Biodiesel supply chain

Energy Conversion and Management 225 (2020) 113345
Contents lists available at ScienceDirect
Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman
Management of animal fat-based biodiesel supply chain under the

paradigm of sustainability
Muhammad Salman Habib a, Muhammad Tayyab b, Sadaf Zahoor a, c, Biswajit Sarkar d, *
a
Department of Industrial and Manufacturing Engineering, University of Engineering and Technology Lahore, Pakistan
b
Suleman Dawood School of Business, Lahore University of Management Sciences, Lahore 54792, Pakistan
c
Mechanical, Automotive and Materials Engineering, Faculty of Engineering, University of Windsor, Ontario N9B 3P4, Canada
d
Department of Industrial Engineering, Yonsei University, 50 Yonsei-ro, Sinchon-dong, Seodaemun-gu, Seoul 03722, South Korea
A R T I C L E I N F O A B S T R A C T
Keywords: The ever-increasing energy demand and harmful environmental and social impacts instigated by fossil fuels have
Waste animal fat led to the development of sustainable energy resources. Biodiesel has been introduced as an alternative owing to
Non-edible feedstock its wide commercial applications. It can be produced from various types of biomass including vegetable oil, used
Sustainable biodiesel supply chain
cooking oil, and animal fat. Of these, animal fat is a promising option because of its low cost and easy avail
Robust possibilistic programming
Social responsibility
ability. For a speedy transition toward large-scale commercially viable biodiesel production, the efficient design
of a biodiesel supply chain is very important. Moreover, because biodiesel supply chain (SC) design decisions
directly influence social and environmental aspects, hence SC optimization within the sustainability paradigm is
necessary. This research proposes a waste-animal-fat-based biodiesel SC optimization model that minimizes
environmental impact and SC cost while maximizing social wellbeing. The complex and dynamic environment of
biodiesel production brings a high level of uncertainty, which compromises the effectiveness of SC decisions;
thus, a modified robust possibilistic chance-constrained programming solution methodology is developed. To
validate the proposed model and solution methodology, a computational analysis of a case study is performed.
The results demonstrate that by paying 1.13% higher costs, a desirable level of social and environmental pro
tection can be achieved for the proposed supply chain. It is also found that with an increase of 5% in the eco
nomic objective, 6% increase in the environmental objective, and 7% decrease in the social objective, the risk of
epistemic uncertainty can be completely ruled out. Further, from the numerical analysis it is inferred that among
all tiers of biodiesel SC, the decisions of the biorefinery tier play a crucial role in achieving sustainable devel
opment goals. Furthermore, a higher priority of economic and environmental objectives makes the SC decisions
more centralized, whereas a higher priority of the social objective makes them more decentralized. The proposed
model can be useful to investors and policymakers involved in biodiesel production and distribution.
1. Introduction with 1990 [7]. Among the various fossil fuel substitutes, biofuel from
biomass has gained worldwide attention in the last two decades.
Recently, problems such as global warming, energy crisis, and scar Biomass from energy crops and agricultural and forest residues has
city of food as well as the quest for a higher living standard have significant potential as a replacement for conventional fuels [8]. Bio
compelled the researchers to search for alternative sustainable energy diesel can be obtained from biomass, which in turn can be used as fuel.
resources [1,2]. With the technological advancement and industrial Biodiesel is combined with fossil diesel for commercial use [9]. Biodiesel
revolution, dependency on fossil fuels has increased exponentially [3,4]. is a renewable and biodegradable resource with lower sulfur content,
However, they are not sustainable, and the remaining reserves are very higher cetane number, greater efficiency, and better lubricant properties
limited. Moreover, the burning of fossil fuels significantly influences the [6], thus making it superior to its counterpart fossil diesel.
environment by increasing greenhouse gas (GHG) emissions and the The sustainability of biodiesel produced from the first-generation
Earth’s ambient temperature[5,6]. To minimize climate impacts, GHG source is still questionable because it competes with the food market
emissions are targeted to be reduced by 50% until 2050 as compared and is costly compared to fossil diesel [10]. Almost 80% of the
* Corresponding author.
E-mail address: [email protected] (B. Sarkar).
https://doi.org/10.1016/j.enconman.2020.113345
Received 26 April 2020; Received in revised form 12 August 2020; Accepted 13 August 2020
Available online 17 October 2020
0196-8904/© 2020 Elsevier Ltd. All rights reserved.
M.S. Habib et al. Energy Conversion and Management 225 (2020) 113345
production cost of biodiesel is currently associated with feedstocks; consideration of uncertainty in the input parameters of these
therefore, it is important to produce it from cheaper sources [11,12]. assessment techniques is another aspect that makes this study novel.
Recently animal fat has drawn the attention of researchers (see:
[6,13–15]) as a potential economically sustainable feedstock. Mutton, The remainder of this paper is structured as follows. Discussion
beef, and chicken fats are rich sources of animal fat. Animal fat as a regarding the existing literature on biofuel SC optimization and the
feedstock for biodiesel is attractive owing to its lower prices [16]. Ani significance of the study are presented in Section 2. A detailed
mal fat has a limited market demand currently and is mainly used as raw description of the WAF-SCND model is provided in Section 3. Section 4
materials in the cosmetics and soap industries [12]. Waste animal fat explains the solution methodology employed. Section 5 introduces the
(WAF) as a substitute feedstock will not only solve its disposal issue but computational analysis of the case study and presents the findings.
