Sustainable Production and Consumption 27 (2021) 1183–1194

Contents lists available at ScienceDirect

Sustainable Production and Consumption

journal homepage: www.elsevier.com/locate/spc

Research article

A Multi-Objective Green Hub Location Problem with Multi Item-Multi

Temperature Joint Distribution for Perishable Products in Cold Supply
Chain
Maryam Golestani a, Seyed Hamed Moosavirad b,∗, Yasaman Asadi c, Shirin Biglari d
a
Department of Industrial Engineering, Faculty of Engineering, Shahid Bahonar University of Kerman, Kerman, Iran
b
Department of Industrial Engineering, Faculty of Engineering, Shahid Bahonar University of Kerman, Kerman, Iran
c
Department of Industrial Engineering, Faculty of Engineering, Shahid Bahonar University of Kerman, Kerman, Iran
d
Technology Foresight Group, Department of Management, Science and Technology, Amirkabir University of Technology, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: This paper investigates a bi-objective green hub location problem, in which multiple perishable prod-
Received 3 November 2020 ucts with various storage temperatures can be distributed simultaneously in a cold supply chain (CSC).
Revised 13 February 2021
The objectives of this problem include minimizing the system’s total cost (including transportation, hub
Accepted 17 February 2021
establishment, adjustment of the storage compartments’ temperatures, and carbon emission costs) and
Available online 20 February 2021
maximizing the quality of the delivered product to the customer via the proposed model. Mixed-integer
Keywords: linear programming (MILP) in the GAMS software was employed to formulate this problem. Then, the
Hub Location ε -Constraint method was adopted to solve the presented bi-objective model to obtain the Pareto frontier
Perishable Products and consequently, a numerical example based on the CAB (Civil Aeronautics Board) database is presented
Cold Supply Chain to validate the applicability of the model. The solutions of the model provide information regarding the
Transportation hub location (HL), allocating customers to the hubs, allocating customers to the vehicles, and the se-
Multi-objective
quence of vehicles’ services for the Multi Item-Multi Temperature Joint Distribution of perishable prod-
ucts in CSCs. Moreover, the final results revealed the existence of a contradictory exchange between the
two objectives of this paper, implying that the higher is the quality of the delivered perishable product to
the customer, the greater is the system’s total cost. The novelty of the proposed model compared to other
hub location problems (HLPs) lies in the integration of the tactical/operational decisions with strategic
decisions to provide logistic solutions in CSCs by considering the carbon emissions as an environmental
factor in the transportation systems for the simultaneous distribution of dissimilar storage temperatures
perishable products within a CSC. The proposed model in this research can help the distributers of per-
ishable products by maintaining the quality of the delivered items and reducing the system’s total costs
and considering the carbon emissions of transportation systems. This study has practical implications
for the logistics and CSCs managers to not only establish a distribution network for multiple perishable
products on the basis of the findings, but also respond to the environmental sustainability.
© 2021 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction tries and consumption of consumers has emerged (Barber, 2007;

Vergragt et al., 2014). This mediator role is played by the "trans-
As theories and practices of sustainable production and con- portation sector" in supply chains, as the procedures of planning,
sumption become an increasingly important matter of discussion conducting, and controlling activities for the conducive storage and
for researchers, managers and policymakers (Subramanian et al., distribution of products and services from the manufacturer to
2017), an intermediary set of activities that have significant ef- consumers for the ultimate goal of meeting the customers’ require-
fects on the strategies and initiatives of both production of indus- ments (Paciarotti and Torregiani, 2020).
As one of the transportation challenges of sustainable
supply chains in various industries and production sectors
∗
Corresponding author.
(Ghaffarinasab et al., 2018), Hub location problem (HLP) was
E-mail addresses: [email protected] (M. Golestani), introduced by Goldman (Damgacioglu et al., 2015). HLPs are
[email protected], [email protected] (S.H. Moosavirad), solved through two types of solutions; I) strategic decisions and II)
[email protected] (Y. Asadi), [email protected] (S. Biglari).

https://doi.org/10.1016/j.spc.2021.02.026
2352-5509/© 2021 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

operational or tactical decisions. Strategic decisions are as re-

Nomenclature gards to the allocation of hubs and assigning the customers to
them. Tactical or operational decisions are related to vehicle
Indices scheduling (Musavi and Bozorgi-Amiri, 2017).
l, k, j, i Nodes index varying from 0 to n + 1 Another supply chain challenge pertains to the health, fresh-
c Vehicles index varying from 1 to c ness, and timely delivery of perishable foods (Daofang et al., 2015;
q Product index varying from 1 to Q Hsu et al., 2013). Precise control over the time and temperatures is
t Storage temperature index varying from 1 to s a prerequisite to ensure the timely delivery of healthy, fresh, and
b Storage spaces per vehicle index varying from 1 to high-quality perishable products to customers (Hsiao et al., 2017).
B The advent of cold supply chain (CSC) has contributed significantly
to prevent food spoilage during the distribution processes, reduce
Parameters
the distribution costs, and increase the sales value (Daofang et al.,
a Reduction coefficient in inter-hub transport
2015). Cold storages, insulation boxes, and Multi-Temperature Joint
H The number of hubs to be located
Distribution (MTJD) are technologies that are employed in CSCs
M A large number
for setting specific temperature to minimize the adverse effects of
β Transportation cost per distance unit
high temperatures on the food quality (Hsu et al., 2013).
εf Coefficient for carbon emissions of full loaded-
Although multi-temperature refrigerated compartments sys-
vehicles
tem technology offers practical solutions to the real-world chal-
εe Coefficient for carbon emissions of empty vehicles
lenges of the CSCs by maintaining the quality of perishable prod-
Vl Maximum portable weight for a vehicle
ucts such as pharmaceuticals, frozen and fresh foods (Cho and
Si Life span of the product ordered by customer i
Hsu, 2008; Stellingwerf et al., 2018), the environmental conser-
QRPi The starting point of spoilage of the product or-
vation and one of its key implications namely as global con-
dered by customer i
trol of carbon emissions has been defined as one of the consti-
Di j The distance between node i and j
q tuting pillars of a sustainable supply chain (Accorsi et al., 2017;
Wi j The product flow rate of type q from node i to node
Musavi and Bozorgi-Amiri, 2017). More specifically, temperature-
j controlled transportation produces even more emissions than non-
ϕi j Travel time between node i and j temperature-controlled transportation due to the additional fuel
f Transportation cost of foodstuffs at a certain tem- prerequisites for refrigeration and as a result of refrigerant leak-
perature age (Stellingwerf et al., 2018). Corroboratively, a study revealed
h The cost of storage temperature change by one de- that 15% of the global energy from fossil fuels is consumed in
gree Celsius the temperature-controlled food transportation and the environ-
Fk Fixed cost of a hub deployment mental impacts of this type of transportation constitute 40% of
LT Number of storage space per vehicle the global greenhouse emission effects (Adekomaya et al., 2016).
N The total number of nodes Nevertheless, the ever-increasing number of distribution compa-
C Number of vehicles per hub nies has led to the growth of greenhouse gas emissions, particu-
λ Cost of carbon emissions larly CO2 (Musavi and Bozorgi-Amiri, 2017). According to a study
Zb Capacity of each storage compartment in 2017, the supply chains (SCs) footprints of 2500 leading compa-
Tmax The highest storage temperature of vehicle nies across the globe generate around 20% of the total greenhouse
Decision making variables gas emissions (Dubey et al., 2017). Therefore, the environmental
gi Quality of product delivered to customer i impacts of temperature-controlled transportation vehicles on the
eikl The flow rate of node i products transferred to hub environment must be considered when planning SCs and network
l through hub K developments.
Ti Customer i order arrival time As a response to this research agenda, this paper investigates a
Xii A zero-one variable, the value of which equals 1 if bi-objective green hub location problem, in which multiple perish-
node i is chosen as a hub, or else it equals 0. able products with various storage temperatures can be distributed
xik A zero-one variable, the value of which equals 1 if simultaneously in a CSC and aims to minimize the system’s total
node i is assigned to hub k, if not it equals 0. cost (including transportation, the hub establishment, adjustment
dki A zero-one variable, the value of which equals to 1 of the storage compartments’ temperatures, and carbon emission
if node j is served by hub k, or else it equals 0. costs) and maximize the quality of the delivered product to the
Edik A zero-one variable, the value of which equals 1 if customer via the proposed model. The distinction of the presented
the vehicle d belongs to hub k service customer i, if study with previous studies includes;
not it equals 0.
Xi jk A zero-one variable, the value of which equals 1 if
customer j is serviced just subsequent to customer i 1) Responding to the two main challenges of supply chain man-
and if both customers are allocated to hub k, or else agers and policy makers, who seek to minimize the total trans-
it equals 0. portation cost (the first objective function of the presented
Yqt A zero-one variable, the value of which equals 1 if model), while maximize the quality of the delivered products
food q is delivered at temperature t, if not it equals with short shelf lifetime to the customers (the second objective
0. function of this study).
MTt A zero-one variable, the value of which equals 1 if 2) Combining the tactical/operational decisions with strategic de-
there is a compartment with temperature t in the cisions to present a novel hub location (HL) and fleet schedul-
vehicle, or else it equals 0. ing problems.
Y Mtq A zero-one covariate used to linearize the model. 3) Incorporation of simultaneous distribution of multiple perish-
able products (q; product index varying from 1 to Q) with dif-
ferent storage temperatures (t; storage temperature index vary-
ing from 1 to s) into the HLP model.

