Download as pdf or txt
Download as pdf or txt
You are on page 1of 48

Comput Math Organ Theory

DOI 10.1007/s10588-017-9247-3

MANUSCRIPT

A comprehensive approach in designing a sustainable


closed-loop supply chain network using cross-docking
operations

Saeid Rezaei1 • Amirsaman Kheirkhah1

 Springer Science+Business Media New York 2017

Abstract In today’s competitive world, it is unavoidable to provide a new efficient


approach in the cycle of production and supplying. The problem of designing a
supply chain network is included in strategic decisions in this area, and short-term
changing the structure and configuration of logistics networks is almost impossible
due to assigned much time and cost. This paper develops a comprehensive model
for designing a sustainable closed-loop supply chain network based on economic,
environmental and social requirements, both with applying cross-docking opera-
tions in the mentioned network. Utilizing the cross-docking system—as a new
strategy of supply chain—along with simultaneous considering the above triple
dimensions—economic, environmental and social requirements—in a comprehen-
sive and sustainable approach offers a novel research scope in the wide range of
problems related to supply chain network design, and in this regards, helps orga-
nizations improve their competitive advantage in different industries. For these
reasons, in this study, a multi-objective mixed-integer linear programming model is
developed, and in order to solve this NP-hard problem, a cuckoo optimization
algorithm is utilized—as the first attempt in this area. Finally, to test the efficiency
of the proposed metaheuristics, it is compared with other strong algorithms illus-
trating a quite well performance.

Keywords Closed-loop supply chain network  Environmental requirements 


Social responsibility  Cross-docking operations  Sustainability  Multi-objective
cuckoo search (MOCS) algorithm

& Saeid Rezaei


saeidrezaei7ie@gmail.com
1
Department of Industrial Engineering, School of Engineering, Bu-ali Sina University,
Hamedan, Iran

123
S. Rezaei, A. Kheirkhah

1 Introduction

Nowadays, accurate delivery of products in ordered time and cost plays not only an
important role in competitive success of an organization, but also is considered a
key element in its survival. Therefore, in order to utilize a new strategy of supply
chain, recognition of customers’ satisfaction features and market restrictions are
critical factors in this area, thereby, an organization is able to provide, elicit and
develop an applicable strategy to meet the supply chain needs.
The immense problem of supply chain design contains extensive issues related to
the supply chain operations in triple dimensions of strategic (logistics network
configuration and facilities locating), tactical (coordination of decisions related to
inventory and transportation) and operational (production scheduling, determination
of delivery type and vehicles routing) decisions (Bender et al. 2002; Simchi-Levi
et al. 2004). The terms ‘‘Network Design (ND)’’ and ‘‘Supply Chain Network
Design (SCND)’’, used in a single concept in associated literatures, are synonyms
for strategic subset of the supply chain design (Chopra and Meindl 2007; Meixell
and Gargeya 2005; Simchi-Levi and Kaminsky 1999). Generally, SCND encom-
passes the determination of locations, numbers, facilities capacity and also material
flow through the network, and all these matters have significant impacts on
flexibility, efficiency and also performance of supply chains.
While for many years, minimizing the total cost or maximizing the profit were
the main objectives of supply chains, but today, they are responsible for the
environmental and social impacts of their operations (Rogers and Tibben-Lembke
1998). Increasing attention to the reverse logistics (RL) and closed-loop supply
chain (CLSC) was first originated from the society and people awareness, and then,
governments’ legislations forced the producers to utilize the used products. This
kind of attitude towards supply chains made the RL/CLSC be an income
opportunity not just a cost reduction approach for manufacturers. Indeed, based
on the new considerations, closed-loop supply chain management includes
designing, controlling and implementation of an integrated system for the reason
of maximizing the earned value of a product throughout its life cycle by ongoing
retrieving the value from different volumes of returned products.
With reviewing the associated literatures, in can be acknowledged that most of
papers have considered forward or reverse logistics separately, and few researches
have been oriented towards sequential integrated supply chain networks design. In
today’s viewpoint, designing an individual supply chain network based on forward
processes not only results in ignoring sustainability aspects, but rather, causes a
declined market share and competitive advantage. Moreover, SCND has been
generally formulated in a single objective programming model leading to an
inefficient solution for covering the needs of the real world.
According to the aforementioned considerations, this paper addresses the
problem of designing a sustainable closed-loop supply chain network along with
applying cross-docking operations. The proposed study can be preferable and
distinguished from previous researches due to considering the sustainability and
cross-docking operations making the presented comprehensive approach more

123
A comprehensive approach in designing a sustainable…

indispensable. The sustainability of this thorough network is achieved by


considering environmental and social requirements in the network design. In this
regards, firstly, all the environmental and social aspects of SCND are quantified
beside the economic dimensions in a three-objective model. Secondly, the cross-
docking operational system is utilized in the mentioned network configuration
illustrating the other prominent aspect of our study.
Furthermore, to solve this problem, belonging to NP-Hard class of the
optimizations (Devika et al. 2014; Soleimani and Kannan 2015), we have developed
an efficient metaheuristic called multi-objective cuckoo search (MOCS) algo-
rithm—as the first attempt in SCND problems—and compared it with two strong
metaheuristics of multi-objective imperialist competitive algorithm (MOICA) and
the multi-objective particle swarm optimization (MOPSO) method to find the
Pareto-optimal solutions.
The remainder of the paper is organized as follows. In Sect. 2, all the researches
carried out in the associated literature are classified and reviewed. In Sect. 3, the
problem is described in whole details, and the mixed-integer linear programming
(MILP) model is presented through objective functions and constraints. Section 4 is
devoted to solution approach and other relevant contents. In this regard, the applied
multi-objective technique and also the considered metaheuristic algorithm are
explained. In this section, the obtained results of the proposed algorithm are
presented to investigate its performance against the other stated metaheuristics in
terms of different criteria. Finally, in Sect. 5, the main and overall conclusions and
also the offers for future researches are described.

2 Background and literature review

Due to increasing importance of sustainability aspects in supply chains, the recent


studies have directed towards the design of closed-loop networks. Indeed, the
problem of SCND has been followed in a developing procedure during the different
years through which the applicability aspect of the network has been considerably
improved. The initial point of this procedure includes the classical configuration of
the supply chain networks only containing suppliers, manufacturers of the end
products, distribution centers, retailers and the products customers (Fig. 1). In this

Suppliers Plants Distribution Customers


centers

Fig. 1 Classical configuration of supply chain network design (Melo et al. 2008)

123
S. Rezaei, A. Kheirkhah

structure, the locating decisions of SCND are based on fixing existing facilities, and
then, adding new ones in the network selected from a predetermined candidates set.
The other strategic decisions of this area consist of procurement, manufacturing,
distribution, capacity development and satisfying customers’ demands, and in this
regard, the bill of material (BOM) is flowed through the network for raw materials,
semi-manufactured and the end products (Melo et al. 2008).
As previously mentioned, the general inputs of SCND process encompass the
customer zones to be satisfied, set of products, assessment of the demands in
different zones, the information related to future conditions, manufacturing and
transportation costs, manufacturing resources and the products flow.
As there are lots of articles conducted in different aspects of SCND, the
associated papers have been appropriately classified—based on the main features
demonstrated in Table 1—and reviewed in Table 2.
Developing procedure in logistics network design indicates that merely
considering the economic aspects in SCND cannot meet the today’s needs, and
defining the other dimensions (such as environmental and social aspects)—known
as sustainability dimensions of supply chains—and reflecting them in associated
models, while becoming a necessity, are critical for competitiveness of
organizations.
In modern structure of supply chain network design, the forward and reverse
logistics are considered in an integrate platform—rather than separate and
sequential modeling—as a closed-loop network (Pishvaee et al. 2010). This
developed structure, which is based on sustainability dimensions of network design,
has been achieved gradually (and) by considering other aspects and constraints in
the primal and classical models of SCND (such as constraints related to
environmental impacts and also the ones associated with the network echelons
especially in the reverse flow).

2.1 Supply chain network design

The problem of supply chain network design comprises three subsets of forward,
reverse and closed-loop networks, and in this regards, those of closed-loop have
attracted much more attentions due to the sustainability importance.
In this way, Faccio et al. (Faccio et al. 2014) studied the new approach of social
responsibility against the classical view in SCND through the complete reprocessing
of used products. They proposed a linear programming model to minimize the total
cost in a supply chain, and then, assessed the economic sustainability of the model
by parametric analysis in comparison to the classical forward supply chain
formulation. Pointing to the consideration of disposal costs in the social
responsibility approach, they utilized the statistical variance analysis and the Pareto
diagram in this comparison.
Cardoso et al. (2013) considered a supply chain network design in terms of
potential demands in their study in which the reverse logistics was also included
simultaneously. Their proposed supply chain network comprised four echelons of
factories, warehouses (in which the end products are assembled and stored), retailers
and the market. They tried to maximize the net present worth of products through

123
A comprehensive approach in designing a sustainable…

Table 1 Coding of SCND papers (Devika et al. 2014)


Aspect Subset Type Code

Type of network Forward F


Reverse R
Closed-loop CL
Objectives Min cost/max profit C
Min environmental impacts EI
Max social benefits S
Modeling Continuous Continuous approximation CA
Discrete Stochastic mixed integer programming SMIP
Fuzzy mixed integer programming FMIP
Mixed integer non-linear programming MINLP
Mixed integer linear programming MILP
Network stages Forward logistics stages Supply centers SC
Production centers PC
Distribution centers DC
Cross-docks CD
Warehouses WH
Customer zones (retail outlets) CZ
Reverse logistics stages Collection/inspection centers CIS
Dismantlers DSM
Disassembly centers DAC
Redistribution centers RDC
Recovering centers RCC
Remanufacturing centers RMC
Recycling centers RYC
Disposal/incineration centers DPC
Solution method Exact E
Branch and bound B&B
Lagrangian relaxation-based LR
Genetic algorithm-based GA
Simulated annealing-based SA
Tabu search-based TS
Interactive fuzzy solution approach F
Others heuristics H
Outputs Suppliers/orders S
Facilities location L
Facility capacity FC
Allocation AL
Production amount PQ
Production assignment to production centers PA
(for multi-product problems)
Utilization of production centers (i.e. days UT
allocated for the production of products)

123
S. Rezaei, A. Kheirkhah

Table 1 continued

Aspect Subset Type Code

Production technology PT
Transportation amount TA
Transportation mode TM
Number of vehicles NV
Inventory I
Price of products P
Treatment (i.e. disposal, recycling and etc.) TT
technology
Amount of used products which are US
processed
Amount of Incentive ordered (acquisition INC
price) for a used product
Carbon credits sold/purchased CC
Quantity of non-satisfied demand ND
Quantity of non-satisfied return of product NR

the reverse logistics in which the used products were collected and stored by
retailers from the market. Ramezani et al. (2014) investigated the use of fuzzy sets
in designing a multi-period multi-product closed-loop supply chain network. They
considered a triple-objective formulation based on maximizing profit, minimizing
delivery time and maximizing the quality, and in this regards, utilized the fuzzy
approach in constraints, coefficients and objectives.
Regarding the main triple considerations in SCND, Mota et al. (2014) criticized
the lack of study in applying the economic, environmental and social dimensions
simultaneously. In this regards, they proposed a general multi-objective model by
which the social metrics were investigated.

2.1.1 Different approaches in modeling SCND

In order to achieve a comprehensive view of supply chain network design, almost


400 scientific papers have been surveyed and reviewed in different aspects of
modeling, solving methods and planning approaches. Accordingly, there have been
various modeling approaches in the papers of SCND classified in 13 categories:
1—conceptual models, 2—linear and integer programming, 3—non-linear
programming, 4—convex and concave programming, 5—dynamic programming,
6—queuing models, 7—Markov decision process, 8—graph theory, 9—game
theory, 10—fuzzy theory, 11—simulation models, 12—multi-criteria decision
approaches, 13—other approaches such as Neural networks (Mazhar et al. 2007),
discrete programming (Zhang et al. 2011), dynamic regression models (Gallego and
Cueto 2009), statistical models (Pati et al. 2010), Bayesian networks (Shevtshenko
and Wang 2009) and engineering economy techniques (Krikke 2010).

