Robust Aggregate Production Planning in A Green Supply Chain Under Uncertainty Considering Reverse Logistics: A Case Study
Robust Aggregate Production Planning in A Green Supply Chain Under Uncertainty Considering Reverse Logistics: A Case Study
Robust Aggregate Production Planning in A Green Supply Chain Under Uncertainty Considering Reverse Logistics: A Case Study
DOI 10.1007/s00170-016-9459-6
ORIGINAL ARTICLE
Received: 16 June 2016 / Accepted: 6 September 2016 / Published online: 27 September 2016
# Springer-Verlag London 2016
Abstract In response to the environmental problems in the computational results demonstrate the effectiveness and ro-
world, organizations need to apply green indicators to gain bustness of the model. The cost analysis is carried out to
competitive advantages. Moreover, manufacturers need to ful- provide useful managerial insights. The total costs and the
fill the demands in such a way as to generate a stable plan in profit in collection and recycling system, is also analyzed to
the real world. In this paper, a robust optimization approach is indicate its performance.
developed to solve a multi-site, multi- period, multi-product
aggregate production planning problem in a green supply Keywords Aggregate production planning . Green supply
chain considering potential collection and recycling centers chain management . Reserve logistics . Uncertainty . Robust
under uncertainty. The specific number of collection and optimization
recycling centers among potential centers can be constructed
to manufacture the second-class products. Some green princi-
ples such as waste management, greenhouse gas emissions 1 Introduction
related to transportation modes, and production methods are
embedded in the model. The objective function sets out to Nowadays, there has been an increasing awareness in the issue
minimize total losses of considered supply chain. Demand of world’s environmental problems including air pollution,
fluctuations and cost parameters are subject to the uncertainty. climate changing, global warming, and decreasing in the
A set of discrete scenarios are employed to illustrate the un- non-renewable resources. In the face of such thread, consid-
certainties. The robust optimization approach is chosen to re- erable attempts have been made to determine how green sup-
duce the impacts of fluctuations of uncertain parameters ply chain management (GSCM) has any role to play in con-
concerning all possible scenarios in the future. A case study trolling pollution and preserving natural resources. In last few
from an Iranian Wood and Paper Industries Company is stud- years, there has been a growing interest in GSCM models (see,
ied to indicate the practicability of the proposed model. The e.g., [1–13]). At the same time, as regards the significance of
environmental performance, green principles have been
adopted in several organizations [14]. Green supply chain
* Arezoo Entezaminia
management has been featured as a set of environmental man-
a_entezaminia@alumni.iust.ac.ir
agement principles which could be utilized to taken into ac-
Mahdi Heidari count environmental considerations in the forward and reverse
Mheydari@iust.ac.ir logistics process [15–18]. GSCM can be defined as “integrat-
Donya Rahmani
ing environmental thinking into supply chain management,
drahmani@kntu.ac.ir including product design, material sourcing and selection,
manufacturing processes, delivery of the final products to
1
Department of Industrial Engineering, Iran University of Science and the consumers, and end-of-life management of the product
Technology, Tehran, Iran after its useful life” [19]. A wide range of research on
2
Department of Industrial Engineering, K. N. Toosi University, GSCM from more than 500 papers have been reviewed by
Tehran, Iran Min and Kim [20].
1508 Int J Adv Manuf Technol (2017) 90:1507–1528
In recent years, the concepts of green logistics and reverse optimization model and considered the tradeoff between ex-
logistics have become a central issue in green supply chain pected cost and its variability. C.H. Leung, O.S. Tsang [42]
management [19]. In the literature review of green logistics developed a robust optimization model to deal with a multi-
reported by Dekker, Bloemhof [21], it has been stressed that in site APP problem. In their research, a robust model was pro-
regard to environment issues, transportation plays an essential posed to make a production plan for perishable products in an
part in supply chains. As for transportation, four choices in uncertain manufacturing environment. Kazemi Zanjani, Ait-
operation research models, namely mode choice, equipment Kadi [43] took into account two robust optimization models
choice, fuel choice as well as the intermodal transport have with variability of one form or another for a multi-product,
been taken into account. Each transportation mode enjoys multi-period production planning problem considering the un-
different cost, accessibility, transit time, and environmental certainty in the quantity of materials. Pan and Nagi [44] fo-
performance. In the present study, mode choice is used to cused on optimization of production and logistics costs related
incorporate the environmental considerations into the model. to supply chin entities. A robust model was formulated to
Reverse logistics process makes effort to get the products handle the uncertain demand. Niknamfar and colleagues
back, from their final destinations to the suppliers, after use, [45] presented a mathematical model for a multi-period
for the purpose of remanufacturing, or proper disposal. multi-product production-distribution problem in a three-
Nowadays, RL is of strategic importance for organizations. level supply chain. Mirzapour Al-E-Hashem et al. [46, 47]
Efficient RL, as an important component of GSCM, can pro- presented multi-site APP problems in an uncertain environ-
vide competitive advantage and social image (see, e.g., ment. Multi-objective robust optimization models were pro-
[22–27]). Taking everything into consideration, the potential posed for the considered problems.
collection and recycling centers are considered in our pro- In reviewing the literature, far too little attention has been
posed model to collect products from demand points and re- paid to adopting green concepts in APP problems under un-
cycle them as the second-class products, after separating the certainty. The first study of multi-site APP models in a green
recyclable ones. supply chain under uncertainty is presented by Mirzapour Al-
What is beyond question is that one of the prominent prob- e-hashem, Baboli [48]. Some green indicators including
lems that need to be addressed in green supply chain manage- greenhouse gas (GHG) emissions arising from transportations
ment is aggregate production planning (APP). The purpose of and waste management are embedded in their proposed mod-
APP is to determine the optimal quantity of production, in- el. This study sets out with the aim of dealing with a multi-
ventory as well as workforce levels to meet the demand for period multi-product multi-site aggregate production system
available resources and capacities during a planning horizon, in a green supply chain considering collection and recycling
in advance of 3 to 18 months [28, 29]. Researchers have centers. A mixed-integer programming model is proposed to
shown an increased interest in APP problems from several minimize the total costs. Finally, using robust optimization
years ago (see, e.g., [30–36]). Nam and Logendran [37] pre- approach, a case study is used to validate the efficiency and
sented a literature review of the studies on the APP models effectiveness of the proposed model.
consist of 140 journal articles and 14 books. The major contributions of this study that differentiate it in
As aggregate production planning in many real-world the related literature can be summarized as follows:
problems draw on some critical parameters such as cost, de- Production method:
mand, price, and capacity with uncertain values, it is worth-
while to focus on the way in which uncertainty effects the & Apart from waste management, greenhouse gas emissions
planning process. A variety of approaches have been proposed related to transportation, multiple production methods are
to address various forms of uncertainty. Stochastic approaches considered to adopt green principles in the model. It is
were used in production planning models to consider some assumed that the level of greenhouse emission related to
parameters as random variables with determined probability differing production methods is varied and it is limited by
distributions [38]. In fuzzy approach, variables have been con- a specific level.