1184
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

4) Introducing the limited capacity of vehicles in the hubs (Zb ; ca-

pacity of each storage compartment) as a new parameter to
GHLP model which responds to one of the real-world chal-
lenges of transportation.
5) Meeting one of the indicators of green supply chains (GSC) by
defining carbon emissions’ cost as a parameter (ε f ; coefficient
for carbon emissions of full-loaded vehicles and εe ; coefficient
for carbon emissions of empty vehicles) in the presented bi-
objective GHLP model.

The remainder of this paper is structured as follows: Section 2

reviews the literature by focusing on the most recent studies re-
garding HL and CSCs. Then, the mathematical formulation is devel-
oped and discussed in section 3, whereas the validity of the model
is tested through a numerical example in section 4. Finally, in sec-
tion 5, the key findings of the paper are concluded and future di-
rections of this paper are outlined based on several recommenda-
tions. Fig. 1. Multiple location (Ghaffarinasab et al., 2018).

1.1. Literature review

As a response to the global environmental awareness, many re-

searchers have been focusing on various theories and concepts to
transform the production and consumption systems to a more sus-
tainable mode (Subramanian et al., 2017; Vergragt et al., 2014). On
the one hand, sustainable consumption systems focus on enhanc-
ing the sustainability understandings of the consumers about the
kinds of pollution and health hazards they face in their societies
because of the industries and incorporate entities and transform
their consumption patterns towards more sustainable behaviors
(Barber, 2007). On the other hand, sustainable production systems
pertain to the process of producing goods and providing services to
the consumers by considering the exploited natural resources, the
greenhouse emissions, pollution and waste caused by the associ-
ated processes (Barber, 2007). As a mediator between the sustain-
able production and consumption systems (Vergragt et al., 2014),
the transportation activities stand as the procedures for preparing, Fig. 2. Single location (Ghaffarinasab et al., 2018).
performing and managing operations for the conducive storage and
delivery of goods and services from the producer to the customers
with the ultimate objective of satisfying the expectations of the Figs. 1 and 2 demonstrate examples of single and multiple
customers (Paciarotti and Torregiani, 2020). Transportation activi- locations.
ties in the supply chain of industries includes but are not limited The HL network connections are categorized into three types: 1)
to the physical packaging, hub establishment, warehousing, load- the set of customer nodes- to -hubs connections, 2) inter-hub con-
ing and unloading, vehicle and route scheduling, and adjustment nections, 3) the hubs- to -customer nodes distribution (Validi et al.,
of the storage compartments’ in the distributing vehicles to ulti- 2014). Fig. 3 represents an example of an HL network.
mately deliver the products from the producers to the consumers. Hub networks are employed in different industries includ-
This study intends to investigate the hub location problem as ing airlines (Soylu and Katip, 2019; Tiwari et al., 2020),
one of the key challenges of sustainable supply chain designs transportation (Etemadnia et al., 2015; Fernández and Sgalam-
which aims to minimize the system’s total cost, considers the bro, 2020; Ghaffarinasab et al., 2018), goods and products delivery
carbon emissions within the distribution network, and maximize (Ghodratnama et al., 2018; Kartal et al., 2017; Rieck et al., 2014),
the quality of the delivered perishable products from the pro- and telecommunications (Kim and O’Kelly, 2009). Since the intro-
ducer to the consumers. Therefore, the relevant literature to this duction of HL as one of the most efficient transportation systems,
study have been discussed in the following sections as hub loca- it has drawn the attention of many scholars.
tion, perishable products, cold supply chains, multi-product, multi- The main reason for employing an HLP in an industry is to
temperature, and GSC. minimize the transportation network costs, by locating the hubs
as well as the customer nodes- to -hubs (Validi et al., 2014).
1.2. Hub location problem Considerable number of studies have investigated HLP to opti-
mize the total transportation costs (Ghodratnama et al., 2018;
HLP was introduced firstly by Goldman in 1969 Mohammadi et al., 2019; Soylu and Katip, 2019). Moreover, another
(Damgacioglu et al., 2015). This type of problem is defined as implication of HLP in an industry is to minimize the total emis-
a network consisting of n nodes and p transport center called sions cost (Dukkanci et al., 2019; Musavi and Bozorgi-Amiri, 2017).
hubs (Damgacioglu et al., 2015). Two main types of hub net-
work models are existed in the literature: single allocation and 1.3. Perishable products and cold supply chain
multiple-allocation. Within single allocations, only one hub serves
each non-hub node whereas, in multiple allocations, more than Perishable products require precise temperature control and
one hub serves each non-hub node (Ghaffarinasab et al., 2018). shorter travel time compared to other types of products in the

1185
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

Fig. 3. Schematic view of the Hub Location Problem.

