Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Solar Energy 184 (2019) 417–425

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

On the comparison between computational fluid dynamics (CFD) and T


lumped capacitance modeling for the simulation of transient heat transfer in
solar dryers

Petros Demissie Tegenaw, Mekonnen Gebreslasie Gebrehiwot, Maarten Vanierschot
KU Leuven, Group T Leuven Campus, Mechanical Engineering Technology Cluster TC, A. Vesaliusstraat 13, B-3000 Leuven, Belgium

A R T I C LE I N FO A B S T R A C T

Keywords: In this paper transient heat transfer in a solar food drier is modeled by using both computational fluid dynamics
Solar food drier (CFD) and lumped capacitance modeling techniques. The CFD model is used to simulate the air flow and
Transient heat transfer transient heat transfer within the dryer. A lumped capacitance model, developed from energy exchanges be-
CFD modeling tween various components of the drying chamber, is extensively compared with the results of the CFD model.
Lumped capacitance modeling
The two models predict both a 40 °C rise in steady-state temperature within the drying chamber. Before reaching
this temperature, there is about 8 °C RMS deviation between the transient temperature prediction of the two
models. This deviation is mainly caused by the heat transfer coefficient around the food rack shelfs. Fitting these
coefficients give rise to a minimal RMS deviation between the two models of 1.8 °C and 0.9 °C on the lower and
upper rack shelfs respectively. The transient heat transfer phenomenon from CFD modeling also showed that the
spatial distribution of temperature after 30 min is uniform thus validating the very assumption of lumped ca-
pacitance modeling. This demonstrates that the lumped capacitance model can supplant the computationally
demanding CFD modeling technique in predicting transient heat transfer phenomenon in solar food dryers.

1. Introduction the air velocity field for meat dryers (Mirade, 2003). Another field of
application of this technique is for designing novel heating devices for
In solar food drying, it is necessary to ensure a fast and homogenous infusion fluids in vitrectomy (Mauro et al., 2018).
thermal evolution towards steady operating temperature. This is com- As argued by Dasgupta et al. (1997), Romdhana et al. (2015),
pelled due to the fluctuating nature of solar radiation during the day- transient CFD modeling and simulations require large computational
time. The faster the drying chamber components and the food to be effort and they were not practical to be carried over a large number of
dried reach to equilibrium temperature, the more effective it will be to simulations or a wide range of operating conditions. Thus, significant
use the peak hour solar radiation for drying the food. Thus, design of a number of studies performed on design and development of driers
solar food drier should optimize various components of the drying usually do not involve prediction of transient heat transfer phenomena.
chamber to realize this design goal. This resulted in design decisions to be made merely based on steady
Recently, computational fluid dynamics (CFD) is being utilized in temperature calculations. In the work of Amanlou and Zomorodian
myriads of solar food drier design efforts, proving itself to be a pro- (2010), seven different geometries of cabinet fruit dryers were com-
mising design and modeling tool as a substitute to costly experimental pared for selecting a better concept design. In their work a steady CFD
trials (Amanlou and Zomorodian, 2010; Darabi et al., 2015; Norton modeling was used for predicting temperature distribution within the
et al., 2013; Tegenaw et al., 2017). The technique is successfully used cabinets and the one with the most uniform distribution was chosen as
for predicting distribution of flow velocity and temperature within the best concept design. Here transient temperature distribution was
drying chambers (Norton and Sun, 2006; Rek et al., 2012). It is also not reported which would give better insight in terms of temporal
used to predict drying uniformity of a new design of a commercial tray temperature distribution within the cabinet. Another study on the ap-
dryer for agricultural products by analyzing temperature and velocity plication of CFD on for designing new generation of heating oven
profiles (Misha et al., 2013). The usefulness of CFD for performance steady calculations were employed to predict best possible backing
assessment of food processing applications is highlighted by predicting conditions (Rek et al., 2012). A transient CFD model for predicting


Corresponding author.
E-mail address: maarten.vanierschot@kuleuven.be (M. Vanierschot).

https://doi.org/10.1016/j.solener.2019.04.024
Received 7 March 2018; Received in revised form 9 November 2018; Accepted 5 April 2019
Available online 10 April 2019
0038-092X/ © 2019 International Solar Energy Society. Published by Elsevier Ltd. All rights reserved.
P.D. Tegenaw, et al. Solar Energy 184 (2019) 417–425

