-conjugated models • Simply formulated • CTCs-related simplifications:
• Mechanistic and flexible problematic use of heat and mass
• Easily extended to simplified transfer analogy, derived for
geometries and flow properties simplified geometrical
• Low computational cost configurations, no spatial and/or
temporal variation considered
16.5 COMPUTATIONAL MODELING OF HEAT AND MASS

TRANSFER IN CYLINDRICAL FRUIT SLICES
As an application of drying modeling, a non-conjugated model aimed at simulating the heat and
mass transfer during the convective drying of a cylindrical quince slice is discussed in the following
paragraphs.
In the model under consideration, the heat and the mass transfer equations are fully coupled,
taking into account the latent heat of vaporization of water at the surface of the slice. Fick’s law of
diffusion accounts for the diffusion inside the slice. The effective diffusion coefficient is estimated
by fitting the analytical solution of the Fick’s law to experimentally derived drying curves, on the
basis on an Arrhenius-type temperature dependence. The necessary CTCs are obtained from CFD
calculations of the turbulent flow field around the slice using a commercial CFD package. The
model is validated against experimental data, for air drying velocity of 1 m/s and 2 m/s, temperature
of 40°C, 50°C and 60°C and relative humidity 17%, 9% and 3%.
16.5.1 Model Formulation
The flow configuration along with the position and orientation of the quince slice is presented in
Figure 16.3. The air stream impacts perpendicularly the slice while the equations of heat and mass
transfer are solved in the direction of the flow. The following assumptions are made:
i. the transport of moisture through the product material is dominated by liquid diffusion
ii. the evaporation of water is occurred from the surface
iii. no phase change, i.e., evaporation, occurs inside the quince
iv. the radiation effects are neglected
v. the liquid water diffusion inside the product is assumed to follow Fick’s law
vi. the concentration gradient developed by the evaporation of water on the surface, due to the
difference of the water vapor partial pressure between the surface and the air, is the driving
force for water diffusion inside the quince slice.
FIGURE 16.3 Configuration of flow and 1D numerical model.

Considering the above assumptions and mechanisms, the one-dimensional heat and mass trans-
fer equations can be presented as follows:
∂ρbc p, pT ∂  ∂ T 
=  kp  (16.1)
∂t ∂z  ∂ z 
∂ρs X ∂  ∂X 
=  ρs Deff  (16.2)
∂t ∂z  ∂z 
where z ∈ [0,Lp] is the axial coordinate (Figure 16.3) and t is the drying time.
Constant values of temperature and moisture content of the quince slice are used for the initial
conditions. These values were also used to the experiments that were organized for the validation of
the model. The initial moisture content was measured using the AOAC [51] method and it was found
to be 81.04% (w.b) or equivalently, 4.274 kg water / kg dry (d.b). The initial product temperature (Tin)
was 17.12°C, based on Pt100 probe measurements.
The following boundary conditions are applied to account for the heat and mass transfer
exchange, at the surface of the slice. The hot drying air stream provides the proper heat energy to
the quince slice, a part of which is consumed for the evaporation of the water leaving the surface,
while the remaining is conducted through the slice.
Heat transfer boundary conditions:
∂T ∂X
front face : − k p = CHTCavg,f (T0 − Tair ) − h fg ρs Deff (16.3)
∂z z=0 ∂z z =0

∂T ∂X
back face : − k p = CHTCavg,b (Tair − TL p ) − h fg ρs Deff (16.4)
∂z z= Lp ∂z z= Lp

Mass transfer boundary conditions:
∂X
front face : − ρs Deff = CMTCavg,f (C0 − Cair ) (16.5)
∂z z=0

∂X
back face : − ρs Deff = CMTCavg,b (Cair − C L p ) (16.6)
∂z z= Lp

The vapor concentration at the surfaces is calculated from the ideal gas law as a function of the
water activity aw. Hence:
pvs (T0 )
C0 = 2.1667 × 10 −3 × aw (16.7)
T0
C L p = 2.1667 × 10 −3 × aw
( ) (16.8)
pvs TL p
TL p

Equations (16.1) and (16.2) are numerically discretized using the implicit finite volume method
while the derived algebraic equations are solved using the tridiagonal matrix algorithm. Internal
iterations are performed in each time step in order to face with the non-linearities associated with
the coupling of the heat and mass transfer equations through the evaporation term in the boundary
conditions, the dependence of thermophysical properties on moisture content and temperature and
the dependence of the slice volume and density on the moisture.
16.5.2 CFD Simulations for the Determination of CTCs

Steady-state CFD simulations are performed to estimate the values of the CTCs for the spe-
cific flow configuration. The flow and the temperature fields of the drying air stream are ana-
lyzed by solving the momentum and energy equations in the commercial CFD package Fluent®
[52].
Figure 16.4 presents a sketch of the computational domain used in the simulations. Due to
the symmetry, the chamber and the cylindrical quince slice are modeled as 2-D axis-symmetric
domains. In this way, only the half of the chamber and the cylindrical quince are modeled.
Constant velocity and temperature values are used at the inlet. At the outlet, constant pressure
and temperature values are imposed. The values of velocity, temperature and pressure used for
the boundary conditions are those measured during experiments designed to evaluate the com-
putational model.
The flow around the cylindrical quince slice is considered as a flow around a bluff body. In
order to predict the fluctuations of the flow and thermal field in the present model, the SST k-ω
model is used for turbulence modeling. This model is considered very efficient considering the
prediction of viscous and turbulent stresses distributions close to the wall and it is considered
typical for simulating external flows [34,53]; however, particular attention must be paid at the
dimensions of the cells close to the surfaces of the slice, according to the wall functions used.
More details of the computational models are given in Tzempelikos et al. [41].
FIGURE 16.4 Computational domain used in the simulations.

The local CHTC is then determined from the calculated air temperature field, using the balance
between conductive and convective heat transfer at the boundary, i.e.:
 ∂T 
− kair  = CHTC(Tps − T∞ ) (16.9)
 ∂ n  s

In Figure 16.5, the calculated spatial distributions of the Nusselt number at the front and the back
sides of the slice are shown. The values are in general higher at the front face of the cylindrical quince
slice. The magnitude of the Nusselt number is higher at the center of the slice and decreases moving
radially towards the edge. At the back face, the distribution presents an inversed trend: the maximum
value is observed at the edge of the cylindrical slice, while the minimum values appear near the center.
The weighted average CHTC, at the front and back side of the slice, which are necessary for the
one-dimensional model of heat and mass transfer inside the cylindrical quince, are calculated from
the local values according to the following equation:
1

CHTCavg =
A ∫ CHTC dA (16.10)
After the determination of the CHTCs, the CMTCs were calculated, by using the analogy between
the thermal and concentration boundary layers [46,48,54]:
 D Le1 3 
CMTC = CHTC  va (16.11)
 kair 

where Le is the Lewis number representing a measure of the relative thermal and concentration
boundary layer thickness:
aair
Le = (16.12)
Dva
350
Front side
300 Back side
250
200
Nu
150
100
50
0
0.00 0.01 0.02 0.03 0.04
rp (m)
FIGURE 16.5 Spatial distribution of the Nusselt number produced by CFD simulations.
200
0.657 1/3
Nu=0.472xRe Pr
100
90
80
70
60 Nu=0.15xRe
0.71 1/3
Pr
Nu
50
40
Back face
30
20
Front face Front face
Air flow Back face
10
2 3 4 5 6 7 8 9 10 11 12
-3
Re x10
FIGURE 16.6 Correlations for the Nu number at the front and the back side of the cylindrical slice.
The weighted average CMTC at the front and back side of the slice, which are necessary for the
one-dimensional model of heat and mass transfer inside the quince, are calculated from the local
values according to the following equation:
1

CMTCavg =
A ∫ CMTC dA (16.13)
Nusselt number is typically expressed as a function of Prandtl and Reynolds numbers of
the flow around the cylindrical slice. Figure 16.6 shows the corresponding distributions of the
Nusselt number at the front and the back side of the slice, for the conditions examined in the
simulations.
16.5.3 Thermophysical Properties
In the model, the water activity aw was calculated using the formula proposed by Chen [55], where
k1, k2 and k3 are experimentally estimated constants [56]:
aw = exp  k1 − k2 exp(− k3 M )  (16.14)

The second term on the right-hand side of Equations (16.3) and (16.4) accounts for the heat of
evaporation of the water leaving the surface, where hfg (J/kg) is the water latent heat of evaporation,
given by the following equation [57]:
0.3298
 647.3 − T 
h fg = 2501.05 × 10 ×  3
(16.15)
 647.3 − 273.15 

The concentration of water vapor in air (Cair) is derived from the ideal gas law, according to the
following equation:
RH pvs (Tair )
Cair = 2.1667 × 10 −3 × (16.16)
100 Tair
where pvs (Pa) is the water vapor pressure, given by [57]:
 −5.8 × 10 3 
pvs = exp  + 1.391 − 4.864 × 10 −2 T + 4.176 × 10 −5 T 2 − 1.445 × 10 −8 T 3 + 6.545 ln(T ) 
 T 

(16.17)

As already noticed, the values of the effective moisture diffusivity (Deff) are necessary corre-
lated with available experimental data. A typical method used extensively for the correlation is the
so-called slope-method in combination with an Arrhenius type dependence on the temperature and/
or moisture content in the product. Details of the applications of method can be found for example in
Tzempelikos et al. [41]. In this work, a temperature dependence of the effective diffusion coefficient
was determined from the experimental data, following an Arrhenius type equation:
 4431.71 
Deff = 5.492 × 10 −4 × exp  −  (16.18)
 T 
The thermal conductivity (kp) and the specific heat (cp,p) were estimated as functions of the mois-
ture content [58]:
k p ( M ) = 0.148 + 0.493 × M (16.19)
c p, p ( M ) = (1.26 + 2.97 × M ) × 1000 (16.20)
The ratio of the dry material mass to the total volume of the slice is equal to the density of the
solid material ρs. For the determination of the variation of the volume of the slice, a simple mecha-
nistic approach was employed, based on the assumption that the volume shrinkage at any time step
can be approximated using the removed water volume in that time step (negligible porosity and air
mass in the cylindrical quince slice).
The dry mass was directly derived from the known initial moisture content and the total slice
mass, while the removed water volume was calculated from the moisture change during a time step.
Both these variables were treated as inputs of the model.
In order to include the shrinkage effect in the model, the thickness of the cylindrical slice was
estimated as the average of the initial and the final thickness of the product. The initial thickness
of the cylindrical slice was known, while the final thickness can be approximated a priori, based
on the assumption that the dimensions of the cylindrical slice varied only in the axial direction (z),
i.e., only the thickness of the slice changes during the drying process, while the diameter remains
constant.
The final thickness of the slice was calculated from the ratio of the final slice volume, which is
approximated to be the initial volume minus the volume of the initial water content, over the area of
the cylindrical quince slice. The latter was assumed constant during drying.
The model has been validated by experimental data, for air drying velocities 1 m/s and 2 m/s
and temperatures from 40°C, 50°C and 60°C [41]. Experiments had been performed by incor-
porating an in-house developed convective batch dryer (Figure 16.7) with a test section of
dimensions 0.5 m × 0.5 m × 0.66 m. A heat exchanger connected to a boiler-hot water system
was used to adjust the temperature of the drying air. The air stream was recirculating inside
the whole device via a centrifugal fan with speed adjustment. The facility was equipped with
temperature, velocity sensors and relative humidity sensors, in order to monitor the conditions
inside the drying chamber and the temperature of the product. A system of four load cells
was used for monitoring the weight loss of the quince slide. A data acquisition system was
developed in order to store the signals of all the above sensors, and present on-line the tem-
perature and relative humidity measurements during experiments. A detailed description of the
Laboratory Thermal Convective Dryer (LTCD) unit is presented in Tzempelikos et al. [59] and
Tzempelikos et al. [60].
To achieve a satisfactory agreement with the measurements, the effective diffusion coefficient
(Deff) given by equation (16.18) should be corrected by a factor equal to 0.45 to provide satisfactory
agreement for all the cases examined. This correction is considered necessary, due to the different
level of approximation inherent in the approach used for the experimental derivation of the effective
diffusion coefficient, in comparison with the developed numerical model. In the diffusion-based
model it is assumed that the effective diffusion coefficient describes the combined effects of air and
liquid transport inside the drying material. For that reason, the value of the diffusion coefficient
cannot be derived by the mass transport theory; instead, it can only be derived by a semi-empirical
method, which is based on the fitting of the analytical solution of the Fick’s diffusion equation to
the experimental drying curves. Nevertheless, the analytical solution used for the post-experimental
FIGURE 16.7 Isometric drawing of LTCD.

processing is a result of a series of approximations (e.g., isothermal conditions, negligible surface

resistance, constant properties), attributing “artificially” the effect of a number of neglected param-
eters to the diffusion coefficient itself. These approximations are not engaged in the developed
numerical model, and it is thus reasonable that the effective diffusion coefficient derived by fitting
the analytical solution to the measured data needs additional matching in order to account for more
detailed models.
In Figures 16.8 and 16.9, the spatially-averaged dimensionless moisture content and the dimen-
sionless temperature at the center of the cylindrical quince slice are plotted as functions of drying
time, in comparison with the experimental data. Apart from the uncertainties related to the effective
diffusion coefficient and the CTCs, the deviations between measured and simulated data can be
explained by additional parameters, such as the change of the shape geometry of the slice that can
produce additional effects, influencing mainly the values of the CTCs.
The model can be furthermore used to predict the spatial and temporal variation of the
moisture content inside the quince slice, as shown in Figure 16.10. At all the time steps, a
moisture profile is formed in the interior part of the product, while near the surface boundary,
the moisture content drops sharply and creates significant gradients, as the time of the experi-
ments increases. This trend has been also observed by Aregawi et al. [61] and indicates that the
internal water towards to the surface diffuses slower due to the high internal mass resistance,
resulting with that way to a considerably slower progress of the water removal [32,34]. In addi-
tion, the symmetrical profile of the moisture content inside the slice indicates the weak effect
of the different CTCs used at the front and the back side of the slice.
Figure 16.11 illustrates the predicted spatial, dimensionless distribution of the temperature
inside the product for a short period (up to 2000 s) after the beginning of drying. The tem-
perature profile shows small gradients along the quince slice, which disappear after about
25 minutes.
1.0
0.9 u = 2 m/s
Tair= 40 C Num.
o
0.8
Tair= 50 C Num.
o
0.7
Tair= 60 C Num.
o
0.6
Tair= 40 C Exp.
o
0.5
X*
Tair= 50 C Exp.
o
0.4 Tair= 60 C Exp.

o
0.3
0.2
0.1
0.0
0 1 2 3 4 5 6 7
-4
t (s) x 10
FIGURE 16.8 Calculated temporal variation of the dimensionless moisture content compared to
measurements.
1.0
0.9
o
Tair = 60 C, u = 2 m/s
0.8
Numerical model
T*
0.7 Experiments
0.6
0.5
0.4
1.0
0.9
o
Tair = 50 C, u = 2 m/s
0.8
Numerical model
T*
0.7 Experiments
0.6
0.5
0.4
1.0
0.9
o
0.8 Tair = 40 C, u = 2 m/s
Numerical model
T*
0.7 Experiments
0.6
0.5
0.4
0 1 2 3 4 5 6 7
-4
t (s) x 10
FIGURE 16.9 Calculated temporal variation of the dimensionless product temperature product compared
to measurements.
Figure 16.12(a) demonstrates the time evolution of the energy balance during drying in terms of
the convective energy, i.e., the energy provided from the drying air to the cylindrical quince slice,
the latent heat of evaporation, i.e., the energy consumed for the evaporation of the water at the
surface of the quince slice and the conductive energy, i.e., the energy conducted in the interior of
the slice. The above quantities correspond to the three terms in the boundary conditions of the heat
equation (Equations 16.3 and 16.4).
The figure indicates that the energy consumed for heating the slice is a small portion of the total
energy provided by the air stream. At the first, short period of drying, even this small portion of
energy is adequate to increase substantially the temperature of the slice. During the drying process,
the energy of the air stream is practically consumed for the water evaporation.
1.0
1/2 h
0.9
1h
0.8
0.7
2h
0.6
X*
0.5
4h
0.4
0.3
8h
0.2
0.1
16 h
0.0
0.00 0.25 0.50 0.75 1.00
z/L
FIGURE 16.10 Temporal and spatial variation of the dimensionless moisture profiles inside the cylindrical
quince slice.
0.9
2000 s
1200 s
0.8
800 s
600 s
0.7
400 s
*
T
300 s
0.6
200 s
0.5
100 s
0.4
0.00 0.25 0.50 0.75 1.00
z/L
FIGURE 16.11 Temporal and spatial variation of the dimensionless temperature profiles inside the cylindri-
cal quince slice.
40
Convection
30 Water evaporation
20 Conduction
Heat (kW)
10
0
-10
-20
-30
(a)
-40
0 1 2 3 4 5 6 7
1.0
0.8
0.6
T*
0.4
0.2 (b)
0.0
0 1 2 3 4 5 6 7
4
t (s) x 10
FIGURE 16.12 Predicted temporal variation of (a) the energy exchange at the surface of the slice and (b) the
dimensionless temperature of the slice
The above statements are furthermore supported by the temporal variation profile of the tem-
perature of the slice, shown in Figure 16.12(b). The figure indicates that the temperature of the slice
increases dramatically at the initial phase of the drying process, while it continues to increase but
with a significantly lower rate, as the time of experiment increases.
16.6 CONCLUSIONS
Drying of fruits and agricultural products is very important in food engineering and processing, and
in the perfume and healthcare industries. Among the various methods, convective drying is consid-
ered one of the most popular, due to the simplicity of the involved mechanisms and the low cost of
the development of the required infrastructure. The process is governed by the strong coupling and
the exchange of energy and mass between the drying stream and the product.
The chapter was focused on the so-called diffusion-based models. In these models, Fick’s diffu-
sion is assumed to approximate the water transport inside the product. Diffusion-based models are
further distinguished in two main categories: conjugated and non-conjugated. In conjugated mod-
els, the fully coupled unsteady transport equations for both the drying air and the product are solved
in parallel in two or more computational domains, while in non-conjugated models, the calculations
of the heat and mass transfer inside the product are performed independently from the velocity,
temperature and concentration fields of the drying.
Simulating with high accuracy the drying rate of fruits is crucial to meet the requirements of
dried products and utilize optimized conditions that fit the morphological characteristics and the
thermophysical properties of the products. On the other hand, computational modeling is a very
important tool, as it contributes to the better understanding of heat and mass transport, and, when
simply formulated, it is also a very useful tool in the improvement of the design and the control
of the process. As a paradigm of application, the performance of a non-conjugated model for the
convective drying of cylindrical quince slices is reviewed in this chapter. All the models analyzed
in this chapter are characterized by weaknesses, associated with the complexity of the formula-
tion or the level of approximations. Further development, refinement and validation of the mod-
els is needed to achieve a more satisfactory and consistent treatment of the drying process. CFD
can play an important role since can provide the computational framework, wherein the coupled
heat and mass transport involved in drying can be solved, in both conjugated and non-conjugated
models.
NOMENCLATURE
A Total area of product (m2)
aw Water activity
Bi Biot number
C Concentration of water vapor (kg m–3)
cp Specific heat (J kg–1 K–1)
D Product diameter (m)
Deff Effective moisture diffusivity (m2 s–1)
Dair Diffusivity of air (m2 s–1)
Dva Diffusivity of water vapor in air (m2 s–1)
DR Drying rate (kg water kg–1 dry products)
hfg Water latent heat of evaporation (J kg–1)
L Thickness (m)
Le Lewis number
M Moisture content in wet basis (kg water kg–1 product)
m Mass rate (kg s–1)
MR Moisture ratio
n Unit normal vector at the surface
Nu Nusselt number
p Pressure (Pa)
pvs Saturated water vapor pressure at surface (Pa)
Pr Prandtl number
r Radial coordinate (m)
Re Reynolds number
RH Relative humidity of air (%)
Sh Sherwood number
Sc Schmidt number
t Drying time (s)
T Absolute temperature (K)
T* Dimensionless temperature (Tp/Tref )
u Velocity (m s–1)
V Volume (m3)
X Moisture content in dry basis (kg water kg–1 dry product)
X* Dimensionless moisture content (Xp/Xin)
z Axial coordinate (m)
Greek symbols
a Thermal diffusivity (m2 s–1)
ρb Bulk density of product = total mass / total volume (kg m–3)
ρs Density of solid product = dry mass / total volume (kg m–3)
Subscripts
air Drying air
p Product
ps Product surface
ref Reference
s Surface
∞ Bulk
Acronyms
CFD-Tu CFD-turbulent flow
CFD-La CFD-laminar flow
CHTC Convective heat transfer coefficient (W m–2 K–1)
CHTCavf Weighted average convective heat transfer coefficient (W m–2 K–1)
CMTC Convective mass transfer coefficient (m s–1)
CMTCavg Weighted average convective mass transfer coefficient (m s–1)
CTC Convective transfer coefficient
HMT Heat and mass transfer
HMTA Heat and mass transfer analogy
LTCD Laboratory thermal convective dryer
ORA Optimization and regression analysis
TDC Time-dependent coefficient
UDF User defined function
REFERENCES
1. Datta A.K., Porous media approaches to studying simultaneous heat and mass transfer in food pro-
cesses, I: Problem formulations, Journal of Food Engineering, 80(1): 80–95, 2007a.
2. Datta A.K., Porous media approaches to studying simultaneous heat and mass transfer in food pro-
cesses, II: Property data and representative results, Journal of Food Engineering, 80(1): 96–110,
2007b.
3. Defraeye T., Advanced computational modelling for drying processes—A review, Applied Energy, 131:
323–344, 2014.
4. Lemus-Mondaca R.A., Vega-Gálvez A., Moraga N.O., Computational simulation and developments
applied to food thermal processing, Food Engineering Reviews, 3(3–4): 121–135, 2011.
5. Defraeye T., Blocken B., Derome D., Nicolaï B.M., Carmeliet J., Nicolai B., Carmeliet J., Convective
heat and mass transfer modelling at air–porous material interfaces: Overview of existing methods and
relevance, Chemical Engineering Science, 74: 49–58, 2012.
6. Curcio S., Aversa M., Chakraborty S., Calabrò V., Iorio G., Formulation of a 3D conjugated multiphase
transport model to predict drying process behavior of irregular-shaped vegetables, Journal of Food
Engineering, 176: 36–55, 2016.
7. Caccavale P., De Bonis M.V., Ruocco G., Conjugate heat and mass transfer in drying: A modeling
review, Journal of Food Engineering, 176: 28–35, 2016.
8. Khan F.A., Straatman A.G., A conjugate fluid-porous approach to convective heat and mass transfer
with application to produce drying, Journal of Food Engineering, 179: 55–67, 2016.
9. Lemus-Mondaca R.A., Vega-Gálvez A., Zambra C.E., Moraga N.O., Modeling 3D conjugate heat
and mass transfer for turbulent air drying of Chilean papaya in a direct contact dryer, Heat and Mass
Transfer, 1–14, 2016.
10. Cârlescu P.-M., Arsenoaia V., Roşca R., ţenu I., CFD simulation of heat and mass transfer during apri-
cots drying, LWT–Food Science and Technology, 85, 2017.
11. Defraeye T., Radu A., Convective drying of fruit: A deeper look at the air-material interface by conju-
gate modeling, International Journal of Heat and Mass Transfer, 108: 1610–1622, 2017a.
12. Defraeye T., Verboven P., Derome D., Carmeliet J., Nicolaï B.M., Stomatal transpiration and droplet
evaporation on leaf surfaces by a microscale modelling approach, International Journal of Heat and
Mass Transfer, 65: 180–191, 2013.
13. Defraeye T., Verboven P., Moisture barriers to control drying of fresh-cut fruit: Quantifying their impact
by modeling, Food and Bioproducts Processing, 101: 205–213, 2017.
14. Manickathan L., Defraeye T., Allegrini J., Derome D., Carmeliet J., Parametric study of the influence of
environmental factors and tree properties on the transpirative cooling effect of trees, Agricultural and
Forest Meteorology, 248: 259–274, 2018.
15. Curcio S., Aversa M., Calabrò V., Iorio G., Simulation of food drying: FEM analysis and experimental
validation, Journal of Food Engineering, 87(4): 541–553, 2008.
16. De Bonis M.V., Ruocco G., A generalized conjugate model for forced convection drying based on an
evaporative kinetics, Journal of Food Engineering, 89(2): 232–240, 2008.
17. Lamnatou C., Papanicolaou E., Belessiotis V.G., Kyriakis N., Conjugate heat and mass transfer from
a drying rectangular cylinder in confined air flow, Numerical Heat Transfer, Part A, 56(5): 379–405,
2009.
18. Lamnatou C., Papanicolaou E., Belessiotis V.G., Kyriakis N., Finite-volume modelling of heat and mass
transfer during convective drying of porous bodies—Non-conjugate and conjugate formulations involv-
ing the aerodynamic effects, Renewable Energy, 35(7): 1391–1402, 2010a.
19. Lamnatou C., Papanicolaou E., Belessiotis V.G., Kyriakis N., Numerical study of the interaction among
a pair of blunt plates subject to convective drying—A conjugate approach, International Journal of
Thermal Sciences, 49(12): 2467–2482, 2010b.
20. Marra F., De Bonis M.V., Ruocco G., Combined microwaves and convection heating: A conjugate
approach, Journal of Food Engineering, 97(1): 31–39, 2010.
21. Curcio S., A multiphase model to analyze transport phenomena in food drying processes, Drying
Technology, 28(6): 773–785, 2010.
22. Defraeye T., Herremans E., Verboven P., Carmeliet J., Nicolaï B.M., Convective heat and mass exchange
at surfaces of horticultural products: A microscale CFD modelling approach, Agricultural and Forest
Meteorology, 162–163: 71–84, 2012.
23. Sabarez H.T., Computational modelling of the transport phenomena occurring during convective drying
of prunes, Journal of Food Engineering, 111(2): 279–288, 2012.
24. Defraeye T., Verboven P., Nicolaï B.M., CFD modelling of flow and scalar exchange of spherical food
products: Turbulence and boundary-layer modelling, Journal of Food Engineering, 114(4), 495–504,
2013.
25. Kurnia J.C., Sasmito A.P., Tong W., Mujumdar A.S., Energy-efficient thermal drying using impinging-
jets with time-varying heat input—A computational study, Journal of Food Engineering, 114(2): 269–
277, 2013.
26. Agrawal S.G., Methekar R.N., Mathematical model for heat and mass transfer during convective drying
of pumpkin, Food and Bioproducts Processing, 101: 68–73, 2017.
27. Aversa M., Curcio S., Calabrò V., Iorio G., An analysis of the transport phenomena occurring during
food drying process, Journal of Food Engineering, 78(3): 922–932, 2007.
28. Kaya A., Aydin O., Dincer I., Experimental and numerical investigation of heat and mass transfer dur-
ing drying of Hayward kiwi fruits (Actinidia Deliciosa Planch), Journal of Food Engineering, 88(3):
323–330, 2008.
29. Oztop H.F., Akpinar E.K., Numerical and experimental analysis of moisture transfer for convective
drying of some products, International Communications in Heat and Mass Transfer, 35(2): 169–177,
2008.
30. Villa-Corrales L., Flores-Prieto J.J., Xamán-Villaseñor J.P., García-Hernández E., Numerical and
experimental analysis of heat and moisture transfer during drying of Ataulfo mango, Journal of Food
Engineering, 98(2): 198–206, 2010.
31. Esfahani J.A., Majdi H., Barati E., Analytical two-dimensional analysis of the transport phenomena
occurring during convective drying: Apple slices, Journal of Food Engineering, 123: 87–93, 2014.
32. Chandra Mohan V.P., Talukdar P., Three dimensional numerical modeling of simultaneous heat and
moisture transfer in a moist object subjected to convective drying, International Journal of Heat and
Mass Transfer, 53(21–22): 4638–4650, 2010.
33. Lemus-Mondaca R.A., Zambra C.E., Vega-Gálvez A., Moraga N.O., Coupled 3D heat and mass transfer
model for numerical analysis of drying process in papaya slices, Journal of Food Engineering, 116(1):
109–117, 2013.
34. Ateeque M., Udayraj, Mishra R.K., Chandramohan V.P., Talukdar P., Numerical modeling of convec-
tive drying of food with spatially dependent transfer coefficient in a turbulent flow field, International
Journal of Thermal Sciences, 78: 145–157, 2014.
35. Esfahani J.A., Vahidhosseini S.M., Barati E., Three-dimensional analytical solution for transport prob-
lem during convection drying using Green’s function method (GFM), Applied Thermal Engineering, 85:
264–277, 2015.
36. Pasban A., Sadrnia H., Mohebbi M., Shahidi S.A., Spectral method for simulating 3D heat and mass
transfer during drying of apple slices, Journal of Food Engineering, 212: 201–212, 2017.
37. Barati E., Esfahani J.A., A new solution approach for simultaneous heat and mass transfer during con-
vective drying of mango, Journal of Food Engineering, 102(4): 302–309, 2011.
38. Barati E., Esfahani J.A., Mathematical simulation of convective drying: Spatially distributed tempera-
ture and moisture in carrot slab, International Journal of Thermal Sciences, 56: 86–94, 2012.
39. Da Silva W.P., e Silva C.M.D.P.S., Farias V.S.O., Gomes J.P., Diffusion models to describe the drying
process of peeled bananas: Optimization and simulation, Drying Technology, 30(2): 164–174, 2012.
40. Da Silva W.P., e Silva C.M.D.P.S., Gama F.J.A., Estimation of thermophysical properties of products with
cylindrical shape during drying: The coupling between mass and heat, Journal of Food Engineering,
141: 65–73, 2014.
41. Tzempelikos D.A., Mitrakos D., Vouros A.P., Bardakas A.V., Filios A.E., Margaris D.P., Numerical
modelling of heat and mass transfer during convective drying of cylindrical quince slices, Journal of
42. Da Silva W.P., Rodrigues A.F., e Silva C.M.D.P.S., de Castro D.S., Gomes J.P., Comparison between
continuous and intermittent drying of whole bananas using empirical and diffusion models to describe
the processes, Journal of Food Engineering, 166: 230–236, 2015.
43. Vahidhosseini S.M., Barati E., Esfahani J.A., Green’s function method (GFM) and mathematical solu-
tion for coupled equations of transport problem during convective drying, Journal of Food Engineering,
187: 24–36, 2016.
44. Kaya A., Aydin O., Dincer I., Numerical modeling of forced-convection drying of cylindrical moist
objects, Numerical Heat Transfer, Part A, 51(9): 843–854, 2007.
45. Perry R.H., Green D.W., Maloney J.O., Perry’s Chemical Engineers Handbook, 7th ed., McGraw Hill,
46. Bird R.B., Stewart W.E., Lightfoot E.N., Transport Phenomena, John Wiley & Sons, Hoboken, NJ,
USA, 1960.
47. Janjai S., Lamlert N., Intawee P., Mahayothee B., Haewsungcharern M., Bala B.K.K., Müller J., Finite
element simulation of drying of mango, Biosystems Engineering, 99(4): 523–531, 2008.
48. Incropera F.P., DeWitt D.P., Bergman T.L., Lavine A.S., Principles of Heat and Mass Transfer, 7th ed.,
John Wiley & Sons, Hoboken, NJ, USA, 2012.
49. Da Silva Júnior A.F., Da Silva W.P., de Farias Aires J.E., Farias Aires K.L.C.A., Numerical approach
to describe complementary drying of banana slices osmotically dehydrated, Heat and Mass Transfer/
Waerme- und Stoffuebertragung, 1–10, 2017.
50. Defraeye T., Radu A., Insights in convective drying of fruit by coupled modeling of fruit drying, defor-
mation, quality evolution and convective exchange with the airflow, Applied Thermal Engineering,
2017b.
51. AOAC, Official Methods of Analysis, 15th ed., Association of Official Analytical Chemists, Arlington,
VA, 1990.
52. Fluent Inc., FLUENT 6.3.26 User’s Guide, Lebanon, NH, USA, 2006.
53. Defraeye T., Blocken B., Carmeliet J., CFD analysis of convective heat transfer at the surfaces of a cube
immersed in a turbulent boundary layer, International Journal of Heat and Mass Transfer, 53(1–3):
297–308, 2010.
54. Chilton T.H., Colburn A.P., Mass transfer (absorption) coefficients prediction from data on heat transfer
and fluid friction, Industrial & Engineering Chemistry, 26(11): 1183–1187, 1934.
55. Chen C.S., Equilibrium moisture curves for biological materials, Transactions of the ASABE, 14(5):
924–926, 1971.
56. Noshad M., Shahidi F., Mohebbi M., Ali Mortazavi S., Mortazavi S.A., Desorption isotherms and ther-
modynamic properties of fresh and osmotic-ultrasonic dehydrated quinces, Journal of Food Processing
and Preservation, 37(5): 381–390, 2012.
57. ASHRAE, ASHRAE Handbook-Fundamentals, American Society of Heating, Refrigiration and Air-
Conditioning Engineers Inc., Atlanta, 2009.
58. Rao A., Rizvi S.S.H., Datta A.K., Ahmed J., Engineering Properties of Foods, 4th ed., CRC Press,
Boca Raton, FL, USA, 2014.
59. Tzempelikos D.A., Vouros A.P., Bardakas A.V., Filios A.E., Margaris D.P., Analysis of air velocity
distribution in a laboratory batch-type tray air dryer by computational fluid dynamics, International
Journal of Mathematics and Computers in Simulation, 6(5): 413–421, 2012.
60. Tzempelikos D.A., Vouros A.P., Bardakas A.V., Filios A.E., Margaris D.P., Design, construction and
evaluation of a new laboratory convective dryer using CFD, International Journal of Mechanics, 7(4):
425–434, 2013.
61. Aregawi W., Defraeye T., Saneinejad S., Vontobel P., Lehmann E., Carmeliet J., Verboven P., Derome D.,
Nicolaï B.M., Understanding forced convective drying of apple tissue: Combining neutron radiography
and numerical modelling, Innovative Food Science & Emerging Technologies, 24: 97–105, 2014.
17 CFD Modeling of Convection
Flow in Pan Cooking
Yvan Llave and Noboru Sakai
CONTENTS
17.1 Introduction............................................................................................................................ 365
17.2 Cooking Appliances................................................................................................................ 366
17.2.1 Gas-Range Heating (GRH)...................................................................................... 366
17.2.2 Induction Heating (IH)............................................................................................... 367
17.2.3 Electrical Resistance Stoves....................................................................................... 368
17.3 Pan Cooking Materials...............................................................................................................368
17.4 Visualization Approaches of Convection Flow for Different Cooking Appliances............ 371
17.5 Visualization Approaches of Convection Flow for Different Pan Materials....................... 373
17.6 Mathematical Modeling by CFD of Convection Flow in Pan Cooking............................... 373
17.6.1 Basic Model Equations and Solution Procedure..................................................... 374
17.6.1.1 Physical Model........................................................................................ 374
17.6.1.2 Assumptions Used in the Numerical Simulation..................................... 374
17.6.1.3 Governing Equations............................................................................... 374
17.6.1.4 Boundary Conditions................................................................................ 375
17.6.1.5 Analysis of Electromagnetic Field................................................................. 376
17.6.1.6 Solution Strategy....................................................................................... 377
17.6.1.7 Computational Grid................................................................................. 378
17.6.2 CFD Applications in Convection Flow in Pans Using IH...................................... 378
17.6.3 CFD Applications in Convection Flow in Pans Using GRH.................................. 380
17.6.4 Additional CFD Applications in Convection Flow..................................................... 383
17.7 Approaches for Continuous Cooking Systems Optimization.............................................. 383
17.8 Suggestions for Future Work................................................................................................... 385
17.9 Conclusions.......................................................................................................................... 385
References....................................................................................................................................... 386
17.1 INTRODUCTION
For the selection of cooking appliances by consumers, several factors should be taken in consid-
eration, such as ease of use, quick response rate, low maintenance, consistency, energy efficiency,
ease of cleaning, and heating capacity; all these factors are relevant in combination with aesthetics.
Significantly more consideration is required now for the selection of cookware type (pan cooking)
to purchase, compared to previous years.
Domestic cooking appliances have evolved significantly over the last few decades, not only those
based on the traditional technology of gas burners, but also the newest cooking hobs with induc-
tion heating (IH) elements. The main improvements that have been introduced are related to the
efficiency, total heating power, and number and configuration of burners, inductors, hotplates, etc.
[1]. In the case of the cooking of liquid foods, the difference in the heating location may affect the
convection in the pan. It has been reported that in IH, only the bottom of the pan is heated because
the region on the coil generates heat by electromagnetic induction; in contrast, in natural gas-range
365
heating (GRH), the bottom and sides of the pan are heated by the flame [2]. In addition, differences
in the energy efficiency and temperature distribution during heating and cooling are related to the
configuration of the pan and the different materials comprising them.
Therefore, the convection flow of liquid foods depends on two main factors: (a) the cooking
appliance and (b) the type of pan cooking used. The natural convection of the different fluids may
affect the microbiology, safety, food quality, and cooking times of the cooking process. Therefore,
an appropriate analysis of the natural convection phenomena in fluid-filled pans is required to evalu-
ate the effect of both factors on temperature distribution, flow velocities, and power distribution
(especially in the bottom of the pan). Understanding the thermal natural convection by simulation
might be necessary to improve the cooking appliances (with better designs and functionalities) and
the materials of the pan used as cookware. To study the performance of these cooking appliances
and cookware, the manufacturers conduct various tests of their products, including cooking tests
with real foods, emulating a normal operation at the user’s home. However, these tests require time
and money that could be saved by appropriate modeling and simulation of the cooking processes. In
recent years, computational fluid dynamics (CFD) has been increasingly applied to food processing
operations, coupled with experimental techniques of flow visualization, and temperature measure-
ments for these purposes [3]. CFD is a simulation tool that uses powerful computers in combination
with applied mathematics to model fluid flow situations and aid the optimal design of industrial
processes [4]. The method comprises solving equations for the conservation of mass, momentum,
and energy; and using numerical methods to predict the velocity, temperature, and pressure pro-
files inside the system. This chapter summarizes the recent developments in this research area and
reviews the applications of CFD as a tool for predicting the convection flow pattern and particle
histories such as temperature and velocity in pan cooking.
17.2 COOKING APPLIANCES
Some authors in the current literature have studied the behavior of cooking appliances using theoreti-
cal models or experimentation. Different technologies of stoves (gas, electric hotplate, induction) were
studied experimentally to determine the performance in terms of heating speed and temperature distri-
bution in the pan frying process [5]. In one study [6], the thermal efficiency of pots in a particular type
of electric stove was calculated and measured. The heat transfer characteristics of different sauce pans
on electric coil, natural gas, and induction cooktops were analyzed and the most appropriate pan for
each cooking appliance was identified [7]. They found that the cooking efficiency was approximately
28%, 39.3% and 70%, respectively, for natural gas, electric coil, and induction cooking appliances.
Further authors have developed different theoretical models of various cooking techniques; e.g.,
a thermal model of the elaboration of frozen hamburger patties was proposed [8], and a model of the
contact baking of thick pancakes was presented [9]. These models combine heat and mass transfer
phenomena and include the food state variation as a result of the temperature rise. They are centered
in the behavior of the food and assume that the heat source is well known and controlled, and that
the temperature distribution in the pan is uniform. However, this assumption is vastly different from
a real cooking situation in a domestic environment, where the power distribution that determines
the heating uniformity depends on the type of technology, i.e., cooker design or pan cooking. To
develop theoretical models that more closely approximate real scenarios and analyze the effects of
non-uniform heating on the elaboration of the food, it is of considerable interest to study the power
distribution and convection flow in pan cooking as a function of the type of cooking appliance. The
characteristics of the most common cooking appliances are presented as follows.
17.2.1 Gas-Range Heating (GRH)

In GRH, heat is generated in a flame or in the heating element, which is then applied to the pan. In
a gas cooker (Figure 17.1A), there is a heat exchange between the gas produced in the combustion
CFD Modeling of Convection Flow in Pan Cooking 367
FIGURE 17.1 Schemes of the thermal model in a (A) gas cooker, (B) induction hob, and (C) electric stove.
(Adapted from Sanz-Serrano, F., Sagues, C., and Llorente, S., Appl. Therm. Eng., 92, 2016.)
process and the bottom surface of the pan. In one study [10], the relationship between the configu-
ration of the burner, the flame, and the power distribution generated in a gas burner cooker was
analyzed by CFD simulation. Another study [1] reported an inverse thermal model that allows the
calculation of the power distribution generated in the bottom of the pan from the measurement of
the surface temperature. This method was used to establish comparisons between the different
power distributions and to achieve a deeper understanding of the behavior of the pan temperature
during the cooking process in each situation with the three considered technologies (induction,
electric resistance, and gas). It was reported that the gas cooker is suitable for a wider range of pan
sizes because the power is more concentrated in the outer area of the pan base. Similar results for
the convection flow in pan cooking were reported by [2], concluding that, in addition to the bottom
of the pan, the side is also heated by the flame.
17.2.2 Induction Heating (IH)

IH is a non-contact heating process (Figure 17.1B). The IH principle in an induction stove works
owing to the alternating current flow generated in the coil, which produces an alternating magnetic
field that is induced on the ferromagnetic bottom of the cookware [2,11]. It uses high-frequency
electricity to heat materials that are electrically conductive. In an induction cooker, a coil of copper
wire is placed under the cooking pot and an alternating electric current is passed through it. The
resulting oscillating magnetic field induces a magnetic flux that repeatedly magnetizes the pan,
treating it like a loss magnetic core of a transformer [12]. This produces large Eddy currents (loops
of electrical current induced within conductors by a changing magnetic field in the conductor, due
to Faraday’s law of induction) in pan cooking. Because of the resistance of the pan, these Eddy cur-
rents heat the pan by the Joule effect. Since this process has lower energy losses and the material
magnetically agitates, the absorbed energy releases as energy heating the pan. For these reasons, IH
lends itself to some unique applications in industry.
The increase in the demand for induction stoves at a household level is because they have several
advantages in comparison to conventional liquefied petroleum gas (LPG)-based stoves and electric
coil stoves [2,11]. However, the appropriate performance of these devices for IH depends directly
on the coupled system cooking appliance–pan cooking. The most important advantages in the use
of IH are: i) energy efficiency as a result of the reduction in heating losses in households due to the
induced magnetic field during pan cooking; ii) enhanced safety as there is no flame during cooking;
iii) no waste of energy when the pan is removed from the hob; iv) automatic cut-off function in case
of overheating; v) no emission of harmful fuel gases; and vi) easy to clean [2,11,13].
17.2.3 Electrical Resistance Stoves

In electrical resistance stoves (Figure 17.1C), the power is transferred by radiation from a hot electric
resistance to a ceramic glass, which is heated to temperatures in the region of 400°C, and transfers
the heat by conduction to the pan placed on the glass. The heating source in the electric stove gener-
ates a power distribution in the glass. Hence, the heat flux in the vertical direction flows from the
glass to the pan. In one study [1], the heating uniformity of the cooking vessels in domestic cooking
appliances (induction, electric resistance, and gas burner) was evaluated by the inverse modeling
method. Among other results, they reported that the electric stove showed the worst performance
in terms of heating uniformity, which could be improved by increasing the power supplied by the
outer resistance in the double cooker.
17.3 PAN COOKING MATERIALS

The determination of cooking efficiency is challenging owing to variations in pans (size and com-
position) and cooking appliances (electrical coil, natural gas, induction; size of the burner, power/
wattage), as well as methods for determining efficiency (water boiling test, wet brick test, alumi-
num and anodized aluminum block test, and carbon steel block test) and the reporting thereof [14].
The composition, size, and shape of the pan also influence the energy consumption. A pan, simply
stated, is the physical delivery system for transferring heat from the source such as, electrical coil,
natural gas, and induction, to the contact surface of the food to be cooked. Conduction of thermal
energy via the food contact surface determines the cooking behavior. The conduction of heat is a
combination of heat transfer through the pan materials and heat transfer from the heat source to the
base of the pan. Therefore, the construction materials of the pan largely determine the manner in
which the thermal energy is transferred from the pan to the food [7].
Material selection has great importance in the design and development of cooking pan materi-
als; the selection of an appropriate material is necessary for improve the efficiency of the system
cooking appliance (pan cooking). Differences in the energy efficiency and temperature distribution
during heating and cooling are related to the configuration of the pan and the different materi-
als thereof. For cooking processes, the natural convection of the different fluids may affect to the
microbiology, safety, food quality, and cooking time. Figure 17.2 shows several thermal images of
different pan materials (equal in size) captured by an IRT camera during IH under the same heat-
ing conditions. It can be observed that the temperature distribution results vary between the pans,
FIGURE 17.2 Difference in temperature distribution in different pan materials, captured by IRT camera.
(Courtesy of Tokyo Electric Power Company (TEPCO) – Japan.)
according to the material type. From these results, a more homogeneous temperature distribution
was observed for a pan manufactured with a complex combination of nine layers of different alumi-
num and stainless steel materials (image 6).
Table 17.1 summarizes the properties of the most common pan materials [15–19]. Thermal con-
ductivity of metals is relevant to cookware selection for obvious reasons; heat must pass from the
pan to its contents in order to boil, simmer, sauté, and fry foods. Density is related to the weight of
the pan. Specific heat provides information regarding the quantity of energy required to raise the
temperature by one unit. The coefficient of thermal expansion yields information regarding the level
of expansion expected, which contributes to the curvature of flat surfaces during heating. Because
of the high thermal diffusivity of metals, pans comprising metals (particularly with copper bases)
can be heated on the stove more efficiently than those with less conductive properties (i.e., glass and
ceramics) [20].
Many pans are manufactured from copper or aluminum, or contain a core of these two highly
conductive materials. The widespread use of aluminum alloys in pans is predominantly attributed
to its high thermal conductivity and low density. This core is sandwiched between stainless steel,
which has the advantages of being corrosion and stain resistant, as well as usable on an induction
appliance (owing to its magnetism). These metals help to better distribute the heat produced in, or
applied to, the pan. However, it has been also reported that stainless steel alone is not a very good
thermal conductor and can produce uneven heating during cooking [21]; moreover, it was reported
that plates thicker than ¼" improve the uniformity of heating by allowing the heat to spread thermal
energy laterally. Cast iron has a greater heat capacity than alternative metals. In addition, pans with
this material are easy to use and care for over a wide range of cooking types. These attributes make
cast iron a good cookware material [22]. Although laminated plates improve the application qual-
ity of utensils, they have disadvantages such as body deformation. This is because materials with
different coefficients of thermal expansion and stiffness are bonded to form laminated plate [23];
370
TABLE 17.1
Properties of Materials Commonly Used in Pan Construction
Coefficient of
Thermal Thermal Electric Relative Thermal
Diffusivity (α) Density ( ρ) Conductivity Specific Heat Conductivity Permeability Yield Strength Expansion,
Material [10−6·m2·s−1] [g·cm−3] (λ) [W·m·K−1] (Cp) [kJ·kg−1·K] (σ) [10−7·S·m−1] ( μ) [−] (Y) [MPa] [μm·m−1·°C−1]
AISI 430 6.95 7.80 24.90 0.46 1.60 850 513.5 –
AISI 410 6.90 7.80 24.70 0.46 1.70 850 1180 –
Cast Iron 7.40 7.18 26.60 0.505 9.00 425 385 –
Co 21.85 8.80 84.60 0.44 1.60 155 225 –
Ni 19.85 8.89 79.00 0.455 1.125 670 52 –
SS 304 – 8.03 16.2 0.50 – – – 17.3
SS 430 – 7.94 26.1 0.46 – – – 10.4
SS 316 – 7.99 16.2 0.50 – – – 16.2
SS 439 – 7.70 24.2 0.46 – – – 11.4
Al 1050 – 2.08 227 0.90 – – – 25.5
Al 3003 – 2.73 159 0.893 – – – 25.1
Al 3004 – 2.72 163 0.893 – – – 25.1
Cu – 8.96 401 0.385 – – – 17.0
ASTM A677 Steal 18.20 7.75 65.00 0.46 2.70 11550 300 –
M-47 No oriented
Electrical Steel
Mumetal (77% Ni, 4% 7.65 8.74 33.50 0.495 1.695 127500 715 –
Mo, 14% Fe, 5% Cu
Permalloy 80 (80% 7.90 8.24 32.50 0.50 1.71 190000 321 –
Ni, 4% Mo, 16% Fe)
Sources: Martínez-Gómez, J., Ibarra, D., Villacis, S., Cuji, P., and Cruz, P.R., Food Policy, 59, 2016; Hayashida, T., Nishizaki, I., Ueda, Y., and Honda, H., Journal of Multi-Criteria Decision
Analysis, 19, 2012; Matweb, Property search, http://www.matweb.com/ (accessed September 20, 2017), 2017; Oral, M., and Kettani, O., Journal of Multi-Criteria Decision
Analysis, 23, 2015; Ribas, J.R., and da Silva, M., Journal of Multi-Criteria Decision Analysis, 22, 2015; Villacis, S., Martínez, J., Riofrío, A.J., Carrión, D.F., Orozco, M.A.,
and Vaca, D., Energy Procedia, 75, 2015.
thus, there is interfacial stress in bi-metal structures. A study [24] of a bi-layer cookware of dif-
ferent metals demonstrated that a bi-layer of copper and stainless steel provides a more uniform
temperature distribution and maximum temperature, and has a better heat retaining ability than
other applied metals such as Al/SS. A further study [25] reported that stainless steel and titanium
provide an almost equivalent temperature distribution over the surface of the pan containing food.
For the selection of the best suitable cookware material for IH, several approaches including
energy efficiency, cost saving, high heating rate, and workability have been reported [12]. According
to [26], a pan cooking must be made of, or contain, a ferromagnetic metal such as cast iron or some
stainless steels. However, copper non-magnetic stainless steel, and aluminum vessels can be placed
on a ferromagnetic interface disk to improve the thermal diffusivity of a conventional hotplate. A
study [12] reported that Ni is the best material for the base of an induction pan, based on its high
permeability, to induce Eddy currents and generate heat. In another study [7], the heat transfer
characteristics of different sauce pans on electric coil, natural gas, and induction cooktops were
evaluated and the most appropriate pan for each cooking appliance was identified. They tested
seven pans (A–G) on different cooktops until the base reached 450°F and cooled to 88°F, measured
through thermocouples fixed at the bottom and wall of the pan, and thermal images were taken at
10 s intervals. Statistical analyses revealed that pans have a greater impact on cooking efficiency
than cooking appliances. Results showed that differences in cooking efficiencies are due to differ-
ences in heating principles and cooking appliance wattage, pan size and shape, composition, base
thickness, and mass. Moreover, it was reported that hot spots, which affect overall pan performance,
are reduced by combining various metals—resulting in different thermal conductivity values—as
well as thickening the base of the pan. Combining a metal with a low thermal conductivity with
other metals with higher values causes the heat transfer rate to decrease, thus saturating the layer
prior to transfer and reducing irregular heating [21]. Metals with higher thermal conductivity, such
as aluminum and copper, are commonly combined with those with lower values, such as stainless
steel. This allows the pan to have high heat transfer performance to increase the speed of cooking,
but also an overall evenness of temperature for increased uniformity.
17.4 VISUALIZATION APPROACHES OF CONVECTION

FLOW FOR DIFFERENT COOKING APPLIANCES
A flow arising naturally owing to the effect of a density difference in a body force field (e.g., gravity)
is known as natural convection. A density difference may arise as a result of a temperature gradient
or heterogeneous multiphase flow [3]. The phenomenon of natural convection in fluid-filled rect-
angular enclosures, square cavities, and cylindrical enclosures has received considerable attention
in recent years. In all of these cases, natural convection has been investigated by either numerical
analysis using in-house, open source, or commercial CFD software sometimes coupled with the
experimental techniques of flow visualization and temperature measurement. Flow visualization is
most commonly effected by the techniques of particle image velocimetry (PIV), hot film anemom-
etry (HFA), and laser-Doppler anemometry (LDA). In the published literature, scant attention has
been given to natural convection accompanied by a phase change (boiling). Natural convection
accompanied by a phase change has gained importance in many hazard-mitigation/safety applica-
tions, e.g., behavior of filled vessels in accidental fire cases, mechanisms of rapid discharge of heat
to stagnant liquid pools, etc. CFD modeling of this phenomenon is highly complex as the turbulence
is anisotropic and a phase change is also involved. A satisfactory model must incorporate complex
mass and momentum transfer terms, including sources and sinks, associated with phase change,
drag, lift, wall lubrication, and turbulent dispersion [3]. Heat transfer from the wall occurs by three
mechanisms: (a) single-phase heat transfer by natural convection, (b) boiling heat transfer, and
(c) enhanced heat transfer by bubble motion in the vicinity of the heat transfer surface.
Some recently published literature focusing on the visualization of convection flow for different
cooking appliances is summarized as follows. The visualization and quantification of the flow in
the pan during IH and GRH were studied using tracer particles [27]. They reported that when the
amount of energy entering the pan was the same during both IH and GRH, the flow velocity and the
direction of convection of the two types of heating were different. In their study, water flow in the
pan was visualized using a developed tracer particle made by imitating the method for making arti-
ficial salmon roe [28]. The density of the tracer particle was adjusted to be the same as that of water
(1000 kg·m–3) by choosing a tracer particle of 0.0025 m in diameter and adjusting it to 5.8% of the
olive oil quantity used to prepare the aqueous solution of the tracer particle. In addition, the invari-
ability of the tracer particle density value in the heating range (25–95°C) was confirmed. During IH
and GRH, the 3D position and flow velocity of the tracer particle were measured by taking pictures
using two video cameras set above and inside the pan, respectively. For details of the measurement
scheme of the visualization approach, readers are suggested to refer to [27]. Figure 17.3 shows an
example of the 3D coordinates obtained from a PIV study [27]. These results were obtained after 1.2 s
of the 12 and 18 s thermal treatment of IH and GRH, respectively, for the initial stage of heating
(Figure 17.3A). The results after 6 min of heating (Figure 17.3B) were 10.8 and 13.2 s for IH and
GRH, respectively. The acquisition time coordinates of IH and GRH are different because in the
coordinate acquisition, the tracer particles leave the field of view of the camera. Further, the circle
on the XY plane in the figure represents the bottom pan. The results show that if the same amount
of heat is applied by IH and GRH, as shown in Figure 17.3A for the initial stage of heating, the flow
rate towards the IH is greater than in GRH. However, the convection stabilized after approximately
6 min in both IH and GRH (Figure 17.3B). In Figure 17.3B, it can be observed that the flow rate of
the evaluated tracer particles fluctuates in a similar manner; however, for IH the flow occurred near
the center on the bottom of the pan, whereas for GRH it occurred close to the pan wall.
In another study [29], the entropy generation and heat-flow visualization within an inclined porous
square cavity with a hot wall, cold wall, and adiabatic walls were analyzed. FEM analysis was used
for the effect of Darcy number for heat-flow visualization (via headlines) and entropy generation
due to heat transfer and fluid friction within inclined porous cavities. [In fluid dynamics, Darcy
number represents the relative effect of the permeability of the medium versus its cross-sectional
area—commonly the diameter squared.] On the basis of the entropy generation minimization
approach, it was concluded that cavities with inclination angles φ ≤ 30° for lower Prandtl number
(Pr is a dimensionless number defined as the ratio of momentum diffusivity to thermal diffusivity)
FIGURE 17.3 Trajectory of motion of tracer particle during heating. (A) At initial stage of heating: 12 s in
IH (black squares) and 18 s in GRH (white circles), and (B) after 6 min of heating: 10.8 s in IH (black squares)
and 13.2 s in GRH (white circles). (From Kawakami, H., Tou, Z., Fukuoka, M., and Sakai, N., J. J. Food Eng.,
11, 2010.)
fluids (Pr = 0.025) and φ = 15° for higher Pr (Pr = 998) may be an efficient strategy for optimal
thermal processing.
17.5 VISUALIZATION APPROACHES OF CONVECTION

FLOW FOR DIFFERENT PAN MATERIALS
A series of test procedures to evaluate the heating performances of domestic appliances for case
of pan-frying (continuous measurement of pan temperature) and oven-baking (heat flux measure-
ments) has been described [5]. To assess the efficiency of these test procedures, they were applied to
a series of domestic appliances (different combinations of pans and hobs; different types of ovens).
For heat treatments of comparable intensity, the results showed considerable variability in effective
operating conditions (as experienced directly by the product during heating) based on both the type
of materials used and the behavior of the consumer when cooking at home. Thus, it was concluded
that the operating conditions stipulated by food product manufacturers may lead to a great variabil-
ity in heating conditions. Depending on the intensity of the heat treatment (medium or high heat),
during domestic pan-frying, the pan temperature declined continuously (medium heat) or rose (high
heat) to values that may be higher than 330°C. Likewise, the oven-baking tests showed that the
effective air temperature could be very different for heat treatments at the same air-temperature set
point. These differences were due to the type of oven used, the mode of convection selected, and the
level of convection reached (related to the power of the fan installed in the baking chamber), as well
as the level of sophistication of the air-temperature control systems.
Another study [13] evaluated the visualization of convection flow in three pots made from differ-
ent materials (enameled iron body and bottom, AISI 304 stainless steel body and AISI 430 stainless
steel bottom, and an aluminum body and AISI 430 stainless steel bottom) heated in an induction
cooker filled with olive oil. In this study, particle tracers were used for the visualization of olive
oil flow in the cookware by imitating the method reported in [2]. The particle tracer density was
adjusted to the density of the olive oil (970 kg·m−3) and the tracer solution was exactly balanced to
5.8% isobutyl isopropyl ether. The particle tracer resulted in a 0.0015 m diameter. They reported
that the convection flow in the enameled iron cookware showed the highest flow velocity at the early
stage (2 s intervals, from 5 to 31 s) of heating and after 300 s of heating.
17.6 MATHEMATICAL MODELING BY CFD

OF CONVECTION FLOW IN PAN COOKING
The power distribution transferred to the pans using gas-range stoves or electric hotplates is limited
in the literature, despite their being traditional technologies used globally in kitchens. In one study
[30], a cooking efficiency of induction, halogen, and resistance hobs of 67%, 60% and 30%, respec-
tively, was reported. They found that the efficiency of the pan depends on the pan mass, external
surface emissivity, and amount of water added (at least half full). They also reported that the aspect
ratio and configuration index affect the efficiencies considerably. In particular, in the case of induc-
tion cooking appliances, which is still a recent technology in the field of home appliances, there are
limited previous works related to thermal modeling. The majority of the reported works focused
on the different methods of calculating the induced power distribution, generally using FEM soft-
ware [31–33]. An evaluation of the stove functionalities (which depends on the type of heat source
(induction, electric resistance, gas burner)) in terms of heating uniformity of the cooking vessels
in domestic stoves, was recently reported [1]. In this study, inverse modeling was used to evaluate
the manner in which the power is transferred to the pan. They reported an approach to calculate
the power distribution generated in the bottom of the pan from the measurement of the surface
temperature.
In this section, some basic information related to model equations and solution procedure is intro-
duced, and selected CFD approaches to model the convection flow in pan cooking are summarized.
17.6.1 Basic Model Equations and Solution Procedure

17.6.1.1 Physical Model
The model described here was reported in a previous study [2]; for more details, please refer to this
study. The analytical model was a pan of 0.39 m in diameter and 0.39 m in height. The pan con-
tained 20 L, and the height of the water in the pan was 0.167 m. To simulate the same conditions as
in the experiment, the height of the liquid level from the bottom of the pan was set as 0.167 m. The
pan had a three-layer structure composed of stainless steel (0.6 mm), iron (2.8 mm), and stainless
steel (0.6 mm). The physical properties of the fluid were assumed to be those of water with an initial
temperature of 27°C.
17.6.1.2 Assumptions Used in the Numerical Simulation

The following assumptions were made to simplify the problem:
a. Axi-symmetry: The shape of the pan was assumed to be a cylinder. The convection of fluid
may show axial symmetry in the pan because the heating condition is axi-symmetrical.
b. Heat generation due to viscous dissipation is negligible.
c. Boussinesq approximation was applied for the buoyancy.
d. The assumption of no-slip condition at the inside wall of the pan is valid.
e. Convection in the pan was a laminar flow, because it was a natural convection induced by
heating with a Rayleigh number: Ra = 6.05 × 106 to 9.90 × 106 for IH, and Ra = 3.95 × 106
to 8.63 × 106 for GRH. [Ra of a fluid is a dimensionless number associated with buoyancy-
driven flow, also known as free convection or natural convection in fluid mechanics.
When the Ra is below a critical value for that fluid, heat transfer is primarily in the form
of conduction; when it exceeds the critical value, heat transfer is primarily in the form of
convection].
f. The gravity force of −9.8 m·s−2 was added axially to the pan.
The thermophysical properties of water were considered to be temperature-dependent, and their

values change during the CFD simulation.
The partial differential equations governing natural convection motion in a cylinder are the Navier–
Stokes equations in cylindrical coordinates [34], as shown below:
a.
Continuity equation
1 ∂ 1 ∂ ∂
(r ρ vr ) + ( ρ vθ ) + ( ρ vz ) = 0 (17.1)
r ∂r r ∂θ ∂z
where r, θ, and z are the radial, angular, and vertical direction of the pan (m), respectively;
ρ is the density (kg·m−3); and v the velocity (m·s−1).
b.
Energy conservation equation
∂T ∂T vθ ∂T ∂T k  1 ∂  ∂T  1 ∂2 T ∂2 T 
+ vr + + vz =  r + +  (17.2)
∂t ∂r r ∂θ ∂z ρC p  r ∂r  ∂r  r 2 ∂θ 2 ∂z 2 

where T is the temperature (°C), t the time (s), k the thermal conductivity (W·m−1·°C−1), and
Cp the specific heat (J·kg−1·°C−1).
To simplify the analysis, the buoyancy force caused by the density variation with tem-
perature was governed by the Boussinesq approximation used in the body force term of the
momentum equation in the radial direction [Equation (17.3)].
c.
Momentum equation
In the radial direction (r)
 ∂v ∂v v ∂v v2 ∂v 
ρ  r + vr r + θ r − θ + vz r 
 ∂t ∂r r ∂θ r ∂z 

∂p  ∂ 1 ∂  1 ∂ 2 vr 2 ∂vθ ∂2 vr 
=− +µ  (rvr ) + 2 − + 2  + ρref g 1 − β (T − Tref )  (17.3)

∂r  ∂r  r ∂r  r ∂θ 2 r 2 ∂θ ∂z 
where β is the thermal expansion coefficient of the liquid (K−1); Tref and ρref are the refer-
ence temperature and density, respectively; μ is the apparent viscosity (Pa·s); and g is the
gravitational acceleration (m·s−2).
In the vertical direction (z)
 ∂v ∂v v ∂v ∂v  ∂ρ  1 ∂  ∂vz  1 ∂2 vz ∂2 vz 
ρ  z + vr z + θ z + v z z  = − +µ r + 2 2 
(17.4)
 ∂t ∂r r ∂θ ∂z  ∂z  r ∂r  ∂r  r vθ ∂z 
2

In the angular direction (θ)
 ∂v ∂v v ∂v vv ∂v 
ρ  θ + vr θ + θ θ + r θ + vz θ 
 ∂t ∂r r ∂θ r ∂z 

1 ∂p  ∂ 1 ∂  1 ∂2 vθ 2 ∂vr ∂2 vθ 
=−
r ∂θ
+µ 
 ∂r  r ∂r
( )
rvθ  + 2 +
 r ∂θ 2 r 2 ∂θ
+ 2 
∂z 
(17.5)

17.6.1.4 Boundary Conditions
The boundary conditions used at the pan surface (r = R, z = 0) were:

ν r = 0, vθ = 0, vz = 0. (17.6)
Initially, the fluid is at rest and a uniform temperature:
T = Tref = 27°C , vr = 0, vθ = 0, vz = 0 at 0 ≤ r ≤ R , 0 ≤ z ≤ H. (17.7)

a. IH: Because the heat is released from the side wall of the pan, this was set as the heat
transfer boundary. The heat transfer coefficient was 10 J·m−2·s−1.
b. GRH: In the case of GRH, the measured heat flux was used as a boundary condition of the
CFD analysis, and it was set to the outer layer of stainless steel.
17.6.1.5 Analysis of Electromagnetic Field

The electromagnetic model is classically based on the set of Maxwell’s equations. This system is
based on the four following equations:
Magnetic flux equation
→ →
∇⋅ B = 0 (17.8)
Maxwell–Gauss equation
→ →
∇⋅ E = 0 (17.9)
Maxwell–Faraday equation
→
→ →
∂B
∇× E = − (17.10)
∂t
Maxwell–Ampere equation
→
→ → →
∂D
∇× H = J + (17.11)
∂t
→ → →
where H is the magnetic field (A·m), B the magnetic induction (N·s·C–1·m–1), E the electric field
→ →
(V·m–1), D the electric flux density (C·m–2), and J the electric current density associated with free
charges (A·m–2). In addition, the following relations were used for the intrinsic material properties:
→ →
D = ε E (17.12)

→
( ) →
B = µm T , H H (17.13)
→
Ohm’s law
→ →
J = σ (T ) E (17.14)
where ε is the dielectric constant (F·m–1), μm the magnetic permeability (N·A–2), and σ the electrical
conductivity (S·m–1). For low and medium frequencies, the displacement currents in the Maxwell–
Ampere equation (magneto-quasi-static approximation) can be neglected.
(A–V) formulation
To model the IH process, the Maxwell’s equations can be approximated using the magneto-
quasi-static approximation hypothesis [35]. This equation system is rewritten using a single non-
linear equation of the A–V formulation [36]. This formulation is based on defining a magnetic
→
vector potential A (V·s·m–1) such that:
→ → →
B = ∇ × A (17.15)
Using this expression in Maxwell–Ampere and Maxwell–Faraday equations leads us to rewrite

the electric field:
→
→
∂A →
E=− − ∇ V (17.16)
∂t
where V is the electric scalar potential (V). Finally, the following equation was obtained:
→
∂ A →  1 → → →
σ +V×  ∇× A = −σ ∇ V (17.17)

∂t  µm 
→
where σ and μm both depend on temperature and H , the magnetic field. In Equation (17.17), the
right-hand term represents the input current. To ensure the uniqueness of the solution, an additional
equation known as the Coulomb gauge was considered:
→ →
∇ ⋅ A = 0 (17.18)
17.6.1.6 Solution Strategy
The software used for the simulations was STORM/CFD2000 (Simunet Corp., USA), which is
based on the finite volume method. To shorten the computation time, calculations were done for
only half of the pan. Because, for the same reason, all of the surroundings of the pan do not need
to be calculated, only a 10° fraction of the 360° pan was used for the calculations (Figure 17.4).
However, for convection in two kinds of fluids, such as water and air, in a single model, the calcula-
tion is complex and required considerable time. In such cases, it is assumed that there is no convec-
tion, but only conduction, in the air layer since the target fluid in this analysis is water.
The value of the heat flux or heat generation, as the heat source, is required for the CFD analysis.
Since the heat flux changes according to the heating conditions, it is necessary to measure the heat
flux whenever the heating condition changes. However, if the heat generation in IH can be estimated
using electromagnetic field analysis (EFA), the heat flux need not be measured. Thus, in IH, the
heat generation distribution in the pan was estimated by EFA using commercial software (JMAG
Studio, USA). CFD was customized by FORTRAN programming to introduce the heat generation
FIGURE 17.4 Diagram of model for simulation of convection flow in a cooking pan: (A) Full model and
(B) 10° vertical cross-section of the cooking pan approach.
into the CFD calculation. By a so-called skin effect, the high-frequency current or the electromagnetic
field is limited to the surface of the conductor. In addition, since the boiling calculation was not possible
with the software used at the time of the study (STORM/CFD2000), it was customized in FORTRAN
programming by using the temperature recovery method as follows. The temperature fluctuation with
the boiling point (Tb), ΔT = T − Tb, was calculated at each point in the pan. Thus, if ΔT > 0, the sensible
heat over the boiling point was considered to be released as latent heat, and this heat was added to the
energy conservation law. Nowadays, the features of available computer simulation software, such as
COMSOL Multiphysics, enable simulation of these phenomena without additional customization
works.
17.6.1.7 Computational Grid
A non-uniform grid system in the radial and axial directions is used in the calculation and simula-
tion. The whole domain was divided into 1500 cells: 50 in the radial direction, 30 in the axial direc-
tion, and only one in the angular direction (each cell represents 10° of the cylindrical pan), graded
with a finer grid near the wall in the radial direction. The calculation of the natural convection
heating of the water was conducted for 600 s. It took 60000 steps to reach 600 s at 0.01 s intervals.
For the animation of the convection flow of water, 100 time steps were used to cover the total time
of heating (600 s) at 1 s intervals to achieve 600 images.
17.6.2 CFD Applications in Convection Flow in Pans Using IH

In the last 15 years of IH modeling, some studies have been conducted to understand the principles
of IH by numerical methods and CFD. There was an evaluation of the differences in convection
flow of a single-phase heat transfer, by natural convection, within a pan filled with water heated by
IH and GRH using flow visualization and simulations by 3D STORM/CFD2000 [2]. The non-linear
and transient magneto-thermal coupled problem with a moving coil, which is considered a moving
heat source, was dealt with using the multi analytic software described in Section 17.6.1. Some of
their results are discussed as follows. In Figure 17.5, the visualization of the temperature distribu-
tion in the pan heated by IH and GRH are shown after 60 s and 360 s of heating process. For IH, the
temperature at the bottom of the pan, excluding its edge, increased as soon as the pan was heated.
Water was heated as the bottom was heated, and it can be seen that some of the heated water rose
(at positions close to the location of the coils at the bottom of the pan) until it reached the top water
surface after 360 s. The temperatures at the side wall of the pan did not increase significantly. In
the case of GRH, the whole area on the bottom and the corner of the pan were heated, unlike in IH.
The temperature of the water was more uniform than in IH, and the temperature change occurred
until the heated water reached the top water surface after 360 s. Figure 17.6 shows the calculation
results for the resultant velocity convection flow in the pan heated by IH and GRH at 10, 60, and
360 s. The length and direction of the arrow represents the magnitude and direction, respectively,
of convection. For IH, while the convection had barely occurred at 10 s of heating, the convection
had risen from inside of the pan and had moved outside by 60 s. The tendency of convection from
the inside to the outside was strengthened at 360 s of heating. Comparing Figures 17.6 and 17.5, it
can be seen that the water rose in the region of the pan where the temperature was high, and moved
outside the vicinity of the surface of the water. In other words, the water adjacent to the bottom
surface above the coils received a higher induction heat. As the liquid is heated, it expands and thus
becomes lighter. Liquid away from the coils position remains at a considerably lower temperature.
The buoyancy force created by the change in liquid density due to temperature variation produces
an upward flow, near the coil position (from the bottom of the pan to the top). The rising hot liquid
is deflected at the top, and then travels radially towards the side of the pan. Being heavier, the water
near the sides moves downwards and then towards the coil position. In the case of GRH, over the
whole fluid the convection had barely occurred at 10 s of heating; however, there was slight convec-
tion at the area of the corner of the pan. At 60 s of heating time, convection began from the vicinity
FIGURE 17.5 Visualization of temperature distribution of water in pan cooking. (A) Heated by IH, and
(B) heated by GRH. The right and left extremities of the figure are the side and center of the pan, respectively.
(From Kawakami, H., Llave, Y., Fukuoka, M., and Sakai, N., J. Food Eng., 116, 2013.)
FIGURE 17.6 Calculated convection flow in pan cooking. (A) Heated by IH, and (B) heated by GRH.
(From Kawakami, H., Llave, Y., Fukuoka, M., and Sakai, N., J. Food Eng., 116, 2013.)
above the gas flame, rose to the surface of the water, and diverged in the internal and external direc-
tions. At 360 s of heating time, the convection became irregular convection through which the fluid
inside the entire pan was mixed. Thus, after a few minutes of GRH, the general behavior of the fluid
motion became irregular in the entire pan.
Another study [13] investigated the temperature and the distribution of natural convection
flow of three different cookware materials composed of stainless steel, aluminum, and enameled
iron, when heating olive oil in an induction stove. A CFD module of COMSOL Multiphysics was
used to analyze the heat transfer process for this study. The simulated results were validated with
measurements of temperature, heat transfer, and flow velocity using a thermographic camera
and a particle tracer approach. As shown in Figure 17.7, the biggest contrails of hotter fluid ris-
ing from the bottom to the top were detected in the enameled iron (Figure 17.7A) and aluminum
cookware (Figure 17.7C). At the enameled iron cookware material, the density of the contrails
increased around the midpoint of the cookware. In the case of stainless steel (Figure 17.7B), the
temperature of the fluid was lower than that of the other two cookware materials, but the number
of contrails was higher. To explain these results in more detail, after 300 s the enameled iron
(Figure 17.8A) presented the biggest contrails and oil velocity convection flow in the center of the
cookware compared to aluminum (Figure 17.8C) and stainless steel (Figure 17.8B). This factor,
and the increment of the temperature of the fluid, caused the authors to decide that the enameled
iron was the best choice from the three evaluated materials for pan cooking material in induction
stoves. The authors attributed the results to the intrinsic properties of the cookware materials
(as shown in Table 17.1). They concluded that magnetic permeability and thermal diffusivity
are decisive for the choice of material, especially for the material of the base of the pans. Thus,
magnetic permeability influences the magnetic field concentration in the base of the cookware,
whereas thermal diffusivity contributes to the speed at which the heat is transferred through and
out of the material.
Additional CFD applications in convection flow in pans using IH are summarized as follows.
A numerical analysis of a new coil design for IH and its experimental verification with globalization
evaluation has been performed [37]. In another study [38], the effect of multi-layered induction coils
on the efficiency and uniformity of surface heating was studied using multiple physical coupling
analyses in ANSYS software, to model the thermal effect of IH on the surface of the heated target.
A numerical analysis and thermographic behavior of IH using thermographic cameras to validate
the temperature distribution and other dynamic changes of the target surface has been investigated
[39]. They demonstrated that thermographic measurements are useful tools for observing tempera-
ture distributions of work-pieces during IH. A non-linear and transient magneto-thermal coupled
problem with a moving coil, which is considered a moving heat source, was evaluated by an efficient
finite-element procedure [40]. The effect of velocity, initial position of the inductor, and inner to outer
radius ratio on temperature distribution were investigated. In another study [41], new approaches for
designing and optimizing IH treatment processes by numerical methods were introduced.
17.6.3 CFD Applications in Convection Flow in Pans Using GRH

The convection flow of a single-phase heat transfer by natural convection, within the pan heated
by GRH, was also analyzed [2] in terms of the temperature history and flow velocity of convec-
tion using flow visualization and simulations by 3D STORM/CFD2000 (results also observed in
Figures 17.5 and 17.6, explained in Section 17.6.2). Results confirmed that in GRH, the bottom and
sides of the pan are heated by the flame, and so the heated location is different from that in IH. This
is especially important because thermal stratification has been reported to be undesirable during
cooking. Thus, when a pan cooking is heated from the sides (by steam), thermal convection cur-
rents are feeble owing to the presence of cooking items and may cause food in the central region to
remain uncooked [42].
FIGURE 17.7 Visualization of temperature distribution of olive oil in a pan heated by IH at 300 s.
(A) Enameled iron, (B) stainless steel, and (C) aluminum. The right and left extremities of the figure are the
side and center of the pot, respectively. (From Kastillo, J.P., Martínez, J., Riofrio, A.J., Villacis, S.P., and
Orozco, M.A., Proceedings of the 1st Pan-American Congress on Computational Mechanics – PANACM
2015, 2015.)
The cooking process for burners using LPG as the fuel was analyzed by CFD simulation [43] to
develop an improved cooking device. The major strategy for the optimization of energy utilization
was to design an appropriate insulation that was obtained by using two cover vessels. Thus, the flow
and temperature patterns in the air gap were extensively analyzed using CFD, in order to select an
air gap. The flow pattern and heat transfer in pan cooking have also been analyzed by CFD to design
382
FIGURE 17.8 Calculated convection flow in pots with olive oil heated by IH at 10, 60, and 300 s (left, center, and right, respectively). (A) Enameled iron, (B) stain-
less steel, and (C) aluminum. The right and left extremities of the figure are the side and center of the pot, respectively. (From Kastillo, J.P., Martínez, J., Riofrio, A.J.,
Villacis, S.P., and Orozco, M.A., Proceedings of the 1st Pan-American Congress on Computational Mechanics – PANACM 2015, 2015.)
suitable internal structures to minimize the stratification of temperature. The understanding of fluid
mechanics has also provided a basis for the selection of heat flux, gap between burner tip and cooker
bottom, and temperature of flue gases leaving the cooker. They concluded that to optimize the air
gap, it is important to analyze the fluid mechanics for different cooker sizes.
In one study [21], several aspects of thermal and mechanical properties of pan materials were
evaluated. The numerical results of temperature distribution, heat retention, thermal stress, and body
deformation were summarized. A finite element method using the ANSYS program was employed
to describe the thermal and mechanical results for Al/Cr-Ni, Al/SSt, Al/Ti, Cu/Cr-Ni, Cu/SSt, Cu/
Ti, gray cast iron (GCI), carbon steel (CSt), and iron. Laminated plates provide improved thermal
and chemical properties in comparison with single layer plates. From this analysis, it was suggested
that the Cu/SSt bi-metal structure provided the best pan structure for cooking.
17.6.4 Additional CFD Applications in Convection Flow

The applications of CFD in the food industry have provided food engineers with new insights and
understanding of the likely performance of food equipment at the design stage and confidence in
the quality or safety of food products [44]. Some of the most relevant applications of CFD in the
food industry have been summarized in [45–49] and in the remaining chapters in this book. Some
of the most relevant CFD applications to simulate convection flow in a range of foods, which can be
used as references for further developments, are summarized as follows. The convection behavior,
temperature change, and change in airflow in fluid foods during retort sterilization were analyzed
using CFD [50]. The temperature distribution and the flow velocity distribution of the fluid in 3D
pouches during heating sterilization were evaluated using CFD [51]. Further, these authors pre-
dicted in another study [52] the transient temperature and velocity profiles in a horizontal can during
sterilization by numerical simulation using CFD.
The change in the flow rate and temperature history in the sterilization of fluid foods in a tube-
type heat exchanger were studied [53]. Another study [3] investigated the single-phase and two-
phase (boiling) natural convection accompanied by thermal stratification, using CFD simulations
and PIV measurement. The flow information obtained from PIV was analyzed for insights into the
dynamics of turbulent flow structures. They used the signal processing technique of discrete wave-
let transform (DWT) for this purpose. Thus, from the analysis, they were able to estimate the size,
velocity, and energy distribution of turbulent structures in their flows. This data was used to esti-
mate wall heat transfer coefficients. In another study [54], a 3D CFD model for forced air cooling of
eggs placed in trays was developed. The continuity, momentum, and energy equations were solved
along with the standard k − ε turbulence model using PHOENICS software. The model was vali-
dated by conducting experiments in a wind tunnel at various air temperatures (7–11°C) and veloci-
ties (0.3–0.7 m·s–1). The root mean square error for predicting yolk temperatures was within 1°C.
The CFD model was integrated with a microbial growth model to estimate the risk of Salmonella
enteritidis growth during cooling.
17.7 APPROACHES FOR CONTINUOUS COOKING SYSTEMS OPTIMIZATION