will also offer a very low-cost source for the biodiesel industry [16]. Finally, Section 6 suggests potential future research prospects and con
Further, the transesterification cost of WAF ($ 0.4–0.5/liter) is a cheaper cludes this research.
option in comparison with vegetable oil ($ 0.6–0.8/liter) [6]. In addition
to using cheap feedstock and processing technology, the efficient design 2. Literature review
of the biodiesel SC network is very important in making the large-scale
biodiesel production commercially feasible. In this background, the Herein, the latest studies in the field of biofuel production are
rapid expansion of the biofuel industry has encouraged the policymakers reviewed to highlight the significance and contribution of this research.
to use SC network design optimization models (see, [8–10,17]) to attain The use of biodiesel as an alternative fuel has attracted increasing
the lowest per-unit production cost. interest owing to its advantages over petroleum diesel, in terms of en
In the past, the dreadful effects of commercialization on the society gine performance and environmental aspect. Such actions can also help
and environment have pushed all industries toward sustainable devel in reducing dependency on foreign oil and greenhouse effects [20]. In
opment while considering economic, environmental, and social aspects, this perspective, many authors have discussed biofuel properties ob
simultaneously [17]. Sustainable development is an integral part at all tained from various types of animal fats. According to Kirubakaran et al.
levels of the supply chain (SC), including strategic development, tactical [12], the properties of chicken fat-based biodiesel are as per ASTM
decision-making, and operation-level activities [17]. Since the inte D6751 standard. Hoekman et al. [21] compared the composition and
grated SC network links all the activities from feedstock suppliers to the properties of biodiesel extracted from WAF and vegetable oil. Sander
final distribution of biodiesel to consumers, decisions made along the et al. [22] examined the impact of different types of animal fats and
way have remarkable environmental and social effects on the society processing conditions on the quality of biodiesel. Rajak et al. [23]
[17]. The key to a sustainable biodiesel SC framework is the tradeoff investigated the emission characteristics of biodiesel derived from waste
among these factors to approach a triple bottom line (TBL) [18,19]. The oil, WAF, alcohol, vegetable oil, and compared them with the identical
TBL emphasizes on a balanced approach to profit, the planet, and peo environment of fossil diesel emissions. Ren et al. [24] proposed a
ple. In this perspective, development of an inclusive decision making mathematical model with the objective to maximize the sustainability
tool is the need of the hour that can facilitate the policymaker in index in a biodiesel supply network design by using soybean, sunflower,
achieving both commercialization and sustainable development goals in and rapeseed oil as feedstock. Gebremariam et al. [25] provided
a WAF-based biodiesel production system. In this scenario, there is no different process alternatives to improve the techno-economic feasibility
research that has addressed these uncertainties in a waste animal fat- of biodiesel production using sulfuric acid as a catalyst during the
based biodiesel SC while considering the triple lines of sustainability. transesterification process. Some researchers, e.g., Zuo et al. [26] and
Considering the aforementioned gaps of the biodiesel SC in the existing Zhang et al. [27], discussed the effect of the composition of biodiesel on
literature, this study proposes a WAF-based biodiesel SC network design the emission characteristics of diesel engines. The experimental results
(WAF-SCND) model for its production and distribution under an un indicate that, in all engine operating conditions, the particle emissions
certain environment. The goal of this model is to minimize the SC from WAF-based biodiesel having a proportion greater than 40% in the
operation cost and environmental effect of processes in the animal fat- blend are less than those from other types of biodiesel having the same
based biodiesel SC while maximizing the social impact. proportion [28].
The major theoretical and practical contributions of this research are Most of the existing optimization models proposed in this area
as follows: address the sustainability framework by considering environmental and
economic dimensions, including the uncertainties related to feedstocks
Practical: This research contributes in two aspects from the practical and biofuel logistics, and different costs. In this domain, Ghaderi et al.
perspective; firstly, it provides a comprehensive mathematical model [10] presented an optimization model for a switchgrass-based biodiesel
for the WAF-SCND problem, while considering triple lines of sus SC under the uncertainty of feedstock supply, biodiesel demand, logis
tainability (economy, environment, and society) as model objectives tics costs, social, and environmental objective related parameters. To
by considering uncertainty in constraints and objective function, tackle the uncertainty, a novel RPP-based solution approach that mini
simultaneously. The developed model is expected to assist policy mizes risk but maximizes the SC performance and feasibility robustness
makers to find optimal locations, numbers, and capacity levels of were introduced. Babazadeh [11] developed an optimization model to
biodiesel processing facilities. Secondly, it proposes a highly effec design a biodiesel SC using Jatropha curcas L. and used cooking oil as
tive solution methodology named as modified robust possibilistic feedstocks with the goal of reducing SC cost and environmental impact
chance-constrained programming (MRPCCP) to solve the given under epistemic uncertainty. To minimize the uncertainty influence, the
problem. MRPCCP is very interactive and provides flexibility to the authors proposed a novel possibilistic programming-based solution
decision-makers to adjust uncertainty-modeling parameters methodology to minimize the total risk values and mean of the model
following the real-time situation. The results of the case addressed in objectives. Ahmed et al. [29] propounded an optimization model for
this research also endorse the robustness and efficiency of MRPCCP. biofuel SC using crop residues as biomass under the TBL framework in
Theoretical: This research utilizes the latest developed social and an uncertain environment and employed the augmented ε-constraint
environmental impact techniques to model the social and environ approach to solve it. Mousavi Ahranjani et al. [30] also provided an
mental burdens of SC configuration. In this study, SLCAP and EI99 optimization model for a lignocellulosic-based biofuel SC by incorpo
methodologies are used to estimate the social and environmental rating sustainability dimensions. They applied a robust possibilistic
impact, respectively. To the best of authors’ knowledge, this is the programming (RPP) technique to minimize the impact of uncertainties
first study to use these latest techniques to calculate social and associated with biofuel demand, feedstock supply, and GHG emissions.
environmental impact in an animal fat-based biodiesel SC. Further, Bairamzadeh et al. [31] recommended an optimization model to manage
2
Table 1
Summary of studies in the field of biofuel SC.