123
A comprehensive approach in designing a sustainable…

Table 2 Review of previously published literature (Devika et al. 2014)


Reference papers Objectives Network stages Modeling Outputs Solution
method

Type of network: forward


Jayaraman and C SC, PC, WH, CZ MILP L, AL, PQ, LR
Pirkul (2001) TA
Dasci and Verter C PC, CZ CA L, AL E
(2001)
Jayaraman and C PC, CD, WH MILP L, AL, TA SA
Ross (2003)
Miranda and C PC, WH, CZ MINLP L, AL, TA, I LR
Garrido (2004)
Wang et al. C, EI SC, PC, CZ MILP TA E
(2011)
Yeh (2005) C SC, PC, DC, CZ MILP L, TA H
Georgiadis et al. C PC, DC, WH, CZ MILP L, AL, TA, I, B&B
(2011) PQ
Tsiakis and C PC, DC, CZ MILP L, Al, FC, E
Papageorgiou TA, PA, UT
(2008)
Pishvaee et al. C, EI PC, DC, CZ FMIP L, AL, PT, F
(2012b) TM, FC, PQ
Syarif et al. C SC, PC, DC, CZ MILP L, TA, SS, PQ GA
(2002)
Elhedhli and C, EI PC, DC, CZ MINLP L, AL, TA LR
Merrick
(2012)
Pishvaee et al. C, EI, S PC, DC, CZ FMIP L, TA, PT E
(2012a)
Type of network: reverse
Krikke et al. C CZ, CIC, RDC, RCC MILP AL, TA E
(1999)
Jayaraman and C CZ, CIC, RCC MILP L, TA H
Ross, (2003)
Min and Ko C CZ, CIC, RCC MINLP L, AL, I GA
(2008)
Listeş and C CZ, CIC, RYC SMIP L, TA E
Dekker (2005)
Aras et al. C CZ, CIC MINLP L, AL, TA, TS
(2008) NV, INC
Fonseca et al. C, EI CZ, CIC, RCC, DPC SMIP L, TA, TT E
(2010)
Kannan et al. C, EI CZ, CIC, RCC, DPC MILP L, TA E
(2012)
Min et al. (2006) C CZ, CIC MINLP L, AL, TA, I GA
Dehghanian and C, EI, S CIC, RYC, DPC MILP L, TA, FC, GA
Mansour TT
(2009)

123
S. Rezaei, A. Kheirkhah

Table 2 continued

Reference papers Objectives Network stages Modeling Outputs Solution


method

Cruz-Rivera and C CZ, CIC MILP L, AL, TA E


Ertel (2009)
Type of network: closed-loop
Marin and C PC, CZ, RMC MILP L LR
Pelegrı́n
(1998)
Fleischmann C PC, WH, CZ, CIC, RMC, MILP L, AL E
et al. (2001) DPC
Lu and Bostel C PC, CZ, CIC, RMC, DPC MILP L, AL LR
(2007)
Salema et al. C PC, WH, CZ, DAC, RMC MILP L, AL, PQ, I, B& B
(2009) TA, ND
Ko and Evans C PC, WH, CZ, RMC MINLP L, TA, FC GA
(2007)
Demirel and C PC, DC, CZ, CIC, DAC MILP L, PQ, TA GA
Gökçen (2008)
Pishvaee and C, EI PC, CZ, CIC, RCC (Steel), FMIP L, TA, PQ, F
Razmi (2012) RCC (Plastic), DPC US
Chaabane et al. C, EI SC, PC, DC, CZ, RYC MILP L, PQ, I, PT, E
(2012) TM CR
Pishvaee et al. C CZ, CZ, CIC, RDC, RCC, SMIP L, TA, ND E
(2011) DPC
Wang and Hsu C SC, PC, DC, CZ, DSM, MILP L, PQ, TA GA
(2010) RMC, RDC, DPC
Devika et al. C, EI, S SC, PC, DC, CZ, CIC, RDC, MILP S, L, AL, PQ, H
(2014) RCC, RMC, RYC, DPC PT, TA, US

There have been also utilized different methodologies for solving RL/CLSC
problems in associated papers classified as follows:
1—Exact methods (that are complicated in case of large scale problems), 2—
general solvers such as LINGO, GAMS or CPLEX, 3—approximate mean sampling
approaches (for stochastic optimization problems), 4—heuristics approaches, 5—
metaheuristics algorithms (such as GA, SA, TS and ACO), 6—simulation methods
and 7—multi-objective/multi-criteria solving methodologies (such as GP, AHP,
ANP and TOPSIS).
Figures 2 and 3 show the distribution of different modeling approaches and also
aforementioned solving methods in SCND problems (in the reviewed papers).

2.1.2 Specific approaches to planning SCND

There are also different procedures about single/multiple objective, single/multiple


product and single/multiple period approaches in SCND scientific papers. The

123
A comprehensive approach in designing a sustainable…

35.00%
30.00%
25.00%
20.00%
15.00%
10.00%
5.00%
0.00%

Fig. 2 Distribution of different modeling approaches in the field of SCND

40.00%
35.00%
30.00%
25.00% 33.50%
20.00% 26.20%
15.00%
10.00% 15.20%
5.00%
0.00% 8.60% 8.60%
4.50% 2.60% 0.80%

Fig. 3 Distribution of utilized solution methodologies in the papers included in SCND

Single objecve
Mul objecve
40

30

20

10

0
2007 2008 2009 2010 2011 2012 2013 2014

Fig. 4 Utilization of single and multiple objective problems in different years

figure below (Fig. 4) illustrates the utilization trend of single or multiple objective
approaches in various years of study (single objective models 87.6% and the multi-
objective models 12.4%).
Figure 5 shows the single or multiple period approaches illustrating the
importance of multi-period planning in recent years.
According to the reviewed papers, there are few studies in the field of multi-part
products (only 5.4%) and almost few researches in multi-product models (only
29.3%). But the single-product models have a larger share in associated studies
(65.4%) (Fig. 6).

123
S. Rezaei, A. Kheirkhah

Single period
Mul period
25
20
15
10
5
0
2007 2008 2009 2010 2011 2012 2013 2014

Fig. 5 Single and multiple period approaches in different years

Single product
Mul product
Mul part products
25
20
15
10
5
0
2007 2008 2009 2010 2011 2012 2013 2014

Fig. 6 Single and multiple product analysis in different years

2.1.3 Environmental and social aspects of SCND problem

As shown in Fig. 7, considering environmental and social aspects in SCND is now


unavoidable due to highly competitive requirements. Following the increasing
importance of collecting and recovering the used products, some studies have been
recently carried out in the associated literature. Accordingly, Pishvaee et al. (2010)
proposed a mixed-integer programming (MIP) model for designing a reverse
logistics network, and in order to solve this NP-hard problem, used a Simulated
Annealing (SA) algorithm with special neighborhood structures. In this way,
Kannan et al. (2012) also studied the previously suggested model (Pishvaee et al.
2010) with considering the environmental impacts due to carbon emissions. Fonseca
et al. (2010) suggested a bi-objective model in reverse supply chain based on the
total costs and environmental impacts by considering a two-phase stochastic
programming.
To avoid the sub-optimal solutions arising in separate modeling of forward and
reverse networks, most of recent researches have focused on integrated modeling of
such networks known as closed-loop supply chain (CLSC) network design
(Soleimani et al. 2013a, b). Dekker et al. (2013) showed that the reverse logistics
can be efficiently integrated into the forward flow by proposing a mixed integer
linear programming (MILP) model. Elhedhli and Merrick (2012) considered the
environmental costs resulting from CO2 emissions alongside fixed and variable

123
A comprehensive approach in designing a sustainable…

Sustainability
Anticipation

Emersion of supply chain


Emersion Sustainable supply Economic
chain management
Environmental
Risk management Responsibility
of supply chain management Social

Emersion of supply
chain

Competitive advantage

Fig. 7 Improvement of competitive advantage in supply chain due to utilizing environmental and social
aspects in network design

location and production costs in a SCND problem. As it is not possible to solve this
problem by exact methods, they developed a lagrangian relaxation approach to do it.

2.2 Applying cross-docking operations in distribution systems

Cross-docking is a highly functional strategy which has caused a growing attention


by experts and industrialists, and according to the recent researches trend, it has
been intensively applied due to unpredictable economic conditions. Implementation
of these intermediate facilities in distribution systems provides a possibility through
which the indirect shipment of products from distribution centers to customer zones
results in economic, environmental and even social benefits, justifying the use of
cross-docking operations in the configuration of supply chains.
For example, in the absence of cross-docking centers in detergents industries, all
the received shipments in distribution centers (from manufacturing facilities) must
be sent to the customer zones by assigned trailers based on the direct shipment or
milk-run policies leading to inappropriate procedures such as transportation loops
and also Less-than-Truckload (LTL) shipments, and accordingly, the economic and
environmental needs are never responded.
There is no need for a significant investment or main infrastructure changes in
applying a cross-docking system, and therefore, it is assumed an efficient
operational option for organizations. In this strategy (as shown in Fig. 8), inbound
trailers to the cross-docking center are first unloaded in assigned area and then, the

Receiving Dock Shipping Dock

Suppliers Receiving Sorting Shipping Customers

Inbound Trucks Cross-docking Center Outbound Trucks

Fig. 8 Visual illustration of a typical flow in a cross-docking system (Stadtler and Kilger 2000)

123
S. Rezaei, A. Kheirkhah

products are scanned, sorted and moved according to the order destinations. Finally,
the prepared cargoes are shipped to customer zones by dedicated trucks. There is not
any long-term storage in this strategy—up to the maximum of 24 h—and due to the
lack of such a possibility in a cross-docking center, balancing the operations (of this
echelon) is essentially needed in supply chain networks more than ever.
Generally, different aspects and dimensions of the cross-docking are depicted in
Table 3, through which all the associated papers published until 2014 have been
investigated and reviewed (Table 4).

3 Problem statement

To design the cited sustainable closed-loop supply chain network, we introduce a


multi-echelon configuration including suppliers, manufacturing centers, distribution
centers, cross-docking centers, customer zones, collection/inspection centers,

Table 3 Clustering of different cross-docking decision problems


Problem Decision problems

Network design Network configuration and facilities type


Number of cross-docking centers
Location of cross-docking centers
Cross-dock design Shape of the cross-docks
Number of dock doors
Capacity of staging area
Design of staging area
Automation of material handling equipment
Network planning Capacity planning for network routes
Freight flow allocation
Shipment to destination assignment
Network scheduling Shipment dispatching
Collection and delivery vehicle routing
Cross-dock planning Dock door specification
Receiving door assignment
Shipping door assignment
Workforce and equipment capacity planning
Cross-dock scheduling Offline scheduling of inbound trailers
Online scheduling of inbound trailers
Offline scheduling of outbound trucks
Online scheduling of outbound trucks
Scheduling of internal cross-dock workforce
Utilization of staging area/shipment allocation
Shipment to outbound truck assignment

123
Table 4 Clustering and reviewing of papers in cross-docking
Shape of the Number of Capacity of Design of Automation of Freight flow Shipment to Shipment
cross-docks dock doors staging area staging area material handling allocation destination dispatching
equipment assignment

Bartholdi and Gue (2004) 0 1 0 0 0 1 1 0


Carlo and Bozer (2011) 0 1 0 0 1 1 1 0
Vis and Roodbergen (2011) 1 0 1 1 1 1 1 0
Tsui and Chang (1990) 1 1 0 0 1 1 1 0
Tsui and Chang (1992) 1 1 0 0 1 1 1 0
Cohen and Keren (2009) 1 1 0 0 1 1 1 0
Stephan and Boysen (2011) 1 1 0 1 1 1 1 0
Oh et al. (2006) 1 1 0 0 1 1 1 0
Bartholdi and Gue (2000) 1 1 0 0 1 1 1 0
Yu et al. (2008) 1 1 0 0 1 1 1 0
A comprehensive approach in designing a sustainable…

Chmielewski et al. (Chmielewski 1 1 1 0 1 1 1 1


et al. 2009)
Luo and Noble (2012) 1 1 1 1 1 1 1 1
Arabani et al. (2012) 0 1 0 0 1 1 0 1
Boysen et al. (2010) 0 1 0 0 0 1 0 1
Forouharfard and Zandieh (2010) 0 1 0 0 0 1 0 1
Larbi et al. (2011) 0 1 0 0 0 1 0 1
Vahdani and Zandieh (2010) 0 1 0 0 1 1 0 1
Yu and Egbelu (2008) 0 1 1 0 1 1 0 1
Vahdani et al. (2010) 0 1 1 0 1 1 0 1
Soltani and Sadjadi (2010) 0 1 0 0 1 1 0 1
Chen and Lee (2009) 0 1 0 0 0 1 1 1
Briskorn et al. (2010) 0 1 0 0 0 1 0 1