transportation process, because of their shorter shelf lifetime and 1.4. Multi-products – multi-temperature distribution
reduced qualities over time (Hsu and Chen, 2014). These thermal
sensitive products are classified into two categorizations. The first In recent years, the application MTJD has increased due to the
group includes the products that should be consumed within a ongoing demands for perishable products. Reports indicate that
specified period of their shelf lifetime, and will be out of the con- global demand for perishable products such as refrigerated foods,
sumption cycle after being expired. Dairy products such as milk fresh fruits, flowers, and etc., has escalated from 42 million tons
and yogurt, pharmaceuticals such as vaccines and blood products in 1987 to 44 million tons in 1990 and 53 million tons in 20 0 0
fall into this category as they will be non-usable after their ex- (Cho and Hsu, 2008). More specifically, the simultaneous deliv-
piry dates (Coelho and Laporte, 2014; Liu et al., 2020). The second ery of foodstuffs with the need to transport at different temper-
group pertains to the products which will lose their quality gradu- atures is an important issue for the urban transportation system
ally over time until they become completely perished by the time (Hsu and Chen, 2014). Maintaining the quality of these products
of expiration. Fruits and vegetables are examples of this type of requires the transportation fleet to provide optimal temperature
perishable products (Coelho and Laporte, 2014). controls for the multi-temperature goods across the SC processes.
Therefore, the supply chain management (SCM) of perishable Nevertheless, most of the freezer transportation fleets that are cur-
products has certain characteristics and requirements which in- rently utilized in CSCs can solely deliver products with equal stor-
cludes the short shelf lifetime of products, the importance of age temperatures which implies that products with different stor-
timely delivery of products to the customers, quality assurance age temperatures have to be distributed through dissimilar deliv-
for products, special storage conditions, and special equipment eries (Hsu and Chen, 2014). Consequently, the disparity among the
and facilities for production, maintenance, and distribution (Soto- products’ storage temperatures requires the procurement of two or
Silva et al., 2017). These requirements combined with the ex- more transportation fleets, which increases the investment and op-
pansion of markets from the domestic to the global level, rigor- erational costs and decreases the efficiency of the available vehicles
ous competitive markets, and the necessity of meeting the ever- (Cho and Hsu, 2008).
growing customers’ demands in terms of the timely delivery of Therefore, the question about the optimal solution of distribut-
higher quality perishable products has led to the advent of CSC ing multi-temperature goods at the same time with lower logis-
(Hsu and Chen, 2014). tic costs has given rise to the emergent of a concept called multi-
Cold chain (CS) refers to the chain of refrigerated products, a product-multi-temperature simultaneous distribution. The Taiwan
particular type of goods which must always be controlled ther- Industrial Technology Research Center developed the MTJD to dis-
mally through warehousing and transportations (Li et al., 2006; tribute foods using vehicles capable of delivering multiple prod-
Saif and Elhedhli, 2015). The failure to deliver products promptly ucts with different storage temperatures simultaneously (Hsu and
could result in the imposed extra costs in the SCs including delay Chen, 2014; Kuo and Chen, 2010). The MTJD services have gained
fine, dismissing the employees, suspension of production lines, and more popularity to be used in stores, supermarkets, and cold per-
loss of market shares, whereas these challenges are more intensi- ishable goods markets (Hsu and Chen, 2014).
fied in the cold chain due to the characteristics of the distributed Not only the technological advancement in transportation sys-
products (Li et al., 2006). tems and multi-temperature refrigerated compartments have been
CSC and the supply chain management of perishable prod- offering new practical business solutions, but also the applica-
ucts have been growing fields of research among researchers. tion of MTJD has been a matter of discussion in several studies.
These studies mainly investigate models to maintain the qual- These studies mostly explore the opportunities that MTJD provides
ity of the products (Devapriya et al., 2017; Lin et al., 2020; to the perishable products industry in retaining the quality dur-
Ma et al., 2020), minimize the production and transportation ing the distribution processes (Agustina et al., 2014; Chen et al.,
cost (Bortolini et al., 2016; Soysal et al., 2014; Stellingwerf et al., 2018; Hsiao et al., 2017), minimizing the costs (transportation cost,
2018), and maximize the earned profit (Blanco et al., 2005; spoilage cost, etc.) (Hsiao et al., 2017; Pasandideh et al., 2015), and
de Keizer et al., 2017). maximizing the earned profit (Hsu and Chen, 2014), and reduce
the carbon emissions (Chen and Hsu, 2019).

1186
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

1.5. Green supply chain

Hub location Green supply chain Perishable products Limitations on the number of vehicles Transportation-Fleet Planning Cold supply chain Multi products-multi temperatures
In the past decades, the world has witnessed rapid changes due
to the advancement of technologies, the globalized world economy,
the expansion of global markets, accelerated growth in the produc-
tion sectors (Shafiei et al., 2020; Subramanian et al., 2017). How-
ever, these economic developments have brought about serious en-
vironmental ramifications namely climate change, global warming,
ozone depletion, deforestation, exhaustion of non-renewable re-
sources, hazardous waste, converted habitat, and threats to biodi-
versity (Khan et al., 2020). Hence, on the grounds of global en-

√
√
√

√
vironmental awareness, GSC was introduced to incorporate en-
vironmental considerations into SCs by decreasing the raw ma-
terial flows and minimizing the adverse effects of consump-
tion and production operations such as perilous wastes and CO2
emissions of transportation vehicles (Badi and Murtagh, 2019;

√
√

√
√
Zhalechian et al., 2017). Additionally, temperature-controlled trans-
portation produces even more emissions than non-temperature-
controlled transportation due to the additional fuel prerequi-
sites for refrigeration and as a result of refrigerant leakage
(Stellingwerf et al., 2018). Therefore, the environmental impacts
of temperature-controlled transportation vehicles on the environ-
ment must be taken into account when planning GSC of perishable
products. Several studies have investigated the GSC of perishable
products from the standpoint of optimizing the fuel and emission

√
√

√
√

√
√

√
√
√
√

√
cost through the distribution process (Babagolzadeh et al., 2020;
Govindan et al., 2014; Zhang et al., 2019).
To the best knowledge of the authors, although there have been
many studies addressing the HLP in the literature, none of them
have addressed it along with carbon emissions and consideration
of the simultaneous transportation and delivery of dissimilar stor-
age temperatures items within a CSC, in order to improve the qual-
ity of the delivered products to the customers and also minimize
the total transportation costs. Therefore, this paper intends to fill
this research gap and presents an investigation of the HLP by com-
bining the strategic decisions with tactical/operational decisions
to propose a model, which is capable of simultaneous delivery of
√

√
multiple perishable products with dissimilar storage temperatures,
and concentrates on meeting the bi- objectives of minimizing the
system’s total cost and maximizing the product quality delivered
to the customer. In order to have a better understanding of the
presented research gap, Table 1 provides a thematic comparison
between the reviewed the related studies and the current paper.
√

√
√

√
√
√

√
√
√

√
2. Methods

The proposed model in this paper is classified within the cate-

gory of location problem. There are two levels in a location prob-
lem: the first level is subjected to the potential locations of distri-
√

√
√

√
√
Review of and comparison between recent research papers.

bution centers (hubs) and the second one is concerned with the
customer-related locations. In order for the model to be solved,
the bi-objective problem is firstly defined. Then, the research as-
sumptions in which the presented model is based on are out-
√

√
√

√
√
√
√
√
√
√

lined. Finally, this bi-objective MILP optimization model is delin-

(Musavi and Bozorgi-Amiri, 2017)

(Fernández and Sgalambro, 2020)

eated through mathematical equations and constrains. The outputs

of the proposed model after being solved are information regard-
(Martins de Sá et al., 2015)

(Ghaffarinasab et al., 2018)

(Babagolzadeh et al., 2020)

(Ghodratnama et al., 2018)
(Mohammadi et al., 2019)
(Stellingwerf et al., 2018)

ing hub location, allocation of customers to hubs, and allocation of

(Etemadnia et al., 2015)

(Soylu and Katip, 2019)

(Devapriya et al., 2017)

(de Keizer et al., 2017)

(Dukkanci et al., 2019)

(Agustina et al., 2014)

(Bortolini et al., 2016)

(Hsu and Chen, 2014)

(Daofang et al., 2015)

customers to vehicles, and sequence of vehicles’ services.