Author Type of Biofuel Edible/ Economic Environmental Social LA CF TT FM IP Uncertainty Solution
feedstock type Non- objective objective objective approach
edible
You et al. [32] Cellulosic Bioethanol Edible ✓ ✓ ✓ ✓ ✓ ✓

Gonela et al. Energy crops, Bioethanol Edible ✓ ✓ ✓ ✓ ✓ ✓ Stochastic
[34] lignocellulosic and approach
non-
edible
de Jong et al. Forest-based Bioethanol Non- ✓ Geographically-
[35] biomass edible explicit cost
optimization
Cambero et al. Forest-based Bioethanol Non- ✓ ✓ ✓ ✓ ✓ ✓ ✓ AUGMECON
[36] biomass edible method
Santibañez- Energy crops Bioethanol Edible ✓ ✓ ✓ ✓ ✓ ✓ ✓ Stochastic
Aguilar et al. and approach
[37] non-
edible
Sarkar et al. [38] Forest-based Bioethanol Non- ✓ ✓ ✓ ✓ ✓ classical
biomass edible optimization
technique
Bairamzadeh Corn stover and Bioethanol Non- ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ RPP
et al. [31] wheat straw edible
Ren et al. [24] Soybean, Biodiesel Edible ✓ ✓
sunflower, and
rapeseed
Azadeh et al. Soybean Biodiesel Edible ✓ ✓ ✓ ✓ ✓ ✓ Stochastic
[39] approach
Zhang et al. [40] Waste cooking Biodiesel Non- ✓ ✓ ✓ ✓
oil edible
Babazadeh et al. Jatropha oil Biodiesel Non- ✓ ✓ ✓ ✓ ✓ ✓ ✓ Possibilistic
[17] and used edible programming
cooking oil
Babazadeh [11] Jatropha oil Biodiesel Non- ✓ ✓ ✓ ✓ ✓ ✓
and used edible
cooking oil
Ghaderi et al. Switchgrass Bioethanol Non- ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ RPP
[10] edible
Mousavi Lignocellulosic Bioethanol Non- ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ RPP
Ahranjani et al. edible
[30]
Orjuela-Castro Palm oil Biodiesel Edible ✓ ✓ ✓ ✓ ✓
et al. [41]
Hosseinalizadeh Waste cooking Biodiesel Edible ✓ ✓ ✓ ✓
et al. [42] oil, soya, and
sunflower, and non-
rapeseed edible
Ahmed et al. [8] Agricultural Biofuel Non- ✓ ✓ ✓ ✓ ✓ ✓ Possibilistic
residue edible programming
Mirhashemi et al. Wastewater Biodiesel Non- ✓ ✓ ✓ ✓ ✓ Data-driven
[43] sludge edible robust
optimization
This study Waste animal Biodiesel Non- ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ MRPCCP
fat edible
CF: Capacity of facilities, IP: Inventory planning, FM: Flow of material, LA: Location allocation, TT: Type of technology.
the lignocellulosic-based biofuel SC while considering uncertainties in optimization model GHG emission saving, net present value, and social
fuel demand, fuel price, and environmental impact coefficients of SC benefit (in terms of number of jobs created) in a forest based biofuel SC.
activities. You et al. [32] provided a mixed-integer linear programming Santibañez-Aguilar et al. [37] provided a stochastic optimization model
model that considered the SC of bioethanol obtained from cellulose, for biomass conversion system considering environmental and economic
under economic, social, and environmental dimensions. Ziolkowska dimensions. In this research, economic perspective was evaluated in
[33] developed a multi-objective framework to rank different feedstocks terms of net annual profit while environmental impact was estimated
based on their potential of achieving sustainability goals in an unclear using EI99 technique. A summary of the studies discussed is provided in
decision environment and concluded that currently, feedstocks with Table 1.
higher sustainability potentials have a low level of commercialization Based on the above-discussed gaps in the field of biodiesel SCND, this
potential. Gonela et al. [34] presented a stochastic model to optimize the paper presents a WAF-based sustainable biodiesel SCND model that
bioethanol SC and minimize costs under social and environmental ob addresses sustainability under the TBL framework in an uncertain
ligations. In their research, GHG emissions were considered to evaluate environment. The proposed study differs from the existing ones in the
the environmental impact, while the social aspect was assessed by following aspects:
considering the land used for irrigation. de Jong et al. [35] assessed the
impact of facilities integration, scale, SC configuration, and trans i. Developing a multi-objective MILP model for multi-feedstock,
portation on the biofuel production cost by using a geographically multiperiod animal fat-based biodiesel SCND problem in which
explicit cost optimization model. Cambero et al. [36] formulated an both strategic level decisions (locations, numbers, economies of
3
Fig. 1. Proposed framework of a sustainable waste animal fat-based biodiesel supply chain.