Nomenclature turbulence to overall dissipation rate, dimensionless

T temperature, K Greek symbols


p pressure, Pa
v velocity, m s−1 μ viscosity, Pas
t time, s λ heat conductivity, W m−1 K−1
A surface area, m2 α thermal diffusivity, m2 s−1
Q heat flux, W m−2 σ turbulent Prandtl number, dimensionless
E total specific energy, m2 s−2 ε bulk porosity, dimensionless
c1ε , c2ε , and c3ε turbulence model constants, dimensionless ρ density, kg m−3
C heat capacity, J K−1
RMS root mean squared Subscripts
2D two dimensional
r radial distance, m r radial direction
x axial distance, m x axial direction
S momentum source term, N m−3 a air
keff effective conductivity, W m−1 K−1 c solar collector
G generation of turbulence kinetic energy, kg m−1 s−3 w internal wall
Sh energy source term, Pa s−1 s1 lower shelf
h convective heat transfer coefficient, W m−2 K−1 s2 upper shelf
 constant matrix p food product
 stiffness matrix n number of nodes in the food slice
 column vector of temperatures of drying chamber com- κ due to mean velocity gradient
ponents, K b due to buoyancy
YM contribution of fluctuating dilatation in compressible t turbulent

transient heat transfer phenomenon in a commercial forced convention small geometric dimensions and thus did not take into consideration
oven was developed (Verboven et al., 2000). Better qualitative insight convection. It was applied for studying heat and mass transfer of wood
into transient heat distribution were observed. Here also it was reported and peanut.
that the commercial success of CFD in the food processing industry In an effort to simplify the prediction of transient temperature dis-
could be improved by adopting modeling techniques with better com- tribution within an infinitely long cylindrical banana an inverse ap-
putational advantage. proach was proposed (Mariani et al., 2008). In this numerical analysis,
Thus, there is a need for a simpler and less computationally de- transient temperature measurements at the central nodes of a banana
manding transient heat transfer modeling. An appropriate technique for cylinder were used to predict the overall transient heat transfer phe-
this is lumped capacitance modeling, where the dryer is divided into nomenon within the cylinder. This analysis focused on a single banana
many compartments, and integral heat transfer interaction between not a slice of it with a finite length and spread over a rack shelf, which is
them is modeled. In a study by Romdhana et al. (2015), two or three the case in solar food dryers.
compartmental models for simulating drying of spherical solid foods In this paper, a transient lumped capacitance model for a solar food
were made and compared with finite volume models. The compart- drying chamber is developed and extensively compared by detailed
mental model was found to be less computationally demanding yet as simulation of the dryer using a transient CFD model. To the authors
accurate as the finite volume model. In another study, a lumped ca- knowledge, such a model is currently lacking in literature. The com-
pacitance model is formulated to predict thermal boundary conditions parison between the models is illustrated to show that the accuracy of
and source terms in the validation of temperature distribution in a the lumped model is similar compared to the transient CFD simulation.
forced convention oven (Verboven et al., 2000). Another example However, the lumped model reduces the computational time drastically
where lumped models are used is to simplify modeling and simulation compared to CFD. Therefore, it can help in optimizing solar dryers
of dynamic space heating systems with controlled radiators (Xu et al., much faster.
2008).
A simplified heat and mass transfer modeling was proposed for
drying biological materials (Kulasiri and Samarasinghe, 1996). This 2. Materials and properties
model was developed for capillary-porous biological materials with
A truncated pyramid shaped solar sweet potato drier was used in

a) Fan Solar collector


b)

Fig. 1. Schematic representation of the solar food drier: (a) isometric view, (b) partial section view (Tegenaw et al., 2017).