(Tiwari et al., 2020)
(Soysal et al., 2014)

(Kartal et al., 2017)

(Hsiao et al., 2017)
(Rieck et al., 2014)

The current study

(Hsu et al., 2013)

(Lin et al., 2020)

2.1. Problem definition

Research

The presented bi-objective problem in this study is modeled

firstly, to minimize the total costs in the supply chain including
transportation, the hub establishment, adjustment of the storage
Table 1

No.

compartments’ temperatures, and carbon emission costs and sec-

11

13
14
15
16
17
18
19

21
22
23

24
25
10

12

20
1
2
3
4
5
6
7
8
9

ondly, to maximize the products’ qualities to meet the customers’

1187
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

expectations. Additionally, this GHLP model considers simultane- These two objective functions are subjected to several con-
ous distribution of multiple products with different storage tem- strains. Eq. (3) represents the quality function, also known as mar-
peratures. Fig. 3 demonstrates the schematic of a viable solution ket purchase probability function, which is a conducive linear func-
to an example of the discussed problem. In this example, amongst tion to capture the relationship between the quality of perishable
the total number of 7 customers, 3 of which are assigned to be products and time (Bortolini et al., 2016).
served by the first hub and the remaining customers are decided Constraints (4) and (5) have been used to linearize the relation-
to be served by the second hub. The existed vehicles in the hubs ships among variables of Eq. (3).
are equipped with temperature-controlled cold chambers. Accord-  
Ti
ing to Fig. 3, initially, the first vehicle goes to the first customer’s 1− Si
gi = min ,1 (3)
place and delivers his/her order. Then, it returns to the hub, goes to 1 − QRPi
the second customer, and delivers his/her order. The second vehi-
Ti
cle only has the third customer to serve. This problem is modeled 1− Si
based on a number of assumptions, the most important of which gi ≤ i = 1, ..., n (4)
1 − QRPi
are as follows:

1. The considered supply chain network has two levels (hubs and gi ≤ 1i = 1, ..., n (5)
customers). Constraint (6) denotes that the number of p hubs is located and
2. The number of hubs is predetermined. constraint (7) makes sure that each number of customer nodes is
3. The hubs have infinite capacity. allocated to only one hub. Constraint (8) states that any of the cus-
4. The vehicles in each hub are equipped with temperature- tomer nodes must be assigned nodes that are located as hubs.
adjustable cold chambers, the quantities of which are equal in
all hubs. Also, unlike most HLPs in which the number of vehi- 
n
xii = H (6)
cles in each hub is unlimited, this value is hypothesized as a
i=1
limited number.
5. Decrease in the products’ qualities occurs only at the time of 
n

transporting the items from hubs to customers. xik = 1i = 1, ..., n (7)

6. The customer’s demand is clear and constant. The demand of k=1

all customers should be met and each customer should be vis-

xik ≤ xkk i, k = 1, ..., n (8)
ited only once.
7. The customer’s demand is equal for each product. Constraint (9) implies the balance of flows among the nodes.
8. Vehicles of all the hubs are similar. Constraints (10) and (11) states that each one of the customer
9. The proposed model is multi-product and all parameters are nodes is allocated just to a hub that responds to its request.
definite and constant.    
10. The problem is a single HLP 
n 
n 
Q 
n 
Q 
n
eikl − eilk = wqi j xik − wqi j x jk i, k = 1, ..., n
l=1 l=1 q=1 j=1 q=1 j=1
2.2. Mathematical formulation l=k l=k

(9)
The presented bi-objective MILP optimization model is math-
ematically formulated via two equations and several constrains.

Q 
n
This model has two objective functions. The first objective function wqi j x jk ≤ Mdk j k, j = 1, ..., n; k = j (10)
(Eq. (1)) aims at the minimization of the total costs which includes q=1 i=1
transportation costs (first part of Eq. (1)), adjustment of the stor-
age compartments’ temperatures costs (second part of Eq. (1)), hub 
Q 
n

deployment costs (third part of Eq. (1)), and carbon emission costs dk j ≤ wqi j x jk k, j = 1, ..., n; k = j (11)
that has two portions (last part of Eq. (1)). The first and second q=1 i=1

portions of carbon emission costs function present the associated Constraint (12) points out that none of the customer nodes is
carbon emission when vehicles are less than full loaded and full served by itself. Constraint (13) shows the relationship between
loaded respectively. the customer nodes that are allocated to the hub which is served
 by the carrier fleet.

n 
n 
n 
Q 
n 
n
Min f1 = β Dik ( wqi j )xik + α β Dkl eikl dkk = 0k = 0, 1, ..., n + 1 (12)
i=1 k=1 j=1 q=1 k=1 l=1
  

n 
n 
Q 
Q s 
 
+ β Dki ( wqi j )xik + Y Mqt × [ f + h × (Tmax − t )] Ecik ≤ dki i, k = 1, ..., n; i = k; c = 1, ..., C (13)
k=1 j=1 q=1 q=1 t=1
⎡ ⎡ Q n
⎡ Q n
⎤⎤ ⎤ Constraint (14) shows the order of servicing for the fleet if both
( ε f − εe ) wqi j wqi j customer nodes i and j are allocated to a similar hub. Constraint

n ⎢n 
n ⎢ ⎢ q=1 i=1 ⎥⎥ ⎥
Fk xkk + λ⎢ Dk j ⎢ + εf ⎢ ⎥⎥dk j ⎥
q=1 i=1 (15) makes sure that each one of the customer nodes that is as-
+
⎣ ⎣ Vl ⎣ Vl ⎦⎦ ⎦ signed to the hub k is serviced by just a single fleet.
k=1 k=1 j=1

2Xi jk ≤ dki + dk j i, j, k = 1, ..., n; i = j (14)

(1)

The second objective function (Eq. (2)) focuses on the delivery 

C
Ecik = dki i, k = 1, ..., n + 1; i = k (15)
of higher quality products to the customers, as follows:
c=1

n
Max f2 = gi (2) Constraint (16) indicates that for each one of the customer
nodes there is exactly one customer node which has been serviced
i=1

1188
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

before if both of the customer nodes are allocated to the similar

hub. Constraint (17) states that there is exactly one customer node
for each one of the customer nodes that will be served afterward
if both are allocated to a similar hub.

n
Xi jk = dk j j, k = 1, ..., n; i = k (16)
i=0
i= j

n+1

Xi jk = dki i, k = 1, ..., n; i = k (17)
j=0
i= j

Constraint (18) makes sure that the imaginary customer node 0

is the first customer served by any transportation fleet in hub k, Fig. 4. Product quality function.
while constraint (19) controls for the imaginary customer n + 1 to
be the last one served by any transportation fleet in hub k.
Constraint (26) indicates that each q-type product is trans-

n
ported only at one temperature, while constraint (27) points out
X0 jk ≤ C xkk k = 1, ..., n (18) that only one q-type product is stored at each temperature t. Con-
j=1
j=k
straint (28) states that whether a q-type food is transported at
temperature t, there must be some storage space existed in the

n transportation fleet.
Xi,n+1,k ≤ C xkk k = 1, ..., n (19)

s
i=1
i=k Yqt = 1q = 1, ..., Q (26)
t=1
Constraint (20) indicates that whereas customer node j is
served just subsequent to customer node i and both of which are 
Q

assigned to hub k, then, either of these customer nodes must be Yqt = 1t = 1, ..., s (27)
q=1
served by a d transportation fleet.