scales, of fat preprocessing platns, biorefineries, and distribution the economic objective function in terms of the total SC cost (TSC)
centers) and tactical level decisions (feedstock, fat oil, biodiesel reduction, the environmental objective in terms of a decrease in envi
flows in each period among processing facilities) are made in an ronmental impact, and the social objective in terms of an increase in
integrated manner by taking into account their economic, social economic impact and economic development. The suggested model
and environmental impacts on the society. enables the managers to obtain both strategic and tactical level decisions
ii. Accounting for the epistemic uncertainty in the input parameters associated with the considered SC. The strategic decisions considered
of WAF-SCND model within robust paradigm by applying RPP are (1) number and locations of fat preprocessing center/rendering
approach, which ensures providing robust results while enabling units, (2) number and locations of biorefineries, (3) number and loca
the managers to model the uncertain environment and decide risk tions of biodiesel storage and distribution facilities, (4) production
attitudes following real-time dimensions. technology used in biorefineries, and (5) capacity levels of preprocessing
iii. Employing the latest social (SLCAP) and environmental impact plants, biorefineries, and distribution centers. The following tactical
(SimaPro) assessment tools for the quantification of social and decisions are considered in each planning phase: (1) amount of WAF to
environmental aspects under a well-to-wheel framework in an be sent to each fat preprocessing center, (2) amount of animal fat oil to
animal fat based biodiesel SC. be sent to each biorefinery, (3) amount of biodiesel to be extracted by
iv. Proposing an interactive MRPCCP methodology using a rolling each facility, (4) amount of extracted biodiesel to be supplied to each
planning horizon approach that can efficiently update the SC storage and distribution facility, and (5) amount of biodiesel transported
decision in each planning period. from each distribution center to the demand zone.
In addition, due to inherent uncertainty associated with the collected
3. Mathematical model information the important parameters of the model are considered
fuzzy/uncertain. Based on the available some of the historical data and
3.1. Introduction of WAF-SCND problem experts’ opinion, trapezoidal possibility distribution is considered for
each uncertain parameter. This study considers the following sources of
The problem addressed in this research is the design of an efficient uncertainties: (i) biodiesel demand triggered by variations in petroleum
animal fat-based SC network integrating all the echelons while consid diesel production, (iii) variation in purchasing cost of biomass arising
ering both commercialization and sustainable development aspects. The from market fluctuations, (iv) environmental impact at different points
considered WAF SC network consists of five echelons: i) fat collection of the biodiesel SC, e.g., in WAF collection, biodiesel extraction process,
points, (ii) preprocessing plants, (iii) biorefineries, (iv) storage and and transportation of biodiesel, and (v) economic impact and job crea
distribution centers, and (v) demand zones. The required WAF from tion potential highlighting from insufficient information.
slaughterhouses and collection centers is collected and transported to Due to a lack of available data, RPP technique is employed to
the preprocessing plants (rendering units), where it is converted into a minimize the impact of the abovementioned sources of uncertainties.
purified form of animal fat oil. Then, it is transferred to biorefineries, The notations used in the WAF-SCND model are listed in Appendix A.
where biodiesel is produced during the transesterification reaction and
glycerol is obtained as a byproduct. The biodiesel is transported from the 3.2. Assumptions
biorefineries to the storage and distribution centers. Finally, from dis
tribution centers biodiesel is supplied to the customer to fulfill the i. A homogeneous fleet of trucks is considered at each required
required demand. Road transportation is considered for transporting the stage of the animal fat-based biodiesel SC.
WAF, animal fat oil, and biodiesel across the network. Fig. 1 illustrates ii. Only chicken (poultry) fat, mutton (goat and sheep) fat, and beef
the underlying structure of the animal fat-based biodiesel SC. (bull, ox, and calf) fat are considered as WAF.
In this research, a multi-objective optimization model is proposed for iii. A planning horizon of three years is considered.
the aforementioned SC by considering economic, environmental, and iv. Biodiesel demand in each demand zone is uncertain for which a
social dimensions as objectives. The presented model takes into account trapezoidal possibility distribution is considered.
4
v. Based on historical data of fossil diesel consumption, the demand tions Industrial Development Organization (UNIDO). By using ecoinvent
for B10 type blend (10% biodiesel, 90% diesel) is considered in database 3.6, the environmental impact of industrial processes can be
each planning horizon. obtained without the support of an LCA expert team. EI99 encompasses
three types of environmental damages: i) human health, ii) ecosystem,
3.3. Formulation of objective functions and iii) earth resource diminution. In EI99 tool, based on the priority
weights of the aforementioned damage categories, three perspectives, i.