sidered as fuzzy numbers to embed uncertainty in manufactur- & Some potential collection and recycling centers are taken
ing systems [39]. Robust optimization was extensively used in into consideration in order to collecting and recycling re-
production planning models considering various scenarios cyclable returned products and producing the second-class
which illustrate the uncertainty of parameters. This strong products. The quantity of products that will be returned to
and efficient technique generates stable solutions in response customers’ zones are considered to be determined in ad-
to any changes in uncertain parameters [40]. vance. It can be identified by forecasting methods based
The concept of stochastic robust optimization was present- on information available in the previous periods. The frac-
ed by M. Mulvey and Vanderbei [40]. C. H. Leung and Wu tion of returned products will be transported to collection
[41] took an aggregate production planning into consideration centers and those are not collected will be penalized by the
with uncertain parameters. They proposed a robust penalty costs. After identifying recyclable products in
Int J Adv Manuf Technol (2017) 90:1507–1528 1509
each collection center, they will be transported to deterministic. Band C,the coefficient matrixes of control con-
recycling centers. Moreover, the recycled products are straint, are random and subject to noise. eis the right-hand side
supplied to the demand points as the second-class vector.
products. In robust optimization model, a finite set of scenarios
& To the best of our knowledge, the present study is among Ω = {1, 2, ... , ξ} is considered to model uncertain parameters.
the first attempts to embed the concepts of green and re- The subset { dξ; Bξ; Cξ; eξ} represents the uncertain parameters
verse logistics in aggregate production planning under un- under the scenarioξ. pξ is the probability of scenarioξ and
certainty. A set of discrete scenarios are considered to ∑ξ pξ ¼ 1. The infeasibility of the model is denoted by δξ for
illustrate the uncertainty of customer demands, procure- each scenario. On condition that the model is feasible,δξ will
ment costs, production costs, inventory costs, shortage be equal to 0.The scenario based robust optimization model is
costs, opening costs of collection and recycling centers, formulated as follows:
handling costs in collection centers, as well as recycling
costs. A robust optimization model is presented to make Min σ x; y1 ; y2 ; :::; yξ þ ωρ δ1 ; δ2 ; :::; δξ ð5Þ
stable production and recycling plans that shows less sen-
sitivity to the fluctuation of input data. subject to
Ax ¼ b; ð2Þ
Bx þ Cy ¼ e; ð3Þ Equation (10) is a nonlinear objective function. C.H.
Leung, O.S. Tsang [42] converted the problem to a linear
x; y ≥ 0; ð4Þ model by developing two deviational non-negative variables.
Equation (10) is reformulated in the following way:
where xrepresents the vector of decision variables and y shows
" ! #
that of control variables. The structural and control constraints X X X
are respectively shown in constraints (2) and (3). A, the coef- Min pξ ψ ξ þ λ pξ ψ ξ − pξ0 ψξ0 þ 2θξ ð11Þ
ficient matrix of structural constraint, is free of noise and ξ∈Ω ξ∈Ω ξ0 ∈Ω
1510 Int J Adv Manuf Technol (2017) 90:1507–1528
X
ψξ − pξ ψξ þ θξ ≥ 0; ∀ξ∈Ω; ð12Þ method. It is assumed that only one raw can be used to
ξ∈Ω produce each product. Each plant is featured by its inven-
tory and production capacities. Production costs in differ-
θξ ≥ 0; ∀ξ∈Ω; ð13Þ ing plants and procurement costs in differing suppliers
and can be different. Moreover, plants are capable of pro-
Broadly, the objective function can be shown as follows in ducing the products in both regular and over time. It is
whichρ(δ1, δ2, ... , δξ) is considered to penalize the violation of supposed that, each collection center is characterized by
control constraints as a penalty function. The tradeoff between its opening cost, handling cost and minimum service lev-
the model and solution robustness is modeled by the weightω. el. Furthermore, there are some potential recycling centers
with differing opening and recycling costs. All entities of
" ! # supply chain including suppliers, plants, customer’s
X X X
Min pξ ψ ξ þ λ pξ ψξ − pξ 0 ψ ξ 0 þ 2θξ zones, collection centers, and recycling centers are geo-