123
Table 4 continued

Shape of the Number of Capacity of Design of Automation of Freight flow Shipment to Shipment
cross-docks dock doors staging area staging area material handling allocation destination dispatching

123
equipment assignment

Alpan et al. (2011) 0 1 0 0 0 1 1 1


Boysen and Fliedner (2010) 1 1 0 0 0 1 1 1
Boysen et al. (2013) 1 1 0 0 0 1 1 1
Rosales et al. (2009) 1 1 0 0 1 1 1 1
Acar et al. (2012) 0 1 0 0 0 0 1 1
Konur and Golias (2013) 0 1 0 0 0 0 0 1
McWilliams (2010) 0 1 1 0 1 1 0 1
McWilliams et al. (2008) 0 1 1 0 1 1 1 1
McWilliams and McBride (2012) 0 1 1 0 1 1 1 1
Wang and Regan (2008) 1 1 0 0 1 1 1 1
Miao et al. (2009) 1 1 1 0 1 1 1 1
Chen and Song (2009) 0 1 0 0 0 1 1 1
Boysen (2010) 0 1 1 0 0 1 1 1
Joo and Kim (2013) 0 1 0 0 0 1 0 1
Van Belle et al. (2013) 1 1 0 0 0 1 1 1
Shakeri et al. (2012) 1 1 0 1 1 1 1 1
Li et al. (2004) 0 1 0 0 1 1 1 1
Alvarez-Perez et al. (2009) 0 1 0 0 1 1 1 1

Collection and Dock door Receiving door Shipping door Workforce and Offline scheduling Online scheduling
delivery specification assigning assigning equipment capacity of inbound trailers of inbound trailers
vehicle routing planning

Bartholdi and Gue (2004) 0 1 0 0 0 1 0


Carlo and Bozer (2011) 0 1 0 1 0 1 0
S. Rezaei, A. Kheirkhah
Table 4 continued

Collection and Dock door Receiving door Shipping door Workforce and Offline scheduling Online scheduling
delivery specification assigning assigning equipment capacity of inbound trailers of inbound trailers
vehicle routing planning

Vis and Roodbergen (2011) 0 0 0 1 0 1 0


Tsui and Chang (1990) 0 1 1 1 0 1 0
Tsui and Chang (1992) 0 1 1 1 1 1 0
Cohen and Keren (2009) 0 1 0 0 1 1 0
Stephan and Boysen (2011) 0 1 0 0 1 1 0
Oh et al. (2006) 0 1 0 0 1 1 0
Bartholdi and Gue (2000) 0 1 0 0 1 1 0
Yu et al. (2008) 0 1 0 0 1 1 0
Chmielewski et al. (2009) 0 1 0 0 1 1 0
Luo and Noble (2012) 0 1 0 1 0 0 0
A comprehensive approach in designing a sustainable…

Arabani et al. (2012) 1 1 0 1 1 0 0


Boysen et al. (2010) 1 1 0 1 1 1 0
Forouharfard and Zandieh (2010) 1 1 0 1 1 1 0
Larbi et al. (2011) 1 1 0 1 1 1 1
Vahdani and Zandieh (2010) 1 1 0 1 0 1 1
Yu and Egbelu (2008) 1 1 0 1 0 1 0
Vahdani et al. (2010) 1 1 0 0 0 1 0
Soltani and Sadjadi (2010) 1 1 0 0 0 1 0
Chen and Lee (2009) 1 1 0 0 0 1 0
Briskorn et al. (2010) 1 1 0 1 0 0 0
Alpan et al. (2011) 1 1 0 1 0 0 0
Boysen and Fliedner (2010) 1 1 0 0 1 0 0
Boysen et al. (2013) 1 1 0 0 1 0 0

123
Table 4 continued

Collection and Dock door Receiving door Shipping door Workforce and Offline scheduling Online scheduling
delivery specification assigning assigning equipment capacity of inbound trailers of inbound trailers

123
vehicle routing planning

Rosales et al. (2009) 1 0 0 0 1 0 0


Acar et al. (2012) 1 0 0 0 1 1 0
Konur and Golias (2013) 1 1 0 0 1 1 0
McWilliams (2010) 1 1 0 0 1 1 0
McWilliams et al. (2008) 1 1 0 1 1 1 0
McWilliams and McBride (2012) 1 1 0 1 1 1 0
Wang and Regan (2008) 1 1 0 1 1 1 0
Miao et al. (2009) 1 1 0 1 1 1 0
Chen and Song (2009) 1 1 0 1 1 1 0
Boysen (2010) 1 1 0 1 1 1 1
Joo and Kim (2013) 1 1 0 1 0 1 1
Van Belle et al. (2013) 1 1 0 1 0 1 1
Shakeri et al. (2012) 1 0 0 1 0 1 0
Li et al. (2004) 1 1 0 1 0 1 0
Alvarez-Perez et al. (2009) 1 1 1 0 0 1 0
S. Rezaei, A. Kheirkhah
A comprehensive approach in designing a sustainable…

SMij MDjk DCRkq Zqlcrz

Suppliers Manufacturing Distribution Cross-docking Customer


(i) centers (j) centers (k) centers (q) zones (l)

zc
Zlm
RVDnk
RMMpj CRVmn
RCSri Recovery Collection /
centers (n) inspection
centers (m)
CRMmp

Reuse Remanufacture
market centers (p)
CRCmr
Recycling
centers (r)
CDSms
Fixed location
Disposal centers
(s)

Potential location

Forward flow
Reverse flow

Fig. 9 Structure and configuration of the supply chain network

recovery centers, remanufacturing centers, recycling centers, disposal centers and


the reuse market (Fig. 9).
In this problem, as it is obvious, we have significantly focused on recovery and
treatment processes of used products. These treatment processes consist of
recovering (the used products are recovered for reuse), remanufacturing (the used
products are remanufactured and prepared for reuse), recycling (the used products
are recycled for manufacturing new products) and disposal (the used product with
too low quality are completely disposed). Moreover, applying cross-docking
operations in the aforementioned network design illustrates more special aspect of
our study and distinguishes it from previous researches. Indeed, establishing cross-
docking operational systems in supply chain networks results in environmental and
economic benefits including:
• Decreased maintenance and storage costs.
• Decreased transportation costs.
• Decreased fuel consumption and increased the air quality as a result.

All in all, considering cross-docking centers in the network configuration causes


a significant decrease in costs and also environmental emissions.

123
S. Rezaei, A. Kheirkhah

3.1 Assumptions

The following assumptions are considered in formulation:


• The customers’ demands must be satisfied.
• The number of facilities in each echelon is limited by pre-determined values due
to their operational capacity and also the potential sites.
• There is no flow between the facilities in the same echelon.
• The capacity of freight vehicles is limited by pre-defined values.
• The receiving and shipping capacity of the cross-docking doors are certain and
limited.
• The amount of used products returned to the collection/inspection centers is a
fraction of the customers’ demands and is allocated to the appropriate treatment
center according to its quality.
• Each customer zone encompasses only one retailer.

3.2 Proposed model notations

3.2.1 Sets

i: Set of suppliers i 2 f1; 2. . .I g


j: Set of manufacturing centers j 2 f1; 2. . .Jg
k: Set of potential distribution centers (facilities) k 2 f1; 2. . .Kg
q: Set of potential cross-docking centers q 2 f1; 2. . .Qg
l: Set of customer zones l 2 f1; 2. . .Lg
m: Set of potential collection/inspection centers m 2 f1; 2. . .Mg
n: Set of recovery centers n 2 f1; 2. . .Ng
p: Set of remanufacturing centers p 2 f1; 2. . .Pg
r: Set of recycling centers r 2 f1; 2. . .Rg
s: Set of disposal centers s 2 f1; 2. . .Sg
t: Set of different manufacturing technologies t 2 f1; 2. . .Tg

3.2.2 Parameters

pci: The cost of purchasing raw materials from supplier i


fcdk : Fixed cost of setting up distribution center k
fccr
q : Fixed cost of setting up cross-docking center q
fccm : Fixed cost of setting up collection/inspection center m
fcjt: Fixed cost of manufacturing at center j with technology t
mcjt: Unit cost of manufacturing at center j with technology t
vcdk : Unit cost of operating at distribution center k
vccr
q : Unit cost of operating at cross-docking center q
vccm : Unit cost of operating at collection/inspection center m
vcrv
n : Unit cost of operating at recovery center n

123
A comprehensive approach in designing a sustainable…

vcrm
p : Unit cost of operating at remanufacturing center p
vcrc
r : Unit cost of operating at recycling center r
vcds
s : Unit cost of operating at disposal center s
tcsm
1ij : Unit cost of vehicle type 1 (trailer) for shipping products from supplier i to
manufacturing center j
tcmd
1jk : Unit cost of vehicle type 1 (trailer) for shipping products from manufac-
turing center j to distribution center k
tcdcr
1kq : Unit cost of vehicle type 1 (trailer) for shipping products from distribution
center k to cross-docking center q
tccrz
2ql : Unit cost of vehicle type 2 (truck) for shipping products from cross-docking
center q to customer zone l
tczc
2lm : Unit cost of vehicle type 2 (truck) for shipping returned used products from
customer zone l to collection/inspection center m
tccrv
2mn : Unit cost of vehicle type 2 (truck) for shipping used products from
collection/inspection center m to recovery center n
tccrm
2mp : Unit cost of vehicle type 2 (truck) for shipping used products from
collection/inspection center m to remanufacturing center p
tccrc
2mr : Unit cost of vehicle type 2 (truck) for shipping used products from
collection/inspection center m to recycling center r
tccds
2ms : Unit cost of vehicle type 2 (truck) for shipping used products from
collection/inspection center m to disposal center s
tcrvd
2nk : Unit cost of vehicle type 2 (truck) for shipping recovered products from
recovery center n to distribution center k
tcrmm
2pj : Unit cost of vehicle type 2 (truck) for shipping remanufactured products
from remanufacturing center p to manufacturing center j
tcrcs
2ri : Unit cost of vehicle type 2 (truck) for shipping recycled products from
recycling center r to supplier i
acdcr
kq : Unit cost of assigning distribution center k to cross-docking center q
accrz
ql : Unit cost of assigning cross-docking center q to customer zone l
k1, k2: Capacity of assigned vehicles type 1 (trailer) and type 2 (truck) k1 [ k2
eojt: Fixed environmental impacts of manufacturing at center j with technology t
eodk : Fixed environmental impacts of setting up the distribution center k
eocr
q : Fixed environmental impacts of setting up the cross-docking center q
eocm : Fixed environmental impacts of setting up the collection/inspection center m
emjt: Unit environmental impacts of manufacturing products at center j with
technology t
ehdk : Unit environmental impacts of operating (handling products) at distribution
center k
ehcr
q : Unit environmental impacts of operating (handling products) at cross-
docking center q
ehcm : Unit environmental impacts of operating (handling products) at collection/
inspection center m

123
S. Rezaei, A. Kheirkhah

ehrvn : Unit environmental impacts of operating (handling products) at recovery


center n
ehrmp : Unit environmental impacts of operating (handling products) at remanu-
facturing center p
ehrcr : Unit environmental impacts of operating (handling products) at recycling
center r
etsm
1ij : Unit environmental impacts of vehicle type 1 (trailer) for shipping products
from facility i to j
etmd
1jk : Unit environmental impacts of vehicle type 1 (trailer) for shipping products
from facility j to k
etdcr
1kq : Unit environmental impacts of vehicle type 1 (trailer) for shipping products
from distribution center k to cross-docking center q
etcrz
2ql : Unit environmental impacts of vehicle type 2 (truck) for shipping products
from cross-docking center q to customer zone l
etzc
2lm : Unit environmental impacts of vehicle type 2 (truck) for shipping used
products from customer zone l to collection/inspection center m
etcrv
2mn : Unit environmental impacts of vehicle type 2 (truck) for shipping used
products from collection/inspection center m to recovery center n
etcrm
2mp : Unit environmental impacts of vehicle type 2 (truck) for shipping used
products from collection center m to remanufacturing center p
etcrc
2mr : Unit environmental impacts of vehicle type 2 (truck) for shipping used
products from collection/inspection center m to recycling center r
etcds
2ms : Unit environmental impacts of vehicle type 2 (truck) for shipping useless
used products from collection/inspection center m to disposal center s
etrvd
2nk : Unit environmental impacts of vehicle type 2 (truck) for shipping recovered
products from facility n to distribution center k
etrmm
2pj : Unit environmental impacts of vehicle type 2 (truck) for shipping
remanufactured products from facility p to manufacturing center j
etrcs
2ri : Unit environmental impacts of vehicle type 2 (truck) for shipping recycled
products from facility r to supplier i
eds: Unit environmental impacts of disposing products at center s
fjdk : The number of fixed job opportunities caused by setting up distribution center
k
fjcr
q : The number of fixed job opportunities caused by setting up cross-docking
center q
fjcm : The number of fixed job opportunities caused by setting up collection/
inspection center m
vjjt: The number of variable job opportunities caused by manufacturing at center j
with technology t
vjdk : The number of variable job opportunities caused by working at distribution
center k
vjcr
q : The number of variable job opportunities caused by working at cross-
docking center q