418
P.D. Tegenaw, et al. Solar Energy 184 (2019) 417–425

this study (see Fig. 1). For a detailed description, the reader is referred ∂ ∂ ∂ ⎛ ∂T ⎞
to Tegenaw et al. (2017). Table 1 shows physical and thermal proper- (ρE) + [υi (ρE + p)] = ⎜k eff ⎟ + Sh
∂t ∂xi ∂xj ⎝ ∂xj ⎠ (6)
ties of the sweet potatoe (Ipomoea batatas) (Farinu and Baik, 2007) and
the other materials in the flow domain used in formulating the model. Eqs. (1)–(6) are nonlinear, thus could not be solved explicitly using
closed-form analytical methods. CFD modelling was used extensively
3. Model formulation and solution procedures because of its capability to solve these equations by using numerical
methods to predict the temperature, velocity, and pressure profiles in
3.1. CFD modeling the drying chamber.
It was found from experiments that temperature differences in a
3.1.1. Computational domain typical solar food dryer could be as large as 50 °C (Benhamou et al.,
A 2D axisymmetric simulation flow domain was used in this study 2014). This was too large for the linear Boussinesq approximation to be
(Fig. 2). Relatively thin components of the chamber were excluded from valid (Harral and Boon, 1997; Verboven et al., 2000), thus a weakly
the flow domain because these components do not influence the air compressible formulation of the governing equations was applied. Since
flow significantly. Moreover, this simplification was motivated by the the solar food drier operates at ambient pressure with small pressure
primary interest in this modeling namely the calculation of the airflow fluctuations and the kinetic energy is low compared to the internal
and corresponding heat transfer phenomena at the vicinity of the two energy, the weakly compressible formulation is valid for this study.
rack shelfs.
3.1.3. Rack shelfs and sliced sweet potatoes simulation
3.1.2. Flow and energy equations The sliced sweet potato was assumed to be evenly distributed over
In this work, the thermal and fluid dynamic field in a solar food the rack and is modeled as a porous medium. The pressure drop in the
drier was modeled and simulated to predict the time evolution of flow past this porous media was determined by its viscous and inertial
temperature of the food to be dried, which was placed on the lower and resistance forces represented by the Darcy–Forchheimer (Lage, 1998)
upper rack shelves. The transient flow and heat transfer equations for expression in Eq. (7).
continuity, and momentum conservation are shown in Eqs. (1), (2) and μ
Δp = − v − Cρ |v| v
(3). As the flow domain is axisymmetric these equations were written in α (7)
a cylindrical (x, r) coordinate system.
In Eq. (8) the Forchheimer extension to the law of Darcy was re-
Continuity:
quired when the particle Reynolds number exceeds 1, which corre-
∂ρ ∂ ∂ ρv sponds to most practical situations for food processing applications
+ (ρ vx) + (ρ vr) + r = 0
∂t ∂x ∂r r (1) (Verboven et al., 2004). Porous media are modeled by the addition of a
momentum source term to the standard fluid flow equations. This
Axial (x) momentum:
source term is composed out of two parts: a viscous friction loss term
∂ 1 ∂ 1 ∂ (the first term on the right-hand side of the equation, and an inertial
(ρ vx) + (rρ vx vx) + (rρ vr vx)
∂t r ∂x r ∂r resistance term (the second term on the right-hand side of the equation
∂p 1 ∂ ⎡ ⎛ ∂vx 2 1 ∂ ⎡ ⎛ ∂vx ∂v (ANSYS, 2013).
=− + rμ 2 − (∇ . → v )⎞⎤ + rμ + r ⎞⎤
∂x r ∂x ⎢
⎣ ⎝ ∂x 3 ⎠⎥
⎦ r ∂r ⎢
⎣ ⎝ ∂r ∂x ⎠ ⎥
⎦ μ 1
Si = −⎛ v + C2 ρ |v| v⎞
+ Sx (2) ⎝α 2 ⎠ (8)

Radial (r) momentum: This momentum sink contributes the pressure gradient in the porous
cell, creating a pressure drop as a function of the fluid velocity in the
∂ 1 ∂ 1 ∂ ∂p 1 ∂ ⎡ ⎛ ∂vr ∂v
(ρ vr) + (rρ vx vr) + (rρ vr vr) = − + rμ + x ⎞⎤ cell. To determine the constants (α and C2) in Eq. (8), Ergun’s equa-
∂t r ∂x r ∂r ∂r r ∂x ⎢
⎣ ⎝ ∂x ∂x ⎠ ⎥
⎦ tion was applied with an appropriate shape factor for adopting the
1 ∂ ⎡ ⎛ ∂vr 2 v relation into cylindrically sliced sweet potatoes (Amanlou and
+ rμ 2 − (∇ . →
v ) ⎞ ⎤ − 2μ 2r
r ∂r ⎢
⎣ ⎝ ∂r 3 ⎠⎦⎥ r Zomorodian, 2010; Gaskell, 2012). The porous media simulation in-
2μ corporates a special treatment when it comes to solving the energy
+ (∇ . →
v ) + Sr
3r (3) conservation equation. In most cases of flow in fluid saturated porous
media the assumption of local thermal equilibrium is valid (Nakayama
The k − ε turbulence model remained an industrial standard and its
and Kuwahara, 2005), and this assumption has been used by several
successful applications were reported in recent literature on drying
previous studies (Chourasia and Goswami, 2007; Delele, 2009; Mauro
processes (Amanlou and Zomorodian, 2010; Darabi et al., 2015;
et al., 2018; Nakayama and Kuwahara, 2005). As such, in this model a
Verboven et al., 2003, 2000). It is a semi-empirical model based on
local equilibrium between the air and the porous solid matrix was as-
transport equations for the turbulent kinetic energy (k) and its dis-
sumed.
sipation rate(ε ) as in Eqs. (4) and (5). In this study the standard k − ε
model was used with standard wall functions.
3.1.4. Boundary conditions
∂ ∂ ∂ ⎡⎛ μ ∂k ⎤ The axisymmetric geometry of the flow domain was numerically
(ρk) + (ρkvi) = ⎜μ + t⎞ ⎟ + Gk + G b − ρε − YM + Sk
∂t ∂xi ∂xj ⎢
⎣⎝ σk ⎠ ∂xj ⎥