ξ∈Ω ξ∈Ω ξ0 ∈Ω graphically scattered. So, transportation cost may be dif-
X ferent from one site to the other sites. The unit transpor-
þω pξ δ ξ : ð14Þ tation cost determines based on the vehicle type that need
ξ∈Ω to be selected among available vehicle types. It is possible
to store the first-class products in each factory. In the
same way, it is possible to store the second-class products
in each recycling center. Customers are able to store the
3 Model description first-class and the second-class products.
Taking everything into account, the related decisions
In this section, the proposed multi-product multi-site multi- consist of: (1) identifying the quantity of product type n
period aggregate production planning problem in a green sup- manufactured at plant j in regular and over time to fulfill
ply chain considering collection and recycling centers can be the demand of demand point i in each period; (2) deter-
described as follows: mining the amount of raw material n provided by supplier
As can be seen from Fig. 1, the supply chain under k to fulfill the requirements of factory j regarding lead
consideration in this study consists of particular sites, in- times; (3) determining the amount of returned product n
cluding suppliers, factories (plants), customers, collection recycled at recycling center j to meet the demand of de-
and recycling centers. We grouped some products into N mand point i in each period; (4) identifying the potential
product families and called them products (indexed by n) collection and recycling centers that need to be opened;
in this paper. The proposed mathematical model considers (5) identifying the number of fired and hired workers in
a set of K suppliers (indexed by k), a set of J plants each time period; (6) identifying the quantity of product n
(indexed by j), a set of I customers (indexed by i), a set that should be stored in factory j, customer i and recycle
of C collection centers (indexed by c), and a set of R center r in each period of time; (7) specifying the mode of
recycling centers (indexed by r), a time horizon of T pe- the transportation utilized for shipping the products; (8)
riods (indexed by t). Moreover, there are G (indexed by g) identifying the quantity of demand that fails to be fulfilled
transportation mode and L (indexed by l) production at each demand point in each period. It is worth noting
that the objective of the problem is to minimize the total d4cr The distance between collection center c and recycling center r
losses of supply chain. d5ri The distance between recycling center r and demand point i
P1ni Sale price of the first-class product n in customer zone i
P2ni Sale price of the second-class product n in customer zone i
3.1 Notations
AFnjl Greenhouse gas emission associated with manufacturing the
first-class product n at factory j by production method l
ALjt Allowed greenhouse gas emission related to production in plant
j in each period of time
Parameters Snit The amount of returned product n from demand point i in each
period of time
D1nit Demand for the first-class product n in customer zone i in pe- ηc Construction cost for each collection center
riod of time t
CCc Unit handling cost at each collection center
D2nit Demand for the first-class product n in customer zone i in pe-
Ci Penalty cost per unit of uncollected returned product from
riod of time t
customer zone i
anjl Manpower to manufacture each unit of product n in factory j by
RCnr Unit recycling cost of returned product n at recycling center r
the method l
νn The volume of per unit product n (m3)
α Allowed fraction of overtime
Vg Capacity of vehicle type g
LT1kjg The lead time needed to transport the raw material from supplier
k to factory j via vehicle type g GHGg CO2 emissions of per distance unit for transportation mode g
LT2jig The lead time needed to transport the first-class product from AG1jt Allowed greenhouse gas emission of transportation related to
plant j to customer zone i via vehicle type g factory j in each period
LT3crg The lead time needed to transport the returned product from AG2t Allowed greenhouse gas emission of transportation related to
collection center c to recycling center r via vehicle type g collection and recycling centers in each period of time
LT4rig The lead time needed to transport the second-class product from WTj Allowed amount of waste produced in each factory
recycling center r to customer zone i via vehicle type g WPnjl Waste percentage of product n manufactured at factory j by
CAPPj Storage capacity in plant j method l
CAPDi Storage capacity in demand point i β Recyclable percentage of returned products
CAPRr Storage capacity in recycling center r ECr Construction cost for recycling center r
YRMnr Maximum production of the second-class product n in recycling ωc Maximum capacity of collection center c
center r NC Maximum number of collection centers that can be opened
XRMnjl Maximum production of the first-class product n at factory j by NR Maximum number of recycling centers that can be opened
method l in regular time λ Minimum service level in each collection center
XOMnjl Maximum production of the first-class product n at factory j by πfni Storage cost of the first-class product n in customer zone i
method l in overtime
πsni Storage cost of the second-class product n at customer zone i
CSnk The number of product n provided by supplier k in each period
Continuous variables
wn Maximum shortage fraction of product n
YRnrt The amount of the second-class product n produced at recycling
CRnjl Variable production cost of product n at plant j by method l in center r in period of time t
regular time
XRnjlt The quantity of the first-class product n produced at plant j by
COnjl Variable production cost of product n at plant j by method l in production method l in regular time of period t
overtime
XOnjlt The quantity t of the second-class product n produced at plant j
PCnjl Fixed production cost of product n at plant j by method l by production method l in overtime of period t
LCj Labor cost in plant j XTnkjgt The amount of product n from supplier k to plant j by vehicle
FCj Firing cost in factory j type g in time period t
HCj Hiring cost in factory j YTnjigt The amount of product n from plant j to demand point i by
CIPnj Inventory holding cost for the first-class product n at plant j vehicle type g in time period t
CIDni Inventory holding cost for product n in the first and second class ZTnicgt The amount of product n from customer zone i to collection
at demand point i center c by vehicle type g in time period t
CIRnr Inventory holding cost for the second-class product n at RTncrgt The amount of product n from collection center c to recycling
recycling center r center r by vehicle type g in time period t
VTCg Variable transportation cost of mode g PTnrigt The amount of product n from recycling center r to demand
point i by vehicle type g in time period t
FTCg Fixed transportation cost of chosen mode g
XVkjgt The amount of transportation mode g required for transportation
CMnk Procurement cost of product n that can be provided by supplier
between supplier k and factory j in period t
k
YVjigt The amount of transportation mode g required for transportation
d1kj The distance between the supplier k and factory j
between factory j and demand point i in period t
d2ji The distance between the plant j and customer zone i
ZVicgt The amount of transportation mode g required for transportation
d3ic The distance between customer zone i and collection center c between demand point i and collection center c in period t
1512 Int J Adv Manuf Technol (2017) 90:1507–1528
RVcrgt The amount of transportation mode g required for transportation bsnit Shortage of the second-class product n at demand point i in
between collection center c and recycling center r in period t period t
PVrigt The amount of transportation mode g required for transportation Binary variables
between recycling center r and demand point i in period t Ynjlt 1 if product n produced at factory j by method l in period t; 0
FWjt Number of fired workers at plant j in period t otherwise.
HWjt Number of hired workers at plant j in period t Znjl 1 if product n produced at factory j by method l; 0 otherwise.
AWjt Available workers at plant j in period t Rc 1 if collection center c is opened; 0 otherwise.
IPnjt Inventory of the first-class product n at the end of period t at Xr 1 if recycling center r is opened; 0 otherwise.
plant j
ID1nit Inventory of the first-class product n in demand point i at the
end of period t
ID2nit Inventory of the second-class product n in demand point i at the 3.2 Problem formulation
end of period t
IRnrt Inventory of the second-class product n at recycling center r at The multi-objective multi-period multi-product multi-site ag-
the end of period t gregate production planning model can be formulated as
bfnit Shortage of the first-class product n at demand point i in period t follows:
!
X X X X
MinTC ¼ PCnjl Z njl þ CRnjl XRnjlt þ COnjl XOnjlt þ CMnk XTnkjgt
n; j;l n; j;l;t n; j;l;t
! n;k; j;g;t
X X X
þ LC j AW jt þ FC j W jt þ HC j HWjt
j;t j;t j;t
!
X X X
þ IPnjt CIPnj þ ID1nit þ ID2nit CIDni þ IRnrt CIRnr
n; j;t n;i;t n; j 0 ;t
X X
þ FTCg XVkjgt þ YVjigt þ ZVicgt þ RVcrgt þ PVrigt þ VTCg d1kj XVkjgt ð1Þ
k; j;i;c;r;g;t k; j;g;t
!
X X X X
þ VTCg d2ji Y Vjigt þ VTCg d3ic ZVicgt þ VTCg d4cr RVcrgt þ VTCg d5ri PVrigt
i; j;g;t i;c;g;t
! c;r;g;t r;i;g;t
X X X X
þ ZTnicgt CCc þ η c Rc þ RCnr YRnrt þ X r ECr
i:c;g;t c n;r;t
! r !