123
A comprehensive approach in designing a sustainable…

vjcm : The number of variable job opportunities caused by working at collection/


inspection center m
vjrv
n : The number of variable job opportunities caused by working at recovery
center n
vjrm
p : The number of variable job opportunities caused by working at remanu-
facturing center p
vjrc
r : The number of variable job opportunities caused by working at recycling
center r
vjds
s : The number of variable job opportunities caused by working at disposal
center s
fldk : The lost days caused by work’s damages during the setting up distribution
center k
flcr
q : The lost days caused by work’s damages during the setting up cross-docking
center q
flcm : The lost days caused by work’s damages during the setting up collection/
inspection center m
vljt: The lost days caused by work’s damages during the manufacturing at center j
with technology t
vldk : The lost days caused by work’s damages during the products operations at
distribution center k
vlcr
q : The lost days caused by work’s damages during the products operations at
cross-docking center q
vlcm : The lost days caused by work’s damages during the products operations at
collection/inspection center m
vlrv
n : The lost days caused by work’s damages during the products operations at
recovery center n
vlrm
p : The lost days caused by work’s damages during the products operations at
remanufacturing center p
vlrc
r : The lost days caused by work’s damages during the products operations at
recycling center r
vlds
s : The lost days caused by work’s damages during the products operations at
disposal center s
capsi : Capacity of supplier i
capjt: Capacity of manufacturing center j with technology t
capdk : Capacity of distribution center k
capcr q : Capacity of cross-docking center q
capcm : Capacity of collection/inspection center m
caprv n : Capacity of recovery center n
caprm p : Capacity of remanufacturing center p
caprc r : Capacity of recycling center r
capds s : Capacity of disposal center s
maxdk : Maximum intended number of established facilities in echelon k
maxcr q : Maximum intended number of established facilities in echelon q

123
S. Rezaei, A. Kheirkhah

maxcm : Maximum intended number of established facilities in echelon m


d1: Demand of customer zone l
dmgjt: Percentage of the products harming the customers (manufactured at center
j with technology t)
bjt: Percentage of the broken products manufactured at center j with technology t
a1: Percentage of the used products returned from customer zone l
bnm ; bpm ; brm ; bsm : The percentage of recoverable, reusable, recyclable and wasted
products in collection/inspection center m, respectively (bnm þ bpm þ brm þ
bsm ¼ 1)
crv rm rc
n , cp , cr : Percentage of the products transferred from a recovering,
remanufacturing and recycling center to reuse market, respectively
scd, scm, scr: Unit cost saving resulted from utilizing recovered, remanufactured
and recycled used-products, respectively
scu: Unit sales price of used products in reuse market
sed, em, ser: Unit environmental benefits resulted from utilizing recovered,
remanufactured and recycled used-products, respectively
wem, wdp: The weights assigned to the elements of environmental objective
function including:

• Environmental emissions related to the transportation and setting up


operations
• Damaging products
• ejo, eld: The weights assigned to the elements of social objective function
including:
• Created fixed and variable job opportunities
• Lost working days

3.2.3 Variables

3.2.3.1 Continuous variables


SMij: Flow of products from supplier i to manufacturing center j
MDjk: Flow of products from manufacturing center j to distribution center k
DCRkq: Flow of products from distribution center k to cross-docking center q
CRZql: Flow of products from cross-docking center q to customer zone l
ZClm: Flow of products from customer zone l to collection/inspection center m
CRVmn: Flow of products from collection/inspection center m to recovery center n
CRMmp: Flow of products from collection center m to remanufacturing center p
CRCmr: Flow of products from collection/inspection center m to recycling center r
CDSms: Flow of products from collection/inspection center m to disposal center s
RVDnk: Flow of recovered products from recovery center n to distribution center k
RMMpj: Flow of remanufactured products from remanufacturing center p to
manufacturing center j
RCSri: Flow of recycled products from recycling center r to supplier i

123
A comprehensive approach in designing a sustainable…

Hjt: The amount of manufactured products at center j with technology t

3.2.3.2 Discrete variables


numsm1ij : Number of assigned vehicles type 1 (trailer) from supplier i to
manufacturing center j
nummd1jk : Number of assigned vehicles type 1 (trailer) from manufacturing center j
to distribution center k
numdcr
1kq : Number of assigned vehicles type 1 (trailer) from distribution center k to
cross-docking center q
numcrz
2ql : Number of assigned vehicles type 2 (truck) from cross-docking center q
to customer zone l
numzc2lm : Number of assigned vehicles type 2 (truck) from customer zone l to
collection/inspection center m
numcrv
2mn : Number of assigned vehicles type 2 (truck) from collection/inspection
center m to recovery center n
numcrm
2mp : Number of assigned vehicles type 2 (truck) from collection/inspection
center m to remanufacturing center p
numcrc
2mr : Number of assigned vehicles type 2 (truck) from collection/inspection
center m to recycling center r
numcds
2ms : Number of assigned vehicles type 2 (truck) from collection/inspection
center m to disposal center s
numrvd
2nk : Number of assigned vehicles type 2 (truck) from recovery center n to
distribution center k
numrmm
2pj : Number of assigned vehicles type 2 (truck) from remanufacturing center
p to manufacturing center j
numrcs
2ri : Number of assigned vehicles type 2 (truck) from recycling center r to
supplier i

3.2.3.3 Binary variables



1; if distribution center k is established
Ydk :
0; otherwise

1; if cross docking center q is established
Ycr
q : 0; otherwise

1; if collection center m is established
Ycm :
0; otherwise

1; if facility k is assigned to cross docking q
Qdcr
kq : 0; otherwise

123
S. Rezaei, A. Kheirkhah


1; if cross docking q is assigned to zone l
Zcrz
ql : 0; otherwise

1; if zone l is assigned to collection center m
Zzc
lm : 0; otherwise

3.3 Objective functions


XX X X X XX
Min OBJ1 ¼ fcjt þ fcdk Ydk þ fccr cr
q Yq þ fccm Ycm þ pci SMij
j t k q m i j
XX XX XX
þ mcjt Hjt þ vcdk MDjk þ vccr
q DCRkq
j t j k k q
XX XX XX
þ vcrv
n CRVmn þ vcrm
p CRMmp þ vcrc
r CRCmr
m n m p m r
XX XX XX
þ vcds
s CDSms þ tcsm sm
1ij num1ij þ tcmd md
1jk num1jk
m s i j j k
XX XX XX
þ tczc zc
2lm num2lm þ tccrv crv
2mn num2mn þ tccrm crm
2mp num2mp
l m m n m p
XX XX XX
þ tccrc crc
2mr num2mr þ tccds cds
2ms num2ms þ tcrvd rvd
2nk num2nk
m r m s n k
XX XX XX
þ tcrmm rmm
2pj num2pj þ tcrcs rcs
2ri num2ri þ tcdcr dcr
1kq num1kq
p j r i k q
XX XX XX
þ tccrz crz
2ql num2ql þ acdcr cr
kq capq Qkq þ accrz crz
ql dl Zql
q l k q q l
!
XX XX XX
þ aczc c zc
lm capm Zlm þ vccm al dl Zzc
lm  scd RVDnk
l m l m n k
! !
XX XX
 scm RMMpj  scr RCSri
p j r i
!
XX XX XX
 scu crv
n CRVmn þ crm
p CRMmp þ crc
r CRCmr
m n m p m r

ð1Þ

In above objective function—economic objective, the total costs of the


network are minimized. In this function, the first to forth terms indicate the fixed
costs of manufacturing with a certain technology and that of setting up facilities.
The fifth to twenty-eighth terms are associated with purchasing, manufacturing,

123
A comprehensive approach in designing a sustainable…

operating (in different facilities), shipping, assigning and collecting costs. The
last six terms are the savings resulted from reusing products in the manufac-
turing centers, redistribution of remanufactured or recovered products and selling
them in reuse market.
"
XX X X X
Min OBJ2 ¼ wem eojt þ eodk Ydk þ eocr rc
q Yq þ eocm Ycm
j t k q m
XX XX XX
þ emjt Hjt þ etsm sm
1ij num1ij þ etmd md
1jk num1jk
j t i j j k
XX XX XX
þ etzc zc
2lm num2lm þ etcrv crv
2mn num2mn þ etcrm crm
2mp num2mp
l m m n m p
XX XX XX
þ etcrc crc
2mr num2mr þ etcds cds
2ms num2ms þ etrvd rvd
2nk num2nk
m r m s n k
XX XX XX
þ etrmm rmm
2pj num2pj þ etrcs rcs
2ri num2ri þ etdcr dcr
1kq num1kq
p j r i k q
XX XX XX
þ etcrz crz
2ql num2ql þ ehdk MDjk þ ehcr
q DCRkq
q l j k k q
XX XX XX
þ ehrv
n CRVmn þ ehrm
p CRMmp þ ehrc
r CRCmr
m n m p m r
!
XX XX XX
þ ehcm al dl Zzc
lm þ eds CDSms  sed RVDnk
l m m s n k
! !#
XX XX
sem RMMpj  ser RVSri
p j r i
" #
XX
þ wdp dmgjt Hjt ð1  bjt Þ
j t

ð2Þ

In second objective function——environmental impacts of the network, the first


to forth terms indicate the fixed environmental impacts resulted from manufac-
turing with a certain technology and establishing facilities. The fifth to twenty
forth terms are associated with manufacturing, transportation, operating (in
different facilities) and disposal environmental impacts. The twenty-fifth to
twenty-seventh summations show the environmental benefits due to reusing the
End-Of-Life (EOL) products. The last summation is also the damage caused by
the products.

123
S. Rezaei, A. Kheirkhah

"
X X X XX
Max OBJ3 ¼ ejo fjdk Ydk þ fjrc cr
q Yq þ fjcm Ycm þ vjjt Hjt =capjt
k q m j t
XX XX
þ vjdk MDjk =capdk þ vjcr cr
q DCRkq =capq
k j q k
XX XX
n
þ vjcm Zzc c
lm dl al =capm þ vjrv rv
n CRVmn bm =capn
ml n m
XX XX
p r
þ vjrm rm
p CRMmp bm =capp þ vjrc rc
r CRCmr bm =capr
p m r m
"
XX X X
s
þ vjds ds
s CDSms bm =caps  eld fldk Ydk þ flcr cr
q Yq ð3Þ
s m k q
X XX XX
þ flcm Ycm þ vljt Hjt =capjt þ vldk MDjk =capdk
m j t k j
XX XX
þ vlcr cr
q DCRkq =capq þ vlcm Zzc c
lm dl al =capm
q k m l
XX X X vlrm p
n p CRMmp bm
þ vlrv rv
n CRVmn bm =capn þ
n m p m
caprm
p
#
X X vlrc CRCmr br XX
s
þ r m
þ vlds ds
s CDSms bm =caps
r m
caprc
r s m

The third objective function—social objective—tries to maximize the social


benefits of the network. In this objective, the first to third terms are the fixed job
opportunities caused by establishing facilities in different echelons (such as
managers). The forth to eleventh summations indicate the created variable job
opportunities (such as workers in facilities). The terms twelfth to twenty-second are
associated with work’s damages caused during the facilities establishment,
manufacturing and operating of products.