Large scale/community cooking has become a necessity at many places, such as hostels, jails,
industrial kitchens, religious places, and kitchens, for preparing mid-day meals for school chil-
dren. Although the scale of cooking is large in these applications, the method of cooking is largely
traditionally and practically independent of scale. Although various heating methods, such as
three-stone fires, biomass and gas stoves, electrical, microwaves, and induction ovens, have been
used over the years for cooking, the most widely used method is open pan cooking. This method
has shown a thermal energy efficiency of less than 20%. In addition, it has been reported to be
less hygienic, more time consuming, and wasteful of resources like water and fuel [55]. To solve
these problems, several studies on the optimization of large scale cooking processes, such as the
continuous cooking process, have been developed to improve the energy efficiency, ease of opera-
tion, and efficient resource utilization by experimentation and modeling. Some of these studies are
summarized as follows.
One study measured the temperature field in their system using thermocouples and simulated the
same using commercial software CFX4, using the wall-boiling model included therein [56]. It was
assumed that neither temperature nor gas fraction influences the density of the liquid, and that the
momentum of the gas phase was negligible. These assumptions are rather restrictive. In another study
[57], the thermal mixing in a rectangular water tank heated by steam was evaluated and the effect of
tube orientation (i.e., horizontal or vertical) on natural convection was investigated experimentally.
Thus, the time required to achieve the saturation temperature of water was found to be independent of
the tube orientation. Natural convection in large pools has been investigated using temperature mea-
surements and CFD simulations using commercial code CFX-4 [58]; they used a laminar flow model for
the system to investigate temperature oscillations due to the presence of a plume near the tube surface.
In another study [59], experimental (thermocouples for temperature measurements and holo-
graphic interferometer for flow measurements) and numerical analysis of air in a square enclosure,
approximating a 2D vertical slot, were carried out. The air was partially heated from the bottom by
a centrally located heating element, with the vertical walls acting as cooling elements. The other
surfaces were considered adiabatic. The numerical simulations were performed using the commer-
cial code FLUENT 6, assuming 2D geometry. They concluded that in a vertical slot, for a Ra less
than 104, the heat transfer took place mainly by conduction, whereas for Ra equal to 105, heat trans-
fer took place by convection. In another study [60], the turbulent natural convection in a rectangular
enclosure heated from the bottom was evaluated by numerical simulations of a 2D geometry using
commercial codes (FLUENT, PHONIX, STAR-CD) and an in-house code. The working fluid, air,
was considered with Ra ranging from 108 to 1012. They investigated the effect of different bound-
ary conditions (constant temperature and constant heat flux) on natural convection by performing
steady state and transient simulations. They reported two counter-rotating cells for the entire range
of Ra numbers and for various values of heated width, and found that for isothermal heating, the
Nusselt number (Nu) increases with an increase in the heated width. In heat transfer at a boundary
(surface) within a fluid, the Nu is the ratio of convective to conductive heat transfer across (normal
to) the boundary. The opposite was found to be true for isoflux heating [60]. They concluded that the
Nu is a weaker function of heated width for isoflux heating compared to isothermal heating. They
further concluded that the intensity of buoyancy-induced flow in the case of isoflux heating is lower
compared to that from isothermal heating at identical values of Ra. They proposed correlations for
the Nu for both boundary conditions.
CFD simulations and experimental measurements for a centrally heated tube placed in a cylin-
drical enclosure were performed [61]. They measured the flow patterns by PIV measurement and
HFA. Temperature patterns were obtained using thermocouples. They performed CFD simulations
for a 2D axi-symmetric geometry using commercial software FLUENT 6.3 with a user-defined
function (UDF) for incorporation of their phase change (boiling) model. The value of Ra varied in
the range from 4.98 × 1012 to 3.764 × 1013. They considered water as a working fluid, and density
variation was modeled using the Boussinesq approximation. In another study [43], the optimiza-
tion design and scale up of continuous cookers were evaluated using CFD. They reported that, for
the development of continuous cookers, knowledge of kinetics, thermal mixing, axial mixing, and
optimum selection of insulation is relevant. In another study [42], flow (using PIV) and temperature
(using thermocouples) measurements were carried out in a rectangular tank (0.8 × 0.6 × 0.6 m3) fit-
ted with a central tube (forming the heat transfer surface), in addition to CFD simulations using the
commercial software FLUENT-6.3 to evaluate the reduction in the stratification of several imple-
mented systems. By simulation, they were able to estimate the size, velocity, and energy distribution
of turbulent structures.
17.8 SUGGESTIONS FOR FUTURE WORK

The following research areas have not yet been studied and are recommended to be prioritized in
further studies of convection flow in pan materials.
a. Investigations are required to understand the sensitivity of pan diameter, extent of

filling, ratios of solid food to liquid, and the extent of cooking. It should be noted that
the porosity, average particle size, and level of cooked material change with respect to
time.
b. CFD simulations need to take into account the flow through porous media with variable
thickness.
c. Detailed analysis needs to be carried out regarding temperature stratification. This infor-
mation should enable the development of suitable internal structure to reduce the extent of
stratification to permit cooking at all locations.
d. The flow pattern around the particle should also be investigated using CFD with direct
numerical simulation (DNS).
e. It was observed that the heat flux absorbed by the pan cooking diminishes as the tem-
perature approaches boiling point [43]. To implement this observation, a control strategy,
simple in construction and operation, needs to be devised.
f. Finally, to evaluate the convection flow of real foods such as liquid foods and solids mixed
with liquid foods (e.g., pasta products like spaghetti), systematic research work is required
to estimate the flow velocity, temperature, pressure profiles, and water diffusion of a vari-
ety of foods cooked in practice.
17.9 CONCLUSIONS
CFD is an attractive and useful technology that, if applied with cognizance of the limitations of
some of its weaker aspects and a willingness to perform the appropriate mesh convergence stud-
ies, can be a valuable tool for the design of new systems or the optimization of existing processes.
In this chapter, the convection in a pan heated by IH and by GRH was analyzed for several pan
materials, using CFD. In addition, some experimental visualization studies of the flow in the pan
were presented using tracer particles. The simulation of the convection within the pan showed
almost the same tendency as the movement of the tracer particles in the flow visualization. The
direction and resultant velocity of convection differed between heating appliances, especially
at the initial stage of the heat; however, at high heating temperature, the resultant velocities
are similar under both heating types. These results, in addition to the velocity and direction of
convection flow in the pan, were estimated successfully using CFD analysis in a number of stud-
ies summarized in this chapter. In contrast, it was confirmed that material selection is of great
importance in the design and development of pan material; this is because the selection of an
appropriate material is necessary to improve the efficiency in the system cooking appliance (pan
cooking). Differences in the energy efficiency and temperature distribution during heating and
cooling are related to the configuration of the pan and the different materials thereof. From the
results of some CFD simulation approaches, it was concluded that the magnetic permeability and
thermal diffusivity are decisive in terms of choice of material, especially for the material of the
base of the pans. Thus, magnetic permeability influences the magnetic field concentration in the
base of the pan, whereas thermal diffusivity contributes to the speed at which the heat is trans-
ferred through the material. The data described herein is valuable as a tool for research on the
heating patterns during convection flow of solids mixed with liquids for several cooking appli-
ances and pan materials.
REFERENCES
1. Sanz-Serrano F., Sagues C., Llorente S., Inverse modeling of pan heating in domestic cookers, Applied
2. Kawakami H., Llave Y., Fukuoka M., Sakai N., CFD analysis of the convection flow in the pan during
induction heating and gas range heating, Journal of Food Engineering, 116: 726–736, 2013.
3. Gandhi M.S., Sathe M.J., Joshi J.B., Vijayan P.K., Two phase natural convection: CFD simulations and
PIV measurement, Chemical Engineering Science, 66: 3152–171, 2011.
4. Kuriakose R., Anandharamakrishnan C., Computational fluid dynamics (CFD) applications in spray
drying of food products, Trends in Food Science & Technology, 21: 383–398, 2010.
5. Cernela J., Heyd B., Broyart B., Evaluation of heating performances and associated variability of domes-
tic cooking appliances (oven-baking and pan-frying), Applied Thermal Engineering, 62: 758–765, 2014.
6. Cadavid F.J., Cadavid Y., Amell A.A., Arrieta A.E., Echavarría J.D., Numerical and experimental meth-
odology to measure the thermal efficiency of pots on electrical stoves, Energy, 73: 258–263, 2014.
7. Karunanithy C., Shafer K. Heat transfer characteristics and cooking efficiency of different sauce pans
on various cooktops, Applied Thermal Engineering, 93: 1202–1215, 2016.
8. Zorrilla S.E., Singh R.P., Heat transfer in double-sided cooking of meat patties considering two-
dimensional geometry and radial shrinkage, Journal of Food Engineering, 57: 57–65, 2003.
9. Feyissa A.H., Gernaey K.V., Ashokkumar S., Adler-Nissen J., Modelling of coupled heat and mass
transfer during a contact baking process, Journal of Food Engineering, 106: 228–235, 2011.
10. Shin J.G., Woo J.H., Analysis of heat transfer between the gas torch and the plate for the application of
line heating, Journal of Manufacturing Science and Engineering, 125: 794–800, 2003.
11. Sung Y.T., Hwang S.J., Lee H.H., Huang D.Y., Study on induction heating coil for uniform mold
cavity surface heating, Advances in Mechanical Engineering, 6: 349078, 2014, http://dx.doi.org
/10.1155/2014/349078.
12. Martínez-Gómez J., Guerrón G., Narvaez A.E., Cookware material selection by multi-criteria deci-
sion making (MCDM) methods, International Journal of Engineering Trends and Technology, 34(8):
394–399, 2016.
13. Kastillo J.P., Martínez J., Riofrio A.J., Villacis S.P., Orozco M.A., Computational fluid dynamic anal-
ysis of olive oil in different induction pots, In: Proceedings of the 1st Pan-American Congress on
Computational Mechanics – PANACM 2015, Argentine, 2015.
14. Hager T.J., Morawicki R., Energy consumption during cooking in the residential sector of developed
nations: A review, Food Policy, 40: 54–63, 2013.
15. Hayashida T., Nishizaki I., Ueda Y., Honda H., Multi-criteria evaluation for collaborative circulat-
ing farming with collective operations between Arable and Cattle farmers, Journal of Multi-Criteria
Decision Analysis, 19(5–6): 227–245, 2012.
16. Matweb, Property search, http://www.matweb.com/ (accessed September 20, 2017), 2017.
17. Oral M., Kettani O., Conceptualizing a research paradigm for multi-objective modelling in supply chain
management, Journal of Multi-Criteria Decision Analysis, 23: 5–6, 2015.
18. Ribas J.R., da Silva M., A decision support system for prioritizing investments in an energy efficiency
program in favelas in the city of Rio de Janeiro, Journal of Multi-Criteria Decision Analysis, 22(1–2):
89–99, 2015.
19. Villacis S., Martínez J., Riofrío A.J., Carrión D.F., Orozco M.A., Vaca D., Energy efficiency analysis of
different materials for cookware commonly used in induction cookers, Energy Procedia, 75: 925–930,
2015.
20. Newborough M., Probert S.D., Energy-conscious design improvements for electric hobs, Journal of
Foodservice Systems, 4: 233–257, 1987.
21. Sedighi M.R., Dardashti B.N., A review of thermal and mechanical analysis in single and bi-layer plate,
Materials Physics and Mechanics, 14: 37–46, 2012.
22. Barr T., Cast Iron Cooking for Dummies, John Wiley & Sons, Hoboken, NJ, USA, 2011.
23. Wen Y., Basaran C., An analytical model for thermal stress analysis of multi-layered microelectronic
packaging, Mechanics of Materials, 36(4): 369–385, 2004.
24. Sedighi M.R., Dardashti B.N., Heat transfer modeling in multilayer cookware using finite element
method, International Journal of Science and Engineering Investigations, 1(1): 190–196, 2012.
25. Dardashti N., Sedighi M.R., Numerical solution of heat transfer for single and multi-metal pan, Applied
Mechanics and Materials, 227: 148–149, 2012.
26. Martínez-Gómez J., Ibarra D., Villacis S., Cuji P., Cruz P.R., Analysis of LPG, electric and induction
cookers during cooking typical Ecuadorian dishes into the national efficient cooking program, Food
Policy, 59: 88–102, 2016.
27. Kawakami H., Tou Z., Fukuoka M., Sakai N., The visualization and quantification of the flow in the
pan during induction heating and gas heating, Japan Journal of Food Engineering, 11: 51–58, 2010. In
Japanese.
28. Kamio K., Man-made salmon roe, The Chemical Education, 35: 309–311, 1987. In Japanese.
29. Basak T., Singh A.K., Richard R., Roy S., Finite element simulation with heatlines and entropy gen-
eration minimization during natural convection within porous tilted square cavities, Industrial &
Engineering Chemistry Research, 52(23): 8046–8061, 2013.
30. Newborough M., Probert S.D., Newman M., Thermal performances of induction halogen and conven-
tional electric catering hobs, Applied Energy, 37: 37–71, 1990.
31. Byun J.K., Choi K., Roh H.S., Hahn S.Y., Optimal design procedure for a practical induction heating
cooker, IEEE Transactions on Magnetics, 36: 1386–1389, 2000.
32. Pham H.N., Fujita H., Ozaki K., Uchida N., Estimating method of heat distribution using 3-D resis-
tance matrix for zone-control induction heating systems, IEEE Transactions on Power Electronics, 27:
3374–3382, 2012.
33. Souley M., Egalon J., Caux S., Pateau O., Lefevre Y., Maussion P., Optimization of the settings of mul-
tiphase induction heating system, IEEE Transactions on Industry Applications, 49: 2444–2450, 2013.
34. Bird R.B., Stewart W.E., Lighfoot E.N., Transport Phenomena, John Wiley & Sons, Hoboken, NJ, USA, 1976.
35. Bay F., Labbe V., Favennec Y., Chenot J.-L., A numerical model for induction heating processes coupling
electromagnetism and thermomechanics, International Journal of Numerical Methods in Engineering,
58: 839–867, 2000.
36. Chari M., Konrad A., Palmo M. et al., Three-dimensional vector potential analysis for machine field
problems, IEEE Transactions on Magnetics, 18(2): 436–446, 1982.
37. Kang, C.G., Seo P.K., Jung H.K., Numerical analysis by new proposed coil design method in induction
heating process for semi-solid forming and its experimental verification with globalization evaluation,
Materials Science and Engineering, 341(1): 121–138, 2003.
38. Huang M.S., Huang Y.L., Effect of multi-layered induction coils on efficiency and uniformity of surface
heating, International Journal of Heat and Mass Transfer, 53: 2414–2423, 2010.
39. Kranjc M., Zupanic A., Miklavcic D., Jarm T., Numerical analysis and thermographic investigation of
induction heating, International Journal of Heat and Mass Transfer, 53: 3585–3591, 2010.
40. Shokouhmand H., Ghaffari S., Thermal analysis of moving induction heating of a hollow cylinder with
subsequent spray cooling: Effect of velocity, initial position of coil, and geometry, Applied Mathematical
Modelling, 36: 4304–4323, 2012.
41. Naar R., Bay F. Numerical optimization for induction heat treatment processes, Applied Mathematical
Modelling, 37(4): 2074–2085, 2013.
42. Gandhi M.S., Joshi J.B., Nayak A.K., Vijayan P., Reduction in thermal stratification in two phase natu-
ral convection in rectangular tanks: CFD simulations and PIV measurements, Chemical Engineering
Science, 100: 300–325, 2013.
43. Joshi J.B., Pandit A.B., Patel S.B. et al., Development of efficient design of cooking systems—II.
Computational fluid dynamics and optimization, Industrial & Engineering Chemistry Research, 51:
1897–1922, 2012.
44. FRPERC, CFD in the food industry, Food Refrigeration and Process Engineering Research Centre
Newsletter, 10: 1–2, 1995.
45. Scott G., Richardson P., The application of computational fluid dynamics in the food industry: A review,
Trends in Food Science and Technology, 8: 119–124, 1997.
46. Xia B., Sun D.-W., Applications of computational fluid dynamics (CFD) in the food industry: A review,
Computers and Electronics in Agriculture, 34: 5–24, 2002.
47. Wang L., Sun D.-W., Recent developments in numerical modeling of heating and cooling process in the
food industry: A review, Trends in Food Science and Technology, 14: 408–423, 2003.
analysis tool for the food industry: A review, Trends in Food Science and Technology, 17: 600–620,
2006.
49. Erdoğdu F., Computational fluid dynamics for optimization in food processing, In: Optimization in Food
Engineering, Erdoğdu F., Ed., CRC Press, Boca Raton, FL, USA, 219–227, 2009.
50. Kumar A., Bhattacharya M., Transient temperature and velocity profiles in a canned non Newtonian
liquid food during sterilization in a still-cook retort, International Journal of Heat and Mass Transfer,
34: 1083–1096, 1991.
51. Ghani A.G.A., Farid M.M., Chen X.D., Richards P., Thermal sterilization of canned food in a 3-D pouch
using computational fluid dynamics, Journal of Food Engineering, 48: 147–156, 2001.
52. Ghani A.G.A., Farid M.M., Chen X.D., Numerical simulation of transient temperature and velocity
profiles in a horizontal can during sterilization using computational fluid dynamics, Journal of Food
Engineering, 51: 77–83, 2002.
53. Ghani A.G.A., Farid M.M., Zarrouk S.J., The effect of can rotation on sterilization of liquid food using
computational fluid dynamics, Journal of Food Engineering, 57: 9–16, 2003.
54. Kumar V., Wee A.P., Birla S., Subbiah J., Thippareddi H., A 3-D computational fluid dynamics model
for forced air cooling of eggs placed, Journal of Food Engineering, 108: 480–492, 2012.
55. Shinde Y.H., Gudekar A.S., Chavan P.V., Pandit A.B., Joshi J.B., Design and development of energy
efficient continuous cooking system, Journal of Food Engineering, 168: 231–239, 2016.
56. Aszodi A., Krepper E., Presser H.M., Investigation of heating up and evaporation processes of fluids
in storage tanks by experiments and by numerical simulation, In: Proceedings of the 2nd International
Symposium on Two-Phase Flow Modeling and Experimentation, 3: 1667–1674, 1999.
57. Kang M., Thermal mixing in a water tank during heating process, International Journal of Heat and
Mass Transfer, 45: 4361–4366, 2002.
58. Krepper E., Hicken E.F., Jaegers H., Investigation of natural convection in large pools, International
Journal of Heat and Fluid Flow, 23(3): 359–365, 2002.
59. Calcagni B., Marshili F., Paroncini M., Natural convective heat transfer in square enclosures heated
from below, Applied Thermal Engineering, 25: 2552–2531, 2005.
60. Sharma A.K., Velusamy K., Balaji C., Turbulent natural convection in an enclosure with localized heat-
ing from below, International Journal of Heat and Mass Transfer, 46: 1232–1241, 2007.
61. Ganguli A.A., Sathe M.J., Pandit A.B., Joshi J.B., Vijayan P.K., Hydrodynamics and heat transfer char-
acteristics of passive decay heat removal systems: CFD simulations and experimental measurements,
Chemical Engineering Science, 65: 3457–3473, 2010.
18 CFD Analysis of Thermal
Processing of Intact Eggs
Behzad Abbasnezhad, Mohsen Dalvi-Isfahan,
and Nasser Hamdami
CONTENTS
18.1 Introduction............................................................................................................................ 389
18.2 Thermophysical and Rheological Properties of Intact Eggs During Thermal
Processing: Measurement and Modeling............................................................................. 391
18.2.1 Density Model Development.................................................................................. 391
18.2.2 Effective Thermal Conductivity............................................................................. 393
18.2.3 Surface Heat Transfer Coefficients............................................................................ 396
18.2.4 Effect of Temperature on Flow Behavior............................................................... 397
18.3 Experimental Methodology..................................................................................................... 401
18.4 CFD Model Development....................................................................................................402
18.4.1 Computational Grid and Geometry Construction...................................................... 402
18.4.2 Governing Equations and Boundary Conditions...................................................... 403
18.4.3 Model Parameters....................................................................................................... 404
18.4.4 Assumptions Used in the Simulation......................................................................404
18.4.5 Pasteurizing Value (F) Calculation.........................................................................405
18.4.6 Numerical Solution of the Model............................................................................405
18.4.7 Model Validation................................................................................................................405
18.5 Application of CFD Model..................................................................................................406
18.5.1 Effect of Air Cell on Slowest Heating Zone (SHZ) and Flow Patterns..................406
18.5.2 Effect of Air Cell Volume on Pasteurization Value.................................................406
18.5.3 Effect of Shell Thickness on Pasteurization Value.................................................. 407
18.5.4 CFD Simulation of the Whole Egg in Rotational Method....................................... 409
18.5.5 Optimization of Intact Egg’s Thermal Pasteurization............................................ 410
18.6 Conclusions............................................................................................................................ 412
Nomenclature........................................................................................................................................412
References....................................................................................................................................... 413
18.1 INTRODUCTION
This chapter concentrates on recent advances in the application of Computational Fluid Dynamics
(CFD) for assessing batch pasteurization processes for intact eggs in stationary and rotational mode.
Although CFDs are now considered to be standard numerical tools and widely used within the
industry for process design and process analysis, they are not yet fully integrated into the egg-
processing sector, which could profit from the flexibility offered by CFD [1–5]. Eggs are widely used
as a highly nutritive food. Although they are available in different forms, such as frozen, powdered,
and liquid, consumers tend to buy intact eggs. Food products containing egg have a high risk of
microbial spoilage due to being contaminated by different microorganisms, especially Salmonella
enteritidis and Escherichia coli. Since the 1980s, an increase in reported cases of egg-associated
salmonellosis has been noted both in Europe and the United States [6–10]. Previous studies have
389
demonstrated that S. enteritidis in raw eggs can be adequately killed by thermal pasteurization.
Thermal pasteurization of liquid egg products has been in practice since the 1960s. Pasteurization
has been an important area of research for the past several years; it is a process intended to reduce
the microbial growth in foods and thus extend its shelf life. Salmonella enteritidis is the most heat
resistant spoilage factor in egg (D60 = 0.17min, zvalue = 4.08°C). The American Food and Drug
Association has recommended a 5D thermal processing for eggs based on Salmonella enteritidis
[8,10–13].
Thermal energy is normally used for pasteurization since the number of pathogenic micro-
organisms present is drastically reduced by applying heat through hot water or steam. The
thermal pasteurization of individual components of egg, such as egg white and yolk is carried
out routinely using heat exchanger device. Industrial processes of pasteurization have to ensure
the prolongation of food shelf life while preserving the quality of the product. The accom-
plishment of both purposes depends on process conditions with the assurance of the adequate
temperature course during processing, where consideration of the temperature profiles within
the product has great importance [14]. During thermal processing, the temperature inside the
food depends on time as well as on the position inside the food system. Therefore, there is an
urgent need to develop a model for predicting the temperature distribution during pasteuriza-
tion of intact eggs.
The design of the thermal process is always based on the temperature course in that position
in food that receives the least intense heat treatment, known as the slowest heating point (SHP)
or the slowest heating zone (SHZ). The slowest heating zone (SHZ) is defined as the location that
receives minimum heating and can be tracked from temperature distribution predictions. During
thermal pasteurization of eggs, heat transfer occurs through conduction as well as convection.
Heat transfer mechanisms in foods are conduction for solid foods; natural convection, especially
for low viscosity liquid foods; natural convection plus conduction for liquid foods with solid
particles; and natural convection followed by conduction for liquid foods containing starch or
high viscosity modifiers. Temperature profiles are determined using analytical and numerical
solutions of partial differential equations governing the process. For realistic and more compli-
cated heat transfer problems, usually no analytic solution is available, and a numerical solution
becomes mandatory [12,15–17].
In recent years, a rapid development in the application of CFD in food processing operations has
been witnessed. CFD is a numerical technique used for the solution of the equations governing fluid
flow and heat transfer problems inside a defined flow geometry. CFD works by dividing the physical
environment of interest into a two- or three-dimensional (3-D) grid or mesh. It contains a number
of discrete cells and can evaluate fluid velocities, temperature, and pressure inside every one of the
cells where fluid flows. This is done by the simultaneous solution of the equations describing fluid
flow, heat, and mass transfer.
CFD proved a powerful design and investigative tool to the food processor, especially in
the field of thermal pasteurization and sterilization of canned food and retort pouch products.
However, at present, little use of this technology has been reported in the thermal processing
of eggs.
Denys et al. studied the heat transfer and velocity profiles of egg shells filled with viscous liquids
(CMC) with thermophysical properties similar to those of egg whites. The slowest heating point
(SHP) was reported to be about the geometrical center of the samples in the conduction mode of
heat transfer. Furthermore, results showed that natural convection tends to force the slowest heating
zone towards the bottom of the egg, and when a conductive heating yolk is present, the cold spot
was found in the yolk at a location near its geometrical center [18]. Ramachandaran developed a
three-dimensional model to simulate the thermal pasteurization of eggs in stationary and rotational
conditions of heat treatment. The results showed that rotation of eggs highly decreased the time
needed for thermal treatment [17]. Fabbri et al. studied thermal pasteurization of eggs in hot air
currents using CFD modeling. They took into account the air cell of eggs in the model. The results
CFD Analysis of Thermal Processing of Intact Eggs 391
showed that air cells have heat resistant characteristics [16]. However, to the best of our knowledge,
there has been no published study available on modeling eggs pasteurization considering all parts
of an egg, such as white, yolk and air cell.
The aim of this chapter is to develop a mathematical model using computational fluid dynam-
ics (CFD) for intact egg pasteurization to allow prediction of temperature profile, SHZ loca-
tion and pasteurization efficiency. Also, given that the heat penetration into the slowest heating
zone in the egg is controlled by thermophysical properties of the egg itself, knowledge of these
parameters is essential for designing the model. Hence, density, thermal conductivity, surface
heat transfer coefficient, and flow behavior as major thermophysical properties are measured and
modeled in this chapter.
18.2 THERMOPHYSICAL AND RHEOLOGICAL PROPERTIES OF INTACT EGGS

DURING THERMAL PROCESSING: MEASUREMENT AND MODELING
Thermophysical properties of food products depend primarily on composition and temperature and,
in the next stage, on porosity and cell orientation. Table 18.1 shows the chemical analysis of intact
eggs [12,13,15,19–22].
18.2.1 Density Model Development

The simplest and most appropriate method to measure the density (ρ) of material is the use of a
pycnometer.
The principle of this method is as follows: The volume of the pycnometer (Vpy), at a particular
temperature (T°C), is calculated using the following equation:
Ww
Vpy = (18.1)
ρw
where Vpy is the volume of the pycnometer at a given temperature (m3), and Ww and ρw are the weight
of water that fills the pycnometer (kg) and the density of water (kg/m3) at the above temperature,
respectively.
Density values of water, ρw (kg/m3), at desired temperatures, T (°C), were taken from the data
presented by Abbasnezhad et al. [6]. Then density of samples (ρs) at measured temperature was
calculated using the following equation:
Ws
ρs = (18.2)
Vpy

TABLE 18.1
Average Amount of Chemical Components of Egg White and Yolk
Water Ash Protein Fat Carbohydrate
(%) (%) (%) (%) (%)
White 86.5 0.055 10.25 0.05 0.975
Yolk 51.5 1.5 16 26 1.075
Source: Abbasnezhad, B. et al. J. Food Meas. Charact. 8(4), 259, 2014.

where Ws is the weight of the sample that fills the pycnometer at the above temperature (kg), and ρs
is the density of the sample at the above temperature (kg/m3).
The density values of yolks and egg whites at temperatures above their initial freezing points can
be obtained from the following equation:
1
ρ= (18.3)

∑ xi
ρi
The parameters xi and ρi are the mass fraction and density value of the components, and the latter
is obtained from the following equations, in which subscript i is fat ( f ), protein (p), water (w), ash
(a), and carbohydrate (c).
ρc = 1.5991 × 10 3 − 0.31046 × T (18.4)
ρa = 2.4238 × 10 3 − 0.28063 × T (18.5)
ρ f = 9.2559 × 10 3 − 0.41757 × T (18.6)

ρ p = 1.3299 × 10 3 − 0.51480 × T (18.7)

ρw = 997.18 + 3.1439 × 10 −3 T − 3.7574 × 10 −3 T 2 (18.8)
The density of both groups decreased with increasing temperature. Density (ρ) data were cor-
related by the Equations 18.4 to 18.8 (Figure 18.1). Fitted functions presented r2 values higher than
0.99 and RMSE lower than 0.0012 [6].
FIGURE 18.1 Density of egg whites and yolks as a function of temperature. (From Abbasnezhad, B. et al.
J. Food Meas. Charact. 8(4), 259, 2014.)
18.2.2 Effective Thermal Conductivity

There are several methods to measure thermal conductivity of food materials and they can be
categorized in three general methods: steady state, quasi-steady state and transient methods
(Figure 18.2). But among them, transient methods are more common due to the short time neces-
sary and high accuracy. The most commonly used transient method is a line heat source probe
(Figure 18.3).
FIGURE 18.2 Thermal conductivity measurement. (From Dalvi-Isfahan, M. Thermophysical measurement

and Heat transfer modeling of Iranian Ultrafiltrated white cheese. MS thesis. In Food Science and Technology,
Isfahan University of Technology: Iran, 2008.)
FIGURE 18.3 Set-up of the thermal conductivity measurement system. (From Abbasnezhad, B. et al.
J. Food Meas. Charact. 8(4), 259, 2014.)
The thermal conductivity measurement system has been described in detail by Hamdami
et al. [24]. In brief, thermal conductivity of the egg can be calculated by using the following
equation:
λ =Q (18.9)
4π S
where Q is calculated by:
Q = RI 2 (18.10)
where λ is the thermal conductivity (W m−1 °C−1), Q is the power generated by the probe heater
(W m−1), S is the slope of the linear portion of the plot of the temperature versus ln (time), R is
the perfect gas constant, 8.314510 J mol−1 °C−1 and I is the current through the heater wire (A).
To obtain satisfactory linearity of temperature versus ln (time) plot, the procedure was standard-
ized by (1) choosing a power level to increase the temperature up to 10°C (initial temperature
basis), (2) using a duration of 10 s, and (3) accepting thermal conductivity values, measured only
when r2 > 0.98. To obtain these conditions, a current range of 0.14−0.24 A was necessary. Thus,
the chosen power levels varied between 1.2 W.m−1 to 5.4 W.m−1 and were sufficiently weak so as
not to disturb the effective value of thermal conductivity of material. Seven measurements were
done for each sample. Reported thermal conductivity values for yolks and whites are the mean
thermal conductivity values. Experiments were performed at different temperatures (20, 30, 40,
50°C). Furthermore, calibration of the probe was realized by measuring the thermal conductiv-
ity of glycerin and of a gel of 0.5% agar and water at selected temperatures. This calibration
permitted determining the heater wire resistance (R = 185.2 Ωm−1) used in the heater power
calculation.
Thermal Conductivity Model Development

Thermal conductivity can be analyzed by physical models based on chemical composition (pro-
teins, fats, carbohydrates, ash and water content). Modified Krischer’s and parallel models were pro-
posed for egg white and yolk, respectively. Krischer’s model recognizes that there are two extremes
in thermal conductivity values: one derived from the parallel model and the other derived from the
series model, whilst the real value of thermal conductivity should be somewhere in between these
two extremes. A conceptual diagram is shown in Figure 18.4. The distribution factor fλ is a weighing
factor between these extremes.
FIGURE 18.4 Krischer model development.

Egg can be analyzed as a multicomponent system formed from a protein (p), water (w), fat (f),
ash (a), and carbohydrate (c), Krischer, series and parallel models are described by the following
equation:
1
λeff (T ) = (18.11)
1− f λ fλ
+
λ pa (T ) λse (T )

λ pa (T ) = ε f (T )λ f (T ) + ε p (T )λ p (T ) + ε w (T )λw (T ) + ε a (T )λa (T ) + ε c (T )λc (T ) (18.12)

1
λse (T ) = (18.13)
ε a (T ) ε p (T ) ε w (T ) ε f (T ) ε c (T )
+ + + +
λa (T ) λ p (T ) λw (T ) λ f (T ) λc (T )

where λ eff, λpa, and λse are the effective thermal conductivity by the Krischer, Parallel, and Series
models, respectively. λf, λp, λw, λ a and λ c are the thermal conductivities of fat, protein, water, ash and
carbohydrate, respectively, obtained from the following equations:
λw (T ) = 0.57109 + 1.7625 * 10 −3 T − 6.703610 * 10 −6 T 2 (18.14)
λc (T ) = 0.20141 + 1.3874 * 10 −3 T − 4.33 * 10 −6 T 2 (18.15)
λ f (T ) = 0.18071 + 2.7604 * 10 −4 T − 1.7749 * 10 −6 T 2 (18.16)

λ f (T ) = 0.17881 + 1.1958 * 10 −3 T − 2.7178 * 10 −6 T 2 (18.17)

λa (T ) = 0.32961 + 1.4011 * 10 −3 T − 2.9069 * 10 −6 T 2 (18.18)
εi as a function of temperature is obtained from the following equation:
ρappx (T )
εi = i
(18.19)
ρi
where xi and ρi are the mass fraction (kg/kg product) and density (kg/m3) of the components, respec-
tively [6,24,25].
Figure 18.5 shows the effective thermal conductivity values, measured for egg white and yolk,
in the temperature range of 20−50°C. The figure shows that the effective thermal conductivity is
dependent on the chemical components (especially water content) and temperature. It increases
with increasing water content due to the higher thermal conductivity of water, above the freezing
point than other egg components. As can be seen, the effective thermal conductivity increases with
increasing the temperature due to the increase of the thermal conductivity of the components [26].
0.7
(a) 0.5 (b)
Thermal Conductivity (W/M.°C)
Thermal Conductivity (W/M.°C)

0.65
0.45
0.6 0.4
0.55 0.35
0.5 0.3
Seri model
Parallel model Seri model
0.45 0.25
Measurement Parallel model
Krischer's model Measurement
0.4 0.2
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Temperature (°C) Temperature (°C)
FIGURE 18.5 Experimental and predicted thermal conductivity values by series and parallel and Krischer
model (a) whites (b) yolks. (From Abbasnezhad, B. et al. J. Food Meas. Charact. 8(4), 259, 2014.)
Effective Thermal Conductivity Models

To clarify, Figure 18.5 presents the comparison between the experimental and predicted effective
thermal conductivity values of the egg yolk and white by the series, parallel, and Krischer models,
respectively, as a function of temperature and chemical composition.
As can be seen, the predicted thermal conductivity values of egg white, using Krischer model
(fλ = 0.20) where RMSE = 0.0061 and r2 = 0.996, and yolk, using parallel model in which RMSE =
0.0032 and r2 = 0.992, were in agreement with the experimental values in temperature range of
20−50°C.
18.2.3 Surface Heat Transfer Coefficients

The surface heat transfer coefficient is defined by Newton’s law of cooling. This parameter is depen-
dent on the thermophysical properties of the fluid and solid, characteristics of the solid shape, char-
acteristics of fluid flow and the system. Several methods have been proposed for measuring or
estimating this parameter, most notably: Transient temperature measurement, mass-loss-rate and
psychrometry method, and heat flux sensors. The methods mentioned above relate to determination
of heat transfer coefficient by experiment. However, this parameter can be obtained by estimation
methods through empirical relationships. The methods for determining the heat transfer coefficient
based on the estimation methods are: dimensional analysis, boundary analysis and by using the
similarity of heat and momentum transfer. But among them, transient temperature measurement,
which is based on the lumped capacitance method, is more common in thermal processing and this
approach is investigated in this section [23].
A prolate spheroid of pure aluminum 6 × 4.5 cm, weighing 178.1 g, was used to estimate the
surface heat transfer coefficient during the warm-up with hot water at 50, 55, and 60°C. A 0.5 mm
diameter, 1.85 cm deep hole, was drilled toward the center of the aluminum block from the end.
The aluminum block was placed in the egg holder, and a 0.5 mm diameter K-type thermocouple
was inserted into the hole (Figure 18.6). The aluminum block and the holder were then immersed
into the hot water at 50, 55, and 60°C. The temperature values of the inner side of the block were
recorded in 1 s intervals for 180 s, using the same recording system for the thermal experiments.
The coefficients of the regression line of time versus ln (Ta − Tal) were calculated, and the slope (m)
was used in Equation 18.20 to find the surface heat transfer coefficient (h).
M alCpal
hm = (18.20)
Sm
FIGURE 18.6 Egg shaped aluminum used to determine surface heat transfer coefficient. (From Abbasnezhad, B.
et al. J. Food Meas. Charact. 8(4), 259, 2014.)
TABLE 18.2
Surface Heat Transfer Coefficients in Different Temperatures
Temperature (°C) 50°C 55°C 60°C
Surface heat transfer 1744 1982 2072
coefficients (W/ m2.°C)
This equation is as follows:
∂T
hm A(T , ±T∞ ) M Cp (18.21)
∂t
In this equation, Cp, Mal, S, and m are the heat capacity of aluminum, weight of aluminum, alu-
minum surface area, and slope of the ln (Ta − Tal) over time graph, respectively [15].
The surface heat transfer coefficient, measured by transient temperature method, is presented
in Table 18.2 (results are the mean values of 15 tests). The results showed increasing surface heat
transfer coefficients due to the temperature increase.
Denys et al. also reported a combined CFD and experimentally based approach for determin-
ing the surface heat transfer coefficient acting at the egg boundary. Their methodology comprises
(1) determining the contour of eggs from digital photographs, (2) generating a computational geom-
etry, and (3) predicting the transient temperature profiles in eggs filled with a conductive heating
medium (Agar gel). Finally, CFD profiles were compared with experimentally observed data, and
the surface heat transfer coefficient was obtained through an optimization procedure [18].
18.2.4 Effect of Temperature on Flow Behavior

Viscosity measurements were conducted at different temperatures, using concentric cylinder viscom-
eter. An Arrhenius-type equation was used to define the effect of temperature on the apparent viscosity:

η or K = A0 exp Ea( ) RT
(18.22)
where A0 is an empirical constant, Ea, the activation energy for flow (J/mol), R, the gas constant
(j/mol°C), and T the absolute temperature. The suitability of the fitted models was appraised by the
coefficient of determination (r2) as well as the root mean square error (RMSE).
( k pre − kexp )2
RMSE = (18.23)
n −1
where kpre and kexp refer to predicted and experimented values of apparent viscosity and the letter (n)
indicates the number of values in the mentioned temperature range.
Rheological Parameters
Non-Newtonian models have been developed by correlation of experimental values to specific fun-
damental equations. Power law and Herschel–Bulkley models were used to study flow behavior of
egg yolk and whites, respectively.
Power-law model:
σ = K γ n (18.24)
where K is the consistency coefficient (Pa.sn), and n is the flow behavior index (dimensionless).
When ln (apparent viscosity) was plotted against ln (shear rate), the apparent viscosity could be
estimated by Equation (18.24):
ln ηa = a + lnγ (18.25)

which can be rewritten as:
ηa = K γ b (18.26)
where ηa is the apparent viscosity (Pa.sn) and K = exp (a). Instead of b, we commonly use (n − 1) for
the exponent and write a result for the apparent viscosity as follows:
ηa = K γ n−1 (18.27)
Herschel–Bulkley:
σ = K γ n + σ 0 (18.28)
where σ 0 is the yield stress value (Pa). The apparent viscosity decreases with an increase in shear
rate after the initial stress has been overcome. The respective expressions for the apparent viscosity
as function of shear rate are [10,14,21,22,27–30]:
ηa = K γ n−1 + σ 0γ −1 (18.29)
The presence of a hysteresis area between the curves representing shear stress versus shear rate
data for increasing and decreasing shear rates indicates that the sample’s flow is time-dependent.
Figures 18.7 and 18.8 show the flow curves (shear stress versus shear rate) of liquid egg white and
yolk at different temperatures (25, 50, 55 and 60°C) for both increasing and decreasing shear rates
(forward and backward direction). It is evident from Tables 18.3 and 18.4 that the largest hysteresis
0.6 forward backward 0.6 forward backward

model forward model backward model forward model backward
0.5 0.5
25°C 50°C
Shear Stress (Pa)

Shear Stress (Pa)
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Shear Rate (S-1) Shear Rate (S-1 )
0.6 0.6
forward backward forward backward
model forward model backward model forward model backward
0.5 0.5
55°C 60°C
Shear Stress (Pa)

Shear Stress (Pa)
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Shear Rate (S-1) Shear Rate (S-1 )
FIGURE 18.7 Experimental data and Herschel–Bulkley model of (shear rate–shear stress) for whites t 25,
50, 55, and 60°C. (From Abbasnezhad, B. et al. J. Food Meas. Charact. 8(4), 259, 2014.)
16.0 forward
backward forward
14.0 4.5
model forward
Shear Stress (Pa)
4.0 backward
12.0 model backward
25°C 50°C
3.5
10.0 model
Shear Stress (Pa)
3.0 forward
8.0 2.5 model
backward
6.0 2.0
1.5
4.0
1.0
2.0
0.5
0.0 0.0
0 10 20 30 40 0 10 20 30 40
Shear Rate (S-1) Shear Rate (S-1)
4.5 forward
5.0 forward
4.0 backward
55°C backward 60°C
3.5 4.0
model forward
model forward
Shear Stress (Pa)
3.0
Shear Stress (Pa)
model backward 3.0

2.5 model backward
2.0
2.0
1.5
1.0
1.0
0.5
0.0 0.0
0 10 20 30 40 0 10 20 30 40
Shear Rate (S-1) Shear Rate (S-1)
FIGURE 18.8 Experimental data and power law model of (shear stress–shear rate) for yolks at 25, 50, 55,
and 60°C. (From Abbasnezhad, B. et al. J. Food Meas. Charact. 8(4), 259, 2014.)
TABLE 18.3
Herschel–Bulkley Model Parameters and Hysteresis Area of Egg Whites at Different
Temperatures (Shear Stress vs. Shear Rate)
Forward Measurements Backward Measurements
Hysteresis Yield
T Area Stress K K
(°C) (Pa·s) (pa) (Pa·sn) n r2 RMSE (Pa·sn) n r2 RMSE
25 11.92 0.0027 0.038 0.64 0.996 0.0241 0.014 0.87 0.999 0.0117
50 4.39 0.0041 0.039 0.55 0.996 0.0204 0.022 0.68 0.999 0.04481
55 3.67 0.0044 0.072 0.47 0.993 0.0035 0.046 0.55 0.998 0.01604
60 3.01 0.0054 0.079 0.4 0.998 0.0119 0.055 0.46 0.992 0.01694
TABLE 18.4
Power Law Model Parameters and Hysteresis Area of Yolk at Different Temperatures
(Shear Stress vs. Shear Rate)
Forward Measurements Backward Measurements
Hysteresis K K
T (°C) area (Pa·s) (Pa·sn) n r2 RMSE (Pa·sn) n r2 RMSE
25 3.775 0.682 0.87 0.999 0.08720 0.62 0.89 0.9996 0.02568
50 0.721 0.184 0.89 0.993 0.03758 0.173 0.91 0.9993 0.03689
55 0.547 0.144 0.90 0.992 0.06824 0.146 0.915 0.9991 0.03708
60 0.458 0.13 0.91 0.999 0.04836 0.118 0.92 0.9989 0.03379
loop is shown at 25°C for both egg white and yolk, that hysteresis area decreased with increasing
temperature, and that this area is larger in egg white than it is in egg yolk; therefore, thixotropy in
egg white samples is more pronounced. The developed power-law model for yolks had satisfac-
tory correlation coefficients (r2) greater than 0.99 for all samples and RMSE lower than 0.025.
Rheological parameters of the yolks are presented in Table 18.4. The behavior index showed values
varying from 0.87 to 0.91 for forward measurement and 0.89 to 0.92 for backward measurement,
indicating that, in this range of temperature, the egg yolk is pseudoplastic. The consistency index
varied from 0.13 to 0.68 Pa.sn for forward measurement and 0.118 to 0.62 Pa.sn for backward.
A yield stress was observed in the case of egg whites. The egg white flow curves could be well
fitted by the Herschel–Bulkley model, at a range of 25–60°C. The rheological parameters of egg
whites are presented in Table 18.3. The behavior index showed values varying from 0.4 to 0.64 for
forward measurement and 0.46 to 0.87 for backward measurement, indicating that, in this range of
temperature, the egg white is pseudoplastic in nature. The consistency index varied from 0.038 to
0.079 Pa.sn for forward measurement and 0.14 to 0.055 Pa.sn for backward measurement.
In order to confirm these findings, apparent viscosity versus shear rate in different temperatures
was recorded and modeled. The power-law model (Figure 18.8) was tested for egg yolk and the
Herschel–Bulkley model (Figure 18.7) for egg white. Results are summarized in Table 18.4 for egg
yolk and Table 18.3 for egg white. Models could predict experimental data satisfactorily with cor-
relation coefficients (r2) greater than 0.99 for all samples and RMSE lower than 0.025.
The effect of temperature on the apparent viscosity at constant shear rate and different tempera-
ture is illustrated in Figure 18.9. An Arrhenius-type equation was used to quantify the effect of
temperature on apparent viscosity. A linear form of Equation (18.22) was fitted to experimental data
0.035 0.9
(a) 0.8
(b)
0.030
0.7 25 °C 50 °C
25 °C 50 °C
Apparent viscosity (Pa.sn)

55 °C 60 °C
Apparent viscosity (Pa.sn)
0.025 55 °C 60 °C 0.6
0.020 0.5
0.4
0.015
0.3
0.010
0.2
0.005
0.1
0.000 0.0
0 10 20 30 40 50 60 0 5 10 15 20 -1 25 30 35
Shear rate (s-1) Shear rate (s )
FIGURE 18.9 Apparent viscosity–shear rate diagram of (a) whites and (b) yolks at different temperatures.
(From Abbasnezhad, B. et al. J. Food Meas. Charact. 8(4), 259, 2014.)
TABLE 18.5
Activation Energy for Yolks and Whites
Ea (j/mol) r2 RMSE
Yolks 36496 0.997 0.0002
Whites 4432 0.998 0.0003
Source: Abbasnezhad, B. et al. J. Food Meas. Charact. 8(4),

259, 2014.
to find the values of the flow activation energies tested. Table 18.5 shows the estimated activation
energy values for each sample in specific shearing speeds (yolk: 9.18 s−1, white: 14.52 s−1). It was
observed that apparent viscosities of liquid egg yolk and white decreased with increasing tempera-
ture (Figure 18.9). Activation energy (Ea) value for egg yolk was more than that for egg white, which
means that egg yolk is more sensitive to temperature than egg white. Great care needs to be taken
when the thermal pasteurization process is applied [14,19−22,28].
18.3 EXPERIMENTAL METHODOLOGY
Thermal processes were simulated using intact fresh eggs. Eggs, not older than three days, were
bought from local stores and kept at 4°C until the time of experiments. The eggs were classified into
three groups: large (heavier than 60 g), medium (55−60 g) and small (less than 55 g). The volume
percentage of air cells studied in this paper ranged from zero to 7%. The volume percentage of the
yolk, white, and shell were split according to Table 18.6. The eggs were placed at 20°C for 2h before
pasteurization. Pasteurization was carried out in a circulating water bath (Memmert, Germany)
at 60°C for 35 min. The water bath used was equipped with an automatic temperature controller
(±0.1°C). K-type thermocouples (Omega, Stamford, USA) with 0.3 mm diameter (±0.1°C) were
placed in the egg and water bath to measure the temperature inside the egg and surrounding water.
Before processing, a calibrated thermocouple was inserted through a hole on the egg (thermocouple
tip positioned from the top surface of the eggshell to 40 mm inside the egg in a central axial verti-
cal position) and sealed with glue. Thermocouples were connected to a temperature data logger
(Datalog 20, AOIP, Evry, France) with 15 channels to carry out the measurements. The egg was
TABLE 18.6
Average Volume Percent of Egg Components
Yolk White Shell Air Cell
Volume percent 26.5 63 7.5 0−7
Source: Abbasnezhad, B. et al. Food Sci. & Nutr. 4(1), 42, 2016.
immersed in the water bath and the temperature data acquisition was started. The data were used to
validate the model. To determine the thermocouple probe position, after completion of the thermal
process, the temperature of the water bath was raised and the eggs were cut and exact location of
the probe determined [6,8,14,15,17].
18.4 CFD MODEL DEVELOPMENT

The basic steps involved in any CFD analysis consist of the following:
18.4.1 Computational Grid and Geometry Construction

The impact of variability in egg shape on the results of the CFD have been studied by various
researchers [31]. However, in this study, the effects of a variability of egg shapes were neglected.
The geometry of the egg was designed as a 3D oval. The length and the radius of the oval were 6 cm
and 4.5 cm, respectively. A sphere with radius of 1.6 cm was designed in the geometric center as the
yolk. The air cell was designed according to the measured sizes of the real ones above and at the
bottom of the shell as shown in Figure 18.10. After defining the physical geometry of the environ-
ment in which the fluid flows, the second step is to discretize the solution domain into finite element;
that is, subdivide the problem into subdomain (meshing). Mesh constitutes a critical step because
it may strongly affect the accuracy of the numerical solution. If the subdomains or cells have a
similar size, the mesh structure is called uniform and is non-uniform otherwise. Furthermore, the
fine mesh-grid (reduction of cell sizes) and mesh skewness are important parameters in the thermal
processing of eggs. Figure 18.10 shows the uniform trigonal type meshing. The total count of trigo-
nal elements was 288186 [15].
z
y x
FIGURE 18.10 Designed and meshed geometry. (From Abbasnezhad, B. et al. Food Sci. & Nutr. 4(1), 42, 2016.)
18.4.2 Governing Equations and Boundary Conditions

The partial differential equations (PDEs) governing natural convection motion in a three-dimensional
x, y, and z coordinate system are the Navier–Stokes equations and, are presented as follows:
Continuity equation:
∂ux ∂u y ∂uz
+ + = 0 (18.30)
∂x ∂y ∂z
Momentum equation:
 ∂ρV 
 + ∇.ρV ⊗ V  = ∇ ( − pδ + µ (∇V + (∇V )t ) ) + S M (18.31)
σt 

Energy equation:
∂T ∂T ∂T ∂T  ∂ 2T ∂ 2T ∂ 2T 
+ ux + uy + uz = ∝  2 + 2 + 2  (18.32)
∂t ∂x ∂y ∂z  ∂x ∂y ∂z 

where u (m/s) is the velocity, p (Pa) is pressure, ρ (kg/m3) is the density, μ (m2/s) is the kinematic
viscosity of the fluid, g (m/s2) is the gravitational acceleration acting in the negative z-direction, and
∝ (m2/s) is the thermal diffusivity. To consider buoyancy, the force driving the convective motion of
the egg white, a gravitational force was comprised in the equations, and the variation of the density
with temperature was expressed using the Boussinesq approximation that was very accurate to
model the natural convection during thermal processing of egg bodies.
S M = ρ g (18.33)

ρ = ρref 1− β (T − Tref )  (18.34)

where Tref (°C) and ρref (kg/m3) are the reference temperature and corresponding density, respec-
tively, and β (1/K) is the volumetric thermal expansion coefficient of the liquid [12,15,17]. The
dimensionless Rayleigh number, which measures the strength of buoyancy driven flows, was cal-
culated. The Rayliegh number was lower than 108 and showed laminar flow behavior during the
process.
Initial and Boundary Conditions:
Yolk and white interface,
T = Ts , u = 0 ν = 0 for 0 ≤ r ≤ R. (18.35)

Outside of eggshell,
∂T ∂u
= 0, = 0, ν = 0 for 0 ≤ z ≤ H . (18.36)
∂r ∂r
Inside of eggshell,
∂T
= 0, u = 0, ν = 0, for 0 ≤ r ≤ R. (18.37)
∂z
Ts is water bath temperature that (333 for heating part).

Initially the egg yolk and white were at rest and at a uniform temperature
T = Ti , u = 0, ν = 0 at 0 ≤ r ≤ R, 0 ≤ z ≤ H . (18.38)

The yolk wall was given as a coupled wall to the egg white for uniform heating.
18.4.3 Model Parameters
The inputs of the mathematical model are: the water bath temperature (constant), product dimen-
sions, density, component mass fractions, surface heat transfer coefficient of egg, initial product
temperature, number of nodes in the space, and heating and cooling time of pasteurization. Based
on these inputs, the model will determine the values of temperature at each node for each time step.
The thermophysical properties used in the model are shown in Table 18.7.
18.4.4 Assumptions Used in the Simulation

The egg is placed into the water bath and its surface, directly exposed to hot water, is heated by
natural convection. The heating of the surface creates a temperature gradient, which is the origin of
TABLE 18.7
Thermal Properties of Egg White, Yolk, and Egg Shell Used in the Modeling
Value r2 Source
Thermal conductivity Yolk k = 0.0008 T + 0.395
a 0.998 –
(W/m.°C) White k = 0.0013 T + 0.5125 0.991 –
Egg shell 2.25 –
Density Yolk ρ = −0.0023 T2 − 0.1386 T+ 1037.1 0.997 –
(kg/m3) White ρ = −0.0041 T2 − 0.0115 T+ 1043.3 0.995 –
Egg shell 2300 – [18]
Specific heat capacity Yolk 3120 – [32]
(J/ kg.°C) White 3800 – [32]
Egg shell 888 – [32]
Surface heat transfer – 2072b – –
coefficient (h) (W/m2.°C)
Viscosity (Pa.s) White η = −0.0006 T + 0.0438 0.989 –
Yolk η = −0.0118 T + 0.7915 0.996 –
Source: Abbasnezhad, B. et al. Food Sci. & Nutr. 4(1), 42, 2016.
a T represents temperature in (°C).
b Calculated from data measured at 60°C.

the driving force behind the heat transfer in the egg toward the center. Heat transfer inside the shell
can be treated as heat conduction.
In order to simplify the problem, the following assumptions were made:
a. Egg white and yolk are homogeneous and isotropic during pasteurization.
b. The initial temperature throughout the egg is constant and equal.
c. Thermophysical properties were used as functions of temperature, excluding the volumet-
ric thermal expansion coefficeint and specific heat.
d. Viscous dissipation was neglected because of low shear rates.
e. No-slip boundary condition was assumed for velocity components relative to boundaries.
f. The natural convectional flow of fluid was assumed in gravitational direction.
g. The moisture transfer was neglected during the process.
18.4.5 Pasteurizing Value (F) Calculation

The purpose of this calculation is to arrive at an appropriate process time under a given set of heat-
ing conditions to result in a given process lethality, or alternatively to estimate the process lethality
of a given process. In order to characterize the effect of temperature evolution on micro-organism
destruction at a given location during pasteurization, a so-called pasteurizing value (F) can be
calculated:
∫ 10
( T − Tref )/ z
F= dt (18.39)
0
where Tref is 60°C, and Z is 4°C (based on the thermal resistance of Salmonella enteritidis). A 5D
inactivation of the mentioned microorganism (Dvalue = 0.17 min) is assumed as a good thermal pro-
cess [7,10–12,15–17].
18.4.6 Numerical Solution of the Model

In this study, the resulting system of the above partial differential equations was solved by the
COMSOL software. The solution of the equations was obtained in the “Direct UMFPACK” mode
and time intervals were set with the help of the algorithm present in the software. The total process
time was 2000 seconds and time step was 1 second. Maximum relative error of 10 −3 was selected as
a criterion for convergence in the relative tolerance window [15].
The model was validated by comparing the predicted temperature histories with experimental
results. Figure 18.11 shows a typical graph of the comparison between the experimental and
predicted temperature evolutions for both heat transfer mechanisms supposed (conductive or
conductive–convective) in egg whites. Considering a natural convection-conduction mode in
egg whites and conduction mode in yolk, a better agreement is obtained (RMSE = 0.0023 and
r 2 = 0.991) than a conductive mode for both (RMSE = 1.256 and r 2 = 0.89). It can be concluded
that the assumptions applied to develop the model (conductive–convective and conductive heat
transfer in whites and yolk, respectively, and other assumptions) describe the heat transfer mech-
anisms well during egg pasteurization.
60
50
m erature (°C)
40
Temperature
30
Temp
20
10
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (s)
FIGURE 18.11 Comparison of egg temperature profiles between measurement and CFD predictions for
convection–conduction and conduction only heating. Positions of the thermocouple probes and different
mechanism for heat transfer are indicated. Exact location of thermocouple was (0.001, 0.006, 0.001). (From
Abbasnezhad, B. et al. Food Sci. & Nutr. 4(1), 42, 2016.)
18.5 APPLICATION OF CFD MODEL

18.5.1 Effect of Air Cell on Slowest Heating Zone (SHZ) and Flow Patterns
The developed model used to determine the SHZ of the egg during pasteurization in a water bath
(at 60°C) of the intact eggs without and with the air cell in three different positions is shown in
Figure 18.12. As can be seen, the air cell is able to influence the heat transfer during thermal pasteuriza-
tion of intact eggs. The SHZ of the eggs with an air cell located in the bottom of the shell was located
between the geometric center of the sample and upper zone of the air cell, as shown in Figure 18.12a.
Assuming the absence of an air cell, the SHZ would be located in the geometric center of the egg
(Figure 18.12b). The SHZ of the eggs with two air cells, on top and bottom, was in the geometric
center (Figure 18.12c). The air cells located at the top of the egg resulted in the displacement of the
SHZ towards the center of the yolk (Figure 18.12d).
Figure 18.13 shows that the fluid in egg is warmed and flows upward in the boundary layer near the
lateral eggshell. The warm liquid coming from the boundary layer almost stagnates at the top of the egg.
In fact, according to mass balance, it flows slowly inward and downward. The area with downward flow
is much greater than the area with upward flow, as can be seen on the horizontal cross-section. As the
liquid descends, its temperature decreases because of conduction and mixing with the lower and colder
liquid layers. Arriving near the bottom, the liquid moves again toward the lateral wall and begins a new
cycle. It can be seen that existence of air cells results in a flow pattern change of egg whites. The egg with
an air cell on the bottom has stronger flow movements than those with air cells on the top and bottom.
As shown in Figure 18.14, the natural convectional flows streams may decrease during the heat
treatment, due to the fact that at the initial stages of the process, the high temperature gradient of
hot water and the inner fluid reinforce the buoyancy driven flows.
18.5.2 Effect of Air Cell Volume on Pasteurization Value

The pasteurizing value can be calculated using the predicted temperatures at the slowest heating
point (SHP) of the egg. Figure 18.15 shows the impact of air cell volume of the egg on the pasteur-
izing value. The results of simulation studies on the required heating time to achieve Fvalue = 2.85
min, for three egg models with different air cell volumes (0, 3, and 7%) at 60°C demonstrated that
Max: 60.0 Max: 60.0

(a) 60 (b) 60
55 55
50 50
45 45
40 40
5s 10 s 50 s 5s 10 s 50 s
35 35
30 30
25 25
20 20
80 s 150 s 300 s Min: 17.0 Min: 17.0
Max: 60.0 80 s 150 s 300 s
(c) 60 (d) Max: 60.0
60
55
55
50
50
45
45
5s 10 s 50 s 40 40
5s 10 s 50 s
35 35
30
30
25 25
20 20
80 s 150 s 300 s Min: 17.0 80 s 150 s 300 s Min: 17.0
FIGURE 18.12 Temperature distribution of eggs having different air cell locations during pasteurization
process: (a) the air cells located at the bottom; (b) absence of air cell; (c) two air cells on top and bottom; and
(d) the air cells located at the top. (From Abbasnezhad, B. et al. Food Sci. & Nutr. 4(1), 42, 2016.)
FIGURE 18.13 White flow pattern of eggs having different positions of air cells at 50 s of the thermal pro-
cess at 60°C. (From Abbasnezhad, B. et al. Food Sci. & Nutr. 4(1), 42, 2016.)
the eggs with larger air cells need more processing time to obtain the same Fvalue. The required time
for 0, 3, and 7% air cell volume was 1680, 1770, and 1920 s, respectively. The SHP of the eggs with
medium air cells was between the geometric center of the sample and the upper zone of the air cell.
The SHP of the eggs with the minimum percentage of air cell, the SHP located in the geometric
center zone, and the SHP of the eggs with maximum air cell were in the upper zone of the air cell.
18.5.3 Effect of Shell Thickness on Pasteurization Value

The heating times to reach an equal pasteurization value (F60°C = 2.87 min) with different egg shell thick-
nesses were simulated by Abbasnezhad et al. For thicknesses of 0.175, 0.4, 0.25 mm, the required times
were 1670, 1695 and 1710 s, respectively. This shows that an increase in shell thickness leads to a reduction
of heat transfer and consequently an increase of required heating time to achieve a satisfactory F value.
FIGURE 18.14 Velocity contours of models that having different air cell positions at 5, 10, and 80 s of the
thermal process. (From Abbasnezhad, B. et al. Food Sci. & Nutr. 4(1), 42, 2016.)
FIGURE 18.15 Effect of different air cell sizes on pasteurization efficiency. (From Abbasnezhad, B. et al.
Food Sci. & Nutr. 4(1), 42, 2016.)
18.5.4 CFD Simulation of the Whole Egg in Rotational Method

The static egg simulation was extended for rotation of the egg at two different speeds (2.5 and
5 rpm) in the conduction mode of heating. Figure 18.16(a–f) gives a comparison between the veloc-
ity vector plots of a stationary (Figure 18.16a–c) and rotating (Figure 18.16d–f) egg during the
heating process at the time intervals of 30 s, 150 s and 300 s. The velocity magnitude of the egg in
stationary processing was low (maximum velocity was 5 × 10−4 m/s) compared to the rotational
(maximum velocity was 2.5 × 10−2 m/s). Moreover, the circumstantial component also increased
(Figure 18.16d–f) in the rotational modes. The egg white moves upwards towards the egg shell and
is reflected downwards around the yolk surface (Figure 18.16a–c). In contrast, during rotation of the
egg (Figure 18.16d–f), the flow of egg white moved in a curved path due to the Coriolis acceleration.
In the rotating egg case, added acceleration to the rotating coordinate system and centripetal
acceleration produces an egg white in acceleration relative to a fixed coordinate system. The rota-
tional process ensured the mixing of egg white that led to heat penetration throughout the egg white,
which reached a uniform temperature within 7 min.
The thermal pasteurization process at various process times of 30 s, 150 s, and 300 s is shown in
Figure 18.17(a–f). It gives the contours and shows that the egg white reaches the set pasteurization
temperature much faster in the rotating egg due to heat penetration throughout the egg. However, the
yolk portion took more time to heat up than the egg white in the rotating position (Figure 18.17d–f),
but less time than the stationary egg (Figure 18.17c). The SHZ was concentrated at the center of
the yolk throughout the rotating process and maintained a circular shape (Figure 18.17d,e), unlike
the SHZ in the stationary egg (Figure 18.17b). Thus, a rotating egg achieves an almost uniform
temperature in approximately 300 s (Figure 18.17f), while a stationary egg still possesses a SHZ at
FIGURE 18.16 Velocity vector plots of a stationary (a–c) and 5 rpm rotating (d–f) egg heated at 55.6°C.
(From Ramachandran, R. et al. Innov. Food Sci. & Emerge. Technol. 12(1), 38, 2011.)
FIGURE 18.17 Comparison of CFD simulated temperature contours of stationary (a–c) and 5 rpm rotating
(d–f) egg heated at 55.6°C. (From Ramachandran, R. et al. Innov. Food Sci. & Emerge. Technol. 12(1), 38, 2011.)
a temperature of 37°C at the end of 300 s processing time (Figure 18.17c). Thus, rotation improves
the efficiency of a heating process by ensuring more uniform heating compared to static heating.
During the entire thermal pasteurization process, the temperature at two different points inside the
egg (from center to bottom 0.018 m and center) was extracted and plotted together with stationary
and rotation modes in Figure 18.18(a,b). It can be observed from Figure 18.18(a,b) that the pasteuriza-
tion temperature was reached much faster in rotation modes than in the stationary egg heating. The
egg white (point 1—Figure 18.18a) reached maximum temperature in 6.5 min at 5 rpm and 7 min at
2.5 rpm, compared to the stationary egg, which took 30 min. The point inside the yolk (Figure 18.7b)
took a longer time to reach the pasteurization temperature compared to the point on the egg white
(Figure 18.18a). The yolk point (Figure 18.18b) indicated that at 5 rpm the rotated egg took 9 min
to reach maximum pasteurization temperature, whereas when rotated at 2.5 rpm the egg took about
13 min. However, in a stationary position the yolk needs approximately 30 min to reach pasteurization
temperature (Figure 18.18b). Thus, the pasteurized egg can be obtained in much less time if it is rotated.
18.5.5 Optimization of Intact Egg’s Thermal Pasteurization

The effectiveness of a heating process like pasteurization can be estimated by the F values (Equation
18.39). The F value (min) can be used to quantify the survival of targeted microorganism; in this
study inactivation of S. Enteritidis was estimated using (Equation 18.39). The F value was calcu-
lated for stationary and rotating modes for every one-minute interval of the thermal process. The
F value results obtained at the top point (from center to top 0.025 m) and center points are plotted
in Figure 18.19(a) and (b) respectively. Figure 18.18(a) and (b) show that the stationary position
egg attained maximum F value (highest inactivation of Salmonella) in 30 min of heating (Figure
18.19a), whereas the rotated egg attained it within 7 min of heating for 5 rpm and 13 min for
FIGURE 18.18 Comparison of CFD simulated temperature profiles of stationary and rotated egg conduc-
tively heated at 55.6°C: (a) point 1—bottom point (from center to bottom 0.018 m) and (b) point 2—center
point. (From Ramachandran, R. et al. Innov. Food Sci. & Emerge. Technol. 12(1), 38, 2011.)
FIGURE 18.19 F value predictions for stationary and rotated egg conductively heated at 55.6°C: (a) top
point (0.025 m from center) and (b) center point. (From Ramachandran, R. et al. Innov. Food Sci. & Emerg.
Technol. 12(1), 38, 2011.)
2.5 rpm rotation. However, inactivation of S. Enteritidis in the yolk (Figure 18.19b) took slightly
more time. Denys et al. (2005) also observed inactivation of Salmonella was slowest in the yolk.
Results indicated that the rotation of the egg reduced the processing time.
18.6 CONCLUSIONS
In this chapter, recent advances in the application of CFD in analyzing, assessing, and designing
batch pasteurization processes for intact eggs were illustrated. The general conclusion is that like in
other food engineering areas, CFD can be a valuable and promising tool for analysis, design, and
assessment of batch processes aimed at the pasteurization of intact eggs. A numerical model was
developed to simulate 3D heat transfer in intact egg to predict the local temperature and F value
during pasteurization. The model accommodates the effects of air cell volume and temperature-
dependent variables such as density, thermal conductivity, rheology and surface heat transfer coef-
ficient. The model was validated by comparison of the experimental temperature profiles during
pasteurization of eggs with the predicted values. The main results that can be drawn from this study
are the following: (i) the position and size of air cells affect the heat transfer patterns and charac-
teristics during egg pasteurization; (ii) the thickness of eggshell affects the required pasteurization
time; (iii) the effect of eggshell thickness on the required pasteurization time was less than air cell
volume; and (iv) the model is useful to describe the heat transfer phenomenon during egg pasteuri-
zation and thermal pasteurization under rotated egg reduced the time of pasteurization.
NOMENCLATURE
Vpy Volume of the pycnometer (m3)
Ww Weight of water that fills the pycnometer (kg)
ρw Density of water (kg m–3)
Ws Weight of sample that fills the pycnometer (kg)
ρ The density (kg m–3)
x Mass fraction
λ Thermal conductivity (W m−1 °C−1),
Q Power generated by probe heater (W m−1)
R Perfect gas constant (8.314510 J mol−1 °C−1)
I Current through heater wire (A)
f λ Weighing factor
λeff The effective thermal conductivity
λpa The effective thermal conductivity parallel model
λse The effective thermal conductivity by series model
Cp, Heat capacity of aluminum (J kg−1 °C−1)
Mal Weight of aluminum (kg)
Ea The activation energy for flow (J mol−1)
K The consistency coefficient (Pa sn)
n The flow behavior index
ηa Apparent viscosity (Pa sn)
σ Shear stress (Pa)
γ Shear rate (s−1)
σ 0 Yield stress value (Pa)
u The velocity (m s−1)
p Pressure (Pa)
μ The kinematic viscosity of the fluid (m2 s−1)
g The gravitational acceleration (m s−2)
∝ The thermal diffusivity (m2 s−1)
β The volumetric thermal expansion coefficient (K−1)

Tref The reference temperature (°C)
ρref The reference density (kg m–3)
h Surface heat transfer coefficient (W m−2 °C−1)
Fvalue Pasteurizing value (min)
REFERENCES
1. Juliano P., Knoerzer K., Versteeg C., Computational fluid dynamics applied in high-pressure high-
temperature processes: Spore Inactivation distribution and process optimization, In: Innovative Food
Processing Technologies: Advances in Multiphysics Simulation, Blackwell Publishing Ltd., Oxford,
UK, 75–100, 2011.
2. Juliano P., Trujillo F.J., Barbosa-Cánovas G.V., Knoerzer K., The need for thermophysical properties
in simulating emerging food processing technologies, In: Innovative Food Processing Technologies:
Advances in Multiphysics Simulation, Blackwell Publishing Ltd., Oxford, UK, 23–38, 2011.
3. Fryer P.J., Porras-Parral G., Bakalis S., Multiphysics modeling of ohmic heating, In: Innovative Food
Processing Technologies: Advances in Multiphysics Simulation, Blackwell Publishing Ltd., Oxford,
UK, 155–169, 2011.
4. Barbosa-Cánovas G.V., Albaali A.G., Juliano P., Knoerzer K., Introduction to innovative food process-
ing technologies: Background, advantages, issues, and need for multiphysics modeling, In: Innovative
Food Processing Technologies: Advances in Multiphysics Simulation, Blackwell Publishing Ltd.,
Oxford, UK, 3–21, 2011.
large particles in a cylindrical vertical can, Journal of Food Engineering, 82(4): 409–417, 2007.
6. Abbasnezhad B., Hamdami N., Shahedi M., Vatankhah H., Thermophysical and rheological properties
of liquid egg white and yolk during thermal pasteurization of intact eggs, Journal of Food Measurement
and Characterization, 8(4): 259–269, 2014.
7. Hema A.L.A., Effect of different pasteurization methods of egg in shell on survival of Salmonella
enteritidis, In: Foreign Agricultural Relations (FAR): Egypt, 107–120, 2011.
8. Hou H., Singh R.K., Muriana P.M., Stadelman W.J., Pasteurization of intact shell eggs, Food
Microbiology, 13(2): 93–101, 1996.
9. Stadelman W.J., Cotterill O.J., Egg Science and Technology, 4th ed., The Haworth Press, Inc., New
York, USA, 1995.
10. Shenga E., Singh R.P., Yadav A.S., Effect of pasteurization of shell egg on its quality characteristics
under ambient storage, Journal of Food Science and Technology, 47(4): 420–425, 2010.
11. American Egg Board, Homepage service: http://www.aeb.com. 2012.
12. Koen D., Siegfried D., Jan P., CFD Analysis of thermal processing of eggs, In: Computational Fluid
Dynamics in Food Processing, CRC Press, Boca Raton, FL, USA, 347–380, 2007.
13. Awuah G.B., Ramaswany H.S., Economides, Thermal processing and quality, Journal of Chemical
Engineering and Processing, 46: 587–602, 2007.
14. Abbasnezhad B., Hamdami N., Khodaei D., Modeling of rheological characteristics of liquid egg white
and yolk at different pasteurization temperatures, Journal of Food Measurement and Characterization,
9(3): 359–368, 2015.
15. Abbasnezhad B., Hamdami N., Monteau J.-Y., Vatankhah, H., Numerical modeling of heat transfer and
pasteurizing value during thermal processing of intact egg, Food Science & Nutrition, 4(1): 42–49,
2016.
16. Fabbri A., Cevoli C., Giunchi A., Validation of a simplified numerical model for hot-air treatment of egg
shell surface, Journal of Food Process Engineering, 35(5): 695–700, 2012.
17. Ramachandran R., Malhotra D., Anishaparvin A., Anandharamakrishnan C., Computational fluid
dynamics simulation studies on pasteurization of egg in stationary and rotation modes, Innovative
Food Science & Emerging Technologies, 12(1): 38–44, 2011.
943–951, 2003.
19. Polachini T.C., Basso R.C., Telis-Romero J., Rheology and fluid dynamic of egg white: Effect of thixot-
ropy on engineering design, Journal of Food Process Engineering, 40(1): e12277, 2017.
20. Mohtashami Y., Hagness S.C., Behdad N., A hybrid slot/monopole antenna with directional heating
patterns for microwave ablation, IEEE Transactions on Antennas and Propagation, 65(8): 3889–3896,
2017.
21. Bozkurt H., Icier F., The change of apparent viscosity of liquid whole egg during ohmic and conven-
tional heating, Journal of Food Process Engineering, 35(1): 120–133, 2012.
22. Atılgan M.R., Unluturk S., Rheological properties of liquid egg products (LEPS), International Journal
of Food Properties, 11(2): 296–309, 2008.
23. Dalvi-Isfahan M., Thermophysical measurement and heat transfer modeling of Iranian ultrafiltrated
white cheese, MS thesis, Isfahan University of Technology: Iran, 2008.
24. Hamdami N., Monteau J.-Y., Le Bail A., Effective thermal conductivity of a high porosity model food at
above and sub-freezing temperatures, International Journal of Refrigeration, 26(7): 809–816, 2003.
25. Baghe-Khandan M.S., Choi Y., Okos M.R., Improved line heat source thermal conductivity probe,
Journal of Food Science, 46(5): 1430–1432, 1981.
26. Dalvi M., Hamdami N., Characterization of thermophysical properties of Iranian ultrafiltrated white
cheese: Measurement and Modeling, Journal of Agricultural Science and Technology, 13(1): 67–78,
2010.
27. Severa L., Nedomová Š., Buchar J., Influence of storing time and temperature on the viscosity of an egg
yolk, Journal of Food Engineering, 96(2): 266–269, 2010.
28. Elena C.-P., Non-newtonian models, In: Encyclopedia of Agricultural, Food, and Biological
Engineering, 2nd ed., CRC Press, Boca Raton, FL, USA, 1134–1137, 2010.
29. Telis-Romero J., Thomaz C.E.P., Bernardi M., Telis V.R.N., Gabas A.L., Rheological properties and
fluid dynamics of egg yolk, Journal of Food Engineering, 74(2): 191–197, 2006.
30. KuŞ S., Altan A., Kaya A., Rheological behavior and time-dependent characterization of ice cream mix
with different salep content, Journal of Texture Studies, 36(3): 273–288, 2005.
31. Narushin V.G., The avian egg: Geometrical description and calculation of parameters, Journal of
Agricultural Engineering Research, 68(3): 201–205, 1997.
32. Denys S., Pieters J.G., Dewettinck K., Computational fluid dynamics analysis of combined conductive
and convective heat transfer in model eggs, Journal of Food Engineering, 63(3): 281–290, 2004.
19 Applications of CFD for
Optimization of Cabinet Dryers
Yasaman Amanlou and Mohammad Hadi Khoshtaghaza
CONTENTS
19.1 Introduction..................................................................................................................... 415
19.2 Cabinet Dryer Definition..................................................................................................... 416
19.2.1 Different Type of Cabinet Dryers and CFD Simulation Considerations................ 417
19.2.2 The Critical Aspect of Cabinet Dryer Design..................................................... 418
19.3 Operative Factors on Drying Process.................................................................................. 419
19.3.1 Drying Air Conditions............................................................................................ 419
19.3.2 Nature of Food........................................................................................................ 419
19.3.2.1 Drying Rate Characteristic Curve........................................................... 419
19.3.2.2 Diffusion Theories of Drying................................................................. 421
19.3.2.3 Thermal Properties and Conventional Heating....................................... 421
19.3.3 Dryer Design........................................................................................................... 421
19.4 Application of CFD to Simulate Cabinet Dryers................................................................ 422
19.4.1 Appropriate Physical Model Selection................................................................... 422
19.4.2 Basic Governing Equations for Designing the Cabinet Dryer................................ 422
19.4.3 Pressure Drop Through the Tray and Its Simulation.............................................. 422
19.4.4 Modeling Air Humidity and Product Moisture Content (Wet Porous Media)....... 423
19.5 Cabinet Dryer Optimization................................................................................................ 426
19.5.1 Calculation of Cabinet Dryer Efficiency................................................................. 426
19.5.2 Dryer Geometrical Design...................................................................................... 427
19.6 Model Validation................................................................................................................. 431
19.7 Conclusions.......................................................................................................................... 432
Nomenclature.................................................................................................................................. 432
References....................................................................................................................................... 432
19.1 INTRODUCTION
In food processing, optimization of the drying stage is very important due to the need to control
energy consumption and waste products. Cabinet dryers (also called tray dryers) are the favorite
equipment of farms and food industry for drying fruit, vegetable and other materials. These dryers
are simple in structure, low in cost installation and can be employed in almost any environmental
conditions. However, erratic design of this equipment leads to spoil, mold, over dried and burned
product. Likewise, if the capacity of the dryer is not proportional to input energy, time and money
is wasted [1–3].
Therefore, cabinet dryers, as common equipment in the food industry, should be carefully
designed to reduce waste and optimize energy consumption. The efficiency of a cabinet dryer
is strongly affected by inappropriate air flow distribution inside the dryer. Furthermore, non-
uniformity in the moisture content of the end product is the main disadvantage of this drying
415
system [4–7]. Heat and moisture distribution in a cabinet dryer depend on a number of param-
eters, including temperature and relative humidity of the drying air, product moisture content,
flow rate, airflow pattern, geometrical features of dryer, tray porosity, direction of airflow and
distances of trays from each other. To manipulate the dryer conditions to the optimum airflow
pattern, the relationship among these important parameters needs to be established. Researchers
and manufacturers can take advantage of computational fluid dynamics (CFD) to improve per-
formance of cabinet dryers. In the recent literature, different levels of complexity are chosen to
simulate the drying process and cabinet dryer geometries. Some scientists recommend particular
codes to accurately simulate real drying processes in transitory conditions and others simply use
commercial codes to simulate airflow profiles in steady state conditions to find defects in the geo-
metrical design.
This chapter deals with different ideas from recently published papers and experiments for read-
ers who want to become familiar with the optimization of cabinet dryers by CFD.
19.2 CABINET DRYER DEFINITION

Cabinet dryers consist of an insulated cabinet fitted with shallow mesh or perforated trays, each
of which contains a thin (2–6 cm deep) layer of food. Hot air is blown at 0.5–5 (m.s–1) through
a system of ducts and baffles to promote uniform air distribution over and/or through each tray.
Additional heaters may be placed above or alongside the trays to increase the rate of drying. Tray
driers are used for small-scale production (1–20 ton per day) or for pilot-scale work. They have
low capital and maintenance costs and are flexible in operation for different foods. However, they
have relatively poor control and produce more variable product quality as food dries more rapidly
on trays nearest to the heat source. Non-uniformity in the moisture content of the end product is
an inherent drawback in applying the cabinet dryer; hence producers are not usually interested
in utilizing this drying system [2]. As shown in Figure 19.1, the cabinet dryer chamber can be
systematically divided into four zones: inlet, plenum chamber, trays and outlet. Actually, there
are no solid walls between the zones, though this classification can be useful for computational
simulation.
FIGURE 19.1 Common cabinet dryer structure.