3.3.1. Cost objective function e., a) individualist, b) average hierarchist, and c) egalitarian, are
Total animal fat-based biodiesel SC cost considered. In this research, the hierarchist perspective is used, wherein
weightages of 40%, 40%, and 20% are allocated to the ecosystem,
Minimize feconomic = Costopen +Costfeed +Costhand +Costinv +Costprod +Costtrans human health, and resource diminution, respectively. Details of the
(1) environmental objective of the proposed WAF-SCND model are provided
Fixed biodiesel processing facilities installation cost below:
L ∑
∑ S O ∑
∑ R ∑
A P ∑
∑ B Minimize fenvI = EnvIopen + EnvIhand + EnvIinv + EnvIprod + EnvItrans (7)
Costopen = PP̃Cls uls + BĨRora xora + BD̃Dpb ypb
l s o r a p b Environmental impact of establishing biodiesel SC facilities:
(2) L ∑
∑ S O ∑
∑ R ∑
A P ∑
∑ B
EnvIopen = EPC̃Hls uls + EB̃IRora xora + EB̃DDpb ypb
Feedstock purchase costs and handling costs l s o r a p b
I ∑
∑ L ∑
Q ∑
T I ∑
∑ L ∑
Q ∑
T (8)
Costfeed +Costhand = FP̃Ciqt πilqt + H P̃PClq πilqt
i l q t i l q t Environmental impact of feedstock handling:
(3) I ∑
∑ L ∑
Q ∑
T
EnvIhand = EH P̃PClq πilqt (9)
Inventory holding cost i l q t
∑ Environmental impact of holding inventory at different stages of the

L ∑
Q ∑
T O ∑
∑ R ∑
T
Costinv = I H̃Plqt IFlqt + I H̃Bort IBort
l q t o r t
biodiesel SC:
O ∑
∑ R ∑
T P ∑
∑ T L ∑
∑ Q ∑
T O ∑
∑ R ∑
T
+ I H̃Gort IGort + I H̃Dpt IDpt (4) EnvIinv = EĨHPlqt IFlqt + EĨHBort IBort
o r t p t l q t o r t
Animal fat oil, biodiesel, and glycerol production cost ∑

O ∑
R ∑
T P ∑
∑ T
+ EĨHGort IGort + EĨHDpt IDpt (10)
Q ∑ o r t p t
L ∑
∑ T O ∑
∑ R ∑
T
Costprod = CP̃PClqt Φlqt + CB̃Port Ψort
l q t o r t
Environmental impact of animal fat oil, biodiesel, and glycerol
production:
∑
O ∑
R ∑
T
+ CG̃Port Πort (5) L ∑
∑ Q ∑
T O ∑
∑ R ∑
T
o r t
EnvIprod = EP̃PClqt Φlqt + EB̃Port Ψort
Transportation cost l q t o r t
I ∑
∑ L ∑
Q ∑
T L ∑
∑ O ∑
Q ∑
T O ∑
∑ P ∑
R ∑
T P ∑
∑ D ∑
T O ∑
∑ G ∑
R ∑
T
Costtrans = CC̃Tilq πilqt + CC̃OTloq π loqt + CB̃DTop πoprt + CB̃DTpd πpdt + CG̃Tog π ogrt (6)
i l q t l o q t o p r t p d t o g r t
3.3.2. Environmental impact objective function O ∑

∑ R ∑
T
To make the biodiesel SC sustainable, the environmental impact of + EG̃Port Πort (11)
all SC processes should be quantified. In this study, parameters related to o r t
the environmental objective are quantified by using the Eco-indicator 99 Environmental impact of transporting material among SC facilities:
(EI99) methodology and SimaPro 9.0 (www.pre-sustainability.com)
software tool. EI99 is an LCA-based tool proposed by Goedkoop [44] and
now available as software. It measures the environmental impact in
terms of points (pt), which denotes 1/1000th of the annual environ 3.3.3. Social impact objective function
mental impact of an average European citizen [17]. SimaPro 9.0 is an The social dimension is the third major pillar of the TBL approach.
effective tool that can efficiently estimate the environmental effect of Quantifying the social impact of a specific activity in an SC is intricate
most industrial processes. It is highly recommended by the United Na due to the involvement of several stakeholders and multidisciplinary
5
I ∑
∑ L ∑
Q ∑
T L ∑
∑ O ∑
Q ∑
T O ∑
∑ P ∑
T ∑
R P ∑
∑ D ∑
T O ∑
∑ G ∑
T ∑
R
EnvItrans = EC̃Tilq πilqt + EC̃OTloq πloqt + EB̃DTop πoprt + EB̃DTpd πpdt + EG̃Tog π ogrt
i l q t l o q t o p t r p d t o g t r
(12)
issues [45]. Several guidelines and methods are proposed by researchers equally important to create a balance in job opportunities among all the
to measure the social influence of SC activities. Of these, a framework regions. For this purpose, employment indicator (EI) is formulated as
named SLCAP developed by Benoît et al. [46] is used to assess the social follows:
impact in the WAF-SCND model because of its following advantages. (1) L ∑
∑ S O ∑
∑ R ∑
A P ∑
∑ B
It is an LCA-based product-oriented social impact evaluation tool; EI = ŨEl J̃ ls uls + ŨEo J̃ ora xora + ŨEp J̃ pb ypb (14)
therefore, it follows the SC dynamics, thereby supporting the WAF- l s o r a p b
SCND model formulation. (2) It mainly focuses on social issues and
To calculate the total social impact, the estimated values of ED and EI
not on the organizational governance and environmental aspect; hence,
must be normalized and assigned priority weights. Equal weights are
the social concerns through the SC are addressed more effectively in a
assigned for both indicators. A normalization procedure is provided
TBL paradigm. (3) This framework is newly developed; therefore, it is
below in Eqs. (15)–(19):
equipped with recent advances in the field of social impact assessment.