3.4 Constraints

The constraints of the model are formulated as follows:


X X
SMij ¼ MDjk 8j ð4Þ
i k
X X
MDjk ¼ DCRkq 8k ð5Þ
j q

X X
DCRkq ¼ dl Zcrz
ql 8q ð6Þ
k l
X
Zcrz
ql ¼ 1 8l ð7Þ
q

123
A comprehensive approach in designing a sustainable…

X
Zzc
lm ¼ 1 8l ð8Þ
m
X X
CRVmn ¼ bnm al dl Zzc
lm 8m ð9Þ
n l
X X
CRMmp ¼ bpm al dl Zzc
lm 8m ð10Þ
p l
X X
CRCmr ¼ brm al dl Zzc
lm 8m ð11Þ
r l
X X
CDSms ¼ bsm al dl Zzc
lm 8m ð12Þ
s l
X  X
RVDnk ¼ 1  crv
n CRVmn 8n ð13Þ
k m

X  X
RMMpj ¼ 1  crm
p CRMmp 8p ð14Þ
j m

X  X
RCSri ¼ 1  crc
r CRCmr 8r ð15Þ
i m

The above constraints indicate that the products flow is maintained and the
customers’ demands are satisfied.
X X
Hjt ¼ SMij 8j ð16Þ
t i

The amount of manufactured products (by each technology) is calculated by the


above constraint.
SMij  k1 numsm
1ij 8i; j ð17Þ

MDjk  k1 nummd
1jk 8j; k ð18Þ

al dl Zzc zc
lm  k2 num2lm 8l; m ð19Þ

CRVmn  k2 numcrv
2mn 8m; n ð20Þ

CRMmp  k2 numcrm
2mp 8m; p ð21Þ

CRCmr  k2 numcrc
2mr 8m; r ð22Þ

CDSms  k2 numcds
2ms 8m; s ð23Þ

RVDnk  k2 numrvd
2nk 8n; k ð24Þ

123
S. Rezaei, A. Kheirkhah

RMMpj  k2 numrmm
2pj 8p; j ð25Þ

RCSri  k2 numrcs
2ri 8r; i ð26Þ

DCRkq  k1 numdcr
1kq 8k; q ð27Þ

dl Zcrz crz
ql  k2 num2ql 8q; l ð28Þ

The above constraints ensure that the flow of products—in different echelons—
doesn’t exceed the determined capacity of freight vehicles.
X
SMij  capsi 8i ð29Þ
j

The amount of raw materials procured by each supplier is constrained by above


inequality based on its capacity.
Hjt  capjt 8j; t ð30Þ

A manufacturing center can produce according to its capacity only when it is


opened.
X
MDjk  capdk Ydk 8k ð31Þ
j

X
DCRkq  capcr cr
q Yq 8q ð32Þ
k
X
al dl Zzc c c
lm  capm Ym 8m ð33Þ
l
X
CRVmn  caprv
n 8n ð34Þ
m
X
CRMmp  caprm
p 8p ð35Þ
m
X
CRCmr  caprc
r 8r ð36Þ
m
X
CDSms  capds
s 8s ð37Þ
m

The products flow is established through a center only when it is operating and
also has enough capacity.
X
Ydk  maxdk ð38Þ
k

123
A comprehensive approach in designing a sustainable…

X
Ycr cr
q  maxq ð39Þ
q

X
Ycm  maxcm ð40Þ
m

The number of facilities in each echelon is constrained by a pre-determined


value.
0  DCRkq  MQdcr
kq 8k; q ð41Þ

The flow of products from distribution centers to cross-docking facilities is


established only when the associated centers are assigned to each other.
Ydk ; Ycr c crz zc
q ; Ym ; Qkq ; Zql ; Zlm 2 f0; 1g ð42Þ

SMij ;MDjk ;DCRkq ;CRZql ;CRVmn ;CRMmp ;CRCmr ;CDSms ;RVDnk ;RMMpj ;RCSri 0
ð43Þ

numsm md zc crv crm crc cds rvd rmm


1ij ; num1jk ; num2lm ; num2mn ; num2mp ; num2mr ; num2ms ; num2nk num2pj ;
þ
ð44Þ
numrcs dcr crz
2ri ; num1kq ; num2ql 2 Z

The above constraints are related to the decision variables through which the
integrality, non-negativity and binary restrictions are ensured.

4 Solution approach

The complex and large-scale optimization problems require the efficient techniques
to profoundly search the solution space. In this paper, we are firstly going to optimize
this CLSC network design in small-scale sample problems using an appropriate
multi-objective approach by the GAMS software. But, as the proposed problem is
highly NP-hard, utilizing the metaheuristic solving methods is inevitable. In this
regard, through a comprehensive analysis of used metaheuristics in similar issues, we
have intended to apply a MOCS algorithm to solve the problem. Due to the specific
procedure of Cuckoo Optimization Algorithm (COA), which will be proposed in the
following sections, the global and local search features are fully covered in this
method indicating its privilege compared with even hybrid algorithms.
Accordingly, to test the efficiency of the algorithm, we have tried to compare its
performance with the following metaheuristics in different metrics:
(1) Multi-objective imperialist competitive algorithm (MOICA).
(2) Multi-objective particle swarm optimization (MOPSO) algorithm.

123
S. Rezaei, A. Kheirkhah

4.1 The multi-objective technique

The proposed problem in this paper contains three conflicting objectives, and there
is no possibility to reach a single optimum solution for them simultaneously. In
these kinds of multi-objective optimization problems, a set of optimal solutions are
obtained at the end of optimization process called Pareto optimal solutions, and as
there is a trade-off between the objectives, it is impossible to improve any of them
without deteriorating at least one of the others. In this regards, to achieve the set of
Pareto optimal solutions, we first calculate the productivity matrix in the Step-
method (known as STEM) (Fig. 10).
In this method, each objective is solely optimized without considering the others.
These solutions are defined as anchor points called f*1, f*2 and f*3 in the objectives
space. Then, through considering the triple columns in the STEM matrix and
attaining the associated intervals for each objective, the improved e-constraint
method is utilized to obtain the effective solutions. One of the main advantages of
this methodology is the possibility to control and handle the number of solutions and
intervals based on the decision-maker’s (DM’s) criteria.
The steps of the improved e-constraint methodology are as follows: 1—the
objective function f1 is considered as the main objective according to the DM’s
priority, 2—now according to the STEM matrix obtained previously, the intervals
related to each objective are divided by the number of b. Following the above sub-
steps, the amount of ea is calculated based on the formula (45) and (46), 3—the
objective functions f2 and f3 are considered as the constraints (47) and (48) in the
formulation and the model is optimized based on the main objective (f1), 4—The
model is optimized in various iterations according to the obtained values for ea, and
the Pareto optimal solutions are reported. In this regards, the Decision-Maker (DM)
is able to select any of the intervals as the optimum solution.
e2a ¼ minf 2 þ ½ðmax f 2  min f 2 Þ=bðaÞ 8a ¼ 1; . . .; b ð45Þ

e3a ¼ minf 3 þ ½ðmax f 3  min f 3 Þ=bðaÞ 8a ¼ 1; . . .; b ð46Þ

f 2  e2a ð47Þ

⎡ f* f 21 f i1 f k1 ⎤ f1 f2 f3
⎢ 1 ⎥

⎢ ⎥ f1 585532* 20113 240.84


⎢ ⎥
⎢ i ⎥ f2 1119419 15671* 304.92
⎢ f1 f 2i f i* f ki ⎥
⎢ ⎥

f3 1532992 74798 505.26*


⎢fk f k
fi k
f * ⎥
⎢⎣ 1 2 k
⎦⎥
Fig. 10 Pay-off matrix in STEM method

123
A comprehensive approach in designing a sustainable…

600 40000
30000
f3 (max)

f2 (min)
400
20000
200 10000
0 0
585587
585587
586758
602671
602671
602671
602671
602733
602733
602733
709891

602733
605733
615700
631287
639912
651284
665874
669910
671285
681005
689954
704010
f1 (min) f1 (min)

(b) The sensitivity analysis in case ε2α = 21584 (a) The sensitivity analysis in case ε3α = 298.8

Fig. 11 Sensitivity analysis on different aspects of objective functions in small scale problems

f 3  e3a ð48Þ

Given the undeniable complexity of the problem, the small-scales of it are


optimized using the GAMS solver in the computer with CPU 2.13 GHz and the
memory 2 GB. And according to the above-mentioned approaches (STEM and
improved e-constraint), the sensitivity analysis results are as Fig. 11.
As stated earlier, the sensitivity analysis indicates the trade-off between
objectives through which improving one of them makes the other to be deteriorated
illustrating a multi-objective optimization problem.

4.2 The proposed metaheuristics algorithm

After introducing the Imperialist Competitive Algorithm (ICA), a new methodology


called Cuckoo Optimization Algorithm (COA) was developed for better searching
the global optimal solutions. This algorithm is one of the most powerful
evolutionary optimization methodologies which have been ever proposed.
Cuckoo algorithm has been inspired by the life of a bird family, called cuckoo,
firstly presented by Yang and Deb (Yang and Deb 2009) at a conference in India.
But later, this metaheuristics was proposed in an article with complete details by
Rajabioun (Rajabioun 2011), which became a basis for the application of this
algorithm on different issues.
Afterwards, Yang and Deb (Yang and Deb 2013) considered the multi-objective
mode of Cuckoo Search (CS) in their study, and in this regards, investigated its
performance in different fields of engineering.
Like other evolutionary algorithms, the COA also starts with an initial population
of cuckoos. These initial cuckoos lay some eggs in some host bird’s nests. A number
of these eggs, which are more similar to the eggs of the host birds, have a greater
chance to grow up and become a mature cuckoo, and the others are distinguished
and then killed by the host birds. The number of grown eggs specifies the suitability
of the nests in that area. The more eggs survive in an area, the more profit is
assigned to that area. Therefore, the position in which more eggs survive will be the
term that COA is going to optimize (Rajabioun 2011).

123
S. Rezaei, A. Kheirkhah

4.2.1 Generating the initial habitats of cuckoos (the first candidate solutions)

To solve an optimization problem, it is critical to define the problem variables as an


array. In GA and PSO algorithms, these arrays are called ‘‘Chromosome’’ and
‘‘Particle Position’’, respectively. But hear in COA, this array is called ‘‘habitat’’. In
an n-dimensional optimization problem, a habitat is an array of 1*n, showing the
current living position of cuckoos. This array is defined as follows:
Habitat ¼ ½x1 ; x2 ; . . .; xn 

The profit (or the suitability) of the current habitat is obtained by evaluation of
associated profit function fp;
Profit ¼ f p ðhabitatÞ ¼ f p ðx1 ; x2 ; . . .; xn Þ:

In order to start the algorithm, a habitat matrix of size Npop*n is generated. Then,
a number of eggs are assigned to each of these habitats randomly. In nature, each
cuckoo lays from 5 to 20 eggs, and these numbers are considered as the upper and
lower limits of egg allocation to each cuckoo at different iterations. Another natural
habitude of each cuckoo is that it lays (eggs) within a specific domain (distance)
called the maximum ‘‘Egg Laying Radius (ELR)’’. ELR is proportional to the total
number of eggs, number of current cuckoo’s eggs and also the problem variable
limits. So ELR is calculated as follows:
 
Number of current cuckoo0 s eggs
ELR ¼ a   ðvarhi varlow Þ
ðTotal number of eggsÞ
where a is a parameter for handling the maximum value of ELR, and both varhi and
varlow represent the upper and lower limits of variables, respectively.

4.2.2 Cuckoos’ procedure in egg laying

Each cuckoo lays some eggs in some other host birds’ nests randomly, within its
ELR (Fig. 12). After all cuckoos laid their eggs in the host nests, some of them are
detected and then thrown out of the nest due to the non-similarity to host birds’

Fig. 12 Random egg laying in


ELR (central red star is the
initial habitat of the cuckoo with
5 eggs: pink stars are the eggs’
new nests) (Rajabioun 2011)

123
A comprehensive approach in designing a sustainable…

eggs. Therefore, after each egg laying process, p % of all eggs (usually 10%), with
less profit value, are killed. The rests grow and are fed by nest birds.