Table 1
(4) Physical and thermal properties.
and Material Thermal conductivity Specific heat Density
(W m−1 K−1) (J kg−1 K−1) (kg m−3 )
∂ ∂ ∂ ⎡⎛ μ ∂ε ⎤ ε ε2
(ρε ) + (ρε ui) = ⎜ μ + t⎞ ⎟ + c1ε (Gk + c3ε G b) − c2ε ρ
∂t ∂xi ∂xj ⎢
⎣⎝ σε⎠ ∂x ⎥
j⎦ k k Sweet potato 0.49 3660 1212
Air 0.0242 1006 1.225
+ Sε (5) Collector 0.7 750 2880
glazing
The turbulent heat transport was modeled using the concept of Aluminum 222 896 2719
Reynolds’ analogy to the turbulent momentum transfer Eq. (6).

419
P.D. Tegenaw, et al. Solar Energy 184 (2019) 417–425

Axial (x)

Radial (r)

Fig. 2. 2D axisymmetric representation of the solar food drying machine and the boundary conditions employed. (All dimensions in mm).

modeled by means of an unstructured 2D mesh by tetrahedral cells with ∂Tp ∂ ∂Tp


p = λp
a boundary layer near the surfaces. The ambient air will get into the ∂t ∂x ∂x (9)
domain with an inlet velocity of 3 m s−1 and will be heated up by
Note that in Eq. (9), the latent heat of the evaporation of the water
convection on its motion beneath the absorber plate. The air will be
in the food slices is not taken into account as the focus of this paper is
heated when it is going up and down within the collector. This will
on the heat transfer and mass transfer is not taken into account. The
increase the efficiency of the solar collector by 15 percent
convective heat exchange was assumed to obey Newton’s law of
(Satcunanathan and Deonarine, 1973). Therefore, the peak net heat flux
cooling:
collected from the available 5.59 kWh day −1 m−2 will be 918W m−2
(Mahmud et al., 2014). The outlet was a pressure outlet with gauge Qconv = hi Ai (Ti − Ta) (10)
pressure of 0 Pa. The Table 2 summarizes these and other boundary
where hi is the average heat transfer coefficient of the i th component of
conditions applied in the CFD model.
the drying chamber, Ai andTi represent the surface area and surface
temperature of the i th component respectively and Ta , implies the
temperature of the drying air.
3.1.5. Numerical procedure
The drying air within the chamber would experience flow velocities
The numerical simulations were performed in ANSYS fluent 17.1.
of different magnitude at each point within the chamber. Thus, the
The SIMPLEC algorithm was used as a numerical solution procedure. As
expected convective heat transfer will be a combination of natural and
initial condition for the transient simulations, a converged cold flow
forced convection modes. To determine which mode is dominant, the
field without the energy equation was used. This was done to have
ratio between the Grashoff number from the natural convection and
accurate initial conditions which are just as important as boundary
Reynolds number from the forced convection was used (Çengel and
conditions for transient problems. The initial conditions were also ex-
Ghajar, 2015), this is summarized in Table 3.
pected to be physically realistic when it comes to a transient simulation.
The convective heat transfer coefficient between the food to be
Then, appropriate source terms of heat fluxes were applied on re-
dried and the drying air was calculated by using the correlation from
spective walls of the flow domain. The simulation was performed with
literature (Achenbach, 1995).
gradual increments in the time step, i.e. initially using very small-time
The heat exchanges per component i of the solar drier can be written
step size and then gradually increasing it as the flow reached a steady
as follows:
state solution. A convergence criterion was set for all residuals of
n
pressure, velocity, k − ε turbulence terms, and energy to be less than dTi
10−6 . The mesh of the flow domain was significantly refined to accu- i = ∑ Qi ± hijAi (Ti − T)j
dt i=1 (11)
rately resolve the flow and thermal boundary layers. A mesh refinement
study was carried out with successively smaller mesh sizes up to Ci is the heat capacity, Q i an external heat source of thei th com-
158,664 cells, this mesh was selected because the result from it and ponent and hij the convective heat transfer coefficient of the i th com-
from the most refined mesh is differing by less than 1%. Furthermore, to ponent with the jth component in Eq. (11).
capture the near wall flow parameters inflation layers were applied Boundary conditions on both sides of the food slice(p) are given in
near the walls, see Fig. 3. (12b) and (12c). The initial temperature conditions of each arbitrary
node m (starting from 0 to M) is given by (12a). The finite difference
formulation applied on the food slice of thickness D is represented in
3.2. Lumped capacitance modeling Fig. 5.