Applications of CFD for Optimization of Cabinet Dryers 417
19.2.1 Different Type of Cabinet Dryers and CFD Simulation Considerations

The cabinet dryers known as hot air-batch dryers can be classified according to various criteria,
such as: type of air replacement (active or passive); air circulation mode (open or close loop); capac-
ity (lab, pilot or industrial scale); the type of material for which it was designed (such as tender
and heat sensitive medicine material or heavy moist and insensitive animal feed); energy resource
(solar, electrical or fossil resources); and hybrid of cabinet dryer with other techniques of dry-
ing, such as infrared or microwaves. The schematics of a solar cabinet dryer, the simplest cabinet
dryer, and an industrial cabinet dryer, a more progressive one, are shown in Figures 19.2 and 19.3,
FIGURE 19.2 Schematic solar cabinet dryer.
FIGURE 19.3 Schematic diagram of industrial tray dryer. (From Sreekumar A., Manikantan P.E., and
Vijayakumar K.P. Performance of indirect solar cabinet dryer. Energy Conversion and Management, 2008,
Volume 49(6), Pages 1388–1895.)
TABLE 19.1
Some Important Aspects of Each Type of Cabinet Dryer
Criterion Explanation Simulation Considerations References
Type of air Some types of tray dryers don’t have a blower Natural convection models such [8,9]
replacement to move air inside the dryer. In these cases as Boussinesq and Non-
natural convection is the only source of air Boussinesq approximation
replacement
Air circulation To avoid energy loss and heat recovery, portion In inlet (or outlet) zone, some [10,11]
mode of exhaust air could be returned into the equipment should be considered
drying chamber to control the relative humidity
and temperature of circulating air
Capacity Air relative humidity increases rapidly inside Different assumptions are needed [7,10,12,13]
the dryer and intense mole grows and for each class of dryer
spoilage.
Intensity of Airflow regime (laminar, transitional and Turbulence equations, heat loss of [14,15]
drying process turbulent) dryer through the walls
Energy resource • Electrical heater is the simplest resource Constant temperature inlet air for [16,17]
for simulation thermostatic electrical heaters
• Fossil resources usually need an exchange Time dependent equation of solar
chamber to transfer heat from burning fuel irradiant
to drying air
• Fluctuation of solar irradiation must be
taking account during the day
Hybrid dryers Depending on the techniques of drying Radiation source definition and [17–22]
active radiation equation
Solar dryers • Direct Opaque [23,24]
• Indirect Semi opaque (transparent) walls
respectively. For simulation and accurate design, different assumptions for each particular cabinet
dryer should be taken account. Some important aspects of each type of cabinet dryer are sum-
marized in Table 19.1.
19.2.2 The Critical Aspect of Cabinet Dryer Design

Uniform air flow distribution inside dryers is of paramount importance because it determines both
the efficiency and the homogeneity of the products being dried [25]. Producers usually have to over-
come this problem by periodically replacing the lowest tray in the stack at conventional cabinet dry-
ers. This requires that a skilled worker estimate the right time for replacing the trays. Furthermore,
the open door during the tray replacement causes intensive wasting of energy.
Food engineers and dryer designers can take advantage of computational fluid dynamics to
improve the structure of cabinet dryer to achieve uniform air flow and temperature distribution
inside the cabinet dryer. Geometry of plenum chamber, distance between trays, inlet air veloc-
ity and turbulence, use of deflectors and baffles, use of an axillary heater inside the cabin and
its position, different shape and type of air inlet and outlet, density of materials on the each
tray, trays mesh, effect of walls and trays conductivity on the dryer efficiency and heat recovery
from exhaust air are all important in the design of a cabinet dryer and make it complicated.
Controlling all of these parameters experimentally is very tedious and difficult. Although com-
putational fluid dynamics techniques (CFD) cannot replace physical experiments completely,
they can significantly reduce the amount of time needed for experimental works. With the help
of information such as temperature, velocity and Reynold’s number of the entrance air to the
dryer, as well as the right physical model for describing dryer conditions, researchers can get
the contours of air distribution inside the cabinet. The closer initial data is to reality and the
more suitable the assumptions are for solving the governing equations, the more accurate the
results will be.
CFD has been widely used for predicting the air velocity and temperature in drying chambers.
Additionally, a number of studies have shown the capability of CFD techniques to accurately pre-
dict air relative humidity and material moisture content in convectional drying processes. With the
development of low cost, powerful computers and commercial software packages such as Fluent,
Star-CD and CFX4 in last decade, CFD has been increasingly deployed in the food industry [26].
19.3 OPERATIVE FACTORS ON DRYING PROCESS

There are a large number of factors that control the rate of the food drying process. Operative fac-
tors in the drying process can be grouped into the following categories [2]:
• Those related to the processing conditions

• Those related to the nature of the food
• Those related to the drier design.
These factors are dealt with in the following sections.
19.3.1 Drying Air Conditions

When hot air is blown over a wet food, water vapor diffuses through a boundary film of air sur-
rounding the food and is carried away by the moving air [2]. A water vapor pressure gradient is
established from the moist interior of the food to the dry air. This gradient provides the “driving
force” for water removal from the food. In addition, there are three inter-related factors that control
the capacity of air to remove moisture from food:
1. The amount of water vapor already carried by the air

2. The air temperature
3. The amount of air that passes over the food.
In fact, the first and second factors are representative of the psychrometric processes of drying
air. Psychrometric processes include physical and thermodynamic properties, such as dry and wet
bulb, humidity, enthalpy, and air density. The third factor is the quantity of drying air that flows
through the dryer. According to the designer’s decision, the simulation can consist of all air qualita-
tive characteristics or simply indicate flow distribution inside the dryer.
19.3.2 Nature of Food
There are several methods to explain the behavior of the material during the drying process, but the
method of a drying rate characteristic curve is the easiest method to understand due to its empirical
quiddity. The fundamental concepts of this method are stated in the following section. The appli-
cation of drying rate characteristic curves in CFD simulation and other methods are explained in
Section 19.4.4.
19.3.2.1 Drying Rate Characteristic Curve

Van Meel (1985) postulated that when working with convective batch driers, a single characteristic
drying curve could be deduced for the material being dried. This model reflects the nature of the
drying rate curves shown in Figure 19.2. It is empirical but has success in correlating drying kinetics
of small particles. The model assumes that for any given sample water content, a corresponding
specific drying rate exists [27].
This expression has been used extensively as the basis for understanding the behavior of indus-
trial drying plants, because of the simplicity of the parameters that are used to define the drying
process. Two distinguished stages of characteristic drying curves are [24]:
1. Constant-rate period
When food is placed into a drier, there is a short initial settling down period as the surface
heats up to the wet-bulb temperature (A–B in Figure 19.4(a)). Drying then commences
and, provided water moves from the interior of the food at the same rate as it evapo-
rates from the surface, the surface remains wet. This is known as the constant-rate
period and continues until a certain critical moisture content is reached (B–C in Figure
19.4(a) and (b)).
The surface temperature of the food remains close to the wet-bulb temperature of the dry-
ing air until the end of the constant-rate period, due to the cooling effect of the evapo-
rating water. In practice, different areas of the food surface dry out at different rates
and, overall, the rate of drying declines gradually towards the end of the “constant”-
rate period.
2. Falling-rate period
When the moisture content of the food falls below the critical moisture content, the rate of
drying slowly decreases until it approaches zero at the equilibrium moisture content
(that is, the food comes into equilibrium with the drying air). This is known as the
falling-rate period. Non-hygroscopic foods have a single falling-rate period (C–D in
Figure 19.4(a) and (b)), whereas hygroscopic foods have two or more periods. In the
first period, the plane of evaporation moves from the surface to inside the food, and
water vapor diffuses from solids to the drying air. The second period occurs when the
partial pressure of water vapor is below the saturated vapor pressure, and drying is by
desorption.
There are many drying curve models in the literature that are successfully applied for dry-
ing process prediction and simulation [28–31]. Nevertheless, deciding whether to use
them in CFD simulation depends on the needed level of accuracy.
FIGURE 19.4 Drying curves (a) moisture variations, (b) drying rate.
19.3.2.2 Diffusion Theories of Drying

As soon as the pure water evaporation-like behavior is finished (i.e., the so-called constant drying
flux period), the falling-rate period starts and the internal moisture transfer or the internal resis-
tance to moisture transfer becomes important. Fick’s diffusion law has been suggested to apply.
This theory is popular as its complexity is in between that of comprehensive theories, such as the
Luikov theory, and the totally empirical model, such as the Page model. It also possesses fundamen-
tal information—an effective Fickian-type diffusivity. Furthermore, the related partial differential
equation, governing mass transfer process, is the simplest mathematical model of drying that could
predict the transient water content distribution realistically across the material being dried [27].
The mass diffusivities in food processes are sometimes difficult to determine and the variability
is large. The moisture diffusivity is dependent on temperature, water content and porosity, and a
number of models exist for describing moisture diffusivities. These models, as mentioned in regard
to drying rate characteristic curves, can be employed in CFD simulation [32–34].
19.3.2.3 Thermal Properties and Conventional Heating

For food materials, two thermal properties are of particular interest: heat conductivity and spe-
cific heat capacity, as they are directly replaced in cooking. Heat conduction is the most important
mechanism of heat transfer within the food materials during drying. It occurs wherever there is
a “material” medium and a temperature gradient within. The conductivity of food material is a
unique topic in food engineering. Food usually consists of fat, protein, mineral, moisture and air
components, and other minor components like volatiles. Specific heat capacity such as conductiv-
ity is another important property of foodstuffs in the calculation of the temperature rise or drop.
The composite, mixture or native food’s specific heat capacity may be estimated using the equation
based on weight fraction [27].
19.3.3 Dryer Design
Usually determining the effect of the geometry on the fluid flow distribution to design a new struc-
ture is the main purpose of cabinet dryer simulation by CFD [5,7,17,35]. Theoretically, dividing the
dryer into four zones allows you to check out more details about each zone. Structural changes in
each zone can lead to a new design of a dryer. To simulate the dryer effectively, initial and boundary
conditions must be defined around the boundary of system (domain). Some important parameters
and aspects of simulation in the case of a cabinet dryer are listed below:
Inlet and outlet zones: Size and shape of hatchway, air velocity and turbulence factors, exis-
tence of heater or thermal exchanger, blower type or ventilating equipment qualifications
like gauge pressure and performance, existence of orifice or plate to control the air flow,
upward ducts bending and roughness, location of outlet hatch and hood type ceiling.
Plenum chamber zone: Geometrical shape of this zone is very important. Divergence angle of
plenum chamber and its direction to trays have excellent influence on flow distribution on
the trays zone. Using deflectors, baffles or straighteners lead the air flow to the desirable
direction.
Trays zone: Distance between trays, density of material on the tray, tray mesh, air flow pres-
sure drop through the trays, thermal properties of the trays and food material such as con-
ductivity and thermal capacity are consequential details for estimating fluid flow.
Further, the use of an axillary heater inside the cabin and its place, heat recovery mechanism,
thermal properties of walls and heat lost coefficient from walls to ambient are other factors that
should be taken account in dryer design.
19.4 APPLICATION OF CFD TO SIMULATE CABINET DRYERS

As previously mentioned, non-uniformity in the moisture content of the end product is an inherent
drawback in applying the cabinet dryer; hence, producers are not usually interested in utilizing this
drying system. Air flow and temperature distribution simulation inside a cabinet dryer is a valuable
method for improving the structure of cabinet dryer or introducing a new design. CFD is a powerful
tool for researchers and engineers, who can take advantage of this tool to improve the final dried
product quality and drying efficiency. The influence of different geometries in cabinet dryers on the
uniformity or air flow distribution inside the dryer can be considered by CFD simulation before the
manufacture.
In the following, the most important applications of CFD to cabinet dryers are briefly described
and some numerical modeling techniques are summarized.
19.4.1 Appropriate Physical Model Selection

Simplification of the problem and selection of a physical model to express the dryer conditions are
the most important stages of simulation. The researcher should carefully decide between steady
state and transient calculation. Predetermining strangle depends on the demanded results and their
desired accuracy. In general, transient simulations are complicated and time consuming, and it is
better to employ them just in the cases where the time evolution of phenomena is under investiga-
tion, such as product local moisture content estimation during the drying.
Due to complexity of the dynamic coupling of phase-heat and mass transfer in wet porous media,
some authors prefer to simply use a steady state model to estimate air flow and temperature inside
the cabinet. Amanlou & Zomorodian [5] took into account Darcy’s law for simulating the pressure
drop of air flow during drying air transit through the trays. They studied effects of different geomet-
rical shapes of the cabinet on the drying air distribution inside the dryer. In the same way, Darabi
et al. [35] computed air flow distribution inside the new-designed cabinet dryer and evaluated the
CFD code by experimental results. Their results indicated that neglecting product moisture content
had no sensible influence on air flow pattern computation inside the dryer.
19.4.2 Basic Governing Equations for Designing the Cabinet Dryer

From the point of view of the designer, the effective simulation of the airflow pattern and tempera-
ture distribution inside the tray dryer is the most important result. However, there are too many
factors regarding the geometry of the cabinet dryer, the operating conditions of the cabinet, and the
air inlet and outlet conditions, that interact and influence the performance of the dryer. Most heat
transfer is by convection from the drying air to the surface of the food, but there may also be heat
transfer by radiation. If the food is dried in solid trays, there will also be conduction through the tray
to the food. Calculation of heat transfer is therefore often very complex in drying systems.
There are three categories of basic equations, derived from three basic laws of mass and energy
conservation. The mass, momentum and energy conservation result in the continuity equation,
Navier–Stokes equation and energy equation, respectively [26].
19.4.3 Pressure Drop through the Tray and Its Simulation

When air flows through a porous bed of materials like agricultural products, the air pressure will
drop. To recover the pressure, fans are necessary and the energy demand for the fan depends highly
on the pressure drop. Inattention to relationship between air velocity, bed type, material moisture con-
tent, bed depth and filling method in any drying process could result in excessive water loss, shrink-
age, quality degradation and also cause large pressure drops that require more powerful fan systems.
However, lower air flow rates result in increasing product temperature and risk of insect influx.
The prediction of air flow resistance in a selected bed of agriculture material has been studied
widely for more than 70 years. Comprehensive studies on the effect of different factors such as air
flow rate, bed type, moisture content of bulk, bed depth, filling method, amount and type of foreign
materials, and direction of air flow through the bulk on air flow resistance have been conducted on
grains and other agricultural products. The effects of these parameters on air flow resistance have
been closely reviewed for grains and other food materials [36–39].
In many cases, the pressure drop can be analyzed by well-known empirical relationships, such
as Shedd (1953) and Hukill and Ives (1955) models. These non-linear models easily fit many experi-
mental data sets. However, the constants in these equations have a purely empirical nature without
physical considerations [36].
Fruit trays are porous media for air flow. Porous media are modeled by the addition of a momen-
tum source term to the standard fluid flow equations. The source term is composed of two parts: a
viscous loss term (the first term on the right-hand side of Equation (19.1) and an inertial loss term
(the second term on the right-hand side of Equation (19.1)).
 3 3 
∑ ∑
1
Si = −  Dij µν j + Cij ρν magν i  (19.1)
 j =1 j =1
2 

where Si is the source term for the ith (x, y, or z) momentum equation and D and C are prescribed
matrices. This momentum sink contributes to the pressure gradient in the porous cell, creating a pres-
sure drop that is proportional to the fluid velocity (or velocity squared) in the cell. Some commercial
software also allows the source term to be modeled as a power law of the velocity magnitude:
C1
Si = −C0 ν (19.2)

where C0 and C1 are user-defined empirical coefficients related to bed pressure drop of the porous
media. The same way other empirical equations could be replaced by S term in the momentum
equation.
19.4.4 Modeling Air Humidity and Product Moisture Content (Wet Porous Media)
Fruits, vegetables and almost all food stuffs dried in cabinet dryers are categorized as porous mate-
rials based on their hygroscopic nature. The modeling interest here is directed towards moist porous
materials. A moist porous material is considered to contain a solid matrix, liquid water and void or
moist air space (i.e., dry air and water vapor) [40,41]. As shown in Figure 19.5, two sources of liquid
water exist within the porous material: free liquid water that occupies the interconnected pores, and
bound water within the solid constituents’ closed micro-pores. As the moist material loses its mois-
ture content, the closed pores open up and the bound water gradually becomes free.
Complete modeling of a drying process requires the consideration of moist air passing across the
material to be dried with dynamic coupling between the constituents to enable exchanges of heat
and moisture to evolve based on local conditions (i.e., temperature, water activity, etc.).
Studies that involve heat and mass transfer in porous media are of significant practical interest
and involve various applications. The importance of the subject has grown over the past decade
and the capability of computational fluid dynamics (CFD) to treat the microscopic heat and mass
exchanges between the different phases of porous materials in an accurate and timely fashion is an
active research topic [40]. There are different numerical methods to simulate porous material in the
convective drying process. Khan and Straatman [41] broadly classified numerical models into cate-
gories based on their level of sophistication. According to their classification, the most basic models
FIGURE 19.5 Illustration showing the different constituents in the porous media. (From Elhalwagy M.M.,
and Straatman A.G. Dynamic coupling of phase-heat and mass transfer in porous media and conjugate fluid/
porous domains. International Journal of Heat and Mass Transfer, 2017, Volume 106, Pages 1270–1286.)
are drying curve models, which utilize a single moisture equation to evaluate moisture loss from a
porous material as a function of temperature and other parameters. There are many mathematical
and semi-empirical models for describing dehydration phenomena for various vegetables, fruits and
food stuffs. These models provide overall drying rates; however, they are not capable of accounting
for the local variations of different quantities inside the porous domain.
To apply this method, the drying rate curve is normalized with initial conditions of product and
equilibrium data as:
M (t ) − M e
MR(t ) = (19.3)
Mc − Me
where MR, M, Me, and t are moisture ratio, moisture content (dry bases), equilibrium moisture,
critical moisture and drying time, respectively. For long drying processes, Me values are negligible
compared to M. Therefore, Equation (19.3) can be simplified to:
M (t )
MR(t ) = (19.4)
Mc
The drying time and moisture ratio variation are related to drying air velocity and temperature.
This equation can convert to a differential equation:
dX
MR(t ) = X (t ); X = (19.5)
dt
The boundary conditions are: X (t = 0) = 0, X (t = ∞) = 0.

Several mathematical equations have been suggested by scientists to explain the relationship
between moisture ratio and drying time at a given drying air temperature and velocity. Some of the
famous models are listed in the Table 19.2. The mathematical models of the drying curve have one
or more empirical coefficients (constants) that are identified by experimental tests for each prod-
uct. The coefficients of some models were calculated for numerous agricultural products and food
materials in previous studies; this makes it easy to apply the particular model in CFD simulations.
Generally, drying characteristic curves are obtained for thin layers of material. Considering that
materials are going to dry in thin layers in the tray dryer, drying characteristic curves have accept-
able efficiency to simulate the drying process in trays.
The next category of models uses differential energy and moisture transport equations to model
the drying process inside the porous domain. As a result, these models predict the local variations
TABLE 19.2
Famous Models Applied to Drying Curve
Model
Number Model Name Model Constants References
1 Newton MR = exp (−kt) k [42]
2 Page MR = exp (−ktn) k, n [43]
3 Modified Page MR = exp (−(kt)n) [44]
4 Henderson and Pabis MR = a exp (−kt) a, k [45]
5 Logarithmic MR = a exp (−kt) + C a, C [46]
6 Two term MR = a exp (−k0t) + b exp (−k1t) a, b, k0, k1 [47]
7 Exponential two term MR = a exp (−kt) + (1 − a) exp (−kat) a, k [48]
8 Wang and Singh MR = 1 + at + bt2 a, b [31]
9 Thompson t = a ln(MR) + b [ln(MR)]2 a, b [49]
10 Approximation of diffusion MR = a exp (−kt) + (1 − a)exp (−kbt) a, b, k [31]
11 Verma et al. MR = a exp(−kt) + (1 − a)exp (−gt) a, g, k [45]
12 Modified Henderson and Pabis MR = a exp(−kt) + b exp(−gt) + c exp (−ht) a, b, c, g, h, k [49]
13 Midilli et al. MR = a exp (−ktn) + bt a, b, k, n [50]
of temperature and moisture content inside the drying material by solving a single equation each for
energy and moisture transport. The transport equations employed, in general, contain only unsteady
and diffusion terms, and Fourier and Fick’s Laws are used to model diffusion in the energy and
moisture equations, respectively. In such models, convective boundary conditions are imposed at
the surface of the material to be dried. In this respect, convective heat and mass transfer coefficients
are evaluated using empirical correlations based on Nu = f (Re, Pr) and Sh = f (Re, Sc), which means
that the simulation results are dependent upon the empirical correlations used.
The third category of models enables improvement in the evaluation of convective heat and
mass transfer inside the porous domain. Such models solve single energy and moisture transport
equations inside the drying material, but also consider the fluid region surrounding the material,
although not in a direct coupled, conjugate manner. The surrounding airflow is first resolved by
solving the mass and momentum transport equations along with the transport equation of energy to
calculate the heat transfer coefficient at the surface of the material. The thermal and concentration
boundary layer analogy is then used to compute the surface mass transfer coefficient. The evaluated
coefficients are then utilized to impose convective boundary conditions at the material’s surface to
obtain a solution inside the drying material. The term “non-conjugate approach” is often used to
refer to this category of models [41,51].
Some scientists have succeeded recently in using a conjugate approach, except that moisture
transfer inside the drying material was modeled by combining the vapor and water transport into a
single moisture transport equation.
The airflow surrounding the porous material plays a crucial role in the convective drying pro-
cess. From the numerical modeling perspective, moist air can be considered a mixture of dry air
(comprised of all the gaseous components) and water vapor. The vapor content of the moist airflow
is evaluated by solving a transport equation in the form of vapor mass fraction. The moist air is then
treated as a mixture of dry air and water vapor, and its flow is evaluated by solving the conventional
mass-momentum transport equations. The moist air density (ρf) used in the transport equations is
continuously updated to account for the moisture gain/loss of the air by the expression:
Pa P
ρ f = ρa + ρ v = + v (19.6)
Ra T Rv T
where Pa is the partial pressure of the dry air, Pv is the vapor partial pressure, and Ra and Rv are the
air and vapor gas constants, respectively. The moist air temperature (T) is obtained by solving the
energy transport equation.
19.5 CABINET DRYER OPTIMIZATION

Before going into details of process calculations, we need to determine when such calculations are
necessary in industrial practice. The following typical cases can be distinguished [3]:
Design—(a) selection of a suitable dryer size for a given product to optimize the capital and
operating costs within the range of limits imposed, (b) specification of all process parameters and
dimensioning of a selected dryer type so the set of design parameters or assumptions is fulfilled.
Simulation—for a given dryer, calculation of dryer performance including all inputs and outputs,
internal distributions, and their time dependence.
Optimization—in design and simulation an optimum for the specified set of parameters is
sought. The objective function can be formulated in terms of economic, quality, or other factors,
and restrictions may be imposed on ranges of parameters allowed.
Process control—for a given dryer and a specified vector of input and control parameters the out-
put parameters at a given instance are sought. This is a special case when not only the accuracy of
the obtained results but the required computation time is important. Although drying is not always
a rapid process, in general for real-time control, calculations need to provide an answer almost
instantly. This usually requires a dedicated set of computational tools like neural network models.
19.5.1 Calculation of Cabinet Dryer Efficiency

Energy efficiency in drying is of obvious importance as energy consumption is such a large com-
ponent of drying costs. Basically, it is a simple ratio of the minimum energy needed to the energy
actually consumed. But because of the complex relationships of the food, the water, and the drying
medium which is often air, a number of efficiency measures can be worked out, each appropriate
to circumstances and therefore selectable to bring out special features important in the particular
process. Efficiency calculations are useful when assessing the performance of a dryer, looking for
improvements, and in making comparisons among the various classes of dryers that may be alterna-
tives for a particular drying operation.
Heat has to be supplied to separate the water from the food. The minimum quantity of heat that
will remove the required water is that needed to supply the latent heat of evaporation, so one mea-
sure of efficiency is the ratio of that minimum to the energy actually provided for the process. The
quantity of energy required to vaporize product moisture calculated by [52]:
Q w = h fg. Mwd (19.7)

where Qw is energy spent on evaporation water from food stuff (kJ), hfg is latent heat of vaporization
(kJ/kg) and Mwd is evaporated water mass (kg).
If dryer CFD simulation consists of estimation of food material moisture content, it can be
advantageous to calculate the thermal efficiency of the cabinet dryer. The thermal efficiency of
dryer is calculated by:
Qw
ηth = (19.8)
Qin
where Qin is the overall thermal energy consumed by the dryer.

Sensible heat can also be added to the minimum, as this added heat in the food often cannot be
economically recovered.
Unlike efficiency, performance indicates the quality of drying process in a special dryer. There
is no regular method to express performance in a dryer but by making comparisons among some
properties—such as thermal efficiency, specific rate of evaporation (SRE) and quality of end product—
performance could be commented on.
19.5.2 Dryer Geometrical Design

To obtain a uniform distribution of drying air flow and temperature, taking into account the overall
operative factors on drying process, different geometries of the cabinet dryer can be envisaged theo-
retically using CFD. Amanlou and Zomirodian [5,53] have studied geometry of a new fruit cabinet
dryer (lab scale) with three trays and suggested a side mounted plenum chamber with a trapezoid
cross-section (Figures 19.6 and 19.7). The side mounted plenum chamber prepared a flow of fresh
drying air between upper trays to keep the air drying potential high and uniformly distributed
within trays. It is believed that this change would result in more even air flow distribution and more
uniform moisture removal from all the trays simultaneously in the cabinet dryer.
In order to find the most appropriate geometrical shape of cabinet dryer for achieving more uni-
form distribution of drying air flow inside the dryer, after investigating many sketches with different
shapes and dimensions, seven designs were selected. It is obvious that the geometry of trapezoid
plenum chamber (air inlet size, angle of side plenum chamber with respect to horizon (divergence
angle, α) and width of side plenum chamber) would affect the air flow pattern in the dryer chamber.
In many conventional dryers, installing a hood-type ceiling and air deflector are common, and thus
the effects of these two parameters were also studied. The geometrical configurations are displayed
in Figure 19.8.
To determine the suitability of the design of the new cabinet dryer, data analysis of seven pro-
posed plans were carried out using CFD. The most feasible design can be assessed by comparing
the uniformity of air flow distribution in the drying chamber (Figure 19.9).
They concluded that changing the geometrical shape of the drying chamber displayed more
influence on air velocity distribution than on air temperature distribution. In design (a), the air
FIGURE 19.6 3D diagram of three trays fruit cabinet dryer with side mounted plenum chamber. (From
Amanlou Y. Optimizing a solar cabinet dryer for Estahban green fig for uniform dehydration. Msc Thesis,
“In Persian,” Shiraz University, Iran, 2009.)
FIGURE 19.7 Inside view of the dryer. (From Amanlou Y. Optimizing a solar cabinet dryer for Estahban
green fig for uniform dehydration. Msc Thesis, “In Persian,” Shiraz University, Iran, 2009.)
FIGURE 19.8 Various designs of cabinet dryer; (a) small air inlet and outlet; (b) big divergence angle and air
outlet in the center of ceiling; (c) air outlet on the right side; (d) air deflector; (e) air deflector and hood-type ceil-
ing; (f) high ceiling; and (g) short ceiling by Amanlou and Zomorodian. (From Amanlou Y., and Zomorodian A.
Applying CFD for designing a new fruit cabinet dryer. Journal of Food Engineering, 2010, Pages 8–15.)
FIGURE 19.9 Air velocity distribution profiles simulated by CFD for the seven designs; (a) poor distribution
of hot air; (b and c) low air velocity on the trays; (d) appropriate air velocity on the trays; (e, f, and g) could not
improve the air distribution in the dryer chamber more than design (d) does. (From Amanlou Y., and Zomorodian A.
Applying CFD for designing a new fruit cabinet dryer. Journal of Food Engineering, 2010, Pages 8–15.)
flow canalizes through the bottom of the drying chamber to the air exit. This canalization can be
attributed not only to the small size of the air inlet and exit in comparison with chamber dimensions
but also due to small divergence angle of α. Therefore, design (a) is not recommended for the dryer
chamber. Minimum air velocities at the tray zones happened in designs (a and d). Design (b) seems
to be the best, taking into account all considerations such as the simplicity in structure, uniformity
in air temperature and velocity distribution.
They used the k − ε model as a turbulence model and taken advantage of an empirical model to
describe the pressure drop through the fruit trays. Even though they did not simulate air humidity
and moisture content, the statistical analysis showed that their results had a good correlation with
experimental results. They reported that correlation coefficients of 99.9% and 86.5% for drying air
temperature and air velocity in the drying chamber, respectively.
In the same method, several different three-dimensional geometries of a cabinet dryer with three
fruit holding trays were studied using CFD by Darabi et al. [35]. They suggested another plenum
FIGURE 19.10 A design recommended by Darabi et al. (From Darabi H., Zomorodian A., Akbari M.H., and
Lorestani A.N. Design a cabinet dryer with two geometric configurations using CFD. Journal of Food Science
and Technology, 2013.)
chamber and changed the place of the outlet so that each tray had its own outlet (Figure 19.10).
Their recommended geometry provided more uniform air flow on the trays but the heat dissipation
seemed high.
Design optimization of a drying chamber is necessary to achieve higher heat/mass transfer
rates and uniform drying by avoiding an unfavorable aerodynamic phenomenon in the chamber.
Misha et al. [7] used CFD simulation to predict the drying uniformity for a new industrial tray
dryer design for agricultural products. The variation in air velocity above the trays was selected
as a benchmark for the uniformity of air distribution inside the dryer. The drying efficiency of
this system may be increased with a high air velocity inlet (Figure 19.11).
Actually, knowing about airflow distribution inside the dryer is insufficient, because air
velocity is just one of the factors affecting uniformity of the drying process. Misha et al. [54], in
another study, observed the variation in final moisture content of kenaf core in an experimental
test with a dryer similar to that mentioned above. They had to remove the dried product from
the first tray and shift the second and third to the front position. Their results indicated that air
relative humidity and product moisture content should be considered to achieve more accurate
results. Finally, Misha and his coauthors recommended that the uniformity of the drying may
be improved by using additional baffle to direct the air flow to each tray, exchange the positions
of the tray during the drying process or conduct the drying in semi-continuous mode.
FIGURE 19.11 Velocity distribution profiles at industrial tray dryer. (From Misha S., Mat S., Ruslan M.h.,
Sopian K., and Salleh E. The prediction of drying uniformity in the tray dryer system using CFD simulation.
International Journal of machine Learning and Computing, 2013, Volume 3(5).)
19.6 MODEL VALIDATION
It is common idea that CFD results must be validated by experimental outputs. For this purpose,
some important and measurable variables are selected. Experimental outputs are collected from a
specified location and then compared to simulated values at known conditions. Usually air velocity,
air temperature, air relative humidity and weight loss of food materials are selected for comparison.
Three statistical criteria—namely coefficient of determination (R2), root mean square error (RMSE)
and mean relative percentage deviation modulus (P %)—are frequently used in literature as indices of
goodness of the CFD simulation output. Scientists are also interested in showing the measured values
on one axis and corresponding CFD predicting values on the other axis (as in the chart in Figure 19.12).
As the data points are closer to the bisector of the coordinate axis, the accuracy of the model is better.
FIGURE 19.12 Comparison of the results of experimental and CFD predicted drying air temperature.
19.7 CONCLUSIONS
Cabinet dryers are the most extensively used drying method because of their simple structure and
economic design. The main drawback of this dryer is non-uniformity in the desired moisture con-
tent of the end product due to poor air flow distribution in the drying chamber. It is essential to
use CFD to surmount this problem and improve the dryer efficiency. The role of CFD analysis is
important as it reduces the number of experimental tests needed to achieve the optimum operating
conditions of drying and the most appropriate geometry of the dryer.
Simplification of the problem and selection of a physical model to express the dryer conditions
are the most important stages of CFD simulation. The researcher should carefully decide between
different models for simulating porous media and drying processes. Most researchers prefer to sim-
ply simulate dryers in the steady state condition rather than take a conjugate approach. Nevertheless,
simulating air humidity changing during the drying process is required to approach optimum condi-
tions in a tray dryer.
NOMENCLATURE
ρ Density of fluid
ρf Moist air density
μ Dynamic viscosity
C, D Prescribed matrices
C0, C1 Empirical coeficients
hfg Latent heat of vaporization (kJ kg–1)
M Moisture content (dry bases)
MR Moisture ratio
Mwd Evaporated water mass (kg)
Mc Critical moisture content
Me Equilibrium moisture content
Nu Nusselt number
Pr Prantel number
Pa Partial pressure of the dry air (Pa)
Pv Vapor partial pressure (Pa)
Qw Energy spent on evaporation water from food stuff (kJ)
Ra The air constants
Rv The vapor gas constants
Re Reynolds number
Sc Schmidt number
Sh Sherwood number
Si Source term for ith momentum equation (m s–1)
T Temperature of moist air
t Drying time
vi Velocity vector
vmag Velocity magnitude
REFERENCES
1. Singh R.P., Heldman D.R., Introduction to Food Engineering, 5th ed., Academic Press, Cambridge,
MA, USA, Chapter 12, 2014.
2. Fellows P.J., Food Processing Technology: Principles and Practice, 3rd ed., Woodhead Publishing,
Sawston, Cambridgeshire, UK, Chapter 16, 2009.
3. Imre L., Handbook of Industrial Drying, 4th ed., Mujumdar A.S., Ed., CRC Press, Boca Raton, FL,
USA, Chapter 13, 2014.
4. Prieto W.H., Iguti A.M., Nitz M., Drying evaluation of green coconut pulp for obtaining a snack-like
product, In: 11th International Congress on Engineering and Food (ICEF11), Procedia Food Science,
1618–1627, 2011.
5. Amanlou Y., Zomorodian A., Applying CFD for designing a new fruit cabinet dryer, Journal of Food
Engineering, 8–15, 2010.
6. Sreekumar A., Manikantan P.E., Vijayakumar K.P., Performance of indirect solar cabinet dryer, Energy
Conversion and Management, 49(6): 1388–1895, 2008.
7. Misha S., Mat S., Ruslan M.h., Sopian K., Salleh E., The prediction of drying uniformity in the tray
dryer system using CFD simulation, International Journal of machine Learning and Computing, 3(5):
2013.
8. Kouroudis I., Saliakellis P., Yiantsios S.G., Direct numerical simulation of natural convection in a
square cavity with uniform heat fluxes at the vertical sides: Flow structure and transition, International
Journal of Heat and Mass Transfer, 115(B): 428–438, 2017.
9. Bouafa M., Hamimid S., Guellal M., Non-Boussinesq convection in a square cavity with surface ther-
mal radiation, International Journal of Thermal Sciences, 96: 236–247, 2015.
10. Fiaschi D., Manfrida G., Russo L., Talluri L., Improvement of waste heat recuperation on an indus-
trial textile dryer: Redesign of heat exchangers network and components, Energy Conversion and
Management, 150: 924–940, 2017.
11. Aktas M., Sevik S., Amini A., Khanlari A., Analysis of drying of melon in a solar-heat recovery assisted
infrared dryer, Solar Energy, 137: 924–940, 2016.
12. Yi Y., Salonitis K., Tsoutsanis P., Litos L., Patsavelas J., Improving the curing cycle time through the
numerical modeling of air flow in industrial continuous convection ovens, Procedia CIRP, 63: 499–504,
2017.
13. Alexeenko A.A., Ganguly A., Nail S.L., Computational analysis of fluid dynamics in pharmaceutical
freeze-drying, Journal of Pharmaceutical Sciences, 98(9): 3483–3494, 2009.
14. Jin G., Zhang M., Fang Z., Cui Z., Song C., Numerical study on spout elevation of a gas-article spout
fluidized bed in microwave-vacuum dryer, Journal of Food Engineering, 143: 8–16, 2014.
15. Llerena-Chavez H., Larachi F., Analysis of flow in rotating packed beds via CFD simulations—Dry pres-
sure drop and gas flow maldistribution, Chemical Engineering Science, 64(9): 2113–2126, 2009.
16. Smolka J., Nowak A.J., Rybarz D., Imporoved 3-D temperature uniformity in a laboratory drying oven
based on experimentally validated CFD, Journal of Food Engineering, 97(3): 373–383, 2010.
17. Amanlou Y., Fadaia D., Zomorodian A., Investigating a gaseous assisted solar cabinet dryers using
CFD, In: XXXIV CIOSTA CIGR V Conference, 29th Jun–1st July 2011.
18. Computational fluid dynamic analysis of innovative design of solar-biomass hybrid dryer: An experi-
mental validation, Renewable Energy, 92: 185–191, 2016.
19. Saxena A., Varun, El-Sebaii, A thermodynamic review of solar air heaters, Renewable and Sustainable
Energy Reviews, 43: 863–890, 2015.
20. Bao Y., Wang Y., Thermal and moisture analysis for Tabacco Leaf fluecuring with heat pump technol-
ogy, Procedia Engineering, 146: 481–493, 2016.
21. Yuan Y., Tan L., Xu Y., Yuan Y., Three-dimensional CFD modeling and simulation on the perfor-
mance of steam ejector heat pump for dryer section of the paper machine, Vacuum, 122(A): 168–178,
2015.
22. Lin S., Broadbent J., McGlen R., Numerical study of heat pipe application in heat recovery systems,
Applied Thermal Engineering, 25(1): 127–133, 2005.
23. Romero V.M., Cerezo C., Garcia M.I., Sanchez M.H., Simulation and validation of vanilla drying pro-
cess in an indirect solar dryer prototype using CFD fluent program, Energy Procedia, 57: 1651–1658,
2014.
24. Prakash O., Laguri V., Pandey A., Kumar A., Kumar A., Review on various modelling techniques for
the solar dryers, Sustainable Energy Reviews, 62: 396–417, 2016.
25. Mirade P.S., Prediction of the air velocity field in modern meat dryers using unsteady computational
fluid dynamics (CFD) models, Journal of Food Engineering, 60: 41–48, 2003.
analysis tool for the food industry: A review, Trends in Food Science and Technology, 17: 600–620,
2006.
27. Chen X.D., Mujumder A.S., Drying Technologies in Food Processing, Blackwell Publishing Ltd., Oxford,
UK, Chapters 1–2, 2008.
28. Mabrouk S., Belghith A., Simulation and design of a tunnel dryer, Renewable Energy, 5: 469–473, 1994.
29. Artnaseaw A., Theerakulpist S., Benjapiyaporn C., Drying characteristics of shiitake mushroom and
jinda chili during vacuum heat pump drying, Food and Bioproducts Processing, 88(2–3): 105–114,
2010.
30. Defraeye T., Verboven P., Convective drying of fruit: Role and impact of moisture transport properties
in modeling, Journal of Food Engineering, 193: 95–107, 2017.
31. Amanlou Y., Tavakoli Hashjin T., Ghobadian B., Najaf G., Mathematical modeling of thin-layer
solar drying for yarrow, coriander and hollyhock, International Journal of Food Engineering,
11(5): 691–700, 2015.
32. Van Belleghem M., Steeman M., Janssens H., Janssens A., De Paepe M., Validation of a coupled heat,
vapour and liquid moisture transport model for porous materials implemented in CFD, Building and
Environment, 81: 340–353, 2014.
33. Farid M., Handbook of Farm, Dairy and Food Machinery Engineering, 2nd ed., Academic Press,
Cambridge, MA, USA, Chapter 12, 2013.
34. Le Page J.F., Chevarin C., Kondjoyan A., Daudin J.D., Mirade P.S., Development of an approximate
empirical-CFD model estimating coupled heat and water transfers of stacked food products placed in
airflow, Journal of Food Engineering, 92(2): 208–216, 2009.
35. Darabi H., Zomorodian A., Akbari M.H., Lorestani A.N., Design a cabinet dryer with two geometric
configurations using CFD, Journal of Food Science and Technology, 52(1): 359–366, 2015.
36. Amanlou Y., Zomorodian A., Evaluation of air flow resistance across a green fig bed for selecting
an appropriate pressure drop prediction equation, Food and Bioproducts Processing, 89: 157–162,
2011.
37. Kashanineja K., Tabil L.G., Resistance of bulk pistachio nuts (Ohadi variety) to airfow, Journal of Food
Engineering, 90(1): 104–109, 2009.
38. Yue R., Zhang Q., A pore-scale model for predicting resistance to airflow in bulk grain, Biosystems
Engineering, 155: 142–151, 2017.
39. Olatunde G., Atungulu G.G., Sadaka S., CFD modeling of air flow distribution in rice bin storage system
with different grain mass configurations, Biosystems Engineering, 151: 286–297, 2016.
40. Elhalwagy M.M., Straatman A.G., Dynamic coupling of phase-heat and mass transfer in porous media
and conjugate fluid/porous domains, International Journal of Heat and Mass Transfer, 106: 1270–1286,
2017.
41. Khan F.A., Straatman A.G., A conjugate fluid-porous approach to convective heat and mass transfer
with application to produce drying, Journal of Food Engineering, 179: 55–67, 2016.
42. Başlar M., Karasu S., Kiliçli M., Us A.A., Sağdiç O., Degradation kinetics of bioactive compounds and
antioxidant activity of pomegranate arils during the drying process, International Journal of Food,
10(4): 839–848, 2014.
43. Wang Z., Sun J., Liao X., Chen F., Zhao G., Wu J. et al., Mathematical modeling on hot air drying of
thin layer apple pomace, Food Research International, 40: 39–40, 2007.
44. Darvishi H., Zarein M., Minaei S., Khafajeh H., Exergy and energy analysis, drying kinetics and math-
ematical modeling of white mulberry drying process, International Journal of Food, 10(2): 269–280,
2014.
45. Dandamrongrak R., Young G., Moson R., Evaluation of various pretreatments for the dehydration of
banana and selection of suitable drying models, Journal of Food Engineering, 95: 139–146, 2002.
46. Pabis S., Jayas D., Cenkowski S., Grain Driers Theory and Practice, John Wiley & Sons, Hoboken,
NJ, USA, 110–112, 1998.
47. Alibas I., Microwave, air and combined microwave-air drying of grape leaves (vitis vinifera l.) and the
determination of some quality parameters, International Journal of Food, 10(1): 69–88, 2014.
48. Tavakolpour H., Mokhtarian M., Neural network approaches for prediction of pistachio drying kinetics,
International Journal of Food, 8(3): 2012.
49. Sharma G.P., Verma R.C., Pankaj P., Mathematical modeling of infrared radiation thin layer drying of
onion slices, Journal of Food Engineering, 171: 28–286, 2005.
50. Balbay A., Engin A., Omer S., Coteli R., Modeling of drying process of bittim nuts (pistacia terebin-
thus) in a fixed bed dryer system by using extreme learning machine, International Journal of Food
Engineering, 8(4): 2012.
51. Defraeye T., Blocken B., Derome D., Nicolai B., Carmeilet J., Convective heat and mass transfer
modelling at air-porous material interfaces: Overview of existing methods and relevance, Chemical
Engineering Science, 74: 49–58, 2012.
52. Earle R.L., Unit Operation in Food Processing, 2nd ed., Pergamon Press, Oxford, UK, 1983.
53. Amanlou Y., Optimizing a solar cabinet dryer for Estahban green fig for uniform dehydration. Msc
Thesis, (In Persian), Shiraz University, Iran, 2009.
54. Misha S., Mat S., Ruslan M.h., Sopian K., Salleh E., Simulation of air flow distribution in a tray
dryer by CFD, In: 4th International Conference on Renewable Energy Research and Applications,
Italy, 2015.
Section III
CFD Applications in Other
Food Processes
20 CFD Design and
Optimization of Biosensors
in the Food Industry
Agnese Piovesan, Jeroen Lammertyn,
Bart Nicolai, and Pieter Verboven
CONTENTS
20.1 Introduction..................................................................................................................... 439
20.2 Mathematical Models of Biosensors.................................................................................... 441
20.2.1 Regime.................................................................................................................... 441
20.2.2 Flow Mechanisms....................................................................................................... 441
20.2.3 Governing Equations.....................................................................................................442
20.2.3.1 Pressure Driven Flow.................................................................................... 443
20.2.3.2 Electroosmotic Flow...............................................................................444
20.2.3.3 Capillary Flow......................................................................................... 445
20.2.3.4 Mass Transfer............................................................................................. 448
20.2.3.5 Reaction Kinetics...................................................................................... 449
20.2.3.6 Reduced Order Models............................................................................ 450
20.3 An Electrokinetic Biochip for Simultaneous Sugars Detection.......................................... 451
20.3.1 Validation of Reduced Order Models.........................................................................451
20.3.2 Lab-on-a-Chip Design............................................................................................ 452
20.3.3 Assay Description........................................................................................................453
20.3.4 Lab-on-a-Chip Optimization.................................................................................. 454
20.4 Future Perspectives.............................................................................................................. 456
20.5 Conclusions.......................................................................................................................... 458
Nomenclature.................................................................................................................................. 458
References....................................................................................................................................... 459
20.1 INTRODUCTION
In recent years concern and attention towards improved food quality and safety has greatly increased
both amidst producers and consumers. Food quality is mostly related to sensory properties, such
as taste, aroma, texture, appearance and, more recently, functionality; the main concern about food
safety is foodborne illness caused by microbiological pathogens such as bacteria, parasites, and
viruses. In addition, food safety is greatly affected by a number of harmful components, such as
allergens, toxins, heavy metals, pesticide residues and additives. Several analytical methods have
been developed in order to detect and identify these substances. The most commonly used methods
are liquid chromatography, gas chromatography, mass spectrometry and multistep assays. Although
they provide sensitive and precise results, the abovementioned techniques are tedious and time-
consuming; moreover, skilled personnel and expensive instrumentation are required, thus dras-
tically hampering the possibility of performing routine quality measurements on food products.
439
Therefore, there is a need for simple, rapid, accurate, portable and inexpensive devices to test food
in order to ensure quality and safety.
A solution is offered by biosensors. Biosensors are analytical devices able to detect and recog-
nize biochemical components and translate this detection into a measurable signal (Figure 20.1).
Bio-recognition molecules, such as enzymes, antibodies, aptamers, microorganisms or even whole
cells, interact with the target chemical component and an integrated physiochemical transducer [1]
converts this biochemical interaction into an electronic signal. Electrochemical, optical, thermal
and mass sensitive transduction mechanisms have been tested in biosensor development over the
past decade.
Biosensors have shown their potential in a number of applications, such as supervising food com-
positions (carbohydrates, triglycerides and amino acids) and freshness [2–4], as well as monitoring
aging and fermentation processes [5–7]. Moreover, biosensors have been applied to investigate food
safety in terms of foodborne pathogens [7–9], metal ions [10], residual pesticides [7,11] and anti
biotics [7,11–13].
Despite the demonstrated potential, commercialization of biosensors has not often been achieved;
this is mainly related to technical factors such as stability (the limited lifetime of the biological
component), mass production, storage and sensitivity (practicality in handling the sample) [14] that
must be optimized before any biosensor is commercialized. In order to ensure a successful practical
implementation, a few basic performance criteria need to be considered in biosensor design. These
include calibration characteristics (i.e., sensitivity, detection and quantitative determination limits,
operational and linear concentration range), selectivity and reliability, response time, high sam-
ple throughput, reproducibility, stability and operating life [15]. Moreover, as already mentioned,
the resulting biosensor should preferably be cheap, small portable and easy to use by semiskilled
operators.
Over the past years, great efforts have been made to fulfill these requirements through miniatur
ization. Miniaturization aims not only at increasing portability of biosensors but also at allowing
low-cost mass production and improving the analytical performance of the process [16,17]. There
are two important technical advantages of miniaturization. First, for a cylindrical microtube with
50 μm radius, the surface area to volume ratio reaches 4 × 10 –4 m–1, which is a huge interface for a
reaction to take place [16]. Second, since the diffusion time scales with the square of the travelling
distance of molecules, it is possible to significantly decrease the molecular diffusion time by han-
dling reduced volumes of fluids in small channels.
These advantages have opened a new field of research called microfluidics. Microfluidics is
defined as the science of designing, manufacturing and formulating devices and processes that deal
with volumes of fluid in the order of nanoliters (dimensions in order of micrometers), which is a key
factor when the necessary reagents are expensive. On the other hand, fluid motion through micro-
metric channels represents a major challenge in microfluidics. One of most common approaches to
FIGURE 20.1 Principle of a biosensor.

CFD Design and Optimization of Biosensors in the Food Industry 441
achieve fluid motion is applying pressure gradients [18]; however, small channels require external
pumps able to apply sufficiently high pressure gradients, thus limiting portability and increasing
costs. Another common approach consists of combining microfluidics with microelectronics
[18–20] to obtain the so-called electro-osmotic flow, but again external actuators are needed, affect-
ing the biosensor portability. A different solution for fluid flow is offered by capillarity-driven flow,
in which surface forces, which are utterly powerful in miniaturized devices, are exploited for spon-
taneous fluid flow [20]. Whatever fluid motion mechanism is chosen, optimization of the geometric
and operational parameters becomes a crucial step for biosensor development. Numerical prototyp-
ing by means of multiphysics computational fluid dynamics meets this need. Although many com-
putational fluid dynamics software packages are commercially available, some numerical models
are too complex and computationally expensive to be directly solved; in this case simplifications
need to be made, as in the case of reduced-order-models (ROM) [19].
This chapter will deal with aspects such as fluid flow, mass transfer, and chemical kinetics in
biosensors. The main focus will be on analytical microfluidic devices, also called lab-on-a-chip.
Through microfluidics it is possible to fabricate highly integrated devices with different functions
on one substrate chip. One of the long-term goals in the field of microfluidics is to create integrated,
portable diagnostic devices for home and bedside use, thereby eliminating expensive and time-
consuming laboratory analysis procedures.
This chapter will present the state-of-the-art models that are currently used in biosensor design,
followed by an explanatory example of a biosensor for glucose quantification; moreover, models that
are expected to be further implemented in the future will also be discussed.
20.2 MATHEMATICAL MODELS OF BIOSENSORS

20.2.1 Regime
The flow of a fluid through a microfluidic channel can be characterized by the Reynolds number,
which represents the ratio between inertial and viscous forces. Due to the reduced dimensions of
microchannels, the Reynolds number is usually smaller than 100 and often much smaller than 1,
meaning viscous forces are highly dominant. In this Reynolds number regime, the flow is com-
pletely laminar. This means that in long, straight, uniaxial channels the fluid flows parallel to the
channel walls in what is called uniaxial flow. The consequence of this is that transport of momen-
tum, mass and heat in the direction normal to the flow is not enhanced by turbulent eddies, but is
just achieved by molecular mechanisms: diffusion becomes the main method to move particles,
mix fluids, and control reaction rates in biochips [20]. On the one hand, this represents a problem
when fast mixing is needed; in this case specific solutions need to be applied, such as designing
the microchannels in a way to increase contact time or contact area between reagents. On the other
hand, as the channel dimension diminishes, the relative importance of surface and interfacial phe-
nomena (such as surface tension, roughness, and electrokinetic effects) increases. Exploiting these
latter phenomena in order to develop cheap and efficient fluid transport mechanisms is what makes
microfluidics an interesting and attractive field of study [7].
20.2.2 Flow Mechanisms
In about 90% of the developed microfluidic devices [18], fluid is transported through the channels
by applying a pressure or a voltage difference. The first is known as pressure-driven flow (PDF)
and, despite being customarily used, it has a few limitations when employed at the microscale.
Primarily, in small channels, PDF exhibits a parabolic velocity profile with near-zero velocity at
the channel walls. This seriously prevents devices with recognition elements on these walls from
working properly. Furthermore, the pressure drop is inversely proportional to the second power of
the transverse dimension of the channels and will consequently be very large and cause the method
to be impractical for some applications. The second method is referred to as electroosmotic flow
(EOF), which is based on electrokinetics [21]. Electrokinetics is a phenomenon that exploits the
interaction between solid surfaces (e.g., glass or a polymer-based substrate), ionic solutions, and
applied electric fields to actuate fluid motion. EOF is characterized by a plug-like flow profile,
which means the velocity profile is uniform in the cross-section of the microchannels and so is less
dispersive than in PDF. This method is well suited for miniaturization since expensive external
pumps and valves are not required. Nevertheless, EOF as well as PDF requires external devices—
in particular, electric field generators, which must be connected to a power source, thus affecting
portability.
An alternative solution of increasing popularity [20,22] consists of using surface forces to auton-
omously move the fluids. Surface forces become predominant in microchannels and can thus be
exploited to produce capillary flow. In capillary flow the surface interactions between the fluid and
the solid the channels are made of cause the fluid to move through the channels with no need for any
actuation device. Capillary forces also occur in porous substrates and this is the principle behind
the so-called paper-based microfluidics [22,23]. Capillary flow therefore offers great possibilities to
develop cheap, simple and portable microfluidic devices, even though the number of fluids that can
be handled is still smaller than for PDF and EOF.
Here the necessary equations to describe fluid flow and mass transfer in microchannels of biosen-
sors are presented. Considering a viscous, heat conductive fluid that can be approximated as a
continuum, the fluid motion is described by the Naviers-Stokes (N-S) equations. The N-S equations
originate from the momentum conservation and, for incompressible fluids, have the following form:
∂u
ρ + ρu • ∇u = − ∆p + η∇ 2u + ρg + Fb (20.1)
∂t
where ρ (kg m–3) is the fluid density, u (m s–1) is the flow velocity vector, p (Pa) is the pressure field,
η (Pa s) is the fluid viscosity, g (m s–2) is the gravitational acceleration and Fb is a term representing
all the other applied body forces.
The term on the left-hand side describes the flow acceleration, which is composed of time-
dependent and convective accelerations; the right-hand side of the equation collects all the body
forces acting on the fluid, such as gravity, pressure or viscous forces.
However, Equation (20.1) is not sufficient to fully describe the fluid motion and state; to do so it
must be coupled with the continuity (20.2) and energy (20.3) equations, which are derived by impos-
ing mass and energy conservation respectively:
∇u = 0 (20.2)
∂T  ∂p 
cpρ + c p ρu i ∇T − β T  + u i ∇p = k ∇ 2T + q + ηΦ (20.3)
∂t  ∂t 

where cp (kg m2 K–1 s–2) is the heat capacity, T (K) is the temperature, β (k–1) is the coefficient of
thermal expansion, k (W m–1 K–1) is the thermal conductivity, q (W m–2) is the conductive thermic
flux and ηΦ represents the energy dissipation due to the fluid viscosity.
If a steady state flow is considered, the previous equations are simplified by neglecting the
time-dependent terms.
In the particular case of microfluidics, where the Reynolds number is much smaller than 1,
Equation (20.1) is simplified, yielding the so-called Stokes equation:
0 = − ∆p + η∇ 2 u + Fb (20.4)

Coupling these equations with the proper boundary conditions allows the fluid motion to be
resolved; the velocity field is first computed and subsequently the other unknown variables are
obtained.
20.2.3.1 Pressure Driven Flow

The equation of motion for steady, low-Reynolds number pressure driven flow in a microfluidic
biosensor is
0 = − ∆p + η∇ 2 u (20.5)

Equation (20.5) can be computationally solved by applying the desired pressure gradient at the
microchannel ends and imposing a non-slip boundary condition at the channel walls.
The critical issues concerning PDF can be easily noticed by considering a simplified geometry
and solving (20.5) analytically. Examining fluid flow in a parallel plate channel (Figure 20.2), the
fluid velocity can be computed as:
∂P 1 2
u= ( y − L2 ) (20.6)
∂x 2η
where P(Pa) is the applied pressure, L (m) is the channel width and y (m) and x (m) are the distances
from and along the channel centerline, respectively.
The obtained velocity profile is parabolic with maximum velocity at the center of the channel
and zero velocity at the walls. Equation (20.6) can be written as function of the average velocity Uav:
3  y2 
u = U av −  2 − 1 (20.7)
2  L 

By applying the Newton law of viscosity, the shear stress, τ w (Pa), at the channel walls can be
calculated as:
3ηU av
τw = (20.8)
L
FIGURE 20.2 Pressure driven flow profile.

A non-dimensional measure of τ w, namely the friction coefficient Cf, can be computed as follows:
τw 12
Cf = = (20.9)
1 Re
ρU av
2
where Re stands for the Reynolds number.

The main consequences of PDF are, therefore, the following:
• Extremely disperse flow. Equation (20.6) shows that the velocity varies quadratically with
the distance from the walls, leading to a significant dispersion due to the great variations
between walls and centerline velocities. Moreover, this kind of flow is not suitable when
the purpose is performing a wall-bounded reaction, since the fluid tends to flow in proxim-
ity to the channel’s central axis rather than approaching the walls.
• High pressure gradients are needed to actuate the fluid flow. The friction coefficient Cf rap-
idly increases in case of very low Reynolds numbers, which are typical for flow in micro-
fluidic channels because of their micrometric dimensions. As a result, important pumping
powers are needed in order to move the fluid through the channels, requiring expensive and
bulky pumping devices that affect the biosensor’s cost and portability.
20.2.3.2 Electroosmotic Flow
The equation of motion for steady, low-Reynolds number electroosmotic flow in a microfluidic
biosensor is
0 = − ∆p + η∇ 2 u + ρE E (20.10)

where the last term on the right-hand side represents the body force on the fluid as a result of the
applied electric field to a microchannel with an electric double layer (EDL), which requires the
solution of the Poisson–Boltzmann equation, described below.
As most surfaces possess a negative charge, resulting from the ionization of the surface or the
adsorption of the ionic species, a layer of cations builds up near the surface when this surface comes
in contact with polar liquids to maintain the charge balance. This creates an EDL of ions near the
surface and a potential difference between the fixed charges on the wall and the diffuse charges of
the mobile ions called zeta potential (ζ). The EDL is resolved in two regions: a compact and a dif-
fusive layer (Figure 20.3).
The compact layer, which is a few Angstroms thick, is immobile due to a strong electrostatic
force, whereas the diffusive layer has a net charge different from zero. This layer will move when a
longitudinal electric field, E, is applied. As a result of the viscous force on the rest of the fluid, the
bulk starts to flow. This collective movement induces fluid motion in the channel, creating what is
called EOF. Therefore, the magnitude of the EOF depends on the electric field strength and the local
net charge density, which is a function of the EDL field. In simplified systems, namely assuming the
electrolyte to be binary and symmetric (z+ = z0– = z) and assuming densities of cations and anions to
be identical (n 0+ + = n 0– = n 0), the net charge density in the EDL is computed as:
 zeψ 
ρE = −2zen0 sinh  (20.11)
 k B T 

where kB (J K–1) is the Boltzmann’s constant, T (K) is the temperature, z is the valence number of
the ion, e (C) is the proton charge, n 0 (m–3) is the ionic concentration and ψ is the local potential.
FIGURE 20.3 Electrical double layer.
This charge density can be related to the local potential by the Poisson equation, assuming the rela-
tive permittivity εr to be constant:
ρE
∇ 2ψ = (20.12)
ε 0ε r
with ε 0 the permittivity of the free space. This combination yields the Poisson–Boltzmann equation:
2 zen0  zeψ 
∇ 2ψ = sinh  (20.13)
ε 0εr  k B T 

The thickness of the diffusive layer therefore depends on the bulk concentration and the electri-
cal properties of the liquid and not on the surface properties of the microchannels. The thickness of
the EDL is described by the Debye length, namely:
ε RT
λD = (20.14)
2 F 2C B

where F is the Faraday’s constant, CB is the ion concentration in the bulk (mol m−3) and R is the
universal gas constant (J mol–1 K–1). Since the EDL is commonly very thin, it can be represented by
a slip flow boundary condition at the wall for standard flow equations [24].
20.2.3.3 Capillary Flow
Capillary flow is determined by surface tension. In a two-phase system, such as liquid and gas, mol-
ecules from the bulk of the liquid are in equilibrium since they are equally pulled in any direction
by the neighboring liquid molecules. On the other hand, at the fluid-air interface, the pulling force
exerted by other liquid molecules is stronger than the one exerted by gas molecules, resulting in the
formation of a meniscus. The net force acting on the meniscus is counterbalanced by the surface
FIGURE 20.4 Contact angle in a three-phase system: gas, liquid, and solid.
tension γ (N m–1), which is defined as the force per unit length at the fluids interface and is the driv-
ing force leading to the reduction of the liquid surface area.
In a three-phase system of gas, liquid and solid, a contact angle θ is formed between the solid/
liquid and liquid/gas interface (Figure 20.4), defined as
γ sg − γ sl
θ= (20.15)
γ lg

in which γsg, γsl and γlg are the surface tension at the solid-gas, solid-liquid and liquid-gas interface
respectively. The contact angle is thus a function of the interaction between the three phases and
determines the solid material wettability; if a capillary tube with radius small enough [20] is char-
acterized by sufficient wettability, the interaction forces are able to propel the liquid through the
capillary against gravitational forces. This phenomenon is called capillary flow and the force mov-
ing the fluid along the capillary tube is referred to as capillary force.
In the case of a cylindrical tube with radius r 0 (Figure 20.5) a spherical meniscus with curvature
radius Rc
r0
Rc = (20.16)
cos θ
FIGURE 20.5 Capillary rise for a fluid with contact angle θ in a capillary tube with radius r 0.
is defined. The pressure in correspondence of the liquid-gas interface can then be computed accord-
ing to the Young–Laplace equation as:
2γ
p( H ) = p0 − cos θ (20.17)
r0
with P0 the atmospheric pressure and H the vertical meniscus position.

The force balance with gravitational forces can then be written as
p( H ) + ρ gH = p0 (20.18)

from which the spontaneous capillary rise of the fluid in the tube can be obtained from Jurin’s law:
2γ
H= cos θ (20.19)
r0 ρ g
By applying the previous equation to a water-air interface in a glass tube, it is possible to notice
the influence of the tube radius on the generated liquid rise. In a tube with radius of 2 m the water
will rise less than 1 mm, while in a tube of radius equal to 200 μm, the spontaneous capillary rise
will be almost 70 mm. In a microfluidic system, such effect may thus be sufficient to eliminate the
need for external fluid motion actuation.
Capillary forces are also present in porous media. In this case a common approach consists of
modeling porous materials as bunches of parallel cylindrical tubes; the spontaneous liquid pen-
etration can then be described as capillarity-driven phenomenon where gravitational forces can be
neglected. The velocity of the resulting capillary flow is obtained through the Lucas–Washburn [25]
equation, that in the simplified case of horizontal capillaries yields:
rγ t
l2 = cos θ (20.20)
2η
where l (m) in the penetration depth of the liquid, r (m) is the mean pore radius, η (Pa s) is the fluid
viscosity and t (s) is the time. A fluid ability to wick through a porous medium is therefore deter-
mined by the material properties, such as the liquid viscosity and surface tension, as well as the
porous medium wettability and pore dimensions.
All these parameters can be designed and engineered, by means for instance of surface modifica-
tions [26], in order to obtain a capillary flow with desired properties.
In CFD, capillary flow can be described applying the volume of fluid (VOF) method, a numerical
technique designed for tracking the fluid-fluid interface in multiphase flow [27–29].
In this method a single momentum Equation (20.1) is solved to obtain the velocity field; the two
present phases (liquid and air) are considered to be a single medium with a discontinuity at the inter-
face. The material properties in (20.1) are replaced by effective material properties, computed in
each cell of the computational domain considering the volume fraction (α) of the liquid-air system.
ρ and η are then calculated as:
ρ = α l ρl + (1 − α l ) ρa (20.21)
η = α lηl + (1 − α l )ηa (20.22)
where the index l stands for the liquid and index a indicates the air phase.
The position of the interface is tracked by solving the continuity equation for the volume fraction:
∂α l
+ ∇(α l u) = 0 (20.23)
∂t
And by computing αa as:
αa = 1 − αl (20.24)
The surface tension is accounted for as a body force rather than a boundary condition acting on the
interface; Equation (20.1) is therefore rewritten as follows:
∂u
ρ + ρu * ∇u = − ∆p + η∇ 2u + ρg + Fs (20.25)
∂t
where ρ and η are obtained from (20.20) and (20.21), and Fs stands for the surface tension repre-
sented through a continuum surface force model [30]:
ρχ∇α l
Fs = γ (20.26)
1
( ρl + ρa )
2
In the previous equation, χ is the surface curvature computed as
k = −∇( n) → χ = −∇( n) (20.27)
where n is the normal unit vector to the interface, defined as
∇α l
n= (20.28)
∇α l

The contact angle θ is used to correct the normal unit vector in the cells near the walls
n = nt sin θ + n w cos θ (20.29)

nt and nw stand for the unit vectors tangent and normal to the wall respectively. The interface cur-
vature and orientation in the walls proximity are therefore corrected taking the contact angle into
account.
20.2.3.4 Mass Transfer
General mass transfer is governed by the convection–diffusion–reaction equation:
∂Ci
+ ∇(− Di∇Ci − z µepCi∇ψ + Ci u) = Ri (20.30)
∂t
where Ci (mol m–3) is the component concentration in the bulk fluid, Di (m2 s–1) the diffusivity and
Ri (mol m3 s–1) denotes the reaction term that is discussed in the next section. The velocity vector u
is equal to the velocity of the solvent. This can be the result of any of the previously mentioned fluid
flow mechanisms or a combination of all of them. The equation also includes a term for electropho-
retic transport. Applying electrical fields to ionic solutions induces migration transport in addition
to the existing diffusion and convection processes. Migration implies that positive ions migrate
from a positive potential to a negative potential along the direction of the electric field and vice versa
for negatively charged ions. μep (m2 V–1 s–1) is the ionic mobility or electrophoretic mobility.
20.2.3.5 Reaction Kinetics
This chapter will only consider enzyme kinetics. Nevertheless it is important to notice that recently
application of hybridization kinetics [31] in food biosensors has also increased and is exploited for
antibody-antigen interactions and detection or DNA and heavy metal ions [32,33].
In an enzymatic reaction the substrate (S) is converted into a product (P) with the help of the
enzyme (E):
S E
→P
The rate of reaction RP can be expressed in terms of either the change of substrate concentration
CS or the product concentration CP:
dCs dC p
Rp = − = (20.31)
dt dt
It is important to know how the reaction rate is influenced by reaction conditions such as substrate,
product and enzyme concentration. Figure 20.6 shows a typical curve in which the enzymatic con-
version rate is depicted as a function of substrate concentration, given a fixed enzyme concentration.
It can be observed that the reaction rate is proportional to the substrate concentration (first-order
reaction) at low values of substrate concentration, and does not depend on the substrate concentra-
tion (zero-order reaction) at high values of substrate concentration, which means the reaction goes
gradually from first-order to zero-order as the concentration of the substrate is increased (Figure
20.6). This behavior can be described by the well-known Michaelis–Menten kinetics:
V max Cs
V = Rp = (20.32)
K M + Cs
FIGURE 20.6 Reaction rate of an enzymatic reaction as function of substrate concentration.

with Vmax (mol m–3s–1) and K M (mol m–3) the maximum reaction rate and the Michaelis–Menten con-
stant, respectively, which need to be experimentally determined. The maximum reaction rate Vmax
is proportional to the enzyme concentration. This equation describes many experimental results
well. Leonor Michaelis and Maud Menten proposed in 1913 a quantitative theory to support the
observed enzyme kinetics that is still widely used today. Nevertheless, enzyme kinetics has been
extensively studied since then and other kinetics models have been proposed. The interested reader
is then referred to [34,35].
20.2.3.6 Reduced Order Models

Computational fluid dynamics (CFD) is a valuable tool for biosensors design since it permits
one to study variables that cannot be experimentally obtained, clarify experimental results,
determine design feasibility and relatively quickly investigate the effect of design changes on
the final results. In spite of the precise results provided by these numerical simulations, CFD is
generally used to refine nearly-final designs rather than for the entire optimization process of
the complex lab-on-a-chip. The reason for this is that a number of phenomena at different length
scales must be taken into consideration, ranging from the electric double layer thickness of nm
to the channel length of cm. Moreover, several physics phenomena must be modeled, such as
surface tension, electrokinetics and reaction kinetics. Because of this multiscale and multiphys-
ics framework, full three-dimensional computation models are still too time-consuming and
expensive to be used to evaluate the performances of microfluidic devices with complex chan-
nel lay-out, or to test large sets of variables, which is required during an optimization proce-
dure. Compact or reduced order models (ROM) represent a valid alternative to quickly evaluate
microfluidics systems while maintaining a sufficient level of accuracy. In ROM flow equations
are further simplified by representing microfluidics channels through their electric analog and
channel intersections are assumed to obey Kirchhoff laws of current conservation. The channel
network is therefore replaced with the analog integrated electrical circuit. For EOF, the electri-
cal resistance of the electrolyte solution R, in a straight channel with cross-sectional area AC (m 2)
and length L (m), is given by:
L
R= (20.33)
σ Ac
where σ (S m–1) stands for the electrical conductivity of the channel.

The electric current can then be related to the applied potential through the Ohm’s law as:
∆Ψ
I = σ Ac (20.34)
L
Eventually, in the case of homogenous straight channel properties, the fluid flow in each channel
is proportional to the current flow:
∆Ψ
Q = µeo Ac (20.35)
L
with μeo (m2 V–1 s–1) the electroosmotic mobility.

By applying the Kirchhoff law for mass and electric charge conservation at all channel connec-
tions [36], potentials to be applied in order to obtain the desired flow rates can be accurately deter-
mined even in devices with complex channel lay-out.
The ROM approach is therefore used to move from a complex 3D multiphysics environment to a
simplified one-dimensional model of a lab-on-a-chip. In this case the description of mass transport
is also simplified and Equation (20.17) can be rewritten as:
∂Ci ∂ 2 Ci ∂Ci
− Di + ui = Ri (20.36)
∂ ∂x 2 ∂x
where ui = ueo + uep,i is the total flow of species, which is a combination of electroosmotic flow and
electrophoretic flow, respectively. In microfluidic systems, mixing of different species is mostly due
to molecular diffusion. Because of this, very long mixing channels might be required; neverthe-
less it was proved that in microchannels with width up to 50 μm diffusive mixing is sufficient to
trigger enzymatic reactions [24]. In this case Equation (20.23) is valid and can be used to describe
mass transport and reaction in microfluidic systems under the condition that homogenization is not
required to start the chemical reaction.
20.3 AN ELECTROKINETIC BIOCHIP FOR

SIMULTANEOUS SUGARS DETECTION
This section elaborates upon an example of using models to design a microfluidic chip, consider-
ing different mechanisms explained in the previous sections and comparing different approaches,
such as CFD and ROM. Driven by the constantly increasing interest in food safety and quality, new
technologies are required for fast, cheap and in-situ analysis of food components. The gold-standard
techniques, such as liquid and gas chromatography, produce accurate results but are expensive, time-
consuming and require fully-equipped laboratories and well-trained personnel. Miniaturization and
automation of bioassays represent appropriate approaches for developing automated, fast and cheap
analytical devices while maintaining sufficient accuracy. CFD and ROM were compared to design
and optimize a multiplexed multi-enzyme assay microfluidic device for the simultaneous detection
of glucose, sucrose and fructose. Detection of these sugars in food samples is highly important since
they largely influence taste; moreover, monitoring sugars presence is desirable to identify illegal
sugars added to wine and fruit juice.
20.3.1 Validation of Reduced Order Models

The purpose of the presented study was to optimize the amount of enzyme and the applied potential
for maximum substrate conversion; this was ultimately achieved by using a ROM approach. Prior
to this, the reduced order modeling approach had to be validated by comparing ROM, CFD and
experimental results of an enzymatic assay performed on a designed lab-on-a-chip [36]. The chosen
model enzyme assay was β-galactosidase (β-gal), which catalyzes the conversion of Resorufin β-D-
galactopyranoside (RBG) in a fluorescent product.
Mass transport in ROM is only modeled one-dimensionally and complete mixing across the
channel width of different species is assumed to happen instantaneously; the latter is a strong
assumption when dealing with laminar flow and diffusive mass transport. A 2D CFD model that
was developed by applying the equations reported previously was used to investigate the effect of
this assumption. As depicted in Figure 20.7 the enzyme concentrations at different distances down-
stream the mixing point were compared with the results of the ROM; it was found that for micro-
channels that are long enough (millimeters), complete mixing is a valid hypothesis.
Figure 20.8 shows the normalized signal registered by the detector for different β-gal con-
centrations, namely 0.352, 0.685 and 1.37 nM. The solid lines are ROM simulation results and
experimental values are identified by different markers. The diamond marker indicates results of
the 2D CFD model. A good overall agreement was found between simulation and experimental
FIGURE 20.7 Profile of enzyme concentration at different location along the microfluidic channel after
mixing with the substrate.
FIGURE 20.8 Comparison of experimental, CFD and ROM results at different β-gal concentrations.
results except at the beginning of the experiment, which could not be explained. Differences
between CFD and ROM results were negligible but the ROM was 64 times faster in terms of CPU
time.
ROMs were therefore proven to provide accurate results while reducing computational time, thus
representing a valid solution for biosensor modeling and optimization.
20.3.2 Lab-on-a-Chip Design
Microfluidics allows integrating sample injection, mixing, reaction, and detection of the final
product on a single device, requiring minimal human intervention. A miniaturized biosensor was
developed for simultaneous detection, in a single microreactor, of glucose, sucrose and fructose,
which are important components determining food taste. The chip consists of two cross channels
and has a total of six reservoirs. NADH is produced as by-product and is optically detected at
460 nm.
Glucose and fructose are directly detected on the original sample, while sucrose is identified on
a sample previously treated with invertase. The design of the lab-on-a-chip is reported in Figure
20.9(a) while Figure 20.9(b) shows the ROM scheme.
FIGURE 20.9 (a) 2D representation of the lab-on-a-chip channels lay-out. All the channels are 50 µm wide
and 20 µm deep. (b) Equivalent circuit representation.
20.3.3 Assay Description
Samples are filled in reservoir 2, for glucose and fructose detection, and in reservoir 3 for sucrose
analysis. A buffer with two dissolved coenzymes, ATP and NAD+, is deposited at reservoir 1. The
purpose of the mixing channel is to allow mixing of the samples with coenzymes; in the reaction
channels samples are mixed with enzymes from reservoir 4 and 5. In reservoir 4 hexokinase (HK)
and glucose-6-phosphate dehydrogenase (G6PDH) are present, while reservoir 5 is filled with phos-
phoglucose isomerase (PGI), HK and G6PDH.
The sugars are analyzed in sequence to avoid cross-contamination: glucose is analyzed first,
followed by sucrose and eventually fructose. Each assay is divided in three steps: the sample is
loaded in the mixing channel, substrate and enzyme are then injected in the reaction channel and
finally the buffer is injected in both to push all the involved species downstream and to avoid
cross-contamination of the successive analysis. For the glucose assay, a predefined volume of sample
is injected from reservoir 2 and mixed with the buffer from reservoir 1 and HK and G6PDH from
reservoir 4. NADH is formed as a by-product and its quantity is equivalent to the original quantity of
glucose. Downstream, a fluorescence microscope at a wavelength of 460 nm is employed to detect
NADH. Subsequently, to quantify the amount of sucrose, the sample previously treated with inver-
tase is loaded in reservoir 3 and mixed again with buffer and enzymes respectively from reservoirs
1 and 4. The amount of produced NADH is the sum of what is produced by the glucose originally
present in the sample and previously measured, and by the glucose derived from the transformation
of sucrose by means of invertase. The concentration of sucrose can then be quantified by subtracting
the concentration of glucose measured in the previous assay from the total concentration of glucose
measured in the sucrose assay. Finally, in the last step, an extra enzyme, phosphoglucose isomerase
(PGI), is added to HK and G6PDH in reservoir 5. The chemical reaction chains lead first to the
formation of glucose and subsequently to the production of NADH. By subtraction again, the total
concentration of fructose can therefore be determined.
20.3.4 Lab-on-a-Chip Optimization
The flow directions and flow rates in the channels are controlled by the potential applied to the
electrodes in the reservoirs; a high-voltage sequencer is used to control the value and the duration
of the applied potential. The maximum flow rate in the reaction channel is set to 72 nL/min, cor-
responding to an electric field of 250 V cm–1, to avoid possible bubble formation.
A 2D CFD model is developed in order to determine whether the length of the mixing and reac-
tion channels is sufficient to ensure homogenous mixing of the involved species. Injection of glu-
cose and HK in the mixing and reaction channel, respectively, are simulated and the concentration
profiles are examined in a portion corresponding to the two crossings. Figure 20.10(a) shows the
contour plots of glucose and HK in the simulated channel portions. No difference is found between
concentration profiles extracted at three different locations (center line and in correspondence of
the channel wall) at the end of the mixing channel, thus implying the length of the mixing channel
is sufficient since glucose diffuses quickly enough.
Figure 20.10(b) depicts the concentration profiles of NADH produced for glucose, sucrose and
fructose, 20 mm downstream the reaction channel obtained both from the 2D and 1D model. There
is no difference between the simulation results of the two models, thus confirming that diffusion
mixing is sufficient to ensure the start of the enzymatic reaction in the microchannel. As a conse-
quence, Equation (20.23) can be used to describe mass transport and reaction in the lab-on-a-chip
with enough accuracy while reducing the computational time by a factor of 15.
The ROM (Figure 20.9(b)) is employed to optimize the bioassay in terms of flow rate and enzyme
concentration and the obtained results are reported below.
The enzyme quantity in the lab-on-chip reaction channel depends on the length of injection from
the reservoir, normally called injection time, and the concentration of enzyme loaded in the reservoir.
In Figure 20.11 the normalized NADH concentration versus the injection time and versus the
enzyme concentration is shown. At a flow rate of 72 nL/min, 1.2 nL of enzyme solution, of which
40% is coming from the enzyme reservoir, requires 2.5 s to be injected. It is found that for a low
FIGURE 20.10 (a) Contour plots of glucose and HK as they are injected in the mixing and reaction chan-
nel respectively. No difference is shown between normalized concentration of glucose at three different points
(center line and closer to the walls) at the end of the mixing channel, meaning that glucose and fructose are com-
pletely homogenized at the moment they enter the mixing channel. (b) Concentration profiles of NADH with a
flow rate of 72 nL/min for glucose, sucrose and fructose assays. The results of the 2D and 1D show a good match.
FIGURE 20.11 (a) Normalized NADH peak area versus injection time at flow rate of 72 nL/min and enzyme
concentration of 3.42 g/L HK and 3.52 g/L G6PDH. (b) Normalized peak area of NADH versus enzyme con-
centration where the sample has 300 µM of both glucose and fructose, total flow rate in the reaction channel
is still 72 nL/min.
substrate concentration (<100 μM), additional quantity of enzyme has no influence on the output
signal, meaning that the enzyme concentration is sufficient to achieve maximum conversion. On
the other hand, when higher substrate concentrations are present (e.g., 1 mM in Figure 20.11(a)), the
signal increases with the enzyme concentration.
It is known [37] that linear optical detection ranges for glucose, sucrose and fructose are, respec-
tively, 25–833 μM, 20–666 μM and 0.538–7.3 μM, which are much smaller than the values usually
present in food samples. To solve this issue, samples can be diluted directly on the chip by increas-
ing the buffer flow rate.
In Figure 20.11(b) the normalized peak area of NADH concentration as function of enzyme con-
centration (0.06–5.8 g/L) is reported; it is found that the amount of produced NADH will not change
beyond enzyme concentration of 3.42 g/L.
However, the optimal enzyme concentration changes with both flow rate and amount of substrate
in the sample, and thus optimization in terms of enzyme concentration and flow rate is necessary.
Several simulations were carried out with a fixed sample and enzyme solution volume of, respec-
tively, 1.1 nL and 1.2 nL, for different values of enzyme concentration and flow rate, in order to iden-
tify the condition allowing the fastest detection while requiring the minimum amount of reagents.
The simulations results are shown in Figure 20.12, in which the normalized amount of NADH
produced by the enzymatic reaction is plotted as function of flow rate and enzyme concentration.
FIGURE 20.12 Contour plot of the normalized peak area of NADH produced from 100 µM of glucose at
different flow rates and enzyme concentrations combinations.
For all the considered sugars, production of NADH increased with lower flow rate and higher
enzyme concentration. For flow rates lower than 30 nL/min and enzyme concentrations higher than
1 g/L, production of NADH is less dependent on enzyme concentration, meaning that the available
enzyme has enough time to consume the same quantity of substrate. A low flow rate can be used
to reduce enzyme concentration, but higher flow rates reduce the assay time when the detector is
sensitive enough.
If a flow rate of 72 nL/min and the highest enzyme concentration of 3.42 g/L are chosen, both
the assay time and the required amount of enzyme are greatly reduced with respect to classic micro-
plate assays.
According to simulation results, glucose, sucrose and fructose are quantified in about 2.5 min
and the required enzyme amount is reduced by 81%.
For a substrate concentration up to 1 mM a linear detection range is achieved, working with a
flow rate of 27 nL/min and enzyme concentrations of 3.42 g/L, 3.52 g/L and 1.44 g/L, respectively,
for HK, G6PDH and PGI.
The linear detection range can be further widened by reducing the flow rate; at the lowest consid-
ered flow rate of 14.4 nL/min, the lowest quantity of samples can be detected but at the expense of a
higher assay time, which would still be lower than with a conventional microplate assay.
This study evidenced the benefits of computational modeling for microfluidic biosensors opti-
mization. Developing proper and fast computational models, the effects of different geometrical
and operating parameters can be investigated and understood more efficiently with respect to the
classical prototype-based optimization. Computational models allow the assay performances to
be evaluated in different conditions without actually producing any lab-on-a-chip or executing
any experimental tests. In this way the required time and costs of the optimization procedure can
be enormously reduced. In the presented study the process conditions, such as concentration of
enzyme, flow rates and injection time, and geometrical parameters, such as the channel lengths,
were optimized to ensure the maximum output signal in the shortest possible assay time.
20.4 FUTURE PERSPECTIVES
The application above considered electrokinetics as a means of controlling fluid flow in microfluidic
platforms for bio-assays. Although EOF is widely used in microfluidics [18] because of its plug-
profile flow and reduced dispersion, it still requires external devices like electrodes and electric
fields generators that need to be connected to a power source.
Capillary flow, on the other hand, allows motion of fluid to be performed without an external
actuation device. Capillary flow has been widely exploited in the so-called paper-based microfluid-
ics [22], also referred to as lateral-flow microfluidics. In this branch of microfluidics, paper, which
is a porous medium and thus able to autonomously move liquids due to capillary forces, is the main
material used to produce cheap, fast, portable and stand-alone devices.
Despite its interesting potential, paper-based microfluidics is mainly exploited to perform later-
flow assays [38,39], of which pregnancy sticks are a common example, that are less complex than
multistep enzymatic assays like the one illustrated in the previous section. These more complex
assays have not yet been successfully combined with paper-based microfluidics, mainly due to its
two-dimensional nature, which doesn’t allow multistep assays to be implemented. Adding a third
dimension might therefore represent a step towards the implementation of lab-on-chips with fluid
flow completely driven by capillary forces. In literature some attempts are reported in which stacks
of functionalized porous membrane [40] are used to build 3D microfluidic platforms characterized
by capillary flow but an unpredictable functionality is highlighted. A promising solution is offered
by powder-based 3D printing [41], through which porous microfluidics platforms with complex
channels design can be realized.
Not only are 3D porous lab-on-chips difficult to produce, but they are also complex to model
numerically. The complex porous structure would require a huge number of volumetric elements to
be properly described and therefore the realization of computational models of whole lab-on-chips
is not feasible. As a solution a multiscale approach can be used; in a multiscale approach models at
a bigger scale (e.g., device scale, cm) are implemented by using averaged parameters extracted from
models at a lower scale (e.g., pore scale μm). For instance, a small volume of the porous geometry
can be modeled to extract constitutive parameters such as permeability, wettability and contact
angle, that are then plugged into the Washburn equation (Equation 20.20) to describe fluid flow in
the microfluidic device.
Particular attention is now being paid to the modeling of porous media [42]; the precise porous
geometry can be extracted by means of non-destructive imaging, such as X-Ray imaging, and sub-
sequently fluid flow can be directly simulated through CFD on the pore space. Nonetheless, this is
computationally expensive, and this is why a good solution consists of simplifying the complex pore
geometry through pore network modeling (PNM).
In PNM [43] pores and pores connections are replaced by simplified geometries, such as spheres
and cylinders. PNMs have been used successfully in soil science to study capillary flow [44,45].
Two main categories of PNMs exist, single phase models and multiphase models. Single phase
models work in stationary conditions and assume the pore space to be completely saturated with
incompressible fluid. In this situation the porous medium is described by porous bodies of a certain
volume and with hydraulic pressure pi, connected by channels in which the fluid flow is controlled
by viscous forces. The flow rate through a channel connecting two pores, i and j, is computed
according to the following equation:
( pi − p j )rij4
Qij = Sn (20.37)
ηlij

where Qij (m3 s−1) is the volumetric flow rate, Sn is a shape factor and rij and lij are respectively the
radius and the length of channel connecting the two pores. If the channels are modeled as cylinders,
then Equation 20.23 yields the Poiseulle Law:
( pi − p j )π rij4
Qij = (20.38)
8ηlij
Since the fluid is modeled as incompressible, mass conservation is imposed at internal pores,
namely:
∑Q ij =0 (20.39)
i
As a result, a linear system of equation is obtained that, once solved, gives the hydraulic pressure
in each pore. The overall volumetric flow rate through the boundaries is then computed and the
object permeability is estimated by applying Darcy’s law [44]. Application of multiphase models is
more cumbersome since dynamic effects, such as pore filling and air trapping, must be accounted
for. In the case of multiphase flow both viscous and capillary effects must be modeled; invasion
of an element (pore or throat) depends on its capillary pressure, while the invasion time rate is
determined by both viscous and capillary forces. These models are further divided into quasi-static
and dynamic models. In the former, the order of the elements to be invaded is only determined by
their capillary pressure, namely by the size of the connection throats. On the contrary, the latter
provide a full dynamic description of the fluid flow; the position of the fluid-fluid interface is itera-
tively uploaded allowing the modeling of both viscous and capillary phenomena but requiring more
computational time. Implementation of predictive multiphase pore network models is still an issue,
but they have been successfully used to study capillary flow and extract parameters that are hard to
obtain experimentally, such as the substrate relative permeability [37,38].
Parameters obtained from PNMs can subsequently be used to implement device scale continuum
models of microfluid devices that can be employed in the optimization process.
In the near future, thus, numerical design and optimization of biosensors will allow the produc-
tion of 3D microfluidic platforms powered by capillary flow, solving the issue of portability that is
still present for both EOF and PDF.
20.5 CONCLUSIONS
Biosensors have great potential and appeal in food industries, especially as online detection meth-
ods. Clear advantages over conventional methods are the high-sampling rates, increase of analysis
speed and reduction of the amount of sample and reagents needed. The development and commer-
cialization of biosensors for food quality and safety control depends on the elimination of some of
the shortcomings, such as short-term stability, low-measurable range of analytes, shelf life of the
biosensor, and ability to deal with situations resulting from different conditions of temperature and
pH. Computational prototyping is a valuable tool since it reduces the cost linked to the realiza-
tion of many suboptimal prototypes and permits rapid determination of the relevant and optimal
design parameters. CFD is a powerful tool for computational prototyping but is generally too time-
consuming and expensive to be used for modeling entire microfluidic devices. Different approaches,
such as ROM, and in the future, multiscale and pore network modeling, allow proper optimization
to be performed in significantly shorter time while maintaining a sufficient level of accuracy. In this
chapter the role of the correct choice and description of transport phenomena, both in terms of fluid
flow and mass transfer, has been stated. Moreover, the potential of miniaturization coupled with
computational optimization for the development of biosensors for the food industry was highlighted
through the description of an implemented lab-on-a-chip for sugar detection.
NOMENCLATURE
Ac Cross sectional area (m2)
Ci Concentration of specie I (mol m–3)
Cf Friction coefficient (-)
cp Heat capacity (kg m2 K–1 s–2)
Di Diffusivity of specie I (m2 s–1)
E Electric field (V m–1)
E Enzyme
e Charge of a proton (C)
F Faraday’s constant (C mol–1)
I Electric current (A)
K M Michaelis–Menten constant (mol m–3)
kb Boltzmann constant (J K–1)
n Normal vector
n 0 Bulk concentration of ion species (mol m–3)
P Product
p Pressure (Pa)
pi Hydraulic pressure at pore i (Pa)
Q Volumetric flow rate (m3 s–1)
q Conductive thermic flux (W m–2)
R Electric resistance of solution in the channels (Ω)
Rg Gas constant (J mol–1 K–1)