4.2.3 Immigration of cuckoos

When young cuckoos grow and become mature, they live in their own area and society
for some time. But as the time of egg lying is approaches, they immigrate to better
habitats in which the chance of survival is more. After the formation of cuckoo groups
in different areas of the environment (the problem search space), the group with the
best profit value is selected as the goal point for other cuckoos to immigrate.
When immigrating toward the goal point, the cuckoos do not fly all the way to
the destination habitat. They only fly a part of the way (k %), and also have a
deviation of u (as shown in Fig. 13). These two parameters (k and u) help cuckoos
search much more positions in the environment. The k is randomly generated in the
interval [0, 1], and the u is also a random value from –p/6 to p/6.
The immigration formula is defined as follows:
XNextHabitat ¼ XCurrentHabitat þ u ðXGoalPoint  XCurrentHabitat Þ:

4.2.4 The multi-objective cuckoo search (MOCS)

In order to extent the primal cuckoo optimization algorithm (COA) for solving
multi-objective optimization problems, Yang and Deb (Yang and Deb 2013; Yang
2013) modified some early rules included in following terms:
• Each cuckoo lays k eggs in any time and puts them in a host nest randomly. In
this regard, the ith egg is related to ith objective i 2 f1; 2; . . .; kg.
• Each nest is left by the probability of Pa and is replaced by a new one including k eggs.
In other words, a fraction of Pa from n nests is replaced by new ones (new solutions).

The pseudo-code of MOCS in following image depicts a clear view of this


procedure (Fig. 14).

Fig. 13 Emigration of a sample cuckoo toward goal habitat (Rajabioun 2011)

123
S. Rezaei, A. Kheirkhah

Initialize objective functions f1(x) , . . . , fk(x) x = (x1, . . .


, xd)T
Generate an initial population of n host nests xi and each with k cuckoos
While (t ≤ Max Generation) or (stop criterion)
Get a cuckoo (say i) randomly and generate k new solutions by Levy flights
Evaluate and check if it is Pareto optimal
Choose a nest among n (say j) randomly
Evaluate k solutions for nest j
if new solutions of nest j dominate those of nest i,
Replace nest i by the new solution set of nest j
End if
Abandon a fraction (pa) of worse nests
Keep the best solutions (or nests with non-dominated sets) for next
generations
Sort and find the current Pareto optimal solutions
end
Post process results and visualization.

Fig. 14 Pseudo code for multi-objective cuckoo search algorithm (Yang and Deb 2013; Yang 2013)

4.3 Efficiency evaluation of the proposed solution approach

In order to evaluate the efficiency and effectiveness of the proposed algorithm


(MOCS), the metaheuristics of Multi-Objective Imperialist Competitive Algorithm
(MOICA) and Multi-Objective Particle Swarm Optimization (MOPSO)—which are
powerful methodologies in associated researches in comparison to other meta-
heuristics—have been utilized. Accordingly, beside the measures of CPU time and
the number of Pareto solutions, four other metrics are defined as follows to assess
the performance of the algorithm:
I. Mean ideal distance (MID) (measures the closeness between Pareto solutions
and the ideal point—less is better);

Pn
i¼1 ci
MID ¼
n
where n is the number of non-dominated solutions and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ci ¼ ðf 1i  f 1 Þ þ ðf 2i  f 2 Þ þ ðf 3i  f 3 Þ2 .
2 2

II. Set Covering (SC) (measures the ability of an algorithm to dominate the
solutions of others—more is better);

jfb 2 Bj9a 2 A : a  bgj


CðA; BÞ ¼
jBj

123
A comprehensive approach in designing a sustainable…

III. Spacing-spread (SS) (measures the spacing and spread of the solutions
simultaneously—less is better);

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u u M
u1 X jQj uX 2
SS ¼ t t
2 jQj jQj
ðdi  dÞ maxi¼1 f im  mini¼1 f im
jQj i¼1 m¼1

PjQj
P
M di
where di ¼ mink2Q;k6¼i f im  f km , d ¼ i¼1
and |Q| is the number of final
m¼1 jQj
solutions.
IV. Spread of non-dominated solutions (SNS) or diversity (measures the diversity
of non-dominated solutions—more is better);

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u n ðMIDc Þ2
uX i

SNS ¼ t n

i¼1
n1

Given the fact that the tabled problem is the first attempt in the field of supply
chain network design (SCND) and there is no previously studied researches in the
literature, the experimental data of the problem are randomly generated based on the
few related case studies and similar models (Devika et al. 2014; McWilliams 2010;
Yang and Deb 2013) in associated literatures (as shown in Table 6). In this regards,

Table 5 Level and size of sample problems


Level Sample problem Size (I, J, K, Q, L, M, N, P, R, S, T )

Small P1 (8, 4, 8, 3, 12, 8, 2, 3, 3, 2, 2)


P2 (10, 6, 10, 5, 16, 10, 3, 4, 4, 3, 2)
P3 (14, 8, 12, 6, 20, 12, 4, 5, 6, 4, 2)
P4 (18, 10, 14, 8, 24, 14, 5, 6, 8, 5, 3)
P5 (26, 14, 18, 10, 32, 18, 7, 8, 12, 7, 3)
P6 (30, 16, 20, 12, 36, 20, 8, 9, 14, 8, 4)
Medium P7 (34, 18, 22, 14, 40, 22, 9, 10, 16, 9, 4)
P8 (54, 28, 32, 20, 60, 32, 14, 15, 26, 14, 6)
P9 (62, 32, 36, 24, 68, 36, 16, 17, 30, 16, 6)
P10 (66, 34, 38, 26, 72, 38, 17, 18, 32, 17, 7)
P11 (70, 36, 40, 28, 76, 40, 18, 19, 34, 18, 7)
P12 (74, 38, 42, 29, 80, 42, 19, 20, 36, 19, 7)
Large P13 (78, 40, 44, 31, 84, 44, 20, 21, 38, 20, 8)
P14 (82, 42, 46, 32, 88, 46, 21, 22, 40, 21, 8)
P15 (102, 52, 56, 40, 108, 56, 26, 27, 50, 26, 10)
P16 (114, 58, 62, 43, 120, 62, 29, 30, 56, 29, 11)
P17 (122, 62, 66, 46, 128, 66, 31, 32, 60, 31, 11)
P18 (130, 66, 70, 47, 136, 70, 33, 34, 64, 33, 12)

123
Table 6 The experimental data of sample problem 6
Parameter Produced value Parameter Produced value Parameter Produced value Parameter Produced value

123
PC (i) U (200, 270) VLcr (q) U (0.17, 0.19) VCds (s) U (15, 19) ET2rcs (r, i) U (13, 17)
CAPs (i) U (450, 850) FCc (m) U (41, 51) ED (s) U (14, 19) d (l) U (95, 200)
FC (j, t) U (150, 190) VCc (m) U (11, 16) CAPds (s) U (295, 360) alfa (l) U (0.30, 0.65)
MC (j, t) U (105, 125) EOc (m) U (8, 14) VJds (s) U (0.38, 0.55) k1 1000
EO (j, t) U (17, 20) EHc (m) U (11, 16) VLds (s) U (0.19, 0.23) k2 250
EM (j, t) U (15, 27) CAPc(m) U (330, 370) TC1sm (i, j) U (25, 30) MAXdk 17
CAP (j, t) U (400, 950) FJc (m) U (95, 150) TC1md (j, k) U (25, 31) MAXcrq 10
VJ (j, t) U (0.30, 0.55) VJc (m) U (0.48, 0.72) TC1dcr (k, q) U (25, 30) MAXcm 16
VL (j, t) U (0.35, 0.60) FLc (m) U (95, 150) TC2crz (q, l) U (15, 20) Wem 0.4
B (j, t) U (0.01, 0.03) VLc (m) U (0.45, 0.55) TC1zc (l, m) U (25, 29) Wdp 0.6
DMG (j, t) U (0.001, 0.003) VCrv (n) U (14, 16) TC2crv (m, n) U (15, 21) Epjo 0.75
FCd (k) U (48, 65) EHrv (n) U (12, 15) TC2crm (m, p) U (15, 20) Epld 0.25
VCd (k) U (70, 95) CAPrv(n) U (310, 440) TC2crc (m, r) U (15, 22) betta (n, m) U (0.10, 0.30)
EOd (k) U (10, 17) VJrv (n) U (0.55, 0.71) TC2cds (m, s) U (15, 21) betta (p, m) U (0.10, 0.32)
EHd (k) U (16, 21) VLrv (n) U (0.18, 0.23) TC2rvd (n, k) U (15, 22) betta (r, m) U (0.10, 0.35)
CAPd (k) U (750, 900) c rv(n) U (0.28, 0.31) TC2rmm (p, j) U (15, 20) betta (s, m) 1-betta (n, m)-betta
(p, m)-betta (s, m)
FJd (k) U (55, 73) VCrm (p) U (13, 17) TC2rcs (r, i) U (15, 21) ACdcr (k, q) U (28, 45)
VJd (k) U (0.52, 0.65) EHrm (p) U (14, 18) ET1sm (i, j) U (16, 22) ACcrz (q, l) U (31, 55)
FLd (k) U (150, 220) CAPrm (p) U (290, 380) ET1md (j, k) U (16, 22) ACzc (l, m) U (35, 53)
VLd (k) U (0.25, 0.45) VJrm (p) U (0.67, 0.70) ET1dcr (k, q) U (16, 22) SCd 27
FCcr (q) U (37, 45) VLrm (p) U (0.18, 0.22) ET2crz (q, l) U (13, 16) SEd 27
VCcr (q) U (10, 15) c rm (p) U (0.25, 0.40) ET1zc (l, m) U (16, 21) SCm 25
EOcr (q) U (5, 12) VCrc (r) U (14, 18) ET2crv (m, n) U (13, 17) SEm 21
EHcr (q) U (8, 11) EHrc (r) U (14, 19) ET2crm (m, p) U (13, 17) SCr 19
S. Rezaei, A. Kheirkhah
Table 6 continued

Parameter Produced value Parameter Produced value Parameter Produced value Parameter Produced value

CAPcr (q) U (1100, 1300) CAPrc (r) U (390, 450) ET2crc (m, r) U (13, 18) SEr 24
FJcr (q) U (250, 340) VJrc (r) U (0.50, 0.65) ET2cds (m, s) U (13, 17) SCu 31
VJcr (q) U (0.81, 0.84) VLrc (r) U (0.20, 0.25) ET2rvd (n, k) U (13, 16)
FLcr (q) U (90, 125) c rc (r) U (0.29, 0.33) ET2rmm (p, j) U (13, 18)
A comprehensive approach in designing a sustainable…

123
Table 7 The obtained results of metrics for algorithms’ performances
Problem size Test problem Population Insertion Time (s) Number of pareto solutions MID
(P,i) size (n)
MOPSO MOICA MOCS MOPSO MOICA MOCS MOPSO MOICA MOCS

123
Small P.1 50 50 41 27 29 15 14 50 922,052 9,113,453 906,936
P.2 50 50 57 35 35 10 16 36 1,243,901 1,296,516 1,237,186
P.3 100 50 149 100 87 19 14 40 1,633,800 1,641,127 1,629,118
P.4 100 50 178 110 99 13 14 44 1,869,836 1,897,662 1,878,177
P.5 100 60 289 167 165 14 13 53 2,434,745 2,455,293 2,443,160
P.6 100 60 333 186 185 26 24 43 2,889,216 2,753,616 2,853,302
Medium P.7 100 65 421 233 220 19 17 32 3,572,199 3,656,265 3,562,378
P.8 100 65 676 349 360 20 13 37 4,822,963 4,843,504 4,813,879
P.9 100 70 863 461 455 9 13 51 5,538,264 5,584,714 5,564,581
P.10 100 70 949 479 480 24 15 43 5,929,772 6,010,170 5,900,273
P.11 120 70 1243 622 605 19 20 62 6,319,643 6,380,492 6,240,954
P.12 120 80 1348 799 780 -2 12 27 6,394,080 6,441,765 6,415,217
Large P.13 120 80 1565 823 831 -2 16 41 6,940,431 6,984,418 6,940,159
P.14 120 90 1985 1003 1010 16 13 50 7,198,348 7,245,276 7,499,424
P.15 120 90 2713 1340 1299 18 16 48 8,980,550 9,040,652 8,883,187
P.16 120 100 3430 1705 1666 24 12 72 9,713,524 9,751,271 9,713,334
P.17 130 100 6109 1894 1823 17 15 58 10,618,103 10,899,633 10,596,332
P.18 130 110 9859 2112 2002 22 23 68 11,258,964 11,745,632 10,998,640

Problem size Test problem Population Insertion SNS SC SS


(P,i) size (n)
MOPSO MOICA MOCS MOPSO MOICA MOCS MOPSO MOICA MOCS

Small P.1 50 50 962 854 1786 1 0 1 0.1 0.12 0.038


P.2 50 50 790 3225 1578 0.4 0.187 0.9 0.134 0.045 0.07
S. Rezaei, A. Kheirkhah
Table 7 continued