Xi jk − 1 ≤ Ecik − Ec jk ≤ 1 − Xi jk i, j, k = 1, ..., n; i = j = k; c = 1, ..., C Yqt ≤ MTt q = 1, ..., Q; t = 1, ..., s (28)

(20) Constraints (29) and (30) have been used to linearize the tem-
perature regulation part of the objective function.
Constraint (21) indicates that if customer node i is served just
prior to customer node j then, the time of service delivery to cus- 2Y Mqt ≤ Yqt + MTt q = 1, 2, ..., Q; t = 1, 2, ..., s (29)
tomer j will be the minimum value of the returning time of vehicle
from customer i to hub k added to the arriving time of the trans-
2Y Mqt + 1 ≥ Yqt + MTt q = 1, 2, ..., Q; t = 1, 2, ..., s (30)
portation fleet from hub k to customer j.
In line with the quality function in Eq. (3), Fig. 4 shows that if i
T j − Ti + M (3 − Xi jk − Ecik − Ec jk ) T
makes Si < QRPi , the delivered product quality is maximum. Other-
≥ ϕik + ϕ jk i, j, k = 1, ..., n; i = j = k; c = 1, ..., C (21) i
wise, beyond the QRPi value, the quality of the perishable products
Constraint (22) makes sure that if customer i and j, which decreases steadily over time (Bortolini et al., 2016).
are allocated to hub k, are served subsequent to the imaginary
customer 0, they should be served by two various transportation 3. Results and discussion
fleets.
In order to validate the proposed model, this section provides
X0ik + X0 jk + Ecik + Ec jk ≤ 3i, j, k = 1, ..., n; i = j = k; c = 1, ..., C a numerical example via a set of standard data that is avail-
(22) able at the Operations Research Library Database. The employed
standard dataset is known as the O Kelly dataset or the CAB
Constraint (23) states that the time of servicing each customer
(Civil Aeronautics Board) dataset, and has been used by many re-
should not be less than the time of traveling from the hub to the
searchers to test and analyze various models (Musavi and Bozorgi-
customer node.
Amiri, 2017). Since the presented problem is a bi-product one,
Ti ≥ ϕik dki i, k = 1, ..., n; i = k (23) the minimum value of the two products’ life spans should be
determined as the final value of the product life span parame-
Constraint (24) pertains to the capacity of each cooling com- ter in the model solution. Hence, the value of the product life
partment in the transportation fleet. Constraint (25) makes certain span parameter of the presented numerical example is the mini-
that the total number of storage compartments in each transporta- mum value between 70 and 50 time units. Moreover, the starting
tion fleet is equal to LT . point of the product spoilage parameter determined based upon

n the product with the shortest lifespan. The values of other in-
(wqi j )Ecik ≤ Zb xik i, k = 1, ..., n; i = k; q = 1, ..., Q; volved parameters are determined in accordance with the study of
j=1 Musavi and Bozorgi-Amiri and are available in Table 2 (Musavi and
b = 1, ..., B; c = 1, ..., C (24) Bozorgi-Amiri, 2017). They performed a parameter tuning experi-
ence on the CAB database to ensure that the selected parameters
are proper in most examples (Musavi and Bozorgi-Amiri, 2017). To

s
enable this, a set of parameters is considered and a parameter is
MTt = LT (25)
altered while fixing others (Musavi and Bozorgi-Amiri, 2017). Their
t=1

1189
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

Table 2
Parameter values.

Parameter α λ β H εf εe νι Zb QRP h f C FK S LT

Value 0.5 10 1.2 5 1.096 0.772 40 500,000,000 0.4 3 10 3 1623,700,000 50 2

parameter tuning experience revealed that the following parameter

values in Table 2 provides suitable solutions (Musavi and Bozorgi-
Amiri, 2017).

3.1. Problem-solving

Subsequent to the determination of all of the model’s parame-

ters and inputs, the considered problem was coded in GAMS soft-
ware. The adoption of the ε -Constraint method in solving the pre-
sented bi-objective model enables the problem to transform into
a mono-objective and resolvable model. The ε -Constraint method
comprises of two stages; the first step pertains to the individual
optimization of each objective function and the following step is
the application of a bounded constraint on one of the objective
functions to obtain the Pareto frontier. In the present problem, the
bounded constraint was applied to the quality objective function. Fig. 5. Pareto approximation.
The determination of the optimal value of the cost objective
function (F1 ) firstly, requires the problem to be solved as a mono-
objective one, attempting to minimize the total cost. Secondly, the to the customer is at the least satisfactory state (F2 =0). F2∗ is the
quality of the delivered product to the customer needs to be deter- optimal value of the second objective function, which equals to 25
mined at the lowest total cost state, by inserting the values of the monetary units according to the previous section and the value of
variables in the second objective function (F2 ). As such, the optimal K + 1 defines the number of values obtained for Epsilon. Conse-
values of the total cost objective function and the quality of the de- quently, by inserting the defined values into Eq. (32) and randomly
livered product to the customer were obtained as 25,247,20 0,0 0 0 assuming the value of K equals to 3, the step length is obtained as
monetary units and zero, respectively. 8.3333 via the Eq. (33).
In order to determine the optimal value of the quality of the 25 − 0
delivered product to the customer function (F2 ), the problem is re- l= = 8.3333 (33)
3
quired to be solved as a mono-objective one, aiming at the maxi-
Knowing the step length to increase the Epsilon value in line
mization of the quality of the delivered product. Then, the value of
with Eq. (31), the Pareto frontier for this problem is obtained ac-
the total cost is needed to be calculated at the highest quality level
cording to Fig 5. The point (20.462 and 255,120) is selected as the
of the delivered product, by inserting the values of the variables in
balance point.
the first objective function (F1 ). Accordingly, the optimal values of
Finally, the mathematical solution was coded in the GAMS soft-
the objective function of maximizing the quality of the delivered
ware and the results regarding the nodes selected as hubs, the
product to the customers and the total cost of the system were
allocation of customer nodes to the hubs, the vehicle selected to
obtained as 25 and 282,727,20 0,0 0 0 monetary units, respectively.
serve each node, as well as the order of servicing to each hub cus-
3.2. Obtained results from the model solving tomer were derived as shown in Table 3.