A lumped capacitance model was adopted whereby overall heat Table 2


exchanges between the various components of the solar food drying CFD modeling boundary conditions.
machine are taken into account, as shown in Fig. 4. Boundary condition Description
In the lumped model, a uniform spatial distribution of temperature
per component was assumed. As such, the components of the drying Inlet-velocity inlet An inlet velocity of 3 m s−1 and a temperature of 300 K.
chamber were designed so that they have good heat conduction prop- Direction of air flow was normal to the boundary
Outlet-pressure outlet A fixed gauge pressure of 0 Pa and zero gradients in the
erties and a relatively small thickness. The food to be dried, on the other
normal direction of the boundary
hand, does not possess this property thus its temperature varies in Axis boundary The axis boundary is used as the centerline of the domain
space. Therefore, the heat transfer within the food slices was calculated Glass cover wall A heat flux input of 918 W m−2
using a one-dimensional finite element heat conduction model Eq. (9).

420
P.D. Tegenaw, et al. Solar Energy 184 (2019) 417–425

Fig. 3. Partial computational grid with inflation near the walls.

T 0m = 300K (12a) Table 3


Determination of dominant convection heat transfer modes on various com-
∂Tp ponents of the drying chamber.
λp = hp Ap (Ta − Tp)
∂x x=0 (12b) Solar collector Internal wall Upper rack Lower rack
shelf shelf
∂Tp Outside air Inside air Outside air Inside air
λp = hp Ap (Ta − Tp)
∂x x=D (12c) Forced Forced Combined Forced Combined Combined
The resulting equations were combined and rearranged in the fol-
lowing matrix notation:
d
 =∗ +
dt (13)
with  = [Ta , Tc, Tw , Ts1, Ts2, Tp10, ⋯, Tp1n, Tp20, ⋯, Tp2n], which re-
presents the temperatures of the drying chamber constituents and the
nodal temperatures at discrete locations in the food. The matrix equa-
tion (13) was implicitly solved using the MATLAB® solver.

4. Results and discussion

Fig. 5. Schematic representation of the finite difference formulation of speci-


4.1. Cold flow simulation
fied boundary conditions on both sides of a sliced food product to be dried.

The flow velocity distribution near the rack shelfs i.e. upper and
lower from cold flow simulation without heat transfer is shown in 2.15 cm s−1 and 4.83 cm s−1 respectively. Simulation of air flow
Fig. 6. The mean velocity near to the upper rack shelf doubles as within a drying chamber was investigated by Misha et al. (2015) and
compared to that of the lower rack shelf. This is a result of the drying from the report there is a linear relationship between the air flow ve-
chamber design which converges towards a relatively smaller outlet locity and drying rate. Thus, the current result on the lower rack shelf
cross section (Fig. 7). Moreover, this is desirable as the drying air is the drying rate will be slower and there will be homogeneous drying. In
picking moisture on its way up within the drying chamber, it is required contrary, on the upper rack shelf there will be fast drying rate but less
to have a larger flow velocity close to the outlet since the deceleration homogeneity in the drying. Generally, both on the lower and upper rack
of the flow due to the mass transfer (moisture pick-up) needs to be shelfs there is a significant flow velocity variability along the length of
compensated. The magnitude of flow velocity deviation from the mean the rack shelf. This will cause a variation in moisture content of the
flow velocity on the lower rack shelf and upper rack shelf is about final product (Amanlou and Zomorodian, 2010; Mathioulakis et al.,

Fig. 4. Schematic representation of the heat exchanges between components of the solar food drying chamber.