Problem size Test problem Population Insertion SNS SC SS


(P,i) size (n)
MOPSO MOICA MOCS MOPSO MOICA MOCS MOPSO MOICA MOCS

P.3 100 50 1142 815 1571 0.16 0.071 0.9 0.058 0.087 0.048
P.4 100 50 716 1025 2059 0.31 0.07 0.89 0.067 0.054 0.03
P.5 100 60 1010 1359 2630 037 0.077 0.73 0.15 0.13 0.03
P.6 100 60 1647 6756 2540 0.15 0.12 0.86 0.045 0.12 0.039
Medium P.7 100 65 1291 4659 1999 0.32 0.059 0.84 0.066 0.299 0.067
P.8 100 65 1484 1338 2628 0.6 0 0.78 0.063 0.084 0.045
P.9 100 70 1001 1376 2805 1 0 0.57 0.09 0.04 0.03
P.10 100 70 1577 1451 2732 038 0 0.84 0.066 0.077 0.044
P.11 120 70 1819 1709 3810 0.63 0 0.68 0.11 0.06 0.03
P.12 120 80 1634 1499 2375 018 0 1 0.053 0.18 0.05
Large P.13 120 80 1875 1679 2952 0.36 0.06 0.88 0.07 0.07 0.04
A comprehensive approach in designing a sustainable…

P.14 120 90 1546 1478 3376 0.87 0 0.68 0.043 0.092 0.039
P.15 120 90 1620 1730 3947 0.72 0.062 0.73 0.09 0.1 0.036
P.16 120 100 1911 1484 4412 0.71 0.083 0.71 0.043 0.13 0.035
P.17 130 100 1798 1528 4928 0.58 0.074 1 0.078 0.096 0.033
P.18 130 110 1965 1478 5198 0.38 0 0.87 0.18 0.1 0.041

The bold numbers indicate the best values of metrics obtained by different algorithms

123
S. Rezaei, A. Kheirkhah

we have considered 18 sample problems, in three levels of small, medium and large,
to validate the proposed solution approach (Table 5).
Table 7 shows the computational results which have been acquired by the
different algorithms in a same solving condition. By the way, we have tried to
prepare a fair circumstance to better compare the performances of the algorithms,

Fig. 15 Means plot and LSD intervals for algorithms (by 95% confidence) in MID metric

Fig. 16 Means plot and LSD intervals for algorithms (by 95% confidence) in NOS metric

Fig. 17 Means plot and LSD intervals for algorithms (by 95% confidence) in SNS metric

123
A comprehensive approach in designing a sustainable…

and do distinguish the most conquering solution approach. As it is obvious in


table below, the MOCS algorithm has a better performance in almost all metrics in
comparison to the MOICA and MOPSO algorithms illustrating the effectiveness of

Fig. 18 Means plot and LSD intervals for algorithms (by 95% confidence) in SC metric

Fig. 19 Means plot and LSD intervals for algorithms (by 95% confidence) in SS metric

Fig. 20 Means plot and LSD intervals for algorithms (by 95% confidence) in CPU time metric

123
S. Rezaei, A. Kheirkhah

MOCS in different complicated problems. So undoubtedly, it is anticipated to


observe more researches using MOCS in various issues.
The following figures depict the mean and LSD interval plots for algorithms in
the considered metrics. These plots also illustrate the better performance of
MOCS (Figs. 15, 16, 17, 18, 19, 20).
The following figures demonstrate the number of Pareto optimal solutions in any
iterations of a single run obtained by the MOCS algorithm (Fig. 21) and also the
dispersion of Pareto optimum solutions obtained by different algorithms (Fig. 22) in
the sample problem 6, as an example. As it is obvious, with the progress of search

60
50
40
Iteration

30
20
10
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59
Number of pareto solutions

Fig. 21 The number of Pareto optimum solutions in each iteration of sample problem 6 by MOCS
algorithm

Fig. 22 Dispersion of Pareto optimum solutions obtained by different algorithms in sample problem 6

123
A comprehensive approach in designing a sustainable…

process and the problem iterations, the number of Pareto solutions has an ascending
trend illustrating the direction of search towards better points of the solution space.
The fallings in some parts of the diagram shows the direction towards new points in
which the solutions have low quality.

5 Conclusions and future studies

This paper addressed the problem of designing a sustainable closed-loop supply


chain network in 11 echelons, in which the cross-docking operations were applied,
resulted in proposing a comprehensive model to achieve the intended goals. The
problem objectives were (I) minimization of total costs, (II) minimization of
environmental impacts, and (III) maximization of social benefits, considered
through a Mixed Integer Linear Programming (MILP) model. The model consists of
strategic and operational decisions in the supply chain including the network
configuration and structure, facilities location and the transportation.
In order to optimize this NP-hard problem, a MOCS algorithm was utilized as the
first attempt in this area. Regarding the need for evaluation of the proposed
methodology, we compared its performance with two popular metaheuristics of
Multi-Objective Imperialist Competitive Algorithm (MOICA) and Multi-Objective
Particle Swarm Optimization (MOPSO) through different metrics in a same
circumstance. The obtained results from running these algorithms in MATLAB
indicated that the MOCS is almost the most efficient approach based on the various
metrics of CPU time, number of Pareto solutions, mean ideal distance (MID),
diversity, set covering (SC) and spacing-spread (SS). Therefore, it can be concluded
that the MOCS algorithm is a super-efficient approach to optimize different
complex problems and undoubtedly, it is anticipated to observe more researches
using MOCS.
Implementation of cross-docking operational system in the mentioned logistics
network both with simultaneous considering the economic, environmental and
social requirements in a comprehensive and sustainable approach propose a quite
novel research scope in the wide range of problems related to supply chain network
design, and in this regards, help organizations improve their competitive advantage
in different industries.
Generally, the distinct and preferable aspects of this study can outlined as
follows:
• The trade-off between three dimensions of economic, environmental and social
requirements was achieved through suggesting a triple-objective model.
• The forward and reverse logistics were integrated in which the recovered,
remanufactured and recycled used-products were returned to the forward flow.
• All the environmental impacts resulted from opening facilities, manufacturing,
operating and transporting the products were formulated.
• All the social effects including the created job opportunities and the caused
damages were quantified and considered as a separate objective.

123
S. Rezaei, A. Kheirkhah

• Above all, the cross-docking operations were utilized in the logistics network
design.
• As it is impossible to propose an exact solving method for these kinds of
complex optimization problems, we utilized a MOCS algorithm as an efficient
approach.

The most prominent shortcomings of the paper ahead could be noted in some
dimensions such as not considering the uncertainty of the parameters required in
practical issues, not incorporating scheduling decisions in various echelons—
particularly in cross-docking facilities, and also not investigating a multi-product
multi-period model through the mentioned approach. So the proposed method-
ology can be considered as a basis for extending the future associated researches
in different aspects. A first and direct developing approach is to apply the
presented model in real functions and evaluate its performance, in which the
inherent uncertainty of real cases could be included. In another extension, it is
engaging to model a multi-product multi-period problem using the presented
method through which the use of cross-docking operations will find more
attentions. One of the other developing approaches is considering the scheduling
decisions in different echelons of the network, and analyzing the related issues
such as ordering, backlogging and forecasting. Finally, it is also striking to
consider the different decisions related to cross-docking operations such as
designing the operational area and its capacity, determining the number of
receiving and shipping doors, online and offline scheduling of the inbound and
outbound trucks, and planning the routes capacity.

References
Acar K, Yalcin A, Yankov D (2012) Robust door assignment in less-than-truckload terminals. Comput
Ind Eng 63(4):729–738
Alpan G, Ladier AL, Larbi R, Penz B (2011) Heuristic solutions for transshipment problems in a multiple
door cross docking warehouse. Comput Ind Eng 61(2):402–408
Alvarez-Perez GA, González-Velarde JL, Fowler JW (2009) Crossdocking—just in time scheduling: an
alternative solution approach. J Oper Res Soci 60(4):554–564
Arabani AB, Zandieh M, Ghomi SF (2012) A cross-docking scheduling problem with sub-population
multi-objective algorithms. Int J Adv Manuf Technol 58(5–8):741–761
Aras N, Aksen D, Tanuğur AG (2008) Locating collection centers for incentive-dependent returns under a
pick-up policy with capacitated vehicles. Eur J Oper Res 191(3):1223–1240
Bartholdi JJ III, Gue KR (2000) Reducing labor costs in an LTL crossdocking terminal. Oper Res
48(6):823–832
Bartholdi JJ, Gue KR (2004) The best shape for a crossdock. Trans Sci 38(2):235–244
Bender T, Hennes H, Kalcsics J, Melo MT, Nickel S (2002) Location software and interface with GIS and
supply chain management. Facility location: applications and theory. Springer, Berlin, pp 233–274
Boysen N (2010) Truck scheduling at zero-inventory cross docking terminals. Comput Oper Res
37(1):32–41
Boysen N, Fliedner M (2010) Cross dock scheduling: classification, literature review and research
agenda. Omega 38(6):413–422
Boysen N, Fliedner M, Scholl A (2010) Scheduling inbound and outbound trucks at cross docking
terminals. OR Spectrum 32(1):135–161

123
A comprehensive approach in designing a sustainable…

Boysen N, Briskorn D, Tschöke M (2013) Truck scheduling in cross-docking terminals with fixed
outbound departures. OR Spectrum 35(2):479–504
Briskorn D, Choi BC, Lee K, Leung J, Pinedo M (2010) Complexity of single machine scheduling subject
to nonnegative inventory constraints. Eur J Oper Res 207(2):605–619
Cardoso SR, Barbosa-Póvoa APF, Relvas S (2013) Design and planning of supply chains with integration
of reverse logistics activities under demand uncertainty. Eur J Oper Res 226(3):436–451
Carlo HJ, Bozer YA (2011) Analysis of optimum shape and door assignment problems in rectangular
unit-load crossdocks. Int J Logist Res Appl 14(3):149–163
Chaabane A, Ramudhin A, Paquet M (2012) Design of sustainable supply chains under the emission
trading scheme. Int J Prod Econ 135(1):37–49
Chen F, Lee CY (2009) Minimizing the makespan in a two-machine cross-docking flow shop problem.
Eur J Oper Res 193(1):59–72
Chen F, Song K (2009) Minimizing makespan in two-stage hybrid cross docking scheduling problem.
Comput Oper Res 36(6):2066–2073
Chmielewski A, Naujoks B, Janas M, Clausen U (2009) Optimizing the door assignment in LTL-
terminals. Trans Sci 43(2):198–210
Chopra S, Meindl P (2007) Supply chain management. Strategy, planning & operation. Gabler,
Wiesbaden, pp 265–275
Cohen Y, Keren B (2009) Trailer to door assignment in a synchronous cross-dock operation. Int J Logist
Syst Manag 5(5):574–590
Cruz-Rivera R, Ertel J (2009) Reverse logistics network design for the collection of end-of-life vehicles in
Mexico. Eur J Oper Res 196(3):930–939
Dasci A, Verter V (2001) A continuous model for production–distribution system design. Eur J Oper Res
129(2):287–298
Dehghanian F, Mansour S (2009) Designing sustainable recovery network of end-of-life products using
genetic algorithm. Resour Conserv Recycl 53(10):559–570
Dekker R, Fleischmann M, Inderfurth K, van Wassenhove LN (Eds) (2013). Reverse logistics:
quantitative models for closed-loop supply chains. Springer Science & Business Media, Berlin
Demirel NÖ, Gökçen H (2008) A mixed integer programming model for remanufacturing in reverse
logistics environment. Int J Adv Manuf Technol 39(11–12):1197–1206
Devika K, Jafarian A, Nourbakhsh V (2014) Designing a sustainable closed-loop supply chain network
based on triple bottom line approach: A comparison of metaheuristics hybridization techniques. Eur
J Oper Res 235(3):594–615
Elhedhli S, Merrick R (2012) Green supply chain network design to reduce carbon emissions. Transp Res
Part D 17(5):370–379
Faccio M, Persona A, Sgarbossa F, Zanin G (2014) Sustainable SC through the complete reprocessing of
end-of-life products by manufacturers: a traditional versus social responsibility company
perspective. Eur J Oper Res 233(2):359–373
Fleischmann M, Beullens P, Bloemhof-Ruwaard JM, Van Wassenhove LN (2001) The impact of product
recovery on logistics network design. Prod Oper Manag 10(2):156–173
Fonseca MC, Garcı́a-Sánchez Á, Ortega-Mier M, Saldanha-da-Gama F (2010) A stochastic bi-objective
location model for strategic reverse logistics. Top 18(1):158–184
Forouharfard S, Zandieh M (2010) An imperialist competitive algorithm to schedule of receiving and
shipping trucks in cross-docking systems. Int J Adv Manuf Technol 51(9–12):1179–1193
Gallego RC, Cueto EP (2009) Forecasting the returns in reusable containers closed-loop supply chains. A
case in the LPG industry. In: XIII Congreso de Ingenierı́a de Organización: Barcelona, 2–4 de
Septiembre de, pp 311–320
Georgiadis MC, Tsiakis P, Longinidis P, Sofioglou MK (2011) Optimal design of supply chain networks
under uncertain transient demand variations. Omega 39(3):254–272
Jayaraman V, Pirkul H (2001) Planning and coordination of production and distribution facilities for
multiple commodities. Eur J Oper Res 133(2):394–408
Jayaraman V, Ross A (2003) A simulated annealing methodology to distribution network design and
management. Eur J Oper Res 144(3):629–645
Joo CM, Kim BS (2013) Scheduling compound trucks in multi-door cross-docking terminals. Int J Adv
Manuf Technol 64(5–8):977–988
Kannan D, Diabat A, Alrefaei M, Govindan K, Yong G (2012) A carbon footprint based reverse logistics
network design model. Resour Conserv Recycl 67:75–79