After solving the problem for each of the objective functions 3.3. Sensitivity analysis
and determining the optimal values of each objective in the pre-
vious two sections, the Pareto approximation needs to be drawn To conduct the sensitivity analysis and provide difference sce-
through the ε -Constraint method. To enable this, it is necessary to narios for decision makers, three main parameters of the present
select one of the objectives as the primary objective and consider problem including the number of hubs located, demand rate, and
the other objective as the bounded constraint to the model. In this life span of the product were changed to investigate their asso-
study, the minimization of the system’s total costs was selected as ciated impacts on the two objective function values. Then, these
the main objective and the quality of the delivered product to the impacts based on the statistical analysis were presented.
customer was added to the model’s constraints, as demonstrated
in Eq. (31). 3.4. Sensitivity analysis of the number of located hubs
F2 ≥ ε (31)
The first sensitivity analysis concerns the effect of changing the
As an attempt to obtain the Pareto approximation, firstly, the number of hubs to be located on the objective function values. To
Epsilon values in Eq. (31) needs to be determined, which by it- enable this, all the parameters of the problem were kept constant
self requires the prior calculation of the Epsilon values step length. and the number of hubs was changed. Table 4 and Fig. 6 demon-
Eq. (32) represents the formula for calculating the Epsilon values strates the resulted impacts of the change in the number of hubs
step length. to be located on the objective function values.
F2∗ − F2
l= (32)
k 3.5. Sensitivity analysis of the demand value
In order to calculate the step length of increasing the Epsilon
value (L), the first objective function value is considered as the op- The second sensitivity analysis pertains to the effect of chang-
timal (F1 ), which implies that the quality of the delivered product ing the demand value on the values of the objective functions.

1190
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

Table 3
Solutions of the model solving.

Order of servicing Vehicle Allocated customers The node selected as a hub

15.9.11 1 5.6.8.9.11.15.21 4
6.8 2
5.21 3
23 1 19.22.23 12
19.22 2
10 1 1.7.10.16 13
16.7.1 2
24 1 24 14
3.25.2.20 1 2.3.18.20.25 17
18 3

Table 4
The results of sensitivity analysis of the hubs number.

Number of Cost objective function (F1 ) Product quality objective function (F2 )
hubs
Value (One hundred thousand monetary units) Percentage change Value Percentage change

4 256,648 0.60% 20.057 −1.98%

5 255,120 0.00% 20.462 0.00%
6 258,753 1.42% 21.475 4.95%
7 271,349 6.36% 19.969 −2.41%
8 285,128 11.76% 24.477 19.62%

Table 5
The results of sensitivity analysis of the demand value.

Parameters Cost objective function (F1 ) Product quality objective function (F2 )
No. Value
Value (One hundred thousand monetary units) Percentage change Value Percentage change

1 10% increase of demand 276,190 8.26% 19.731 −3.57%

2 30% increase of demand 304,407 19.32% 19.335 −5.51%
3 10% decrease of demand 236,438 −7.32% 21.263 3.91%
4 20% decrease of demand 218,424 −14.38% 21.905 7.05%
5 30% decrease of demand 203,771 −20.13% 21.94 7.22%

Fig. 6. The effect of changing the number of hubs on the objective functions.
Fig. 7. The impact of demand changes on the objective functions.

Similar to the first sensitivity analysis, while all the other param-
eters of the problem were kept constant, the value of the demand
was changed. The resulted effects of the increase in the demand Sensitivity analysis of the increase in the demand values and
value on the objective function values are depicted in Table 5 and the change in the number of hubs
Fig. 7. The fourth sensitivity analysis explores the impact of the
change in the number of hubs to be located on the objective func-
3.6. Sensitivity analysis of the product lifespan tion values in the case of increasing the demand values by 10%. For
this purpose, whilst all the other parameters of the problem were
The third sensitivity analysis delves into the effect of changing kept constant, the demand values increased by 10%, and the num-
the product lifespan on the values of the objective functions. Ac- ber of hubs was changed. The details on the effect of these changes
cordingly, the value of the product lifespan was changed, whereas on the two objective functions values are shown in Table 7 and
other parameters of the problem were unchanged. The results from Fig. 9.
the sensitivity analysis of the product lifespan are displayed in Sensitivity analysis of the decrease in the demand values and
Table 6 and Fig. 8. the change in the number of hubs

1191
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

Table 6
The results of sensitivity analysis of the product lifespan.

Product Cost objective function (F1 ) Product quality objective function (F2 )
lifespan
Value (One hundred thousand monetary units) Percentage change Value Percentage change

40 253,350 −0.69% 19.699 −3.73%

50 25,512 −90.00% 20.462 0.00%
80 255,000 −0.05% 23.189 13.33%
150 258,320 1.25% 25 22.18%
200 259,558 1.74% 25 22.18%

Table 7
The results of sensitivity analysis of the change in the hubs number and the 10% increase in the demand values.

Hubs Cost objective function (F1 ) Product quality objective function (F2 )
number
Value (One hundred thousand monetary units) Percentage change Value Percentage change

4 273,274 7.12% 17.847 −12.78%

5 276,190 8.26% 19.731 −3.57%
7 286,979 12.49% 22.604 10.47%
8 296,732 16.31% 24.199 18.26%

Fig. 10. The impact of the change in the hubs number and the 10% decrease in the
Fig. 8. The impact of the product lifespan changes on the objective functions.
demand values on the objective functions.

The final sensitivity analysis addresses the impact of the change

in the number of hubs to be located on the objective function val-
ues in the case of decreasing the demand values by 10%. To enable
this, all the other parameters of the problem were kept unchanged,
while the demand values decreased by 10% and the number of
hubs were changed. The details on the effect of these changes
on the objective functions’ values are shown in Table 8 and
Fig. 10.

4. Conclusion

This research presented a novel bi-objective MILP optimization

model for the GHLP with the ability of simultaneous distribution
of multiple perishable products at multiple storage temperatures.
Fig. 9. The impact of the change in the hubs number and the 10% increase in the
Furthermore, a numerical example via a set of standard data (CAB
demand values on the objective functions.
database) was coded in the GAMS software to validate the pro-

Table 8
The results of sensitivity analysis of the change in the hubs number and the 10% decrease in the demand value.

Hubs Cost objective function (F1 ) Product quality objective function (F2 )
number
Value (One hundred thousand monetary units) Percentage change Value Percentage change

5 236,438 −7.32% 21.263 3.91%

6 243,276 −4.64% 22.290 8.93%
7 254,005 −0.44% 23.006 12.43%
8 274,115 7.45% 23.348 14.10%