X X X X X
þ ðπf ni bf nit þ πsni bsnit Þ þ C i S nit − ZTnicgt − P1ni YTnjigt þ P2ni PTnrigt
i;n;t i;c;n;t;g g;c n;i; j;g;t n;i;r;g;t
n þ
X X X
ZVicgt GHGg d3ic þ RVcrgt GHGg d4cr
IRnrt ≤ CAPRr ∀r; t ð8Þ i;c;g c;r;g
n X
bf nit ≤ wn D1nit ∀n; i; t ð9Þ þ PVrigt GHGg d5ri ≤ AG2t ∀t ð17Þ
r;i;g
Int J Adv Manuf Technol (2017) 90:1507–1528 1513
X X
XRnjlt þ XOnjlt AFnjl ≤ALjt ∀ j; t ð18Þ RVcrgt −1 V g ≤ νn RTncrgt ≤ RVcrgt
n;l n
X V g ∀r; i; g; t ð30Þ
XRnjlt þ XOnjlt ≤ XTnkjgðt−LT kjg Þ ∀n; j; l; t ð22Þ
k;g
X
PTnrigt ≤ YRnrt ∀r; n; t ð31Þ
X i;g
YTnjigt ≤ XRnjlt þ XOnjlt ∀n; j; l; t ð23Þ
i;g
X
YRnrt ≤ RTncrgðt−LT crg Þ ∀r; n; t ð32Þ
Y njlt c;g
≤ XRnjlt þ XOnjlt ≤ Y njlt M ∀n; j; l; t ð24Þ X
M
X YRnrt ≤ M X r ∀r ð33Þ
Y njlt ≤ MZnjl ∀n; j; l ð25Þ n;t
t X X
X ZTnicgt ≥ β RTncrgt ð34Þ
XVkjgt −1 V g ≤ νn XTnkjgt ≤ XVkjgt n;i;c;g;t n;c;r;g;t
n X
ZTnicgt ≤ ωc Rc ∀c ð35Þ
V g ∀k; j; g; t ð26Þ n;i;g;t
X X
YVjigt −1 V g ≤ νn YTnjigt ≤ YVjigt ZTnicgt ≥ λ S nit ∀n; i; t ð36Þ
n g;c
V g ∀ j; i; g; t ð27Þ X
1≤ Rc ≤ NC ð37Þ
X c
ZVicgt −1 V g ≤ νn ZTnicgt ≤ ZVicgt
n X
1≤ xr ≤ N R ð38Þ
V g ∀i; c; g; t ð28Þ r
YRnrt ; XRnjlt ; XOnjlt ; bf nit ; bsnit ; XTnkjgt ; YTnjigt ; ZTnicgt ; RTncrgt ; PTnrit ; IPnjt ; ID1nit ; ID2nit ; IRnrt ≥ 0 ð39Þ
FWjt ; HWjt ; AWjt ; XVkjgt ; YVjigt ; ZVicgt ; RVcrgt ; PVrigt integer Y njlt ; Z njl ; Rc ; X r ∈f0; 1g ð40Þ
The objective function (1) seeks to minimize the total collection centers and handling cost in collection
cost of the supply chain. The first term of this objective centers. The sixth term is opening cost and recycling
function includes of procurement cost and production cost in recycling centers. The seventh, eighth, ninth
cost which is related to the regular time and overtime terms respectively are shortage cost and penalty costs
production. The second term is total human resources of uncollected returned products from demand points
cost such as salary, hiring and firing cost. The third minus revenues.
term is total inventory cost in plants, customer zones Constraint (2) is the inventory balance equation for
and recycling centers. Transportation cost between the first-class products at the end of each period in
entities of considered supply chain is shown in the factory j. Constraints (3) and (4) respectively are the
fourth term. The fifth term is opening cost of inventory balance equations for the first-class and the
1514 Int J Adv Manuf Technol (2017) 90:1507–1528
second-class products at the end of each period in de- considered supply chain. The quantity of the second-
mand point i. Constraint (5) is also the balance equation class products produced in recycle centers is limited to
at the end of each period for the second-class product the quantity of recyclable products transferred from the
inventory in recycling center r. The inventories at plant collection centers by constraint (31). Constraint (32)
j, custom zone i and recycling center r are limited by limits the amount of transferred the second-class prod-
the upper bounds of inventory capacity Constraints ucts to the amount of products produced in recycling
(6)–(8). The shortage of the first-class and the second- centers in the current time period. Constraint (33) en-
class products are limited to maximum specific levels in sures that recycling is possible in each recycling center,
customer zone i by constraints (9) and (10). Constraint on condition that the related center is opened.
(11) is related to the available workforce in each period Constraint (34) indicates that only recyclable percentage
of time. The upper bounds of regular and over time at of products should be transferred from considered col-
each factory are shown by constraints (12) and (13). lection centers to be recycled. Constraint (35) is a ca-
Constraint (14) ensures that the amount of pacity restriction for each collection center. The mini-
transportations from each supplier should not exceed mum service level needed for each customer zone is
its capacity. The quantity of waste produced at each indicated by constraint (36). The number of potential
plant is limited to an allowed level by constraint (15). collection and recycling centers that could be opened
Constraints (16)–(18) imply that the GHG emissions is respectively restricted to the maximum levels, NC
arising from the production methods and transportation and NR, by constraints (37) and (38). Finally, constraint
are restricted to a specific predetermined level. (39) determines the type of decision variables.
Constraints (19) and (20) restrict the number of the
first-class products manufactured in regular and over
time to an allowed level. The number of the second- 3.3 The proposed robust optimization model
class products produced at each recycling center is lim-
ited to a maximum allowed level by constraint (21). In this paper, a stochastic robust optimization approach
The quantity of the first-class products produced in fac- lied on Mulvey’s model is proposed in which uncertainty
tories is restricted to the quantity of raw materials re- is shown by a set of scenarios (ξ). In the considered aggre-
ceived in the current period of time by constraint (22). gate production system, inventory cost, production cost,
Constraint (23) implies that the amount of the first-class shortage cost, collection and recycling cost as well as de-
transferred products need to be limited to the amount of mand are scenario based uncertain (IC ξ ,PC ξ ,SC ξ ,CC ξ
products produced in regular or over time in the current ,RCξ,Dξ). It is assumed that there is no fluctuation in em-
period of time. Constraint (24) guarantees that if at least ployee wages and there is no significant change in the price
one product is produced in the related factory, binary of fuel during the period under study. So, labor cost and
variable Y njlt equals 1, and otherwise it equals 0. transportation cost are not scenario based uncertain.