The SLCAP framework comprises five stakeholders, i.e., employees, end-
product consumers, value chain actors, society, and the local commu
⎧[{ } ]⎫
⎪
⎪ ∑L ∑S ∑O ∑R ∑A ∑P ∑B ⎪
⎪
⎪
⎨ l s E Ṽ ls uls (1 − RDLl ) + o r a E Ṽ ora xora (1 − RDLo ) + p b E Ṽ pb ypb (1 − RDLp ) − EDmin ⎪
⎪
⎪
⎬
EDnorm = γ (15)
⎪
⎪ EDmax − EDmin ⎪
⎪
⎪
⎪ ⎪
⎪
⎩ ⎭
⎧[{ } ]⎫
⎪
⎪ ∑L ∑S ∑O ∑R ∑A ∑P ∑B ⎪
⎪
⎪
⎨ l s ŨEl J̃ ls uls + o r a ŨEo J̃ ora xora + p b ŨEp J̃ pb ypb − EImin ⎪
⎪
⎪
⎬
EInorm = ζ (16)
⎪
⎪ EImax − EImin ⎪
⎪
⎪
⎪ ⎪
⎪
⎩ ⎭
nity. The WAF-SCND model decisions affect only “society” and “local
community.” Therefore, indicators for the evaluation of these stake ED − EDmin
EDnorm = (17)
holders are developed in this research. EDmax − EDmin
In the stakeholder category of “society,” the “level of economic
development in a specific society” is affected by the SCND decisions. In EInorm =
EI − EImin
(18)
this perspective, balanced economic development is a major concern. EImax − EImin
The decision of biodiesel processing facilities significantly affects bal
ance. Hence, installing biodiesel processing plants in less developed Maximize fsocial = γEDnorm + ζEInorm (19)
areas is the priority of the “society” stakeholder so that a mutual balance
with respect to economic development among all areas can be achieved. 3.4. Formulation of constraints
In this context, economic development (ED) indicator is designed as
follows [47]: Constraint (20) ensures that the quantity of chicken, mutton, and
beef fat transported to preprocessing centers must be less than the
∑ available quantity at fat supply points during each planning period t.
∑
L ∑
S O ∑
R ∑
A
ED = EṼ ls uls (1 − RDLl ) + EṼ ora xora (1 − RDLo )
l s o r a
∑
L
∑
P ∑ πilqt ⩽σ̃ iqt ∀i, q, t (20)
+ B
b EṼ pb ypb (1 − RDLp ) (13) l
p
Constraints (21) and (22) fulfill biodiesel and glycerol demands for
In the category of “local community” stakeholder, the regional demand points in each period, respectively.
employment level is affected by the WAF-SCND model decisions. The
location of biodiesel processing facilities and their capacity levels ∑
P
πpdt ⩾μ̃dt ∀ d, t (21)
directly affect the employment level of a particular region. Thus, it is p
6
∑ ∑
O
o
R
r πogrt ⩾μ̃gt ∀ g, t (22) ∑
R ∑
R ∑
R R ∑
∑ P
IBort = IBo,r,t− 1 + Ψort − π oprt ∀o, t|t⩾2 (33)
Constraint (23) provides the total amount of chicken, mutton, and r r r r p
beef fat oil produced at the respective processing center in a specific

period. The quantity of raw animal fat is reduced by a conversion factor ∑ ∑ ∑
R R R ∑
G
IGort = Πort − πogrt ∀o, t|t = 1 (34)
when fat is converted into animal oil. r r r g
∑
I
φq πilqt = Φlqt ∀ l, q, t (23) ∑
R ∑
R R ∏
∑ R ∑
∑ G
i IGort = IGo,r,t− 1 + − πogrt ∀o, t|t⩾2 (35)

r r r ort r g
Constraint (24) ensures that the total animal fat oil sent from each fat
preprocessing plant to all of the biorefineries is less than the total animal Eqs. (36) and (37) are the biodiesel inventory constraints for distri
fat oil produced. bution centers.
∑
O
O ∑
∑ R ∑
D
πloqt ⩽Φlqt ∀l, q, t (24) IDpt = πoprt − πpdt ∀p, t|t = 1 (36)
o r d
o
Constraints (25) and (26) provide the total produced volume of O ∑

∑ R ∑
D
biodiesel and glycerol, respectively, at a biorefinery in a specific period. IDpt = IDp,t− 1 + π oprt − π pdt ∀p, t|t⩾2 (37)
The quantity of animal fat oil is reduced by the yield factor when bio
o r d
diesel and glycerol are produced from animal fat oil. Eqs. (38) and (39) ensure that at most, one capacity level and one
technology for each operational biorefinery can be selected.
∑
L ∑
Q ∑
R
πloqt ηq = Ψort ∀o, t (25) ∑
A
l q r xora ⩽1 ∀o, r (38)
a
∑ Q
L ∑ ∑
R
πloqt (1 − ηq ) = Πort ∀o, t (26) ∑
R
l q r xora ⩽1 ∀o, a (39)
r
Constraints (27) and (28) guarantee that the total volume of bio
diesel and glycerol transported from a biorefinery is less than their Eqs. (40) and (41) guarantee that only one capacity level is chosen
respective total produced amounts. for each operational preprocessing plant and biodiesel storage and dis
tribution center.
∑
R R ∑
∑ P
Ψort ⩾ πoprt ∀o, t (27) ∑
S
r r p uls ⩽1 ∀l (40)
s
∑
R R ∑
∑ G
Πort ⩾ πogrt ∀o, t (28) ∑
B
r r g ypb ⩽1 ∀p (41)
b
Constraint (29) states that in each period, the total quantity of bio
Constraints (42)–(44) are the capacity constraints for the animal fat
diesel transported from biorefineries to each distribution center must be
preprocessing centers, biorefineries, and storage distribution centers,
greater than the total amount distributed to the biodiesel demand
respectively.
points.