421
P.D. Tegenaw, et al. Solar Energy 184 (2019) 417–425

Fig. 6. Flow velocity distribution around the food (rack shelfs). (a) Velocity distribution near upper rack shelf; (b) velocity distribution near the lower rack shelf.

1998). In practice this may be solved by putting product of relatively


low moisture in the areas of low air velocities. The flow velocity
magnitude close to the tips of both rack shelfs is changing curvature to
an increasing trend because of less resistance to the flow from the gap
between the internal wall and the tip of the rack shelf.

4.2. Lumped capacitance modeling and CFD simulation with heat transfer

4.2.1. CFD modeling prediction


Fig. 8 shows the time evolution of temperature on the lower rack
shelf and distribution of temperature within the entire flow domain at
various flow times. It can be observed that the entire flow domain
evolves to steady operating temperature after about 1 h. After this time
on the average temperature of the drying chamber will be uniform and
340 K. That means 40 °C increase in temperature which is also the
value of temperature rise reported by (Kumar et al., 2016) who pre-
sented various progresses in solar dryers for drying various commod-
ities.
Fig. 8. Food temperature evolution and contour plot as predicted using the CFD
model.

4.2.2. Lumped capacitance modeling prediction


Fig. 9 shows the temperature evolution of the various solar food 4.3. Comparison of CFD model and lumped capacitance model in predicting
drying machine components as predicted from the lumped capacitance food temperature evolution
model. The door is designed to be constructed from aluminum sheet
metal, thus its thermal resistance is very low. Thus, it stabilizes at a Fig. 10 shows comparison between CFD model and lumped capa-
lower temperature compared to the other components. The slower citance model for the food on the lower rack shelf and upper rack shelf.
temperature evolution of the food is due to the higher heat capacity The RMS deviation between the CFD model and the lumped capacitance
associated with it, i.e. sweet potato in this case. model prediction is about 8 °C and 6 °C at each point in time in the

Fig. 7. Contour plot of flow total velocity distribution within the solar food drying chamber.

422
P.D. Tegenaw, et al. Solar Energy 184 (2019) 417–425

modified heat transfer coefficients showed 1.8 °C on the lower rack


shelf and 0.85 °C on the upper rack shelf RMS deviation at each point in
time as compared to the CFD model (Fig. 11).

4.4. Comparison of the models on prediction of the time evolution of


temperature on to the lower and upper rack shelfs

The food temperature prediction using CFD model and lumped


model on the upper and lower rack shelves is shown in Figs. 12 and 13.
Both models predict that the temperature rise on the food to be around
40 °C. The CFD models shows that the food placed on the upper rack
shelf heated up faster than that placed on the lower rack shelf. The time
it takes for the food placed on the lower and upper rack shelf to reach to
50% of the final steady temperature value is 21 and 12 min respec-
tively. This can be explained using the results from flow velocity dis-
tribution shown on Fig. 6, there the vicinity of the upper rack shelf is at
relatively higher mean flow velocity, thus higher heat transfer to the
food. In addition, the result from the transient CFD model simulation
shows on average about 3 °C deviation at each point in time between
Fig. 9. Temperature evolution of various components of the solar food drier
the temperature of the food placed on the upper and lower rack shelfs.
using the lumped capacitance model.
The lumped model also predicts the heating up phenomenon but due to
its assumption i.e. temperature only varies with time, it is not possible
lower and upper rack shelfs respectively. In the discretization of the to see variation of temperature evolution with space, see Fig. 13. From
lumped model the food slice is divided into much less number of nodes the CFD result the observed 3 °C average deviation is not that sig-
as compared to the second order discretization scheme used in the CFD nificant; thus, one can conclude that the assumptions of the lumped
model. This results in higher heat capacities between nodes in the model are valid.
lumped model, thus less temperature rise at same time as compared to Moreover, the transient heat transfer phenomenon predictions of
the CFD. The overall warm up time and corresponding temperature rise the two models are studied on the internal wall of the drying chamber
prediction is the same in both models. see Fig. 14. Both of the models predicted the same rise in temperature
The deviation in the results of the two models in predicting the time on the internal wall with similar trend in the transient characteristics.
evolution of temperature on the food is observed in the transient period
i.e. before coming to a uniform temperature of 340 K. Further in- 5. Conclusion
vestigation is done on the parameters of the lumped model specifically
those used as heat transfer coefficients on various surfaces of the drying In this paper, the transient heat transfer phenomenon within a solar
chamber. The result showed those heat transfer coefficients used for the food drying chamber was modeled and simulated. The flow of the
food on the rack shelfs are highly influencing the overall transient drying air within a 2D axisymmetric domain was calculated using a
phenomenon. Fig. 11 shows the deviation observed between the two k − ε turbulence model with appropriate boundary and initial condi-
models as a function of the heat transfer coefficient on the food placed tions applied. The food to be dried, in this case sweet potato and the
on the rack shelfs. The RMS error is minimal at the heat transfer rack shelfs (lower and upper) were modeled as porous media with
coefficients 30 and 58 W m−1 K−1 on the surface of the food placed on calculated inertial and viscous resistance coefficients. The result from
the lower and upper rack shelfs respectively. This value is 10 times the variations of the flow velocity magnitudes at the vicinity of the two
larger than the one calculated by using the available correlations of rack shelfs was used to explain the variation in the rate of thermal
Achenbach (1995). The lumped capacitance model prediction with the evolution towards steady operating temperature.