Constraint (25) guarantees that if each product is not
assigned to factory j, then it does not need to be pro- LCðlaborcostÞ
duced at factory j in any time period. Constraints X X X
(26)–(30) determine the number of each vehicle type : SC j Ljt þ FC j F jt þ HC j H jt ð41Þ
required for transportation between different entities of j;t j;t j;t
X X
TCðtransportationcostÞ : TFg XVkjgt þ YVjigt þ ZVicgt þ RVcrgt þ PVrigt þ TVg d1kj XVkjgt
k; j;i;c;r;g;t k; j;g;t
X X X
þ TVg d2ji YVjigt þ TVg d1ic XVicgt þ TVg d4cr RVcrgt
i; j;g;t i;c;g;t c;r;g;t
X
þ TVg d5ri XVrigt ð42Þ
r;i;g;t
Int J Adv Manuf Technol (2017) 90:1507–1528 1515
X X CCξ ðcollectioncostÞ
ICξ ðinventorycostÞ : I1njt CI1nj þ I2nit ξ
n; j;t n;i;t X X ξ
X : ηc ξ R c þ ZQnicgt CCc
þ I3nit ξ CI2ni ξ þ I4nrt c i:c;g;t
n; j0 ;t X X
þ C i S nit − ZQnicgt ð46Þ
CI3nr ð43Þ i;c;n;t;g g;c
X X X
PCnjl ξ Z njl þ CRnjl ξ XRnjlt þ COnjl ξ XOnjlt
PCξ ðproductioncostÞ :
n; j;l n; j;l;t n; j;l;t
X X X
þ CMnk ξ XQnkjgt − P1ni ξ YQnjigt − P2ni ξ PQnrigt
n;k; j;g;t n;i; j;g;t n;i;r;g;t RCξ ðrecyclingcostÞ
ð44Þ X X
: X r ECr ξ þ RCnr ξ YRnrt ð47Þ
SCξ ðshortagecostÞ r n;r;t
X
: π1ni ξ b1nit ξ þ π2ni ξ b2nit ξ ð45Þ The developed robust optimization model for mentioned
i;n;t problem can be formulated as follows:
X
MinðZ 1 Þ ¼ ρξ LC þ ICξ þ TC þ PCξ þ SCξ þ CCξ þ RCξ
ξ
X X
ξ ξ ξ ξ ξ ξ0 ξ0 ξ0 ξ0 ξ0
þλ ρξ LC þ IC þ TC þ PC þ SC þ CC þ RC − ρξ0 LC þ IC þ TC þ PC þ SC þ CC þ RC
ξ
ξ0
ð48Þ
subject to The first and the second terms of the objective function
X (48) are respectively the mean value and variance of total
YQnjigðt−LT jig Þ þ I2niðt−1Þ ξ −I2nit ξ þ B1nit ξ −B1niðt−1Þ ξ costs. They are considered to measure the solution
j;g
robustness. ρ ξ is the probability of scenario ξ. In the
¼ D1nit ξ ∀n; i; t; ξ ð49Þ proposed model, inventory and shortage in demand
points are taken into account as control variables. I2nitξ
X
PQnrigðt−LT rig Þ þ I3niðt−1Þ ξ −I3nit ξ þ B2nit ξ −B2niðt−1Þ ξ and I3 n i t ξ take specific value for each scenario.
r;g Constraints (49) and (50), as control constraints, guaran-
tees that the demand in the customer zones will be fulfilled.
¼ D2nit ξ ∀n; i; t; ξ ð50Þ
If ∑ YQnjigðt−LT jig Þ þ I2niðt−1Þ ξ −B1niðt−1Þ ξ ¼ D1nit ξ , the in-
X j;g
I2nit ξ þ I3nit ξ ≤ Cap2i ∀i; t; ξ ð51Þ ventory and shortage at period t will be equal to zero. If
n
∑ YQnjigðt−LT jig Þ þ I2niðt−1Þ ξ −B1niðt−1Þ ξ ≥ D1nit ξ , product in-
j;g
B1nit ξ ≤ wn D1nit ξ ∀n; i; t; ξ ð52Þ ventory will be stored in customer zones at period of time
t for each scenario. If
B2nit ξ ≤ wn D2nit ξ ∀n; i; t; ξ ð53Þ ξ ξ ξ
∑ YQnjigðt−LT jig Þ þ I2niðt−1Þ −B1niðt−1Þ ≤ D1nit , we will face
j;g
I2nit ξ ; I3nit ξ ; b1nit ξ ; b2nit ξ ≥ 0 ∀ξ ð54Þ a shortage in customer zones at time period t for each
scenario. The objective function (48) is a non-linear
YRnrt ; XRnjlt ; XOnjlt ; XQnkjgt ; YQnjigt ; ZQnicgt ; RQncrgt ; PQnrit ; I1njt ; I4nrt ≥ 0 ð55Þ function. Non-negative variable (θξ) is used to transform
the problem into a linear programming model with lin-
ear objective function as well as linear constraints as
with Constraints (2), (5), (6), (8), (11)–(38), and (40). follows:
1516 Int J Adv Manuf Technol (2017) 90:1507–1528
X X h
MinðZ 1 Þ ¼ ρξ LC þ ICξ þ TC þ PCξ þ SCξ þ CCξ þ RCξ þ λ1 ρξ LC þ ICξ þ TC þ PCξ þ SCξ
ξ ξ
X 0 0 0 0 0
þ CCξ þ RCξ − ρξ0 LC þ ICξ þ TC þ PCξ þ SCξ þ CCξ þ RCξ
ξ0
i X
ξ
þ 2θ LC þ ICξ þ TC þ PCξ þ SCξ þ CCξ þ RCξ − ρξ LC þ ICξ þ TC þ PCξ þ SCξ þ CCξ
ξ
þ RCξ þ θξ ≥0 ∀ξ ð56Þ
X θξ ≥ 0 ∀ξ ð58Þ
LC þ ICξ þ TC þ PCξ þ SCξ þ CCξ þ RCξ − ρξ
ξ
4 Computational results
LC þ ICξ þ TC þ PCξ þ SCξ þ CCξ þ RCξ
In this section, some computational experiments are presented to
þ θξ ≥ 0 ∀ξ ð57Þ evaluate the effectiveness and robustness of the model. A data set
Int J Adv Manuf Technol (2017) 90:1507–1528 1517
is provided from an Iranian Wood and Paper Industries Company provided by Iranian Wood and Paper Industries Company are
as a real-world industrial case to illustrate the applicability of used. The pulp and paper industry converts fibrous raw mate-
proposed robust optimization model to practical problems. rial into pulp, paper. The processes involved in papermaking
include raw material preparation, pulping (chemical, mechan-
4.1 Case description ical), bleaching, chemical recovery, pulp drying, and paper-
making. A flow diagram of the processes is shown in Fig. 2.
In order to show the efficiency of the proposed robust pro- The most significant energy-consuming processes are pulping
gramming model for the considered supply chain, a set of data and the drying section of papermaking.