∑
I ∑
Q ∑
S
∑ (42)
O ∑
R ∑
D
πoprt ⩾ πpdt ∀p, t (29)
πilqt ⩽ Capis uls ∀l, t
i q s
o r d
Constraints (30) and (31) are the inventory balance constraints at the ∑
O ∑
B
animal fat preprocessing center. Constraint (30) states that the animal πoprt ⩽ Cappb ypb ∀p, r, t (43)
fat oil inventory during the first period is equal to the amount of animal
o b
fat oil produced at preprocessing centers minus the quantity of animal ∑

L ∑
Q R ∑
∑ A
fat transported to the biorefineries, while constraint (31) provides the πloqt ηq ⩽ Capoa xara ∀o, t (44)
animal fat oil inventory balance at preprocessing centers for the second l q r a
and succeeding periods. Eqs. (45)–(47) limit the maximum number of various operational
∑
O processing facilities owing to budget limitations.
IFOlqt = Φlqt − πloqt ∀l, q, t |t = 1 (30)
o ∑
L ∑
S
uls ⩽MaxU (45)
l s
∑
O
IFOlqt = IFOl,q,t− 1 + Φlqt − π loqt ∀l, t t⩾2 (31)
o ∑
O ∑
R ∑
A
xora ⩽MaxO (46)
Eqs. (32)–(35) are the inventory balance constraints for biodiesel and o r a
glycerol at biorefineries.
∑
P ∑
B
∑
R ∑
R R ∑
∑ P ypb ⩽MaxY (47)
IBort = Ψort − πoprt ∀o, t|t = 1 (32) p b
r r r p
Constraints (48) and (49) ensure that all types of decision variables
considered in the WAF-SCND model are non-negative.
7
\; Continuous decision variables⩾0 (48) WAF-SCND may not provide practically feasible optimal solutions. The
literature review stipulates that PCCP, which is an extension of possi
Binary and non-negativity
bilistic programming, can be used to deal with the challenge of opti
uls , xora , ypb ∈ {0, 1} ∀l, s, o, p, r, a, b (49) mizing frameworks having epistemic uncertainties. The generic
functional procedure of PCCP is provided below.
4. Solution methodology Consider the following single-objective uncertain optimization
model:
Business environment of animal fat-based biodiesel SC is highly dy Minimize k = ãq + b̃r
namic in nature as a result, a high level of uncertainty is associated with Subject to Ar⩽B̃,
the optimization model input parameters. Two types of uncertainties are Dr = 0,
(50)
considered in the parametric information. First is the randomness, Fr⩽G̃q,
created by the random trait of the parameters [48]. In this type of un Jr⩾1,
certainty, sufficient knowledge related to probability distribution is r⩾1, q ∈ {0, 1},
present and it can be handled effectively by a stochastic programming
technique [49,50]. Second is the epistemic uncertainty created due to where a, b, B, and G are the imprecise parameters depicting trapezoidal
the unavailability of adequate parametric information, and it can be fuzzy numbers. To change the minimizing objective function in Eq. (50)
handled best by using a possibilistic programming approach [51]. In the to the equivalent PCCP model, credibility measure approach and ex
business environment of a WAF-based biodiesel SC, it is very difficult to pected value (EV) operator are used.
accurately estimate the parameters such as fat supply potential, bio
diesel demand, biofuel prices, and various associated costs. Owing to the i. EV operator
unavailability of past data and lack of information, these WAF SC pa
rameters are contaminated with uncertainties. As mentioned earlier, The EV operator is applied to obtain the expected crisp value of fuzzy
possibilistic programming can efficiently handle these uncertainties. parameters of the model. This study takes into account the signed dis
However, possibilistic programming derives solutions based on the tance approach for the defuzzification of uncertain parameters. The EV
average values of dubious parameters; therefore, it cannot control risks of a fuzzy numberf̃ = (f1, f2, f3, f4 )considering a trapezoidal possibility
effectively. distribution using the signed distance method is as follows:
1
d(f̃ , 0̃1 ) = (f1, f2, f3, f4 ) (51)
4.1. Why RPP is best suitable for this study? 4
To deal with the shortcomings of possibilistic programming, robust

ii. Credibility measure
programming, which was propounded by Pishvaee et al. [45], is
implemented to obtain the solutions for the WAF-SCND model. RPP is a
Credibility measure is considered to capture the chance constraints
risk aversive approach and combination of robust and possibilistic
of the framework. In the literature, two types of measures, namely ne
programming, which provides the following advantages over other
cessity (Nec) and possibility (Pos) measures, are commonly used to solve
methodologies:
the PCCP models. Both measures depict two extremes of an uncertain
parameter. The first type of measure represents the most optimistic level
• RPP enables the managers to select any point between the range of
and second one represents the most pessimistic level. The Nec and Pos
fully pessimistic and optimistic based on real time conditions. RPP is
measures of an uncertain numberf̃considering a trapezoidal possibility
classified in the following categories:
distribution are indicated in Eqs. (53)–(56).
(a) realistic category provides a compromise between the cost of
robustness and achieved level of robustness while increasing the The fuzzy membership function off̃can be formulated as
SC performance ⎧
⎪ r − f1
⎪
(b) hard worst category follows the min–max logic and delivers ⎪
⎪
⎪ f2 − f1
f1 ⩽r < f2
⎪
maximum protection against uncertainty by completely over ⎪
⎪
⎪
⎪
⎪
seeing the possibility of being infeasible ⎪
⎨ 1 if f2 ⩽r⩽f3
(c) soft worst category also minimizes the worst case, however in μf̃ (r) =
⎪
⎪
soft worst case the model constraints are relatively relaxed. ⎪
⎪
⎪
f4 − r
f3 < r⩽f4
⎪
• RPP is best suitable for the cases where historical data are not ⎪ f4 − f3
⎪
⎪
⎪
⎪
available consequently, probability distribution for the uncertain ⎪
⎩ 0 otherwise,
parameter cannot be obtained [52]. The unique plus of using RPP
over stochastic techniques is that it does not demand historical re where r ∈ R. (52)
cords of uncertain parameter; rather, possibility distribution func The Pos and Nec measures for f̃⩽rand f̃⩾rare obtained as follows:
tion is used.