Ri Reaction rate (mol m–3 s–1)
S Substrate
Sn Shape factor
T Temperature (K)
t Time (s)
u Velocity vector (m s–1)
uep,i Electrophoretic flow velocity of specie i (m s–1)
Uav Average velocity (m s–1)
V Reaction rate (mol m–3 s–1)
Vmax Maximum reaction rate (mol m–3 s–1)
z Ion valence number
αi Volume fraction of phase i
β Coefficient of thermal expansion (k –1)
γ Surface tension (N m–1)
ε Permittivity (F m–1)
η Dynamic viscosity (kg m–1 s–1)
θ Contact angle (°)
λD Characteristic EDL thickness (m)
µeo Ionic mobility (m–2 V–1 s–1)
ρE Net charge density (C m–3)
σ Electrical conductivity (S m–1)
τ Shear stress (Pa)
χ Surface curvature
ψ Electrical potential (V)
REFERENCES
1. Larroche C., Sanroman M., Du G., Pandey A., Eds., Current Developments in Biotechnology and
Bioengineering, Elsevier, Amsterdam, Netherlands, 2017.
2. Monosik R., Stredansky M., Tkac J., Sturdik E., Application of enzyme biosensors in analysis of food
and beverages, Food Analytical Methods, 5: 40–53, 2012. doi:10.1007/s12161-011-9222-4.
3. German N., Ramanaviciene A., Voronovic J., Ramanavicius A., Glucose biosensor based on graphite
electrodes modified with glucose oxidase and colloidal gold nanoparticles, Microchimica Acta, 168:
221–229. 2010. doi:10.1007/s00604-009-0270-z.
4. Silva N.F.D., Magalhães J.M.C.S., Freire C., Delerue-Matos C., Electrochemical biosensors for
Salmonella: State of the art and challenges in food safety assessment, Biosensors and Bioelectronics,
99: 667–682, 2017. doi:10.1016/j.bios.2017.08.019.
5. Farina D., Zinellu M., Fanari M., Porcu M.C., Scognamillo S., Puggioni G.M.G., Rocchitta G.,
Serra P.A., Pretti L., Development of a biosensor telemetry system for monitoring fermentation in craft
breweries, Food Chemistry, 218: 479–486, 2017. doi:10.1016/j.foodchem.2016.09.092.
6. Monošík R., Magdolen P., Streďanský M., Šturdík E., Monitoring of monosaccharides, oligosaccha-
rides, ethanol and glycerol during wort fermentation by biosensors, HPLC and spectrophotometry,
Food Chemistry, 138: 220–226, 2013. doi:10.1016/j.foodchem.2012.10.039.
7. Weng X., Neethirajan S., Ensuring food safety: Quality monitoring using microfluidics, Trends in Food
Science and Technology, 65: 10–22, 2017. doi:10.1016/j.tifs.2017.04.015.
8. Singh A., Poshtiban S., Evoy S., Recent advances in bacteriophage based biosensors for food-borne
pathogen detection, Sensors (Basel), 13: 1763–1786, 2013. doi:10.3390/s130201763.
9. Arora P., Sindhu A., Dilbaghi N., Chaudhury A., Biosensors as innovative tools for the detection of food
borne pathogens, Biosensors and Bioelectronics, 28: 1–12, 2011. doi:10.1016/j.bios.2011.06.002.
10. Amaro F., Turkewitz A.P., Martín-González A., Gutiérrez J.C., Whole-cell biosensors for detection
of heavy metal ions in environmental samples based on metallothionein promoters from tetrahymena
thermophila, Microbial Biotechnology, 4: 513–522, 2011. doi:10.1111/j.1751-7915.2011.00252.x.
11. Gaudin V., Biosensors and bioelectronics advances in biosensor development for the screening of
antibiotic residues in food products of animal origin—A comprehensive review, Biosensors and
Bioelectronics, 90: 363–377, 2017. doi:10.1016/j.bios.2016.12.005.
12. Suárez G., Jin Y.-H., Auerswald J., Berchtold S., Knapp H.F., Diserens J.-M., Leterrier Y., Månson
J.-A.E., Voirin G., Lab-on-a-chip for multiplexed biosensing of residual antibiotics in milk, Lab on a
Chip, 9: 1625, 2009. doi:10.1039/b819688e.
13. Scognamiglio V., Arduini F., Palleschi G., Rea G., Biosensing technology for sustainable food safety,
TrAC—Trends in Analytical Chemistry, 62: 1–10, 2014. doi:10.1016/j.trac.2014.07.007.
14. Scognamiglio V., Pezzotti G., Pezzotti I., Cano J., Buonasera K., Giannini D., Giardi M.T., Biosensors
for effective environmental and agrifood protection and commercialization: From research to market,
Microchimica Acta, 170: 215–225, 2010. doi:10.1007/s00604-010-0313-5.
15. Abdullah S., Uddin M., Introduction to food biosensors, Introduction to Food Biosensors, 1–21, 2016.
doi:10.1039/9781782623618-00001.
16. deMello A.J., Control and detection of chemical reactions in microfluidic systems, Nature, 442: 394–
402, 2006. doi:10.1038/nature05062.
17. Whitesides G.M., The origins and the future of microfluidics, Nature, 442: 368–373, 2006. doi:10.1038
/nature05058.
18. Pennathur S., Flow control in microfluidics: Are the workhorse flows adequate?, Lab on a Chip, 8: 383,
2008. doi:10.1039/b801448p.
19. Atalay Y.T., Witters D., Vermeir S., Vergauwe N., Verboven P., Nicolaï B., Lammertyn J., Design and
optimization of a double-enzyme glucose assay in microfluidic lab-on-a-chip, Biomicrofluidics, 3: 1–14,
2009. doi:10.1063/1.3250304.
20. Chakraborty S., Microfluidics and Microfabrication, Springer-Verlag, New York, USA, 2010.
21. Polson N.A., Hayes M.A., Electroosmotic flow control of fluids on a capillary electrophoresis microdevice
using an applied external voltage, Analytical Chemistry, 72: 1088–1092, 2000. doi:10.1021/ac9912698.
22. Li X., Ballerini D.R., Shen W., A perspective on paper-based microfluidics: Current status and future
trends, Biomicrofluidics, 6, 2012. doi:10.1063/1.3687398.
23. Li X., Tian J., Garnier G., Shen W., Fabrication of paper-based microfluidic sensors by printing, Colloids
Surfaces B Biointerfaces, 76: 564–570, 2010. doi:10.1016/j.colsurfb.2009.12.023.
24. Atalay Y.T., Verboven P., Vermeir S., Vergauwe N., Delport F., Nicolaï B.M., Lammertyn J., Design
optimization of an enzymatic assay in an electrokinetically-driven microfluidic device, Microfluidics
and Nanofluidics, 5: 837–849, 2008. doi:10.1007/s10404-008-0291-7.
25. Washburn E.W., The dynamics of capillary flow, Physical Review, 17: 273–283, 1921. doi:10.1103/Phys
Rev.17.273.
26. Xi Z.Y., Xu Y.Y., Zhu L.P., Wang Y., Zhu B.K., A facile method of surface modification for hydrophobic
polymer membranes based on the adhesive behavior of poly(DOPA) and poly(dopamine), Journal of
Membrane Science, 327: 244–253, 2009. doi:10.1016/j.memsci.2008.11.037.
27. Delele M.A., Nuyttens D., Duga A.T., Ambaw A., Lebeau F., Nicolai B.M., Verboven P., Predicting the
dynamic impact behaviour of spray droplets on fat plant surfaces, Soft Matter, 12: 7195–7211, 2016.
doi:10.1039/C6SM00933F.
28. Raeini A.Q., Blunt M.J., Bijeljic B., Modelling two-phase flow in porous media at the pore scale using
the volume-of-fluid method, Journal of Computational Physics, 231: 5653–5668, 2012. doi:10.1016/j
.jcp.2012.04.011.
29. Raeini A.Q., Blunt M.J., Bijeljic B., Direct simulations of two-phase flow on micro-CT images of porous
media and upscaling of pore-scale forces, Advances in Water Resources, 74: 116–126, 2014. doi:10.1016/j
.advwatres.2014.08.012.
30. Brackbill J.U., Kothe D.B., Zemach C., A continuum method for modeling surface tension, Journal of
Computational Physics, 100: 335–354, 1992. doi:10.1016/0021-9991(92)90240-Y.
31. Yang D., Tang Y., Miao P., Hybridization chain reaction directed DNA superstructures assembly
for biosensing applications, TrAC—Trends in Analytical Chemistry, 94: 1–13, 2017. doi:10.1016/j
.trac.2017.06.011.
32. Cai W., Xie S., Zhang J., Tang D., Tang Y., An electrochemical impedance biosensor for Hg2+ detection
based on DNA hydrogel by coupling with DNAzyme-assisted target recycling and hybridization chain
reaction, Biosensors and Bioelectronics, 98: 466–472, 2017. doi:10.1016/j.bios.2017.07.025.
33. Zhang Y., Li H., Chen M., Fang X., Pang P., Wang H., Wu Z., Yang W., Ultrasensitive electrochemi-
cal biosensor for silver ion based on magnetic nanoparticles labeling with hybridization chain reac-
tion amplification strategy, Sensors and Actuators B: Chemical, 249: 431–438, 2017. doi:10.1016/j
.snb.2017.04.108.
34. Johnson K.A., A century of enzyme kinetic analysis, 1913 to 2013, FEBS Letters, 587: 2753–2766, 2013.
doi:10.1016/j.febslet.2013.07.012.
35. Wong M.K.L., Krycer J.R., Burchfield J.G., James D.E., Kuncic Z., A generalised enzyme kinetic
model for predicting the behaviour of complex biochemical systems, FEBS Open Bio, 5: 226–239,2015.
doi:10.1016/j.fob.2015.03.002.
36. Berli C.L.A., Theoretical modelling of electrokinetic flow in microchannel networks, Colloids and Surfaces
A: Physicochemical and Engineering Aspect, 301: 271–280, 2007. doi:10.1016/j.colsurfa.2006.12.066.
37. Vermeir S., Nicolaï B.M., Verboven P., Van Gerwen P., Baeten B., Hoflack L., Vulsteke V., Lammertyn J.,
Microplate differential calorimetric biosensor for ascorbic acid analysis in food and pharmaceuticals,
Analytical Chemistry, 79: 6119–6127, 2007. doi:10.1021/ac070325z.
38. Naik L., Sharma R., Mann B., Lata K., Rajput Y.S., Surendra Nath B., Rapid screening test for detec-
tion of oxytetracycline residues in milk using lateral flow assay, Food Chemistry, 219: 85–92, 2017.
doi:10.1016/j.foodchem.2016.09.090.
39. Bahadır E.B., Sezgintürk M.K., Lateral flow assays: Principles, designs and labels, TrAC—Trends in
Analytical Chemistry, 82: 286–306, 2016. doi:10.1016/j.trac.2016.06.006.
40. Lutz B., Liang T., Fu E., Ramachandran S., Kauffman P., Yager P., Dissolvable fluidic time delays for
programming multi-step assays in instrument-free paper diagnostics, Lab on a Chip, 13: 2840, 2013.
doi:10.1039/c3lc50178g.
41. Shirazi S.F.S., Gharehkhani S., Mehrali M., Yarmand H., Metselaar H.S.C., Adib Kadri N., Osman
N.A.A., A review on powder-based additive manufacturing for tissue engineering: Selective laser
sintering and inkjet 3D printing, Science and Technology of Advanced Materials, 16: 33502, 2015.
doi:10.1088/1468-6996/16/3/033502.
42. Bultreys T., De Boever W., Cnudde V., Imaging and image-based fluid transport modeling at the pore
scale in geological materials: A practical introduction to the current state-of-the-art, Earth-Science
Reviews, 155: 93–128, 2016. doi:10.1016/j.earscirev.2016.02.001.
43. Valvatne P.H., Piri M., Lopez X., Blunt M.J., Predictive pore-scale modeling of single and multiphase
flow, Transport in Porous Media, 8: 23–41, 2005. doi:10.1007/1-4020-3604-3_3.
44. van Marcke P., Verleye B., Carmeliet J., Roose D., Swennen R., An improved pore network model for
the computation of the saturated permeability of porous rock, Transport in Porous Media, 85: 451– 476,
2010. doi:10.1007/s11242-010-9572-1.
45. Sorbie K.S., Skauge A., Can network modelling predict two-phase flow functions?, In: Proceedings of
the International Symposium of the Society of Core Analysts, 1–12, 2011.
21 Analysis and Simulation
of Pasta Dough Extrusion
Process by CFD
Fabrizio Sarghini
CONTENTS
21.1 Introduction..........................................................................................................................463
21.2 Pasta Extrusion Process.......................................................................................................464
21.2.1 The Physics of Pasta Extrusion..............................................................................465
21.2.2 CFD in Pasta Extrusion........................................................................................... 467
21.2.3 Equations.................................................................................................................... 467
21.2.3.1 Mixing and Rheological Models of Pasta Dough................................. 468
21.2.3.2 Equations in Screw Press....................................................................... 473
21.2.3.3 Model Validation.................................................................................... 475
21.2.4 Use of CFD for Extrusion Optimization Process: Optimal Shape Design (OSD)......478
21.3 Conclusions............................................................................................................................... 485
References....................................................................................................................................... 485
21.1 INTRODUCTION
Pasta production is generally considered a mature industrial process, given the worldwide diffu-
sion of the final product and market acceptance. Nonetheless, as in many affirmed food processes,
several problems are still present in industrial production, as the scale-up from small equipment to
pasta lines with a capacity up to 5,000 kg/h was mostly based on a geometrical scaling approach.
As a consequence, pasta dough radial pressure distribution in industrial scale die is not uniform
into the extrusion channels, especially in configuration for short pasta, resulting in uneven extru-
sion velocities and eventually in quality differences in the same production batch. To mitigate this
phenomenon, a hand-made pressure compensation plate with different radial die channel diameters
was introduced, often developed with a heuristics approach. The same problem also arises in long
pasta production.
Moreover, the high viscosity and the non-Newtonian character of pasta dough make possible the
presence of some recirculation zones into the extrusion bell, where molds could generate.
As a matter of fact, the geometries of the extruder bells, apart from small changes, are still similar
to those designed thirty years ago. The presence of previously mentioned hand-made pressure com-
pensation disks to correct flow uneven velocities in the die shows that reduced scale experimental
pilot plants can not completely substitute for the full-scale experiments, due to implicit difficulties
in scaling-up a material with complex rheological properties at different production rates.
After the work of Le Roux and Vernier [1], very little attention has been paid to the process
itself, and industrial plants continue to suffer from the aforementioned problems. The importance
of a homogeneous dispersion in water–semolina mixing has been recognized, and something has
been done by introducing in-line mixers rather than using the traditional paddle mixers, to reduce
the effects on the final products of nonhomogeneous water–semolina distribution, but still design
details of several critical parts are almost the same.
463
A powerful optimization tool, when properly used, is represented by modeling and numerical simula-
tions: such approaches, coupled with a posteriori experimental evidence (validation), can be considered
the modern basis of any process optimization. While 0D and 1D modeling can provide partial knowl-
edge of extrusion process, 2D and particularly 3D models using CFD can be extremely beneficial.
Numerical and experimental comparisons represent a powerful reciprocal validating tool.
Moreover, a reliable numerical model introduces the possibility of reverse engineering optimization
of the pasta final product, i.e., designing the extruder’s die from the shape of the final product.
Although a lot of work has been done in general on numerical modeling of polymer extrusion,
flour dough extrusion and semolina dough extrusion still remain partially investigated topics: only
two simulations were published by Dhanasekharan and Kokini [2], and by Fabbri and Lorenzini
[3], mainly because of the complexity of rheological models involved, and only recently was a com-
prehensive framework published. Sarghini et al. [4], validated experimentally and numerically in
the industrial production a range of process variables.
21.2 PASTA EXTRUSION PROCESS

Extrusion is a thermomechanical process especially tailored for mass production, nowadays applied
to a number of food products, like snacks, ready-to-eat cereals and several other starch-based foods.
It is an inline integrated low-cost process with many unique advantages; it is energy efficient,
usually organized in multiple inputs/single output without collateral waste effluents, and provides
versatility with respect to ingredient selection, shapes and textures of final products.
A correct combination of variables involving product formulation, process and equipment setting
are required to obtain the final product.
Given the extensive number of variables involved, extrusion is a complex multivariate process,
quite challenging to control.
The pasta extrusion process can be generally divided into three functional processes: ingredients’
mixing, pasta dough polymerization in the screw, followed by the final pasta production in the die section.
Although mixing takes place outside the extruder itself, a homogeneous composition of the
mixed product is very important to control the quality of the final product and to establish the rheo-
logical characteristics of pasta dough.
In the first part of the extruder, hydrated semolina can be considered a granular flow and is
exposed to compression and thermal and mechanical stresses generated by the pressure agglomera-
tion and rotation of the screws.
Thermomechanical effects in the screw section strongly affect the rate of these reactions and
therefore play a decisive role on the resulting structural and thermomechanical properties of the
plasticized pasta dough, which is eventually forced to flow through the die section (Figure 21.1) [5].
VARIABLES PROCESS PRODUCT

MIXING SCREW SECTION DIE SECTION
Extrusion Variables Thermomechanical processing

thermal stress profile
Process variables mechanical stress profile
screw speed, feed rate,
barrel temperature
Machine parameters Mixing Structuring &

screw geometry and Relative Humidity, polymerization Final shaping at
configuration, die/barrel spatial distribution Thermomechanical die’s inserts
geometry of RH properties T,p,flow rate,
viscosity, elasticity, viscous dissipation
Structure composition
protein/carbohydrates/fat
Tg, melting point,
type and content, specific heat Structure
water content (physical &
chemical)
FIGURE 21.1 Pasta dough extrusion process. (Adapted from Emin, M.A. 2015. Modeling extrusion pro-
cesses. In: Modeling Food Processing Operations, Woodhead Publishing Series in Food Science, Technology
and Nutrition, edited by Serafim Bakalis, Kai Knoerzer and Peter J. Fryer, Woodhead Publishing, pp. 235–253.)
Analysis and Simulation of Pasta Dough Extrusion Process by CFD 465
The physical and chemical structure of the material through various mechanisms, such as
gelatinization, melting, fragmentation/degradation, protein denaturation, lipid oxidation, and
formation of flavors are influenced by generated stresses, especially due to barrel wall/screw
interaction [6–9].
Strong interactions between mass, energy and momentum transfer, coupled with complex
physico-chemical transformations governing thermomechanical properties, are present, making it
difficult to control the extrusion process.
While pilot plants can provide experimental evidence of the real product behavior, measuring the
thermal and mechanical stresses and/or material properties in extruders is a difficult and expensive
task if performed on real scale plant. This is why the possibility of modeling and simulating the
pasta extrusion process can provide a better understanding of the mechanisms involved.
On the other hand, modeling should be used together with experimental tests to validate the
approach. Users and developers of computational modeling and simulations today must face a criti-
cal issue about how confidence in modeling and simulation be assessed.
The methodology approach relies on verification and validation of computational simulations as
the primary methods for building and quantifying this confidence.
Briefly speaking, verification is the assessment of the accuracy of the solution to a computational
model by comparison with known solutions. Validation is the assessment of the accuracy of a com-
putational simulation by comparison with experimental data. In verification, the relationship of the
simulation to the real world is not an issue: the congruence with a reliable physics is the key point.
In validation, the relationship between computation and the real world, i.e., experimental data,
is the issue. Stated differently, verification is primarily a mathematics issue, while validation is
primarily a physics issue [10].
Sometimes we do not have the opportunity to test completely our model with exact solutions, as
the knowledge of the physics involved is partial, and the extrusion process of complex biopolymers
involving structural changes is one of these cases. We will discuss in detail problems involved in
rheological characterization of pasta dough in a following section.
With modeling alone, in many cases, it is not possible to predict exactly what the material prop-
erties will be at the end of extrusion. An empirical-experimental approach, therefore, cannot be
replaced by modeling alone.
Nevertheless, extrusion modeling and simulation provides many benefits when coupled with reli-
able empirical data.
Among them, we can mention:
– Understanding the effects of single variables;

– Characterization and control of the physics of specific mechanisms (e.g., mixing, reac-
tion kinetics);
– Monitoring and control of processing conditions (pressure, thermal and mechanical
stress profiles);
– Possibility of optimizing extruder design and scale-up;
– Improved understanding of the process to be used for offline training of personnel.
In this chapter the framework for a global numerical simulation of pasta extrusion is provided
together with a validation evidence of the numerical model using experimental data obtained from
a laboratory scale extrusion system (a laboratory scale Sercom press).
21.2.1 The Physics of Pasta Extrusion

The schematic of a pasta dough extrusion system is described in Figure 21.2.
In order to give the pasta the final desired shape, the dough in its polymerized plastic state is forced
through an appropriate die at relatively high pressure, and in the case of continuous extrusion press,
z
WATER SEMOLINA
v scre w
v barrell wall
PADDLE
OR
INLINE
MIXER INSULATED
FEED COOLING
HOPPER SCREW PRESS
JACKET
DRIVE
AND
GEARS
FEED ZONE SOLID CONVEYING ZONE MELTING ZONE METERING ZONE

EXTRUSION
BELL
HOMOGENIZATION
GRID
COMPENSATION
DISK
DIE’S INSERTS
FIGURE 21.2 Schematic of pasta dough extrusion system.
this pressure is generated by a rotating screw. This very important phase of extrusion has a decisive
influence on the quality of the end product and is subsumed under the term “rheology” of the extru-
sion screw. This refers to the physical and technological behavior of the dough. The mixing of the raw
materials—durum semolina and water—creates a protein framework in the dough that depends on the
protein content and quality of the semolina, the retention time in the mixer and the mixer efficiency.
The protein framework is able to provide the final product with a certain resistance to overcooking
and retention of the desired nutritional value, and it should be preserved as much as possible during the
extrusion. This is possible if the kneading of the dough in the screw is gentle, in particular in the meter-
ing zone, and the screw profiles and speeds adopted today are designed to meet these requirements.
Sarghini et al. [11], showed that after polymerization takes place, mixing is minimal if not neg-
ligible in the barrel and pasta dough mostly exhibits a plug-flow behavior.
At the end of this process the most evident result is the structural homogenization of the mixed
material under pressure, and the protein framework that was already created during the mixing
process is further reinforced. While pressure alone does not have any significant damaging effect on
the protein framework, the shear forces occurring mainly during extrusion have a negative influence
on the material and should therefore be prevented to the extent possible. Such shear forces usually
occur when the dough is forced to flow over sharp edges of plate holes, die inserts, etc.
The screws used for pasta production basically have two major zones, the dough intake and pres-
sure build-up zone, and the actual pressure and kneading zone.
These zones can be further detailed; for example, in a single-screw extruder, from a modeling
point of view, up to four fundamentally different regions can be identified:
1. The feed zone, where the screw channel is not completely filled and loose powder is gradu-
ally consolidated to a coherent solid mass;
2. The solids conveying zone where powder is compacted and transported to a good approxi-
mation as a plug;
3. The melting zone, where the material begins to melt and form a paste;
4. The metering or pumping zone, where the material that has been melted or fused reaches
its maximal temperature and pressure.
Due to the toughness (viscosity) of a dough, with a water content of approximately 30%–34%, a
pressure between 80 and 120 bar must be generated to shape the pasta by the press screw, depending
on the die resistance. The dough pressure depends on a number of different factors:
– dough moisture;
– die resistance;
– die protection devices;
– compensastion plate geometry if present;
– dough temperature;
– and, obviously, the screw press geometry.
The influence of the dough moisture content (MC) is widely known. The pressure drops as the
dough moisture rises, and vice versa. There are certain upper and lower limits to the moisture
range, usually between 30 to 34%. Dough with an excessive moisture content results in a sticky
product, in particular in the first drying phase. If the dough is too dry, a temperature rise is created
by the increased extrusion pressure, normally resulting in a degradation of quality and colour (white
specks).
21.2.2 CFD in Pasta Extrusion

Computational fluid dynamics simulations can provide information about flow characteristics that
are difficult, if not impossible, to obtain using experimental measurements. Such detailed informa-
tion on flow characteristics allows us to model the key process mechanisms in extruders, such as
reactions and mixing, determining the final product properties. The typical workflow of computa-
tional simulation involves the following steps:
1. Experimental or bibliographic data acquisition of necessary material properties (e.g., vis-

cosity, thermal conductivity, etc.) required for calculations.
2. Experimental data fitting to transform experimental data into suitable mathematical
models.
3. CAD generation of the computational domain according to the geometry of the extruder
barrel, screws, extrusion bell and dies depending on the problem of interest.
4. Mesh preparation according to the numerical method and required accuracy (it must be
cross checked by grid independence analysis).
5. Solving the governing equations to obtain profiles of variables of interest.
6. Post-processing and calculation of desired physical quantities, such as the viscous dissipa-
tion, mechanical stress profile or mixing characteristics from numerical solution.
7. Validation of the physical reliability of the simulation by comparing results with the exper-
imental data.
21.2.3 Equations
Some general assumption can be adopted in numerical simulation of the pasta dough extrusion
process:
1. The Reynolds number in pasta dough extrusion flow (e.g., starch or proteins) is very small
because of the high viscosity of melts, and therefore the flow is assumed to be laminar and
the fluid is incompressible.
2. The simulated domains using Navier–Stokes equations are assumed to be fully filled.
3. The flow is considered to be 3D in a Cartesian framework.
21.2.3.1 Mixing and Rheological Models of Pasta Dough

A key point in pasta dough simulations is represented by a detailed knowledge of rheological prop-
erties, as a detailed determination of physical properties is a preliminary step in modeling.
Since in complex phenomena such as pasta extrusion, the possibility to achieve validation of
numerical results with experimental data is linked not only to the numerical model, optimal design
and detailed analysis of the experimental setup of both rheological tests and extrusion processes are
required.
In pasta extrusion, semolina/water mixing is a preliminary process the effects of which strongly
influence the final results [12].
To this purpose, prerequisite to a correct numerical simulation is the knowledge of rheological
properties of semolina dough in the industrial range of temperature and moisture content (MC). A lim-
ited number of works have investigated in general the rheological properties of pasta dough in order to
obtain a better understanding of the basic rheology of mixing: among them, the works of LeRoux and
Vernier [1], de La Pena et al. [13], and more recently an extensive investigation of these properties in
real production ranges of variability of temperature and MC published by Sarghini et al. [4].
Rheological characterization of cereal doughs using a capillary rheometer have been
described in literature [14–17]. Most of the studies focused on bread dough, while the behavior
of semolina dough during extrusion has not been extensively investigated [1,18]. A more recent
work of de la Pena et al. [13], partially closed the gap, although the authors did not cover all the
ranges of temperature and RH involved in industrial production, required for numerical simula-
tions. Rheological experiments show that pasta dough is a viscoelastic system, exhibiting com-
plex rheological behavior. The work of de la Pena et al. [13], focused on the effects of moisture
and dough formulation on rheological properties of pasta dough during extrusion at fixed tem-
perature. The authors experienced difficulties in experimental setup related to flow instabilities
caused by heterogeneous hydration pressure oscillations or slip at the capillary. This makes it
difficult to obtain repeatable reliable results, particularly when using small diameter capillar-
ies, unless a meticulous preparation of samples’ hydration is performed. A careful preparation
of experimental conditions, and particularly mixing conditions, was considered in the work of
Sarghini et al. [4], which tried to minimize the pressure oscillation effects and obtain reliable
and reproducible results. A typical experimental diagram of the apparent viscosity vs. shear rate
is shown in Figure 21.3.
As underlined in the work of Hatzikiriakos and Dealy [19], the rise time required to achieve a
steady pressure reading in a capillary rheometer operated at constant piston speed can be very short
or very long depending on the amount of the material in the barrel, its isothermal compressibility,
FIGURE 21.3 Apparent viscosity experimental results for T = 35°C at different values of MC.
the flow rate and the geometrical characteristics of the dies used. When a short die having a large
diameter is used, a maximum can be obtained in the pressure transient, solely due to the com-
pressibility of the material. In any case, experimental results in the work of Sarghini et al. [4],
and in the work of de la Pena et al. [13], at 32% RH and 45°C are quite similar. Rheological
measurements were performed at shear rates in the range 10−400 s−1, corresponding to the range
(∞ 200 s−1) in extrusion operations performed with the laboratory extrusion unit used to validate
results (a Sercom laboratory extrusion unit is described in Figure 21.4).
Although consistent experimental data can be obtained experimentally, it is not straightfor-
ward to develop a unique generalized rheological model taking into account temperature-depen-
dent structural modifications due to starch gelatinization, relative humidity effects and shear rate
influence.
A review of dough rheological models used in numerical applications is presented in the work
of Hosseinalipour et al. [20]. Most of the applications are based on the power-law models, due
to their ability to predict velocity and pressure distributions in uniform flows and represent the
simplest representation of shear thinning behavior. However, they encounter the following short-
comings: these models should be applied over only a limited range of shear rates and therefore
the fitted values of k and n will depend on the range of shear rates [21]; the dimensions of the flow
consistency coefficient, k, depend on the numerical value of n and therefore the k values should
not be compared when the n values differ [21]; they are seldom able to provide accurate predic-
tions of measurements of Specific Mechanical Energy (SME) and Residence Time Distribution
(RTD) [22].
A major drawback of the mentioned rheological models in numerical applications is connected
to the fact that they are not able to predict zero and infinite shear viscosities [21]. Mackey et al.
[23], proposed a model for the viscosity of starch-based products related by using a modified model
from Morgan et al. [24]. This model provides accurate results for predicting the viscosity only
of relatively pure starches such as potato flour [23] and corn starch [25] and does not account for
non-starch components. In the model the effects of shear rate, temperature, moisture content, time-
temperature history, and strain history are incorporated [25]. In the aformentioned work on rheo-
logical analysis of whole wheat flour doughs, results showed a certain lack of accuracy, attributed by
the authors to the presence of flour multiple components such as bran, protein, and lipids, which the
FIGURE 21.4 Schematic of a Sercom press laboratory extrusion unit.

model does not account for and which probably altered the starch gelatinization kinetics. As a mat-
ter of fact, viscosity in pasta dough computed with the Mackey and Ofoli model differ significantly
from experimental rheological data previously obtained in the work of Sarghini et al. [4], and in the
work of de la Pena et al. [13].
Although the temperature varies during extrusion operations in a small range, (35°C–50°C), a
larger temperature interval (35°C–80°C) was explored to gain some information about macroscopic
effects of thermal denaturation of gluten on the mixture rheological response. This is because tem-
perature may change inside the extruder barrel due to viscous dissipation effects, as shown in Figure
21.8 (reported in Section 21.2.3.3 for validation with numerical simulation data), for an extrusion
experiment in which the thermostatic bath was set at 45°C. The trend can be clearly identified, show-
ing that viscosity increases as MC decreases: for example, comparing the fractions at MC = 36% and
at MC = 32%, the former has a viscosity values less than one order of magnitude with respect to the lat-
ter. This sensitivity of the rheological response to the variation of the water content of the semolina
dough can be observed at all considered temperatures. It is accepted that the lubricant role is played
by gliadine because its globular shape allows the movement of the glutenin fibers constituting the
gluten network that entraps wheat granules; the water content influences the degree of network
formation, and in an indirect way plays the role of lubricant, too. Within an industrial production
range, an increase of MC involves a reduction of viscosity. Conversely, the water, within the range
explored, does not seem to interfere with the rheological behavior of pastes that always appear to
be pseudoplastic, following a power-law model, in accordance with results of LeRoux and Vergnes
[1], and de la Pena et al. [13]. In general one can notice that with increasing moisture content and
temperature, the viscosity of the dough decreases; the larger the moisture content; the smaller the
influence of the temperature. The same conclusions can be drawn by considering the parametric
representation shown in Figures 21.5 and 21.6, where the flow index, n, and the consistency, k, rela-
tive to the mixture having different moisture content, were plotted against temperature. Increasing
the temperature, the flow index increases until T = 55–60°C, and then it decreases. This behavior is
related to structure formation processes starting at temperatures above T = 50°C–60°C, which can
reflect in a negative way on the pasta final quality from an industrial point of view.
Consistency index k and flow index n are tabulated for use in numerical simulations in Table 21.1,
while std. deviations are plotted in Figures 21.5 and 21.6.
At this point a complete set of rheological data in industrial production parameters range was
obtained using the experimental test, and can be used to perform a numerical simulation of the real
extrusion process.
Understanding the complex behavior of dough-like materials and interpreting it as one general
equation requires a deep knowledge of the characteristics and formation dynamics of this complex
FIGURE 21.5 Power law k consistency coefficient.

FIGURE 21.6 Power law flow index n coefficient.
TABLE 21.1
Consistency Index k (kPa sn) and Flow Index n with Respect
to Moisture Content (MC) and Temperature
MC T(C) k n T(C) k n
30% 35 109340.70 0.3984 40 119257.38 0.3365
32% 68209.72 0.4020 54333.71 0.4160
34% 27958.86 0.4191 17924.02 0.4393
36% 17066.97 0.4372 11778.92 0.4229
30% 45 81949.88 0.3586 50 41705.23 0.4510
32% 27919.99 0.4888 23307.17 0.4997
34% 9838.48 0.4974 9758.88 0.4936
36% 8879.85 0.4338 6641.63 0.4778
30% 55 33042.81 0.4540 60 34738.72 0.4051
32% 18623.81 0.4910 20992.37 0.4242
34% 9821.87 0.4508 13463.74 0.4293
36% 7372.91 0.4229 9356.96 0.4349
30% 80 33642.60 0.2910
32% 29476.20 0.3071
34% 22777.39 0.3282
36% 15238.78 0.3551
material. While a power law can represent at fixed condition of temperature and relative humid-
ity the rheological behavior, determination of a general unifying law including dependance on
shear rate, temperature and relative humidity is an overhelming task, and as of today nobody has
succeeded.
A possible alternative for numerical simulations is based on the hypothesis that rheological prop-
erties exhibit a continuous trend in the range of interest and under the mentioned assumption they
can be expressed using a local Taylor expansion.
A multivariate data fitting model cross-matching experimental data represents a sufficient and
somehow more accurate approach to provide essential data in an industrial production range of
temperature and MC for numerical modeling of pasta dough rheology. This is because what we
really need in a numerical simulation is the correct apparent viscosity in time and space. For this
reason, in the proposed approach, temperature (and MC if required) effects on viscosity μ are
expressed by using a multivariate third order truncated Taylor expansion relating local shear stress
and temperature:
3
 ∂n µ (γ )  (T − Tref )n
µ (T , γ ( x )) = µ0 (T , γ ) + ∑
n =1

∂T n 

n!
(21.1)

where x is the coordinates vector.
Unknown derivatives ∂n µ (γ )/∂T n and µ0 (T , γ ) can be obtained algebraically (solving sym
bolically the set of equations) by solving for each point x a 4 × 4 linear system at fixed MC imposing

( ) (
µ Ti , γ ( x ) = µ Ti , γ )exp
, Ti = {35°C, 40°C, 45°C, 55°C}} (21.2)
where γ ( x) is obtained by numerically solving Navier–Stokes equations during the simulation and
it is related to the second invariant of the deformation gradient tensor (the symmetric part of the
velocity gradient tensor) S:
γ ( x) = ( 2S : S ) (21.3)
where v(x) is the local velocity vector, and
1  ∂vi ∂v j 
Sij = + (21.4)
2  ∂x j ∂xi 

The numerical procedure to solve the system is the following:
1. for each time step and each control volume compute numerically the temperature and the
local average shear rate γ ( x);
2. using the experimentally derived power-law model
τ = kγ n (21.5)
compute apparent viscosity at different temperatures ( Ti = {35°C, 40°C, 45°C, 50°C})

(using a cubic expansion in temperature we select four temperatures levels out of six up to
60°C, depending on local temperature, in order to perform a centered interpolation and not
extrapolate values);
3. impose the computed values in Taylor expansion equation at four selected temperatures to
match exactly the experimental values;
4. solve the 4 × 4 linear system to obtain the first order term and the three unknown deriva-
tives of the Taylor expansion;
5. solve the Navier–Stokes equations with the new apparent viscosity at computed tempera-
ture and restart the cycle.
Using the previously described approach in each point the local Taylor expansion matches exactly
the experimental viscosity values at the same shear rate and experimental temperatures, as the
constitutive relation between shear rate and stress tensor remains the power-law model. A special
consideration is devoted to pasta dough behavior at low shear rate, as the model predicts an infinite
viscosity at zero shear rate. Close to zero shear viscosities were obtained using creep measurements
at vanishing shear rates in the work of Dus and Kokini [26], for hard wheat flour dough. While
in the experimental analysis inside a cylindrical barrel this situation does not appear, in numerical
analysis of real extruder bell configuration very low shear rate zones can be present. Due to the
relevant differences between flour dough and semolina dough, a cut-off level for low shear rate was
arbitrarily set to 1 in numerical simulations. This resulted in a continuous regular distribution of
viscosity, as shown in Figure 21.17.
21.2.3.2 Equations in Screw Press

Different governing equations are required in different zones depending on the physics involved.
The transport of hydrated semolina as powder in the feeding zone is comparable to the conveying
of polymer powder, and the solid conveying zone and the melting zone can be modeled using the
model derived by Darnell and Mol [27], and improved by Tadmor and Klein [28]. Although several
other improved models are present in literature [29], for our purpose the original formulation can
be used, where the balance of pressure and friction forces on the solid determines the increase in
pressure ΔP on a slab of thickness Δz according to Equation 21.6:
  f W + 2 H   
∆P = P0 exp   1 (cos(φ + γ ) − f2 sin(φ + γ )) − f2 ∆z − 1 (21.6)
 H WH   

where P0 is the pressure in the preceding slab, f1 and f 2 are the friction coefficients on the barrel and
screw (Coulomb wall boundary conditions: the friction coefficients on the barrel and screw surfaces
are, in general, different due to different surface machining and temperatures), H and W are the
screw and internal barrel channel dimensions respectively, γ is the helix angle and ϕ is the solid
conveying angle (see Figure 21.7), defined by the volumetric flow rate Q as:
 sin γ 
Q = U 0WH  cos γ − (21.7)
 tan (φ + γ ) 

where U0 is the longitudinal component of the barrel velocity.
FIGURE 21.7 The velocities and angles used to model pressure in solid conveying zone of the extruder,
where vextruder denotes the motion of the plug relative to the engine as a whole, vscrew denotes the motion of the
plug in the down channel direction relative to the screw, and vbarrel is the velocity difference between the barrel
and the internal plug.
In the proposed framework a value f1 = 1 was adopted to take into account the rugosity of the
internal surface of extruder barrel, while for the screw f 2 = 0.35 was assumed (see LeRoux and
Vergnes [1]) for a discussion about these hypotheses). To identify the beginning of the metering
section, we followed the work of Le Roux and Vernier [1], assuming that the transition between
powdery hydrated semolina and homogeneous dough occurred in the screw when the local pressure
was higher than a characteristic value of 2 MPa, which turned out to be in proximity of the eighth
turn of the screw, so that only 10 out of 18 real extruder turns were numerically modeled for CFD
analysis. In the work of Sarghini et al. [4], this assumption was verified by interrupting an experi-
mental run, extracting the screw and the dough together, and checking the temperature (necessary
to setup correct boundary conditions at the inflow) and the degree of compaction of the semolina
inside. Given the uncertainty in setting the real pressure at the exit of the compression and melting
part of the screw computed with the Tadmor model, corresponding to the inflow of the numerical
domain, mass flow rate at the inflow and screw speed were instead set matching experimental data,
and temperature and pressure in the bell were used for comparison. The pressure inside the bell to
be compared with the experimental probe was computed setting the value of the numerical pressure
to 20 atm at the inflow; in the following results, the beginning of the metering section is assumed to
be the same for all mass flow rates, although a more detailed application of the proposed approach
would require repetition of the same procedure for each mass flow rate. The no slip approxima-
tion at the barrel was adopted as barrel of the extruder was grooved to minimize slip phenomena;
nonetheless a small slip component at the barrel wall is always present, but lack of experimental
data allowed adoption of the no slip assumption. The main force responsible for dough heating is
the viscous dissipation, an always-positive term describing the conversion of mechanical energy to
heat, which can be computed instantaneously as:
 ∂u  2  ∂u  2  ∂u  2 
τ : ∇v = 2µapp  1  +  2  +  3   +
 ∂x1   ∂x 2   ∂x3  
 ∂u (21.8)
∂u  
2 2 2
∂u   ∂u ∂u   ∂u
µapp  2 + 1  +  2 + 3  +  1 + 3   = 0
 ∂x1 ∂x 2   ∂x3 ∂x 2   ∂x3 ∂x1  

For example, only the metering zone was simulated using CFD, in order to use only the continu-
ous Navier–Stokes equations and drop the further difficulty of modeling the compacting phenom-
ena using a different set of equations (granular compressible flows).
Although the complete system is non-homogeneous, continuous equations can be applied in
the final part of the extruder barrel, where compression effects allow one to consider the system
homogeneous from an engineering point of view. The reference continuity, momentum and energy
equations for the general non-Newtonian case are the following:
∂ρ
+ ∇ ⋅ ( ρU ) = 0 (21.9)
∂t
∂
( ρU ) + ∇ ⋅ ( ρU ⊗ U ) = −∇p + ∇ ⋅ τ + ρ g (21.10)
∂t
( ρe) + ∇ ⋅ ( ρUe) = ∇ ⋅ ( k∇T ) − p∇ ⋅ U + τ : ∇U (21.11)

∂
∂t
p
where e = h − is the internal energy, h is the sensible enthalpy, U is the velocity vector, τ is the
ρ
stress tensor related to the strain rate by:
 2 
τ = µ  ∇U + (∇U )T − δ∇ ⋅ U  (21.12)
 3 

and τ: ∇ U is the viscous dissipation term. In our case compressibility effects can be neglected,
as the compacting and melting part of the press was modeled using a theoretical mathematical
approach described in the following paragraph, so that the continuity equation becomes
∇ ⋅ U = 0 (21.13)
Although the gravitational term was included in simulations, in this specific case body forces are
negligible compared to viscous and pressure forces, along with the equally insignificant centripetal
and Coriolis accelerations.
Several possible CFD softwares could be used to perform simulations related to pasta extrusion
process, and results presented are obtained using a reliable and robust open source control volume
software (OpenFoam).
To account for possible mixing effects at the end of the screw barrel, the screw rotation was
modeled applying a moving mesh approach, the Arbitrary Mesh Interface (AMI), for non-conformal
patches, according to the implementation based on the algorithm described in Farrell and
Maddison [30]. This technique allows simulation across disconnected but adjacent mesh domains,
where the domains can be stationary or they can move relative to one another. The implementa-
tion in OpenFoam is fully parallelized, with the AMI being either distributed across several sub-
domains, or confined to a single subdomain by a constrained domain decomposition. A modified
version of pimpleDyMFoam solver for non-Newtonian fluids and heat transfer with moving mesh
was adopted. A centered cubic scheme was adopted for vectors field, an explicit non-orthogonal cor-
rection for surface normal gradient, a fourth order least squares scheme for gradient, an unbounded
fourth order conservative scheme for Laplacian and divergence operators, and a blended second
order implicit scheme in time was used to stabilize the computation.
A blended (PIMPLE) pressure-implicit split-operator (PISO) and semi-implicit method for
pressure-linked equations (SIMPLE) scheme were adopted to solve the resulting system. Grid inde-
pendence of results was checked by using several grid ranging from 104 to 106 control volumes. The
minimum volume was order of 10−15 m3, located near the homogenization screen, and the maximum
was order of 10−9 m3. Results shown in this work were obtained with 1⋅106 control volumes, and the
computation was performed on a 64 cores workstation with 256 Gb of memory.
21.2.3.3 Model Validation
Since the work by Dhanasekharan and Kokini in 2003 focused on numerical simulation of pasta
extrusion, literature on numerical simulation of pasta dough extrusion is almost non-existent,
excluding the work of Fabbri and Lorenzini [3], where the field of investigation was restricted to
the extrusion bell and very limited attention to the rheological properties of pasta dough was paid.
Although Dhanasekharan and Kokini’s work showed that important parameters for extruder’s
optimization could be obtained numerically, several issues were not assessed in Dogan and Kokini
[31]; the rheological model was not validated in the industrial range of temperature and MC, simula-
tions were somehow under-resolved (9000-17.000 elements are not enough to correctly capture 3D
thermal gradients, and under-resolved gradients could have a strong impact on viscous dissipation
calculation; moreover grid independence analysis of numerical results was not performed), and no
comparison with an experimental setup was proposed. Using the previously described numerical
model for the apparent viscosity, the accuracy of the numerical and experimental analysis was cross-
checked in a pasta laboratory extruder. The geometry of the Sercom press barrel and screw used
for experimental test is described in Figure 21.4: under a paddle mixer (not shown), the extruder
barrel is followed by a final cone allowing the 90° bending of the flow into the extrusion bell. At
the bottom of the bell an homogenization screen (1×1 mm square holes) is positioned, followed by a
dual die configuration, with each die extruding seven spaghetti. A temperature and pressure probe
was inserted into the extrusion bell in a non-disturbing position with respect to the main flow. The
constant pitch extruder is untapered, with the following dimensions: barrel internal diameter W =
36 mm, screw channel height H = 9.5 mm, screw pitch p = 20 mm, screw channel width 12.54 mm.
The heat capacity Cp was set to 2 kJ/kg°C and thermal conductivity k = 0.242 W/m°C [32]. The
generalized Brinkman number Br* computed according to the definition in Coelho and Pinhoa,
2009, has a value of −5, so that the the hypothesis of constant thermal properties may be reasonably
accepted.
The water–semolina mixing obtained using the integrated paddle mixer was unsatisfactory,
as the MC was not evenly distributed into the raw material after 20 minutes. For this reason the
hydrated semolina in the experimental test was previously prepared in the batch Kitchen Aid mixer
used for rheological tests, following the same procedure described in the previous section, and the
batch mixer was left on only to facilitate the insertion of hydrated semolina into the screw barrel.
The first numerical-experimental comparison was done using the temperature history into the
extrusion bell during the initial transitory. Results comparing numerical and experimental data
during the transitory startup of the extruder are shown in Figure 21.8; screw rotation set to 20 rpm,
thermostatic bath temperature set at Tbath = 45°C and time 0 was computed synchronizing the
numerical and physical temperature probes for T = 46.5°C.
Several simulations were also performed by varying the screw speed (varying from 10 to 30 rpm)
and the mass flow rate, and results compared with experiments are reported in Figure 21.9 after
20 min of extrusion process (Tbarrel = 45°C), showing a very good agreement.
The following results give us insight into the flow dynamics inside the extruder, and refer to
the case of screw speed set to 20 rpm, Tset = 45°C, corresponding to a low mass rate of 6.45 kg/h.
FIGURE 21.8 Influence of viscous dissipative effects in a extrusion experiment: numerical and experimen-
tal data (thermostatic bath temperature set at T = 45°C).
FIGURE 21.9 Numerical and experimental data comparison for pressure vs. mass flow rate in the extruder
bell.
The pressure contours on a longitudinal plane are shown in Figure 21.10, showing the maximum at
the end of the metering section as expected.
The effect of viscous dissipation is more evident by comparing the global temperature field
with and without thermostatic bath (adiabatic conditions), as shown in Figure 21.11(a) and (b). The
adiabatic simulation was used to computed the Specific Mechanical Energy (SME) of the system,
defined as the amount of mechanical energy dissipated as heat per unit mass inside the material with
and without considering the heat exchange due to the thermostatic jacket.
The mechanical energy input transferred by the mechanical shaft to rotate the screw is mostly
lost as heat through viscous dissipation, and an energy balance indicates that SME must be equal
to the work input from the drive motor into the extruded material, therefore providing a good char-
acterization of the extrusion process [18]. In this simulation the SME of the extruder at 20 rpm was
290 kJ/kg without the thermostatic jacket on, and 130 kJ/kg in the other case, with a clearance of
the screw at the barrel wall set equal to 0.
A interesting insight provided by the numerical simulation can be obtained by comparing the
temperature, viscous dissipation and velocity in different sections of the screw (Figure 21.12) while
the velocity field remains almost unchanged from one turn of the screw to another. In Figure 21.13,
velocity profiles on a normal section (at the center of the channel) at different values of X/L (turns
of the screw) are plotted, while the temperature field shows a more complex pattern.
FIGURE 21.10 Pressure contours on a longitudinal plane.

FIGURE 21.11 Temperature flow field in thermostatic enforced conditions at T = 45°C at the barrel wall and
extrusion bell (a) and adiabatic conditions (b).
The physical mechanism of temperature increase is the following: viscous dissipation (a) gen-
erates heat near the wall, particularly near the barrel where velocity gradients are more intense;
the velocity field (c) is dominated by the axial velocity (d), considering as main axes the helical
path passing through the center of normal sections to the screw shaft. The generated heat is then
transported by convection in an axial direction, while in the radial direction convective exchanges
are limited (e), resulting in a complex temperature field (b) which can not properly be described by
using a single normal profile. The analysis of global velocity field shows that no mixing is present
either in the screw or in the bell; only at the end of the screw is relative intensity mixing present, as
Viscous dissipation is strongly localized on the screw flight near the external barrel and in prox-
imity of homogenizing grid, as shown in Figure 21.15(a) and (b), where the velocity gradients are
more intense. Maximum values were reported inside the die extrusion channels.
The presence of low shear zones involves as a consequence high local viscosity values, opposing
the mixing like in the centre of the channel or inducing local recirculation zones near the homog-
enizing grid, as shown in Figure 21.16.
21.2.4 Use of CFD for Extrusion Optimization Process: Optimal Shape Design (OSD)
The opportunity to model and simulate the pasta extrusion process also opens new possibilities
in the field of optimization. Computational science and engineering techniques have allowed a
major change in the way that products can be engineered, as a computational model that simu-
lates the physical processes can be built rather than building real world prototypes and performing
experiments.
Among such techniques, Optimal Shape Design (OSD) [33] represents an interesting approach.
In OSD, the essential element with respect to classical numerical simulations in fixed geometri-
cal configurations is the introduction of a certain amount of geometrical degrees of freedom as
one of the unknowns, which means that the geometry is not completely defined, but part of it
is allowed to move dynamically in order to minimize or maximize the objective function. The
applications of optimal shape design (OSD) are uncountable, and for systems governed by partial
FIGURE 21.12 Details of temperature (a), viscous dissipation (b), velocity magnitude (c), axial velocity
(d) and radial velocity (e) in a section of the screw (from turn 3 to 7 of the screw flight).
differential equations, they range from structure mechanics to electromagnetism and fluid
mechanics or a combination of the three.
In industry, optimum design is not a once-and-for-all solution because engineering design is
made of compromises, owing to the multidisciplinary aspects of the problems and the necessity of
doing multipoint constrained design [33].
OSD is a branch of differentiable optimization and more precisely of optimal control for dis-
tributed systems [34], where gradient and Newton-based methods are natural numerical tools. The
problem is that OSD is still numerically difficult, because it is computer intensive and moreover
FIGURE 21.13 Velocity profiles on a normal section (at the center of the channel) at different values of X/L.
FIGURE 21.14 Velocity contours field and streamtraces on a normal longitudinal section.
because the “optimal” is a compromise between shapes that are good with respect to several cri-
teria. One possible approach is the use of Pareto optimality, i.e., adopting a mathematical theorem
that says in simple situations Pareto optimal points are minimizers of some convex combination of
all the criteria. The trouble is that such linear combinations can lead to stiff problems with many
suboptimal nodes, requiring a global optimization algorithm such as the genetic one.
Another problem is linked to the existence of solutions and differentiability of the criteria with
respect to shape deformation. This was the focus of most of the 1980s studies [34–38], after it became
clear [39] that oscillations of shapes could lead to nonphysical solutions of the optimization problem
in the limit, a phenomenon known as homogenization, leading to a new class of problems called
topological optimization.
FIGURE 21.15 Viscous dissipation contours in the extruder screw (a) and bell (b) (kg m−1 s−2).
FIGURE 21.16 Viscosity contours in a longitudinal plane (y = 0).
Descent algorithms are generally used to find local or global minima, based on the information
given by the variation of the gradients numerically obtained. The sensitivity analysis allows the
calculation of the objective function gradients, but it is a complex job, fraught with possible errors.
Sometimes sensitivity analysis is not even possible, since the use of industrial solvers does not
allow a proper interaction without extensive code rewriting. In some cases it is preferable to the use
of algorithms that do not use the computation of the gradients, such as genetic algorithms (GA).
Genetic algorithms add a certain robustness, since they do not require the objective function
conditions of regularity. They also allow the development of innovative design compared to conven-
tional methods because they are not based on local minimum criteria. Nonetheless gradient-based
algorithms are quite efficient, like, for example, the Broyden–Fletcher–Goldfarb–Shanno (BFGS,
[40]) algorithm, an iterative method for solving unconstrained nonlinear optimization problems of
the quasi-Newton class.
This variable metric algorithm requires the calculation of the gradient of the objective function
using an in line sub-minimization algorithm.
The methodological approach requires the definition of a set of physical variables xp and a set
of geometrical variables x nk where max(n) = N is the number of geometrical degrees of freedom
(DOF) and k is an iterative index.
The set of geometrical DOF is constrained between the envelope defined by two functions
( ) ( )
s1 x nk and s2 x nk , which delimit the maximum and the minimum variation

{ ( ) ( )}
x nk : s1 x nk ≤ x nk ≤ s 2 x nk . (21.14)
The constraints are assigned in an arbitrary way depending on the problem and, for example, on
mechanical or industrial considerations.
For a given set of partial differential equations (PDE)

(
F x p , x nk = 0 ) (21.15)
we can define an integral weighted target function (objective function)
( ) ∫ δ (x)I ( x , x ) dω (21.16)
Θ x p , x nk = p
k
n
( )
where I x p , x nk represent a required property of the system. Depending on the choice of the weight-
ing function δ(x) we can have an integral or a pointwise condition. The target function represents the
optimal performance required to the system, defined by solving the associated minimum problem
 ∫ ( 
)
min  δ (x) I x p , x nk dω  (21.17)
Ω 

The existence of a local or a global minimum in a mathematical sense is strongly linked to the cho-
sen target function Θ, to the search envelope defined by the geometrical constrains (Equation 21.14),
and to the solution algorithm. To this purpose, when possible, it is preferable to adopt a convex
function Θ as objective function. From an operative point of view, the solution of the constrained
minimization problem can be described in the following way:
a. Start from an initial geometry x1n

( )
b. Solve the balance equations’ set F x p , xnk
∂Θ ( x , x ) k
c. Compute the objective function Θ ( x , x ) and the gradient J =
k p n

( )
p n
d. Find the minimum α with BFGS algorithm, i.e.: ∂ xnk
n
k
n { (
d1. find α such as x = x + α and α : Θ x p , x < Θ x p , x
k +1 k
n ) ( k +1
n )}
( ) ( )
e. if Θ x p , xn( k +1) = Θ x p , xnk + ε , where ε is a defined error, then xnk +1 = xoptional else goto b.
An application of the previously described method is the optimal design of an extrusion bell for
pasta production.
Pasta production is a robust and widely diffused process all over the world. Nonetheless,
extrusion bells for short pasta production are still designed based on a heuristic approach,
mostly based on linear scaling of smaller extrusion units, adapted to a mass production scale
up to 4 tons per hour on a single unit. As a consequence of such design, recirculation zones are
BELL
COMPENSATION
DISK
GRID
INSERTS
DIES
FIGURE 21.17 (a) 3D Full Scale simulation of a short pasta extrusion bell & dies (b) Velocity distribution
and recirculating zones (unpublished results).
present [11] (see Figure 21.17a and b) where the biological flow (pasta dough) remains inside,
possibly generating molds.
Moreover, the uneven pressure distribution on the die insert plane generates an uneven length of
the produced pasta, with secondary quality effects due to the different shear rates. The non-uniform
length problem is addressed by using a hand-made compensation disk (made on a basis of trial and
error for any new industrial plant) with different diameters in the radial direction.
For the purpose of this work, OSD is performed to investigate the optimal shape on a 2D axisym-
metric problem defined in Figure 21.18. Here side l and m are let free to move on the y directions.
In order to reduce the computational weight of the optimization process, a subset of the geometrical
( )
nodes (form l1 to l5 and from m1 to m5) is chosen as geometrical DOF xnk , while all other computational
nodes are recovered at each simulation by using a B-Spline interpolation with xnk as reference nodes. ( )
In 3D the same approach can be used replacing B-Splines with Bezier surfaces. The die’s chan-
nels are here identified using the radial distribution from r1 to r8.
For computational details about the CFD solver and rheological models for pasta dough (non-
Newtonian temperature and shear rate behavior) please see Sarghini et al. [4].
inflo w
m1
l1
m2
l2
m3
l3
l4 m4
m5
l5
r1 r2 r3 r4 r5 r6 r7 r8
FIGURE 21.18 Axisymmetric 2D problem.

In order to minimize such a discrepancy, the quadratic sum of the difference between the effec-
tive mass flow rate and the average optimal mass flow rate computed on all radial channels r1 to r8
is assumed as objective function.
 8
2  8
2
( ) ∑ ( MFR
Θ x p , xnk = 

r =1
k
r − MFR 

) 
 ∑(
r =1
)
MFRr0 − MFR  (21.18)


where MFRrx is the mass flow rate computed in rx channels MFR is the optimal one (MFRinflow/8), and
 8
2

 ∑ ( MFR
r =1
0
r − MFR  (21.19)

)

is the target function computed in the initial geometry.

In order to get a limited range of the target function, the value was adimensionalized by dividing
the computed value at iteration k respect to the initial value, so that the possible range was limited,
( )
Θ x p , xnk ∈[0; 1].
Convergence was set by choosing Θ x p , xnk +1 − Θ x p , xnk < ε = 0.05. ( ) ( )
The optimization process required 182 iterations to converge with an error set to 5% of the initial
value, and convergence graph is shown in Figure 21.19a.
Figure 21.19 shows the initial average mass flow rate respect to the radial distribution and it is
compared with the final distribution on the optimal shape.
1.2 0.03
1 0.025 inial MFR

Adimensional object function
opmall MFR
MFR
FR
0.8 0.02
0.015
0.6
0.01
0.4
5
0.005
0.2
0
r1 r2
0 r3 r4 r5
0 50 100 150 200 r6
r7
Number of iteraons r8
(a) (b)
FIGURE 21.19 (a) Optimization process convergence graph (b) Initial and final (on the optimized geometry)
mass flow rate radial distribution for the axisymmetric 2D problem.
Velocity Magnitude Velocity Magnitude

0.0036 0.0036
0.0032 0.0032
0.0028 0.0028
0.0025 0.0025
0.0021 0.0021
0.0017 0.0017
0.0013 0.0013
0.0010 0.0010
0.0006 0.0006
0.0002 0.0002
FIGURE 21.20 Contour plot of velocity field in initial (a) and optimal configuration (b).
The evident uneven radial pressure and velocity, distribution resulting in different mass flow rate
on each insert in the original geometry was mostly recovered in the final optimal shape.
The final shape with flooded velocity distribution is shown in Figure 21.20b in comparison with
the initial distribution shown in Figure 21.20a.
21.3 CONCLUSIONS
In this chapter a combination of experimental data and numerical models was proposed and vali-
dated to develop a numerical modeling framework using CFD to analyze and optimize the pasta
production process and equipments.
The “natural” combination of experimental evidence and numerical modeling provides a symbi-
otic tool to gain insights into the physics of pasta dough extrusion, together with the possibility to
identify the influence of each variable and possibly optimize product quality and minimize process
handling.
Simulations in the proposed framework were performed by using a modified version of
OpenFoam, an open source finite volumes Navier–Stokes solver written in C++, whose reliabil-
ity has been extensively tested in industrial problems. A non-Newtonian power-law temperature-
dependent model was used to mimic the rheological behavior of pasta dough respect to the local
shear rate, adopting a truncated Taylor expansion to model the temperature dependence, and results
were then validated on an laboratory extruder, solving the unassessed questions about accuracy of
both numerical models and rheological data. The analysis of velocity and temperature flow field
shows that only adopting an accurate numerical discretization of the domain is possible to capture
significant details providing global information like SME or Residence Time Distribution. This
integrated experimental-numerical model can be used not only to obtain a better understanding
of the phenomena involved in pasta extrusion, but also in equipment optimization and reverse die
design.
REFERENCES
1. LeRoux D., Vergnes B., A thermomechanical approach to pasta extrusion, Journal of Food Engineering,
26: 351–368, 1995.
2. Dhanasekharan K.M., Kokini J.L., Design and scaling of wheat dough extrusion by numerical simula-
tion of flow and heat transfer, Journal of Food Engineering, 60: 421–430, 2003.
3. Fabbri A., Lorenzini G. CFD analysis of the semolina dough extrusion for short pasta production,
International Journal of Heat and Technology, 26(1): 35–40, 2008.
4. Sarghini F., Romano A., Masi P., Experimental analysis and numerical simulation of pasta dough extru-
sion process, Journal of Food Engineering, 176: 56–70, 2016.
5. Emin M.A., Modeling extrusion processes. In: Modeling Food Processing Operations, Bakalis S.,
Knoerzer K. and Fryer P.J., Eds., Woodhead Publishing, Sawston, Cambridgeshire, UK, 235–253, 2015.
6. Camire M.E., Camire A., Krumhar K., Chemical and nutritional changes in foods during extrusion,
Critical Reviews in Food Science and Nutrition, 29(1): 35–57, 1990.
7. Lai L.S., Kokini J.L., Physicochemical changes and rheological properties of starch during extrusion
(a review), Biotechnology Progress, 7(3): 251–266, 1991.
8. Riaz M.N., Asif M., Ali R., Stability of vitamins during extrusion, Critical Reviews in Food Science and
Nutrition, 49(4): 361–368, 2009.
9. Singh S., Gamlath S., Wakeling L., Nutritional aspects of food extrusion: A review, International
Journal of Food Science and Technology, 42(8): 916–929, 2007.
10. Roache P.J., Verification and Validation in Computational Science and Engineering, Hermosa Publishers,
Albuquerque, NM, USA, 1998.
11. Sarghini F., Cavella S., Torrieri E., Masi P., Experimental analysis of mass transport and mixing in a
single screw extruder for semolina dough, Journal of Food Engineering, 68(4): 497–503, 2005.
12. Medvedev G.M., Non conventional aspects of pasta production Part 2a—Influence of mixing regimes and
pressure operating conditions on dough properties and pasta quality, Tecnica Molitoria, 12: 1338–1343, 1997.
13. de la Pena E., Manthey F.A., Patel B.K., Campanella O.H., Rheological properties of pasta dough
during pasta extrusion: Effect of moisture and dough formulation, Journal of Cereal Science, 60:
346–351, 2014.
14. Bagley B.E., Dintzis F.R., Chakrabarti S., Experimental and conceptual problems in the rheological
characterization of wheat flour doughs, Rheologica Acta, 37: 556–565, 1998.
15. Singh N., Smith A.C., Rheological behavior of different cereals using capillary rheometry, Journal of
16. Cuq B., Yildiz E., Kokini J., Influence of mixing conditions and rest time on capillary flow behavior of
wheat flour dough, Cereal Chemistry, 79(1): 129–137, 2002.
17. Hicks C.I., See H., The rheological characterisation of bread dough using capillary rheometry,
Rheologica Acta, 49: 719–732, 2010.
18. Chillo S., Civica V., Iannetti M., Mastromatteo M., Suriano N., Del Nobile M.A., Influence of repeated
extrusions on some properties of non-conventional spaghetti, Journal of Food Engineering, 100(2):
329–335, 2010.
19. Hatzikiriakos S.G., Dealy J.M., Start-up pressure transients in a capillary rheometer, Polymer
Engineering and Science, 34: 493–499, 1994.
20. Hosseinalipour S.M., Tohidi A., Shokrpour M., A review of dough rheological models used in numeri-
cal applications, Journal of Computational and Applied Research in Mechanical Engineering, 1(2):
129–147, 2012.
21. Chhabra R.P., Richardson J.F., Non-Newtonian Flow in the Process Industries, Butterworth-Heinemann,
Oxford, UK, 1999.
22. Leonard A.L., Cisneros F., Kokini J.L., Use of the rubber elasticity theory to characterize the viscoelas-
tic properties of wheat flour doughs, Cereal Chemistry, 76(2): 243–248, 1999.
23. Mackey K.L., Ofoli R.Y., Morgan R.G., Steffe J.F., Rheological modeling of potato flour during extru-
sion cooking, Journal of Food Process Engineering, 12: 1–11, 1990.
24. Morgan R.G., Steffe J.F., Ofoli R.Y., A generalized viscosity model for extrusion of protein doughs,
Journal of Process Engineering, 2: 57–78, 1989.
25. Mackey K.L., Ofoli R.Y., Rheology of low- to intermediate-moisture whole wheat flour doughs, Cereal
Chemistry, 67(3): 221–226, 1990.
26. Dus S.J., Kokini J.L., Prediction of the nonlinear viscoelastic properties of a hard wheat flour dough
using the Bird-Carreau constitutive model, Journal of Rheology, 34(7): 1069–1084, 1990.
27. Darnell W.H., Mol E.A.J., Solids conveying in extruders, SPE Journal, 12: 20–27, 1956.
28. Tadmor Z., Klein I., Engineering Principles of Plasticating Extrusion, Van Nostrand, New York, USA, 1970.
29. Weert X., Lawrence C.J., Adams M.J., Briscoe B.J., Screw extrusion of food powders: Prediction and
performance, Chemical Engineering Science, 56: 1933–1949, 2001.
30. Farrell P.E., Maddison J.R., Conservative interpolation between volume meshes by local galerkin
projection, Computer Methods in Applied Mechanics and Engineering, 200: 89–100, 2011.
31. Dogan H., Kokini J.L., Rheological properties of foods. In: Handbook of Food Engineering, 2nd ed.,
Heldman D.R. and Lund D.B., Eds., CRC Press, Boca Raton, FL, USA, 2007.
32. Andrieu J., Gonnet E., Laurent M., Thermal conductivity and diffusivity of extruded durum wheat
pasta, Lebensmittel-Wissenschaft und-Technologie, 22: 6–10, 1989.
33. Mohammadi B., Pironneau O., Shape optimization in fluid mechanics, Annual Review of Fluid
Mechanics, 36: 255–79, 2004.
34. Lions J.L., Controle Optimal de Systemes Gouvernes par des Equations aux Derives Partielles, Dunod/
Gauther - Villars, Paris, France, 1968.
35. Pironneau O., Optimal Shape Design for Elliptic Systems, Springer-Verlag, New York, USA, 1984.
36. Delfour M., Zolezio J.P., Shapes and Geometries: Analysis, Differential Calculus and Optimization,
SIAM, Philadelphia, USA, 2001.
37. Sokolowski J., Zolezio J.P., Introduction to Shape Optimization: Shape Sensitivity Analysis, Springer-
Verlag, New York, USA, 1991.
38. Haslinger L., Makinen R., Introduction to Shape Optimization: Theory, Approximation, and
Computation, SIAM, Philadelphia, USA, 2003.
39. Tartar L., Control problems in the coefficients of PDE. In: Control Theory, Numerical Methods and
Computer Systems Modelling, Bensoussan A. and Lions J.L., Eds., Springer-Verlag, Berlin, Germany,
420–426, 1974.
40. Nocedal J., Wright S.J., Numerical Optimization, 2nd ed., Springer-Verlag, New York, USA, 2006.
22 Computational Modeling
of Radio Frequency Thawing
of Frozen Food Products
Francesco Marra, Tesfaye Faye Bedane, Oriana Casaburi,
Ozan Altin, Rahmi Uyar, and Ferruh Erdogdu
CONTENTS
22.1 Introduction............................................................................................................................ 487
22.2 RF Heating Mechanism....................................................................................................... 488
22.3 Heat Transfer and Electric Field Displacement in RF Heating............................................ 490
22.4 Phase Change During Thawing............................................................................................. 492
22.5 Dielectric Properties............................................................................................................ 493
22.5.1 Dielectric Permittivity............................................................................................ 493
22.5.2 Physical Properties Related to Dielectric Properties............................................... 494
22.5.2.1 Electric Conductivity............................................................................... 494
22.5.2.2 Loss Tangent............................................................................................ 494
22.5.2.3 Penetration Depth.................................................................................... 495
22.5.3 Factors Affecting Dielectric Properties.................................................................. 495
22.5.3.1 Temperature............................................................................................. 495
22.5.3.2 Composition................................................................................................ 495
22.5.3.3 Frequency.................................................................................................... 496
22.5.4 Methods for Determining Dielectric Properties..................................................... 496
22.6 Thermal–Physical Properties.............................................................................................. 497
22.6.1 Specific Heat........................................................................................................... 497
22.6.2 Thermal Conductivity............................................................................................... 497
22.6.3 Density......................................................................................................................... 498
22.7 Modeling Studies for RF Thawing........................................................................................ 498
22.7.1 Electric Field Distribution Inside a RF System........................................................ 500
22.7.2 Effect of Sample Load on the Efficiency of RF Process........................................500
22.7.3 Simulation of Sample Movement within an RF Cavity.......................................... 501
22.8 Conclusions.......................................................................................................................... 505
Acknowledgments...........................................................................................................................506
References.......................................................................................................................................506
22.1 INTRODUCTION
Before eating, cooking or further processing, with the exception of ice cream and frozen desserts,
frozen food products are usually thawed. Sometimes, the temperature might need to be raised to just
below the freezing point for tempering purposes. Conventional methods for thawing and tempering
usually apply air and water, and are based on conduction and convection heat transfer mechanisms.
These methods leave a thawed, low thermally conductive region on the outer surfaces of the product
487
FIGURE 22.1 Center and surface temperature change of a frozen food product during a conventional thaw-
ing process.
while inside regions are still frozen. This low thermal conductivity layer, creating a barrier, reduces
the speed of heat transfer and necessitates a longer time for thawing. In addition, the tempera-
ture difference between surface and center of the product becomes higher while high temperature
regions become open to the risk of microbial growth and increased enzymatic activity, especially
when the temperature exceeds the critical temperature of +4°C. During a conventional process, the
surface of food product is usually at contamination risk from the thawing environment. Besides,
due the increased temperature and humidity, there is also a risk of microbial growth. Figure 22.1
demonstrates temperature change at the center and surface of a frozen food product undergoing a
conventional thawing.
The meat industry uses air thawing or water immersion thawing approaches at the cost of
longer thawing times, increased microbial load and reduced quality. A conventional thawing
process of a large block of meat, for example, might take more than a day, creating the risks
mentioned above. Therefore, a rapid thawing process is required. Radio frequency (RF) is such
a process to obtain rapid and volumetric heating and uniform temperature increase within the
products compared to a conventional thawing process. It offers several advantages over conven-
tional methods, such as increased heat efficiency, rapid and volumetric heating, an even moisture
profile, and avoiding pollution of by-products of combustion. This process also saves operating
space by faster product transfer and reduced system length and automatic adjustment of operating
parameters for variations of the moisture content of the product [1]. While the first RF applica-
tions go back to the 1940s, thawing methods with commercial applications were focused on after
the 1960s, and the recent literature focuses on mathematical modeling of the process for design
and optimization purposes.
Therefore, this chapter discusses the physics and transport phenomena involved in an RF thaw-
ing process. Specifically, as analysis of the critical issues involved is often the main purpose of sim-
ulation of such a process, mathematical representation of all the relevant phenomena involved in the
process was included with the description of transport phenomena involved, equations to describe
boundary conditions and physical properties of all the possible materials related to an RF process.
22.2 RF HEATING MECHANISM

RF technology uses electromagnetic energy within a frequency range of 3 kHz to 300 MHz to
heat target foods. The Federal Communication Commission (FCC) assigned 13.56 ± 0.00678 MHz,
Computational Modeling of Radio Frequency Thawing of Frozen Food Products 489
27.12 ± 0.16272 MHz, and 40.68 ± 0.02034 MHz for RF treatments in industrial, scientific
and medical (ISM) applications to prevent disturbance in telecommunications [2]. RF heating
is based on the transformation of alternating electromagnetic field energy into thermal energy
by affecting polar molecules and charged ions in a material. In an RF application process, the
product forms a dielectric material between two capacitor plates that are alternatively charged
by a high frequency alternating electric field [3]. Dielectric materials are composed of atoms or
molecules irrespective of phase (solid, liquid, or gas) that are composed of negatively charged
electrons and positively charged nucleus. These constituents may be locked into regular struc-
tures of crystals or free to wander through the structure. Hence, several types of electric displace-
ments of these constituents, such as electronic, atomic, molecular, and ionic, are possible [4].
Molecular displacement involves molecules that are initially asymmetrical in the structure and
have a definite electric moment in addition to their own electric field when placed in an external
electric field [3]. When an external electric field is applied, bipolar molecules tend to behave like
microscopic magnets and attempt to align themselves with the field. When the electrical field is
changing millions of times per second (e.g., 27.12 MHz), these molecular magnets are unable to
withstand the forces acting to slow them. This resistance to the rapid movement of bipolar mol-
ecules creates friction and results in heat dissipation in the material exposed to the RF radiation.
This interaction results in translation motions of free or bound charges and rotation of dipoles,
and causes losses that, in turn, produce the volumetric heating within the material [3]. The ability
of RF radiation to penetrate and couple with materials provides an attractive method for obtaining
controlled and precise heating.
A typical RF system consists of two main components: generator and applicator. The generator
part is mainly used to generate the RF energy, and the applicator is a metal structure that directs
the RF field towards the product or load to be heated. Two main kinds of RF systems are used in
commercial applications: open circuit (also called free running oscillator) and 50-Ω technology,
based on their radio wave generation mechanism, components and properties [5]. The open circuit
RF systems are used widely in many industrial applications, while the 50-Ω systems are a relatively
new technology to provide a fixed frequency compared to self-oscillatory circuit. They are also
known for their precise control power and feedback [2]. Thus, the heating rate can be tightly con-
trolled and precise final temperatures can be achieved easily in a 50-Ω system compared to the free
running oscillator systems. Since the power is fed through a standard 50-Ω cable, the generator can
be remotely located and quickly connected. It has a more stable frequency output since the match-
ing system in the RF heater is automatically adjusted to maintain the load impedance at 50 Ω. The
50-Ω systems are more expensive than the open circuit ones and have not been popularly used by
industries. The open circuit RF system generates heat by means of an RF generator that produces
oscillating fields of electromagnetic energy and comprises a power supply and control circuitry, a
parallel plate and a system for supporting processed material.
When a food product is placed between the RF applicator electrodes, complex electric imped-
ance is introduced into the RF electric field, as expressed in equation below:
1 ε ′′ − jε ′
ZC = (22.1)
2π fC0 ε ′′ 2 + ε ′ 2
where ZC is the capacitance of the material, f is the frequency of the electric field, εʹ is dielectric con-
stant and εʺ is dielectric loss factor of the material. C0 is the capacitance of free space and j = −1.
From Equation (22.1), a finite resistance R appeared across the capacitor, expressed as:
1
R= (22.2)
2π fC0ε ′′
v2
Taking the power, P, dissipated in an electric resistance to be equal to in a 50 Ω system, then
for a capacitor containing a dielectric material, R
P = 2π fC0ε ′′V 2 (22.3)

For a parallel plate capacitor, C0 = ε 0 Ap/d where Ap is the plate area, d is the plate separation and
ε 0 is the permittivity of free space. As the voltage V is equal to the electron field strength E multi-
plied by the distance between the two electrodes d. Power P can be written as:
2
P = 2π f ε 0ε ′′ E ( Ap d ) (22.4)

Since (Apd) is the volume, the power dissipation per unit volume or power density, Qv (W/m3) is
then written as
2
Qv = 2π f ε 0ε ′′ E (22.5)

where f is the frequency of the radio wave generator (MHz), ε 0 is the permittivity of free space
(8.85 × 10 –12 F m–1), εf ″ is relative dielectric loss factor of the food material under investigation and
E is the modulus of electric field.
22.3 HEAT TRANSFER AND ELECTRIC FIELD DISPLACEMENT IN RF HEATING

The mathematical description of heat transfer within food materials placed between the electrodes
is given by unsteady heat conduction equation (assuming that a solid like food stuff is processed in
the RF applicator) plus a generation term.
The governing equation of heat transfer inside the sample is described as:
∂T
pc p = ∇ ⋅ k ∇T + Qv (22.6)
∂t
where T is temperature within the sample, t is the process time, ρ is density (kg m–3), Cp is the
specific heat (J kg−1 K−1), k is thermal conductivity (W m–1K–1) and Qv is the power per unit volume
(W m–3) generated by electric field (E-field) distribution (Equation 22.5).
Electric field strength is part of a more complex electromagnetic field that can be described, in
terms of electric field intensity E, magnetic field intensity H and electric flux density D, by the fol-
lowing Maxwell equations in differential form, when the involved media are isotropic, linear and
homogenous [6]:
∂
∇ × E = −µ (H ) (22.7)
∂t
 ∂ 
∇ × H =  ε 0ε r + σ c  E (22.8)
 ∂t 

∇ ⋅ D = ρc (22.9)

∇ ⋅ H = 0 (22.10)

where σ c and ρ c are the effective electrical conductivity and the charge density, respectively.
Equation (22.7) is known as Faraday’s Law, Equation (22.9) is known as Ampere Law, Equation
(22.9) is known as the Gauss’ Law for electric fields and Equation (22.10) is known as the
Gauss’ Law for magnetic fields. Maxwell’s equations need some assumptions regarding the
dependence of fields with respect to time. A usual assumption is to consider the fields to be
time-harmonic [7]:
E = E0 e jwt (22.11)

H = H 0 e jwt (22.12)

When the time-harmonic assumption is made, Equations (22.7) and (22.8) become
∇ × E = − j µ wH (22.13)

∇ × H = jε 0ε r wE (22.14)

Combining Equations (22.13) and (22.14), one obtains the following wave equation that is used
for the calculation of the frequency domain:
1
∇× ∇ × E − w 2 µ0ε 0ε r E = 0 (22.15)
µT
Boundary conditions are needed in order to get a unique and proper solution. External boundar-
ies of the cavity are considered perfect conductors and the following boundary condition is then
used [7]:
n×E = 0 (22.16)

with n being the unit vector normal to the boundary surfaces.