1518 Int J Adv Manuf Technol (2017) 90:1507–1528
Table 1 Demand forecast for the first-class products under scenario 1 Table 3 Purchasing cost under different scenarios (in dollar per unit)
(units)
CM
Customer 1 Customer 2
Scenario 1 Scenario 2 Scenario 3
Period Period
Supplier Supplier Supplier
1 2 3 4 1 2 3 4
1 2 3 1 2 3 1 2 3
Product 1 380 340 320 400 360 570 380 320
2 450 390 460 360 470 480 480 410 Product 1 15 16 10 12 10 14.2 14 12.8 16.4
3 320 340 450 430 450 420 460 460 2 12.5 12 11 13 14.2 12 12 16.4 14
4 350 430 450 440 350 530 370 390 3 10 15 12.5 16 13.1 10.9 11.6 15.2 12.8
5 300 540 320 530 450 270 530 430 4 15.8 18 13 12 16.8 15.3 10.4 11.6 17.6
5 13 17.5 12.5 10 12 14.2 12.8 14 16.4
For scenarios 2 and 3, the estimations are multiplied by 1.05 and 1.1,
respectively
1 2 1 2 1 2 1 2 1 2 1 2
1 2 1 2 1 2 1 2 1 2 1 2
Product 1 37.5 28.4 24.6 31.2 37.3 28.2 25.5 32 37.2 28 25.4 32.1
2 31.2 25.6 39.2 30.6 31 25.5 39 29.5 29.8 25.8 39.1 29.4
3 38.4 31 27.7 28.2 38.5 29.1 28.4 27.3 38.4 28.7 28.5 27
4 35.4 32 24.6 25.8 35.2 32 25.5 27 35.1 32.1 25.4 26.5
5 33.5 28.4 38.4 32.2 33.1 27.5 38.5 32 32.7 28.3 38.4 32.1
1 2 1 2 1 2 1 2 1 2 1 2
Product 1 210 420 410 520 215 445 400 505 220 440 420 500
2 190 480 520 460 195 475 490 485 180 460 510 470
3 290 390 330 350 275 385 325 355 270 375 320 340
4 340 510 440 340 315 500 410 340 325 515 420 355
5 270 345 350 290 275 315 340 300 270 320 355 295
1520 Int J Adv Manuf Technol (2017) 90:1507–1528
17.4
14.4
16.2
16.2
Table 7 Construction and handling cost of collection centers under
15
2
different scenarios
Scenario 3
η ($) C ($/unit)
14.4
16.8
16.8
16.8
14.4
1
Scenario Scenario
15.95
14.85
13.75
14.85
13.2
1 2 3 1 2 3
2
Scenario 2
Collection center 1 39,000 40,000 41,500 17 15 14
13.2
17.5
15.4
15.4
13.2
2 50,000 45,500 44,000 13 14.5 16
13.5
14.5
12.5
13.5
12
2
Scenario1
Customer
Table 8 Construction cost of each potential recycling center under
different scenarios
15.5
CID
14
12
14
12
1
EC ($)
Scenario
33
32
19
15
15
Price and holding cost of the first and second-class products in customer zones under different scenarios (in dollar per unit)
2
Scenario 3
1 2 3
32
35
13
12
14
1
2 62,000 63,200 64,400
3 52,500 53,750 55,000
14.5
14.5
34
35
18
2
Scenario 2
been reported about 88,554,272 tons. About 10 % of
31
34
18
13
16
these emissions are related to the Wood and Paper
1
Industries. According to the plan announced by the gov-
ernment, an emissions reduction target of 3 % is taken
35
33
17
15
14
2
into account for this industry. Based on the cap and
Scenario1
Customer
33
35
17
15
11
1
RC
44
48
23
33
35
1
Customer
1 2 3 1 2 3 3 3
P1
43
47
32
31
29
1
4 10 9 8 11 9 9.5 8.5 10
Product
5 8.2 8 10 8 8 11 8 11.5
Int J Adv Manuf Technol (2017) 90:1507–1528 1521
Table 11 Inventory cost in plants (in dollar per unit) shown in Tables 7, 8, and 9. Table 10 presents the price
Product and holding cost of the first and second-class products
1 2 3 4 5 in customer zones under different scenarios. Inventory
cost in plants and recycling centers, as well as shortage
Plant 1 12.0 12.5 13.5 13.0 14.5 cost are given in Tables 11, 12, and 13. The quantity of
2 11.5 12.5 12.0 14.0 12.0 each product provided by suppliers in each period is
presented in Table 14. Inventory capacity of each
recycling center and maximum capacity of collection
centers are presented in Tables 15 and 16. Table 17
Table 12 Inventory cost shows the salary, firing, and hiring cost and inventory
in potential recycling CIR
capacity in each plant. Penalty cost and inventory ca-
centers (in dollar per
unit) Recycling center pacity in demand point are presented in Table 18.
Transportation cost of each vehicle type and distances
1 2 3 between entities are given in Tables 19, 20, and 21.
Product 1 5.0 4.5 3.5 Waste percentage of production is given in Table 22.
2 6.0 5.0 4.0
Table 23 shows the quantity of returned products in
3 6.5 4.5 4.0
each customer zone. At most two recycling centers can
4 4.0 5.0 3.5
be opened from all potential sites. The lead time re-
quired for transportation from each factory or recycling
5 3.0 2.5 4.0
center to demand points is one period of time.
Customer Customer
1 2 1 2 1 2 1 2 1 2 1 2
Product 1 150 100 165 110 180 120 130 150 143 165 156 180
2 120 110 132 121 144 132 140 110 154 121 168 132
3 130 120 143 132 156 144 130 150 143 165 156 180
4 140 100 154 110 168 120 120 140 132 154 144 168
5 120 150 132 165 144 180 110 120 121 132 132 144
1522 Int J Adv Manuf Technol (2017) 90:1507–1528
Customer Supplier
Table 17 Salary, firing and hiring cost and inventory capacity in each
plant 1 2 1 2 3
161
148
136
134
208
Table 22 Waste percentage of production
4
WP
394
250
357
372
416
Plant 1 Plant 2
3
Production method
455
248
383
390
383
2
1 2 1 2
Method 2
Product 1 0.03 0.04 0.01 0.011
436
374
221
411
375
2 0.04 0.05 0.04 0.03
1
3 0.03 0.03 0.03 0.02
4 0.012 0.011 0.02 0.03
154
200
149
140
192
5 0.01 0.02 0.022 0.042
367
302
370
378
428
customer zones as a cost saving covers about 30 % of
3
total cost. It should be noted that considerable part of
total cost is arising from construction of the centers.
428
270
416
396
337
Results indicated in Table 32 and Fig. 5 indicates per-
2
formance and efficiency of the collection and recycling
Method 1
system.
Plant 2
Period
407
396
254
406
329
1
4.2.1 Comparing the effectiveness of robust optimization Regular time assigned in each plant to produce differing the first-class products by different methods
and mean models
130
160
106
106
166
4
In order to demonstrate the robustness of production
and recycling plan obtained from the proposed model,
242
345
406
304
341
3
the expected values of uncertain parameters in the MIP
model are considered as deterministic values. Hereafter,
this model is called the mean value-based model.