Fig. 10. Comparison of lumped model with CFD simulation results on the food placed on the rack shelf.

423
P.D. Tegenaw, et al. Solar Energy 184 (2019) 417–425

Fig. 11. Comparison of lumped model with CFD model results: (top) the RMS deviations observed between the two model predictions of food temperature; (bottom)
comparison between the two models after modification of the heat transfer coefficients for minimized error;

Fig. 12. Time evolution of area weighted average temperature on the two rack Fig. 13. Time evolution of temperature on the two rack shelfs using Lumped
shelfs using CFD model. capacitance model.

The aim of the present research was to develop a model for pre- simulated. Its results were rigorously validated by simulated results of a
dicting transient heat transfer phenomenon that can supplant the CFD model for the same drying chamber. The lumped model resulted in
computationally arduous CFD modeling. For this, a lumped capacitance a similar steady operating temperature of about 60 °C within the
model for a particular solar food drying chamber was developed and chamber and on the food placed on the rack shelfs. The lumped model
was further modified to minimize a deviation observed in the predicted

424
P.D. Tegenaw, et al. Solar Energy 184 (2019) 417–425

in vertical ducts. Ind. Eng. Chem. Res. 36, 3375–3390.


Delele, M.A., 2009. Engineering design of spraying systems for horticulturalapplications
using computational fluid dynamics. Status Publ 220.
Farinu, A., Baik, O.-D., 2007. Thermal Properties of Sweet Potato with its Moisture
Content and Temperature. Int. J. Food Prop. 10, 703–719.
Gaskell, D.R., 2012. An Introduction to Transport Phenomena in Materials Engineering,
second ed. Momentum press, New Jersey.
Harral, B.B., Boon, C.R., 1997. Comparison of predicted and measured air flow patterns in
a mechanically ventilated livestock building without animals. J. Agric. Eng. Res. 66,
221–228.
Kulasiri, D., Samarasinghe, S., 1996. Modelling heat and mass transfer in drying of bio-
logical materials: a simplified approach to materials with small dimensions. Ecol.
Modell. 86, 163–167.
Kumar, M., Sansaniwal, S.K., Khatak, P., 2016. Progress in solar dryers for drying various
commodities. Renew. Sustain. Energy Rev. 55, 346–360.
Lage, J.L., 1998. The fundamental theory of flow through permeable media from darcy to
turbulence. Transp. Phenom. Porous Media. Elsevier 1–30.
Mahmud, A.M., Kahsay, M.B., Hailesilasie, A., Hagos, F.Y., Gebray, P., Kelele, H.K.,
Gebrehiwot, K., Bauer, H., Deckers, S., De Baerdemaeker, J., Driesen, J., 2014. Solar
energy resource assessment of the geba catchment, Northern Ethiopia. Energy Proc.
57, 1266–1274.
Mariani, V.C., Barbosa de Lima, A.G., dos Santos Coelho, L., 2008. Apparent thermal
diffusivity estimation of the banana during drying using inverse method. J. Food Eng.
85, 569–579.
Fig. 14. Comparison between lumped capacity model and lumped model on the Mathioulakis, E., Karathanos, V.T., Belessiotis, V.G., 1998. Simulation of air movement in
a dryer by computational fluid dynamics: application for the drying of fruits. J. Food
temperature evolution prediction on internal wall.
Eng. 36, 183–200.
Mauro, A., Massarotti, N., Salahudeen, M., Cuomo, F., Costagliola, C., Ambrosone, L.,
Romano, M.R., 2018. Design of a novel heating device for infusion fluids in vi-
evolution towards this temperature. This modification resulted in
trectomy. Appl. Therm. Eng. 128, 625–636.
minimal prediction deviations: about 1.8 °C and 0.9 °C in the lower Mirade, P., 2003. Prediction of the air velocity field in modern meat dryers using un-