123
S. Rezaei, A. Kheirkhah

Ko HJ, Evans GW (2007) A genetic algorithm-based heuristic for the dynamic integrated forward/reverse
logistics network for 3PLs. Comput Oper Res 34(2):346–366
Konur D, Golias MM (2013) Analysis of different approaches to cross-dock truck scheduling with truck
arrival time uncertainty. Comput Ind Eng 65(4):663–672
Krikke H (2010) Opportunistic versus life-cycle-oriented decision making in multi-loop recovery: an eco-
eco study on disposed vehicles. Int J Life Cycle Assess 15(8):757–768
Krikke HR, Kooi EJ, Schuur PC (1999) Network design in reverse logistics: a quantitative model. In New
trends in distribution logistics. Springer, Berlin, pp 45–61
Larbi R, Alpan G, Baptiste P, Penz B (2011) Scheduling cross docking operations under full, partial and
no information on inbound arrivals. Comput Oper Res 38(6):889–900
Li Y, Lim A, Rodrigues B (2004) Crossdocking—JIT scheduling with time windows. J Oper Res Soc
55(12):1342–1351
Listeş O, Dekker R (2005) A stochastic approach to a case study for product recovery network design.
Eur J Oper Res 160(1):268–287
Lu Z, Bostel N (2007) A facility location model for logistics systems including reverse flows: The case of
remanufacturing activities. Comput Oper Res 34(2):299–323
Luo G, Noble JS (2012) An integrated model for crossdock operations including staging. Int J Prod Res
50(9):2451–2464
Marin A, Pelegrı́n B (1998) The return plant location problem: Modelling and resolution. Eur J Oper Res
104(2):375–392
Mazhar MI, Kara S, Kaebernick H (2007) Remaining life estimation of used components in consumer
products: life cycle data analysis by Weibull and artificial neural networks. J Oper Manag
25(6):1184–1193
McWilliams DL (2010) Iterative improvement to solve the parcel hub scheduling problem. Comput Ind
Eng 59(1):136–144
McWilliams DL, McBride ME (2012) A beam search heuristics to solve the parcel hub scheduling
problem. Comput Ind Eng 62(4):1080–1092
McWilliams DL, Stanfield PM, Geiger CD (2008) Minimizing the completion time of the transfer
operations in a central parcel consolidation terminal with unequal-batch-size inbound trailers.
Comput Ind Eng 54(4):709–720
Meixell MJ, Gargeya VB (2005) Global supply chain design: a literature review and critique. Trans Res
Part E 41(6):531–550
Melo T, Nickel S, Saldanha-da-Gama F (2008) Network design decisions in supply chain planning.
Fraunhofer-Institut für Techno-und Wirtschaftsmathematik, Fraunhofer (ITWM), Munich
Miao Z, Lim A, Ma H (2009) Truck dock assignment problem with operational time constraint within
crossdocks. Eur J Oper Res 192(1):105–115
Min H, Ko HJ (2008) The dynamic design of a reverse logistics network from the perspective of
thirdparty logistics service providers. Int J Prod Econ 113(1):176–192
Min H, Ko HJ, Ko CS (2006) A genetic algorithm approach to developing the multi-echelon reverse
logistics network for product returns. Omega 34(1):56–69
Miranda PA, Garrido RA (2004) Incorporating inventory control decisions into a strategic distribution
network design model with stochastic demand. Transp Res Part E Logist Transp Rev 40(3):183–207
Mota B, Gomes MI, Carvalho A, Barbosa-Povoa AP (2014) Towards supply chain sustainability:
economic, environmental and social design and planning. J Clean Prod 105:14–27
Oh Y, Hwang H, Cha CN, Lee S (2006) A dock-door assignment problem for the Korean mail
distribution center. Comput Ind Eng 51(2):288–296
Pati RK, Vrat P, Kumar P (2010) Quantifying bullwhip effect in a closed loop supply chain. Opsearch
47(4):231–253
Pishvaee MS, Razmi J (2012) Environmental supply chain network design using multi-objective fuzzy
mathematical programming. Appl Math Model 36(8):3433–3446
Pishvaee MS, Farahani RZ, Dullaert W (2010) A memetic algorithm for bi-objective integrated forward/
reverse logistics network design. Comput Oper Res 37(6):1100–1112
Pishvaee MS, Rabbani M, Torabi SA (2011) A robust optimization approach to closed-loop supply chain
network design under uncertainty. Appl Math Model 35(2):637–649
Pishvaee MS, Razmi J, Torabi SA (2012a) Robust possibilistic programming for socially responsible
supply chain network design: A new approach. Fuzzy Sets and Syst 206:1–20
Pishvaee MS, Torabi SA, Razmi J (2012b) Credibility-based fuzzy mathematical programming model for
green logistics design under uncertainty. Comput Ind Eng 62(2):624–632

123
A comprehensive approach in designing a sustainable…

Rajabioun R (2011) Cuckoo optimization algorithm. Appl Soft Comput 11(8):5508–5518


Ramezani M, Kimiagari AM, Karimi B, Hejazi TH (2014) Closed-loop supply chain network design
under a fuzzy environment. Knowl-Based Syst 59:108–120
Rogers DS, Tibben-Lembke RS (1998) Going backwards: reverse logistics practices and trends.
University of Nevada, Reno. Center of Logistics Management. Reverse Logistics Executive Council
Rosales CR, Fry MJ, Radhakrishnan R (2009) Transfreight reduces costs and balances workload at
Georgetown crossdock. Interfaces 39(4):316–328
Salema MIG, Póvoa APB, Novais AQ (2009) A strategic and tactical model for closed-loop supply
chains. OR spectr 31(3):573–599
Shakeri M, Low MYH, Turner SJ, Lee EW (2012) A robust two-phase heuristic algorithm for the truck
scheduling problem in a resource-constrained crossdock. Comput Oper Res 39(11):2564–2577
Shevtshenko E, Wang Y (2009) Decision support under uncertainties based on robust Bayesian networks
in reverse logistics management. Int J Comput Appl Technol 36(3–4):247–258
Simchi-Levi E, Kaminsky P (1999) Designing and managing the supply chain: concepts, strategies, and
cases. McGraw-Hill, New York
Simchi-Levi D, Kaminsky P, Simchi-Levi E (2004) Managing the supply chain: the definitive guide for
the business professional. McGraw-Hill Companies, New York
Soleimani H, Kannan G (2015) A hybrid particle swarm optimization and genetic algorithm for closed-
loop supply chain network design in large-scale networks. Appl Math Model 39(14):3990–4012
Soleimani H, Seyyed-Esfahani M, Shirazi MA (2013a) Designing and planning a multi-echelon multi-
period multi-product closed-loop supply chain utilizing genetic algorithm. Int J Adv Manuf Technol
68(1–4):917–931
Soleimani H, Seyyed-Esfahani M, Shirazi MA (2013b) A new multi-criteria scenario-based solution
approach for stochastic forward/reverse supply chain network design. Ann Oper Res 242:399–421
Soltani R, Sadjadi SJ (2010) Scheduling trucks in cross-docking systems: a robust meta-heuristics
approach. Trans Res Part E 46(5):650–666
Stadtler H, Kilger C (2000) Supply chain management and advanced planning. Springer, Berlin
Stephan K, Boysen N (2011) Vis-à-vis vs. mixed dock door assignment: a comparison of different cross
dock layouts. Oper Manag Res 4(3–4):150–163
Syarif A, Yun Y, Gen M (2002) Study on multi-stage logistic chain network: a spanning treebased genetic
algorithm approach. Comput Ind Eng 43(1–2):299–314
Tsiakis P, Papageorgiou LG (2008) Optimal production allocation and distribution supply chain
networks. Int J Prod Econ 111(2):468–483
Tsui LY, Chang CH (1990) A microcomputer based decision support tool for assigning dock doors in
freight yards. Comput Ind Eng 19(1):309–312
Tsui LY, Chang CH (1992) An optimal solution to a dock door assignment problem. Comput Ind Eng
23:1–4
Vahdani B, Zandieh M (2010) Scheduling trucks in cross-docking systems: robust meta-heuristics.
Comput Ind Eng 58(1):12–24
Vahdani B, Soltani R, Zandieh M (2010) Scheduling the truck holdover recurrent dock cross-dock
problem using robust meta-heuristics. Int J Adv Manuf Technol 46(5–8):769–783
Van Belle J, Valckenaers P, Berghe GV, Cattrysse D (2013) A tabu search approach to the truck
scheduling problem with multiple docks and time windows. Comput Ind Eng 66(4):818–826
Vis IF, Roodbergen KJ (2011) Layout and control policies for cross docking operations. Comput Ind Eng
61(4):911–919
Wang JF, Regan A (2008) Real-time trailer scheduling for crossdock operations. Trans J 8:5–20
Wang HF, Hsu HW (2010) A closed-loop logistic model with a spanning-tree based genetic algorithm.
Comput Oper Res 37(2):376–389
Wang F, Lai X, Shi N (2011) A multi-objective optimization for green supply chain network design.
Decis Support Syst 51(2):262–269
Yang XS (ed) (2013) Cuckoo search and firefly algorithm: theory and applications. Springer, Berlin, vol
516
Yang XS, Deb S (2009) Cuckoo search via Lévy flights. In: NaBIC 2009. World Congress on nature &
biologically inspired computing. IEEE, pp 210–214
Yang XS, Deb S (2013) Multi-objective cuckoo search for design optimization. Comput Oper Res
40(6):1616–1624
Yeh WC (2005) A hybrid heuristic algorithm for the multistage supply chain network problem. The
International Int J Adv Manuf Technol 26(5–6):675–685

123
S. Rezaei, A. Kheirkhah

Yu W, Egbelu PJ (2008) Scheduling of inbound and outbound trucks in cross docking systems with
temporary storage. Eur J Oper Res 184(1):377–396
Yu VF, Sharma D, Murty KG (2008) Door allocations to origins and destinations at less-than-truckload
trucking terminals. J Ind Syst Eng 2(1):1–15
Zhang X, Pieter van Donk D, van der Vaart T (2011) Does ICT influence supply chain management and
performance? A review of survey-based research. Int J Oper Prod Manag 31(11):1215–1247

Saeid Rezaei is a Master Graduate at Department of Industrial Engineering, School of Engineering in Bu-
ali Sina University. He holds a Master’s degree in Industrial Engineering from Bu-ali Sina University of
Hamedan and also a Bachelor’s degree in the same field from Islamic Azad University. His research
interest is supply chain network management and all associated topics.

Amirsaman Kheirkhah is an Associate Professor at Department of Industrial Engineering, School of


Engineering in Bu-ali Sina University. He holds a Ph.D. and also a Master’s degree in Industrial
Engineering from Iran Science and Technology University, and a Bachelor’s degree in the same field
from Sharif University of Tehran. His research interests are supply chain management, scheduling and
MODM decision making.

123

You might also like