1192
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

posed model. Then, the ε -Constraint method was adopted to solve Declaration of Competing Interest
the presented bi-objective model by transforming the problem into
a mono-objective and obtaining the Pareto frontier. The solutions The authors declare that they have no known competing finan-
to the model provided information regarding the HL, allocation of cial interests or personal relationships that could have appeared to
customers to the hubs, allocation of customers to the vehicles, and influence the work reported in this paper.
the sequence of vehicles’ services. In the following, several param-
eters of the present problem consisting of the number of hubs lo- Supplementary materials
cated, demand rate, and life span of the product were changed to
investigate their associated impacts on the objective function val- Supplementary material associated with this article can be
ues. Ultimately, the final results revealed the existence of a con- found, in the online version, at doi:10.1016/j.spc.2021.02.026.
tradictory tradeoff between the two objectives of this paper, in-
dicating that the higher is the quality of the delivered perishable References
product to the customer, the greater is the total transportation cost
Accorsi, R., Gallo, A., Manzini, R., 2017. A climate driven decision-support
of the system. Hence, these results suggest that managers and key model for the distribution of perishable products. J Clean Prod 165, 917–
decision makers need to have deep insight about the associated 929.
effects of perishable products’ life span, number of hubs, and de- Adekomaya, O., Jamiru, T., Sadiku, R., Huan, Z., 2016. Sustaining the shelf life of fresh
food in cold chain–A burden on the environment. Alexandria Engineering Jour-
mand rates on the total transportation costs and the quality of the nal 55, 1359–1365.
delivered products within CSCs. Agustina, D., Lee, C., Piplani, R., 2014. Vehicle scheduling and routing at a cross
Theoretically, this study’s contributions pertain to the optimiza- docking center for food supply chains. International Journal of Production Eco-
nomics 152. doi:10.1016/j.ijpe.2014.01.002.
tion of transportation cost and quality objectives, several parame- Babagolzadeh, M., Shrestha, A., Abbasi, B., Zhang, Y., Woodhead, A., Zhang, A., 2020.
ters, and decision variables to the GHLP of the SCs of perishable Sustainable cold supply chain management under demand uncertainty and car-
products. Firstly, the presented model combined tactical/operation bon tax regulation. Transportation Research Part D: Transport and Environment
80, 102245.
decision and strategic decision to provide a novel and more effi-
Badi, S., Murtagh, N., 2019. SC. Journal of Cleaner Production. doi:10.1016/j.jclepro.
cient HL and fleet scheduling problems. Secondly, this model intro- 2019.03.132.
duces new parameters to the HLP that allows for the simultaneous Barber, J., 2007. Mapping the movement to achieve sustainable production and con-
sumption in North America 15, 499–512. 10.1016/j.jclepro.2006.05.010.
distribution of dissimilar storage temperatures perishable products
Blanco, A.M., Masini, G., Petracci, N., Bandoni, J.A., 2005. Operations management of
within a CSC. As far as the author of this paper concerns, no prior a packaging plant in the fruit industry. J Food Eng 70, 299–307. doi:10.1016/j.
studies have addressed these research agendas together. jfoodeng.2004.05.075.
Practically, this novel bi-objective MILP optimization model for Bortolini, M., Faccio, M., Ferrari, E., Gamberi, M., Pilati, F., 2016. Fresh food sustain-
able distribution: cost, delivery time and carbon footprint three-objective opti-
the GHLP provides three specific advantages for managers and pol- mization. J Food Eng 174, 56–67. doi:10.1016/j.jfoodeng.2015.11.014.
icy makers. Firstly, the presented model helps SCs managers to de- Chen, M.-.C., Lu, C.-.C., Liu, Y.-.C., 2018. Optimal consolidation of fresh agricultural
sign a distribution network for multiple products with short lifes- products in a multi-temperature joint distribution system. The International
Journal of Logistics Management.
pan, such as pharmaceutical, blood products, vaccines, fresh foods, Chen, W.-.T., Hsu, C.-.I., 2019. Optimal scheduling for multi-temperature joint dis-
vegetables and diary, which optimizes the quality and freshness of tribution under carbon tax. International Journal of Operations Research 16,
the delivered products to the customers. Secondly, this model aids 45–62.
Cho, Y.-.J., Hsu, Y.-.W., 2008. A study on the Mixed Fleet Multi-Temperature Com-
managers to respond to one of their main goals, as minimizing the mon Distribution: heuristics and Computational Analysis. Wseas Transactions
total transportation cost (transportation, hub establishment, and on Business and Economics 5, 140–149.
adjustment of the storage compartments’ temperatures) in the SCs Coelho, L.C., Laporte, G., 2014. Optimal joint replenishment, delivery and inven-
tory management policies for perishable products. Comput Oper Res 47, 42–52.
of perishable products. Finally, the presented model assists policy
doi:10.1016/j.cor.2014.01.013.
makers to meet one of the indicators of GSC establishments, by Damgacioglu, H., Dinler, D., Evin Ozdemirel, N., Iyigun, C., 2015. A genetic algorithm
minimizing the carbon emissions costs of fleet vehicles into one of for the uncapacitated single allocation planar hub location problem. Comput
Oper Res 62, 224–236. doi:10.1016/j.cor.2014.09.003.
its objective functions.
Daofang, C., Jinfeng, Z., Danping, L., 2015. Cold chain logistics distribution network
In spite of the theoretical contributions and managerial impli- planning subjected to cost constraints. International Journal of Advanced Sci-
cations of the proposed model, this study is subject to a few lim- ence and Technology 75, 1–10.
itations which could lead to further research. Firstly, the proposed de Keizer, M., Akkerman, R., Grunow, M., Bloemhof, J.M., Haijema, R., van der
Vorst, J.G.A.J., 2017. Logistics network design for perishable products with het-
model did not take into account the role of manufacturers in the erogeneous quality decay. Eur J Oper Res 262, 535–549.
distribution process and how the perishable products are firstly de- Devapriya, P., Ferrell, W., Geismar, N., 2017. Integrated production and distribution
livered from the manufacturers to the distribution centers (hubs) scheduling with a perishable product. Eur J Oper Res 259, 906–916.
Dubey, R., Gunasekaran, A., Papadopoulos, T., Childe, S.J., Shibin, K.T., Wamba, S.F.,
and then to the customers, which negatively affects the quality 2017. Sustainable supply chain management: framework and further research
of the delivered products. Therefore, future studies could incorpo- directions. J Clean Prod 142, 1119–1130. doi:10.1016/j.jclepro.2016.03.117.
rate the distribution process from the manufacturer to the hubs Dukkanci, O., Peker, M., Kara, B.Y., 2019. Green hub location problem. Transportation
Research Part E: Logistics and Transportation Review 125, 116–139.
in order to extend the distribution process of the model from the Etemadnia, H., Goetz, S.J., Canning, P., Tavallali, M.S., 2015. Optimal wholesale facil-
manufacturer to the customers and bring the model even closer to ities location within the fruit and vegetables supply chain with bimodal trans-
the real-world transportation problems. Secondly, the capacities of portation options: an LP-MIP heuristic approach. Eur J Oper Res 244, 648–661.
doi:10.1016/j.ejor.2015.01.044.
the hubs were hypothesized to be unlimited in this study. How-
Fernández, E., Sgalambro, A., 2020. On carriers collaboration in hub location prob-
ever, further research in this field may benefit from considering lems. Eur J Oper Res 283, 476–490.
the capacity of the hubs to be limited and bring the presented Ghaffarinasab, N., Motallebzadeh, A., Jabarzadeh, Y., Kara, B.Y., 2018. Efficient simu-
lated annealing based solution approaches to the competitive single and mul-
multi-objective GHLP more veracious. Thirdly, future studies may
tiple allocation hub location problems. Comput Oper Res 90, 173–192. doi:10.
develop an HL-routing problem on the grounds of the presented 1016/j.cor.2017.09.022.
HLP. Fourthly, due to the ever-growing environmental awareness, Ghodratnama, A., Arbabi, H.R., Azaron, A., 2018. A bi˗ objective hub location-alloca-
future research is needed to delimitate other sustainability objec- tion model considering congestion. Operational Research 1–40.
Govindan, K., Jafarian, A., Khodaverdi, R., Devika, K., 2014. Two-echelon multiple-
tives in addition to the investigated CO2 emissions in the presented vehicle location–routing problem with time windows for optimization of sus-
multi-objective GHLP. Finally, future studies could fruitfully explore tainable supply chain network of perishable food. International Journal of Pro-
this problem further by considering other parameters including the duction Economics 152, 9–28. doi:10.1016/j.ijpe.2013.12.028.
Hsiao, Y.-.H., Chen, M.-.C., Chin, C.-.L., 2017. Distribution planning for perishable
demand, the vehicles’ speed, and the time interval between nodes foods in cold chains with quality concerns: formulation and solution procedure.
as uncertain and variable parameters. Trends Food Sci Technol 61, 80–93. doi:10.1016/j.tifs.2016.11.016.