When a quasi-static approach is considered, Maxwell’s equations collapse to the following:
∇ ⋅ (σ + jwε ) ∇V  = 0 (22.17)

where V is the electrical potential, related to the electric field by
E = −∇V (22.18)

In this case, as boundary conditions, a source electric potential V0 is applied to the upper elec-
trode of the capacitor while at the bottom electrical ground conditions is considered:
V = 0 (22.19)
RF applicator walls are electrically insulated, so last boundary conditions are:
∇⋅E = 0 (22.20)

In the air, the heat balance will not take into account the generation due to the interaction with
the electromagnetic field since the air has a negligible loss factor at 27.12 MHz. On the other hand,
contribution to the heat transfer due to forced convection must be taken into account. For this reason
the heat balance equation in the air is the following:
∂T
ρaC pa + ρaC pa v ⋅ ∇T + ∇ ⋅ q = 0 (22.21)
∂t
with
q = − ka∇T (22.22)

where v is the air velocity field (m s–1), ka is the fluid thermal conductivity, Cpa is the air heat capac-
ity (J kg–1 K–1) and ρa is the air density (k gm–3).
22.4 PHASE CHANGE DURING THAWING

Foodstuffs are complex systems of water, solutes and macromolecules but are often considered for
engineering purposes as two fractions: an aqueous solution and a solid component [8]. Phase change
in foods involves intricate interactions between the aqueous and solid fractions. In high-moisture
solid food (like meat, vegetables and fish) water is bounded in a solid matrix. Phase change involves
mainly the aqueous part changing from ice to water or vice versa, with the solid being relatively
inert. The interactions in the food cause continuous freezing point depression as ice separates from
the aqueous phase; the latent heat is released or absorbed over a range of temperature during a phase
change [9]. The thermal properties of food are highly dependent on the fraction of ice in the food
[10,11].
During thawing, thermal properties change gradually before the initial freezing point, whilst in
freezing, thermal properties are relatively constant above the initial freezing temperature and below
this temperature, rapid changes occur as the ice fraction increases. In the energy balance, the phase
change plays an important role and, in low conductivity materials such as foods, it has a direct
impact on the accumulation term. In order to take into account the phase change in the heat transfer
equation, the so-called apparent specific heat method can be used.
The apparent specific heat method still appears to be one of the preferred methods for model-
ing problems in heat conduction with phase change. Apparent specific method solves the original
Fourier equation over the melting range, which is directly proportional to the latent heat of phase
change. In this method, latent heat is merged with sensible heat to produce a specific heat curve with
a large peak around the freezing point. Because of the large variations in specific heat, iterations
must be carried out at every step. It is difficult to obtain convergence with this technique and there
is always a chance that the latent heat is underestimated. This happens when the temperature at a
node steps over the peak in the apparent specific curve; the mean specific heat between the initial
and final temperature is then always less than the peak, and the temperature change will therefore
be over-estimated. For general purposes, commercial software such as COMSOL (COMSOL, AB,
Stockhom, Sweden) is often the only method available to ensure convergence in reasonable time.
The following specific heat expression could be used for the estimation of heat capacity in the
process [12].
C p = C pfro T ≤ Tm1 (22.23)

C punf + C pfro ∆H S − L
Cp = + Tm1 ≤ T ≤ Tm 2 (22.24)
2 Tm 2 − Tm1

C p = C punfr T ≥ Tm 2 (22.25)

where Tm1 and Tm2 are initial and final temperatures of phase change range, ΔHs–L is the latent heat
of the sample food and Cpfro and Cpunf are specific heat of frozen and unfrozen sample, respectively.
This approach was successfully applied by Uyar et al. [13] in the first experimentally applied com-
putational RF thawing study.
22.5 DIELECTRIC PROPERTIES
Dielectric properties are the main parameters that provide information about how materials
interact with electromagnetic energy during dielectric heating of foods and biological products.
Mudgett [14] observed that dielectric properties of food are the principal parameters that deter-
mine the coupling and distribution of electromagnetic energy during dielectric heating. The
dielectric properties of food materials can be divided into two parts known as the permeability
and permittivity [5].
22.5.1 Dielectric Permittivity
Permeability values for foodstuffs are generally similar to that of free space and as a result are not
believed to contribute to heating [15]. However, the permittivity—which determines the dielectric
constant (ε′) and the loss factor (ε″)—influences the RF heating.
The ε′ and ε″, which are the real and imaginary parts, respectively, of ε, are given by
ε = ε ′ − jε ′′ (22.26)
The relative dielectric constant ε′ is a characteristic of any material and is a measure of the
capacity of a material to absorb, transmit and reflect energy from the electric portion of the
electrical field [16]. ε′ is a constant for a material at a given frequency under constant condi-
tions. The relative loss factor ε″ measures the amount of energy that is lost from the electrical
field, which is related to how the energy from a field is absorbed and converted to heat by a
material passing through it [16].
A material with a low ε″ will absorb less energy and could be expected to heat poorly in an electrical
field due to its greater transparency to electromagnetic energy [17]. These values are strong functions
of temperature, and this certainly affects the temperature change during an RF thawing process. Farag
et al. [18] measured the dielectric properties of lean beef over a range of temperature between –18 to
+10°C using the probe method at 27.12 MHz RF frequency. Change of dielectric constant and dielectric
loss factor as a function of temperature is shown in Figure 22.2(a) and (b), respectively. These values
were used by Uyar et al. [13] to experimentally validate a mathematical model developed for an RF
thawing process. Llave et al. [19] also reported the dielectric property change of tuna undergoing a thaw
cycle at 13.56 and 27.12 MHz while Llave et al. [20] carried out a mathematical modeling study for RF
defrosting of frozen tuna at 13.56 MHz.
FIGURE 22.2 Dielectric permittivity for lean beef, at 27.12 MHz, as a function of temperature: (a) dielec-
tric constant; (b) loss factor. (From Farag, K., Lyng, J.G., Morgan, D.J. and Cronin, D.A. 2008. Dielectric
and thermophysical properties of different beef meat blends over a temperature range of –18 to +10 C. Meat
Science. 79: 740–747.)
22.5.2 Physical Properties Related to Dielectric Properties

22.5.2.1 Electric Conductivity
Electric conductivity (σ) is the material property that indicates the ability to conduct electricity. It is
related to the loss factor ε″ and it can be calculated using the following formula:
σ = 2π f ε ′′ (22.27)
22.5.2.2 Loss Tangent
The tangent of dielectric loss angle (tanδ) is often called the loss tangent or the dissipation (power)
factor of the material. For a given material this is equivalent to the ratio of dielectric loss factor to
the dielectric constant:
ε ′′
tanδ = (22.28)
ε′
22.5.2.3 Penetration Depth
The penetration depth (dp) is defined as the depth into a sample where the microwave or RF power
has dropped to 1/e (e = 2.718) or 36.8% of the surface value [21].
c′
dp = (22.29)

2π f 2ε ′ 1 + (tanδ )2 − 1

where C is the speed of propagation of waves in a vacuum (3 × 108 m s–1) and dp is expressed in
meters. When tanδ is low (i.e., far less than 1) penetration depth is well described by the following
simpler form:
c′ 4.47 × 10 7
dp = = (22.30)
2π f ε ′ tanδ f ε ′ tanδ
Equations (22.24) and (22.25) illustrate the effect of f, ε′ and ε″ on dp. Bengtsson and Risman
[21] found that the greatest dp was experienced when both ε′ and ε″ were low. After obtaining the
dielectric properties, the penetration depths of electromagnetic energy in selected material can be
calculated. Deeper penetration corresponds to lower frequencies, and higher frequencies result in
greater surface heating. Thus the penetration depth of RF energy in food can be as large as one order
of magnitude compared with the microwaves [22]. For example, at 27.12 MHz a six times greater
penetration depth in mangoes was calculated compared to microwave energy at 1800 MHz at 20°C
[23]. Farag et al. [18] reported the dp of lean beef at 27.12 MHz significantly decreased with increas-
ing temperature. The values were 54.2 and 17.7 cm at –10°C and +1°C respectively. Similarly, a
significant difference in dp was reported as the frequency increased; the values were 18.4 and 1.6 cm
at 27.12 MHz and 2450 MHz, respectively, at –1°C.
22.5.3 Factors Affecting Dielectric Properties

Some of the factors affecting the dielectric properties during RF heating are related to the nature
of the material (composition, structure), conditions when dielectric heating is applied (temperature,
frequency) and age or maturity of the food material.
22.5.3.1 Temperature
The influence of temperature on the dielectric properties of foods depends on many factors, includ-
ing food composition, especially moisture and salt contents, and the involved frequencies [24].
Temperature dependence of the dielectric constant is quite complex and it may increase or decrease
with temperature depending on the material. Loss factor increases with increasing temperature at low
frequencies due to ionic conductance [25] and decreases with increasing temperature at high frequen-
cies due to free water dispersion [26]. Farag et al. [18] studied the effect of temperature on dielectric
constant (ε′) and dielectric loss factor (ε″). In the study, they considered the effect of temperature in
three ranges: –18 to –10, –10 to –1 and –1 to +10°C. They reported that, at lower temperature (–18 to
–10°C), there were no significant differences in ε″ and ε′ while all blends showed significant increase
in ε′ and ε″ at higher temperatures from –5 to –1°C.
22.5.3.2 Composition
In foods, water is generally the predominant component and the major absorber of microwave and
RF energy. Consequently, the higher the moisture content of the food material, the higher the dielec-
tric heating will be.
In general, higher moisture content results in a higher dielectric constant and loss factor of the
food material [27]. Because of the reduced loss factor with decreasing moisture content, dehydrated
foods have less ability to convert electromagnetic energy into thermal energy.
22.5.3.3 Frequency
The dielectric properties of most materials vary considerably with the frequency of the applied elec-
tric fields (Figure 22.3) [28]. An important phenomenon contributing to the frequency dependence
of the dielectric properties is the polarization of molecules arising from the orientation with the
imposed electric field, which have permanent dipole moments [29].
22.5.4 Methods for Determining Dielectric Properties

It is very important to know the material dielectric properties because they can be used to monitor
the specific properties of the product subjected to the physical and chemical processes. The three
most popular methods used are: open ended coaxial probe, transmission line and resonant cavity
method [22].
• Open ended coaxial probe method

The probe method is based on a coaxial line ending abruptly at the tip that is in contact
with the material being tested. The technique calculates the dielectric properties from
the phase and amplitude of the reflected signal at the end of open-ended coaxial line
inserted into a sample. For low values of dielectric constant and loss factor, as well as
for very low and very high frequencies, care must be taken in order to prevent errors.
It is the easiest method to use because it does not require a particular sample shape
or special containers. It also provides broad band measurements while minimizing
sample disturbance.
FIGURE 22.3 Dielectric constant (1) and loss factor (2) dependence versus alternating electric field frequency.
(From Badr, A.M., Elshaikn, H.A., Ashraf, I.M. 2010. Impacts of Temperature and Frequency on the Dielectric
Properties for Insight into the Nature of the Charge Transports in the Tl_2S Layered Single Crystals; p. 14.)
The only cases in which one needs to pay more attention with this method are:
– lower dielectric constant values,
– lower loss factor values, and
– extremely low or high frequencies.
• The transmission line method
The transmission line method involves placing a sample inside an enclosed transmis-
sion line where the cross section is precisely filled with the sample. During the mea-
surement, a vector network analyzer is used to detect the change of the impedance
and propagation characteristics due to the filling of dielectric material. The dielectric
properties of the tasted material are calculated by software. This method is more accu-
rate and sensitive than the probe method. The way of measurement makes this method
difficult to test liquid and semi-solid sample material.
• Resonance cavity method
The third method is the resonant cavity method, in which the sample of known geom-
etry is placed in a single-mode cavity. The dielectric property of the sample is then
computed using the reflected power of the cavity and the frequency of the resonance.
This method can be accurate and is especially suited for samples with a very low loss
factor. It provides dielectric properties at only fixed frequency [16,22]. The measure-
ment frequency range is from 50 MHz to more than 100 GHz. This method can be
applied to all liquid and solid materials but not to gases because of the low permittivity
of gases.
22.6 THERMAL–PHYSICAL PROPERTIES
22.6.1 Specific Heat
Specific heat is determined by calorimetric methods, including classical differential scanning calo-
rimetry (DSC). In the case study presented in this chapter, the frozen and unfrozen specific heats
of the sample considered in the proposed simulation (lean beef meat) of RF thawing of meat were
approximated by using equations Choi and Okos [30] based on the proximate constituents of the
sample (lean beef meat), Table 22.1.
The values were 1.9352 kJ kg–1 K–1 and 3.4878 kJ kg–1 K–1 for frozen and unfrozen lean beef meat,
respectively. The apparent specific heat is then calculated based on above approximated values and the
latent heat of fusion of lean meat (which is approximated from latent heat of fusion of water).
22.6.2 Thermal Conductivity
The thermal conductivity (Wm–1 °C–1) of lean beef meat varies with temperature and the value
increases with increasing ice content in the frozen food. Thermal conductivity used in simulation,
TABLE 22.1
Equations Used for the Prediction of Specific Heat Values
Specific heat Protein Cp = 2.0082 × 103 + 1.2089T – 1.3129 × 10−3T2
(kJ kg–1 K–1) Fat Cp = 1.9842 × 103 + 1.4733T – 4.8008 × 10−3T2
Carbohydrate Cp = 1.5488 × 103 + 1.9625T – 5.9399 × 10−3T2
Fiber Cp = 1.8459 × 103 + 1.8306T – 4.6509 × 10−3T2
Ash Cp = 1.0926 × 103 + 1.8896T – 3.6817 × 10−3T2
Source: Choi, Y. and Okos, M.R. 1986. Effects of temperature and composition on thermal properties
of foods. Journal of Food Process and Applications. 1(1): 93–101. 1986
FIGURE 22.4 Thermal conductivity as a function of temperature used in this work. (From Farag, K., Lyng,
J.G., Morgan, D.J. and Cronin, D.A. 2008. Dielectric and thermophysical properties of different beef meat
blends over a temperature range of –18 to +10 C. Meat Science. 79: 740–747.)
TABLE 22.2
Density Equations for Food Components
Density Proteins ρ = 1.3299 × 103 – 5.1840 × 10−1T
(kg m–3) Fat ρ = 9.2559 × 102 – 4.1757 × 10−1T
Carbohydrate ρ = 1.5991 × 103 – 3.1046 × 10−1T
Source: Choi, Y. and Okos, M.R. 1986. Effects of temperature and composition
on thermal properties of foods. Journal of Food Process and Applications.
1(1): 93–101. 1986
as shown in (Figure 22.4), considered as a function of temperature, was taken from experimental
data in Farag et al. [18], which was studied in the temperature interval between –18°C and +10°C.
22.6.3 Density
Density of the sample was approximated using the following equation:
1
ρ= (22.31)
xw
∑ i

ρi
where xiw is the weight fraction of the different components. In this study considered components
are: water, proteins, ice, ash, fat and carbohydrates. The ρi value is the pure density calculated using
the equation of Choi and Okos [30], Table 22.2. The composition of lean beef meat used in this study
is: moisture 67.8%; fat 11.8%, protein 17.9%, ash 0.8%. Using Table 22.2 the density of frozen lean
beef meat was estimated to be equal to 961 kg m–3 and the unfrozen density 1053 kg m3.
22.7 MODELING STUDIES FOR RF THAWING

As a typical study on transport phenomena of a process, the mathematical modeling of an RF thaw-
ing process also starts with a description of the system and the surrounding environment and a defi-
nition of the computational geometry in which the system properties, including the bottom and top
electrode with the system walls, are properly defined. The sample is generally assumed to be placed
on top of the bottom electrode or between the two electrodes, simulating the fact that it is placed
on an RF transparent belt (like a Teflon belt used typically in RF systems to vary/move the sample
through the system). This section is preferably not considered in the computational geometry, due
to its transparent nature to the electromagnetic energy and so as not to increase the complexities
in the system. Figure 22.5 shows a typical industrial scale RF system including a sample with the
generated mesh structure for the computational solution.
Before continuing with the simulations, numerical mesh independency studies should be performed
to decide upon the best mesh parameters and mesh structure. Nyquist criteria might be applied for a
proper solution of the electric field, while the mesh structure must be denser inside the sample consid-
ering the phase change. To determine the independency from the grid size, the refinement on the mesh
structure is applied based on, for example, the changes in the absorbed power or temperature within
the sample. The convergence of the electric field solution will be much faster than the convergence of
the temperature solution especially for phase change problems, e.g., thawing.
While modeling of RF heating is a perfect example of a multi-physics problem (where the elec-
tric field interaction with the product to lead to heat transfer and temperature evolution is solved), the
thawing with the phase change phenomena brings additional complexities. The Multiphysics software
COMSOL (COMSOL, AB, Stockhom, Sweden) has been chosen to serve this multi-purpose case,
starting with its earlier versions. The solution of this complex multi-physics follows the given sequence:
Solution or heating or thawing problem is first based on the solution of Maxwell equations
(Equations 22.8–22.11) to determine the electric field distribution.
The calculated electric field distribution through the volume is used to calculate the absorbed
power distribution (Equation 22.7), and
The calculated power distribution is then used in solving the energy equation to determine the
temperature distribution.
For RF heating applications, the so called quasi-static approach is used. The representation of the
electric field with the quasi-static approach is given by Gauss law:
∇ ⋅ (ε E ) = 0 (22.32)

where ɛ is the permittivity (as a function of location – temperature through the sample). To solve the
Gauss law to determine the electric field distribution, the required boundary conditions might be stated:
A certain electric potential value is applied at the top electrode based on the given applied
power of the system (V)—possible top electrode voltage variation is often assumed to be
negligible.
FIGURE 22.5 Computational geometry of a typical industrial scale RF system including a sample with the
generated mesh structure.
The bottom electrode is maintained at the ground condition (V = 0), and

The system walls might be assumed to be electrically insulated.
For the case of heat transfer boundary conditions, the general approach is to apply the third kind
convective boundary condition with a lower heat transfer coefficient (5 to 20 Wm–1-K–1) and medium
temperature. The thermal and physical properties of the sample should be defined to demonstrate
the specific heat change at the phase change temperature as explained in the above sections. A uni-
form constant temperature distribution at the start of the process is a common practice to define the
initial condition for solving the time dependent energy equation.
As explained, for an RF thawing modeling study, simultaneous solution of the Gauss law to
determine the electric field distribution and Fourier conduction heat transfer equation with power
generation to determine the temperature distribution is required after completing the computa-
tional geometry and defining material thermal–physical and dielectric properties and the initial and
boundary conditions. For this simultaneous solution of the time-dependent problem, an implicit
time-stepping scheme to solve the non-linear system of the equations at each time step might be
preferred with the required solvers depending upon the software.
22.7.1 Electric Field Distribution Inside a RF System

Uyar et al. [13] were one of the first to develop a mathematical model for the RF thawing process.
Following this study, Llave et al. [20] carried out another RF simulation for tempering of frozen
food products. Figure 22.6 demonstrates the electric field distribution and voltage contours in an RF
system where a thawing simulation was carried out based on the approaches listed above. For this
specific case, top electrode voltage was 1000 V while the initial temperature of the 25 × 15 × 17 cm
sample was –18°C (255.15 K). Temperature maps of this sample (at different heights in z-direction)
at 600 and 3600 s of the process is shown in Figure 22.7a,b. Due to the attraction of the electric
field by the sample (a magnet-like effect as explained by Erdogdu et al., [31]), temperature contours
resemble the electric field distribution where the hotter areas are likely to occur at the edges and
corners. Due to the increases over the edges and corners, the expected volumetric heating and
thawing based on the temperature evolution is not likely to meet, contrary to common knowledge.
Even though a certain temperature uniformity is assumed to be obtained using RF compared to the
conventional processes, there are still certain temperature differences among the edges, surface and
volume averaged temperatures. Figure 22.8 shows the edge, surface and volume averaged tempera-
ture change of the given sample.
22.7.2 Effect of Sample Load on the Efficiency of RF Process

As demonstrated in Figure 22.6, the electric field distribution inside an RF system is symmetrically
distributed; only a fraction might be used where a considerably smaller size product is used during
processing. To effectively benefit from the distribution of the electric field, there might be certain
sample size or load volume required during processing. Uyar et al. [32] focused on this issue, dem-
onstrating the power absorption by the sample and its temperature evolution while the sample size
was reduced between the top and bottom electrodes. In this study, the initial volume (0.398 L) of
a cubic was considered to be 100% load between the electrodes (where the distance between the
electrodes was 78.557 mm and the distance of the top electrode from the sample top surface was 2.5 cm).
The decrease of the sample volume from 100% to lower volumes led to considerable reduction in
the power absorption and the eventual temperature change where the simulations were carried out
applying a 400 W RF power in a 50 Ω system. Figure 22.9a,b show the volume averaged tempera-
ture evolution and the absorbed power by the sample.
As demonstrated here, the sample load in a RF system must be rather significant to obtain a rea-
sonable temperature increase. The lower the sample size within the RF cavity, the more RF energy
FIGURE 22.6 (a) Example of an electric field distribution in mid-xz cross-section (electric potential contours
with its vectors and direction); (b) Surface temperature distribution of the sample at 3600 s with electric poten-
tial contours (blue to red streamlines where green arrows are vectors for the electric field around the sample).
diminished directly through the ground electrode, while a smaller distance between the sample and
the charged electrode might result in runaway heating leading to unexpectedly high temperatures
through the surface. As concluded by Uyar et al. [32], tuning of the process design conditions is
required to obtain a high heating efficiency without further increasing the edge and surface tem-
peratures. Based on the results of this study, it might be possible to fill the RF cavity to obtain a
higher production volume within a reasonable process time.
22.7.3 Simulation of Sample Movement within an RF Cavity

In a typical industrial process, the movement of samples placed on a Teflon belt is preferred for
continuous RF processing. The movement of samples and the electrodes to make uniform the
temperature distribution inside the samples then becomes a challenging process to simulate. Chen
et al. [33,34] were one of the the first to simulate and validate the RF heating with conveyor move-
ment. In these studies, the movement of one sample was simulated using the constant distanced
electrodes. Bedane et al. [35] carried out a performance comparison between batch and continuous
RF thawing indicating the significant effect of electrode gap and slight improving of the heating
uniformity by continuous approach. Erdogdu et al. [31] demonstrated the effects of the sample
FIGURE 22.7 Temperature maps of a 25 × 15 × 17 cm sample (initial temperature: 255.15 K) after RF thawing
(1000 V top electrode voltage) at different heights in z-direction after (a) 600 and (b) 3600 s of the process.
FIGURE 22.8 Edge, surface and volume averaged temperature change of a 25 × 15 × 17 cm sample during RF
thawing at 1000 V top electrode voltage.
FIGURE 22.9 (a) Volume averaged temperature and (b) power absorption vs. processing time in a meat-based
sample, placed between fixed distanced electrodes during RF processing. (Modified and adapted from Uyar, R.,
Erdogdu, F. and Marra, F. 2014. Effect of load volume on power absorption and temperature evolution during radio-
frquency heating of meat cubes: A computational study. Food and Bioproducts Processing. 92: 243–251.)
movement (1.45 m/h) and the oscillated movement (0.45 m/h; 1-up and 1-down movement while
the product moves through the RF system) of the top electrode to optimize processing conditions
and make uniform the temperature distribution through the sample during an RF thawing process.
While a 1000 V charge at the top electrode was applied at a distance of 10 cm between sample
(25 × 15 × 17 cm tuna block with an initial temperature of –18°C) surface and top electrode,
up-down oscillated movement of the top electrode with sample movement was demonstrated to
increase the temperature uniformity. Change of the thermal–physical and dielectric properties as a
function of temperature might be obtained in Erdogdu et al. [31]. Figure 22.5 showed the geometry
of the RF system used in the simulations, and the two simulations scenarios (stationary and sample
movement through the system with up-down oscillation movement of the electrode) are shown in
Figure 22.10.
The change in the maximum and minimum temperatures of these two cases was demonstrated
in Figure 22.11a,b. Effect of moving sample and oscillation of the electrode was explained with the
FIGURE 22.10 Two simulations scenarios; (a) stationary and (b) sample movement through the system with
up-down oscillation movement of the electrode. (Modified and adapted from Erdogdu, F., Altin, O., Marra,
F. and Bedane, T.F. 2017. A computational study to design process conditions in industrial radio-frequency
tempering/thawin process. Journal of Food Engineering. 213: 99–112.)
FIGURE 22.11 (a) Volume averaged temperature and (b) power absorption vs processing time in a meat-
based sample, placed between fixed distanced electrodes during RF processing. (Modified and adapted from
Erdogdu, F., Altin, O., Marra, F. and Bedane, T.F. 2017. A computational study to design process conditions in
industrial radio-frequency tempering/thawin process. Journal of Food Engineering. 213: 99–112.)
FIGURE 22.12 Change in the power uniformity index (PUI) values of the configurations demonstrated in
Figure 22.10. (Modified and adapted from Erdogdu, F., Altin, O., Marra, F. and Bedane, T.F. 2017. A compu-
tational study to design process conditions in industrial radio-frequency tempering/thawin process. Journal of
Food Engineering. 213: 99–112.)
change in the power uniformity index (PUI), where the smaller the PUI, the more improved tem-
perature uniformity in a well-designed RF process.
1
V
⋅
∫ ( Pabs − Pabs )2 dV
PUI = V (22.33)
Pabs
where P abs (Wm–3), volume average power absorption of the sample is defined by:
1
P abs =
V ∫P abs dV (22.34)
V
The PUI change of the two cases, given in Figure 22.10, was shown in Figure 22.12.
22.8 CONCLUSIONS
RF application for thawing of frozen food products is a promising technology for industrial use, but
the process design and optimization to decide upon the process parameters are rather critical. The
over-heating problem observed, especially in the sharp corners and edges of a material, are constraints
in this process due to the problems in the thawed product quality. This observed temperature increase
might lead to certain microbial problems and even in cooked sections in the thawed product. Therefore,
a mathematical modeling approach to design a process for the given system and target product is a
feasible option. For this purpose, with the complexity of the thawing problem and the RF heating
together, the use of a multi-physics computer-aided modeling approach is suggested. This chapter
focused on different stages of developing such a model with virtual tools based on computer simula-
tion and demonstrated the various scenarios from the literature for designing an industrial process.
ACKNOWLEDGMENTS
Some of the simulated RF thawing data was obtained during the studies of the projects supported
by the Scientific Research Foundation of the Ankara University, Turkey (Project no: 15B0443005 –
PI: F. Erdogdu) and Scientific and Technical Research Council of Turkey (TUBITAK – Project no:
5155029 – PI: F. Erdogdu).
Authors Bedane, Casaburi and Marra acknowledge the support of University of Salerno (Italy),
FARB ORSA163144.
REFERENCES
1. Sun D.-W., Thermal Food Processing: New Technologies and Quality Issues, CRC Press, Boca Raton,
FL, USA, 2012.
2. Jones P.L., Rowley A.T., Dielectric dryers, In: Industrial Drying of Foods, Baker C.G.J., Ed., Blackie
Academic and Professional, London, UK (Chapter 8), 1997.
3. Piyasena P., Dussault C., Koutchma T., Ramaswamy H.S., Awuah G.B., Radio frequency heating of
foods: Principles, applications and related properties: A review, Critical Reviews in Food Science and
Nutrition, 53: 587–606, 2003.
4. Hartshorn L., Radio-Frequency Heating, George Allen & Unwin Ltd., London, UK, 1949.
5. Marra F., Zhan L., Lyng J.G., Radio frequency treatment of foods: Review of recent advances,
Journal of Food Engineering, 91: 497–508, 2009.
6. Chan T.V.C.T., Tang J., Younce F., 3-dimensional numerical modeling of an industrial radio frequency
system using finite elements, Journal of International Microwave Power Institute, 39: 87–105, 2004.
7. Neophytou R.I., Metaxas A.C., Combined 3D FE and circuit modeling of radio frequency heating
systems, Journal of Microwave Power and Electromagnetic Energy, 33: 243–262, 1998.
8. Cleland D.J., Prediction of freezing and thawing times for foods, Ph.D. Thesis, Massey University,
Palmerston North, New Zealand, 1985.
9. Dickerson R.W., Relationships between water content enthalpy, specific heat and thermal diffusivity of
foods, ASHRAE Transactions, 38(1): 525, 1977.
10. Mellor J.D., Seppings A.H., Thermo-physical data for designing a refrigerated food chain, Bulletin of
the IIR, Annexe-1: 349, 1976.
11. Heldman D.R., Food properties during freezing, Food Technology, 362: 92–109, 1982.
12. Huang Z., Marra F., Subbiah J., Wang S., Computer simulation for improving radio frequency (RF)
heating uniformity of food products: A review, Critical Reviews in Food Science and Nutrition, 58(6):
1033–1057, 2008.
13. Uyar R., Bedane T.F., Erdogdu F., Palazoglu T.K., Farag K.W., Marra F., Radio-frequency thawing of
food products—A computational study, Journal of Food Engineering, 146: 16–171, 2015.
14. Mudgett R.E., Electrical properties of foods, In: Engineering Properties of Foods, 2nd ed., Rao M.A.
and Rizvi S.S.H., Eds., Marcel Dekker, New York, USA, 1994.
15. Zhang H., Datta A.K., Electromagnetics of microwave heating: Magnitude and uniformity of energy
absorption in an oven. In: Handbook of Microwave Technology for Food Applications, Datta A.K. and
Anantheskawaran R.C., Eds., Marcel Dekker, New York, USA, 1–28, 2001.
16. Engelder D.S., Bufler C.R., Measuring dielectric properties of food products at microwave frequencies,
Microwave World, 12: 2–11, 1991.
17. Decerau R.V., Food Industry and Trade—Microwave Heating—Industrial Applications, Academic
Press, Cambridge, MA, USA, 1–10, 1985.
18. Farag K., Lyng J.G., Morgan D.J., Cronin D.A., Dielectric and thermophysical properties of different
beef meat blends over a temperature range of –18 to +10 C, Meat Science, 79: 740–747, 2008.
19. Llave Y., Treda Y., Fukuoka M., Sakai N., Dielectric properties of frzen tuna and analysis of defrosting
using a radio-frequency system at low frequencies, Journal of Food Engineering, 139: 1–9, 2014.
20. Llave Y., Liu S., Fukuoka M., Sakai N., Computer simulation of radiofrequency defrosting of frozen
foods, Journal of Food Engineering, 152: 32–42, 2015.
21. Bengtsson N.E., Risman P.O., Dielectric properties of foods at 3 GHz as determined by cavity perturba-
tion technique, Journal of Microwave Power, 6: 107–123, 1971.
22. Marra F., Lyng J., Romano V., McKenna B., Radio-frequency heating of foodstuff: Solution and valida-
tion of a mathematical model, Journal of Food Engineering, 79: 908–1006, 2007.
23. Sosa-Morales M.E., Tiwari G., Wang S., Tang J., López-Malo A., García H.S., Dielectric heating as a
potential post-harvest treatment of disinfesting mangoes I: Relation between dielectric properties and
ripening, Bio-systems engineering, 103, 297 to 303 thawing of frozen beef, Journal of Food Science,
63(2): 272–275, 2009.
24. Tang J., Dielectric properties of foods, In: The Microwave Processing of Foods, Schubert H. and
Regier M., Eds., Woodhead Publishing, Sawston, Cambridgeshire, UK, 22–40, 2005.
25. Guan D., Cheng M., Wang Y., Tang J., Dielectric properties of mashed potatoes relevant to microwave
and radio-frequency pasteurization and sterilization processes, Journal of Food Science, 69: 30–37,
2004.
26. Wang Y., Wig T., Tang J., Hallberg L., Dielectric properties of food relevant to RF and microwave pas-
teurization and sterilization, Journal of Food Engineering, 57: 257–268, 2003.
27. Komarov V., Wang S., Tang J., Permittivity and measurements, In: Encyclopedia of RF and Microwave
Engineering (3693-3711), Chang K., Ed., John Wiley & Sons, Hoboken, NJ, USA, 3693–3711, 2005.
28. Badr A.M., Elshaikn H.A., Ashraf I.M., Impacts of temperature and frequency on the dielectric prop-
erties for insight into the nature of the charge transports in the tl2s layered single crystals, Journal of
Modern Physics, 2(1): Article ID: 3764, 2011, 2010.
29. Venkatesh M.S., Raghavan G.S.V., An overview of dielectric properties measuring techniques, Canadian
Biosystems Engineering, 47(7): 15–30, 2005.
30. Choi Y., Okos M.R., Effects of temperature and composition on thermal properties of foods, Journal of
Food Process and Applications, 1(1): 93–101, 1986.
31. Erdogdu F., Altin O., Marra F., Bedane T.F., A computational study to design process conditions in
industrial radio-frequency tempering/thawin process, Journal of Food Engineering, 213: 99–112, 2017.
32. Uyar R., Erdogdu F., Marra F., Effect of load volume on power absorption and temperature evolution
during radio-frquency heating of meat cubes: A computational study, Food and Bioproducts Processing,
92: 243–251, 2014.
33. Chen L., Huang Z., Wang K., Li W., Wang S., Simulation and validation of radio frequency heating with
conveyor movement, Journal of Electromagnetic Waves and Applications, 30: 473–491, 2016.
34. Chen J., Lau S.K., Chen L., Wang S., Subbiah J., Modeling radio frequency heating of food moving on
a conveyor belt, Food and Bioproducts Processing, 102: 307–329, 2017.
35. Bedane T.F., Chen L., Marra F., Wang S., Experimental study of radio frequency (RF) thawing of foods
with movement on conveyor belt, Journal of Food Engineering, 201: 17–25, 2017.
23 CFD Application for the
Evaluation of Food Texture
Won Byong Yoon and Hwabin Jung
CONTENTS
23.1 Introduction.................................................................................................................................... 509
23.2 Food Texture..............................................................................................................................510
23.3 Classification of Computer Simulation Methods for Analyzing Food Texture................... 511
23.3.1 Finite Element Method (FEM).......................................................................................511
23.3.2 Finite Volume Method (FVM)..........................................................................................512
23.3.3 Discrete Element Method (DEM)........................................................................... 512
23.4 Application of Computer Simulation to Food Texture..............................................................512
23.4.1 Compression Behavior.................................................................................................512
23.4.1.1 Loading and Dropping of Fruits.................................................................513
23.4.1.2 Compression of Nuts............................................................................... 519
23.4.1.3 Compression Behavior of the Cellular Structure of Food....................... 519
23.4.2 Crack Propagation by Tensile Loading................................................................... 521
23.4.3 Oral Processing and Digestion................................................................................ 522
23.4.3.1 Mastication and Swallowing...................................................................... 523
23.4.3.2 Gastric Digestion..................................................................................... 525
23.5 Conclusions.......................................................................................................................... 527
References....................................................................................................................................... 528
23.1 INTRODUCTION
The quality of correlations between sensory and instrumental measurements is a key issue facing
researchers, in both basic and applied areas of texture studies. By definition, texture is a sensory
property [1]. Food texture can be a criterion to judge food quality such as ripeness of fruits and
cheese, or viscosity of drinks [2]. Textural quality attributes of food may be evaluated by descriptive
sensory (subjective) or instrumental (objective) analysis. The combination of a time-consuming pro-
cess and high cost associated with sensory analysis has promoted the development and widespread
use of empirical mechanical tests that correlate with sensory analysis of food texture [3].
Food structure has been studied at the macroscopic level (for example, breakdown through the
measurement of hardness) and at the microscopic and molecular level to understand the structure
fundamentally through rheological mechanisms [4,5]. However, food is frequently heterogeneous,
sensitive to environmental changes, and has fluid-containing structures [6], and therefore, a deeper
analysis is required for understanding the relationship among food texture, structure, and perception.
For a better understanding of food breakdown and change in structure under deformation condi-
tions, computer simulation techniques have been developed. The problem that originates from a
complex structure may be solved by using numerical methods, as this problem is difficult to over-
come through an analytical or experimental approach. Predicting internal stress distribution in food
and food breakdown or swallowing through the mastication process using computer simulations and
computational fluid dynamics (CFD) is advantageous in the food industry. Thus, this chapter will
mainly focus on the application of computer simulations and CFD to food texture and oral processing.
509
23.2 FOOD TEXTURE
Texture is one of the main physical factors of food that correlates with sensory perception and prod-
uct quality. Food texture has been defined by the International Organization for Standardization as
an attribute that encompasses all the rheological and structural (geometrical and surface) attributes
of a food product perceptible by means of mechanical, tactile and, where appropriate, visual and
auditory receptors [7]. Sensory measurements are usually taken to inform the evaluation by trained
and semi-trained descriptive panels, as well as the evaluation and acceptance by consumers. There
is increasing interest in designing instrumental tests that can predict consumer acceptance and
replace descriptive panels [8]. Instrumental methods to measure food texture have been classified
into the following three categories of rheological analysis: fundamental, empirical, and imitative
methods [9] (Figure 23.1).
Fundamental measurement properties are independent of the particular instrument and the size
or geometry of the sample used. These properties, such as viscosity, elasticity, and moduli, are
based on engineering theory and provide the same result for the same material. However, funda-
mental testing of food materials is often laborious, time-consuming, and may not be practical due
to the high complexity of food structure.
Empirical measurements are based on the resistance of the sample to force deformation, and
these methods include most of the instrumental techniques that are useful, quicker, and simpler
than fundamental measurements. The results of empirical measurements are relatively correlated
with those of sensory texture measurements. In spite of the advantages of empirical methods, the
measurements are not well defined and the testing conditions should be the same for making a
comparison. Examples of empirical measurements are a Bostwick consistometer, a Farinograph, a
Mixograph, an Extensograph, and a viscometer.
Imitative measurement attempts to mimic the conditions to which the material is subjected in the
mouth, and thus, it correlates well with a real-life situation. The most typical example of the imita-
tive measurement method is the instrumental texture profile analysis (TPA) and shear cell (Kramer
and Warner-Bratzler). These imitative methods have been developed to determine the association
between texture measurement and oral processes, including chewing and biting.
Regarding the current trend in the evaluation of food texture, food scientists have concluded
that traditional texture measurements, including a range of empirical, imitative, and rheological
measurements, cannot capture the great complexities of oral processing. The mechanical behavior
during chewing (compression, tension, and shear) transforms the food structure and induces het-
erogeneous distribution of mechanical properties. Therefore, to clarify the mechanism in the food
product, computer simulation has been used for recreating the reality of the deformation behavior
and for analyzing the local information on food.
FIGURE 23.1 Instrumental methods to measure food texture.

CFD Application for the Evaluation of Food Texture 511
23.3 CLASSIFICATION OF COMPUTER SIMULATION

METHODS FOR ANALYZING FOOD TEXTURE
In this chapter, the most commonly used numerical methods for computer simulation to analyze
food texture are classified into the finite element method (FEM), the finite volume method (FVM),
the meshless method, and the discrete element method (DEM). An appropriate numerical method
should be chosen to solve the problem. The continuity of the domain of interest cannot be sepa-
rated for the continuum physics-based simulation methods, such as the FEM, FVM, and meshless
method. Thus, the numerical solution is calculated by the finite number of continuous nodes or ele-
ments, which reduces the system complexity. Contrarily, the discontinuum-based model, including
the DEM, represents the food structure as an assemblage of separate elements or discrete blocks that
interact with each other. These continuum- and discontinuum-based models or combined models
have been applied to simulate the continuous and discontinuous deformation behavior or breakage
of food, and are chosen depending on the characteristic to be analyzed (Table 23.1).
23.3.1 Finite Element Method (FEM)

The FEM is the most popular numerical method for solving engineering problems described by
partial differential equations (PDE). In this method, the continuum is subdivided into elements
of variable size and shape that are interconnected in a finite number of nodal points. In every ele-
ment, the unknown solution is expressed as a linear combination of so-called shape functions. In
the next step, the equations are spatially discretized over the finite element mesh using a suitable
technique [10]. This method has been successfully applied in many areas of engineering sciences
to study, model, and predict the behavior of structures [11]. This method has also been found to be
useful in researching the complex structure of organic materials, and food under various bound-
ary conditions is appropriately and accurately estimated using the FEM. Studies relating the use of
FEM applications to estimate internal mechanical behaviors of agricultural products are abundant.
For example, fragmentation of porous food, failure of brittle food, deformation behavior induced
by mechanical damage, and swallowing of food have been numerically simulated using the FEM.
FEM is a powerful tool for engineering analysis and is used to solve problems ranging from the
very simple to the very complex. Time constraints and limited availability of product data call for
many simplifications of the analysis models. On the other hand, specialized analysts implement
FEM to solve very advanced problems, such as crash dynamics, material forming or analysis of
TABLE 23.1
Overview of the Most Notable Numerical Approaches
and Methods for Food Texture
Continuum Methods Applications
• Finite element method (FEM) Fragmentation and damage of brittle food
Deformation by mechanical damage
Swallowing of food
• Finite volume method (FVM) Swallowing of food
Flow phenomena in gastrin
Discontinuum Method
• Discrete element method (DEM) Fragmentation
Failure of agglomerate
Gastric digestion
Mastication behavior
biostructures. Today, with the help of advancing technologies, computers, and CFD software allow
us to use FEM in nearly all disciplines of engineering [12].
23.3.2 Finite Volume Method (FVM)

The FVM is a method for solving PDE in an integral sense in finite volumes. The computational
domain is discretized in finite volumes and the balance principles of this method are the basis for
the mathematical modeling of continuum mechanical problems. The finite volumes formulation
uses small “finite volumes” inside each element, where both physical and chemical properties are
considered constant [13]. The conservation laws underlying the governing equations are imposed at
the level of every finite volume, and applying Green’s theorem then naturally leads to a relationship
between fluxes at the finite volume boundaries. These fluxes are approximated by finite differences,
and the end result is again a system of algebraic or differential equations in the unknowns at the
discretization points [10]. The swallowing process of liquid food and flow phenomena in gastrin
have been numerically simulated using the FVM.
23.3.3 Discrete Element Method (DEM)

The DEM has been used as an alternative method to avoid the drawbacks of the continuum method
since it is a meshless method. DEM is based on the concept that individual material elements are
considered to be separate and are connected only along their boundaries, or through longer range
mechanisms, by appropriate physically based interaction laws [14]. DEM is capable of providing
detailed information about the evolution of the internal geometric structure. The DEM is a powerful
numerical method initially proposed for predicting the behavior of particulate media. It is a popular
and natural technique to study the dynamics of granular geomaterials and powders. The method
consists of describing the material as an assembly of interacting particles. Particle motions obey
Newton’s second law where displacements and rotations are updated at a suitable time increment
[15]. The DEM was conducted for the food breakdown pathways during oral processing and for the
establishment of links between food fragmentation and initial food structure. Recently, this method
has been proposed to tackle several challenges related to food fragmentation, such as breakdown of
foamy and brittle food, due to its ability to represent the material as a large number of discrete units.
23.4 APPLICATION OF COMPUTER SIMULATION TO FOOD TEXTURE

23.4.1 Compression Behavior
In food science, resistance to compressive stresses is an important physical signature of the trans-
formed food item as it represents an intrinsic component of its texture. The uniaxial compression test
has been used in many laboratories. One of the most important experimental methods for determin-
ing the related engineering properties and deformation characteristics of agricultural products is the
compression test. The compression elastic modulus (Young’s modulus) at a small deformation range
may be the first sensation in the mouth when a solid food is ingested and before the mastication pro-
cess begins. In addition, the fracture stress at large deformation in the compression test is directly
related to the stimulus in the mouth at the first bite [16]. The physics and mechanisms of breakdown
of composites in food are relevant to the compressive properties. Low-speed compression, which is
described as quasi-static loading, is commonly performed in the linear elastic region. On the other
hand, actual compression such as high-speed compression and dynamic loading occurs with high
nonlinearity and time-dependent deformation. However, it is difficult to predict the dynamic impact
of non-linear deformation/damage/failure characteristics of agricultural and food products. For this
reason, numerical simulation may be effective because it can predict both linear and non-linear
deformation characteristics.
23.4.1.1 Loading and Dropping of Fruits

Loading and dropping are the major factors involved in mechanical impact during post-harvest and
storage of agricultural products. It is obvious that static and dynamic forces act on agricultural prod-
ucts from origin to end user, as they are exposed to multiple processes. Unfortunately, fresh fruits are
very susceptible to mechanical damage during harvesting, packaging, and transport, and their quality
can be substantially reduced due to poor handling. Impact areas become discolored due to the release
of enzymes from damaged cells; however, the internal damage is not always immediately visible.
There have been efforts to evaluate the effect of static and dynamic loading during harvest
and post-harvest stages on the mechanical damage to fruits, vegetables, and biological matters.
Quantification of deformation and internal stresses is a very complicated task, since they are caused
by external forces and depend on the cell structure of fruits and vegetables. Therefore, prediction of
the level of damage, stress distribution, and deformation of the organic material under the external
forces has become a very important issue. It should also be emphasized that it is a very difficult
to measure or/and describe internal stresses associated with mechanical impact forces, due to the
biological cell structure of fruits and rapid deformation progression during cases of dynamic and
nonlinear impact.
In light of this finding, numerical methods can be utilized as an efficient alternative solution for
predicting the stress distribution that occurs during impact. Researchers have investigated the 3D
modeling methods of fruits and the mechanical damage progression in fruits under external forces.
Researchers are always looking for new ways of developing non-destructive evaluation methods for
agricultural products; one of these ways is to use the FEM, which has been used in areas such as
heat, vibration and structural analysis. The FEM can also be used to model irregular and heteroge-
neous objects. It can solve problems such as a change in the geometrical shape during loading, con-
tact regions, and distribution of stress and strain in fruits and vegetables. Many reports are available
in the scientific literature on the FEM and resonant frequencies of various kinds of near-spherical
agricultural objects, such as tomatoes, pears, apples, and other fruits.
Tomatoes are susceptible to mechanical damage because of their soft texture [17]. The multibody
system of a tomato, which consists of exocarp, mesocarp, and locular gel tissues, has been accu-
rately modeled using FEM by the two types of structural elements that have plasticity, large deflec-
tion, and large strain capabilities, are suitable for the curved boundaries, and can tolerate irregular
shapes (Figures 23.2 and 23.3) [18,19]. These studies confirmed that the FEM model can be used to
FIGURE 23.2 Geometry and structure of a tomato. (From Li, Z., Li, P., Yang, H., and Liu, J., J. Food Eng.,
116, 639, 2013. With permission.)
2.321 23.184 45.176 67.342 85.251
110.266 144.536 176.217 176.217

60.339
285.092 285.092 MX
118.358 197.348 243.896 MX MX MN
MX MX 393.698 393.698
176.376 284.43 343.257
502.843 502.843
234.395 MN 371.511 MN 442.617 MN MN 582
582
292.413 458.593 541.977 611.718 615.495
(a1) F1 (a2) F2 (a3) F3 (a4) F4 (a5) F5

.188674 1.884 3.669 5.475 7.228
6.353 18.905 32.137 32.137 32.137
57.047 57.047 57.047

12.517 35.928 MX
MX MX MX
81.956 81.956 81.956 MX
18.681 52.947
108.886 106.866 106.866
24.845 MN 69.968 MN 122 MN 122 MN 122 MN
31.009 86.889 126.578 129.456 131.776
(b1) F1 (b2) F2 (b3) F3 (b4) F4 (b5) F5

.063578 .646382 1.254 1.885 2.531
4.553 4.553 4.553 4.553

.47588
6.575 6.575 6.575 6.575
.888182 MX
MX MX MX
8.597 8.597 8.597 8.597
1.3 MX
10.619 10.619 10.619 10.619
MN MN MN MN MN
1.713 12 12
12 12
2.125 12.383 12.383 12.537 12.641
(c1) F1 (c2) F2 (c3) F3 (c4) F4 (c5) F5

.01684 .01192 .012477 .012123 .012115
.075247 .356973 .550109 .714777 .859356
.133655 .702026 1.086 1.417 1.707
.192062 1.047 1.625 2.12 2.554
.25047 1.392 2.163 2.823 3.401
.308877 1.737 2.701 3.525 4.246
.367284 2.082 3.238 4.228 5.096
.425692 2.427 3.776 4.931 5.943
.484099 2.772 4.314 5.633 6.79
.542507 3.117 4.851 6.336 7.637
(d1) F1 (d2) F2 (d3) F3 (d4) F4 (d5) F5
FIGURE 23.3 Nodal solutions of stress and displacement vector sum of different types of tissues of three
locular tomatoes compressed at five loading forces. Point MX indicates the position of maximum stress of tis-
sues, and Point MN indicates the position of minimum stress of tissues. Stress unit: KPa. (a1–a5), (b1–b5) and
(c1–c5) are the change processes of stress of exocarp, mesocarp and locular gel tissues respectively. (From Li,
Z., Li, P., Yang, H., and Liu, J., J. Food Eng., 116, 639, 2013. With permission.)
predict the mechanics of the tomato (Figures 23.4 and 23.5) [19]. Also, a sample cherry tomato has
been modeled as a 3D solid, which is assumed to be nearly spherical and solid and possesses skin
that has the same properties over the entire body. Deformation behavior simulation on dropping a
sample cherry tomato with a physical test using FEM showed good agreement [12].
Pear fruit also has a soft tissue that must be protected against mechanical bruising. Structural
deformation/stress progression of time-dependent bruising of pear fruit with an irregular shape
was investigated and the complex three-dimensional bruise volume was described using a nonlin-
ear FEM-based explicit dynamics simulation approach. Instead of common linear elastic material
properties, elastic-plastic nonlinear material properties that were experimentally determined were
used to describe bruising efficiently (Figure 23.6) [11]. The bruised areas of unripe, ripe, and over-
ripe pear fruits dropped from different heights were experimentally determined and modeled by
FIGURE 23.4 Whole tomato fruit compression test (a), section of tomato fruit (b), and multiscale geometri-
cal modeling (c). (From Li, Z., and Wang, Y., Postharvest Biol. Technol., 121, 19, 2016. With permission.)
FIGURE 23.5 Deformation and damage situation of tomato fruit model (a and b) and the corresponding
force-deformation and pressure-deformation curves (c and d). (From Li, Z., and Wang, Y., Postharvest Biol.
Technol., 121, 19, 2016. With permission.)
using the FEM. Different fruit orientations and contact surfaces were considered in the research
performed by Yousefi et al. (Figure 23.7) [20] and Salarikia et al. (Figure 23.8) [21].
Large quantities of apple are destroyed or suffer from quality loss because of bruising and other
types of mechanical damage during harvesting, transportation, storage, and packaging. Lewis et al.
[22] analyzed the static pressure loading between parallel plates based on FEM. To examine the
bruise, apples were modeled completely in a 3D-shape and elastically. They utilized laser scanning
to provide the real geometrical shape of the apple and considered the mechanical properties of apple
to be elastic. Through the video-recording of scenes with a high-frame camera, Celik et al. [23]
simulated the deformation behavior of an apple falling from different heights by using the FEM
(Figure 23.9). Using a combination of energy analytical method and numerical modeling in FEM,
Cherng et al. [24] suggested a new strength index for oval-like elastic fruits. Lu et al. [25] statisti-
cally analyzed the contact stress distribution resulting from the pressure of soft-bioyield probes in
Simulation Plot: Equivalent Stress MPa Impact surface: Wood (Y. Popular)
Impact height : 1 m Impact angle: 0°
Outer Surface Plot
Y Y Y Y Y
Z X Z X Z X Z X Z X
0.00 40.00 80.00 (min) 0.00 40.00 80.00 (min) 0.00 40.00 80.00 (min) 0.00 40.00 80.00 (min) 0.00 40.00 80.00 (min)
20.00 60.00 20.00 60.00 20.00 60.00 20.00 60.00 20.00 60.00
0 Max
0.115
0.152
0.189
0.226
0.263
0.3
0.0009741 Min
0.077994
0.075731
0.11311
0.15049
0.187870.22524
0.22524
0.26262
0.3
0.040993
0.54285 Max
0.3
0.3
0.037606
0.15006
0.00030916 Min
0.037771
0.10811 Max
0.22508
0.00012078 Min
0.15015
0.0039922 Min
0.3
0.3
0.22503
0.45516 Max
0.18762
0.26254
0.038352
0.075235
0.10714 Max
0.26252
0.18755
0 Min
Cross Section Plot
Y Y Y Y Y
Z X Z X Z X Z X Z X
0.00 40.00 80.00 (min) 0.00 40.00 80.00 (min) 0.00 40.00 80.00 (min) 0.00 40.00 80.00 (min) 0.00 40.00 80.00 (min)
20.00 60.00 20.00 60.00 20.00 60.00 20.00 60.00 20.00 60.00
Simulation Time Simulation Time Simulation Time Simulation Time Simulation Time
0s 1.05 10-3 s 2.4 10-3 s 4.4 10-3 s 1 10-2 s
FIGURE 23.6 Simulation visuals of stress progression at different simulation time during dropping of pear.
(From Celik, H.K., Postharvest Biol. Technol., 128, 83, 2017. With permission.)
FIGURE 23.7 The drop test of pear on steel surface at different orientations, (a) 0°, (b) 90°, (c) 30°, (d) 60°
and (e) 45° from horizontal line. (From Yousefi, S., Farsi, H., and Kheiralipour, K., Biosyst. Eng., 147, 17, 2016.
With permission.)
S, Mises (Pa)
296870
272131
247392
222653
Step 1 Step 3 Step 6 Step 9 Step 12 Step 15 197913
173174
148435
123696
98957
74218
49478
Step 30 24739
Step 18 Step 21 Step 24 Step 27 0
Step 16
FIGURE 23.8 The stress distribution in the inner structure of pear at different tie steps after its horizontal
impact on a steel surface. (From Salarikia, A., Ashtiani, S.H.M., Golzarian, M.R., and Mohammadinezhad,
H., Inf. Process. Agric., 4, 64, 2017. With permission.)
FIGURE 23.9 Drop simulation and high speed camera comparison of the apple. (From Celik, H.K., Rennie,
A.E., and Akinci, I., J. Food Eng., 104, 293, 2011. With permission.)
a compression test on the apple by using the FEM. The effect of probes of different size, thickness,
and elastic modulus on contact pressure distribution was investigated in different samples of apple.
In the study by Ahmadi et al. [26], the behavior of various layers of skin, cortex, and core of the
apple was analyzed by applying the impact test to collision of the fruit with a flat rigid plate as well
as the collision of an apple with another apple using FEM (Figures 23.10 and 23.11).
Researchers have used the FEM to predict the mechanical behavior of other agricultural prod-
ucts. Cardenas-Weber et al. [27] studied the mechanical properties and internal stress on melons by
use of FEM. They reported that the relative error in simulation was 11%, which was a significant
approximation of the experimental data. The finite element technique was used by Nourain et al.
to determine the optimum excitation location of the measurement sensor and to analyze the mode
shapes of the melon [28]. They suggested that the middle part on the fruit surface should be used
while estimating the fruit firmness by the resonant frequency of the first-type spherical mode. The
internal bruising of two varieties of watermelon was predicted by Sadrnia et al. using FEM [29].
The FEM applications include non-destructive sensing of the firmness of the melon [30]. Also, the
FIGURE 23.10 The modeling steps of apple: finite element model of fruit with separate sections for each
piece of fruit. (From Ahmadi, E., Barikloo, H., and Kashfi, M., Comput. Electron. Agric., 121, 1, 2016. With
permission.)
FIGURE 23.11 (A) The collision of apple to apple and (B) the stress distribution in the FEM for the collision
of apple to apple with the speed of 1 m/s. (From Ahmadi, E., Barikloo, H., and Kashfi, M., Comput. Electron.
Agric., 121, 1, 2016. With permission.)
mechanical properties of the skin and flesh of cantaloupe were determined by individual experi-
ments, and then, finite element simulation was developed for cantaloupe fruit under compression
loading [31]. The force threshold of internal bruising for cantaloupe was determined by FEM. These
studies show that FEM can be a suitable method for determining the deformation behavior of fruits.
23.4.1.2 Compression of Nuts
A mechanical shelling operation is essential to obtain the pure peeled kernel and this operation
is the most critical and delicate step for achieving high-quality kernels in the shelled edible agri-
cultural product processing industry. In this regard, the engineering properties of shelled edible
agricultural products are a prerequisite for the design and development of cracking/shelling systems
[32]. In addition, most manufacturers avoid the use of chemical extraction methods in order to pre-
serve the natural nutty flavor and nutrients of oil during edible oil production. Therefore, mechani-
cal extraction methods are used to extract oil. This is basically a compression process. During this
process, it is important to determine the mechanical properties and deformation characteristics of
kernels in order to proceed with optimally designed oil extraction systems [33]. Experiment-based
material models of pecan shelling, which were used in the nonlinear time-dependent FEM-based
simulation procedure, were successfully described in the research by Celik et al. [32]. Realistic
high-speed loading cases were successfully simulated as was intended. Simulation outputs exhib-
ited logical structural deformation characteristics. Also, the elastic-plastic deformation case embed-
ded to the material model used in the non-linear time dependent FEM-based simulation procedure
and realistic compression simulation of pecan kernel was accomplished [33] as aimed. The FEM
was used to model the mechanical behavior of the sunflower fruit hull [34]. Petru et al. [35] applied
the FEM to study the mechanical properties of Jatropha curcas seed under compression loading at
three different stages of maturity (ripe, unripe, and over-ripe). They found that FEM can be useful
for optimizing and designing the press machines (Figure 23.12). Khodabakhshian et al. [36] devel-
oped a model to study the mechanical properties of pumpkin seeds by FEM under parallel plate
compression. They reported that FEM can accurately explain the experimental results.
23.4.1.3 Compression Behavior of the Cellular Structure of Food

Materials with voids are found in both nature and engineering. Cellular solids consist of a three-
dimensional interconnected network of solid struts or plates, which form the edges and faces of
cells, respectively [37]. Consumer appreciation of solid food foams like bread, extruded cereals,
biscuits and cakes is strongly linked to the texture. The control of the sensory properties of such
products, which is still a challenge, requires a better understanding of relationships between compo-
sition, cellular structure formation mechanisms, and final texture. Such understanding copes with a
possible control of food mastication by designing food using degradation criteria [38].
Breakfast cereals are airy, transformed cereals that exhibit a highly porous structure. The study
of their fragmentation behavior using simple compression tests can provide information about their
FIGURE 23.12 Principal stress in model of Jatropha curcas L. seed at three stages: unripe, ripe, over-ripe at 75%
strain. (From Petrů, M., Novák, O., Herák, D., and Simanjuntak, S., Biosyst. Eng., 111, 412, 2012. With permission.)
degradation behavior, especially if we relate such a result to oral processing. Compression loading is
also a discriminating test to evaluate the crispness of the transformed product [39]. Also, fragmenta-
tion is viewed as a result of the mastication process. In some studies, the notion of fracturability is
used to relate the food breakdown to the mechanical energy available during the oral process [40].
Different bread crumbs leading to distinct cellular structures have been investigated by Babin
et al. [38]. An in-situ study by tomography is performed to characterize not only the final cellular
structure, but also to follow the whole process of development of microstructure during the fer-
mentation stage. Two finite element approaches are performed on the tomography images of the
final products; the results are compared with the experimental compression test data and they show
good agreement. The study by Mamlouk and Guessasma [39] encompasses compression testing and
fragment size evaluation of airy breakfast cereals using 2D image analysis. Structural information
about the airy structure is then determined using X-ray tomography and related 3D image analysis.
The numerical part has the ambition of assessing the fragmentation process using a damage-based
mechanical model that simulates solid material rupture events as onset and growth of damage up to
brittle failure. The model is based on a finite element scheme in which direct information about the
3D airy structure is encoded in the solid meshing that accounts for the product shape, cell wall, and
airy structure properties (Figure 23.13). FEM has been used by Guessasma et al. to calculate effec-
tive Young’s modulus using a regular meshing scheme of a 3D typical cellular structure of bread
crumbs [41]. It is found that sphere overlap has the larger effect compared to sphere distribution
width for a given density. Experimental evidence of the role of void overlapping is treated in the case
of bread crumbs structures determined using X-ray tomography. Using the FEM, Mohammed et al.
[37] predicted the compressive response of the wafer to a high level of accuracy both qualitatively
and quantitatively at large global strains. The cell walls of the model were assigned a linear elastic
material model and a damage criterion to simulate fracture of the cell walls (Figure 23.14).
The FEM approach is usually accurate when it is used to virtually test porous materials under
small strain conditions. However, for large strains, the method is limited by problems related to
opposing cell wall contact and distortion of elements. The DEM has been used as an alterna-
tive that avoids the above-mentioned drawbacks since it is a mesh-free method. In the study by
FIGURE 23.13 FEM showing the loading conditions, meshing of breakfast cereal product, and the principal
stress counterplot. (From Mamlouk, H., and Guessasma, S., Innov. Food Sci. Emerg. Technol., 19, 190, 2013.
With permission.)
S. Mises
30.00
12.00
11.25
10.50
9.75
9.00
8.25
7.50
4.75
6.00
5.25
4.50
3.75
3.00
0.00
Y
a ε=0 b ε = 0.018
Z X
c ε = 0.036 d ε = 0.091
e ε = 0.164 f ε =0.209
FIGURE 23.14 The damaged wafer at different stages (a–f) in the compression (stress in MPa). The initial
fracture was represented at approximately 10% strain (d). (From Mohammed, I.K., Charalambides, M.N.,
Williams, J.G., and Rasburn, J. Innov. Food Sci. Emerg. Tech., 24, 48, 2014. With permission.)
Hedjazi et al. [42], DEM has been employed to study the fragmentation of breakfast cereals (Figure
23.15). The maximum force values for fragmentation and the size of resulting fragments are in good
agreement with experimental responses. The numerical results show that the observed fragment
size distribution is the consequence of a small number of rupture events of cell walls. This result
highlights the role of the airy structure associated with a particular tendency to form a bimodal
size distribution of fragments. The research described in the study by Hanley et al. [43] aimed to
calibrate a DEM for infant formula agglomerates using experimental data obtained for quasi-static
loading and to use this model to study the mechanics of the particle response in detail. DEM is
capable of providing detailed information about the evolution of the internal geometric structure of
an agglomerate over time.
23.4.2 Crack Propagation by Tensile Loading

Human mastication of food products can be regarded as a complex mechanical process leading to
food particle breakdown. Because of its importance in the other degradation steps, understand-
ing such a process depends on our ability to address fracture mechanisms that are associated with
mastication. The idea of comparing mastication to simple mechanical tests is the first step in under-
standing material fracturing behavior. Thus, in several crispy products, food breakdown operates
FIGURE 23.15 (a) Lateral, (b) perspective, and (c) cross-section views showing fragments of breakfast cere-
als at final stage of compression in the numerical simulation. (From Hedjazi, L., Martin, C.L., Guessasma, S.,
Della Valle, G., and Dendievel, R. Food Res. Int., 55, 28, 2014. With permission.)
following the idea of the opening mode, which requires local tension loading and the presence of a
defect for crack propagation. Upon loading, stress concentration in the vicinity of the defect leads to
crack propagation and material failure if enough energy is delivered [44]. Extensive work has been
performed on the mechanisms leading to crack initiation and propagation in various brittle materi-
als. Numerical methods can help in describing such mechanisms using increasing computational
resources and more elaborated models that are capable of handling local microstructural informa-
tion [15].
A finite element approach has been developed to account for the explicit effect of the void struc-
ture of extruded starch on crack propagation (Figure 23.16) [45]. In these materials, complex stress
distribution takes place upon loading leading to a jagged crack extension. Based on the principle
of stress intensity factors, the maximum energy release criterion was implemented and combined
with a particular 2D meshing of the material. Within the context of linear elastic fracture analysis,
a regular meshing is successfully considered for the prediction of small crack angle variation in a
heterogeneous carbohydrate porous material. In addition, crack propagation in starch material with
a hole under tensile loading has been studied considering the numerical and experimental aspects
(Figure 23.17) [44]. The experimental results show that the crack path is affected by the presence of
the defect. The results predicted by using FEM show that the selected criterion adequately antici-
pates the crack propagation in vitreous starch under linear elastic conditions. Another more recent
application is DEM modeling of dense materials with bonded particles. It is the set of microprop-
erties that enables the dense material to be modeled. DEM has been proven to be more sensitive
to stress heterogeneities than FEM under the criterion of maximum release of elastic energy [15].
DEM gives results that are closer to the considered analytical model, when the crack tip is not too
close to the hole (Figure 23.18). Crack branching predicted by DEM under no specific criterion for
the creation of secondary cracks shows good agreement.
23.4.3 Oral Processing and Digestion

Food oral processing is an essential procedure not only for the ingestion, mastication, swallowing,
and digestion of foods but also for the appreciation and pleasure of food texture and food flavor [46].
To estimate the texture perception and mouth-feel properties of a newly developed product before its
introduction into the market, sensory investigations are necessary in the industry. There is relatively
sparse up-to-date literature in the field of research dealing with numerical simulations of intra-oral
flows during the intake of food. This holds also for mechanically induced stresses due to flow.
a b
0 6.709 26.839
3.354 10.06313.41816.77220.12623.181 30.669 0 1.065 2.131 3.196 4.261 5.326 6.392 7.457 8.522 9.739
c d
0 1.957
.978725 2.936 3.915 4.894 5.872 6.851 7.83 8.948 0 1.852 3.705 5.557 7.41 9.262 11.11412.96714.81916.936
e f
0 2.31 4.62 6.93 9.24 11.55 13.86 16.17 18.48 21.12 0 .912141
1.824
2.736
3.649
1.561 5.473 6.385 7.297 8.31
FIGURE 23.16 Simulated stress distribution of carbohydrate extruded starch as function of crack extension
corresponding to the simulation steps (a–f). (From Hedjazi, L., Guessasma, S., Della Valle, G., and Benseddiq,
N. Carbohydr. Polym., 83, 1696, 2011. With permission.)
Our knowledge regarding how food changes during the oral process is still quite limited due to the
fact that dynamical properties of food breakage are invisible and immeasurable during the actual
process, although it is crucial for human perception of food. Nevertheless, numerical mechanics
offers a very broad spectrum of possibilities for the investigation of mechanical phenomena in the
human body. Treating the mechanical phenomena in the mouth during food intake in the frame of
numerical simulations is a recent approach and it possesses great potential. This leads to highly
necessary and industrially relevant, objectified prediction of the mouth-feel of foods.
23.4.3.1 Mastication and Swallowing

Mastication is a complicated phenomenon in which many morphological and functional compo-
nents act interdependently. For analyzing the dynamical properties of fluid food and mastication
components, the FEM is a useful tool. Multiphase materials (solid food, fluid food, saliva, volatiles
released from food, etc.) are involved in the mastication process. The techniques that combine DEM
with CFD and FEM for numerical studies of multiphase flow systems have gained close attention
of researchers and promise to be a method for mastication system research. The movement and
0 15.52 31.04 46.56 62.081 77.601 93.121 108.641 129.335

(a) (b)
0 18.431 36.863 55.294 73.726 92.157 110.588 129.02 153.595 0 17.556 35.112 52.668 70.225 87.781 105.337 122.893 146.301
(c) (d)
0 13.031 26.062 39.093 52.123 65.154 78.185 91.216 108.591 0 12.242 24.485 36.727 48.97 61.212 73.454 85.697 102.02
(e) (f )
0 18.265 36.53 54.795 73.06 91.324 109.589 127.854 152.207