253
384
406
376
366
2
Customer 2 Customer 1
2
Method 1
Plant 1
Period Period
Period
392
282
380
460
296
1
1 2 3 4 1 2 3 4
Products
Table 25 Overtime assigned in each plant to manufacture differing the first-class products by different methods
Plant 1 Plant 2
Period Period
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
Products 1 100 100 90 90 130 110 110 110 128 128 128 108 101 101 101 101
2 164 164 114 114 128 118 118 118 76 126 76 76 148 148 128 128
3 192 99 72 72 103 89 86 86 103 103 103 103 136 136 116 116
4 72 72 72 72 126 126 86 86 108 108 108 108 114 114 114 114
5 104 104 104 104 132 132 132 102 102 132 102 102 86 86 114 86
5135.96
649,276.6
Table 29 Interactions between customers and collection centers
30 %
652,690
651,770
643,370
Profit
Period t
28,140.93
center c
2,151,764.8
2,168,885.6
100 %
2,153,528
2,201,364
Total cost
1 1 1 310 (2) 208 (1) 216 (2) 241 (2)
1 2 2 208 (2) 240 (2) 224 (2) 192 (2)
2 1 1 224 (2) 216 (2) 232 (2) 208 (3)
2 2 2 240 (2) 224 (3) 224 (2) 216 (2)
22.96 %
5 1 1 271 (2) 300 (2) 316 (1) 316 (3)
2673.16
497,941.6
5 2 2 240 (3) 216 (2) 200 (2) 200 (1)
496,095
501,007
496,723
a
It means that 284 units of type 3 product are shipped by vehicle type 3,
from customer 1 to collection center 1 during period 1
Workforce cost
16.11 %
plan obtained from the proposed model are robust
16,345.89
349,093.3
363,200
352,900
331,180
against the changes in uncertain parameters in the fu-
ture. The robust optimization model is yielded a set of
solutions with less sensitivity against actual realizations
of uncertain input data. It would be better to say the
32.15 %
that of the mean value based model, under different
697,427.26
2405.59
scenarios. It is worthwhile to note that the proposed 696,040.8
696,036
11.57 %
sufficient, but low fluctuation in realistic scenarios in
4907.98
251,120
245,810
255,490
252,060
n center r i
0.79 %
12,213.94
1 2 3 4
17,090
4081
18,877
28,312
Scenario
a
It means that 215 units of type 5 product are shipped by vehicle type 3,
from recycling center 2 to customer 1 during period 3
1
2
3
1526 Int J Adv Manuf Technol (2017) 90:1507–1528
Table 32 Total cost, profit, and cost saving of the collection and 5 Conclusion
recycling system
Scenario Total cost The profit Cost saving In this paper, a robust optimization model was proposed
of collection of collection for a multi-product multi-period multi-site aggregate
and recycling and recycling production planning problem in a green supply chain
center system
considering a reverse logistic system under uncertainty.
1 496,095 170,181 151,230 The objective function was taken into account to mini-
2 496,723 178,310 151,440 mize total losses of the supply chain including purchas-
3 501,007 181,400 153,100 ing cost, production cost, workforce cost, inventory
cost, shortage cost, transportation cost, collection cost
and recycling cost.
Table 33 Total cost of robust and mean models for the scenarios with The limited number of collection and recycling cen-
probabilities 0.35, 0.5, and 0.25
ters among potential centers taken into account to be
Problem number Robust model Main value based model constructed to produce the second-class products. On
the other hand, in addition to waste management, dif-
1 1,976,100 1,981,900 ferent production methods was taken into consideration
2 1,970,300 1,968,000 and the GHG emission related to each method was lim-
3 1,972,900 1,982,100 ited to a maximum specific level. Furthermore, GHG
4 1,973,700 1,983,700 emissions associated with transportation modes were
5 1,975,500 1,991,100 embedded in the model.
6 1,971,500 1,971,800 In the proposed model, demand fluctuation and some
7 1,972,400 1,977,500 of the cost parameters are subject to the uncertainty. In
8 1,968,400 1,968,700 order to reduce the impacts of changes and fluctuations of
9 1,969,200 1,971,100 uncertain parameters and provide a stable plan, the robust
10 1,970,200 1,983,100 optimization is used and a scenario set is employed to
11 1,971,200 1,989,200 illustrate the uncertainties. The computational results re-
12 1,968,800 1,972,000 lated to a data set from an Iranian Wood and Paper
13 1,966,800 1,967,900 Industries Company demonstrated that the proposed mod-
14 1,972,200 1,982,100 el is more practical to handle the uncertain parameters in
15 1,971,300 1,968,700 the real-world production environment. The effectiveness
and robustness of the model is also indicated. In order to
provide helpful managerial insights, cost analysis is per-
the future is more important. All things considered, nu- formed. Furthermore, total costs and the profit in collec-
merical results indicate the effectiveness of the proposed tion and recycling centers, is analyzed to show the perfor-
robust optimization model. mance of this part of considered supply chain.
700000
objective function under different
Purchasing Cost
400000
699000
scenarios 698000
300000
697000
Cost
200000 696000
100000 695000
694000
0 693000
1 2 3 1 2 3
Scenario Scenario
Collecon and Recycling Cost
502000 654000
501000 652000
500000
650000
499000
648000
Profit
498000
646000
497000
496000 644000
495000 642000
494000 640000
493000 638000
1 2 3 1 2 3
Scenario Scenario
Int J Adv Manuf Technol (2017) 90:1507–1528 1527
Fig. 5 Total cost, profit and cost Total cost, profit and cost saving of collecon Total cost, profit and cost saving of collecon
saving of collection and recycling and recycling system under Scenario 1 and recycling system under Scenario 2
system under different scenarios 600000 600000
500000 500000
400000 400000
300000 300000
151230 496095 151440 496723
200000 200000
100000 170181 100000 178310
0 0
System profit and cost Total cost of collecon and System profit and cost Total cost of collecon and
saving recycling centers saving recycling centers
1985000
9. Yeh W-C, Chuang M-C (2011) Using multi-objective genetic algo-
1980000 rithm for partner selection in green supply chain problems. Expert
Syst Appl 38(4):4244–4253
1975000
10. Ghayebloo S et al (2015) Developing a bi-objective model of the
1970000 closed-loop supply chain network with green supplier selection and
disassembly of products: the impact of parts reliability and product
1965000 greenness on the recovery network. J Manuf Syst 36:76–86
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
11. Fahimnia B et al (2015) Policy insights from a green supply chain
Problem No. optimisation model. Int J Prod Res 53(21):6522–6533
Robust model Mean value based model
12. Bhattacharya A, Kumar Dey P, Ho W (2015) Green manufacturing
Fig. 6 Comparison of total costs between mean and robust models with supply chain design and operations decision support. Int J Prod Res