1193
M. Golestani, S.H. Moosavirad, Y. Asadi et al. Sustainable Production and Consumption 27 (2021) 1183–1194

Hsu, C.-.I., Chen, W.-.T., 2014. Optimizing fleet size and delivery scheduling for Rieck, J., Ehrenberg, C., Zimmermann, J., 2014. Many-to-many location-routing with
multi-temperature food distribution. Appl Math Model 38, 1077–1091. doi:10. inter-hub transport and multi-commodity pickup-and-delivery. Eur J Oper Res
1016/j.apm.2013.07.036. 236, 863–878. doi:10.1016/j.ejor.2013.12.021.
Hsu, C.-.I., Chen, W.-.T., Wu, W.-.J., 2013. Optimal delivery cycles for joint dis- Saif, A., Elhedhli, S., 2015. simulation-optimization approach. Eur J Oper Res 0 0 0,
tribution of multi-temperature food. Food Control 34, 106–114. doi:10.1016/j. 1–14. doi:10.1016/j.ejor.2015.10.056.
foodcont.2013.04.003. Shafiei, M., Moosavirad, S.H., Azimifard, A., Biglari, S., 2020. Water consumption
Kartal, Z., Hasgul, S., Ernst, A.T., 2017. Single allocation p-hub median location and assessment in Asian chemical industries supply chains based on input–output
routing problem with simultaneous pick-up and delivery. Transportation Re- analysis and one-way analysis of variance. Environmental Science and Pollution
search Part E: Logistics and Transportation Review 108, 141–159. Research 1–14.
Khan, A., Chenggang, Y., Bano, S., Hussain, J., 2020. The Empirical Relationship Be- Soto-Silva, W.E., González-Araya, M.C., Oliva-Fernández, M.A., Plà-Aragonés, L.M.,
tween Environmental degradation, Economic growth, and Social Well-Being in 2017. Optimizing fresh food logistics for processing: application for a large
Belt and Road Initiative countries. Environmental Science and Pollution Re- Chilean apple supply chain. Computers and Electronics in Agriculture 136,
search International. 42–57.
Kim, H., O’Kelly, M.E., 2009. Reliable p-hub location problems in telecommunication Soylu, B., Katip, H., 2019. A multiobjective hub-airport location problem for an air-
networks. Geogr Anal 41, 283–306. line network design. Eur J Oper Res 277, 412–425.
Kuo, J.-.C., Chen, M.-.C., 2010. Developing an advanced Multi-Temperature Joint Dis- Soysal, M., Bloemhof-Ruwaard, J.M., van der Vorst, J.G.A.J., 2014. Modelling food
tribution System for the food cold chain. Food Control 21, 559–566. doi:10.1016/ logistics networks with emission considerations: the case of an international
j.foodcont.20 09.08.0 07. beef supply chain. International Journal of Production Economics 152, 57–70.
Li, S., Ragu-Nathan, B., Ragu-Nathan, T.S., Subba Rao, S., 2006. The impact of sup- doi:10.1016/j.ijpe.2013.12.012.
ply chain management practices on competitive advantage and organizational Stellingwerf, H.M., Kanellopoulos, A., van der Vorst, J.G.A.J., Bloemhof, J.M., 2018. Re-
performance. Omega (Westport) 34, 107–124. doi:10.1016/j.omega.20 04.08.0 02. ducing CO2 emissions in temperature-controlled road transportation using the
Lin, Q., Zhao, Q., Lev, B., 2020. Cold chain transportation decision in the vaccine LDVRP model. Transportation Research Part D: Transport and Environment 58,
supply chain. Eur J Oper Res 283, 182–195. 80–93.
Liu, W., Ke, G.Y., Chen, J., Zhang, L., 2020. Scheduling the distribution of blood prod- Subramanian, N., Gunasekaran, A., Jose, C., Jabbour, C., Yusuf, Y., Azapagic, A., 2017.
ucts: a vendor-managed inventory routing approach. Transportation Research Sustainable Global Operations Management and Frugal Innovative Sustainable
Part E: Logistics and Transportation Review 140, 101964. Production methods: Advancing theory and Practice For a Truly Sustainable So-
Ma, X., Wang, J., Bai, Q., Wang, S., 2020. Optimization of a three-echelon cold chain ciety. Sustainable Production and Consumption doi:10.1016/j.spc.2017.06.002.
considering freshness-keeping efforts under cap-and-trade regulation in Indus- Tiwari, R., Jayaswal, S., Sinha, A., 2020. Alternate Solution Approaches for Competi-
try 4.0. International Journal of Production Economics 220, 107457. tive Hub Location Problems. Eur J Oper Res.
Martins de Sá, E., Contreras, I., Cordeau, J.-.F., 2015. Exact and heuristic algorithms Validi, S., Bhattacharya, A., Byrne, P.J., 2014. Integrated low-carbon distribution sys-
for the design of hub networks with multiple lines. Eur J Oper Res 246, 186– tem for the demand side of a product distribution supply chain: a DoE-guided
198. doi:10.1016/j.ejor.2015.04.017. MOPSO optimiser-based solution approach. International Journal of Production
Mohammadi, M., Jula, P., Tavakkoli-Moghaddam, R., 2019. Reliable single-allocation Research 52, 3074–3096. doi:10.1080/00207543.2013.864054.
hub location problem with disruptions. Transportation Research Part E: Logistics Vergragt, P., Akenji, L., Dewick, P., 2014. Sustainable production, consumption, and
and Transportation Review 123, 90–120. livelihoods : global and regional research perspectives. J Clean Prod 63, 1–12.
Musavi, M., Bozorgi-Amiri, A., 2017. A multi-objective sustainable hub location- doi:10.1016/j.jclepro.2013.09.028.
scheduling problem for perishable food supply chain. Computers & Industrial Zhalechian, M., Tavakkoli-moghaddam, R., Rahimi, Y., Jolai, F., 2017. An interactive
Engineering 113, 766–778. doi:10.1016/j.cie.2017.07.039. possibilistic programming approach for a multi-objective hub location problem :
Paciarotti, C., Torregiani, F., 2020. The Logistics of the Short Food Supply chain: a economic and environmental design. Applied Soft Computing Journal 52, 699–
Literature Review. Sustainable Production and Consumption. 713. doi:10.1016/j.asoc.2016.10.002.
Pasandideh, S.H.R., Niaki, S.T.A., Asadi, K., 2015. Bi-objective optimization of a multi- Zhang, L.-.Y., Tseng, M.-.L., Wang, C.-.H., Xiao, C., Fei, T., 2019. Low-carbon cold chain
product multi-period three-echelon supply chain problem under uncertain en- logistics using ribonucleic acid-ant colony optimization algorithm. J Clean Prod
vironments: NSGA-II and NRGA. Inf Sci (Ny) 292, 57–74. doi:10.1016/j.ins.2014. 233, 169–180.
08.068.

1194