0 15.816 31.632 47.448 63.264 79.08 94.896 110.712 131.8
(g) (h)
FIGURE 23.17 Stress distribution of carbohydrate extruded starch with a hole as function of crack
growth corresponding to the simulation steps (a–h). (From Hedjazi, L., Guessasma, S., Della Valle, G., and
Benseddiq, N., Eng. Fract. Mech., 78, 1328, 2011. With permission.)
interaction of food particles and the contact forces between the particles and teeth have been ana-
lyzed. Primary biting and swallowing simulations have been conducted to validate the feasibility of
DEM in mastication investigation by Sun and Xu [14]. Results from simulations can help to under-
stand the masticatory performance and to promote food product development.
The mechanism of swallowing is still not fully understood, because the process of swallowing is
a rapid and complex interaction among several involved organs and the food bolus. The process of
swallowing is a rapid and complex interaction between the food bolus and the involved organs, i.e.,
the tongue, palate, pharynx, esophagus, and larynx. Simulation of swallowing enables quantitative
FIGURE 23.18 Comparison between FEM, DEM, and experimental crack propagation for stress distribu-
tion of carbohydrate extruded starch with a hole as function of crack growth for three different experimental
positioning of the hole (a–c). (From Hedjazi, L., Martin, C.L., Guessasma, S., Della Valle, G., and Dendievel,
R., Int. J. Solids Struct., 49, 1893, 2012. With permission.)
and systematic virtual examinations of swallowing under detailed control of the organ motions and
the physical properties of the food bolus [47]. Thus, numerical modeling is suitable for the quantita-
tive analysis of the swallowing process. Swallowing movements have been modeled using an FEM
of an elastic pharynx in the study by Mizunuma et al. [48]. A rheological model of barium jelly and
a friction model between the jelly and the organs were both integrated into the numerical model
[48].
The objective analysis and prediction of mouth-feel attributes during the swallowing of yoghurt
using FVM have been conducted by Rauh et al. [49]. By numerical simulations, fluid mechanical
interactions of yoghurt and mouth surface were determined and evaluated in a 3D tongue-palate
model (Figure 23.19). The numerical simulations revealed time-dependent characteristic pressure
and shear stress distributions at the surface of the tongue and palate induced by the flow of yoghurt
during swallowing. The model fluids were selected to cover a realistic range of different fat and
protein contents, and furthermore, to represent different rheological behaviors (Newtonian fluids
and non-Newtonian fluids). As a result of the numerical simulations of spatiotemporal distributions
of pressure, velocity and stresses were gained.
23.4.3.2 Gastric Digestion
The underpinning knowledge of the fluid-dynamics of gastric contents is critical to under-
stand and model the breakdown of food structures and consequent release of nutrients during
digestion. CFD applications provide a unique and accurate insight into the fluid dynamics
of the gastric contents, offering a fundamental understanding of the mechanisms involved in
the digestion process [50]. Computational modeling has been recently used as an alternative
approach to investigate the dynamics of gastric contents. Use of CFD techniques to develop
a 2D and 3D models of the geometry and motility of the human stomach during digestion
FIGURE 23.19 Wall shear stresses in the mouth produced by Newtonian (water) and non-Newtonian (yoghurt)
fluids. (From Rauh, C., Singh, J., Nagel, M., and Delgado, A., Int. Dairy J., 26, 2, 2012. With permission.)
and to characterize and compare the fluid dynamics of gastric contents of different viscosi-
ties were investigated by Ferrua and Singh [51] and Ferrua et al. (Figure 23.20) [50]. Imai et
al. [52] reconstructed a numerical model of fluid dynamics in the stomach, using an anatomi-
cally realistic geometry and free-surface flow modeling with the moving particle semi-implicit
(MPS) method (Figure 23.21). A numerical model was used in the study by Ferrua et al. [53]
FIGURE 23.20 Streamlines of Newtonian gastric flow (1 cP) colored by velocity (cm/s). (a) Within a 2-D
representation of the stomach (non-slip condition). (b) Within the middle plane of a 3-D model (free slip
condition). (From Ferrua, M.J., Kong, F., and Singh, R.P., Trends Food Sci. Technol., 22, 480, 2011. With
permission.)
to characterize the kinematics and efficiency of advective mixing during gastric digestion
by analyzing the stretching properties of gastric flows with different rheological properties.
Characterizing the fluid dynamics and mixing conditions that develop during digestion using
computer simulation will help to better understand the underlying mechanisms of digestion and
the design of foods for their optimal performance within the body [54].
23.5 CONCLUSIONS
In this chapter, food texture analysis using numerical simulation techniques including CFD has
been reviewed. The results presented in the case studies show good results for analyzing and pre-
dicting the physical characteristics during the change in food texture. It was shown how the numeri-
cal simulation approach can be applied to the handling and processing of horticultural products,
fracture behavior of cellular structured food, oral processing and digestion of the various phases of
food, and so on. A clear advantage of numerical simulations of food texture is the understanding of
relationships between composition, structure, and texture from production to consumption of food.
Various examples presented in this chapter have demonstrated the importance and potentiality of
analyzing and understanding the structural changes in food.
FIGURE 23.21 Instantaneous velocity vectors of gastric flow at different time t = 0 s (A) and t = 6.4 s (B).
The peristaltic contraction velocity of the wall is 10 mm/s. (From Imai, Y., Kobayashi, I., Ishida, S.,
Ishikawa, T., Buist, M., and Yamaguchi, T., Am. J. Physiol. Gastrointest. Liver Physiol., 304, G536, 2013.
With permission.)
For sensory investigations, estimation of texture attributes using instrumental techniques has
been demanded. The use of numerical simulation techniques is believed to be the method for obtain-
ing less deviation in results between instrumental measurement and sensory evaluation.
REFERENCES
1. Szczesniak A.S., Brandt M.A., Friedman H.H., Development of standard rating scales for mechanical
parameters of texture and correlation between the objective and the sensory methods of texture evalua-
tion, Journal of Food Science, 28(4): 397–403, 1963.
2. Rosenthal A.J., Food Texture: Measurement and Perception, Springer-Verlag, New York, USA, 1–17,
1999.
3. Chen L., Opara U.L., Texture measurement approaches in fresh and processed foods—A review, Food
Research International, 51(2): 823–835, 2013.
4. Szczesniak A.S., Texture is a sensory property, Food Quality and Preference, 13(4): 215–225, 2002.
5. Wilkinson C., Dijksterhuis G.B., Minekus M., From food structure to texture, Trends in Food Science
& Technology, 11(12): 442–450, 2000.
6. Peleg M., Food Texture: Instrumental and Sensory Measurement, CRC Press, Boca Raton, FL, USA,
3–33, 1987.
7. Geneva, Switzerland. International Organization for Standardization. ISO 3103: 1–4, 1980.
8. Szczesniak A.S., Correlating sensory with instrumental texture measurements—An overview of recent
developments, Journal of Texture Studies, 18(1): 1–15, 1987.
9. Blair G.S., Rheology in food research, Advances in Food Research, 8: 1–61, 1958.
10. Ho Q.T., Carmeliet J., Datta A.K., Defraeye T., Delele M.A., Herremans E., Opara L., Ramon H.,
Tijskens E., van der Sman R., Liedekerke P.V., Verboven P., Nicolaï B.M., Multiscale modeling in food
engineering, Journal of Food Engineering, 114(3): 279–291, 2013.
11. Celik H.K., Determination of bruise susceptibility of pears (Ankara variety) to impact load by means of
FEM-based explicit dynamics simulation, Postharvest Biology and Technology, 128: 83–97, 2017.
12. Kabas O., Celik H.K., Ozmerzi A., Akinci I., Drop test simulation of a sample tomato with finite element
method, Journal of the Science of Food and Agriculture, 88(9): 1537–1541, 2008.
13. Martins R.C., Simple finite volumes and finite elements procedures for food quality and safety simula-
tions, Journal of Food Engineering, 73(4): 327–338, 2006.
14. Sun Y., Xu W.L., Simulation of food mastication based on discrete element method, International
Journal of Computer Applications in Technology, 39(1–3): 3–11, 2010.
15. Hedjazi L., Martin C.L., Guessasma S., Della Valle G., Dendievel R., Application of the discrete
element method to crack propagation and crack branching in a vitreous dense biopolymer material,
International Journal of Solids and Structures, 49(13): 1893–1899, 2012.
16. Nishinari K., Rheology, food texture and mastication, Journal of Texture Studies, 35(2): 113–124, 2004.
17. Mahmoodi-Eshkaftaki M., Maleki A., Simulation of a tomato sample using finite element method,
International Journal of Agriculture and Crop Sciences, 5(16): 1739–1744, 2013.
18. Li Z., Li P., Yang H., Liu J., Internal mechanical damage prediction in tomato compression using multi-
scale finite element models, Journal of Food Engineering, 116(3): 639–647, 2013.
19. Li Z., Wang Y., A multiscale finite element model for mechanical response of tomato fruits, Postharvest
Biology and Technology, 121: 19–26, 2016.
20. Yousef S., Farsi H., Kheiralipour K., Drop test of pear fruit: Experimental measurement and finite ele-
ment modeling, Biosystems Engineering, 147: 17–25, 2016.
21. Salarikia A., Ashtiani S.H.M., Golzarian M.R., Mohammadinezhad H., Finite element analysis of the
dynamic behavior of pear under impact loading, Information Processing in Agriculture, 4(1): 64–77, 2017.
22. Lewis R., Yoxall A., Marshall M.B., Canty L.A., Characterising pressure and bruising in apple fruit,
Wear, 264(1): 37–46, 2008.
23. Celik H.K., Rennie A.E., Akinci I., Deformation behaviour simulation of an apple under drop case by
finite element method, Journal of Food Engineering, 104(2): 293–298, 2011.
24. Cherng A.P., Ouyang F., Blot L., Zwiggelaar R., An estimation of firmness for solid ellipsoidal fruits,
Biosystems Engineering, 91(2): 257–259, 2005.
25. Lu R., Srivastava A.K., Ababneh H.A.A., Finite element analysis and experimental evaluation of bioyield
probes for measuring apple fruit firmness, Transactions of the ASABE, 49(1): 123–131, 2006.
26. Ahmadi E., Barikloo H., Kashfi M., Viscoelastic finite element analysis of the dynamic behavior of apple
under impact loading with regard to its different layers, Computers and Electronics in Agriculture, 121:
1–11, 2016.
27. Cardenas-Weber M., Stroshine R.L., Haghighi K., Edan Y., Melon material properties and finite element
analysis of melon compression with application to robot gripping, Transactions of the ASAE, 34(3):
920–929, 1991.
28. Nourain J., Ying Y.B., Wang J.P., Rao X.Q., Yu C.G., Firmness evaluation of melon using its vibration
characteristic and finite element analysis, Journal of Zhejiang University Science B, 6(6): 483–490,
2005.
29. Sadrnia H., Rajabipour A., Jafari A., Javadi A., Mostofi Y., Kafashan J., Dintwa E., De Baerdemaeker J.,
Internal bruising prediction in watermelon compression using nonlinear models, Journal of Food
Engineering, 86(2): 272–280, 2008.
30. Chen H., De Baerdemaeker J., Bellon V., Finite element study of the melon for nondestructive sensing
of firmness, Transactions of the ASAE, 39(3): 1057–1065, 1996.
31. Seyedabadi E., Khojastehpour M., Sadrnia H., Predicting cantaloupe bruising using non-linear finite
element method, International Journal of Food Properties, 18(9): 2015–2025, 2015.
32. Celik H.K., Caglayan N., Rennie A.E., Nonlinear FEM based high–speed shell shattering simulation for
shelled edible agricultural products: Pecan fruit shattering, Journal of Food Process Engineering, 40(5):
e12520, 2017.
33. Celik H.K., Non-linear FEM based compression simulation of pecan fruit kernels, Akademik Gıda,
14(1): 1–7, 2016.
34. Hernández L.F., Bellés P.M., A 3-D finite element analysis of the sunfower (Helianthus annuus L.) fruit.
Biomechanical approach for the improvement of its hullability, Journal of Food Engineering, 78(3):
861–869, 2007.
35. Petrů M., Novák O., Herák D., Simanjuntak S., Finite element method model of the mechanical behav-
iour of Jatropha curcas L. seed under compression loading, Biosystems Engineering, 111(4): 412–421,
2012.
36. Khodabakhshian R., Emadi B., Development of a finite element method model to determine mechanical
behavior of pumpkin seed, International Journal of Food Properties, 18(2): 231–240, 2015.
37. Mohammed I.K., Charalambides M.N., Williams J.G., Rasburn J., Modeling the microstructural evolu-
tion and fracture of a brittle confectionery wafer in compression, Innovative Food Science & Emerging
Technologies, 24: 48–60, 2014.
38. Babin P., Della Valle G., Dendievel R., Lassoued N., Salvo L., Mechanical properties of bread crumbs
from tomography based finite element simulations, Journal of Materials Science, 40(22): 5867–5873,
2005.
39. Mamlouk H., Guessasma S., Finite element simulation of the compression behaviour of airy breakfast
cereals, Innovative Food Science & Emerging Technologies, 19: 190–203, 2013.
40. Guessasma S., Hedjazi L., On the fragmentation of airy cereal products exhibiting a cellular structure:
Mechanical characterisation and 3D finite element computation, Food Research International, 49(1):
242–252, 2012.
41. Guessasma S., Babin P., Della Valle G., Dendievel R., Relating cellular structure of open solid food
foams to their Young’s modulus: Finite element calculation, International Journal of Solids and
Structures, 45(10): 2881–2896, 2008.
42. Hedjazi L., Martin C.L., Guessasma S., Della Valle G., Dendievel R., Experimental investigation and
discrete simulation of fragmentation in expanded breakfast cereals, Food Research International, 55:
28–36, 2014.
43. Hanley K.J., O’Sullivan C., Byrne E.P., Cronin K., Discrete element modeling of the quasi-static uni-
axial compression of individual infant formula agglomerates, Particuology, 10(5): 523–531, 2012.
44. Hedjazi L., Guessasma S., Della Valle G., Benseddiq N., How cracks propagate in a vitreous dense bio-
polymer material, Engineering Fracture Mechanics, 78(6): 1328–1340, 2011.
45. Hedjazi L., Guessasma S., Della Valle G., Benseddiq N., Finite element modeling of crack propagation
in carbohydrate extruded starch with open void structure, Carbohydrate Polymers, 83(4): 1696–1706,
2011.
46. Chen J., Food oral processing—A review, Food Hydrocolloids, 23(1): 1–25, 2009.
47. Kikuchi T., Michiwaki Y., Koshizuka S., Kamiya T., Toyama Y., Numerical simulation of interaction
between organs and food bolus during swallowing and aspiration, Computers in Biology and Medicine,
80: 114–123, 2017.
48. Mizunuma H., Sonomura M., Shimokasa K., Ogoshi H., Nakamura S., Tayama N., Numerical modeling
and simulation on the swallowing of jelly, Journal of Texture Studies, 40(4): 406–426, 2009.
49. Rauh C., Singh J., Nagel M., Delgado A., Objective analysis and prediction of texture perception of
yoghurt by hybrid neuro-numerical methods, International Dairy Journal, 26(1): 2–14, 2012.
50. Ferrua M.J., Kong F., Singh R.P., Computational modeling of gastric digestion and the role of food mate-
rial properties, Trends in Food Science & Technology, 22(9): 480–491, 2011.
51. Ferrua M.J., Singh R.P., Modeling the fluid dynamics in a human stomach to gain insight of food diges-
tion, Journal of Food Science, 75(7): R151–R162, 2010.
52. Imai Y., Kobayashi I., Ishida S., Ishikawa T., Buist M., Yamaguchi T., Antral recirculation in the stom-
ach during gastric mixing, American Journal of Physiology-Gastrointestinal and Liver Physiology,
304(5): G536–G542, 2013.
53. Ferrua M.J., Xue Z., Singh R.P., On the kinematics and efficiency of advective mixing during gastric
digestion—A numerical analysis, Journal of Biomechanics, 47(15): 3664–3673, 2014.
54. Ferrua M.J., Singh R.P., Computational modeling of gastric digestion: Current challenges and future
directions, Current Opinion in Food Science, 4: 116–123, 2015.
24 CFD Study of Top-Spray
Fluidized Bed Coating Process
Wasan Duangkhamchan, Frederik Ronsse, and Jan G. Pieters
CONTENTS
24.1 Introduction.....................................................................................................................................532
24.2 Fluidized Bed Coating Principles and Its Application in Food........................................... 532
24.2.1 Fluidized Bed Coating Process............................................................................... 532
24.2.2 Mechanisms in a Top-Spray Fluidized Bed Coating.............................................. 533
24.2.2.1 Fluidization..................................................................................................533
24.2.2.2 Droplet Formation and Atomization.............................................................533
24.2.2.3 Heat Transfer and Layered Growth Mechanism.........................................533
24.3 Multiphase Flow Modeling Approaches.............................................................................. 533
24.3.1 Eulerian–Lagrangian Approach.................................................................................... 534
24.3.2 Eulerian–Eulerian Approach..............................................................................................534
24.4 CFD Descriptions for Mechanisms in a Top-Spray Fluidized Bed Coating Process.......... 536
24.4.1 Airflow Patterns...................................................................................................... 537
24.4.1.1 Governing Equations............................................................................... 537
24.4.1.2 Computational Grid Generation and Boundary Conditions................... 537
24.4.1.3 Numerical Setup...................................................................................... 538
24.4.2 Particles Motion...................................................................................................... 539
24.4.2.1 Governing Equations of Mass and Momentum Conservation................ 539
24.4.2.2 Interphase Momentum Exchange Coefficients........................................... 540
24.4.2.3 Suitable Models of Gas–Solid Momentum Exchange Coefficient
Model for Modeling Solid Hydrodynamics in a Top-Spray
Fluidized Bed Coating Reactor........................................................................542
24.4.3 Droplets Atomization and Dispersion in a Gas–Solid Fluidized Bed.................... 547
24.4.3.1 CFD Model of Droplet Atomization............................................................ 547
24.4.3.2 Experimental Spray Visualization...................................................................549
24.4.3.3 Spray Patterns.......................................................................................... 550
24.4.3.4 Effect of the Presence of Liquid Spray on Gas–Solid Fluidized Bed
Behavior........................................................................................................... 550
24.4.4 Heat and Mass Transfer and Layer-Growth Mechanism........................................ 551
24.4.4.1 Heat and Mass Transfer Models of the Gas–Solid System.......................551
24.4.4.2 Heat Transfer Models for a Droplet Phase.................................................552
24.4.4.3 Gas Temperature and Humidity Distribution as Affected
by the Inclusion of Droplet Evaporation.................................................. 554
24.4.4.4 Layered Growth Mechanism................................................................... 555
24.5 Conclusions.......................................................................................................................... 557
Nomenclature.................................................................................................................................. 557
References....................................................................................................................................... 558
531
24.1 INTRODUCTION
Fluidized bed coating is one of the potential encapsulation methods in which a particulate solid
material is entrapped by spraying a coating solution directly into fluidized particles. With the help
of hot fluidizing air, coating droplets deposited onto the solid surface are evaporated resulting in
a solidified shell wall. Both physical and chemical properties of the particulate solid material are
therefore modified with the purposes of appearance modification, nutritional value addition, func-
tionality improvement, and shelf life extension [1]. The fluidized bed coating process has been
successfully applied in foods due to its versatility and simplicity with low cost [2], as seen in many
works [1,3–7] and summarized in [8].
Among different configurations, the top-spray fluidized bed is appropriate to not only agglom-
erate food ingredients as a method for improving the water dispersibility and flowability [6], but
also to coat food particulate solids with the purpose of adding nutrition and improving function-
ality [3–5,7]. However, the unwanted side effects of premature spray drying of coating materials
and agglomeration still challenge food technologists. Consequently, the complex thermodynamic
and hydrodynamic interactions between three phases (droplet, solid particle and gas) need to be
explored to serve as a tool for further process design and optimization.
The top-spray fluidized bed coating process involves a number of operating variables, e.g., fluid-
izing air flow rate, air temperature, spraying time and rate, etc. [9]. Extensive trial-and-error testing
with time-consuming experiments is therefore unavoidable. Computational Fluid Dynamics (CFD)
has been successfully introduced as an alternative to provide insights into multiphase flow and
related interactions in this process. Consequently, this chapter presents all mechanisms and critical
aspects in the top-spray fluidized bed coating process via the CFD framework to serve as a numeri-
cal tool for understanding and clarifying the impact of operating parameters on coating efficiency.
The outline of this chapter is as follows. In Section 24.2, the overview of fluidized bed coating
and its principles are presented. Section 24.3 covers the multiphase flow modeling approaches.
Furthermore, CFD descriptions of mechanisms taking place in a top-spray fluidized bed coating
process are presented in Section 24.4. Finally, conclusions are proposed in Section 24.5.
24.2 FLUIDIZED BED COATING PRINCIPLES AND ITS APPLICATION IN FOOD

24.2.1 Fluidized Bed Coating Process
Fluidization has been widely used in many food applications due to its high heat and mass transfer
performance and uniformity. In the so-called fluidized bed coating or fluidized bed granulating
processes, fluidization is combined with a spraying process in order to apply coating materials to
particles and/or to agglomerate particles with the goal of modifying the particles’ physical and/or
chemical properties. In these processes, fluidizing air is used not only to lift or suspend the par-
ticles, but also used evaporate the solvent in the coating or granulation liquid resulting in layered
growth (in coating) or bridge formation (in agglomeration) on the particle surface. Purposes in
the food industry of these processes include reduction of fine particles formation, improving the
handling properties and homogeneity, addition of color or flavor, masking undesired off flavors and
protection against environmental conditions (e.g., high humidity) [10].
In industrial applications, different configurations of fluidized bed coating processes are used,
i.e., top-spray, bottom-spray or Wurster and tangential spray [11]. In top-spray fluidized bed coat-
ing, a nozzle is positioned above the fluidized bed with the spray directed downward counter
current to the particle flow. In contrast, the nozzle is situated at the air distributor with the spray
directed upward concurrent with the particle flow in a bottom-spray fluidized bed coating reactor. A
tangential-spray fluidized bed coater provides tangential spray resulting from positioning a nozzle
at the side of the reactor. Even though the top-spray fluidized bed is particularly applied with the
goal of granulating, it is often used to coat granular materials due to its versatility and simplicity
with low cost [2].
CFD Study of Top-Spray Fluidized Bed Coating Process 533
24.2.2 Mechanisms in a Top-Spray Fluidized Bed Coating

24.2.2.1 Fluidization
Fluidization is a process in which air is drawn through a bed of particles with sufficient velocity to
lift or suspend the particles. The fluidized particles move freely, behaving like a fluid. Different flu-
idization regimes occur depending on the gas flow rate: fixed bed, minimum fluidization, bubbling
fluidization, slugging, turbulent fluidization and pneumatic transport. Particles remain fluidized
with the gas velocity ranging from the minimum fluidization velocity at which the drag force upon
the particle bed is equal to its weight and the terminal velocity at which particles are pneumatically
transported out of the reactor [12]. Required fluidizing velocity varies with particle size, shape and
density. As a result of fluidization playing a critical role in achieving a stable and successful coat-
ing process, many attempts have been made to investigate particle hydrodynamics as affected by
fluidization [13–17].
24.2.2.2 Droplet Formation and Atomization

Two-fluid nozzles normally positioned above the bed in the top-spray fluidized bed coating are
typically used to spray coating materials into the bed. With the help of compressed air resulting
in high shear forces at the liquid-gas interface, the coating liquid pumped through the nozzle tip is
atomized as it emerges [18]. Droplet size and distribution vary depending on spray rate, viscosity,
nozzle design and atomizing air pressure. For the coating purpose, droplet size is a critical param-
eter influencing layered growth and subsequent coating efficiency. Therefore, spray characteristics
and its flow pattern have been investigated in many works, as referred to in [19–22].
24.2.2.3 Heat Transfer and Layered Growth Mechanism

Besides fluidizing particles, the inlet process air with temperature ranging from near ambient to
almost water boiling point evaporates volatile components of the coating liquid (i.e., solvent). The
particles that circulate between the bottom and the top of the bed are exposed to hot air entering
in the region just above the air distributor, the so-called “heat transfer zone.” After droplet/particle
impact in the “spraying zone” (i.e., region close to the spraying nozzle), the wetted particle is dried
[23]. The uniformly layered growth takes place after several passages of fluidized particles moving
from and towards the spraying zone where wetting and drying occur almost simultaneously [12].
However, optimal conditions should be of concern in order to avoid undesired mechanisms, includ-
ing agglomeration in which liquid bridges are formed and solidified and spray drying of the droplets
containing the coating liquid before particle contact.
24.3 MULTIPHASE FLOW MODELING APPROACHES

In a multiphase flow, more than one phase, i.e., gas, liquid or solid, flows simultaneously in the same
process. Occurring in many industrial processes (for instance, riser reactors, dryers, bubble column
reactors, scrubbers, etc.) the multiphase flow also takes place in a top-spray fluidized bed coating
process consisting of gas, solid and liquid (droplet) phases. In addition to simultaneous flow of mate-
rials within different phases, the ones with the same state but different properties are also taken into
consideration in the multiphase flow. For example, solid particles with different sizes can be treated
as different phases because each size class of particles particularly responds to and interacts with
the flow field in which it is immersed [24]. The phases considered in the multiphase flow systems are
divided into two major phases including the primary and secondary phase. The continuous phase
is normally considered the primary phase, while the other phase(s) dispersed within the former is
considered the secondary phase.
To provide the basis of or further insight into the multiphase flow dynamics, two numerical
approaches are currently possible: the Eulerian–Eulerian approach and the Eulerian–Lagrangian
approach. In the former, both the continuous fluid phase and the disperse phase are considered as
a continuum, and their appropriate continuum equations are solved. In the Eulerian–Lagrangian
approach, the movement of the disperse phase as a result of forces acting on the particles are calcu-
lated individually, while the continuum equations of the continuous fluid phase are solved.
24.3.1 Eulerian–Lagrangian Approach
The Lagrangian treatment of the dispersed phase combined with the Eulerian approach for the fluid
is also referred to as the discrete phase model (DPM). In this approach, the gas phase is modeled
based on the Eulerian frame of reference in which the mass and momentum conservation equations
for this phase are Reynolds averaged. In the Lagrangian reference frame, a number of particles in
the dispersed phase and their trajectories are tracked individually once the flow field of continu-
ous phase has been calculated. Due to a low volume fraction occupied by the dispersed phase, this
approach is appropriate for dilute systems such as in spray dryers, coal and liquid combustion and
some particle-laden flows [24].
For the dispersed phase, the model equations are derived from the Newton’s second law of
motion, which can be written in the Lagrange reference frame as:
 
du p   g (ρ p − ρ) 
= FD (u − u p ) + + F (24.1)
dt ρp

In Equation (24.1), the term on the left of the equation is the particle acceleration, while the first,
second and last terms on the right are drag force, gravitational force and other forces acting on the
particle, respectively.
The drag force, FD is calculated by
18 µ C D Re
FD = (24.2)
ρ p d p2 24

 
where CD, u, u p, μ, ρ, ρp and dp denote the drag coefficient, the fluid phase velocity, the particle
velocity, the molecular viscosity of the fluid, the fluid density, the density of the particle and the
particle diameter, respectively. Re is the relative Reynolds number, which is defined as [24]
 
ρd p u p − u
Re = (24.3)
µ
The simpler DPM can be made with exclusion of particle-particle interaction, which is described
throughout this chapter. However, interaction between particle phases can be included in the DPM
via a discrete element model (DPM), which will be briefly discussed in the last section.
24.3.2 Eulerian–Eulerian Approach
In contrast to the Eulerian–Lagrangian or DPM approach, a Eulerian treatment is used for each
phase. This is a so-called “multifluid model” (MFM). In the Eulerian–Eulerian approach, the dis-
persed phase is also treated as a continuum, interpenetrating with the continuous fluid phase [23].
Because the volume of a phase cannot be occupied by other phases, the volume fraction of each
phase has to be introduced with the constraint of its summation for all phases being unified [25].
Calculations of mass, momentum and energy conservation are therefore based on the volume frac-
tion of the dispersed phase. In this approach, due to the impossibility of resolving every point in
time or space, time-averaged Navier–Stokes equations for all phases are solved with single pressure
shared for each phase. The conservation equations for each phase have similar structure, as shown
in the following equations:
Conservation equations for the gas phase:

Continuity equation
∂ ∂
(α g ρg ) + ⋅ (α g ρg u) = 0 (24.4)
∂t ∂y
Momentum equations
∂ ∂ ∂  2  ∂
(α g ρg u) + ⋅ (α g ρg uu) = −  P + ρg k  δ ij − ⋅ (α g sg )
∂t ∂y ∂y  3  ∂y (24.5)
+ α g ρg g − β gs (u − v)

  ∂u   ∂u  T  
2   ∂u 
(
where sg = sgm + sijt = −  ξg − µ g    + µ g + µ gt
 3   ∂y 
)   +     (24.6)
 ∂y   ∂y   
Turbulent kinetic energy equation
∂  µ + µ g ∂k 
t
∂ ∂
(α g ρg k ) + ⋅ (α g ρg uk ) = − ⋅  α g g 
∂t ∂y ∂y  σ k ∂y 
 (24.7)
 ∂u   ∂u  T   ∂u  
− α g µ gt   +    :    − α g ρgε
  ∂y   ∂y    ∂y  

Turbulent dissipation equation
∂  µ + µ g ∂ε 
t
∂ ∂
(α g ρgε ) + ⋅ (α g ρg uε ) = − ⋅  α g g 
∂t ∂y ∂y  σ ε ∂y 
  T  (24.8)
ε α g µ gt  ∂u  +  ∂u   :  ∂u   − C2εα g ρg ε
2
−C1ε      
k   ∂y   ∂y    ∂y   k

However, the set of equations seen above has to be closed using constitutive equations, which
are based on the kinetic theory of granular flow [26]. These additional constitutive relations are
presented as follows:
Dissipation of granular temperature by inelastic particle collision [27]
1  4  θ  1/ 2 
2
γ = −ℵc  mC  = 3(e − 1)α s ρs g0θ
2 2
   − ∇ ⋅ v  (24.9)
2   d p  π  

Radial distribution function
 1 + 2.5α s + 4.5904α s2 + 4.515439α s3 

g0 = 1 + 4α s   (24.10)
   α 3 
 1 − s  
 
  α s ,max  
 
Solid phase pressure
Ps = 1 + 2(1 + e)α s g0  α s ρsθ (24.11)

Bulk viscosity solid phase
4 θ
ξs = α s ρs d p g0 (1 + e) (24.12)
3 π
Shear viscosity solid phase
  8  (1 + e  8 
µsk  1 +   α s g0   1 + α s g0 
  5  2   5 
µs = µsc + (24.13)
α s g0
4 θ
µsc = α s ρs d p g0 (1 + e) (24.14)
5 π
5 m θ
µsk = 1.016 (24.15)
16 d P2 π

Turbulent viscosity of gas phase
k2
µ gt = Cµ ρg (24.16)
ε
24.4 CFD DESCRIPTIONS FOR MECHANISMS IN A TOP-SPRAY

FLUIDIZED BED COATING PROCESS
Even though top-spray fluidized bed coating has been widely used in the food industry due to its
simplicity and versatility, many problems, such as the premature spray-drying of the droplets and
agglomeration, are still encountered [24,28]. To solve these problems, all mechanisms taking place
in the top-spray fluidized bed coating system have to be clearly understood. In the system, the fol-
lowing phenomena have to be accounted for: hydrodynamics behavior of fluidizing air, fluidized
particles and coating droplets, heat and mass transfer, and the resulting layer growth mechanism.
Therefore, their descriptions are explored via a CFD framework in this chapter.
24.4.1 Airflow Patterns
Fluidization has been successfully introduced to a coating or granulating application due to its
high performance of mixing and high heat and mass transfer rates [29]. The key factors influencing
this process are airflow and its distribution. Assessment of the fluidizing airflow pattern inside the
equipment is therefore essential to describe the fluidization hydrodynamics [30]. Consequently, in
this section, the design of air distributors and the resulting airflow patterns in the top-spray fluidized
bed coating system are described by means of a CFD model.
For an incompressible fluid (air), the conservative mass (Equation 24.4) and momentum equations
(Equation 24.5) are numerically solved. The additional equation that needs to be solved is the
turbulence scale of which variables are obtained by decomposing into their mean and fluctuating
components, resulting in the Reynolds Averaged Navier–Stokes (RANS) equations that contain
additional unknowns. Many turbulence models are developed to determine these unknowns [30].
Among them, the realizable k-ε model has been widely used in fluid bed applications due to its
consistency with turbulent flow characteristics. In addition, the presence of walls inside the fluid
bed reactor significantly affects turbulent flows, resulting in high intensity turbulence effects. This
viscosity-affected region can be resolved by means of near-wall treatment or wall functions.
24.4.1.2 Computational Grid Generation and Boundary Conditions

24.4.1.2.1 Computational Grid Generation
Based on the finite volume approach, the computational domain is divided into discrete volumes by
means of grid generation (the so-called control volumes). Although Peirano et al. [31] and Depypere
et al. [30] recommended that the reactor’s entire air supply system should be included in the simu-
lation, no significant difference is observed in the model-predicted air flow patterns in the reator
when only the air plenum was modeled. Therefore, the grid of the tapered reactor and inlet air ple-
num, corresponding to a laboratory-scale GPCG-1 fluidized bed coater, is used in order to reduce
the computational expense. Mesh refinement associated with wall functions or near-wall modeling
techniques are applied in the regions in which high turbulence intensity may occur or at near-wall
locations. Prior to solving of the model’s equations, the mesh quality should be evaluated using a
normalized measure of the element skewness in order to reach solution convergence and accuracy.
A mesh consisting of control volumes that are too skewed will impair the solving algorithm in find-
ing a proper solution. The meshed domain specified as an air zone is subsequently subjected to the
CFD solver.
24.4.1.2.2 Boundary Conditions and Simulation Parameters

The boundary conditions and the properties of air and solid material are specified. At the inlet a
pressure-inlet type boundary condition is applied using the Bernoulli equations, while a pressure-
outlet is used for modeling the air outlet. In order to rapidly converge, radial pressure distribution
as well as realizable values of backflow quantities are employed. In addition, the walls are normally
modeled as no-slip stationary boundaries.
24.4.1.2.3 Modeling the Air Distributor

An internal boundary is normally used to model the air distributor, which is the key factor influenc-
ing the fluidization hydrodynamics. Between two porous media models, one with porous jump and
one with porous zone boundary conditions, the latter is recommended when homogeneous airflow
is concerned [13,14,19,23,30]. Therefore, the focus of modeling the air distributor only lies on the
porous zone boundary in this chapter.
TABLE 24.1
Impermeability (Dz) and the Inertial Resistance (Cz) for Two Different
Commercial Distributors (Termed Rob172 and Rob280) Modeled as a
Porous Zone
Distributor Porous Zone Thickness (m) Dz (m−2) Cz (m−1)
Rob172 0.0020 1.19e+09 11,214
0.0035 6.80e+08 6,408
0.0070 3.40e+08 3,204
Rob280 0.0020 2.64e+09 8,263
0.0035 1.51e+09 4,722
0.0070 7.54e+08 2,361
Source: Depypere, F., Pieters, J.G., and Dewettinck, K. Powder Technol. 145, 183, 2004.
With permission from Elsevier.
In the porous zone, which is defined by flow resistance, the momentum sink term is added in the
standard Navier–Stokes equations in order to model the pressure gradient in the porous cell. The
additional term originated from the orifice equation is expressed as follows.
 3 3 
Si = − 
 ∑ D µu + ∑C
j =1
ij j
j =1
ij
1
2
ρ u j u j  (24.17)


In Equation (24.17), Si is the momentum sink term with matrix D and C representing the imper-
meability and the inertial resistance, respectively. The first term on the right of this equation repre-
sents the viscous loss term, while inertial loss is described by the second term.
The impermeability and the inertial resistance in the main flow direction (z-direction) are only
taken into account to provide homogeneity of airflow upstream of the distributor. The distributor
characteristics (Dz and Cz) are shown in Table 24.1. These values can be adjusted corresponding
to a number of sequential identical distributors, for instance, multiplied by two when using two
stacked on top of one another. In CFD simulations with thickness variation, the higher porous zone
thickness is recommended if significant differences in simulated airflow patterns are not observed.
With a smaller size of porous zone thickness in the model, the tiny computational elements lead to
difficulties in obtaining convergence of the continuity equation.
The resistance ratio, RR, is therefore defined as:
Dx D y C x C y
RR = = = = (24.18)
Dz Dz Cz Cz

The suitable RR value is 10, at which the differences in resulting air mass flow rate and corre-
sponding distributor pressure drop are negligible [30].
24.4.1.3 Numerical Setup
A single-precision segregated implicit solver is used to simulate the airflow patterns inside a top-
spray fluidized bed reactor under steady-state circumstances. The continuity equation is solved
using a standard discretization scheme. Turbulent flow behavior is modeled using the realizable
k-ε turbulence model, in which both the turbulence kinetic energy and the turbulence dissipation
rate equation are solved using a first-order upwind scheme. The numerical diffusion in momentum
and energy equations is reduced by means of a second-order upwind scheme. The velocity-pressure
relation is determined using the Semi-Implicit Method for Pressure Linked Equations (SIMPLE)
algorithm. The under-relaxation factors can be adjusted with respect to convergence behavior. All
governing equations are iteratively solved until reaching convergence criteria, at which the scaled
residuals drop to 10 −6 for the energy equation and 10 −3 for the other conservation equations.
24.4.2 Particles Motion
In this section, fluidized particle hydrodynamics as affected by the fluidizing airflow are modeled
by means of a Eulerian–Eulerian approach (Multifluid model: MFM) in which all phases, including
gas and solids, penetrate each other and are incorporated with the concept of phase volume fraction.
In this approach, all conservative equations are solved for each phase, resulting in a set of equations
for each phase.
24.4.2.1 Governing Equations of Mass and Momentum Conservation

The conservative mass equation for phase q, which represents either gas or solid phases, can be
written as:
∂ 
(α q ρq ) + ∇ ⋅ (α q ρq vq ) = 0 (24.19)
∂t
where ρq is the density and vq the velocity of phase q. αq is the volume fraction for each phase, which
is constrained by the summation expressed in the following equation, with Nq being the number of
phases in the model.
Nq
∑α
q =1
q = 1 (24.20)

The conservative momentum equations for gas phase and solid phase are given by equations
(24.21) and (24.22), respectively.
Ns
∑K v
∂    
(α l ρl vl ) + ∇ ⋅ (α l ρl vs vs ) = −α l ∇ ⋅ p + ∇ ⋅ τ l + α l ρl g − ls ls (24.21)
∂t
s= I
∂    
(α s ρs vs ) + ∇ ⋅ (α s ρs vs vs ) = −α s∇ ⋅ p − ∇ p + ∇ ⋅ τ s + α s ρs g + K ls vlss
∂t
Ns
(24.22)
∑

+ K ms vms
m =1,m ≠ s

For conservation of momentum, the equations have a similar structure with different subscript
denoting a different phase: l for gas phase and s for solid phase. However, the solid pressure term,
–∇ps, is found only to the right side of Equation (24.22). In both Equations (24.21) and (24.22), vls is
the slip velocity between gas and solid phase. Kls denotes the coefficient of drag force (the momen-
tum exchange coefficient) relevant to both phases, while Kms represents the solid–solid momentum
exchange coefficient between solid phase s and m (1 ≤ s ≤ Ns, m ≠ s, only applicable when using
more than one solid phase). p is the pressure and τ l the deviatoric effective stress tensor of the fluid
phase. Ns represents the number of solid phases.
24.4.2.2 Interphase Momentum Exchange Coefficients

24.4.2.2.1 Gas–Solid Momentum Exchange Coefficients
Many attempts [13,32–35] have been made to describe the drag force acting on solid particles in
a term of momentum exchange, which is based on the drag coefficient Kls. The momentum trans-
fer between the gas phase and the solid phase is proportionally related to the pressure gradient.
Therefore, most drag coefficient models are derived from this correlation. The gas–solid momen-
tum exchange coefficient models widely used in solid-gas models are described below.
Ergun drag model [36]:

This drag model is obtained by combining the gas momentum balance; pressure gradient
equals gas–solid momentum interaction, and the equation of pressure drop across a dense
bed of spherical particles (αs ≥ 0.8):
α s2 µl ρα  
K ls = 150 + 1.75 l s vs − vl (24.23)
α ls d s
2 2
α l ds

Wen–Yu drag model [37]:

In case of the dilute solid-gas system with αs < 0.8, the momentum exchange coefficient can
be calculated as:
3 α sα l ρl  Re   
K ls = 2
C D  s  vs − vl (24.24)
4 vr ,s d s  vr ,s 

Gidaspow drag model [38]:

In practice, both dilute and dense regimes simultaneously occur in a gas–solid fluid bed system.
Therefore, the Ergun drag model (Equation 24.23) and the Wen–Yu drag model (Equation
24.24) are used for dense phase calculation and dilute phase calculation, respectively.
Even though the Gidaspow drag model includes all models suitable for both gas–solid regimes,
a step change of drag coefficient at the volume fraction of 0.2 [39] is still observed, as
Figure 24.1 shows the drag coefficient as a function of solid volume fraction, calculated from
the Ergun drag model (Equation 24.23) and the Wen–Yu drag model (Equation 24.24). It
can be seen from this figure that the coefficient value changes around a value of the solid
volume fraction of 0.2 in a stepwise manner, which could result in convergence difficulties.
To avoid this discontinuity increasing with higher Res at a solid volume fraction of 0.2,
a weighted average of the two drag models is introduced [39]:
K ls = (1 − ϕ ls ) K ls , Ergun + ϕ ls K ls ,Wen −Yu (24.25)

where φls is the transition function, which is expressed by:
tan −1 150 × 1.75(0.2 − α s ) 

ϕ ls = + 0.5 (24.26)
π
This drag model containing the smoothed transition function is the so-called “modified
Gidapow drag model” and is used throughout this chapter.
FIGURE 24.1 Comparison of drag coefficients as a function of solid volume fraction (αs). (From Huilin, L.,
and Gidaspow D. Chem. Eng. Sci., 58, 3781, 2003. With permission from Elsevier.)
Symlal–O’Brien drag model [40]:

This drag model is based on the measured terminal velocity of particles in a solid-gas
fluidized bed, which is written as:
3 α s ρl  
K ls = C D vs − vl (24.27)
4 vr2,s d s

 4.8 
where C D =  0.63 +  and (24.28)
 Re s 
 vr , s 

Res is the relative Reynolds number for solid phase, expressed by
 
ρl d s vs − vl
Re s = (24.29)
µl
The terminal velocity vr,s found in Equation (6) is determined by the following correlation.

( )
vr ,s = 0.5 A − 0.06 Re s + (0.06 Re s )2 + 0.12 Re s (2 B − A) + A2 (24.30)
with
A = α l4.14 (24.31)

and
B = 0.8α l1.25 ⋅ (for α l ≤ 0.85) and B = α l2.65 ⋅ (for α l > 0.85) (24.32)

Arastoopour drag model [41]:
This model is obtained by modifying a pressure drop correlation and is available for a
wide range of solid volume fractions.
 17.3 ρ  
K ls =  + 0.336  l vs − vl α sα l−2.8 (24.33)
 Re s  ds
24.4.2.2.2 Solid–Solid Momentum Exchange Coefficients

The solid–solid momentum exchange coefficients are obtained by solid collisions, which can be
expressed by the following equation.
Symlal solid–solid drag model [42]:
π π2 
3(1 + esm )  + C fr ,sm  α s ρsα m ρm (d m + d s )2 g0 ,sm
2 8   
K sm = vm − vs (24.34)

( 3
)
2π ρm d m + ρs d s3
where
esm = the coefficient of restitution
Cfr,sm = the coefficient of friction between the mth and sth solid phase
dm = the diameter of the solid phase m
g0,sm = the radial distribution coefficient
24.4.2.3 Suitable Models of Gas–Solid Momentum Exchange Coefficient Model for

Modeling Solid Hydrodynamics in a Top-Spray Fluidized Bed Coating Reactor
A CFD model associated with a Eulerian–Eulerian approach is used to simulate gas–solid flow in
a tapered fluidized bed reactor in the absence of additionally introduced atomized air. Due to the
existence of various drag force models, the most suitable one should be selected for continuing the
study of the hydrodynamics behavior of the fluidized bed.
24.4.2.3.1 Grid Generation
The geometry of the top-spray fluidized bed reactor is defined corresponding to a laboratory-scale
Glatt GPCG-1 fluidized bed coater (Glatt GmbH, Germany) that was used for experimental valida-
tion. The numerical grid consisted of hybrid hexahedral-tetrahedral elements and with near-wall
treatment, this grid was composed of 108,721 elements (control volumes). The geometry and the
numerical grid are presented in Figure 24.2.
24.4.2.3.2 Numerical Setup
In most numerical studies, the common two-phase Eulerian model is used to predict the hydro-
dynamics behavior of solid particles in fluidized bed applications. This approach accounts for the
monodispersity of the solids (i.e., all particles having a single diameter), which may not be consis-
tent with polydisperse particles used in practice (i.e., particle diameter is distributed over a certain
range). Therefore, in order to account for the solids’ polydispersity, a more-than-one-phase Eulerian
model is recommended.
FIGURE 24.2 The modeled half geometry of a laboratory Glatt GPCG-1 fluidized bed reactor (left) and the
corresponding mesh (right). (From Duangkhamchan, W., Ronsse, F., Depypere, F., Dewettinck, K., Pieters,
J.G. Chem. Eng. Sci., 65, 3102, 2010. With permission from Elsevier.)
A Eulerian model with four phases is selected to simulate the gas–solid flow pattern in the
top-spray fluidized bed coating process. In this approach, gas is defined as the continuous phase,
while the polydisperse solid particles were defined using three disperse solid phases with a different
particle size for each phase. The particle size in each solid phase corresponds to the experimentally
measured size distribution. The size distribution is divided into three classes: the 20th, between
the 20th and the 80th, and above the 80th percentile [13]. The definition of these three solid phases
and the particles’ size distribution are shown in Figure 24.3. From this, the right and the left tail
in the particle size distribution are defined as the larger and smaller sized particles, respectively,
in order to take the effect of polydisperse particles into consideration in the modeled fluidized bed
[13]. However, the computational expense should be of concern when using more than one dispersed
phase, as this multiplies the number of constitutive equations that have to be accounted for in the
model.
FIGURE 24.3 Definition of the particle size classes in each solid phase of glass beads in the 4-phase Eulerian
model. (From Duangkhamchan, W., Ronsse, F., Depypere, F., Dewettinck, K., Pieters, J.G. Chem. Eng. Sci., 65,
In addition to the governing equations for the gas phase, the conservative mass and momentum
equations for the three solid phases are solved by means of the finite volume approach. To simulate
the fluidization behavior as a function of process time, an unsteady-state first order implicit solver
is applied with a single precision.
24.4.2.3.3 Experimental Validation
The solid’s motion inside a laboratory-scale Glatt GPCG-1 fluidized bed (Glatt GmbH, Germany)
was measured by means of the positron emission particle tracking (PEPT) [11]. Movement in three
dimensions of a single trace particle is captured by tracking its location and movement over pro-
longed periods. Therefore, time-averaged profiles of solids volume fraction (αs) and voidage (αl)
were obtained [11,13].
24.4.2.3.4 Particle Movement in the Absence of the Spray Nozzle

Commonly, the effect of polydisperse particles is disregarded in the two-phase modeling, in which
only a single particle diameter is considered due to lower complexity and lower computational
expenses. Poor agreement with the experimental data is therefore found, as can be confirmed in
Figure 24.4, in which the void fraction profiles in the reactor’s YZ plane predicted at steady-state
fluidization conditions are shown [13]. As recommended, the more-than-one-phase model should be
employed in order to account for the polydispersity of the gas–solid system. The improved model-
predicted void fraction profiles, obtained by using the four-phase Eulerian model, are shown in
Figure 24.5.
To select the suitable gas–solid momentum exchange coefficient models for a top-spray fluidized
bed equipment, the quantitative analysis is used to compare the sums of squared residuals (SSRs),
which measure the models’s deviation from the experimentally PEPT-measured solid volume frac-
tion. The SSR values are determined by integrating over a 2D plane the symmetric YZ plane. It is
found that the modified Gidaspow drag model is suitable for predicting the solid’s motion inside the
fluidized bed as it gives the lowest SSR value compared with the PEPT results for two different air
flow rates tested in [11]. As a result of adding the transition function (Equations 24.25 and 24.26) at
the solid volume fraction of 0.2 in this model, the abrupt change in drag can be solved continuously,
leading to ease of numerical convergence. Consequently, the four-phase Eulerian model together
with the modified Gidaspow drag model is subsequently applied in the numerical investigation of
mechanisms taking place in the top-spray fluidized bed coating process.
In the voidage plots shown in Figure 24.5, the fluid bed zone can be divided into two distinctive
zones: a dead zone and a dense zone. The former is located at an annulus region between the bed wall
and the distributor, as shown in Figure 24.6. In this zone, the particles move very slowly or do not visit
frequently, resulting in low time-averaged solid volume fractions. On the other hand, the dense zone
has a high number of particle visits. This zone is situated in the annulus region between the central
axis and the tapered wall [11,13]. The annular dead zone is attributed to the circulating particles’
movement. As observed in a spouted bed and Wurster column, the particles move upward with a high
velocity in the center to the freeboard, and subsequently move radially towards the walls and fall along
the vessel wall, where they are not fluidized. As a result, a central region of high-velocity upward par-
ticle motion and a slow-moving annulus of higher occupancy are distinguished as the dense zone [11].
24.4.2.3.5 Effect of Atomization Air on Particle Movement

As described earlier, particles circulate in the vessel, moving upward in the center and falling down
along the walls. However, this circulating solid flow behavior lends itself to a top-spray fluidized
coating, where rate of deposition of the coating onto the particles is essentially controlled by a more
or less ordered circular motion of the particles in the bed. Therefore, the effect of the release of
atomization air contributing to the spray formation may disrupt the solid’s motion and a study of its
effect on fluidization of solid particles is presented here. However, the coating spray is not included
in this section.
FIGURE 24.4 Model-predicted (2-phase) and experimental bed voidage (dimensionless) contour plots in the
XZ plane (y = 0) for glass beads fluidised at 55 m3 h−1 using from (a) PEPT-experiment [11], (b) Wen–Yu model,
(c) Symlal–O’Brien model, (d) Gidaspow model, (e) modified Gidaspow model and (f) Arastoopour model.
(From Duangkhamchan, W., Ronsse, F., Depypere, F., Dewettinck, K., Pieters, J.G. Chem. Eng. Sci., 65, 3105,
2010. With permission from Elsevier.)
With the inclusion of a binary nozzle into the bed, the hexahedral-tetrahedral grid with mesh refine-
ment near the nozzle tip is generated, leading to a mesh complexity up to 209,955 cells. The four-phase
Eulerian model together with the modified Gidaspow drag model is applied, as it has been demon-
strated in the previous section to provide sufficient accuracy in predicting the time-averaged hydrody-
namics behavior of the fluidized bed. The model-predicted as well as the experimental time-averaged
voidages in the YZ-plane as affected by atomization air pressure are presented in Figure 24.7.
From these figures, the time-averaged voidages change with the presence of atomization air. A
hollow cone appears in the center part of the tapered reactor just below the nozzle, which is located
12 cm above the air distributor. In this hollow region, the solid particles moving upward by the
FIGURE 24.5 Model-predicted (4-phase) and experimental bed voidage (dimensionless) contour plots in
the XZ plane (y = 0) for glass beads fluidised at 55 m3 h−1 from (a) PEPT-experiment [11], (b) Wen–Yu model,
(c) Symlal–O’Brien model, (d) Gidaspow model, (e) modified Gidaspow model and (f) Arastoopour model.
(From Duangkhamchan, W., Ronsse, F., Depypere, F., Dewettinck, K., Pieters, J.G. Chem. Eng. Sci., 65, 3106,
2010. With permission from Elsevier.)
fluidizing air face an obstruction zone in which atomization air is directed downward towards the
bed. Therefore, the solid particles redirect to the upper bed region along the wall of the atomization
air cone, leading to a conical voidage at the reactor’s center. Furthermore, as usually found in the
fluid bed regime (see Figure 24.7), the dead zone is minimized when introducing the atomization
air or even eliminated when using higher pressure.
FIGURE 24.6 Definition of zones and regions in the fluidized bed. (From Duangkhamchan, W., Ronsse,
F., Depypere, F., Dewettinck, K., Pieters, J.G. Chem. Eng. Sci., 65, 3108, 2010. With permission from
Elsevier.)
The hollow cone enlarges with increasing atomization air pressure, resulting in a smaller radial
zone between the nozzle atomization cone and the reactor wall. The additional atomization air
directed downward toward the bed also enhances the circulating solid’s motion. In addition, the
expanded-bed height of the fluid bed is found to increase due to higher superficial air velocity of the
particles moving toward the above bed region.
24.4.3 Droplets Atomization and Dispersion in a Gas–Solid Fluidized Bed

24.4.3.1 CFD Model of Droplet Atomization
Among various complex mechanisms taking place in the top-spray fluidized bed operation, coating
spray formulation also plays important role in coating efficiency. This section describes droplets’
formation and their trajectories inside the coating reactor. To simulate the spray pattern presented
here, a Lagrangian approach is used to solve the Newton’s second law of motion for a discrete phase
(droplets), while the Eulerian method solves the transport equations for the continuous gas and solid
phases in different way. This is the so-called “Eularian–Lagrangian approach.”
In a Lagrange frame of reference, the movement of droplets is computed individually, expressed
by Equation (24.1). In this case, the coupling between the gas phase and liquid phase is included via
the drag force calculated by the drag coefficient correlation proposed by Morsi and Alexzander [43]
with the assumption of spherical droplets.
Due to a similar atomization mechanism, droplet formation of a binary or two-fluid nozzle used
in a top-spray fluidized bed operation is modeled using an air-blast/air-assisted atomizer model,
available in CFD softwares such as ANSYS Fluent®. Droplets are normally formed in three steps:
production of a liquid sheet, breakup of the sheet into ligands and finally disintegration of the
ligands into droplets. The breakup of the liquid sheet into droplets is accelerated by the additionally
supplied compressed air. The details of the air-blast/air-assisted atomizer model is available in the
ANSYS manual [24] and is briefly described here.
FIGURE 24.7 Comparison between PEPT and simulated, time-averaged steady-state voidage in the YZ plane
at different atomization air pressures: (a) PEPT-1.5 bar, (b) simulation-1.5 bar, (c) PEPT-2 bar, (d) simulation-2
bar, (e) PEPT-3 bar, (f) simulation-3 bar, (g) PEPT-4 bar, (h) simulation-4 bar. (From Duangkhamchan, W.,
Ronsse, F., Dewettinck, K., and Pieters, J.G. Powder Technol., 212, 108, 2011. With permission from Elsevier.)
In the air-blast/air-assisted atomizer model, the liquid sheet is directly formed by a nozzle, and
thus the sheet formation equations are excluded in this model. However, the sheet thickness, a nec-
essary parameter due to its relevance in various mechanisms in droplet formation, still needs to be
specified in a model of the air-blast atomizer. Assisted by the additional air stream, the liquid sheet
is accelerated to break up into instable ligaments with diameter calculated by:
2π Clg
d lg = (24.35)
K′
where Clg is the ligament constant and Kʹ is the wave number (m−1) corresponding to the maximum
growth rate, Ω, which is obtained from the maximized dispersion relation based upon the growth of
sinuous waves on the liquid sheet [44].
The instable ligaments subsequently break up into droplets based on Weber’s analysis and thus
droplet diameter, d 0, is expressed as:
d 0 = 1.88d lg (1 + 3Oh)1/ 6 (24.36)

In Equation (24.36), Oh, Ohnesorge number, is correlated with the Reynolds number and the
Weber number. Furthermore, a Rosin–Rammler distribution with a spread parameter of 3.5 and dis-
persion angle of 6° is mostly assumed for the droplet diameter, d 0, calculated from Equation (24.36).
For more details concerning the atomizer model, the reader is referred to [24].
24.4.3.2 Experimental Spray Visualization

The schematic of the spray visualization setup is shown in Figure 24.8. In order to visually capture
the spray pattern, a transparent polycarbonate reactor was used with the same size as the Glatt
GPCG-1 fluidized bed coater. The pneumatic nozzle (Schlick Model 970-S1, Germany), normally
used in the Glatt GPCG-1, is positioned 12 cm above the air distributor. Sodium fluorescein salt, as
a fluorescent dye, is added into water and is transported to the nozzle via a peristaltic pump.
FIGURE 24.8 Spray visualization setup. (From Duangkhamchan, W., Ronsse, F., Depypere, F., Dewettinck,
K., and Pieters, J.G. Chem. Eng. Sci., 68, 560, 2012. With permission from Elsevier.)
The UV illuminator located 0.5 m above the reactor outlet directs a 400 W UV spotlight through
the reactor, as shown in Figure 24.8. This manner of illumination proves the most efficient, as the
polycarbonate reactor shell material has some UV absorbing capacity. The illuminated spray is
captured by a digital camera (Olympus i-Speed 1) at 60 fps (800 × 600 pixel size), and stored in an
uncompressed video format (AVI). The captured images are then post-processed to visualize the
spray cone. Details in data post-processing are reported in [19].
24.4.3.3 Spray Patterns
Figure 24.9 shows a comparison of spray patterns visualized from the UV illumination (above) and
from the DPM model (bottom) at atomization air pressures of 1 bar (left) and 3 bar (right). The model-
predicted spray cone presented in this figure is consistent with the measured one, indicating that the
DPM model associated with the air-blast/air-assisted atomizer model can reasonably be used to simu-
late the droplet formation of a binary nozzle. The spray cone is narrower with increasing atomization
air pressure. In addition, higher pressure results in a slightly asymmetric cone, which could be attrib-
uted to the higher volume of expanding atomization air in combination with the asymmetric construc-
tion of the reactor resulting from the one-sided nozzle support [19], as shown in Figure 24.9.
24.4.3.4 Effect of the Presence of Liquid Spray on Gas–Solid Fluidized Bed Behavior

The existing 4-phase Eulerian CFD model is combined with a Lagrange extension in which individ-
ual droplets defined as a disperse phase (or discrete phase) are tracked throughout the computational
domain. The most probable droplet size is calculated based on a Rosin–Rammler size distributor
used in the air-blast/air-assisted atomizer model.
The simulated time-averaged voidage and droplets tracked at a steady state are presented in
Figure 24.10a and 24.10b, respectively. As observed in the gas–solid flow, the presence of a liquid
spray supported by the atomization air leads to the voidage cone under the nozzle, as shown in
Figure 24.10a. As shown in Figure 24.10b, droplets moving downward along with the atomization
air cone contact the counter-current fluidizing particles at the central part of the reactor. Due to
the exclusion within the model of the phenomena of droplet evaporation and adhesion onto particle
FIGURE 24.9 Comparison of visualized and model-predicted spray cones. (From Duangkhamchan, W.,
Ronsse, F., Depypere, F., Dewettinck, K., and Pieters, J.G. Chem. Eng. Sci., 68, 563, 2012. With permission
from Elsevier.)
FIGURE 24.10 Comparison between (a) contour of gas–solid–liquid DPM model-predicted time-averaged
steady-state voidage, (b) gas–solid–liquid DPM model-predicted droplet tracks at t = 15 s. (From Duangkhamchan,
W., Ronsse, F., Depypere, F., Dewettinck, K., and Pieters, J.G. Chem. Eng. Sci., 68, 564, 2012. With permission
from Elsevier.)
surface, the droplets are tracked continuously until they travel through the reactor exit, which sel-
dom exists in reality. Consequently, this model can be used only to numerically investigate the
effect of liquid spray together with atomization air on fluidization hydrodynamics.
24.4.4 Heat and Mass Transfer and Layer-Growth Mechanism

Due to extreme difficulty in modeling the formation of droplets together with their evaporation and
subsequent layer growth or agglomeration mechanisms in the Eulerian framework, the combined
Eulerian–Lagrangian framework is preferred in this section as it is naturally capable of simulat-
ing complex droplet-related processes, especially in a top-spray fluidized bed coating system. The
Lagrangian approach is therefore used to simulate the formation of droplets (as discrete phase)
and their evaporation, while the continuous gas–solid phase flow behavior as well as heat transfer
between phases are modeled based on the Eulerian framework.
24.4.4.1 Heat and Mass Transfer Models of the Gas–Solid System
In order to simulate heat and mass transfer in the top-spray fluidized bed coating process, the
continuous gas–solid phases couple with the discrete (droplet) phase via the combined Eulerian–
Lagrangian CFD model with the inclusion of the energy conservative equations. Presented here are
not only gas–solid fluidization behavior and spray flow pattern but also droplet evaporation with
subsequent vapor transferred to the gas phase. The energy governing equations for the continuous
gas–solid phases based on the Eulerian framework are written as follows:
for the gas phase:

Ns
∂p
∑Q
∂
(α l ρl hl ) + ∇ ⋅ (α l ρl vl hl ) = −α l l + τ l : ∇vl − ∇ ⋅ ql + ls + Qld + m ld hld (24.37)
∂t ∂t s= I

for the solid phase(s):
Ns
∑
∂  ∂p 
(α s ρs hs ) + ∇ ⋅ (α s ρs vs hs ) = −α s s + τ s : ∇vs − ∇ ⋅ qs + Qsl + Qsm + Qsd (24.38)
∂t ∂t m =1,m = s

where hi is the specific enthalpy of each phase (i = l for gas phase and i = s for solid phase), and q
the heat flux resulting from heat conduction, calculated by:
qi = − ki ∇ ⋅ Ti (24.39)

where ki is a thermal conductivity (J/s mK). The heat conduction, as expressed in Equation (3), is
used with the assumption that thermal conductivity is independent of temperature as a result of the
relative narrow range of temperature in the top-spray fluidized bed coating process.
In addition, Equations (24.37) and (24.38) contain the interphase heat exchange, Qij, with sub-
scripts i and j denoting different phases. Qls describes the heat transferred from the gas phase (l) to
the solid phases (s), while Qld represents the gas-droplets heat transfer term. In the energy conserva-
tive equation for the solid phases, Qsm is the heat transfer term between the solid phases, whereas
Qsd is for solid-droplet heat exchange. However, the two terms in Equation (24.38), Qsm and Qsd, are
assumed to be small compared to the others. Hence, they are omitted in the energy conservative
equation (Equation 24.38).
The interphase heat transfer, Qij, is assumed to depend on temperature difference between
phases, given by:
Qij = hij (Ti − Tj ) (24.40)

where hij is the convective heat transfer coefficient (kJ/s m2K) between the ith and jth phase. The gas-
solid heat transfer coefficient, hsg, is calculated by the dimensionless Nusselt number of the solid
phase, Nus, given by:
6 k gα gα s Nus
hsg = (24.41)
d s2
where ds is the solid diameter and kg the gas thermal conductivity. The Nusselt number used in
Equation (24.41) correlates with the Reynolds number (Res), the Prandlt number (Pr) and voidage,
(αl) [45], expressed by (for αl = 0.35-1.00 and Res ≤ 105):

( )( ) ( )
Nus = 7 − 10α g + 5α g2 1 + 0.7 Re s0.2 Pr1/ 3 + 1.33 − 2.4α g + 1.2α g2 Re 0s .72 Pr1/ 3 (24.42)
where the Prandlt number for the gas phase is calculated by:
c p, g µ g
Pr = (24.43)
kg

24.4.4.2 Heat Transfer Models for a Droplet Phase

Droplet evaporation and subsequent mass transferred to the gas phase are modeled under two
regimes: the inert heating regime and the vaporization regime. Not only are droplet trajectories
computed, but gas-to-droplet heat transfer is also predicted once the droplet passes through each
computational cell [24].
When droplet temperature does not reach the vaporization temperature, Tvap, which defines
the onset of vaporization, and in the case of the volatile fraction in the droplet, f v,0, having been
consumed, the droplet temperature as a function of time, Td (t), is calculated using a simple heat bal-
ance equation in which the radiation at the droplet surface is neglected, as expressed in Equation
(24.44).
dTd
M d c p, d = hdg Ad (Tg − Td ) (24.44)
dt
where
cp,d = specific heat of the liquid (J kg−1K−1)
Ad = surface area of the droplet (m2)
Tg = local gas temperature (K)
hdg = convective heat transfer coefficient at the droplet/gas boundary(Wm−2K−1)
The convective heat transfer coefficient in Equation (24.44) is calculated using the correla-
tion of Ranz and Marshall [46] for forced convection around spherical bodies. The correlation is
expressed by:
hdg d d
Nud = = 2 + 0.6 Re1d/ 2 Pr1/ 3 (0 ≤ Re < 200, 0 ≤ Pr < 250 (24.45)
kg

where Nud, dd, kg, Red and Pr denote the Nusselt number, droplet diameter (m), thermal conductivity
of the gas (Wm−1K−1), the relative Reynolds number and the Prandtl number, respectively.
As the thermal energy changes continuously due to heat exchange with the gas phase, the droplet
temperature continuously increases until reaching the boiling point, T bp, or until the volatile fraction
is completely consumed [24]. Subsequently the vaporization regime is initiated. In this regime the
vaporization rate is low, and the diffusion gradient (difference in vapor concentration at the droplet/
gas boundary) is used to calculate the flux of droplet vapor into the gas phase:
N dg = hdg
′ (Cd − Cg ) (24.46)

where Ndg is the molar flux of vapor diffusing into the gas phase (mol/m2s), and Cd and Cg are the
′ is the convec-
vapor concentration at droplet surface and in the gas phase (mol/m3), respectively. hdg
tive mass transfer at the droplet/gas boundary (m/s).
At the droplet surface, the partial vapor pressure is assumed to be equal to the saturated vapor
pressure, psat, at the droplet temperature, Td. The surface vapor concentration is therefore computed
by:
psat , d
Cd = (24.47)
RTd
while the vapor concentration in gas is computed based on the transport equation of vapor:
p
Cg = X g (24.48)
RTg

where Xg, p and Tg represent the local bulk mole fraction of vapor in the gas, the local absolute pres-
sure, and the local bulk gas temperature, respectively.
′ , found in Equation (24.46) is correlated with the

The convective mass transfer coefficient, hdg
Sherwood number, Sh [46]:
′ dd
hdg
Sh = = 2.0 + 0.6 Re1d/ 2 Sc1/ 3 (0 ≤ Re < 200, 0 ≤ Sc < 250) (24.49)
Dv

where Dv (m2/s) denotes the diffusion coefficient of vapor in the gas phase, and Sc is the Schmidt
number, calculated by:
µg
Sc = (24.50)
ρg Dv

In the vaporization regime, the droplet temperature is updated in accordance with a heat balance
given by:
dTd dM d
M d c p, d = hdg Ad (Tg − Td ) + λ (24.51)
dt dt
dM d
Here λ denotes the latent heat (J kg−1) and is the rate of evaporation (kg s−1).
dt
24.4.4.3 Gas Temperature and Humidity Distribution as Affected

by the Inclusion of Droplet Evaporation
Figure 24.11a and 24.11b show the model-predicted gas temperature and relative humidity. As seen
in Figure 24.11a, low gas temperatures are found near the binary nozzle. This region partly covers
FIGURE 24.11 Time-averaged steady-state contour plots of (a) air temperature (°C) and (b) relative humid-
ity (%) from model prediction in XZ-plane. (From Duangkhamchan W., Ronsse F., Siriamornpun S., and
Pieters J.G. J. Food Eng., 146, 89, 2015. With permission from Elsevier.)
the top bed. The lower gas temperature in comparison with inlet air temperature results from the
interphase heat exchange between the gas and droplet phase, which is undergoing evaporation as
well as the releasing cold atomization air.
Gas temperature increases away from the area, resulting in dense isotherms extending from the
outer part of this region towards the reactor wall. At the bottom part of the bed in which the hot
air enters the reactor, a high temperature gradient is observed. In this region, high gas–solid heat
exchange rates occur—because of the higher gas temperatures in this region, in which the colder
particles coming down from the droplet zone absorb heat from the entering hot air.
The contour plot of the model-predicted relative air humidity (Figure 24.11b) shows the influence
of droplet evaporation. A high humidity region is found under the nozzle in which the gas tempera-
ture is low. The droplet evaporation is also attributed to this high humidity zone.
According to gas temperature and humidity profiles, shown in Figure 24.11a and 24.11b, respec-
tively, three zones are obviously observed based on gas temperature and humidity gradients [47],
consequently called the thermal zones. In the “spray zone,” situated near the nozzle at the top bed,
the sprayed liquid and its evaporation result in low air temperature and high relative humidity. Away
from this zone towards the reactor wall, the “isothermal zone” is defined as the region in which
uniformity of air temperature and humidity is found. The narrow region with high gas temperature
gradients, located at the bottom bed just above the air distributor, is called “heat transfer zone.” This
zone is characterized by high heat transfer rates between entering hot gas and colder circulating
particles—hence rapid change in gas temperature.
24.4.4.4 Layered Growth Mechanism

Coupled discrete element-CFD models (DEM-CFD) or discrete phase models (DPM) have proven
to be a natural way for simulating a multiphase flow and its interactions in the top-spray fluidized
bed coating process [48,49]. However, problems regarding computational requirements due to a
huge number of particles in the production scale [50] still challenge the researcher.
Increasing the particle size, resulting in reduced particle interactions, and advances in comput-
ers’ numerical performance contribute to decreased simulation time. However, these methods are
currently not sufficient for solving the complex CFD models.
The scaling method is an alternative method for reducing the computational load by means of
grouping fine particles and considering them as a single particle with modified drag forces and
other interactions (i.e., with other phases in the system). Many attempts have been made to simulate
solids’ and droplets’ flow behavior and their interactions with the gas flow field in fluidized bed
granulations by means of similarity models [48–50]. The similarity model is briefly described here.
In this model, both solids and droplets are defined as discrete elements flowing through a CFD
mesh cell. Computational load is decreased by reducing total numbers of discrete particles while
maintaining the hydrodynamics of the process. Three dimensionless parameters derived from the
governing equations and used for scaling are described below.
1. The Archimedes number Ar is taken in account for considering the buoyance of particles
in the gas flow field:
gd 3p ( ρ p − ρg )
Ar = (24.52)
vg2 ρg

2. The Reynolds number Re respects the ratio of inertia forces over the viscous force of the fluid:
d pu
Re = (24.53)
vg

3. The Stokes number St describes the drag force of particles:
Fd
St = (24.54)
v g ρ g d pu

The total numbers of particles are scaled using the scaling factor, ks:
N p,2
ks = (24.55)
Np

where Np denotes the number of original particles, while Np,2 represents the reduced par-
ticle number.
For scaling, three modified mass, Ar and Re number equations are applied. Similarity conditions
are Mbed,1 = Mbed,2, Ar1 = Ar2 and Re1 = Re2, comprising the particle diameter dp, the particle density
ρp and the gas viscosity vg, respectively. The conservation of the St number is subsequently used to
confirm the flow equality [50].
According to the mass conservation, the particle diameter dp,2 is scaled, expressed as:
6mbed
d p, 2 = 3 (24.56)
πρ p,2 N p,2

Summation of all single particle masses is defined as the bed mass mbed:
π 3
mbed = N p ρ p d p (24.57)
6
The particle density ρp,2 is scaled based on the conservation of Ar number:
d 3p vg2,2
ρ p, 2 = ( ρ p − ρg ) + ρg (24.58)
d 3p,2 vg2

The conservation of Re number is used to scale the gas viscosity vg,2:
d p, 2
vg ,2 = vg (24.59)
dp

The scaled particle density ρp,2 can be simplified with the use of the scaled vg,2 from Equation
(24.59), yielding:
dp
ρ p, 2 = ( ρ p − ρg ) + ρg (24.60)
d p, 2

Besides the solid particles, droplet size is also scaled with interlinking to the gas viscosity. The
scaled droplet size is therefore calculated by:
vg ,2
d d ,2 = d d ,1 (24.61)
vg ,1

and the droplet density is scaled with the use of the Archimedes number for droplets, Ard , as
follows:
vg2, 2 ρg
ρd ,2 = Ard + ρg (24.62)
d d3,2 g

Borner et al. [50] have successfully used the similarity models to achieve reasonable computa-
tion times to simulate production-scale fluidized bed granulation. Their proposed DEM-CFD model
associated with similarity models gives insight into the spraying and drying zones as well as the
residence times of solid particles and droplets in these zones, which can be used as a basis for under-
standing the layered growth mechanisms and eventually used in process optimization. However, a
disadvantage of this approach should be taken into consideration when the local resolution of par-
ticle distribution is concerned.
24.5 CONCLUSIONS
In this chapter, a multiphase Computational Fluid Dynamics (CFD) model was developed to simu-
late the batch fluidized bed coating process. A modeling methodology was presented that included
a stepwise addition of complexity to the CFD model with experimental validation at each step.
Through modeling, it was concluded that a 4-phase model could provide significant model accuracy
in order to account for the polydispersity of the fluidized solids. Also, in tapered fluidized beds, the
modified Gidaspow drag model has been demonstrated to give the most accurate prediction of solid
volume concentrations or voidage profiles.
When droplet behavior has to be included in the model, then a hybrid multiphase Eulerian–
Lagrangian model, whereby droplets are modeled as discrete entities, proved adequate. Considering
the modeling of heat and mass transfer during fluidized bed coating processes, it has been demon-
strated that the model-predicted result was consistent with the experimental tendencies. However, a
discrepancy between the measured outlet air temperature and that predicted by the CFD model could
be seen. In order to resolve these discrepancies, other mechanisms, including droplet/particle adhesion
and mass transfer occurring at the surface of wetted particles, need to be included in the model as well.
Ultimately, the goal in modeling the fluidized bed coating process is to achieve a complete pro-
cess model in which phenomena such as agglomeration, layered growth and spray drying losses can
be predicted.
NOMENCLATURE
dp Particle diameter (m)
F Additional acceleration (force/unit particle mass) (N kg–1)
FD Drag force (N)
g Gravity (m s–2)
k Turbulent kinetic energy of gas phase (m2 s–2)
P Pressure (Pa)
s Viscous stress tensor (kg m–1 s–2)

u Fluid velocity (m s–1)
Greek letters
βgs Interphase momentum transfer coefficient (kg m–3 s–1)
ξ Bulk viscosity (kg m–1 s–1)
σ Eddy parameter (-)
ε Turbulent dissipation energy for gas phase (m2 s–3)
ρ Fluid density (kg m–3)
α Void fraction (-)
μ Fluid viscosity (Pa s)
Superscripts
g Gas
k Kinetic
p Particle
s Solid
REFERENCES
1. Palamanit A., Prachayawarakorn S., Tungtrakul P., Performance evaluation of top-spray fluidized bed
coating for healthy coated rice production, Food and Bioprocess Technology, 9: 1317–1326, 2016.
2. Börner M., Hagemeier T., Ganzer G., Peglow M., Experimental spray zone characterization in top-spray
fluidized bed granulation, Chemical Engineering Science, 116: 317–330, 2014.
3. Solís-Morales D., Sáenz-Hernández C.M., Ortega-Rivas E., Attrition reduction and quality improve
ment of coated puffed wheat by fluidized bed technology, Journal of Food Engineering, 93: 236–241,
2009.
4. Palamanit A., Soponronnarit S., Prachayawarakorn S., Tungtrakul P., Effects of inlet air temperature
and spray rate of coating solution on quality attributes of turmeric extract coated rice using top-spray
fluidized bed coating technique, Journal of Food Engineering, 114: 132–138, 2013.
5. Duangkhamchan W., Siriamornpun S., Quality attributes and anthocyanin content of rice coated by
purple-corn cob extract as affected by coating conditions, Food and Bioproducts Processing, 96:
171–179, 2015.
6. Sakurai Y., Mise R., Kimura S.-I., Noguchi S., Iwao Y., Novel method for improving the water dis-
persibility and flowability of fine green tea powder using a fluidized bed granulator, Journal of Food
Engineering, 206: 118–124, 2017.
7. Duangkhamchan W., Incheun S., Texture and antioxidant properties of puffed pounded-unripe rice
coated by marigold extract as affected by fluidized-bed coating conditions, Key Engineering Materials,
718: 115–119, 2017.
8. Dordević V., Balanč B., Belščak-Cvitanović A., Lević S., Trifković K., Kalušević A., Kostić I., Komes
D., Bugarski B., Nedović V., Trends in encapsulation technologies for delivery of food bioactive com-
pounds, Food Engineering Reviews, 7: 452–490, 2015.
9. Kulah G., Kaya O., Investigation and scale-up of hot-melt coating of pharmaceuticals in fluidized beds,
Powder Technology, 208: 175–184, 2011.
10. Frey C., Fluid bed coating-based microencapsulation, In: Microencapsulation in the Food Industry,
Gaonkar A., Vasisht N., Khare A., Sobel R., Eds., Elsevier, Amsterdam, Netherlands, 65–81, 2014.
11. Depypere F., Pieters J.G., Dewettinck K., PEPT visualization of particle motion in a tapered fluidized
bed coater, Journal of Food Engineering, 93: 324–336, 2009.
12. Guignon B., Duquenoy A., Dumoulin E.D., Fluid bed encapsulation of particles: Principles and practice,
Drying Technology, 20(2): 419–447, 2002.
13. Duangkhamchan W., Ronsse F., Depypere F., Dewettinck K., Pieters J.G., Comparison and evaluation of
interphase momentum exchange models for simulation of the solids volume fraction in tapered fluidized
beds, Chemical Engineering Science, 65: 3100–3112, 2010.
14. Duangkhamchan W., Ronsse F., Dewettinck K., Pieters J.G., CFD study of solids concentration in a
fluidized-bed coater with variation of atomization air pressure, Powder Technology, 212: 103–114, 2011.
15. Qiu G., Ye J., Wang H., Yang W., Investigation of flow hydrodynamics and regime transition in a
gas-solids fluidized bed with different riser diameters, Chemical Engineering Science, 116: 195–207,
2014.
16. Wang H., Qiu G., Ye J., Yang W., Experimental study and modeling on gas-solid flow in a lab-scale
fluidized bed with Wurster tube, Powder Technology, 300: 14–27, 2016.
17. Saidi M., Tabrizi H.B., Influences of the fluidizing and spouting pulsation on particle motion in spout-
fluid beds, Particuology, 36: 139–148, 2018. https://doi.org/10.1016/j.partic.2016.11.007.
18. Ronsse F., Pieters J.G., Dewettinck K., Combined population balance and thermodynamic modeling of
the batch top-spray fluidized bed coating process. Part I—Model development and validation, Journal
19. Duangkhamchan W., Ronsse F., Depypere F., Dewettinck K., Pieters J.G., CFD study of droplet atomi-
zation using a binary nozzle in fluidized bed coating, Chemical Engineering Science, 68: 555–566,
2012.
20. Rybdylova O., Qubeissi M.A., Braun M., Crua C., Manin J., Pickett L.M., de Sercey G., Sazhina E.M.,
Sazhin S.S., Heikal M., A model for droplet heating and its implementation into ANSYS Fluent,
International Communications in Heat and Mass Transfer, 76: 265–270, 2016.
21. Chen Q., Thu K., Bui T.D., Li Y., Ng K.C., Chua K.J. Development of a model for spray evaporation
based on droplet analysis, Desalination, 399: 69–77, 2016.
22. Cheng P., Li Q., Xu S., Kang Z., On the prediction of spray angle of liquid-liquid pintle injectors, Acta
Astronautica, 138: 145–151, 2017.
23. Duangkhamchan W., Ronsse F., Siriamornpun S., Pieters J.G., Numerical study of air humidity and
temperature distribution in a top-spray fluidised bed coating process, Journal of Food Engineering, 146:
81–91, 2015.
24. ANSYS Inc. Ansys Fluent 12.0 Theory’s Guide Manual. Ansys Inc., Canonsburg, PA, USA, 2009.
25. Sau D.D., Biswal K.C., Computational fluid dynamics and experimental study of the hydrodynamics of
a gas-solid tapered fluidized bed, Applied Mathematical Modeling, 35: 2265–2278, 2011.
26. Gidaspow D., Multiphase Flow and Fluidisation, Academic Press, Cambridge, MA, USA, 1994.
27. Das A.K., De Wilde J., Heynderickx G.J., Marin G.B., Vierendeels J., Dick E., CFD simulation of dilute
phase gas – solid riser reactor: Part I—A new solution method and flow model validation, Chemical
Engineering Science, 59: 167–186, 2004.
28. Werner S.R.L., Jones J.R., Paterson A.H.J., Archer R.H., Pearce D.L., Air-suspension coating in the
food industry: Part II—Micro-level process approach, Powder Technology, 171: 34–45, 2007.
29. Jakobean H.A., Fluidized bed reactors, In: Chemical Reactor Modeling, Springer-Verlag, New York,
USA, 1005–1056, 2014.
30. Depypere F., Pieters J.G., Dewettinck K., CFD analysis of air distribution in fluidized bed equipment,
Powder Technology, 145: 176–189, 2004.
31. Peirano E., Delloume V., Johnsson F., Leckner B., Simonin O., Numerical simulation of the fluid dynam-
ics of a freely bubbling fluidized bed: Influence of the air supply system, Powder Technology, 122(1):
69–82, 2002.
32. Lanza A., Islam M.A., de Lasa H., Particle clusters and drag coefficients in gas-solid downer units,
Chemical Engineering Science, 200–202: 439–451, 2012.
33. Friedle M., Niyogi K., Torregrosa M.M., Marin G.B., Heynderickx G.J., A drag model for the gas-solid
vortex unit, Powder Technology, 312: 210–221, 2017.
34. Romero Luna C.M., Carrocci L.R., Arce G.L.A.F., Ávila I., A comparative assessment of empirical and
lattice Boltzmann method-based drag models for simulation of gas-solid flow hydrodynamics in a bub-
bling fluidized bed, Particuology, 33: 129–137, 2017.
35. Qiu X., Wang L., Yang N., Li J., A simplified two-fluid model coupled with EMMS drag for gas-solid
flows, Powder Technology, 314: 299–314, 2017.
36. Ergun S., Fluid flow through packed columns, Chemical Engineering Progress, 48: 89–94, 1952.
37. Wen X., Yu Y.H., Mechanics of fluidization, Chemical Engineering Progress, 62: 100–111, 1966.
38. Ding J., Gidapow D., A bubbling fluidization model using kinetic-theory of granular flow, AIChE
Journal, 36: 523–538, 1990.
39. Huilin L., Gidaspow D., Hydrodynamics of binary fluidization in a riser: CFD simulation using two
granular temperatures, Chemical Engineering Science, 58: 3777–3792, 2003.
40. Symlal M., O’Brien T.J., Computer simulation of bubbles in a fluidized bed, AIChE Symposium Series,
85: 22–31, 1989.
41. Arastoopour H., Pakdel P., Adewumi M., Hydrodynamic analysis of dilute gas-solids flow in a vertical
pipe, Powder Technology, 62: 163–170, 1990.
42. Symlal M., The particle-particle drag term in multiparticle model of fluidization, DOE/MC/21353-2373,
DE87 006500, 1987.
43. Morsi S.A., Alexzander A.J., An investigation of particle trajectories in two-phase flow systems, Journal
of Fluid Mechanics, 55: 193–208, 1972.
44. Schmidt D.P., Nouar I., Senecal P.K., Rutland C.J., Martin J.K., Reitz R.D., Pressure-swirl atomization
in the near field, SAE Transactions, Journal of Engines, 108(3): 471–484, 1999.
45. Gunn D.J., Transfer to heat or mass to particle in fixed and fluidized beds, International Journal of Heat
and Mass Transfer, 21: 467–476, 1978.
46. Ranz W.E., Marshall W.R., Evaporation from drops, Chemical Engineering Progress, 48: 141–146;
173–180, 1952.
47. Jiménez T., Turchiuli C., Dumoulin E., Particles agglomeration in a conical fluidized bed in relation
with air temperature profiles, Chemical Engineering Science, 61: 5954–5961, 2006.
48. Fries L., Antonyuk S., Heinrich S., Palzer S., DEM-CFD modeling of a fluidized bed spray granulator,
Chemical Engineering Science, 66: 2340–2355, 2011.
49. Li L., Remmelgas J., van Wachem B.G., von Corswant C., Johansson M., Folestad S., Rasmuson A.,
Residence time distributions of different size particles in the spray zone of a Wurster fluid bed studied
using DEM-CFD, Powder Technology, 280: 124–134, 2015.
50. Börner M., Bück A., Tsotsas E., DEM-CFD investigation of particle residence time distribution in top-
spray fluidized bed granulation, Chemical Engineering Science, 161: 187–197, 2017.
25 Operation of Biofilm Reactors
for the Food Industry Using
CFD
Luciana C. Gomes, João Miranda, and Filipe J. Mergulhão
CONTENTS
25.1 Introduction......................................................................................................................... 561
25.2 Hydrodynamics and Biofilm Formation.............................................................................. 562
25.3 The Use of CFD in Biofilm Reactors.................................................................................. 565
25.4 Biofilm Reactors.................................................................................................................. 566
25.4.1 Microtiter Plates...................................................................................................... 567
25.4.1.1 96-Well MTPs......................................................................................... 567
25.4.1.2 48-Well MTPs......................................................................................... 568
25.4.1.3 24-Well MTPs......................................................................................... 570
25.4.1.4 12-Well MTPs......................................................................................... 570
25.4.1.5 6-Well MTPs........................................................................................... 575
25.4.2 Flow Cells............................................................................................................... 578
25.4.2.1 Robbins Device and Modifications......................................................... 579
25.4.2.2 Flow Chambers for Direct Inspection of Biofilms.................................. 580
25.4.3 Microfluidic Devices............................................................................................... 582
25.5 Conclusions.......................................................................................................................... 583
Nomenclature.................................................................................................................................. 584
References....................................................................................................................................... 584
25.1 INTRODUCTION
Microbial biofilms were first discovered on the surface of teeth by A. van Leeuwenhoek [1] and
they were initially described by J.W. Costerton in 1978 [2]. Biofilms are microbial communities
commonly attached to a surface, forming three-dimensional structures composed essentially of
cells embedded in a self-produced matrix [3]. This matrix is mainly constituted by water (97%),
polysaccharides (1–2%), proteins (<1–2%) and nucleic acids (<1–2%), and is responsible for biofilm
morphology, functional integrity, cohesion and structure [4,5]. The biofilm mode of living confers
protection against harmful environments (nutrient deprivation, pH changes, oxygen radicals, hydro-
dynamic conditions and antimicrobial agents), enables genetic material transference and facilitates
the colonization of favorable and hostile niches [6]. It is estimated that more than 90% of bac-
teria in natural environments exist within a biofilm [7]. In the food industry, it is believed that
biofilm formation enables bacteria to cope with several stresses that are found in food processing
environments such as refrigeration, acidity, salinity, desiccation and exposure to disinfection [8,9].
Although biofilm formation by some bacteria can be advantageous, particularly in fermented foods
or as means of protection against colonization by spoilage or pathogenic bacteria [8], most biofilms
in the food industry have a detrimental effect on the process, on the final product or on the consumer.
The Center for Disease Control and Prevention (CDC) has reported that between 1996 and 2010,
561
FIGURE 25.1 Impact of biofilms in the food industry: key factors affecting biofilm development, advantages
to the organisms, problems to industry and main biofilm formation sites.
48 million illnesses, 128,000 hospitalizations and 3,000 deaths occurred annually in the U.S. due
to foodborne diseases [10]. Several factors affect the biofilm development (Figure 25.1), including
environmental, microbial and surface-related factors. The control of biofilms formed in different
areas of an industrial plant is a challenging task that includes hygienic design, good manufacture
practices and adequate sanitation programs [11]. Different organisms form biofilms in food indus-
tries and some representative examples can be found in Table 25.1.
25.2 HYDRODYNAMICS AND BIOFILM FORMATION