and upper rack shelfs respectively. The insights gained from this study steady computational fluid dynamics (CFD) models, vol. 60, pp. 41–48.
Misha, S., Mat, S., Rosli, M.A.M., Ruslan, M.H., Sopian, K., Salleh, E., 2015. Simulation of
may be of assistance to model transient heat transfer phenomenon in a
Air Flow Distribution in a Tray Dryer by CFD, pp. 29–34.
less computationally demanding way. Moreover, a lumped capacitance Misha, S., Mat, S., Ruslan, M.H., Sopian, K., Salleh, E., 2013. The prediction of drying
model might be used in parametric studies and dynamic modeling of uniformity in tray dryer system using CFD simulation. Int. J. Mach. Learn. Comput. 3,
solar food dryers for design and optimization. 419–423.
Nakayama, A., Kuwahara, F., 2005. Algebraic model for thermal dispersion heat flux
within porous media. AIChE J. 51, 2859–2864.
Acknowledgement Norton, T., Sun, D.-W., 2006. Computational fluid dynamics (CFD) – an effective and
efficient design and analysis tool for the food industry: a review. Trends Food Sci.
Technol. 17, 600–620.
The Interfaculty Council for Development Cooperation (IRO) of KU Norton, T., Tiwari, B., Sun, D.-W., 2013. Computational fluid dynamics in the design and
Leuven is gratefully acknowledged for funding this research. analysis of thermal processes: a review of recent advances. Crit. Rev. Food Sci. Nutr.
53, 251–275.
Rek, Z., Rudolf, M., Zun, I., 2012. Application of CFD simulation in the development of a
References new generation heating oven. J. Mech. Eng. 58, 134–144.
Romdhana, H., Lambert, C., Goujot, D., Courtois, F., 2015. Model reduction technique for
Achenbach, E., 1995. Heat and flow characteristics of packed beds. Exp. Therm. Fluid Sci. faster simulation of drying of spherical solid foods. J. Food Eng. 170, 125–135.
10, 17–27. Satcunanathan, S., Deonarine, S., 1973. A two-pass solar air heater. Sol. Energy 15,
Amanlou, Y., Zomorodian, A., 2010. Applying CFD for designing a new fruit cabinet 41–49.
dryer. J. Food Eng. 101, 8–15. Tegenaw, P.D., Gebrehiwot, M.G., Vanierschot, M., 2017. Design and CFD Modeling of a
ANSYS, 2013. ANSYS FLUENT User’s Guide. ANSYS FLUENT User’s Guid. 15317, 2498. Solar Food Drier. In: EuroDrying. 6th European Drying Conference, pp. 183–184.
Benhamou, A., Fazouane, F., Benyoucef, B., 2014. Simulation of solar dryer performances Verboven, P., Datta, A.K., Anh, N.T., Scheerlinck, N., Nicolï, B.M., 2003. Computation of
with forced convection experimentally proved. Phys. Proc. 55, 96–105. airflow effects on heat and mass transfer in a microwave oven. J. Food Eng. 59,
Çengel, Y.A., Ghajar, A.J., Afshin, J., 2015. Heat and Mass Transfer: Fundamentals and 181–190.
Applications. p. 968. Verboven, P., Hoang, M.L., Baelmans, M., Nicolaï, B.M., 2004. Airflow through beds of
Chourasia, M.K., Goswami, T.K., 2007. CFD simulation of effects of operating parameters apples and chicory roots. Biosyst. Eng. 88, 117–125.
and product on heat transfer and moisture loss in the stack of bagged potatoes. J. Verboven, P., Scheerlinck, N., De Baerdemaeker, J., Nicolai, B.M., 2000. Computational
Food Eng. 80, 947–960. Fluid Dynamics modelling and validation of the temperature distribution in a forced
Darabi, H., Zomorodian, A., Akbari, M.H., Lorestani, A.N., 2015. Design a cabinet dryer convection oven. J. Food Eng. 43, 61–73.
with two geometric configurations using CFD. J. Food Sci. Technol. 52, 359–366. Xu, B., Fu, L., Di, H., 2008. Dynamic simulation of space heating systems with radiators
Dasgupta, S., Jackson, R., Sundaresan, S., 1997. Developing flow of gas-particle mixtures controlled by TRVs in buildings. Energy Build. 40, 1755–1764.

425

You might also like