probabilities 0.35, 0.5, and 0.25 53(21):6339–6343
1528 Int J Adv Manuf Technol (2017) 90:1507–1528
13. Vachon S (2007) Green supply chain practices and the selection of 35. Kazemi A, Fazel Zarandi MH, Moattar Husseini SM (2008) A
environmental technologies. Int J Prod Res 45(18–19):4357–4379 multi-agent system to solve the production–distribution planning
14. Darnall N, Jolley GJ, Handfield R (2008) Environmental manage- problem for a supply chain: a genetic algorithm approach. Int J
ment systems and green supply chain management: complements Adv Manuf Technol 44(1–2):180–193
for sustainability? Bus Strateg Environ 17(1):30–45 36. Lim SJ et al (2005) A simulation approach for production-
15. Zhu Q, Sarkis J (2004) Relationships between operational practices distribution planning with consideration given to replenishment
and performance among early adopters of green supply chain man- policies. Int J Adv Manuf Technol 27(5–6):593–603
agement practices in Chinese manufacturing enterprises. J Oper 37. Nam S (1992) J. And R. Logendran, Aggregate production plan-
Manag 22(3):265–289 ning—a survey of mdels and methodologies. Eur J Oper Res 61:
16. Zhu Q et al (2008a) Firm-level correlates of emergent green supply 255–272
chain management practices in the Chinese context. Omega 36: 38. Leung SCH, Wu Y, Lai KK (2006) A stochastic programming
577–591 approach for multi-site aggregate production planning. J Oper Res
17. Zhu Q, Sarkis J, Lai K-h (2008b) Green supply chain management Soc 57(2):123–132
implications for “closing the loop”. Transp Res E 44:1–8 39. Fung YK, Tang RJ, Wang D (2003) Multiproduct aggregate production
18. Zhu Q, Sarkis J, Lai K-h (2008c) Confirmation of a measurement planning with fuzzy demands and fuzzy capacities. IEEE Transactions
model for green supply chain management practices implementa- on Systems, Man, and Cybernetics, Part A 33(3):302–313
tion. Int J Prod Econ 111:261–273 40. Mulvey M, Vanderbei JR, Zenios SA (1995) Robust optimization
19. Srivastava SK (2007) Green supply-chain management: a state-of- of large-scale systems. Oper Res 43:264–281
the-art literature review. Int J Manag Rev 9(1):53–80 41. Leung SCH, Wu Y (2004) A robust optimization model for sto-
20. Min H, Kim I (2012) Green supply chain research: past, present, chastic aggregate production planning. Prod Plan Control 15(5):
and future. Logist Res 4(1–2):39–47 502–514
21. Dekker R, Bloemhof J, Mallidis I (2012) Operations research for 42. Leung SCH et al (2007) A robust optimization model for multi-site
green logistics—an overview of aspects, issues, contributions and production planning problem in an uncertain environment. Eur J
challenges. Eur J Oper Res 219(3):671–679 Oper Res 181(1):224–238
22. Blumberge DF (1999) Strategic examination of reverse logistics
43. Kazemi Zanjani M, Ait-Kadi D, Nourelfath M (2010) Robust pro-
and repair service requirements, needs, market size, and opportuni-
duction planning in a manufacturing environment with random
ties. J Bus Logist 20(2):141–159
yield: a case in sawmill production planning. Eur J Oper Res
23. Álvarez-Gil MJ et al (2007) Reverse logistics, stakeholders’ influ-
201(3):882–891
ence, organizational slack, and managers’ posture. J Bus Res 60(5):
44. Pan F, Nagi R (2010) Robust supply chain design under uncertain
463–473
demand in agile manufacturing. Comput Oper Res 37(4):668–683
24. Glenn R et al (2005) Monitoring reverse logistics programs: a
45. Niknamfar AH, Niaki STA, Pasandideh SHR (2014) Robust opti-
roadmap to sustainable development in emerging markets.
mization approach for an aggregate production–distribution plan-
Multinatl Bus Rev 13(3):41–65
ning in a three-level supply chain. Int J Adv Manuf Technol 76(1–
25. Galvez D et al. (2015) Reverse logistics network design for a biogas
4):623–634
plant: an approach based on MILP optimization and Analytical
Hierarchical Process (AHP). Journal of Manufacturing Systems 46. Mirzapour Al-e-hashem SMJ, Malekly H, Aryanezhad MB (2011b)
26. Alshamsi A Diabat A (2015) A reverse logistics network design. A multi-objective robust optimization model for multi-product
Journal of Manufacturing Systems multi-site aggregate production planning in a supply chain under
27. Mishra N, Kumar V, Chan FTS (2012) A multi-agent architecture uncertainty. Int J Prod Econ 134(1):28–42
for reverse logistics in a green supply chain. Int J Prod Res 50(9): 47. Al-e-Hashem SMJM, Aryanezhad MB, Sadjadi SJ (2011) An effi-
2396–2406 cient algorithm to solve a multi-objective robust aggregate produc-
28. Wang R-C, Fang H-H (2001) Aggregate production planning with tion planning in an uncertain environment. Int J Adv Manuf
multiple objectives in fuzzy environment. Eur J Oper Res 133:521–536 Technol 58(5–8):782–765
29. Gomes da Silva C et al (2006) An interactive decision support system 48. Mirzapour Al-e-hashem SMJ, Baboli A, Sazvar Z (2013) A sto-
for an aggregate production planning model based on multiple criteria chastic aggregate production planning model in a green supply
mixed integer linear programming. Omega 34(2):167–177 chain: considering flexible lead times, nonlinear purchase and
30. Orcun S, Uzsoy R, Kempf KG (2009) An integrated production shortage cost functions. Eur J Oper Res 230(1):26–41
planning model with load-dependent lead-times and safety stocks. 49. Yu C-S, Li H-L (2000) A robust optimization model for stochastic
Comput Chem Eng 33(12):2159–2163 logistic problems. Int J Prod Econ 64:385–397
31. Zhang R et al (2012) The activity-based aggregate production plan- 50. Martin N et al. (2000) Opportunities to improve energy efficiency
ning with capacity expansion in manufacturing systems. Comput and reduce greenhouse gas emissions in the U.S. pulp and paper
Ind Eng 62:491–503 industry
32. Xue G et al (2011) Integrated production planning with sequence- 51. Network for Transport and Environment, (NTM). http://
dependent family setup times. Int J Prod Econ 131(2):674–681 www.ntmcalc.se/index.html.
33. Masud ASM, Hwang CL (2007) An aggregate production planning 52. Benn H Milliband E (2009) Guidance on how to measure and
model and application of three multiple objective decision methods. report your greenhouse gas emissions
Int J Prod Res 18(6):741–752 53. Rahmani D et al (2013) A robust optimization model for multi-
34. David AG, Goal Programming A (1974) Approach to aggregate product two-stage capacitated production planning under uncertain-
planning of production and work force. Manag Sci:1569–1575 ty. Appl Math Model 37(20–21):8957–8971
International Journal of Advanced Manufacturing Technology is a copyright of Springer,
2017. All Rights Reserved.