Hydrodynamics have a very strong influence in the process of biofilm formation (Figure 25.1). During
initial adhesion, hydrodynamics affect the transport rate of macromolecules and bacteria to the surface,
the time they reside in close proximity to the surface and the mechanical shear forces at the surface-
fluid interface [12]. Later, during biofilm maturation, hydrodynamics are also extremely important in
the transport of macromolecules since the fluid surrounding a biofilm is the primary source for nutri-
ents and the vehicle for cell by-products removal [13]. An increase in flow velocity promotes the flux
(in and out) of molecules (nutrients, cells, biocides, antibiotics, cellular products, etc.) by changing the
molecule concentration in the fluid biofilm-interface. Hydrodynamics also regulate the physiological
properties of the biofilm by changing the mechanical shear stresses at the interface [12]. It has been
observed that higher shear forces lead to the formation of thinner and denser biofilms [14]. Although
higher flow velocities enhance molecular transport by convection, shear forces lead to denser biofilms,
which in turn reduces the diffusivity of molecules inside them [15,16]. Additionally, a higher flow
velocity promotes stronger shear forces that can promote biofilm sloughing or detachment [14].
Given the importance that shear forces have on cell adhesion and biofilm development, it is
important to characterize them. Two parameters are often used to characterize shear effects: the
shear stress and the shear rate. The shear rate is the derivative of the velocity in the perpendicular
Operation of Biofilm Reactors for the Food Industry Using CFD 563
TABLE 25.1
Organisms Commonly Forming Biofilms in Food Industries
Industry Organism
Dairy Achromobacter Geobacillus thermoglucosidans
Acinetobacter Klebsiella oxytoca
Aeromonas Lactobacillus spp.
Alcaligenes Listeria monocytogenes
Anoxybacillus flavithermus Microbacterium
Arthrobacter Micrococcus
Bacillus coagulans Pseudomonas chlororaphis
Bacillus cereus Pseudomonas fluorescens
Bacillus licheniformis Pseudomonas fragi
Bacillus subtilis Paenibacillus polymyxa
Campylobacter spp. Paenibacillus spp.
Clostridium tyrobutyricum Pediococcus
Corynebacterium Salmonella agona
Cronobacter malonaticus Salmonella infantis
Cronobacter turicensis Salmonella typhimurium
Cronobacter sakazakii Serratia
Escherichia coli Shigella spp.
Flavobacteriumm Staphylococcus aureus
Geobacillus stearothermophilus Streptococcus thermophilus
Fish processing Aeromonas hydrophila Salmonella montevideo
Bacillus spp. Serratia liquefaciens
Listeria monocytogenes Vibrio parahaemolyticus
Neisseriaceae Vibrio alginolyticus
Pseudomonas spp. Vibrio cholerae
Salmonella agona Vibrio vulnificus
Seafood processing Listeria monocyogenes Shigella spp.
Pseudomonas fluorescens Staphylococcus aureus
Pseudomonas putida Vibrio parahaemolyticus
Salmonella spp.
Poultry processing Campylobacter jejuni Salmonella typhimurium
Listeria monocytogenes Salmonella sofia
Pseudomonas fragi
Meat industry Acinetobacter calcoaceticus Flavobacterium
Aeromonas Gamella
Alcaligenes Listeria monocytogenes
Bacillus thermosphacta Lactobacillus
Branhamella Leuconostoc
Brochothrix Micrococcus
Corynebacterium Moraxella
Escherichia coli O157:H7 Proteus
Erysipelothrix Pseudomonas putida
Staphylococcus aureus Streptococcus.
Salmonella enterica Yersinia enterocolitica
Shewanella putrefaciens
Ready to eat foods Escherichia coli O157:H7 Listeria monocytogenes
(Continued )

Organisms Commonly Forming Biofilms in Food Industries
Industry Organism
Brewing Acinetobacter spp. Megasphaera sueciensis
Chryseobacterium species Pectinatus haikarae
Claussenii Pectinatus spp.
Lactobacillus brevis Pediococcus damnosus
Lactobacillus paracollinoides Pseudomonas spp.
Megasphaera paucivorans
Wineries, soft drinks, Escherichia coli O157:H7 Pediococcus
and juice Lactobacillus Salmonella
Leuconostoc oenos
Sources: E. Giaouris, E. Heir, M. Hébraud, N. Chorianopoulos et al., “Attachment and biofilm forma-
tion by foodborne bacteria in meat processing environments: Causes, implications, role of
bacterial interactions and control by alternative novel methods”, Meat Science, vol. 97,
no. 3, pp. 298–309, 2014; S. Srey, I. K. Jahid, and S.-D. Ha, “Biofilm formation in food
industries: A food safety concern”, Food Control, vol. 31, no. 2, pp. 572–585, 2013; P. M.
Fratamico, B. A. Annous, and N. W. Gunther, Biofilms in the Food and Beverage Industries,
P.M. Fratamico, B.A. Annous, and N.W. Gunther, Editors. 2009, Cornwall, UK: Woodhead
Publishing in Food Science, Technology and Nutrition; K. H. Teh, S. Flint, J. Brooks, and
G. Knight, Biofilms in the Dairy Industry. 2015, Chichester, UK: John Wiley & Sons, Ltd.
direction from the surface [17] and is also an indicator of the frequency at which cells contact the
surface. For this reason, it is frequently used when studying phenomena occurring during adhesion
and early biofilm development. Table 25.2 lists typical shear rate values that can be found in the
food industry. The shear stress in Newtonian fluids is proportional to the shear rate where viscosity
is the constant of proportionality [17]. It is a measure of the force of friction (parallel to the surface)
TABLE 25.2
Typical Food Processes and Their Associated Shear Strain Rates
Phenomenon or Equipment Shear Rate (s−1) References
Agitated thin film evaporator with a sugar solution 0.0014−302 [78]
Extrusion 0.1−5,000 [79,80]
Single-screw extruders 5−55 [81]
Pipe flow 1−1,000 [79]
Calendering 10−100 [79]
Pouring 10−100 [79]
Spreading (knife) 1−100 [79]
Anchor mixers in agitation of xanthan gum 0.0841−1,700 [82]
Mixing and stirring 10−1,000 [79]
Wall of a planetary mixer handling cake batter 20−500 [79]
Blender >1,000 [83]
Annular space of a scraped surface heat exchanger <40 [84]
Plate heat exchanger for yogurt processing 50−1,750 [85]
Yogurt filling machine 800−1,250 [85]
Chocolate tempering device 500−12,000 [86]
Scraped surface heat exchanger 50,000 [87]
Two-fluid nozzle of a pilot-scale spray dryer 145 × 106 [88]
TABLE 25.3
Typical Food Processes and Their Associated Shear Stress Ranges
Phenomenon or Equipment Shear Stress (Pa) References
Milk coagulation <0.5 [89]
Pipeline elbows 0.009 [90]
Dead-ends 0.05−18.9 [91]
Removal of deposits from stainless steel tubes 0.09 [92]
Corners of a washing tank <0.1 [93]
Angles of a washing tank 0.1−0.4 [93]
Cleaning of a downstand T-piece 0−5 [94]
Mix proof valve 0−0.25 [94]
Three way valve 0.4−1.7 [95]
Half open butterfly valve 0−190 [91]
Fully open butterfly valves 3−10 [91]
Product fill valve 0−1,180 [96]
Milk spray dryer 0−0.4 [97]
Cleaning-in-place pilot plant 0−5 [98]
Tubular membrane channel 0.25−0.5 [99]
Membrane emulsification process 0−140 [100]
Plate heat exchanger for yogurt processing 6.7 and 20−46 [101,102]
Plate heat exchanger of an ice slurry system 50−100 [103]
Pilot-scale plate heat exchanger for milk treatment 150−450 [101]
from the fluid acting on the adhered cells or on the biofilm. As such, it is frequently used as a
descriptor of shear forces acting on the biofilm during maturation or detachment (by sloughing or
erosion). Table 25.3 lists typical shear stress values that can be found in the food industry.
25.3 THE USE OF CFD IN BIOFILM REACTORS

CFD uses numerical methods to solve problems involving fluid flow. Although the history of CFD
can be traced back to more than 200 years ago when Navier and Stokes integrated viscous transport
into Euler equations and developed the equations for the conservation of mass and momentum,
the development of commercial CFD software only started in the 1980s [18]. Some examples of
commonly used CFD software are ANSYS Fluent®, ANSYS CFX, FLOWIZARD, COMSOL and
OpenFOAM.
In recent years, CFD modeling has become a widely accepted technique to study local charac-
teristics in simple and complex flows. This technique enables the estimation of fluid flow param-
eters that are hard or impossible to measure and the prediction of parameters that can only be
measured by time-consuming techniques or that require optical access, which is not available in
many situations [19].
CFD requires that the geometry to be analyzed is divided into a finite set of volumes called cells,
forming a computational grid called mesh. Then, the differential equations for the conservation
of mass, momentum and energy, representing the balances of accumulation, net inflow (mostly by
convection and diffusion) and volumetric production are set. These equations are then discretized
into algebraic equations that are solved for each cell, resulting in a flow field [18].
Simulations for laminar single phase flows can be obtained by direct resolution of the Navier–
Stokes (NS) equations, whereas turbulence can be tackled with different models. The most commonly
used models are based on the steady-state RANS (Reynolds-averaged Navier–Stokes) equations
since transient models such as LES (large eddy simulation) require much longer computational
times [20]. Current CFD packages include simple zero and single equation models (e.g., the Spalart–
Allmaras model), complex multiple equation models (e.g., the Reynolds stress model) and the two
equation turbulence models of the κ-ε and κ-ω families, which have become the industry standard
[19]. Of particular interest for some applications is the shear-stress transport (SST) κ-ω model,
which combines the accuracy of the κ-ω model in the near-wall region with the free-stream inde-
pendence of the κ–ε mode in the far field [21]. This model is particularly suited for cases of both
rough turbulence and smooth shear flow [19]. Thanks to the increasing computing power, the LES
approach has been increasingly used. In this approach, large eddies are resolved directly, while
small eddies are modeled [19].
For multiphase flows and when tracking of the interphase is required, the most adequate model is
the volume-of-fluid (VOF) model, which is based on the assumption that there is no interpenetration
between the different phases [22]. Assuming that the phases share the same velocity and pressure
field, a single set of mass and momentum equations is solved and topology changes to the interface
induced by the relative motion between the two phases are tracked in time [19]. Surface tension is
usually modeled according to the continuum surface force model [23] and the free surface is often
modeled by the geometric reconstruction scheme [24].
Using CFD to characterize the hydrodynamics occurring inside biofilm formation platforms is
very useful to better understand the biofilm behavior, but one must bear in mind that simulations
are mostly performed for clean surfaces. Thus, these simulations are particularly suited for the study
of initial adhesion or early stages of biofilm development. Since food contact surfaces are cleaned
more frequently than most industrial surfaces (sometimes every 4 hours) [25], this is probably a very
good approximation. When biofilms are formed, the cross sectional flow area is reduced, resulting
in an increase of the bulk flow velocity and wall shear stress. Since surfaces in the food industry are
cleaned so frequently, the thickness of the formed biofilms is unlikely to have a significant impact
on flow dynamics and shear stress distribution [26]. A further assumption is that biofilm structures
are not mobile and therefore a no-slip condition can be used at the flow boundaries [27,28]. The
rheological properties of biofilms have been described in different studies showing that biofilms are
viscoelastic and have a high viscous modulus [29,30], which is often much larger than the viscosity
of the surrounding fluid. Thus, the deformation rates and interface velocity are very small and the
magnitude of the wall shear stress is indistinguishable from that at the interface [26]. These assump-
tions must be verified for each particular case since highly viscous and non-Newtonian fluids are
common in many food processes.
25.4 BIOFILM REACTORS
In the context of this chapter, biofilm reactors are platforms for the study of biofilms in labora-
tory conditions. These reactors are necessary because a detailed knowledge of the steps governing
biofilm formation, development and elimination is only possible if the biofilms under study are
obtained in controlled settings where key variables can be manipulated and the implications of
these manipulations can be readily assayed. Although completely reproducible biofilms are nearly
impossible to obtain, the development of in vitro platforms for biofilm studies is a major step towards
the standardization of procedures and also for a better control of the environmental conditions that
affect their development [31]. There are several reactors available for biofilm studies and select-
ing the most adequate platform determines the type of data that can be obtained from them. This
chapter describes the most widely used biofilm reactors that enable biofilm formation in controlled
hydrodynamic conditions and where CFD has been used to determine relevant flow characteristics.
Examples of the use of these reactors will be given when appropriate for the simulation of the
hydrodynamics (shear stress and shear rate) found in the food industry (Tables 25.2 and 25.3). It is
important to point out that the values of shear stress or shear rate provided here are merely indica-
tive of what has been obtained in physical experiments and do not represent maximum values of
these parameters that can be attained, for instance, using higher shaking frequencies or flow rates.
25.4.1 Microtiter Plates
Microtiter plates (MTPs) are currently the most widely used platform for biofilm formation studies.
They were originally used by Madilyn Fletcher to study bacterial attachment [32] and were later
adopted for biofilm studies [33]. MTPs are flat plates with multiple wells arranged in a rectangular
array with a 2:3 aspect ratio, resulting in 6, 12, 24, 48, and 96 wells. The wells have designs similar
to test tubes, and may have flat bottoms, rounded or tapered, and their volumes can range from tens
of microliters to few milliliters, depending on the number of wells [34]. MTPs are constructed from
different polymers, although the transparent polystyrene (PS) is the most widely used. Plates may also
be white pigmented for use in luminescence assays [35] or black to be used in fluorescence tests [36].
Polypropylene microplates (PP) are also manufactured and they are more resistant to high tempera-
tures [37]. Although many researchers use MTPs in static conditions, they can be placed in orbital
incubators and used for dynamic biofilm studies [38]. Usually, bacterial cells are grown in the wells
of a polystyrene MTP and at different time points the wells are emptied and washed to remove the
planktonic cells [39]. The biomass attached to the surface of the wells can be stained with an aque-
ous solution of crystal violet or the cells can be imaged with or without staining by microscopy [40].
MTPs can be considered as arrays of miniature reactors with identical shapes and hydrodynamic
conditions [41]. Additionally, these devices are easy to handle with the use of robotic systems and mul-
tichannel micropipettes, allowing the rapid development of biofilms in different cultivation conditions.
Therefore, it is possible to study the adhesion of different bacterial strains and consequent biofilm
formation in rapid and inexpensive assays due to their reduced volume [42]. MTPs are commonly used
in screening applications, for instance when testing new antibiofilm compounds [43]. Depending on
the format used (discussed in the following sections), they enable high throughput at an affordable cost
and sometimes non-invasive imaging [44]. In some cases, particularly for larger well dimensions, it
is possible to place coupons at the bottom of the wells so that different surfaces can be tested [45,46].
The main limitations of MTPs are that loosely attached cells may not be measured correctly due to
detachment during washing and biofilms formed in this platform may be affected by sedimentation.
25.4.1.1 96-Well MTPs
This format is currently the most intensively used MTP format, essentially for screening purposes. These
plates are extremely sensitive to sedimentation and direct inspection of the biofilm is possible, but techni-
cally difficult [47,48]. They are particularly suited for short-term experiments as they operate in batch mode
with the inherent exhaustion of nutrients and accumulation of toxic metabolites [44]. Although they enable
high throughput, researchers are forced to use replicates in the same plate or in different plates (as liquid
evaporation can be positional), which greatly reduces Introduce their throughput. Results obtained in this
platform often lack reproducibility, mostly due to pipetting errors and evaporation, which can create micro-
environments of osmolarity. These plates are not compatible with the use of coupons as the bottom surface
is relatively small, so only a limited number of surfaces can be assayed (limited to the construction materi-
als of these plates). Biofilms will form on the bottom and also on the side walls of the wells, and it has been
shown that the majority of the biofilm is sometimes located at the air-liquid interface [47,48]. Since the liq-
uid surface oscillates during agitation [38] and most of the biofilm can be found at the interface, it is impor-
tant to load an adequate volume of crystal violet staining solution (a 20% higher volume than used during
the biofilm formation assay should be enough) when the biofilm amount is determined by this procedure.
We have previously used CFD to estimate the shear rate in 96-well MTPs using different agitation fre-
quencies and orbital diameters (Figure 25.2). As in all agitated MPTs, higher shear values were obtained
next to the air-liquid interface and since the shear stress field was not uniform in the entire well, biofilm
formation was different in different well locations [40]. Although the shear values that can be obtained
are relatively small, industrial plants have many places like crevices, corners, valves and joints where
low fluid velocities are attained, making these zones suitable niches for bacterial accumulation and bio-
film growth. Additionally, these places are sometimes shielded from the main fluid carrying biocides
during cleaning in place (CIP) operations, thus rendering biofilm control in those areas a particularly
50 rpm 100 rpm 150 rpm 200 rpm
1.5 1.5 1.5 1.5
25 mm
1 1 1 1
z/D
z/D
z/D
z/D
0.5 0.5 0.5 0.5
0 0 0 0
–0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5
y/D y/D y/D y/D
1.5 1.5 1.5 1.5
50 mm
1 1 1 1
z/D
z/D
z/D
z/D
0.5 0.5 0.5 0.5
0 0 0 0
–0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5
y/D y/D y/D y/D
1.5 1.5 1.5 1.5
75 mm
1 1 1 1
z/D
z/D
z/D
z/D
0.5 0.5 0.5 0.5
0 0 0 0
–0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5
y/D y/D y/D y/D
1.5 1.5 1.5 1.5
100 mm
1 1 1 1
z/D
z/D
z/D
z/D
0.5 0.5 0.5 0.5
0 0 0 0
–0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5 –0.5–0.25 0 0.25 0.5
y/D y/D y/D y/D
Strain rate / s–1
20 40 60 80 100 120 140 160 180
FIGURE 25.2 Time-averaged shear strain rates on 96-well MTPs at different orbital shaking diameters and
shaking frequencies. The diameter of the circle is proportional to the diameter of the orbit described by each
well of the plate when placed on an orbital incubator with the indicated orbital shaking diameter. Strain rates
below 20 s−1 are not represented. (Reprinted from Gomes, L.C., Moreira, J.M., Teodósio, J.S., Araújo, J.D.,
Miranda, J.M., Simões, M., Melo, L.F., Mergulhão, F.J., 96-well microtiter plates for biofouling simulation in
biomedical settings. Biofouling, 2014. 30(5): p. 535–546.)
difficult task. Although the flow topology in some of these areas is different from what is obtained in
96-well MTPs, we have previously shown that differences in flow topology have a small impact on
biofilm formation at reduced shear stress levels [49]. As expected, the average wall shear rate increases
with the shaking frequency and orbital diameter (Figure 25.3). The information presented in Figure 25.3
is extremely important for biofilm researchers since it defines which operational conditions should be
chosen so that the biofilms formed in 96-well MTPs mimic those found in food processing scenarios.
25.4.1.2 48-Well MTPs
These plates also enable a very high throughput and relatively low shear forces. The use of coupons
is not practical, not only because of the relatively small surface area, but also due to the reduced
150
100 mm
125
S hear rate ( s )
100
-1
75 mm
75
50 mm
50
25 mm
25
0
50 100 150 200
Shaking frequency (rpm)

FIGURE 25.3 Numerical results of the shear rate (s−1) as a function of the shaking frequency (rpm) for dif-
ferent orbital shaking diameters (mm). (Adapted from Gomes, L.C., Moreira, J.M., Teodósio, J.S., Araújo,
J.D., Miranda, J.M., Simões, M., Melo, L.F., Mergulhão, F.J., 96-well microtiter plates for biofouling simula-
tion in biomedical settings. Biofouling, 2014. 30(5): p. 535–546.)
shear forces that can be obtained on the bottom of the wells. In fact, Lv et al. [50] reported shear
rate values below 10 s−1 in this region using CFD simulation (Figure 25.4a) [50]. Thus, MTPs can
be used for biofilm formation in both the side walls as well as on the bottom of the well. Given that
coupons are rarely used with these systems, the types of surfaces that can be assayed are also lim-
ited. As biofilm formation platforms, these MTPs share many features with the 96-well format,
FIGURE 25.4 Shear rates (s−1) on (a) 48-well and (b) 24-well U-bottom MTPs. (Adapted from Lv, X., Song, J., Yu,
B., Liu, H., Li, C., Zhuang, Y., Wang Y., High-throughput system for screening of high L-lactic acid-productivity
strains in deep-well microtiter plates. Bioprocess and Biosystems Engineering, 2016. 39(11): p. 1737–1747.)
FIGURE 25.5 Comparison of the evaporation rates at different positions of a U-bottom and round-bottom
24-well, U-bottom 48-well and round-bottom 96-well MTP. (Reprinted from Lv, X., Song, J., Yu, B., Liu, H.,
Li, C., Zhuang, Y., Wang, Y., High-throughput system for screening of high L-lactic acid-productivity strains
in deep-well microtiter plates. Bioprocess and Biosystems Engineering, 2016. 39(11): p. 1737–1747.)
although slightly higher shear forces can be achieved at the cost of a lower throughput. Lv et al.
[50] have shown that in some conditions the evaporation rate is similar between both plate formats
and significantly higher when compared to 24-well MTPs (Figure 25.5). These authors have used
48-well MTPs for screening of high L-lactic acid productivity strains and were able to reach average
shear rates of 18 s−1 in their experiments (Table 25.4).
25.4.1.3 24-Well MTPs
As the volume of the wells increases, it becomes possible to use these platforms for biofilm forma-
tion in the entire wetted surface or at the bottom by using coupons. In this case, biofilms will still
form on the side walls, but the quantitation of the biofilm formed at the bottom is possible by remov-
ing the coupon from the well after the desired experimental time. Circular or square coupons (up to
1 cm) can be used for this purpose.
Lv et al. [50] have used these plates for cultivation of Bacillus coagulans and their CFD simula-
tions indicated that average shear rates of 27 and 20 s−1 were attained using liquid volumes of 2 and
3 mL, respectively. Barrett et al. [51] have used higher shaking frequencies and were able to attain
average shear rates up to 190 s−1, according to CFD simulations. These are average shear rates con-
sidering the entire wetted surface, which includes the air-liquid interface where higher shear forces
are observed. If only the bottom region is considered, the shear rates would drop to values below
10 s−1 (Figure 25.4b) [50].
25.4.1.4 12-Well MTPs
These MTPs are a very attractive format and although theoretically their throughput is much lower
than the 96-well MTPs, the results obtained with this platform are generally more reproducible due
to the higher liquid volume (which reduces the impact of pipetting errors) decreasing the need for a
large number of replicate wells. The larger working volume also enables longer experimental times
(although medium replacement is often necessary) and the negative impacts of liquid evaporation
TABLE 25.4
Hydrodynamic Parameters Determined by CFD for the Different Types of Biofilm Reactors
Shear Rate
Biofilm Reactor Format Agitation/Flow Software Model (s−1) Shear Stress (Pa) Velocity (m s−1) References
Microtiter plate 96 wells Orbital shaking (ϕ = 25 and FLUENT VOF 23.0 and 0.070 and 0.034** – [38,40,48]
50 mm; υ = 150 rpm) 46.2**
Orbital shaking (ϕ = 25, 50, FLUENT VOF 10−142 ** – – [104]
75 and 100 mm; υ = 50, 100,
150 and 200 rpm) and static
Round-bottom Orbital shaking (ϕ = 25 mm; CFX VOF 6.3** – – [50]
96 wells υ = 200 rpm) RNG k-ε(a)
U-bottom 48 wells Orbital shaking (ϕ = 25 mm; CFX VOF 18.0** – – [50]
υ = 200 rpm) RNG k-ε(a)
U- and round- Orbital shaking (ϕ = 25 mm; CFX VOF 20.2 and – – [50]
bottom 24 wells υ = 200 rpm) RNG k-ε(a) 27.2**
24 wells Orbital shaking (ϕ = 20 mm; CFX VOF 190, 130, 98, – – [51]
υ = 120, 150, 200, 225, 250, 99, 84, 73
300 and 400 rpm) and 150**
12 wells Orbital shaking (ϕ = 25 mm; FLUENT VOF 5−42* 0.008−0.070* – This chapter
υ = 40, 100, 140 and 180 rpm)
Operation of Biofilm Reactors for the Food Industry Using CFD
6 wells Orbital shaking (ϕ = 14 mm; FLUENT VOF 0−334* 0−0.317* ≤ 0.265 [52]
υ = 60 and 120 rpm)
Orbital shaking (ϕ = 19 mm; FLUENT VOF – 0.15−1.02* – [53,54]
υ = 100 and 200 rpm)
Orbital shaking (ϕ = 4 mm; FLUENT VOF 15−58* 0.025−0.140* – This
υ = 160 rpm) chapter
Flow cells Semi-circular flow Continuous flow FLUENT (SST) κ-ω – 0.4 0.275 [61]
cell (F = 5833 mL min−1)
(Continued)
571
572

Biofilm Shear Rate
Reactor Format Agitation/Flow Software Model (s−1) Shear Stress (Pa) Velocity (m s−1) References
Continuous flow FLUENT (SST) κ-ω – 0.6 0.4 [58]
(F = 6900 mL min−1)
Square and Continuous flow FLUENT NS Laminar 2.82−110 0.02−0.14 0−0.0173 [26]
rectangular (F = 2 mL min−1)
cross-sections
Shear stress flow F = 0–420 mL min−1 COMSOL 3D NS − 0−237 − [63]
chamber Laminar
Glass capillary Continuous flow CFX 3D NS − 2.7 × 10−2− − [105]
tubes with square (F = 0.4 mL min−1) Laminar 8.4 × 10−2
cross-sections
Parallel plate flow Continuous flow FLUENT NS equations − 0.07 − [49]
chamber (F = 660 mL min−1) SST κ-ω
Continuous flow (F = 60, 120, FLUENT NS equations 7−100 0.005−0.07 0−3 [106]
240, 360, 480 and SST κ-ω
660 mL min−1)
Continuous flow FLUENT NS equations 32 0.022 0.04 [49]
(F = 240 mL min−1) SST κ-ω
Continuous flow COMSOL 3D NS − 7−43 0.42−2.5 (with 0.42 [107]
(F = 50–300 mL min−1 with Laminar m s−1 step size)
50 mL min−1 steps)
Continuous flow (F = 3 and FLUENT 3D NS − 0.05−1.47 − [108]
5 mL min−1) Laminar
Continuous flow FLUENT 3D NS − < 2.4 − [109]
(F ≈ 0−12 mL min−1) Laminar
(Continued)
Biofilm Shear Rate
Parallel plate flow Continuous flow (F = 4 and FLUENT 3D NS − 0.02−1.74 − [110]
chamber with 5.5 mL min−1) Laminar
microstructured
polymer surfaces
Parallel-plate Continuous flow FIDAP 2D NS 420−32,000 1.3−96 − [111]
perfusion (F = 10 mL min−1) laminar
chambers
Microfluidics Rectangular PDMS Continuous flow FLUENT 3D NS 32 0.022 0.04 [70]

microchannel (F = 0.0012 mL min−1) Laminar
Continuous flow FLUENT 3D NS − 0−0.182 5.07 × 10−11−0.0036 [112]
(F = 0.001 mL min−1) Laminar
Rectangular Continuous flow FLUENT 3D NS − 0.01−0.25 3 × 10−4−0.005 [113]
patterned PDMS (F = 0.0035 mL min−1) Laminar
microchannel
Rectangular PDMS Continuous flow CFX 3D NS 0.1−654 1.98×10−4−0.9 0−0.0055 [114]
Operation of Biofilm Reactors for the Food Industry Using CFD
with microwell
arrays
PDMS Continuous flow FLUENT 3D NS − 0.03−4.30 0.37−1.63 [115]
with four
microchambers
Diffusion-based Continuous flow FLUENT 3D NS − − 0−0.003 [116]
PDMS (F = 2.8 × 10−4 mL min−1) Laminar
microchannel
PDMS funnel Continuous flow FLUENT 3D NS 0−6820 0.0032−0.029 − [117]
device (F = 0.005−0.05 mL min−1) Laminar
(Continued)
573
574

Biofilm Shear Rate
PDMS chamber Continuous flow (F = 0.005 FLUENT 3D NS − −4
7 × 10 −0.6 −5
8 × 10 −0.0014 [118]
with microholes, and 0.035 mL min−1) Laminar
microchambers
and microchannels
PDMS chamber Continuous flow FLUENT 3D NS − 0.00246−0.00833 7.22 × 10−6− [119]
with microstamps (F = 0.01 mL min−1) Laminar 2.71 × 10−4
Two consecutive F = 80 mL min−1 COMSOL 3D NS − 100−400 8.9 [120]
arrays of parallel Laminar
micronozzles
Rectilinear and F = 0.0017 mL min−1 FLUENT 3D NS − 0−0.441 0−0.00136 [121]
funnel shaped Laminar
inlet and outlet coupled with
regions a Brownian/
Stokesian
bacteria
transport
model
Microfluidic chips F = 0.0093 mL min−1 FLUENT 3D NS − 0−0.07 0−1.49 [122]
with two Laminar
different SCRSs

Notes: (a) RNG k-ɛ turbulence transport equations (the RNG k-ɛ model is based on renormalization group analysis of the Navier–Stokes equations); * - simulated at the bottom; ** - considering
the whole wetted area.
FIGURE 25.6 Shear stress magnitudes (Pa) on the bottom of 12-well MTPs with a liquid volume of 3 mL
shaken at (a) 40, (b) 100, (c) 140 and (d) 180 rpm (at 25°C and with an orbital diameter of 25 mm).
are reduced. Large coupons can be used for biofilm formation (square surfaces of up to 1.5 cm can
be placed on the bottom of the wells) and uniform shear forces can be obtained. Even though the
shear stress in the coupon will vary with the radial distance to the center, each coupon will have
identical average shear stress values. Furthermore, the shear stress in each spot of the coupon is not
constant but is periodic (the period is dictated by the shaking frequency), which enables biofilm
formation in a range of shear stresses in the same experiment.
We have simulated 12-well MTPs to assess the effect of shaking frequency on shear stress using
ANSYS Fluent, a grid with 40,441 hexahedral cells and the VOF methodology. Simulations were
performed at 25°C, with an orbit of 25 mm, a liquid volume of 3 mL and shaking frequencies of
40, 100, 140 and 180 rpm (Figure 25.6). As expected, higher shear stress values at the bottom of the
wells were obtained at higher shaking frequencies and values up to 0.07 Pa and shear rates of 42 s−1
were obtained (Table 25.4). These values are much higher than those obtained with 96-well MTPs
and are already within the range of several industrial applications (Tables 25.2 and 25.3).
25.4.1.5 6-Well MTPs
Using 6-well MTPs is advantageous due to the relatively high shear forces that can be obtained with
this platform (Table 25.4). Unlike other MTP formats, these plates do not enable high throughput
screening, but allow the use of large coupons that can be placed on the bottom of the wells,
so that different materials can be tested. This is particularly important if biochemical analysis of
the biofilm is necessary due to the requirement for a relatively large biofilm mass. If using trans-
parent MTPs, initial adhesion can be directly observed on the plates after gentle washing using an
inverted microscope. Since the liquid volume is larger, these plates also sustain bacterial growth for
longer time periods, although medium exchange is required in that case.
This format has been previously used by Azevedo et al. [52] to study the adhesion of Helicobacter
pylori to stainless steel and PP, and by Kostenko et al. [53] and Salek et al. [54] to evaluate the bio-
film formation by Staphylococcus aureus. These last authors have simulated the shear stress in dif-
ferent agitation conditions (shaking frequencies of 100 and 200 rpm) using different liquid volumes
(2 and 4 mL) and they reached average shear stresses on the bottom around 1 Pa (Figure 25.7).
Using ANSYS Fluent, we have also simulated the hydrodynamics inside the wells of 6-well
MTPs using a grid with 62,748 hexahedral cells and the VOF methodology to assess the effect of
temperature and liquid volume on shear stress. Many operations in the food industry occur at ambi-
ent temperature or below and therefore we have simulated agitated MTPs (at a shaking frequency
of 160 rpm and an orbital radius of 4 mm) at 5 and 25°C (Figure 25.8). It was found that higher
shear stresses can be obtained at the lowest temperature (Table 25.4). As expected, the temperature
effect is more noticeable on the shear stress (60% increase was obtained at the lowest temperature)
than on the shear rate (similar values were obtained), reflecting the effect of temperature on fluid
(a) 1
(b) 3.80
0.93 3.54
0.86 3.29
0.79 3.03
0.72 2.77
0.65 1 2.51
0.58 2.26
0.51 2.00
0.44 1.74
0.37 1.49
0.3 1.23
0.23 0.97
0.16 0.71
0.09 0.46
0.02 0.20
x x
z z
(c) 1 (d) 3.80

0.93 3.54
0.86 3.29
0.79 3.03
0.72
0.65
1 2.77
2.51
0.58 2.26
0.51 2.00
0.44 1.74
0.37 1.49
0.3 1.23
0.23 0.97
0.16 0.71
0.09 0.46
0.02 0.20
x x
z z
FIGURE 25.7 Shear stress (Pa) on the bottom of 6-well MTPs with surface ‘‘streamlines’’ superimposed
at t = t0 for (a) 100 rpm, 4 mL; (b) 200 rpm, 4 mL; (c) 100 rpm, 2 mL; and (d) 200 rpm, 2 mL. Curved arrow
indicates sense of rotation. (Reprinted from Salek, M.M., Sattari, P., Martinuzzi, R.J., Analysis of Fluid Flow
and Wall Shear Stress Patterns Inside Partially Filled Agitated Culture Well Plates. Annals of Biomedical
Engineering, 2012. 40(3): p. 707–728.)
FIGURE 25.8 Shear stress magnitudes (Pa) on the bottom of 6-well MTPs with 4 mL at (a) 5 and (b) 25°C
(orbital diameter of 4 mm and shaking frequency of 160 rpm).
FIGURE 25.9 Shear stress magnitudes (Pa) on the bottom of 6-well MTPs at 25°C containing (a) 4 and
(b) 8 mL (orbital diameter of 4 mm and shaking frequency of 160 rpm).
viscosity. The effect of liquid volume is also significant, as it can be observed in Figure 25.9 for
wells filled with 4 or 8 mL. Wells filled with a liquid volume of 8 mL have much lower shear stresses
and shear rate values, and an approximate 4-fold difference was obtained for both parameters in
these situations. This is explained by the increased distance of the air-liquid interface from the
bottom of the well for higher liquid volumes. As the liquid height is decreased, higher shear forces
can be obtained on the bottom of the well and this further increases the range of shear stresses and
shear rate values that can be obtained with this platform for the simulation of different industrial
scenarios (Tables 25.2 and 25.3).
25.4.2 Flow Cells
Flows cells are usually integrated in a system with other components that can be operated in a
number of ways [55]. The most straightforward configuration is that of a bioreactor containing a
batch culture of the desired organism so that the contents of the reactor are pumped through the flow
cell and the effluent drained to waste. This configuration is adequate for initial adhesion studies,
particularly if the flow rates to be tested are relatively low, as the duration of the assay is limited by
the culture volume contained in the bioreactor. An alternative configuration is to place the flow cell
in a recycle loop so that the culture volume is no longer a limitation and assays can last longer and
be performed at higher flow rates (i.e., at increased shear forces). However, the composition of the
batch culture is always changing and this is often undesirable. A third option, which circumvents
this problem, is to use the effluent of a chemostat to feed the flow cell. In this case, the composi-
tion of the culture remains constant and the experimental time can be extended. Nevertheless, the
flow rate that is used is dependent on the dilution rate of the chemostat, which reduces the range
of hydrodynamic conditions that can be tested. A fourth option is to have the chemostat feeding a
recirculation loop so that the nutrient feed flow to the chemostat equals the drain flow from the loop
(Figure 25.10). In this scenario, it is possible to feed the flow cell with a suspension with a constant
concentration of cells and nutrients while decoupling the flow rate going through the flow cell from
the dilution rate in the bioreactor. Thus, it is possible to use very high recirculation flow rates and
attain higher shear stresses that are comparable to those found in some food industries (Table 25.3).
Using very high recirculation rates also avoids the formation of chemical gradients in the system
FIGURE 25.10 Schematic representation of the biofilm producing system comprising a recirculating tank,
one semi-circular flow cell (with ten Perspex coupons), and peristaltic and centrifugal pumps. Five different
zones are identified in the flow cell: inlet duct, entry zone, sampling zone, exit zone and outflow duct.
and minimizes residence times. Therefore, these systems can be assumed as completely mixed and
are considered “chemostats with irregular geometries” [56].
Flow cells can be generally divided into two categories: those that are based on the design of the
Robbins device and those that are built for direct inspection of biofilm development (often called
flow chambers). In both types of flow cells, it is possible to use coupons so that different materials
can be tested.
25.4.2.1 Robbins Device and Modifications

The Robbins device was designed by Jim Robbins and Bill McCoy and was originally used to moni-
tor biofilm formation in a simulated drinking water system [57]. Several modifications were later
introduced to this design, including the use of a square channel pipe containing sample ports that
do not disturb the flow [56]. With this device, the flow can be momentarily stopped to allow direct
access to the coupon so that time-course experiments are also possible [58]. In order to better simu-
late the geometry of piping systems, a semi-circular geometry was later introduced [59].
These flow cells can be operated either in laminar or turbulent regimes [59] but it is important
to guarantee that the flow cell has an entry section which is long enough to allow flow development
before the sampling zone (Figure 25.10). This will ensure that all coupons are subjected to the same
hydrodynamic conditions and that samples can be directly compared [55]. Using CFD, we have
simulated the hydrodynamics inside a semi-circular flow cell and concluded that the shear stress
field was similar in the curved and flat walls in both laminar and turbulent conditions (Figure 25.11)
[60]. With this information, it was possible to identify which coupons could be used in different flow
conditions to ensure a constant shear stress field across the whole sampling area [60].
Using the Darcy friction factor [17], we investigated if a semi-circular flow cell could emulate the
flow in a tube at different Reynolds numbers (Re) since industrial piping for fluid transport usually
has a circular section. The friction factor (f) and the Reynolds number are defined as (Equations
25.1 and 25.2, respectively).
4τ w
f= (25.1)

ρU 2/ 2
ρUDh
Re = (25.2)
µ
where ρ and μ are the density and dynamic viscosity of the fluid, respectively, U is the average
velocity in the flow cell, τw is the wall shear stress and Dh is the hydraulic diameter of the semi-
circular flow cell of diameter D (Equation 25.3).
Dh = π D /(2 + π ) (25.3)
By plotting the friction factor as a function of the Reynolds number (Figure 25.12), we observe
that in the laminar region the behavior is similar in both geometries and agrees with the theoreti-
cal prediction for a smooth pipe ( f = 64/Re). Although some minor deviations are observed in
turbulent flow conditions, the values that were obtained also show a good agreement between
both geometries and with the theoretically predicted values using the Blausius equation for a
smooth pipe [17]. With this analysis, we have demonstrated that a semi-circular flow cell can
efficiently emulate the flow in a pipe with circular section in laminar conditions (for Reynolds
numbers between 10 and 1,000) and also in turbulent flow (for Reynolds numbers between 3,500
and 10,000). Furthermore, using Equation 25.1 and Figure 25.12, it is possible to calculate which
operational conditions should be used in a semi-circular flow cell in order to emulate the flow in
FIGURE 25.11 Normalized wall shear stress contour plots for (a) laminar and (b) turbulent flows in the flat
and curved walls of the semi-circular flow cell represented in Figure 25.10. (Adapted from Teodósio, J.S.,
Silva, F.C., Moreira, J.M.R., Simões, M., Melo, L.F., Alves, M.A., Mergulhão, F.J., Flow cells as quasi-ideal
systems for biofouling simulation of industrial piping systems. Biofouling, 2013. 29(8): p. 953–966.)
piping systems. Our research group has operated these flow cells in different conditions and was
able to attain shear stress values up to 0.6 Pa during Escherichia coli biofilm formation [58,61],
confirming the versatility of this platform and its capacity to mimic the hydrodynamic conditions
that can be found in food industry (Table 25.3).
25.4.2.2 Flow Chambers for Direct Inspection of Biofilms

Despite the many advantages of the described flow cells, they are usually not amenable to direct
analysis of biofilm development. Additionally, they are not sensitive enough to monitor initial cell
adhesion to a surface. For these purposes, flow chambers have been developed and they are com-
monly used with a sampling probe (most often a microscope lens or a microelectrode). Flow cham-
bers can be divided in two different categories: opened or closed type systems. In opened systems,
the probe is placed directly in the liquid surrounding the biomass. This can affect the hydrody-
namics and the system is prone to contamination [44]. In closed systems (Figure 25.13), inspection
windows allow direct observation of biofilm development without contamination issues and in con-
trolled hydrodynamics.
FIGURE 25.12 The friction factor as a function of the Re for the semi-circular flow cell and tube. Comparison
with the theoretical behavior of the friction factor in a circular pipe, represented by the expressions of laminar
flow and the Blasius equation for turbulent flow (smooth pipe). (Reprinted from Teodósio, J.S., Silva, F.C.,
Moreira, J.M.R., Simões, M., Melo, L.F., Alves, M.A., and Mergulhão, F.J., Flow cells as quasi-ideal systems
for biofouling simulation of industrial piping systems. Biofouling, 2013. 29(8): p. 953–966.)
FIGURE 25.13 Detailed view of a shear stress flow chamber (a) before and (b) after assembling. (Adapted
from Mercier-Bonin, M., Dehouche, A., Morchain, J., Schmitz, P., Orientation and detachment dynamics of
Bacillus spores from stainless steel under controlled shear flow. International Journal of Food Microbiology,
2011. 146(2): p. 182–191.)
These flow chambers can be operated in turbulent or laminar flow conditions but, depending on
the dimensions of the system, biofilms can be vulnerable to the passage of air bubbles, which induce
a significant shear stress and may cause biofilm detachment. To alleviate this problem, bubble traps
are often included in the system [44].
We have previously used these flow chambers to study Escherichia coli attachment to different
surfaces [62] and have simulated the hydrodynamics inside them to verify flow stabilization (Figure
25.14). By performing this analysis, we can better define the shear stress in the observation win-
dow, thus ensuring higher reproducibility between experiments. The numerical results showed that
FIGURE 25.14 Wall shear stress on the bottom wall of the parallel plate flow cell. (Reprinted from Moreira,
J.M.R., Araújo, J.D.P., Miranda, J.M., Simões, M., Melo, L.F., Mergulhão, F.J., The effects of surface properties on
Escherichia coli adhesion are modulated by shear stress. Colloids and Surfaces B: Biointerfaces, 2015. 123, p. 1–7.)
despite the existence of a flow instability near the entry of the flow cell, the flow stabilizes along the
length and a stable shear stress field is obtained at the viewing point.
Different groups have used these flow cells in shear rate/shear stress intervals of great amplitude,
covering a huge range of shear values that can be found in the food industry (Tables 25.2 and 25.3).
One group in particular was able to attain shear stress values of up to 237 Pa while studying the
detachment of Bacillus spores from stainless steel [63]. Despite their dynamic versatility, one must
bear in mind that these platforms have much lower throughput than MTPs and larger flow cells
based on the Robbins device. Also the biofilm mass they produce is often not sufficient for subse-
quent biochemical analysis.
25.4.3 Microfluidic Devices
Microfluidic devices handle fluids that are generally confined to small environments, usually at
the sub-millimeter scale [64]. Compared with traditional flow cell systems, microfluidics enable a
FIGURE 25.15 (a) Optical image, (b) designs and (c) shear rate distribution along a microfluidic funnel
device for a flow rate of 50 μl min−1. (Adapted from Shumi, W., Kim, S.H., Lim, J., Cho, K.-S., Han, H.,
Park, S., Journal of Microbiological Methods, 93(2), 85–89.)
greater control over flow conditions, can be used to explore a much wider range of shear rates with
great flexibility in designing flow geometries of interest, and enable the parallelization of experi-
ments (Figure 25.15). Microfluidics also allow testing multiple flow conditions at the same time,
with the same culture and on the same chip, which is fundamental to overcome the unavoidable
biological variability [65].
Microfluidic devices can be fabricated from a range of materials using different methods, such as
photolithography and wet etching methods [66]. Most microfluidic devices are fabricated from glass
[67], duroplastic or thermoplastic materials [68], and from the flexible elastomer polydimethylsilox-
ane (PDMS) [69]. Several other surfaces can be studied using xerographic construction techniques
that enable different polymers to be incorporated into microfluidic flow cells [70]. Introduction of
functional groups can also be accomplished by different methods, including organosilanes deposi-
tion [71,72]. Several detection methods have been developed for microfluidic devices and although
off-chip detection with conventional methods is feasible (usually limited by the sample volume that
can be collected), on-chip detection is favored to observe in situ and real time effects [44]. Optical
detection methods are often employed [73,74], in particular fluorescence [75].
These platforms also have their limitations: the minute liquid volumes used in microfluidics may
preclude further biochemical or genetic analysis, and the spatial confinement [65] may generate dif-
ferent biofilms from those formed in more opened structures like process equipment. Furthermore,
these platforms require special equipment for manufacturing and operation, and system clogging
can occur due to the small dimensions [44].
Microfluidic devices have been intensively used to study biofilm development at a huge range
of shear conditions. Shear rate values up to 6820 s–1 and shear stresses up to 400 Pa were obtained
with these systems (Table 25.4).
25.5 CONCLUSIONS
The importance of CFD simulation has been growing in the last three decades both in the number
and range of applications. This growth was fueled by the increase in computational power and
speed, together with significant developments in numerical methods. CFD simulation provides a
spatial and temporal understanding of general flow phenomena that are difficult to analyze experi-
mentally. Moreover, CFD enables a quick assessment of different alternatives and changes with low
commitment of time and resources.
The use of CFD tools is particularly useful to obtain detailed information about the hydrody-
namics occurring inside biofilm reactors (sometimes with complex geometries) allowing the pre-
diction of different flow variables, such as the shear rates and shear stresses exerted on the walls.
This information is crucial to understanding the influence of flow behavior on the development and
resistance of biofilms.
Despite these advantages, CFD simulations mainly rely on physical models of real situations and
therefore the accuracy of the obtained simulations is limited by the accuracy of the models that are
used. Also, different assumptions and simplifications are used to reduce complexity and computing
time. Since CFD is often used to simulate very complex scenarios, the models are likely to suffer
from inaccuracies when compared to experimental data and this requires physical validation of the
simulation data.
NOMENCLATURE
D Diameter (m)
Dh Hydraulic diameter of the semi-circular flow cell (m)
f Friction factor
F Flow rate (m3 s−1)
Re Reynolds number
U Velocity (m s−1)
Greek Symbols
μ Dynamic viscosity (kg m−1 s−1)
ρ Density (kg m−3)
τw Wall shear stress (Pa)
υ Shaking frequency (rpm)
ϕ Orbital diameter (m)
Acronyms
CIP Cleaning in place
LES Large eddy simulation
MTP Microtiter plate
NS Navier–Stokes
PDMS Polydimethylsiloxane
PP Polypropylene
PS Polystyrene
RANS Reynolds-averaged Navier–Stokes
SCRS Single-cell retention structure
SST Shear-stress transport
VOF Volume-of-fluid model
REFERENCES
1. Agle M.E., Biofilms in the food industry, In: Biofilms in the Food Environment, Blaschek H.P., Wang H.H.,
and Agle M.E., Eds., Blackwell Publishing Ltd., Oxford, UK, 3–17, 2007.
2. Costerton J.W., Geesey G.G., Cheng K.J., How bacteria stick, Scientific American, 238(1): 86–95, 1978.
3. Costerton J.W., Stewart P.S., Greenberg E.P., Bacterial biofilms: A common cause of persistent infec-
tions, Science, 284(5418): 1318–1322, 1999.
4. Branda S.S., Vik Å., Friedman L., Kolter R., Biofilms: The matrix revisited, Trends in Microbiology,
13(1): 20–26, 2005.
5. Sutherland I.W., The biofilm matrix—An immobilized but dynamic microbial environment, Trends in
Microbiology, 9(5): 222–227, 2001.
6. Nikolaev Y.A., Plakunov V.K., Biofilm—“City of microbes” or an analogue of multicellular organisms?,
Microbiology, 76(2): 125–138, 2007.
7. Petrova O.E., Sauer K., Sticky situations: Key components that control bacterial surface attachment,
Journal of Bacteriology, 194(10): 2413–2425, 2012.
8. Giaouris E., Heir E., Hébraud M., Chorianopoulos N., Langsrud S., Møretrø T., Habimana O.,
Desvaux M., Renier S., Nychas G.J. Attachment and biofilm formation by food-borne bacteria in meat
processing environments: Causes, implications, role of bacterial interactions and control by alternative
novel methods, Meat Science, 97(3): 298–309, 2014.
9. Esbelin J., Santos T., Hébraud M., Desiccation: An environmental and food industry stress that bacteria
commonly face, Food Microbiology, 69: 82–88, 2018.
10. Srey S., Jahid I.K., Ha S.-D., Biofilm formation in food industries: A food safety concern, Food Control,
31(2): 572–585, 2013.
11. Stanga M., Sanitation: Cleaning and Disinfection in the Food Industry, Wiley-VCH, Weinheim,
Germany, 2010.
12. Martinuzzi R.J., Salek M.M., Numerical simulation of fluid flow and hydrodynamic analysis in
commonly used biomedical devices in biofilm studies, In: Numerical Simulations—Examples and
Applications in Computational Fluid Dynamics, Angermann L., Ed., IntechOpen Ltd., London, UK,
193–212, 2010.
13. Gjersing E.L., Codd S.L., Seymour J.D., Stewart P.S., Magnetic resonance microscopy analysis of
advective transport in a biofilm reactor, Biotechnology and Bioengineering, 89(7): 822–834, 2005.
14. Liu Y., Tay J.-H., The essential role of hydrodynamic shear force in the formation of biofilm and granu-
lar sludge, Water Research, 36(7): 1653–1665, 2002.
15. Stewart P.S., A review of experimental measurements of effective diffusive permeabilities and effective
diffusion coefficients in biofilms, Biotechnology and Bioengineering, 59(3): 261–272, 1998.
16. Stewart P.S., Diffusion in biofilms, Journal of Bacteriology, 185(5): 1485–1491, 2003.
17. Munson B.R., Young D.F., Okiishi T.H., Fundamentals of Fluid Mechanics, John Wiley & Sons,
Hoboken, NJ, USA, 2002.
18. Sharma C., Malhotra D., Rathore A.S., Review of computational fluid dynamics applications in biotech-
nology processes, Biotechnology Progress, 27(6): 1497–1510, 2011.
19. Werner S., Kaiser S.C., Kraume M., Eibl D., Computational fluid dynamics as a modern tool for engi-
neering characterization of bioreactors, Pharmaceutical Bioprocessing, 2(1): 85–89, 2014.
20. Kelly W.J., Using computational fluid dynamics to characterize and improve bioreactor performance,
Biotechnology and Applied Biochemistry, 49(4): 225–238, 2008.
21. Menter F.R., Two-equation eddy-viscosity turbulence models for engineering applications, AIAA Journal,
32(8): 1598–1605, 1994.
22. Hirt C.W., Nichols B.D., Volume of fluid (VOF) method for the dynamics of free boundaries, Journal of
Computational Physics, 39(1): 201–225, 1981.
23. Brackbill J.U., Kothe D.B., Zemach C., A continuum method for modeling surface tension, Journal of
Computational Physics, 100(2): 335–354, 1992.
24. Youngs D.L., Time-dependent multi-material flow with large fluid distortion, In: Numerical Methods
for Fluid Dynamics, Morton K.W. and Baibnes M.J., Eds., Academic Press, Cambridge, MA, USA,
273–285, 1982.
25. Moreira J.M.R., Fulgêncio R., Alves P., Machado I. et al., Evaluation of SICAN performance for
bio-fouling mitigation in the food industry, Food Control, 62: 201–207, 2016.
26. Salek M.M., Jones S.M., Martinuzzi R.J., The influence of flow cell geometry related shear stresses on
the distribution, structure and susceptibility of Pseudomonas aeruginosa 01 biofilms, Biofouling, 25(8):
711–725, 2009.
27. Picioreanu C., van Loosdrecht M.C., Heijnen J.J., A theoretical study on the effect of surface roughness
on mass transport and transformation in biofilms, Biotechnology and Bioengineering, 68(4): 355–69,
2000.
28. Picioreanu C., Van Loosdrecht M.C.M., Heijnen J.J., Effect of diffusive and convective substrate
transport on biofilm structure formation: A two-dimensional modeling study, Biotechnology and
Bioengineering, 69(5): 504–515, 2000.
29. Klapper I., Rupp C.J., Cargo R., Purvedorj B. et al., Viscoelastic fluid description of bacterial biofilm
material properties, Biotechnology and Bioengineering, 80(3): 289–96, 2002.
30. Characklis W.G., Bioengineering report: Fouling biofilm development: A process analysis, Biotechnology
and Bioengineering, 23(9): 1923–1960, 1981.
31. Simões M., Mergulhão F.J., Biofilms in Bioengineering, Nova Science Publishers, New York, USA, 2013.
32. Fletcher M., The effects of culture concentration and age, time, and temperature on bacterial attachment
to polystyrene, Canadian Journal of Microbiology, 23(1): 1–6, 1977.
33. O’Toole G.A., Kolter R., Flagellar and twitching motility are necessary for Pseudomonas aeruginosa
biofilm development, Molecular Microbiology, 30(2): 295–304, 1998.
34. Kumar C.G., Anand S.K., Significance of microbial biofilms in food industry: A review, International
Journal of Food Microbiology, 42(1): 9–27, 1998.
35. Tauriainen S., Karp M., Chang W., Virta M., Luminescent bacterial sensor for cadmium and lead,
Biosensors and Bioelectronics, 13(9): 931–8, 1998.
36. Tani C., Inoue K., Tani Y., Harun-ur-Rashid M. et al., Sensitive fluorescent microplate bioassay using
recombinant Escherichia coli with multiple promoter-reporter units in tandem for detection of arsenic,
Journal of Bioscience and Bioengineering, 108(5): 414–420, 2009.
37. Fukazawa T., Yamazaki Y., Miyamoto Y., Reduction of non-specific adsorption of drugs to plastic con-
tainers used in bioassays or analyses, Journal of Pharmacological and Toxicological Methods, 61(3):
329–33, 2010.
38. Moreira J.M.R., Gomes L.C., Araújo J.D.P., Miranda J.M. et al., The effect of glucose concentration and
shaking conditions on Escherichia coli biofilm formation in microtiter plates, Chemical Engineering
Science, 94: 192–199, 2013.
39. Djordjevic D., Wiedmann M., McLandsborough L.A., Microtiter plate assay for assessment of Listeria
monocytogenes biofilm formation, Applied and Environmental Microbiology, 68(6): 2950–2958, 2002.
40. Gomes L.C., Moreira J.M.R., Simões M., Melo L.F. et al., Biofilm localization in the vertical wall of
shaking 96-well plates, Scientifica, 2014(6): 2014.
41. Kumar S., Wittmann C., Heinzle E., Minibioreactors, Biotechnology Letters, 26(1): 1–10, 2004.
42. Stepanović S., Vuković D., Hola V., Di Bonaventura G. et al., Quantification of biofilm in microtiter
plates: Overview of testing conditions and practical recommendations for assessment of biofilm produc-
tion by staphylococci, APMIS, 115(8): 891–9, 2007.
43. Borges A., Lopez-Romero J.C., Oliveira D., Giaouris E. et al., Prevention, removal and inactivation of
Escherichia coli and Staphylococcus aureus biofilms using selected monoterpenes of essential oils,
Journal of Applied Microbiology, 123(1): 104–115, 2017.
44. Azeredo J., Azevedo N.F., Briandet R., Cerca N. et al., Critical review on biofilm methods, Critical
Reviews in Microbiology, 43(3): 313–351, 2017.
45. Gomes L.C., Silva L.N., Simões M., Melo L.F. et al., Escherichia coli adhesion, biofilm development
and antibiotic susceptibility on biomedical materials, Journal of Biomedical Materials Research
Part A, 103(4): 1414–23, 2015.
46. Moreira J.M.R., Gomes L.C., Simões M., Melo L.F. et al., The impact of material properties, nutrient
load and shear stress on biofouling in food industries, Food and Bioproducts Processing, 95: 228–236,
2015.
47. Gomes L.C., Moreira J.M.R., Simões M., Melo L.F. et al., Biofilm localization in the vertical wall of
shaking 96-well plates, Scientifica, 2014, article 231083, 2014.
48. Gomes L.C., Moreira J.M.R., Miranda J.M., Simões M. et al., Macroscale versus microscale methods
for physiological analysis of biofilms formed in 96-well microtiter plates, Journal of Microbiological
Methods, 95(3): 342–349, 2013.
49. Gomes L.C., Moreira J.M.R., Araújo J.D.P., Mergulhão F.J., Surface conditioning with Escherichia coli
cell wall components can reduce biofilm formation by decreasing initial adhesion, AIMS Microbiology,
3(3): 613–628, 2017.
50. Lv X., Song J., Yu B., Liu H. et al., High-throughput system for screening of high l-lactic acid-produc-
tivity strains in deep-well microtiter plates, Bioprocess and Biosystems Engineering, 39(11): 1737–1747,
2016.
51. Barrett T.A., Wu A., Zhang H., Levy M.S. et al., Microwell engineering characterization for mamma-
lian cell culture process development, Biotechnology and Bioengineering, 105(2): 260–275, 2010.
52. Azevedo N.F., Pinto A.R., Reis N.M., Vieira M.J. et al., Shear stress, temperature, and inoculation
concentration influence the adhesion of water-stressed Helicobacter pylori to stainless steel 304 and
polypropylene, Applied and Environmental Microbiology, 72(4): 2936–2941, 2006.
53. Kostenko V., Salek M.M., Sattari P., Martinuzzi R.J., Staphylococcus aureus biofilm formation
and tolerance to antibiotics in response to oscillatory shear stresses of physiological levels, FEMS
Immunology & Medical Microbiology, 59(3): 421–431, 2010.
54. Salek M.M., Sattari P., Martinuzzi R.J., Analysis of fluid flow and wall shear stress patterns inside par-
tially filled agitated culture well plates, Annals of Biomedical Engineering, 40(3): 707–728, 2012.
55. Teodósio J.S., Simões M., Mergulhão F.J., Platforms for in-vitro biofilm studies, In: Biofilms in Bioengineering,
Simões M. and Mergulhão F.J., Eds., Nova Science Publishers, New York, USA, 45–62, 2013.
56. Stoodley P., Warwood B.K., Use of flow cells an annular reactors to study biofilms, In: Biofilms in
Medicine, Industry and Environmental Biotechnology: Characteristics, Analysis and Control, Lens P.
et al., Eds., IWA Publishing, Cornwall, UK, 197–213, 2003.
57. McCoy W.F., Bryers J.D., Robbins J., Costerton J.W., Observations of fouling biofilm formation,
Canadian Journal of Microbiology, 27(9): 910–917, 1981.
58. Teodósio J.S., Simões M., Melo L.F., Mergulhão F.J., Flow cell hydrodynamics and their effects on E. coli
biofilm formation under different nutrient conditions and turbulent flow, Biofouling, 27(1): 1–11, 2011.
59. Pereira M.O., Kuehn M., Wuertz S., Neu T. et al., Effect of flow regime on the architecture of a
Pseudomonas fluorescens biofilm, Biotechnology and Bioengineering, 78(2): 164–171, 2002.
60. Teodósio J.S., Silva F.C., Moreira J.M.R., Simões M. et al., Flow cells as quasi-ideal systems for biofoul-
ing simulation of industrial piping systems, Biofouling, 29(8): 953–966, 2013.
61. Teodósio J.S., Simões M., Alves M.A., Melo L.F. et al., Setup and validation of flow cell systems for
biofouling simulation in industrial settings, The Scientific World Journal, 2012(10): 2012.
62. Moreira J.M.R., Simões M., Melo L.F., Mergulhão F.J., Escherichia coli adhesion to surfaces—A ther-
modynamic assessment, Colloid and Polymer Science, 293(1): 177–185, 2015.
63. Mercier-Bonin M., Dehouche A., Morchain J., Schmitz P., Orientation and detachment dynamics of
Bacillus spores from stainless steel under controlled shear flow: Modelling of the adhesion force,
International Journal of Food Microbiology, 146(2): 182–191, 2011.
64. Kim J., Park H.D., Chung S., Microfluidic approaches to bacterial biofilm formation, Molecules, 17(8):
9818–34, 2012.
65. Yawata Y., Nguyen J., Stocker R., Rusconi R., Microfluidic studies of biofilm formation in dynamic
environments, Journal of Bacteriology, 198(19): 2589–2595, 2016.
66. Madou M.J., Fundamentals of Microfabrication and Nanotechnology, CRC Press, Boca Raton, FL,
USA, 2011.
67. Iliescu C., Taylor H., Avram M., Miao J. et al., A practical guide for the fabrication of microfluidic
devices using glass and silicon, Biomicrofluidics, 6(1): 16505–1650516, 2012.
68. Becker H., Gärtner C., Polymer microfabrication technologies for microfluidic systems, Analytical and
Bioanalytical Chemistry, 390(1): 89–111, 2008.
69. McDonald J.C., Duffy D.C., Anderson J.R., Chiu D.T. et al., Fabrication of microfluidic systems in
poly(dimethylsiloxane), Electrophoresis, 21(1): 27–40, 2000.
70. Moreira J.M.R., Ponmozhi J., Campos J.B.L.M., Miranda J.M. et al., Micro- and macro-flow systems to
study Escherichia coli adhesion to biomedical materials, Chemical Engineering Science, 126: 440–445,
2015.
71. Pallandre A., de Lambert B., Attia R., Jonas A.M. et al., Surface treatment and characterization:
Perspectives to electrophoresis and lab-on-chips, Electrophoresis, 27(3): 584–610, 2006.
72. Glass N.R., Tjeung R., Chan P., Yeo L.Y. et al., Organosilane deposition for microfluidic applications,
Biomicrofluidics, 5(3): 036501–036501–7, 2011.
73. Mariana T.M., Varnika R., William E.B., Reza G., Development and validation of a microfluidic
reactor for biofilm monitoring via optical methods, Journal of Micromechanics and Microengineering,
21(5): 054023, 2011.
74. Zhu H., Isikman S.O., Mudanyali O., Greenbaum A. et al., Optical imaging techniques for point-of-care
diagnostics, Lab on a Chip, 13(1): 51–67, 2013.
75. Johnson M.E., Landers J.P., Fundamentals and practice for ultrasensitive laser-induced fluorescence
detection in microanalytical systems, Electrophoresis, 25: 21–22, 3513–3527, 2004.
76. Fratamico P.M., Annous B.A., Gunther N.W., Biofilms in the Food and Beverage Industries, Woodhead
Publishing, Sawston, Cambridgeshire, UK, 2009.
77. Teh K.H., Flint S., Brooks J., Knight G., Biofilms in the Dairy Industry, John Wiley & Sons, Hoboken,
NJ, USA, 2015.
78. Pawar S.B., Mujumdar A.S., Thorat B.N., CFD analysis of flow pattern in the agitated thin film
evaporator, Chemical Engineering Research and Design, 90(6): 757–765, 2012.
79. Chesterton A.K.S., Moggridge G.D., Sadd P.A., Wilson D.I., Modelling of shear rate distribution in two
planetary mixtures for studying development of cake batter structure, Journal of Food Engineering,
105(2): 343–350, 2011.
80. Emin M.A., Schuchmann H.P., Analysis of the dispersive mixing efficiency in a twin-screw extrusion
processing of starch based matrix, Journal of Food Engineering, 115(1): 132–143, 2013.
81. Kokini J.L., Dhanasekharan M., Ashokan B.K., Use of CFD for optimization, design, and scale-up of
food extrusion, In: Computational Fluid Dynamics in Food Processing, Sun D.-W., Ed., CRC Press, Boca
Raton, FL, USA, 2007.
82. Pakzad L., Ein-Mozaffari F., Upreti S.R., Lohi A., Evaluation of the mixing of non-Newtonian biopoly-
mer solutions in the reactors equipped with the coaxial mixers through tomography and CFD, Chemical
Engineering Journal, 215: 279–296, 2013.
83. Bourne M.C., Food Texture and Viscosity: Concept and Measurement, Academic Press, Cambridge,
MA, USA, 2002.
84. Yataghene M., Pruvost J., Fayolle F., Legrand J., CFD analysis of the flow pattern and local shear rate in
a scraped surface heat exchanger, Chemical Engineering and Processing: Process Intensification, 47(9):
1550–1561, 2008.
85. Fernandes C.S., Nóbrega J.M., Maia J.M., Dias R.P., CFD simulation of stirred yoghurt processing in
plate heat exchangers, In: Computational Fluid Dynamics in Food Processing, Sun D.-W., Ed., CRC
Press, Boca Raton, FL, USA, 2007.
86. Stapley A.G.F., Tewkesbury H., Fryer P.J., The effects of shear and temperature history on the crystal-
lization of chocolate, Journal of the American Oil Chemists’ Society, 76(6): 677–685, 1999.
87. Dumont E., Fayolle F., Legrand J., Flow regimes and wall shear rates determination within a scraped
surface heat exchanger, Journal of Food Engineering, 45(4): 195–207, 2000.
88. Ghandi A., Powell I.B., Howes T., Chen X.D. et al., Effect of shear rate and oxygen stresses on the sur-
vival of Lactococcus lactis during the atomization and drying stages of spray drying: A laboratory and
pilot scale study, Journal of Food Engineering, 113(2): 194–200, 2012.
89. Konuklar G., Gunasekaran S., Rennet-induced milk coagulation by continuous steady shear stress,
Journal of Colloid and Interface Science, 250(1):149–158, 2002.
90. Piepiórka-Stepuk J., Mierzejewska S., Simulation tests of liquid flow in the pipeline elements,
Agricultural Engineering, 1(149): 175–184, 2014.
91. Jensen B.B.B., Friis A., CFD prediction of hygiene in food processing equipment, In: Computational
Fluid Dynamics in Food Processing, Sun D.-W., Ed., CRC Press, Boca Raton, FL, USA, 2007.
92. Bergman B.-O., Tragardh C., An approach to study and model the hydrodynamic cleaning effect,
Journal of Food Process Engineering, 13(2): 135–154, 1990.
93. Cunault C., Faille C., Bouvier L., Föste H. et al., A novel set-up and a CFD approach to study the biofilm
dynamics as a function of local flow conditions encountered in fresh-cut food processing equipments,
Food and Bioproducts Processing, 93: 217–223, 2015.
94. Jensen B.B.B., Stenby M., Nielsen D.F., Improving the cleaning effect by changing average velocity,
Trends in Food Science & Technology, 18, Supplement 1: S58–S63, 2007.
95. Lelièvre C., Legentilhomme P., Legrand J., Faille C. et al., Hygienic design: Influence of the local wall
shear stress variations on the cleanability of a three-way valve, Chemical Engineering Research and
Design, 81(9): 1071–1076, 2003.
96. Rahaman M.F., Bari S., Veale D., Flow investigation of the product fill valve of filling machine for
packaging liquid products, Journal of Food Engineering, 85(2), 252–258, 2008.
97. Jin Y., Chen X.D., A fundamental model of particle deposition incorporated in CFD simulations of an
industrial milk spray dryer, Drying Technology, 28(8): 960–971, 2010.
98. Lelièvre C., Legentilhomme P., Gaucher C., Legrand J. et al., Cleaning in place: Effect of local wall
shear stress variation on bacterial removal from stainless steel equipment, Chemical Engineering
Science, 57(8): 1287–1297, 2002.
99. Ahmed S., Taif Seraji M., Jahedi J., Hashib M.A., CFD simulation of turbulence promoters in a tubular
membrane channel, Desalination, 276(1–3): 191–198, 2011.
100. Charcosset C., Limayem I., Fessi H., The membrane emulsification process—A review, Journal of
Chemical Technology & Biotechnology, 79(3): 209–218, 2004.
101. Afonso I.M., Hes L., Maia J.M., Melo L.F., Heat transfer and rheology of stirred yoghurt during cooling
in plate heat exchangers, Journal of Food Engineering, 57(2): 179–187, 2003.
102. Fernandes C.S., Dias R., Nóbrega J.M., Afonso I.M. et al., Simulation of stirred yoghurt processing in
plate heat exchangers, Journal of Food Engineering, 69(3): 281–290, 2005.
103. Nørgaard E., Sørensen T.A., Hansen T.M., Kauffeld M., Performance of components of ice slurry
systems: Pumps, plate heat exchangers, and fittings, International Journal of Refrigeration, 28(1):
83–91, 2005.
104. Gomes L.C., Moreira J.M., Teodósio J.S., Araújo J.D. et al., 96-well microtiter plates for biofouling
simulation in biomedical settings, Biofouling, 30(5): 535–546, 2014.
105. Klayman B.J., Volden P.A., Stewart P.S., Camper A.K., Escherichia coli O157:H7 requires coloniz-
ing partner to adhere and persist in a capillary flow cell, Environmental Science & Technology, 43(6):
2105–2111, 2009.
106. Moreira J.M.R., Araújo J.D.P., Miranda J.M., Simões M. et al., The effects of surface properties on
Escherichia coli adhesion are modulated by shear stress, Colloids and Surfaces B: Biointerfaces, 123:
1–7, 2014.
107. Dimartino S., Mather A.V., Alestra T., Nawada S. et al., Experimental and computational analysis of a
novel flow channel to assess the adhesion strength of sessile marine organisms, Interface Focus, 5(1):
2015.
108. Kaur H., Carriveau R., Mutus B., A simple parallel plate flow chamber to study effects of shear stress on
endothelial cells, American Journal of Biomedical Sciences, 4(1): 70–78, 2012.
109. Brown A., Meenan B.J., Investigating the effects of fluid shear forces on cellular responses to pro-
filed surfaces in-vitro: A computational and experimental investigation, In: 29th Annual International
Conference of the IEEE Engineering in Medicine and Biology Society, 2007.
110. Brown A., Burke G., Meenan B.J., Modeling of shear stress experienced by endothelial cells cultured on
microstructured polymer substrates in a parallel plate flow chamber, Biotechnology and Bioengineering,
108(5): 1148–1158, 2011.
111. Barstad R.M., Roald H.E., Cui Y., Turitto V.T. et al., A perfusion chamber developed to investigate
thrombus formation and shear profiles in flowing native human blood at the apex of well-defined
stenoses, Arteriosclerosis, Thrombosis, and Vascular Biology, 14(12): 1984–1991, 1994.
112. Lee J.-H., Kaplan J.B., Lee W.Y., Microfluidic devices for studying growth and detachment of
Staphylococcus epidermidis biofilms, Biomedical Microdevices, 10(4): 489–498, 2008.
113. Mosier A.P., Kaloyeros A.E., Cady N.C., A novel microfluidic device for the in situ optical and mechani-
cal analysis of bacterial biofilms, Journal of Microbiological Methods, 91(1): 198–204, 2012.
114. Halder P., Nasabi M., Lopez F.J.T., Jayasuriya N. et al., A novel approach to determine the efficacy of
patterned surfaces for biofouling control in relation to its microfluidic environment, Biofouling, 29(6):
697–713, 2013.
115. Salta M., Capretto L., Carugo D., Wharton J.A. et al., Life under flow: A novel microfluidic device for
the assessment of anti-biofilm technologies, Biomicrofluidics, 7(6): 064118, 2013.
116. Edlich A., Magdanz V., Rasch D., Demming S. et al., Microfluidic reactor for continuous cultivation of
Saccharomyces cerevisiae, Biotechnology Progress, 26(5): 1259–1270, 2010.
117. Shumi W., Kim S.H., Lim J., Cho K.-S. et al., Shear stress tolerance of Streptococcus mutans aggregates
determined by microfluidic funnel device (μFFD), Journal of Microbiological Methods, 93(2): 85–89,
2013.
118. Leclerc E., David B., Griscom L., Lepioufle B. et al., Study of osteoblastic cells in a microfluidic
environment, Biomaterials, 27(4): 586–595, 2006.
119. Leclerc E., El Kirat K., Griscom L., In situ micropatterning technique by cell crushing for co-cultures
inside microfluidic biochips, Biomedical Microdevices, 10(2): 169–177, 2008.
120. Wurm M., Zeng A.-P., Mechanical disruption of mammalian cells in a microfluidic system and its
numerical analysis based on computational fluid dynamics, Lab on a Chip, 12(6): 1071–1077, 2012.
121. Alejandro Bonilla F., Kleinfelter N., Cushman J.H., Microfluidic aspects of adhesive microbial dynamics:
A numerical exploration of flow-cell geometry, brownian dynamics, and sticky boundaries, Advances in
Water Resources, 30(6): 1680–1695, 2007.
122. Shen F., Li X., Li P.C.H., Study of flow behaviors on single-cell manipulation and shear stress reduction
in microfluidic chips using computational fluid dynamics simulations, Biomicrofluidics, 8(1), article
014109, 2014.
Index
Page numbers followed by f and t indicate figures and tables, respectively.
A Ambient conditions, display cabinets, 13

ambient air movement, 13–14
Activation energy (Ea), value for egg yolk, 401 radiation, 13
Agricultural products using conjugated models, convective American Food and Drug Association, 390
drying of, 342, 343t Ampere law, 491
Air–carcass interface, 69–70 ANSYS CFX, 334, 565
Air cell ANSYS Fluent®, 330, 334, 547, 565, 575, 576
on SHZ and flow patterns, 406, 407f, 408f Apples, 515, 517f, 518
volume on pasteurization value, 406, 407, 408f cooling inside ventilated packaging, 136–138,
Air circulation, inside cabinet, 7 137f–138f
Air conditions, drying, 419 Appliances, cooking, 366–368
Air curtain(s), 9–10, 12 electrical resistance stoves, 367f, 368
for display cabinets, CFD optimization, 24–55 GRH, 366–367
aims and objectives, 26–27 IH, 367–368
airflow design and research methodology, 25–26 overview, 366
CFD numerical modeling, 34–37 visualization approaches of convection flow for,
energy consumption and food safety aspects, 24 371–373
experimental setup and procedure, 27–33 Applications
with external lateral flow, 45–55 cabinet dryers, 422–426
global modeling approach, 37–41 governing equations for designing cabinet dryer,
numerical modeling, 24–25 422
numerical modeling approaches, 33–41 modeling air humidity and product moisture
without external perturbation, 42–45 content (wet porous media), 423–426
global exchanges through, 44, 45f, 45t physical model selection, 422
ELF effect on, 53–55, 53f–54f, 55t pressure drop through tray and its simulation,
RDC, 27–29, 27f, 28f, 29t 422–423
role of, 24 computer simulation to food texture, 512–527
use of, 24 cellular structure of food, compression behavior of,
Air distributor, modeling, 537–538 519, 520–521, 522f
Airflow compression behavior, 512–521
design, RDC, 25–26 compression of nuts, 519
measurement, fresh produce cooling processes, 125, 126t crack propagation by tensile loading, 521, 522,
modeling, 11–13 523f, 524f, 525f
Airflow patterns gastric digestion, 525, 526–527, 528f
air curtains, RDC, 42–44, 42f–44f loading and dropping of fruits, 513–519
ELF effect on, 45–47, 46f mastication and swallowing, 523, 524, 525, 526f
top-spray fluidized bed coating process, 537–539 oral processing and digestion, 522, 523
boundary conditions and simulation parameters, 537 in convection flow in pans using
computational grid generation, 537 GRH, 380, 381, 383
governing equations, 537 IH, 378–380, 381f, 382f
modeling air distributor, 537–538 evaluation of food texture, see Evaluation, of food
numerical setup, 538, 539 texture
Airflow velocity, baking process, 195 fluidized bed coating, 532–533
Airflow visualization, with different ventilation ports, ISM, 489
152–157; see also Ventilation ports model, thermal processing of intact eggs, 406–412
Air fluidization, 87, 88 air cell on SHZ and flow patterns, 406, 407f, 408f
Air humidity modeling, 423–426 air cell volume on pasteurization value, 406,
influence of, 13 407, 408f
Air impingement baking oven, 220–221, 220f shell thickness on pasteurization value, 407
Air temperature profile, baking, 195 simulation of whole egg in rotational method,
AISI 304 and 430 stainless steel, 373 409–410, 411f
Alginate, 305 thermal pasteurization, optimization, 410, 411f, 412
Alkaline phosphatase, 262 for optimization of cabinet dryers, see Cabinet dryers,
Aluminum, pan materials, 371 optimization of
Ambient air movement, 13–14 OSD, 478, 479, 480, 481–485
display cabinets operation and, 13–14
591
592 Index
Applicator part, RF system, 489 cooling air flow characteristics, 72–73, 72f
Arastoopour drag model, 542 cooling characteristics of
Arbitrary Mesh Interface (AMI), 475 carcasses temperature distribution, 74–75, 74f, 75f
Archimedes number (Ar), 555, 556, 557 carcass glucose level, 75–76, 75f
Arrhenius equation, 293 heat and mass transfer coefficients, 73, 73f
Arrhenius type equation, 354, 400 heat and mass transfer mechanisms during cold air
Aseptic filling, 245–246; see also Filling cooling of
Asparagus, velocity profiles heat and mass transfer inside carcass, 67–68
in brine, 308, 309f surface heat and mass transfer, 67
number on, 309f methods
Assay description, simultaneous sugars detection, cold air cooling, 63–64, 64f
453–454 enhanced by hot boning, 66
Atomization spray cooling, 64–65
air, on particle movement, 544, 545, 546–547, 548f using high thermal conductivity inserts, 65–66, 66f
droplets, dispersion in gas–solid fluidized bed and, vascular perfusion cooling, 65, 65f
547–551 overview, 62
CFD model, 547, 549 postmortem metabolism and meat quality, 62–63, 63f
experimental spray visualization, 549–550 Benard convection cells, 307, 326–327
liquid spray on, 550, 551 β-galactosidase (β-gal), 451
spray patterns, 550 Biochip, electrokinetic, for simulatneous sugars detection,
ATP hydrolysis, 62 451–456
Axi-symmetry approach assay description, 453–454
heating condition, 374 lab-on-a-chip design, 452, 453f
natural-convection heating processes, 327–328 lab-on-a-chip optimization, 454–456
ROMs, validation, 451–452
B Biofilm reactors, operation of, 561–583
CFD in, 565–566
Bacillus coagulans, 570 defined, 561
Bacillus spores, 582 flows cells, 578–582
Baking; see also Continuous industrial baking oven chambers for direct inspection of biofilms,
airflow velocity, 195 580–582
air temperature profile, 195 overview, 578–579
apparent interactions during, 194 Robbins device and modifications, 579–580
changes during, 198–200 hydrodynamics and biofilm formation, 562, 564–565
color development, 199–200 microfluidic devices, 582–583
flavor development, 200 MTPs, 567–577
solidification, 198–199 overview, 567
volume expansion, 198 6-well, 575, 576–577
defined, 194, 226 12-well, 570–575
heat transfer mechanism, 196–197 24-well, 570
mass transfer mechanism, 197–198 48-well, 568, 569–570
optimal temperature, 194 96-well, 567–568, 569f
pressure gradient, 194–195 overview, 561–562, 563t–564t
stages, 194 Biosensors, design and optimization, 439–458
temperature gradient, 194 electrokinetic biochip for simultaneous sugars
water vapor migration, 195 detection, 451–456
Baking index, 199; see also Starch gelatinization assay description, 453–454
minimum, 228, 235 lab-on-a-chip design, 452, 453f
simulation of, 212–214 lab-on-a-chip optimization, 454–456
Barium jelly, rheological model of, 525 ROMs, validation, 451–452
Batch baking ovens, 225–226 future perspectives, 456–458
Batch pasteurization, 264 mathematical models of, 441–451
Beef carcasses (large), industrial cooling of, 61–81 capillary flow, 445–448
CFD model application in assessment of, 76–77, 77f EOF, 444–445
cooling air temperature, 79, 80f flow mechanisms, 441–442
cooling air velocity direction and magnitude, governing equations, 442–451
77–79, 78f–79f mass transfer, 448–449
cooling room design, 81, 81f pressure driven flow, 443–444
evaporative vs. convective heat flux, 80, 80f reaction kinetics, 449–450
CFD modeling of, 68 regime, 441
carcass geometry, 71, 71f ROM, 450–451
governing equations, 68–70 overview, 439–441
model parameters, 70–71 Blausius equation, 579
simulat ion procedure, 71–72, 72f Blockage ratio (BR), 314
Index 593
Bostwick consistometer, 510 C

Bottled beer, thermal pasteurization of, 271–274, 273f,
274t, 275f, 276f Cabinet dryers, optimization of, 415–432
Boundary condition(s) application of CFD, 422–426
airflow pattern, top-spray fluidized bed coating governing equations for designing cabinet dryer, 422
process, 537 modeling air humidity and product moisture
CFD numerical modeling for RDC, 36–37, 37f, 37t content (wet porous media), 423–426
convection flow in pan cooking, 375 physical model selection, 422
heat and mass transfer, 350 pressure drop through tray and its simulation, 422–423
heat and mass transfer modeling cold store, 108–111 definition, 416–419
inlet, outlet, and fan, 109 design, critical aspect of, 418–419
wall, 109–111 drying process, operative factors on, 419–421
heat and mass transfer simulation, partially loaded air conditions, 419
cold store, 113 design, 421
heat transfer, 500 diffusion theories, 421
heat transfer and electric field displacement, 491 drying rate characteristic curve, 419–420
settings, during baking, 205–207 nature of food, 419–421
convection heat transfer from duct surfaces, thermal properties and conventional heating, 421
205–206 efficiency, calculation of, 426–427
exhaust box, 207 geometrical design, 427–431
heat loss through oven walls, 206 model validation, 431
radiation heat transfer from all metal surfaces, overview, 415–416
206 type, CFD simulation considerations and, 417–418
travelling trays, 206–207 Canned liquid, natural convection heating, 330–332
thermal processing of intact eggs, 403–404 with solid particles, 332–335
Boussinesq approach, 323, 327, 330 Canned liquid foods, thermal sterilization of
Bread-baking process, 225–239 CFD modeling, 293–298
baking, 227–228 assumptions in numerical simulation, 294
CFD modeling, 229, 229f–230f boundary conditions, 294–295
challenge in, 226 geometry and orientation, 296–297, 296f, 297f–298f
as complex process, 226 solid-liquid food mixtures, 297, 298
cooling, 227 vertical can simulation results, 295–296, 295f
emulsifiers, 227 fluid flow patterns analysis, 290
evaporation condensation mechanism, 232–235 F value, 290–292
SHMT, 233, 234f determining, 299–302
specific heat of bread, 234 reaction kinetics of quality changes, 292–293
steps involved in, 233–234 Canned milk, thermal pasteurization of, 267–271, 268f,
thermal conductivity, 234 269f–270f, 269t, 271f, 272f
fermentation, 226 Canning, 289–290
historical background, 225 Capillary flow, governing equations, 445–448
leavening process, 226–227 Capillary force, defined, 446
mixing process, 226 Capsules, coffee packaging in, 255, 256t
overview of, 226–228, 227f Caramelization, 199
proofing, 227 Carboxy-methyl-cellulose (CMC), 324–325, 332
radiation models, 230–231 Carry bags, 171
discrete ordinate (DO), 231 cooling and moisture loss, 187
DTRM, 230 CT image, 184–185, 185f
radiative transfer equation, 230 Cell adhesion, shear forces on, 562, 564–565
surface to surface (S2S), 231 Cellular structure of food, compression behavior of, 519,
shortening agents, 227 520–521, 522f
starch gelatinization and bread softness, 235–238 Center for Disease Control and Prevention (CDC), 561
surface browning, 238–239, 238f–239f Cereals, breakfast, 519, 520, 521
temperature profile of oven, 231, 233f CFD, see Computational fluid dynamics (CFD)
velocity profile of oven, 231, 232f CFD-based optimization, of vent hole design, 178–184
Bread crumbs, 3D typical cellular structure, 520 citrus fruit, 178–181, 179f–180f
Bread softness, starch gelatinization and, 235–238 pome fruit, 181–184, 181f–182f, 183f, 183t, 184f
Breakfast cereals, 519, 520, 521 CFD codes
Brinkman number (Br), 476 mass transfer, 16–17
Browning, bread-baking process, 238–239, 238f–239f methodology, 15–16
Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm, postprocessing, 16
481 preprocessing, 15–16
Bulk viscosity solid phase, Eulerian–Eulerian approach, solving, 16
536 turbulence models, 16
Buoyancy force, defined, 320 validation, 17–18
594 Index
CFD modeling effect on jet decay, 50–51, 51f

bread-baking process, see Bread-baking process effect on jet deflection, 49–50, 49f–50f
continuous industrial baking process, see Continuous effect on velocity profiles, 51–53, 52f
industrial baking oven turbulence modeling performance, 47–49, 48f
in fresh produce cooling processes, 128 global modeling approach, 37–41
discretization of model geometry, 129–132, 131f, air curtain performance, 40–41
131t, 132t equations, 37–40, 38f
geometry and flow domain, 129, 130f numerical modeling, 24–25
mass transfer, 132–133 numerical modeling approaches, 33–41
mathematical model, 128–129 without external perturbation, 42–45
model validation, 135–136, 136f global exchanges through air curtain, 44, 45f, 45t
produce quality modeling, 134–135, 135f jet characteristics and airflow patterns, 42–44,
refrigeration heat load calculation, 133–134, 134f 42f–44f
pasteurization process, 266–285 CFD simulation, validation, for strawberry packaging,
energy equation, 266–267 150, 151f
mass equation, 266 CFX4 software, 384, 419
momentum equation, 266 Chambers, flow, for direct inspection of biofilms, 580–582
PEF pasteurization, 279–285 Chemical agents, sterilization, 249–250
thermal pasteurization, 267–279 ethanol, 250
thermal sterilization of canned liquid foods, 293–298 hydrogen peroxide, 249–250
assumptions in numerical simulation, 294 ozone, 250
boundary conditions, 294–295 peracetic acid, 250
geometry and orientation, 296–297, 296f, Chemostats
297f–298f dilution rate of, 578
solid-liquid food mixtures, 297, 298 with irregular geometries, 578–579
vertical can simulation results, 295–296, 295f CHTC (convective heat transfer coefficient), 341, 344–348,
vent hole design; see also Vent hole design for cartons 352
airflow, heat, and mass transfer, 175–176, 175f CHTF, see Convective heat transfer factor (CHTF)
multi-scale approach, 173–175, 174f Citrus fruit
quantifying cooling performance, 176–178, 176f, cooling inside ventilated packaging, 138–139
178f vent hole design for, 178–181, 179f–180f
CFD modeling, of beef carcass cooling, 68 Clam shells, 171–172
application in process assessment, 76–77, 77f CMC (carboxy-methyl-cellulose), 324–325, 332
cooling air temperature, 79, 80f CMTC (convective mass transfer coefficient), 341,
cooling air velocity direct ion and magnitude, 344–348, 352, 353
77–79, 78f–79f Coating process, top-spray fluidized bed, see Top-spray
cooling room design, 81, 81f fluidized bed coating process
evaporative vs. convective heat flux, 80, 80f Coffee packaging in capsules, 255, 256t
carcass geometry, 71, 71f Cold air cooling, of beef carcasses, 63–64, 64f
governing equations, 68–70 heat and mass transfer mechanisms during
model parameters, 70–71 heat and mass transfer inside carcass, 67–68
simulat ion procedure, 71–72, 72f surface heat and mass transfer, 67
CFD numerical modeling, RDC “Cold feet effect,” 14
boundary conditions, 36–37, 37f, 37t Cold storage room, 123, 124f
equations and hypothesis, 34–35 Cold store
RSM, 35 heat and mass transfer modeling
standard k-ε model, 34–35 boundary conditions, 108–111
CFD optimization, of perturbed air curtains for display equations, 106–107
cabinets, 24–55 loaded boxes assuming porous media, 107–108
aims and objectives, 26–27 1-MCP gas distribution, 106
airflow design and research methodology, 25–26 overview, 105–106
CFD numerical modeling, 34–37 partially loaded, heat and mass transfer simulation in,
boundary conditions, 36–37, 37f, 37t 111, 112t
equations and hypothesis, 34–35 boundary conditions, 113
energy consumption and food safety aspects, 24 equations, 111–113
experimental setup and procedure, 27–33 ethylene control in, 117, 118f
air curtain facility, 27–29, 27f, 28f, 29t loading patterns effects, 113–115, 114f–115f
flow field, general features, 32–33, 32f–33f radiation effects, 116–117, 116f, 117t
non-intrusive experimental facilities, 29–30, performance improvement by CFD, 105–118
30f–31f Color development, during baking process, 199–200
with external lateral flow, 45–55 Composition, dielectric properties of foods, 495, 496
effect on airflow patterns, 45–47, 46f Compression behavior, food texture, 512–521
effect on global exchanges through air curtain, cellular structure, 519, 520–521, 522f
53–55, 53f–54f, 55t loading and dropping of fruits, 513–519
Index 595
nuts, 519 Conjugated drying models

overview, 512 overview, 341–342, 343t
Computational fluid dynamics (CFD), 4 strengths and weaknesses, 348
applications, retail cabinet display and, 10 Conservation equations, for gas phase, 535–536
air curtain, 12 Conservation of energy, natural convection heat transfer,
airflow modeling, 11–13 322, 323
product temperature distribution modeling, 10–11, 11f Conservation of momentum
Computational fluid dynamics (CFD), in food processing equations for gas and solid phases, 539
analysis of thermal processing of intact eggs, natural convection heat transfer, 322, 323
see Thermal processing of intact eggs Consistency index, 470, 471
applications, see Applications Constant-rate period, 420
biofilm reactors, operation of, see Biofilm reactors, Continuity equation
operation of convection flow in pan cooking, 374
biosensors, design and optimization, see Biosensors Eulerian–Eulerian approach, 535
convection flow in pan cooking, see Convection flow in natural convection heat transfer, 322, 323
pan cooking thermal processing of intact eggs, 403
convective drying of cylindrical fruit slices, Continuous cooking systems optimization, approaches for,
see Convective drying 383–384
examples, 330–335 Continuous flow pasteurization, 264–265
canned liquid, natural convection heating, 330–332 Continuous industrial baking oven
canned liquid with solid particles, 332–335 description, 200
natural-convection heating processes, see Natural- indirect-heating system, 200
convection heating processes monitoring system, 200, 202f
pasta dough extrusion process, see Pasta dough thermocouples, 200, 202f
extrusion process travelling sensors, 200, 202f
RF thawing, see Radio frequency (RF) thawing, of predicting temperature profiles, 209–221
frozen food products air impingement baking oven, 220–221, 220f
thermal processing of particulate foods, see Thermal baking index simulation, 212–214
processing of particulate foods indirect gas-fired radiant burner for tunnel baking
top-spray fluidized bed coating process, see Top-spray oven, 217–219, 219f
fluidized bed coating process optimum tin temperature profiles, operating
Computational grid condition, 215–216, 216f
convection flow in pan cooking, 378 process controller design, 216–217
generation, top-spray fluidized bed coating process, radiant tube, operational condition, 219–220
537 simulation of oven operation under increasing oven
geometry construction and, 402 load, 211–212, 212f
Computational modeling, of heat and mass transfer, schematic diagram, 201f
349–359 thermal properties of dough, 200, 202–209
CFD simulations for determination of CTCs, 351–353 boundary and initial condition settings, 205–207
formulation, 349–351 model assumption, 204–205
thermophysical properties, 353–355 modeling, 203–204
validation, 355–359 model validation, 207–209
Computer simulation methods, for analyzing food texture solving model, 207
applications, 512–527 volume condition settings, 205
cellular structure of food, compression behavior of, travelling trays and tins, 200, 201f
519, 520–521, 522f Continuum surface force model, 448
compression behavior, 512–521 Convection, 196–197; see also Heat transfer mechanism,
compression of nuts, 519 during baking
crack propagation by tensile loading, 521, 522, Convection–diffusion–reaction equation, 448–449
523f, 524f, 525f Convection flow in pan cooking, 365–385
gastric digestion, 525, 526–527, 528f approaches for continuous cooking systems
loading and dropping of fruits, 513–519 optimization, 383–384
mastication and swallowing, 523, 524, 525, 526f cooking appliances, 366–368
oral processing and digestion, 522, 523 electrical resistance stoves, 367f, 368
DEM, 512 GRH, 366–367
FEM, 511–512 IH, 367–368
FVM, 512 overview, 366
COMSOL Multiphysics, 328, 330, 378, 380, 405, 492, 499, visualization approaches, 371–373
565 materials, 368–371
Condensation, 197–198; see also Mass transfer visualization approaches, 373
mechanisms, during baking mathematical modeling by CFD, 373–383
Conduction, 196, 197; see also Heat transfer mechanism, additional CFD applications, 383
during baking applications in convection flow in pans using GRH,
Conduction heat transfer, 320 380, 381, 383
596 Index
applications in convection flow in pans using IH, carcasses temperature distribution, 74–75, 74f, 75f
378–380, 381f, 382f carcass glucose level, 75–76, 75f
assumptions, in numerical simulation, 374 heat and mass transfer coefficients, 73, 73f
basic model equations and solution procedure, Cooling performance, of ventilated cartons, 176–178, 176f,
374–378 178f
boundary conditions, 375 Cooling uniformity, among packaged fruit, 177
computational grid, 378 Copper, pan materials, 371
EFA, 376–377 Corrugated paperboard (CP), 169–170
governing equations, 374–375 Coulomb gauge, defined, 377
overview, 373 Coxiella burnetti, 262
physical model, 374 Crack propagation, by tensile loading, 521, 522, 523f, 524f,
solution strategy, 377–378 525f
overview, 365–366 Crust color intensity, 199–200
suggestions for future work, 385 CTCs (convective transfer coefficients)
Convection heat transfer, 319, 320–321 CFD simulations for determination, 351–353
from duct surfaces, 205–206 literature for derivation, 344
Convective boundary condition, 324f–325f, 326 for non-conjugated modeling, 342, 344–348
Convective drying, of cylindrical fruit slices, 339–359 overview, 340, 341
advantages and disadvantages, 348 temporal variation of, 342
CHTC and CMTC, 344–348 Cylindrical fruit slices
computational modeling of heat and mass transfer in, computational modeling of heat and mass transfer in,
349–359 349–359
CFD simulations for determination of CTCs, CFD simulations for determination of CTCs,
351–353 351–353
model formulation, 349–351 model formulation, 349–351
model validation, 355–359 model validation, 355–359
thermophysical properties, 353–355 thermophysical properties, 353–355
general description of diffusion-based models for, convective drying of, see Convective drying
340–344
agricultural products using conjugated models, D
342, 343t
conjugated drying models, 341–342, 343t Darcy friction factor, 579
non-conjugated drying models, 342, 344f Darcy number, for heat-flow visualization, 372
overview, 340–341 Darcy’s law, 422, 457
overview, 339–340 Death rate curve, 291, 291f
Convective heat flux, during beef carcass cooling, 80, 80f Debye length, 445
Convective heat transfer coefficient (CHTC), 341, Decimal reduction time (DT), 291, 293
344–348, 352 Defogging, glass doors, display cabinets, 14
Convective heat transfer factor (CHTF), 177 Degrees of freedom (DOF), geometrical, 482
Convective mass transfer coefficient (CMTC), 341, DEM (discrete element method), 316, 512, 520, 521, 522,
344–348, 352, 353 523, 524, 555
Convective transfer coefficients (CTCs) Density, thermal–physical property, 498, 499
CFD simulations for determination, 351–353 Density model development, 391, 392
literature for derivation, 344 Dielectric materials, RF thawing
for non-conjugated modeling, 342, 344–348 application, 489
overview, 340, 341 properties, 493–497
temporal variation of, 342 composition, 495, 496
Conventional heating, 421 electric conductivity, 494
Cooking, pan, convection flow in, see Convection flow in factors affecting, 495–496
pan cooking frequency, 496
Cooking appliances, 366–368 loss tangent, 494
electrical resistance stoves, 367f, 368 methods for determining, 497
GRH, 366–367 penetration depth, 495
IH, 367–368 permittivity, 493
overview, 366 physical properties related to, 494–495
visualization approaches of convection flow for, temperature, 495, 496f
371–373 Differential scanning calorimetry (DSC), 497
Cooling, industrial, of large beef carcasses, see Industrial Diffusion-based models for convective drying
cooling, of large beef carcasses general description of, 340–344
Cooling air, beef carcass cooling agricultural products using conjugated models,
flow characteristics, 72–73, 72f 342, 343t
temperature, 79, 80f conjugated drying models, 341–342, 343t
velocity direction and magnitude, 77–79, 78f–79f non-conjugated drying models, 342, 344f
Cooling characteristics, of large beef carcass overview, 340–341
Index 597
Diffusion coefficient, effective, 354, 355, 356 nature of food, 419–421

Diffusion theories, of drying, 421 diffusion theories, 421
Digestion drying rate characteristic curve, 419–420
gastric, 525, 526–527, 528f thermal properties and conventional heating, 421
oral processing and, 522, 523 Drying rate characteristic curve, 419–420
Dimensionless temperatures DTRM, see Discrete transfer radiation model (DTRM)
distributions in fluid domain, 314, 315–316
power law index and BR on, 314, 314f E
Direct inspection of biofilms, flows chambers for,
580–582 Eddy currents, 368
Direct numerical simulation (DNS), 385 EFA (electromagnetic field analysis), 376–378
“Direct UMFPACK” mode, 405 Effective thermal conductivity, eggs, 393–396
Discrete element method (DEM), 89, 316, 512, 520, 521, Eggs
522, 523, 524, 555 CFD analysis of thermal processing of intact,
Discrete ordinate (DO), 231 see Thermal processing of intact eggs
Discrete phase model (DPM), 89, 316, 534, 555 thermal pasteurization of, 274, 275, 277–279, 277f,
Discrete transfer radiation model (DTRM), 230 278f–279f, 280f, 281f
Discrete wavelet transform (DWT), signal processing Electrical resistance stoves, 367f, 368
technique, 383 Electric conductivity, RF thawing, 494
Discretization, of model geometry, fresh produce cooling Electric double layer (EDL), 444, 445
processes, 129–132, 131f, 131t, 132t Electric field displacement, in RF heating, 490–492
Dispersion in gas–solid fluidized bed, droplets atomization Electric field distribution, RF system, 500–501, 502f
and, 547–551 Electrokinetic biochip, for simultaneous sugars detection,
CFD model, 547, 549 451–456
experimental spray visualization, 549–550 assay description, 453–454
liquid spray on, 550, 551 lab-on-a-chip design, 452, 453f
spray patterns, 550 lab-on-a-chip optimization, 454–456
Display cabinets, CFD applications, 10; see also ROMs, validation, 451–452
Refrigerated display cabinet (RDC); Retail Electrokinetics, defined, 442
cabinet Electromagnetic field analysis (EFA), 376–378
air curtain, 12 Electroosmotic flow (EOF)
airflow modeling, 11–13 defined, 442
product temperature distribution modeling, 10–11, 11f governing equations, 444–445
shelves, 12–13 ELF, see External lateral flow (ELF)
Display cartons, 170, 170f; see also Ventilated cartons Emulsifiers, bread-baking process, 227
Dough; see also Baking Energy conservation equation, convection flow in pan
physical properties, 205, 205f cooking, 374–375
thermal properties during baking, see Thermal Energy consumption
properties, of dough during baking cabinet, 7
DPM, see Discrete phase model (DPM) and food safety, in RDC, 24
Drag models fresh produce cooling processes, 128, 128f
Arastoopour, 542 Energy equation, thermal processing of intact eggs, 403
Ergun, 540 EN 23953 Standard, 8–9, 8t, 9t
Gidaspow, 540, 541f, 544 Entropy generation, heat-flow visualization and, 372
Symlal–O’Brien, 541–542 EOF (electroosmotic flow)
Symlal solid–solid, 542 defined, 442
Wen–Yu, 540 governing equations, 444–445
Droplet evaporation, inclusion of, 554–555 Ergun drag model, 540
Droplet formation and atomization, top-spray fluidized Escherichia coli, 389, 580, 581
bed coating, 533 Ethanol, sterilization with, 250
Droplet phase, heat transfer models for, 552–554 Ethylene, 117
Droplets atomization and dispersion in gas–solid fluidized control in partially loaded shipping container, 117, 118f
bed, 547–551 Eulerian–Eulerian approach, 533, 534–536, 539
CFD model, 547, 549 Eulerian–Lagrangian approach, 533, 534, 547
experimental spray visualization, 549–550 Evaluation, of food texture, 509–527
liquid spray on, 550, 551 application of computer simulation to, 512–527
spray patterns, 550 compression behavior, 512–521
Dropping, of fruits, 513–519 compression behavior of cellular structure of food,
Dryers, cabinet, see Cabinet dryers, optimization of 519, 520–521, 522f
Drying, convective, of cylindrical fruit slices, compression of nuts, 519
see Convective drying crack propagation by tensile loading, 521, 522,
Drying process, operative factors on, 419–421 523f, 524f, 525f
air conditions, 419 gastric digestion, 525, 526–527, 528f
design, 421 loading and dropping of fruits, 513–519
598 Index
mastication and swallowing, 523, 524, 525, 526f investigating product fill valve for packaging liquid
oral processing and digestion, 522, 523 products, 251–252
classification of computer simulation methods for PET-bottles with volumetric swirl chamber valve,
analyzing, 511–512 252–254
DEM, 512 hot filling, 245
FEM, 511–512 post-packaging treatment, 244–245
FVM, 512 room temperature, 244
defined, 510 ultra clean, 246–247
overview, 509 Finite element method (FEM), 511–512, 513–515, 518, 519,
Evaporation, 197–198; see also Mass transfer mechanisms, 522, 523
during baking Finite volume method (FVM), 512
Evaporation condensation mechanism, bread-baking Flash pasteurization, 263
process, 232–235 Flavor development, during baking process, 200
SHMT, 233, 234f Flow behavior, of egg, 397–401
specific heat of bread, 234 Flow chambers, defined, 579
steps involved in, 233–234 Flow diversion valve (FDV), 264
thermal conductivity, 234 Flow field, RDC, 32–33, 32f–33f
Evaporative heat flux, during beef carcass cooling, 80, 80f Flow index, 470–471
Evaporator ducts, cabinets, 13 FLOWIZARD, 565
Exchange coefficients, interphase momentum Flow mechanisms, mathematical model of biosensors,
gas–solid, 540–542 441–442
solid–solid, 542 Flow patterns, air cell on, 406, 407f, 408f
Exhaust box, 207 Flows cells, biofilm reactors, 578–582
Experimental methodology, thermal processing of intact chambers for direct inspection of biofilms, 580–582
eggs, 401–402 overview, 578–579
Extensograph, 510 Robbins device and modifications, 579–580
External lateral flow (ELF), 27 Fluent®, 327, 351, 384, 419
effect on airflow patterns, 45–47, 46f Fluid dynamic behavior, during filling processes, 250–251
effect on global exchanges through air curtain, 53–55, Fluid flow, inside can with large particles, 305, 306f
53f–54f, 55t Fluidization
effect on jet decay, 50–51, 51f airflow pattern, 537
effect on jet deflection, 49–50, 49f–50f in food applications, 532
effect on velocity profiles, 51–53, 52f mechanism in top-spray fluidized bed coating, 533
RDC, 27, 27f Fluidized bed coating process, top-spray, see Top-spray
turbulence modeling performance, 47–49, 48f fluidized bed coating process
External perturbation, air curtain without Fluid thermal conductivity, 492
global exchanges through air curtain, 44, 45f, 45t Fogging, glass doors, display cabinets, 14
jet characteristics and airflow patterns, 42–44, Food, nature of, 419–421
42f–44f diffusion theories, 421
Extrusion/blowing, heat sterilization, 248 drying rate characteristic curve, 419–420
Extrusion process, pasta dough, see Pasta dough extrusion thermal properties and conventional heating, 421
process Food and Drug Administration (FDA), 264
Food packaging, see Filling; Sterilization
F Food pasteurization process, see Pasteurization
Food safety, in RDC, 24
FAC, see Forced-air cooling (FAC) Food texture
Falling-rate period, 420 application of computer simulation to, 512–527
Fan, cold store, boundary conditions, 109 cellular structure of food, compression behavior,
Faraday’s law, 491 519, 520–521, 522f
Farinograph, 510 compression behavior, 512–521
FDA, see Food and Drug Administration (FDA) compression of nuts, 519
FDV, see Flow diversion valve (FDV) crack propagation by tensile loading, 521, 522,
Federal Communication Commission (FCC), 488 523f, 524f, 525f
FEM (finite element method), 511–512, 513–515, 518, 519, gastric digestion, 525, 526–527, 528f
522, 523 loading and dropping of fruits, 513–519
Fermentation, bread-baking process, 226 mastication and swallowing, 523, 524, 525, 526f
Fick’s diffusion law, 340, 349, 355, 356, 421 oral processing and digestion, 522, 523
FIDAP–Fluid Dynamics International, Inc., 327 classification of computer simulation methods for
Filling, 244–247 analyzing
aseptic, 245–246 DEM, 512
CFD applications, 250–256 FEM, 511–512
coffee packaging in capsules, 255, 256t FVM, 512
fluid dynamic behavior during filling processes, defined, 510
250–251 evaluation of, see Evaluation, of food texture
Index 599
Forced-air cooling (FAC), 122, 141, 169 Gas–solid fluidized bed, dispersion in, 547–551
FORTRAN programming, 377–378 CFD model, 547, 549
48-well MTPs, 568, 569–570 experimental spray visualization, 549–550
Fourier and Fick’s laws, 425 liquid spray on, 550, 551
Fourier conduction heat transfer equation, 500 spray patterns, 550
Fourier equation, 68 Gas–solid momentum exchange coefficient(s), 540–542
Fracturability, notion of, 520 model, for solid hydrodynamics, 542–547
Free running oscillator, 489 atomization air on particle movement, 544, 545,
Frequency, dielectric properties of foods, 496 546–547, 548f
Fresh produce cooling processes experimental validation, 544
CFD modeling in, 128 grid generation, 542, 543f
discretization of model geometry, 129–132, 131f, numerical setup, 542, 543–544
131t, 132t particle movement in absence of spray nozzle, 544,
geometry and flow domain, 129, 130f 545f, 546f
mass transfer, 132–133 Gas–solid system, heat and mass transfer models of, 551–552
mathematical model, 128–129 Gas temperature and humidity distribution, 554–555
model validation, 135–136, 136f Gastric digestion, 525, 526–527, 528f
produce quality modeling, 134–135, 135f Gauss law, 491, 500
refrigeration heat load calculation, 133–134, 134f Generator part, RF system, 489
experimental approaches Genetic algorithms (GA), 481
airflow measurement, 125, 126t Geometrical characteristics, carton vent hole, 172–173
energy consumption, 128, 128f experimental studies, 172
temperature and relative humidity (RH) modeling studies, 172–173
measurements, 125–127, 127t Geometry
inside ventilated packaging, modeling beef carcass, CFD modeling of cooling, 71, 71f
apples, 136–138, 137f–138f cabinet dryers, 427–431
citrus fruit, 138–139 construction, computational grid and, 402
pomegranate, 139, 139f retail cabinet, 5–6, 5f–6f
strawberries, 140, 140f horizontal display, 6f
table grape, 140–141 vertical, multi-deck display, 5f
package stacking pattern effect, 141–142, 141f–143f of top-spray fluidized bed reactor, 542, 543f
types of, 122–125 Gidaspow drag model, 540, 541f, 544, 545
cold storage room, 123, 124f Glass doors fogging and defogging, display cabinets, 14
precooling, 122, 123f Glatt GPCG-1 fluidized bed coater, 542, 549
refrigerated display cabinet, 124–125, 125f Gliadine, 470
transport (shipping) cooling, 124 Global exchanges, through air curtain, 44, 45f, 45t
Friction coefficient, 94, 444 ELF effect on, 53–55, 53f–54f, 55t
Friction factor ( f), defined, 579 Global modeling approach, RDC
Froud (Fr) number, 320 air curtain performance, 40–41
Frozen food products, RF thawing of, see Radio frequency equations, 37–40, 38f
(RF) thawing, of frozen food products Glucose level, large beef carcass, 75–76, 75f
Fruits, loading and dropping of, 513–519 Glucose-6-phosphate dehydrogenase (G6PDH), 453
Fruits, packaging of; see also Internal packaging, Glutenin fibers, 470
ventilated cartons Grashof number, 321
citrus fruit, 178–181, 179f–180f Green’s theorem, 512
pome fruit, 181–184, 181f–182f, 183f, 183t, 184f GRH (gas–range heating)
strawberry, see Strawberries, packaging applications in convection flow in pans using, 380,
Fruit slices, cylindrical, convective drying of, 381, 383
see Convective drying cooking appliance, 365–367
Furosine amino acid, 199 visualization and quantification of flow in pan, 371–372
F value(s) Grid
for fruits, 310, 312t computational
during thermal processing, 290, 291–292 convection flow in pan cooking, 378
degradation of spores, 301–302, 302f generation, top-spray fluidized bed coating process,
determining, 299–302 537
slow heating point, 299, 300f geometry construction and, 402
subzone, 299, 300, 300f generation, gas–solid momentum exchange coefficient
model, 542, 543f
G
H
Gas-range heating (GRH)
applications in convection flow in pans using, 380, 381, 383 Heat and mass transfer analogy (HMTA), 344, 348
cooking appliance, 365–367 Heat and mass transfer coefficients, during air cooling of
visualization and quantification of flow in pan, 371–372 large beef carcass, 73, 73f
600 Index
Heat and mass transfer models/modeling, 551–555 Hot air-batch dryers, see Cabinet dryers
cold store Hot boning, large beef carcasses cooling enhanced by, 66
boundary conditions, 108–111 Hot filling, 245; see also Filling
equations, 106–107 Hot film anemometry (HFA), 371
loaded boxes assuming porous media, 107–108 HTST, see High-temperature-short-time (HTST)
for droplet phase, 552–554 pasteurization
of gas–solid system, 551–552 Humidification, display cabinets, in 14–15
gas temperature and humidity distribution, inclusion of Hydrated semolina, 464, 473, 476
droplet evaporation, 554–555 Hydrodynamics, biofilm formation and, 562, 564–565
Heat and mass transfer simulation Hydrofluidization (HF), 87–88, 315–316, 317f
partially loaded cold store advantages, 88
boundary conditions, 113 background, 87–88
equations, 111–113 CFD in, 88–90, 90f
ethylene control in, 117, 118f Hydrofluidization (HF) system
loading patterns effects, 113–115, 114f–115f CFD in, 88–90, 90f
radiation effects, 116–117, 116f, 117t laboratory-scaled, 87, 88f
Heating mechanism, RF, 488–490 mathematical modeling of transport phenomena
Heating processes, natural-convection, see Natural- general approach, 90–91
convection heating processes heat, mass and momentum transfer in liquid
Heat loss through oven walls, 206 medium, 91–97
Heat sterilization, 247–248 heat and mass transfer in solid medium, 97–100,
extrusion or blowing, 248 100f
overheated steam and hot air, 247–248 overview, 87–88
saturated vapor, 247 Hydrogen peroxide, sterilization with, 249–250
Heat transfer Hydroxymethylfurfural, 199, 200
boundary conditions, 500
computational modeling of, 349–359 I
CFD simulations for determination of CTCs,
351–353 ICs, see Integrated circuits (ICs)
model formulation, 349–351 IH, see Induction heating (IH)
model validation, 355–359 Immersion, 87, 88
thermophysical properties, 353–355 Indirect gas-fired radiant burner, for tunnel baking oven,
of fruit salad, 310, 311f 217–219, 219f
HF system Individual quick freezing (IQF) technologies, 87–88
in liquid medium, 91–97 Induction heating (IH)
in solid medium, 97–100, 100f applications in convection flow in pans using, 378–380,
mechanism in top-spray fluidized bed coating, 533 381f, 382f
mechanisms, during cold air cooling of beef carcass cooking appliance, 367–368
heat and mass transfer inside carcass, 67–68 cookware material for, 371
surface heat and mass transfer, 67 elements, 365
in RF heating, 490–492 heat generation in, 377–378
Heat transfer, natural-convection, 320–330 principle, 367–368
examples from studies in literature, 327–330 process, Maxwell’s equations, 376
fundamentals, 320–322 visualization and quantification of flow in pan,
mathematical background, 322–327 371–372
Benard convection cells, 326–327 Industrial, scientific and medical (ISM) applications, 489
CMC, 324–325 Industrial cabinet dryer, 417
convective boundary condition, 324f–325f, 326 Industrial cooling, of large beef carcasses, 61–81
2-dimensional square cavity, 324–325 CFD model application in assessment of, 76–77, 77f
Heat transfer mechanism, during baking, 196–197 cooling air temperature, 79, 80f
conduction, 196, 197 cooling air velocity direct ion and magnitude,
convection, 196–197 77–79, 78f–79f
radiation, 196 cooling room design, 81, 81f
Heat transfer zone, defined, 555 evaporative vs. convective heat flux, 80, 80f
Helicobacter pylori, 576 CFD modeling of, 68
Herschel–Bulkley models, 398–401 carcass geometry, 71, 71f
HF, see Hydrofluidization (HF) governing equations, 68–70
High-temperature-short-time (HTST) pasteurization, 263 model parameters, 70–71
High thermal conductivity inserts, large beef carcasses simulat ion procedure, 71–72, 72f
cooling using, 65–66, 66f cooling air flow characteristics, 72–73, 72f
HMTA (heat and mass transfer analogy), 344, 348 cooling characteristics of
Homogenization, defined, 480 carcasses temperature distribution, 74–75, 74f, 75f
Horizontal display cabinet, 6f carcass glucose level, 75–76, 75f
Horticultural carton, see Ventilated cartons heat and mass transfer coefficients, 73, 73f
Index 601
heat and mass transfer mechanisms during cold air Layout optimization, ventilation ports, 157–164
cooling of analysis sequence, 157, 157f
heat and mass transfer inside carcass, 67–68 CFD simulation model, 157–158, 158f
surface heat and mass transfer, 67 estimation of ventilation ability, 158–164
methods LDV, see Laser Doppler velocimetry (LDV)
cold air cooling, 63–64, 64f Leavening process, bread-baking process, 226–227
enhanced by hot boning, 66 LES (large eddy simulation), 565, 566
spray cooling, 64–65 Lewis number, 352
using high thermal conductivity inserts, 65–66, 66f Liner bags, 171–172
vascular perfusion cooling, 65, 65f air velocity, 188f
overview, 62 cooling and moisture loss, 187
postmortem metabolism and meat quality, 62–63, 63f cooling performance, 187
Infrared rays, sterilization with, 248 CT image of, 184–185, 185f
Inlet, cold store, boundary conditions, 109 geometry of, 185–186
In-line anemometer, oven monitoring system, 200 temperature profiles, 186, 186f, 187f
Intact eggs, CFD analysis of thermal processing of, Liquid medium, heat and mass transfer in, HF system in,
see Thermal processing of intact eggs 91–97
Integrated circuits (ICs), 125, 126 Liquid products, packaging of
Internal packaging, ventilated cartons, 171–172 canning, see Canned liquid foods, thermal sterilization
carry bags, see Carry bags of
effects of, 184–188 investigating product fill valve for, 251–252
liner bags, see Liner bags Liquid spray, on gas–solid fluidized bed behavior, 550, 551
International Organization for Standardization, 510 Loading, of fruits, 513–519
Interphase momentum exchange coefficients Loading patterns, effects on cooling process, cold store,
gas–solid, 540–542 113–115, 114f–115f
solid–solid, 542 Logarithmic order of death, 291
Ionizing radiation, sterilization with, 248–249 Logarithmic order of inactivation/destruction, 292
Isothermal zone, defined, 555 Loss tangent, RF thawing, 494
“Low temperature” cabinets, 4
J Low-temperature-long-time (LTLT) pasteurization, 263
LTLT, see Low-temperature-long-time (LTLT)
Jatropha curcas, 519 pasteurization
Jet, air curtains, RDC, 42–44, 42f–44f Lucas–Washburn equation, 447
Jet decay, ELF effect on, 50–51, 51f Luikov theory, 421
Jet deflection, ELF effect on, 49–50, 49f–50f Lumped capacitance method, 396
Joule effect, 368
Jurin’s law, 447 M
K Mackey and Ofoli model, pasta extrusion, 470

Magnetic flux equation, EFA, 376
Kinematic viscosity, defined, 321 Magnitude, of cooling air, during beef carcass cooling,
Kinetics, reaction, 449–450 77–79, 78f–79f
Kirchhoff laws, 450 Maillard reaction, 199
Kitchen Aid mixer, 476 Mass transfer
Krischer’s models, thermal conductivity, 394–396 CFD codes, 16–17
K-type thermocouples, 401 during cold air cooling of beef carcasses, 67
Κ-ω model, 566 computational modeling of, 349–359
CFD simulations for determination of CTCs,
L 351–353
model formulation, 349–351
Lab-on-a-chip model validation, 355–359
defined, 440 thermophysical properties, 353–355
design, 452, 453f convection–diffusion–reaction equation, 448–449
optimization, 454–456 HF system
Laboratory-scaled hydrofluidization (HF) system, 87, in liquid medium, 91–97
88f in solid medium, 97–100, 100f
Laboratory thermal convective dryer (LTCD) unit, 355 mechanisms, during baking, 197–198
Large beef carcasses, see Beef carcasses (large) in RF heating, 490–492
Large eddy simulation (LES), 565, 566 Mastication, 523, 524, 525, 526f
Laser-Doppler anemometry (LDA), 371 Mathematical modeling
Laser Doppler velocimetry (LDV), 29–30 biosensors, 441–451
Lateral-flow microfluidics, 456 capillary flow, 445–448
Layered growth mechanism, 555–557 EOF, 444–445
mechanism in top-spray fluidized bed coating, 533 flow mechanisms, 441–442
602 Index
governing equations, 442–451 approaches, multiphase flow, 533–536

mass transfer, 448–449 Eulerian–Eulerian approach, 533, 534–536
pressure driven flow, 443–444 Eulerian–Lagrangian approach, 533, 534
reaction kinetics, 449–450 overview, 533–534
regime, 441 convection flow in pan cooking, see Convection flow in
ROM, 450–451 pan cooking
by CFD, convection flow in pan cooking, 373–383 convective drying of cylindrical fruit slices,
additional CFD applications, 383 see Convective drying
applications in convection flow in pans using GRH, development, thermal processing of intact eggs, 402–406
380, 381, 383 assumptions, in simulation, 404, 405
applications in convection flow in pans using IH, computational grid and geometry construction, 402
378–380, 381f, 382f governing equations and boundary conditions,
assumptions, in numerical simulation, 374 403–404
basic model equations and solution procedure, numerical solution, 405
374–378 parameters, 404
boundary conditions, 375 pasteurizing value (F) calculation, 405
computational grid, 378 validation, 405, 406f
EFA, 376–377 drag models, see Drag models
governing equations, 374–375 drying process, 423–426
overview, 373 gas–solid momentum exchange coefficient model, for
physical model, 374 solid hydrodynamics, 542–547
solution strategy, 377–378 atomization air on particle movement, 544, 545,
natural-convection heat transfer, 322–327 546–547, 548f
transport phenomena of HF system experimental validation, 544
general approach, 90–91 grid generation, 542, 543f
heat, mass and momentum transfer in liquid numerical setup, 542, 543–544
medium, 91–97 particle movement in absence of spray nozzle, 544,
heat and mass transfer in solid medium, 97–100, 545f, 546f
100f heat and mass transfer, 551–555
Maxwell–Ampere equation, 376, 377 for droplet phase, 552–554
Maxwell–Faraday equation, 376, 377 of gas–solid system, 551–552
Maxwell–Gauss equation, 376 gas temperature and humidity distribution,
Maxwell’s equations, 376, 491, 499, 500 inclusion of droplet evaporation, 554–555
1-MCP gas distribution, Cold store, 106 mathematical, see Mathematical modeling
“Medium temperature” cabinets, 4–5 mixing and rheological, pasta extrusion process, 468–473
Melanoidine, 199 natural-convection heating processes, see Natural-
Meshing, uniform trigonal type, 402 convection heating processes
MFM (multifluid model), 534, 539 studies, for RF thawing, 499–505
Michaelis–Menten constant, 69 electric field distribution, 500–501, 502f
Michaelis–Menten kinetics, 449–450 overview, 499–500
Microbial population sample load on efficiency of RF process, 501, 503
death rate curve, 291, 291f simulation of sample movement within RF cavity,
decimal reduction time (DT), 291, 293 503–505
thermal death time curve, 291, 292f Tadmor, 474
time and, 293 thermal processing of particulate foods, see Thermal
Microfluidics processing of particulate foods
challenge in, 440 validation
defined, 440 cabinet dryers, 431
devices, biofilm reactors, 582–583 pasta dough extrusion process, 475–478, 479f,
lateral-flow, 456 480f, 481f
paper-based, 456 Modified Gidapow drag model, 540, 541f
Microtiter plates (MTPs), 567–577 Moisture content (MC), pasta extrusion, 467, 468–473
overview, 567 Moisture diffusivity, effective, 354
6-well, 575, 576–577 Moisture loss, during baking process, 197–198
12-well, 570–575 Momentum equations
24-well, 570 convection flow in pan cooking, 375
48-well, 568, 569–570 Eulerian–Eulerian approach, 535
96-well, 567–568, 569f thermal processing of intact eggs, 403
Mist cooling (humidification), in display cabinets, 14–15 Momentum exchange coefficients, interphase
Mixing model, pasta extrusion process, 468–473 gas–solid, 540–542
Mixing process, bread-baking process, 226 solid–solid, 542
Mixograph, 510 Movement of samples, simulation, within RF cavity, 503–505
Models/modeling; see also specific models Moving particle semi-implicit (MPS) method, 526–527
air distributor, 537–538 MTPs, see Microtiter plates (MTPs)
Index 603
Multifluid model (MFM), 534, 539 Open ended coaxial probe method, 497
Multiphase flow modeling approaches, 533–536 OpenFoam software, 475, 565
Eulerian–Eulerian approach, 533, 534–536 Optimal shape design (OSD), 478, 479, 480, 481–485
Eulerian–Lagrangian approach, 533, 534 Optimization; see also CFD-based optimization
overview, 533–534 approaches for continuous cooking systems, 383–384
Multi-region models, defined, 340 biosensors, see Biosensors
Multi-scale modeling approach, to vent hole design, cabinet dryers, see Cabinet dryers, optimization of
173–175, 174f lab-on-a-chip, 454–456
Mycobacterium tuberculosis, 262 process, pasta dough extrusion process, 478, 479, 480,
481–485
N thermal pasteurization, intact eggs, 410, 411f, 412
Optimizing ventilation efficiency, 149–150, 150f
Natural convection CFD simulation, 150
defined, 371 validation, 150, 151f
by phase change, 371 Oral processing and digestion, 522, 523
Natural-convection heating processes, 319–336 OSD (optimal shape design), 478, 479, 480, 481–485
CFD examples, 330–335 Outlet, cold store, boundary conditions, 109
canned liquid, 330–332 Oven monitoring system, in continuous industrial baking
canned liquid with solid particles, 332–335 oven, 200, 202f
heat transfer, 320–330 Oven rise, 194
Benard convection cells, 326–327 Ovens, 225–226; see also Continuous industrial baking
CMC, 324–325 oven
convective boundary condition, 324f–325f, 326 bread-baking process, see Bread-baking process
2-dimensional square cavity, 324–325 heat distribution inside, 194, 226
examples from studies in literature, 327–330 temperature profile of, 231, 233f
fundamentals, 320–322 U-turn movement, 205
mathematical background, 322–327 velocity profile of, 231, 232f
overview, 319–320 Oven spring, 194
Nature of food, 419–421 Oven walls, heat loss through, 206
diffusion theories, 421 Overheated steam and hot air, sterilization with,
drying rate characteristic curve, 419–420 247–248
thermal properties and conventional heating, 421 Ozone, sterilization with, 250
Navier–Stokes equations, 133
convection flow in pan cooking, 374 P
DPM, 316
microchannels of biosensors, 442–443 Package stacking pattern, effects of, 141–142, 141f–143f
natural convection heat transfer, 322 Packaging
pasta extrusion process, 467, 472, 474 food, see Filling; Sterilization
thermal processing of intact eggs, 403 fruit, see Fruits, packaging of
Newton law of viscosity, 443 ventilation ports optimization for, see Ventilation ports
Nickle, pan materials, 371 Page model, 421
96-well MTPs, 567–568, 569f Pan cooking, convection flow in, see Convection flow in
Non-conjugate approach, 425 pan cooking
Non-conjugated drying models Parallel models, thermal conductivity, 394–396
agricultural products using, 345t–347t Pareto optimality, 480
CHTC and CMTC, 344–348 Partial differential equations (PDEs), 403, 482, 511, 512
overview, 342, 344f Partially loaded cold store; see also Cold store
strengths and weaknesses, 348 heat and mass transfer simulation, 111, 112t
Non-enzymatic browning reactions, 199 boundary conditions, 113
Numerical modeling, RDC, 24–25; see also specific equations, 111–113
models ethylene control in, 117, 118f
Nusselt (Nu) number loading patterns effects, 113–115, 114f–115f
corresponding distributions of, 353 radiation effects, 116–117, 116f, 117t
as function of Ra number, 321 Particle image velocimetry (PIV), 30, 30f–31f, 371, 383
for isothermal heating, 384 Particles motion, top-spray fluidized bed coating process,
of solid phase, 552 539–547
spatial distributions of, 352 gas–solid momentum exchange coefficient model, for
Nuts, compression of, 519 solid hydrodynamics, 542–547
Nyquist criteria, 499 absence of spray nozzle, 544, 545f, 546f
atomization air on particle movement, 544, 545,
O 546–547, 548f
experimental validation, 544
Ohm’s law, 376, 450 grid generation, 542, 543f
Ohnesorge number, 549 numerical setup, 542, 543–544
604 Index
governing equations of mass and momentum Poisson–Boltzmann equation, 444, 445

conservation, 539 Poisson equation, 445
interphase momentum exchange coefficients Polydimethylsiloxane (PDMS), 583
gas–solid, 540–542 Polypropylene microplates (PP), 567
solid–solid, 542 Polystyrene (PS), 567
Pasta dough extrusion process, 463–485 Pome fruit, vent hole design for, 181–184, 181f–182f, 183f,
CFD in, 467 183t, 184f
equations, 467–478 Pomegranate, cooling inside ventilated packaging, 139,
mixing and rheological models, 468–473 139f
model validation, 475–478, 479f, 480f, 481f Pore network modeling (PNM), 457
in screw press, 473, 474–475 Porous media, heat and mass transfer within, 107–108
optimization process, OSD, 478, 479, 480, 481–485 Positron emission particle tracking (PEPT), 544
overview, 463–465 Postharvest, CFD simulation techniques in, 132t
physics of, 465–467 Power law model
verification and validation, 465 index
Pasteur, Louis, 261 on dimensionless temperatures, 314, 314f
Pasteurization, 261–285 flow index, 470–471
CFD modeling, 266–285 temperature on flow behavior, 398–401
energy equation, 266–267 Power uniformity index (PUI), 505
mass equation, 266 Prandtl (Pr) number, 321, 372
momentum equation, 266 for gas phase, 552
PEF pasteurization, 279–285 Precooling, 122, 123f
thermal pasteurization, 267–279 Preservation techniques, 289
defined, 261 Pressure-driven flow (PDF), 441, 442
overview, 261–262 microfluidic biosensor, 443–444
pulsed electric field (PEF), 261, 262, 265–266, 265f Pressure drop, through tray and its simulation, 422–423
CFD modeling, 279–285 Pressure gradient, baking process, 194–195
sterilization vs., 262 Pressure-implicit split-operator (PISO), 475
thermal, 262–265 Process controller design, for continuous bread baking
CFD modeling, 267–279 process, 216–217, 217f
methods, 263–265 Product moisture content, 423–426
Pasteurization–sterilization Product temperature distribution, modeling, 10–11, 11f
canned food products, 323 Properties
canned liquids, 330 mechanical
Pasteurization value (F), thermal processing of intact eggs Jatropha curcas seed, 519
air cell volume on, 406, 407, 408f pumpkin seeds, 519
calculation, 405 pan materials, 369, 370t
shell thickness on, 407 physical, related to dielectric properties, 494–495
stationary and rotating modes, 410, 411f, 412 electric conductivity, 494
Pasteurizing values (PV), difference in, 306 loss tangent, 494
PDEs (partial differential equations), 403, 482, 511, 512 penetration depth, 495
Pear fruit, 514, 515, 516f RF thawing, dielectric, 493–497
Pecan shelling, 519 composition, 495, 496
Penetration depth, RF energy, 495 electric conductivity, 494
Peracetic acid, sterilization with, 250 factors affecting, 495–496
Permittivity, dielectric, RF thawing, 493 frequency, 496
Phase change, during thawing, 492–493 loss tangent, 494
PHOENICS software, 327, 383 methods for determining, 497
Phosphoglucose isomerase (PGI), 453, 454 penetration depth, 495
Photolithography, 583 permittivity, 493
Physical model physical properties related to, 494–495
convection flow in pan cooking, 374 temperature, 495, 496f
selection, 422 thermal–physical, 497–499
Physical properties, related to dielectric properties, density, 498, 499
494–495 specific heat, 497, 498t
electric conductivity, 494 thermal conductivity, 498
loss tangent, 494 thermophysical, computational modeling of heat and
penetration depth, 495 mass transfer, 353–355
Physics, pasta dough extrusion process, 465–467 thermophysical and rheological, intact eggs,
PimpleDyMFoam, 475 391–401
Pineapple, thermal sterilization process, 308, 310, 311f chemical analysis, 391t
PIV, see Particle image velocimetry (PIV) density model development, 391, 392
Plates, MTPs, see Microtiter plates (MTPs) effective thermal conductivity, 393–396
Poiseulle law, 457 rheological parameters, 398–401
Index 605
surface heat transfer coefficients, 396–397 specific heat, 497, 498t

temperature on flow behavior, 397–401 thermal conductivity, 498
Protein denaturation, 199, 200, 228 RANS equations, see Reynolds-averaged Navier–Stokes
Pulsed electric field (PEF) pasteurization, 261, 262, (RANS) equations
265–266, 265f Rayleigh (Ra) number, 321, 374, 384, 403
advantages of, 265 RDC, see Refrigerated Display Cabinet (RDC)
CFD modeling, 279–285 Re, see Reynolds (Re) numbers
design and optimization of treatment chamber, Reaction kinetics, in food biosensors, 449–450
280–285, 282f, 283f, 284f Reactors
governing equation for electrostatics, 279, 280 biofilm, operation of, see Biofilm reactors, operation of
components of, 265 top-spray fluidized bed coating, gas–solid momentum
efficacy of, 265–266 exchange coefficient model, 542–547
Pulsed light, sterilization, 249 Rear ducts, cabinets, 13
Pumpkin seeds, mechanical properties, 519 Reduced order models (ROM)
Pycnometer, volume of, 391, 392 governing equation, 450–451
validation, 451–452
R Refrigerated display cabinet (RDC), 24
aims and objectives, 26–27
Radial distribution function, Eulerian–Eulerian approach, airflow design and research methodology, 25–26
536 CFD numerical modeling
Radiant tube, operational condition design of, 219–220 boundary conditions, 36–37, 37f, 37t
Radiation, 319 equations and hypothesis, 34–35
during baking, 196 energy consumption and food safety, 24
display cabinets operation and, 13 experimental setup and procedure
heat and mass transfer simulation in partially loaded air curtain facility, 27–29, 27f, 28f, 29t
cold store and, 116–117, 116f, 117t flow field, general features, 32–33, 32f–33f
sterilization, 248–249 non-intrusive experimental facilities, 29–30,
infrared rays, 248 30f–31f
ionizing radiation, 248–249 with external lateral flow
pulsed light, 249 effect on airflow patterns, 45–47, 46f
UV rays, 248 effect on global exchanges through air curtain,
Radiation heat transfer from all metal surfaces, 206 53–55, 53f–54f, 55t
Radiation models, bread-baking process, 230–231 effect on jet decay, 50–51, 51f
discrete ordinate (DO), 231 effect on jet deflection, 49–50, 49f–50f
DTRM, 230 effect on velocity profiles, 51–53, 52f
radiative transfer equation, 230 turbulence modeling performance, 47–49, 48f
surface to surface (S2S), 231 fresh produce cooling processes, 124–125, 125f
Radiative transfer equation, bread-baking process, 230 global modeling approach
Radio frequency (RF) thawing, of frozen food products, air curtain performance, 40–41
487–505 equations, 37–40, 38f
dielectric properties, 493–497 numerical modeling, 24–25
composition, 495, 496 approaches, 33–41
electric conductivity, 494 without external perturbation
factors affecting, 495–496 global exchanges through air curtain, 44, 45f, 45t
frequency, 496 jet characteristics and airflow patterns, 42–44,
loss tangent, 494 42f–44f
methods for determining, 497 Refrigeration equipment, retail cabinet, 6
penetration depth, 495 Refrigeration heat load, calculating, 133
permittivity, 493 Regime, mathematical model of biosensors, 441
physical properties related to, 494–495 Relative dielectric constant, material, 493
temperature, 495, 496f Relative humidity (RH) measurements, fresh produce
heating mechanism, 488–490 cooling processes, 125–127, 127t
heat transfer and electric field displacement, Research methodology, RDC, 25–26
490–492 Residence time distribution (RTD)
modeling studies for, 499–505 measurement, 469
electric field distribution, 500–501, 502f by solids fraction, 312
overview, 499–500 Resistance ratio, defined, 538
sample load on efficiency of RF process, 501, 503 Resistive temperature detectors (RTDs), 125–126
simulation of sample movement within RF cavity, Resonance cavity method, 497
503–505 Resorufin β-Dgalactopyranoside (RBG), 451
overview, 487–488 Retail cabinet
phase change during thawing, 492–493 applications of CFD to display, 10
thermal–physical properties, 497–499 air curtain, 12
density, 498, 499 airflow modeling, 11–13
606 Index
product temperature distribution modeling, 10–11, 11f Sample movement, simulation, within RF cavity, 503–505
shelves, 12–13 Saturated vapor, sterilization with, 247
CFD codes Schmidt number (Sc), 554
mass transfer, 16–17 Screw press, equations in, 473, 474–475
methodology, 15–16 Semi-implicit method for pressure-linked equations
postprocessing, 16 (SIMPLE)
preprocessing, 15–16 algorithm, 539
solving, 16 scheme, 475
turbulence models, 16 Semolina, 464, 466, 468, 470, 473, 476
validation, 17–18 Sercom press laboratory extrusion unit, 469f, 476
classification, 4–7 Series models, thermal conductivity, 395–396
air circulation, 7 Shear forces, on cell adhesion and biofilm development,
energy consumption, 7 562, 564–565
geometry, 5–6, 5f–6f Shear viscosity solid phase, Eulerian–Eulerian approach,
refrigeration equipment, 6 536
storage temperature, 4–5 Shell thickness on pasteurization value, 407
evaporator and rear ducts, 13 Shelves, cabinets, 12–13
features, 4 Sherwood number (Sh), 554
glass doors fogging and defogging, 14 Shortening agents, bread-baking process, 227
influence of air humidity, modeling, 13 SHP (slowest heating point)
interactions with ambient conditions, modeling of, 13 defined, 390
ambient air movement, 13–14 eggs, 406, 407
radiation, 13 pasteurizing value, 406
mist cooling - humidification, 14–15 Shrinkage effect, 354
standardized temperature tests, 7–10, 8f, 8t–9t SHZ (slowest heating zone), 310
air curtains, 9–10 air cell on, 406, 407f, 408f
Reynolds-averaged fluid flow equations, 106 defined, 390
Reynolds-averaged Navier–Stokes (RANS) equations, 68, simulation of whole egg in rotational method, 409,
106–107, 112, 128–129 410f
airflow pattern, top-spray fluidized bed coating Signal processing technique, of DWT, 383
process, 537 SIMPLE (semi-implicit method for pressure-linked
steady-state, 565 equations)
Reynolds (Re) numbers algorithm, 539
defined, 321, 534, 579 scheme, 475
layered growth mechanism, 555, 556 Simulation(s); see also Heat and mass transfer simulation
microfluidic channel, 441, 444 cabinet dryers
for solid phase, 541 applications, 422–426
Reynolds stress model (RSM), 34, 35, 323 considerations, 417–418
Rheological model, pasta extrusion process, 468–473 determination of CTCs, 351–353
Rheological properties, intact eggs, 391–401 FEM-based, 519
chemical analysis, 391t F values for fruits, 310, 312t
density model development, 391, 392 parameters, airflow pattern, 537
effective thermal conductivity, 393–396 pasta dough extrusion process, see Pasta dough
parameters, 398–401 extrusion process
surface heat transfer coefficients, 396–397 of sample movement within RF cavity, 503–505
temperature on flow behavior, 397–401 thermal processing of intact eggs
Rheology, of extrusion screw, 466 assumptions in, 404, 405
Robbins device and modifications, 579–580 whole egg in rotational method, 409–410, 411f
ROM (reduced order models) Simulatneous sugars detection, electrokinetic biochip for,
governing equation, 450–451 451–456
validation, 451–452 assay description, 453–454
Room designl, beef carcass cooling, 81, 81f lab-on-a-chip design, 452, 453f
Root mean square error (RMSE), 398, 400 lab-on-a-chip optimization, 454–456
Rosin–Rammler distribution, 549 ROMs, validation, 451–452
RTD (residence time distribution) 6-well MTPs, 575, 576–577
measurement, 469 Slowest heating point (SHP)
by solids fraction, 312 defined, 390
RTDs, see Resistive temperature detectors (RTDs) eggs, 406, 407
pasteurizing value, 406
S Slowest heating zone (SHZ), 289, 290
air cell on, 406, 407f, 408f
Saccharomyces cerevisiae, 226 defined, 390
Salmonella enteritidis, 383, 389, 390, 405, 410, 412 as reference zone, 290
Sample load, on efficiency of RF process, 501, 503 simulation of whole egg in rotational method, 409, 410f
Index 607
Sodium fluorescein salt, 549 CFD applications, 256–258

“Soft sphere” approach, 317 decontamination by atmospheric plasmas, 256–257
Solar cabinet dryer, 417 VHP pouch, 257–258, 258t
Solid hydrodynamics, gas–solid momentum exchange chemical agents, 249–250
coefficient model for, 542–547 ethanol, 250
absence of spray nozzle, 544, 545f, 546f hydrogen peroxide, 249–250
atomization air on particle movement, 544, 545, ozone, 250
546–547, 548f peracetic acid, 250
experimental validation, 544 heat, 247–248
grid generation, 542, 543f extrusion or blowing, 248
numerical setup, 542, 543–544 overheated steam and hot air, 247–248
Solidification, during baking process, 198–199 saturated vapor, 247
protein denaturation, 199 pasteurization vs., 262
starch gelatinization, 198–199 radiation, 248–249
Solid medium, heat and mass transfer in, HF system in, infrared rays, 248
97–100, 100f ionizing radiation, 248–249
Solid particles, canned liquid with, natural convection pulsed light, 249
heating, 332–335 UV rays, 248
Solid phase pressure, Eulerian–Eulerian approach, Sterilization–pasteurization, of canned food products, 323
536 Still cans, sterilization of, see Canned liquid foods,
Solid–solid momentum exchange coefficients, 542 thermal sterilization of
Solution strategy, convection flow in pan cooking, Stokes number (St), 556
377–378 Storage temperature, retail cabinet, 4–5
Spalart–Allmaras model, 566 STORM/CFD2000, 377, 380
Spatial variation Strawberries
moisture profiles, dimensionless, 356, 358f cooling inside ventilated packaging, 140, 140f
temperature profiles, dimensionless, 356, 358f packaging
Spatiotemporal temperature measurement data, 126 CFD simulation, validation of, 150, 151f
Specific heat ventilation ports, see Ventilation ports
thermal–physical property, 497, 498t Sugars detection, electrokinetic biochip for, 451–456
Specific heat capacity, food materials, 421 assay description, 453–454
Specific mechanical energy (SME), measurement of, 469, lab-on-a-chip design, 452, 453f
477 lab-on-a-chip optimization, 454–456
Specific rate of evaporation (SRE), 427 ROMs, validation, 451–452
Spray cooling, large beef carcasses, 64–65 Sums of squared residuals (SSRs), 544
Spray nozzle, particles movement in absence of, 544, 545f, Surface browning, bread-baking process, 238–239,
546f 238f–239f
Spray patterns, 550 Surface heat, during cold air cooling of beef carcasses, 67
Spray visualization, experimental, 549–550 Surface heat transfer coefficients, eggs, 396–397
Spray zone, 553, 555 Surface to surface (S2S) radiation models, 231
S2S, see Surface to surface (S2S) radiation models Survivor curve, see Death rate curve
Stacking, 141 Swallowing, 523, 524, 525, 526f
Stainless steel, pan materials, 371 Symlal–O’Brien drag model, 541–542
Standardized temperature tests Symlal solid–solid drag model, 542
display cabinet during, 8f Symmetry approach, natural-convection heating
EN 23953 Standard, 8–9, 8t, 9t processes, 327
retail cabinet, 7–10, 8f, 8t–9t
air curtains, 9–10 T
specific conditions for, 7
Standard k-ε model, 34–35 Table grape, cooling inside ventilated packaging, 140–141
Staphylococcus aureus, 576 Tadmor model, 474
Star-CD, 419 Taylor expansion, 471, 472
Starch gelatinization Taylor (Ta) number, 320
baking process, 195 Teflon belt, 499, 503
bread-baking process, 228, 235–238 Telescopic cartons, 170, 170f, 171; see also Ventilated
CFD modeling/simulation, 212–214, 214f, 215f cartons
crumb formation, 202 Temperature
emulsifiers, 199 cooling air, during beef carcass cooling, 79, 80f
solidification, 198–199 dielectric properties of foods, 495, 496f
temperature profiles, 215–216, 216f distribution, during air cooling of large beef carcass,
variation, 213–214, 214f, 215f 74–75, 74f, 75f
Sterilization on flow behavior, eggs, 397–401
canned liquid foods, see Canned liquid foods, thermal measurements, fresh produce cooling processes,
sterilization of 125–127, 127t
608 Index
profiles canned milk, 267–271, 268f, 269f–270f, 269t, 271f,

of particulates in three distinct locations, 313f 272f
in rotating can, 311, 313f eggs in stationary and rotation modes, 274, 275,
storage, retail cabinet, 4–5 277–279, 277f, 278f–279f, 280f, 281f
stratification heat transfer mechanisms, 267
inside can, 305, 306f continuous flow pasteurization, 264–265
natural convection effects by, 307, 308f flash pasteurization, 263
Temperature gradient, baking process, 194 HTST, 263
Temperature measuring sensors, 125–126, 127t intact egg, optimization of, 410, 411f, 412
Temperature profiles of liquid egg products, 390
for baked bread quality estimation, 209–221; see also LTLT, 263
Continuous industrial baking oven methods, 263–265
air impingement baking oven, 220–221, 220f tunnel pasteurization, 263–264, 263f
baking index simulation, 212–214 ultra-pasteurization, 264
indirect gas-fired radiant burner for tunnel baking Thermal–physical properties, 497–499
oven, 217–219, 219f density, 498, 499
optimum tin temperature profiles, operating specific heat, 497, 498t
condition, 215–216, 216f thermal conductivity, 498
process controller design, 216–217 Thermal processing
radiant tube, operational condition, 219–220 canning, 289–290; see also Thermal sterilization of
simulation of oven operation under increasing oven canned liquid foods
load, 211–212, 212f defined, 289
of oven, 231, 233f Thermal processing of intact eggs, CFD analysis, 389–412
Temporal variation experimental methodology, 401–402
moisture content, dimensionless, 356 model, application of, 406–412
moisture profiles, dimensionless, 356, 358f air cell on SHZ and flow patterns, 406, 407f, 408f
predicted, energy exchange at surface of slice, 357, air cell volume on pasteurization value, 406, 407, 408f
359 shell thickness on pasteurization value, 407
product temperature product, dimensionless, 356, simulation of whole egg in rotational method,
357f 409–410, 411f
temperature profiles, dimensionless, 356, 358f thermal pasteurization, optimization, 410, 411f, 412
Tensile loading, crack propagation by, 521, 522, 523f, 524f, model development, 402–406
525f assumptions, in simulation, 404, 405
Texture, food computational grid and geometry construction, 402
application of computer simulation to, 512–527 governing equations and boundary conditions,
compression behavior, 512–521 403–404
compression behavior of cellular structure of food, numerical solution, 405
519, 520–521, 522f parameters, 404
compression of nuts, 519 pasteurizing value (F) calculation, 405
crack propagation by tensile loading, 521, 522, validation, 405, 406f
523f, 524f, 525f overview, 389–391
gastric digestion, 525, 526–527, 528f thermophysical and rheological properties during,
loading and dropping of fruits, 513–519 391–401
mastication and swallowing, 523, 524, 525, 526f chemical analysis, 391t
oral processing and digestion, 522, 523 density model development, 391, 392
classification of computer simulation methods for effective thermal conductivity, 393–396
analyzing parameters, 398–401
DEM, 512 surface heat transfer coefficients, 396–397
FEM, 511–512 temperature on flow behavior, 397–401
FVM, 512 Thermal processing of particulate foods, 305–317
defined, 510 dimensionless temperatures
evaluation of, see Evaluation, of food texture distributions in fluid domain, 314, 315–316
Texture profile analysis (TPA), 510 power law index and BR on, 314, 314f
Thawing, RF, of frozen food products, see Radio fluid flow, inside can with large particles, 305, 306f
frequency (RF) thawing, of frozen food heat transfer, of fruit salad, 310, 311f
products hydrofluidization, 315–316, 317f
Thermal conductivity PV, difference in, 306
eggs, 393–396 RMSE, experimental, and simulated F values for
thermal–physical properties, 498 fruits, 310, 312t
Thermal death time (TDT), 291 temperature
Thermal pasteurization, 262–265 of particulates in three distinct locations, 311, 313f
batch pasteurization, 264 in rotating can, 311, 313f
CFD modeling, 267–279 stratification inside can, 305, 306f
bottled beer, 271–274, 273f, 274t, 275f, 276f velocity around sphere and variation of viscosity, 314
Index 609
velocity profiles compression test, 515f

for asparagus in brine, 308, 309f deformation and damage situation, 515f
asparagus number on, 308, 309f geometry and structure, 513f
of canned peas, 306, 307f loading and dropping of fruits, 513–514
in rotating can, 311, 313f mechanics, 514f
VOF multiphase approach, 310 Topological optimization, defined, 480
Thermal properties, food materials, 421 Top-spray fluidized bed coating process, 532–557
Thermal properties, of dough during baking, 200, CFD descriptions for mechanisms in, 536–557
202–209 airflow patterns, 537–539
boundary and initial condition settings, 205–207 droplets atomization and dispersion in gas–solid
convection heat transfer from duct surfaces, fluidized bed, 547–551
205–206 heat and mass transfer and layer-growth
exhaust box, 207 mechanism, 551–557
heat loss through oven walls, 206 particles motion, 539–547
radiation heat transfer from all metal surfaces, 206 droplet formation and atomization, 533
travelling trays, 206–207 heat transfer and layered growth mechanism, 533
model assumption, 204–205 mechanisms in, 533
three-dimensional flow, 204 multiphase flow modeling approaches, 533–536
transient state computation, 205 Eulerian–Eulerian approach, 533, 534–536
turbulent flow, 204–205 Eulerian–Lagrangian approach, 533, 534
modeling, 203–204 overview, 533–534
baking oven geometry, simplification of, 203–204, overview, 532
204f principles and its application in food, 532–533
model validation, 207–209 Total ventilated area (TVA), 141
solving model, 207 Transient state computation, continuous baking process, 205
volume condition settings, 205 Transmission line method, 497
dough properties, 205 Transport (shipping) cooling, fresh produce, 124
flow source, 205 Travelling tray oven, 194
heat source, 205 Travelling trays, boundary condition settings, 206–207
Thermal sterilization of canned liquid foods Tray dryers, see Cabinet dryers
CFD modeling, 293–298 Tunnel baking oven, indirect gas-fired radiant burner for,
assumptions in numerical simulation, 294 217–219, 219f
boundary conditions, 294–295 Tunnel pasteurization, 263–264, 263f
geometry and orientation, 296–297, 296f, Turbulence models(ing), 351
297f–298f CFD codes, 16
solid-liquid food mixtures, 297, 298 performance, 47–49, 48f
vertical can simulation results, 295–296, 295f Turbulent dissipation equation, Eulerian–Eulerian
fluid flow patterns analysis, 290 approach, 535
F value, 290–292 Turbulent flow, continuous baking process, 204–205
determining, 299–302 Turbulent kinetic energy equation, Eulerian–Eulerian
reaction kinetics of quality changes, 292–293 approach, 535
Thermal zones, defined, 555 Turbulent viscosity of gas phase, Eulerian–Eulerian
Thermocouples, 126 approach, 536
K-type, 401 TVA, see Total ventilated area (TVA)
Thermophysical properties 24-well MTPs, 570
computational modeling of heat and mass transfer,
353–355 U
intact eggs, 391–401
chemical analysis, 391t Ultra clean filling, 246–247; see also Filling
density model development, 391, 392 Ultra-pasteurization, 264
effective thermal conductivity, 393–396 Uniaxial flow, defined, 441
rheological parameters, 398–401 United States Department of Agriculture (USDA)
surface heat transfer coefficients, 396–397 pasteurization defined by, 261
temperature on flow behavior, 397–401 User-defined function (UDF), 384
3D CFD discretization, 131t U-turn movement, in oven, 205
Three-dimensional approach, natural-convection heating UV rays, sterilization with, 248
processes, 327
Three-dimensional CFD modeling, see Continuous V
industrial baking oven
Three-dimensional flow, continuous baking process, 204 Validation, CFD codes, 17–18
Tin temperature profiles, operating condition design for, Validation, model
215–216, 216f cabinet dryers, 431
Titanium, pan materials, 371 computational modeling of heat and mass transfer,
Tomatoes 355–359
610 Index
pasta dough extrusion process, 465, 475–478, 479f, Ventilation ability, estimation of, 158–164
480f, 481f optimized port layouts, 161–164, 162f, 163t, 164f
ROMs, 451–452 various port layouts, 158–161, 159f, 160t, 161f
thermal processing of intact eggs, 405, 406f Ventilation ports, 152–157
Vascular perfusion cooling, large beef carcasses, 65, 65f airflow inside the box and, 152–157, 153f–154, 155f, 156f
Velocity direction, of cooling air during beef carcass CFD simulation conditions, 152, 153f
cooling, 77–79, 78f–79f layout optimization, 157–164
Velocity profile(s) analysis sequence, 157, 157f
around sphere and variation of viscosity, 314 CFD simulation model, 157–158, 158f
for asparagus in brine, 308, 309f estimation of ventilation ability, 158–164
asparagus number on, 308, 309f Vertical, multi-deck display cabinet, 5f
of canned peas, 307, 307f Viscometer, 510
ELF effect on, 51–53, 52f Visualization approaches, of convection flow
of oven, 231, 232f cooking appliances, 371–373
in rotating can, 311, 313f pan materials, 373
Vent hole design for cartons; see also Ventilated cartons VOF (volume of fluid) model, 310, 329, 447, 566
CFD-based optimization, 178–184 Volume average power absorption, of sample, 505
citrus fruit, 178–181, 179f–180f Volume condition settings, during baking, 205
pome fruit, 181–184, 181f–182f, 183f, 183t, 184f dough properties, 205
CFD modeling, 173–178 flow source, 205
airflow, heat, and mass transfer, 175–176, 175f heat source, 205
multi-scale approach, 173–175, 174f Volume expansion, during baking process, 198
quantifying cooling performance, 176–178, 176f, Volume of fluid (VOF) model, 310, 329, 447, 566
178f Volume of pycnometer, 391, 392
geometrical characteristics, 172–173
experimental studies, 172 W
modeling studies, 172–173
Ventilated cartons Wall, cold store, boundary conditions, 109–111
display cartons, 170, 170f Washburn equation, 457
future prospects, 188 Water fluidization (WF), 88; see also Hydrofluidization (HF)
internal packaging, 171–172 Water mixing, pasta extrusion process, 468–473
effects, 184–188 Weber number, 549
overview, 169–170 12-well MTPs, 570–575
sub-category designs, 170–171 Wen–Yu drag model, 540
telescopic cartons, 170, 170f, 171 Wet etching method, 583
types of, 170–171 Wet porous media, 423–426
vent hole design, see Vent hole design for cartons Wheat gluten, 199
Ventilated packaging, fresh fruit cooling inside,
modeling X
apples, 136–138, 137f–138f
citrus fruit, 138–139 X-ray tomography, airy breakfast cereals, 520
pomegranate, 139, 139f
strawberries, 140, 140f Y
table grape, 140–141
Ventilation, optimizing efficiency, 149–150, 150f Young–Laplace equation, 447

Da-Wen Sun (Editor) - Computational Fluid Dynamics in Food Processing-CRC Press (2018)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Da-Wen Sun (Editor) - Computational Fluid Dynamics in Food Processing-CRC Press (2018)

Uploaded by

Copyright:

Available Formats

Computational Fluid

Dynamics in Food Processing

For more information about this series, please visit: https://www.crcpress.com/

© 2019 by Taylor & Francis Group, LLC

No claim to original U.S. Government works

Printed on acid-free paper

International Standard Book Number-13: 978-1-138-56831-0 (Hardback)

Visit the Taylor & Francis Web site at

and the CRC Press Web site at

Section I CFD Applications in Cold Chain Facilities

Chapter 1 CFD Aided Retail Cabinets Design..............................................................................3

Chapter 3 CFD Modeling to Improve the Performance of Industrial Cooling of Large

Chapter 4 CFD Modeling of Heat and Mass Transfer in a Hydrofluidization System

Chapter 5 Improving the Performance of a Partially Loaded Cold Store by CFD.....................105

Chapter 6 CFD Investigation of Fresh Produce Cooling Processes and Effects

Chapter 8 Optimization of Horticultural Carton Vent Hole Design by CFD..............................169

Section II CFD Applications in Thermal

Chapter 9 Three-Dimensional CFD Modeling of Continuous Industrial Baking Process........ 193

Chapter 10 Improving Bread-Baking Process Under Different Oven Load Conditions

Chapter 11 CFD Applications of Food Packaging Sterilization and Filling............................... 243

Chapter 12 CFD Analysis of Food Pasteurization Processes.........................................................261

Chapter 14 CFD Modeling of Thermal Processing of Particulate Foods.................................... 305

Chapter 15 CFD Modeling of Natural-Convection Heating Processes....................................... 319

Chapter 16 CFD Modeling of Convective Drying of Cylindrical Fruit Slices..............................339

Chapter 17 CFD Modeling of Convection Flow in Pan Cooking................................................ 365

Chapter 18 CFD Analysis of Thermal Processing of Intact Eggs................................................. 389

Chapter 19 Applications of CFD for Optimization of Cabinet Dryers........................................ 415

Section III CFD Applications in Other Food Processes

Chapter 20 CFD Design and Optimization of Biosensors in the Food Industry..........................439

Chapter 23 CFD Application for the Evaluation of Food Texture...............................................509

Chapter 24 CFD Study of Top-Spray Fluidized Bed Coating Process........................................ 531

Filippo Ferrari Jeroen Lammertyn

Nasser Hamdami Filipe J. Mergulhão

Bart Nicolai Noboru Sakai

Giuseppe Vignali Weibiao Zhou

Alexandros Vouros Susana E. Zorrilla

Won Byong Yoon

1.2 THE RETAIL CABINET

• Food should be displayed in the most efficient way to promote selling;

• Closed (in the presence of doors or sliding covers/glasses) or open;

FIGURE 1.1 Vertical, multi-deck display cabinet.

FIGURE 1.2 Horizontal display cabinet.

1.2.2 Standardized Temperature Tests

FIGURE 1.3 A display cabinet during a standardized temperature test.

1.3 APPLICATIONS OF CFD TO DISPLAY CABINETS

1.3.1 Modeling Product Temperature Distribution

1.3.2.3 Evaporator and Rear Ducts

1.3.3 Modeling the Influence of Air Humidity

1.3.4 Modeling Interactions with the Ambient Conditions

1.3.4.2 Ambient Air Movement

1.3.5 Glass Doors Fogging and Defogging

2.2 ENERGY CONSUMPTION AND FOOD SAFETY ASPECTS OF RDC

2.4 AIRFLOW DESIGN AND RESEARCH METHODOLOGY OF RDC

2.5 AIMS AND OBJECTIVES

2.6 EXPERIMENTAL SETUP AND PROCEDURE

2.6.2 Non-Intrusive Experimental Facilities

2.6.2.1 Laser Doppler Velocimetry (LDV)

2.6.2.2 Particle Image Velocimetry (PIV)

2.6.3 General Features of the Flow Field

The normalised velocity half-widths are given by

where Ky is the jet spreading rate and x0 is the virtual origin.

FIGURE 2.6 Schematic view of the plane turbulent jet regions.

2.7 NUMERICAL MODELING APPROACHES

2.7.1 CFD Numerical Modeling

Section II CFD Applications in Thermal

1.2 THE RETAIL CABINET

1.2.2 Standardized Temperature Tests

1.3 APPLICATIONS OF CFD TO DISPLAY CABINETS

1.3.1 Modeling Product Temperature Distribution

1.3.2.3 Evaporator and Rear Ducts

1.3.3 Modeling the Influence of Air Humidity

1.3.4 Modeling Interactions with the Ambient Conditions

1.3.4.2 Ambient Air Movement

1.3.5 Glass Doors Fogging and Defogging

2.2 ENERGY CONSUMPTION AND FOOD SAFETY ASPECTS OF RDC

2.4 AIRFLOW DESIGN AND RESEARCH METHODOLOGY OF RDC

2.5 AIMS AND OBJECTIVES

2.6 EXPERIMENTAL SETUP AND PROCEDURE

2.6.2 Non-Intrusive Experimental Facilities

2.6.2.1 Laser Doppler Velocimetry (LDV)

2.6.2.2 Particle Image Velocimetry (PIV)

2.6.3 General Features of the Flow Field

2.7 NUMERICAL MODELING APPROACHES

2.7.1 CFD Numerical Modeling

2.7.2 Global Modeling Approach

2.7.2.2 Air Curtain Performance

2.8 RESULTS AND DISCUSSION

2.8.1 Without External Perturbation

2.8.1.2 Global Exchanges Through Air Curtain

2.8.2 With External Lateral Flow

2.8.2.2 Turbulence Modeling Performance

2.8.2.3 Effect of the ELF on the Jet Deflection

2.8.2.4 Effect of ELF on the Jet Decay

2.8.2.5 Effect of ELF on the Velocity Profiles

2.8.2.6 Effect of ELF on the Global Exchanges Through Air Curtain

3.2 POSTMORTEM METABOLISM AND MEAT QUALITY

3.3 BEEF CARCASS COOLING METHODS

3.3.1 Cold Air Cooling

3.3.3 Vascular Perfusion Cooling

3.3.4 Cooling Using High Thermal Conductivity Inserts

3.3.5 Cooling Enhanced by Hot Boning

3.4 HEAT AND MASS TRANSFER MECHANISMS DURING

3.4.1 Surface Heat and Mass Transfer

3.4.2 Heat and Mass Transfer Inside the Carcass

3.5 CFD MODELING OF BEEF CARCASS COOLING

3.6 COOLING AIR FLOW CHARACTERISTICS

3.7 COOLING CHARACTERISTICS OF A LARGE BEEF CARCASS

3.7.2 Carcass Temperature Distribution

3.7.3 Carcass Glucose Level

3.8 CFD MODEL APPLICATION IN ASSESSING

3.8.1 Cooling Air Velocity Direction and Magnitude

3.8.2 Cooling Air Temperature

3.8.3 Evaporative vs. Convective Heat Flux

3.8.4 Cooling Room Design

4.2 CFD IN THE HYDROFLUIDIZATION PROCESS

4.3 MATHEMATICAL MODELING OF THE TRANSPORT PHENOMENA

4.3.2 Heat, Mass and Momentum Transfer in the Liquid Medium

4.3.3 Heat and Mass Transfer in the Solid Medium