{ }
• RPP technique is highly interactive; therefore, by incorporating the Nec f̃ ⩽r ⩾ℓ ⇔
r − f3
⩾ℓ ⇔ r⩾(1 − ℓ)f3 + ℓf4, (53)
values of interactive parameters such as constraint violation penalty f4 − f3
levels, optimistic-pessimistic parameters, and confidence level, it can { }
be implemented for a given problem. Nec f̃ ⩾r ⩾ℓ ⇔
f2 − r
⩾ℓ ⇔ r⩽ℓf1 + (1 − ℓ)f2 , (54)
f2 − f1
Hereafter, generic formulation of possibilistic programming and RPP { } r − f1
are described. Pos f̃ ⩽r ⩾ℓ ⇔ ⩾ℓ ⇔ r⩾(1 − ℓ)f1 + ℓf2, (55)
f2 − f1
{ }
4.2. Possibilistic chance-constrained programming (PCCP) f4 − r
Pos f̃ ⩾r ⩾ℓ ⇔ ⩾ℓ ⇔ r⩽ℓf3 + (1 − ℓ)f4 , (56)
f4 − f3
In real-life situations of the WAF SC, most of the collected data is
The Nec and Pos measures indicate extreme values, whereas in real-
corrupted with epistemic uncertainty. The utilization of such data for
8
life situations, the attitude of the decision-makers lies somewhere be

1 1
tween the mentioned range. Therefore, Liu et al. [53] proposed a Minimize E[k] = (a1 + a2 + a3 + a4 ) q + (b1 + b2 + b3 + b4 )r
4 4
credibility (Cr) measure to obtain a compromise between these extreme
levels, which is practically more suitable. Eq. (57) gives the credibility Subject to Ar⩽(2ℓ1 − 1)B1 + (2 − 2ℓ1 )B2 ,
measure of the proposed uncertain case. Dr = 0, (63)
{ } 1 { } Fr⩽[(2ℓ2 − 1)G1 + (2 − 2ℓ2 )G2 ]q,
Cr f̃ ⩾r = (Pos f̃ ⩾x + Nec{f ⩾r}) (57)
2 Jr⩾1,
The fuzzy membership functions of a trapezoidal fuzzy numberf̃ r ≥ 1, q ∈ {0, 1}, 0.5⩽ℓi ⩽1.
using the Cr measure are stated as:
⎧ 4.3. Modified PCCP (MPCCP)
⎪
⎪ 1, r ∈ ( − ∞, f1 ]
⎪
⎪
⎪
⎪ r − f
⎪
⎪
⎪
⎪
1
, r ∈ (f1 , f2 ] Though the credibility measure in the PCCP approach gives a
⎪ 2 (f2 − f1 )
⎪
⎪
⎪ compromise solution between both extremes, the decision-maker is
{ } ⎨
Cr f̃ ⩽r =
1
, r ∈ (f2 , f3 ] (58) bound to select a specific point. This limitation reduces the reliability of
⎪
⎪
⎪ 2 the results obtained using the Cr measure. To address this issue, Xu et al.
⎪
⎪
⎪
⎪
⎪
r − 2f3 + f4 [54] proposed a fuzzy Me measure. A convex merger of the Nec and Pos
⎪ , r ∈ (f3 , f4 ]
⎪
⎪
⎪ 2 (f4 − f3 ) measures provides flexibility to decision-makers in choosing any point
⎪
⎪
⎩ 1, r ∈ (f4 , +∞] within the given range as below:
{ } { } { }
⎧ Me f̃ ⩾r = λPos f̃ ⩾r + (1 − λ)Nec f̃ ⩾r ,
⎪
⎪ 1, r ∈ ( − ∞, f1 ] { } [ { } { }]
⎪
⎪
⎪
⎪
⎪ 2f 2 − f1 − r = Nec f̃ ⩾r + λ Pos f̃ ⩾r − Nec f̃ ⩾r , (64)
⎪
⎪ , r ∈ (f1 , f2 ]
⎪
⎪ 2(f2 − f1 )
⎪
⎪
{ } ⎪⎨ where λ is the combined pessimistic-optimistic parameter and its value
1
Cr f̃ ⩾r = , r ∈ (f2 , f3 ] (59) may vary between 0 and 1. Hence, it can be concluded that the Nec, Pos,
⎪ 2
⎪
⎪
⎪ and Cr measures are special cases of the Me measure described as
⎪
⎪ f4 − r
⎪
⎪
⎪
⎪ , r ∈ (f3 , f4 ] follows:
⎪
⎪ 2(f2 − f3 )
⎪
⎪
⎩ 0, r ∈ (f4 , +∞] • λ = 0 denotes that the decision-maker is opting for a pessimistic
Using Eqs. (58) and (59), the Cr measure for an uncertain constraint approach for which Me = Nec holds.
can be derived as follows: • λ = 0.5 indicates that the decision-maker is adopting a compromise
approach for which Me = Cr holds.
{ }
Cr f̃ ⩽r ⩾ℓ ⇔
r − 2f3 + f4
⩾ℓ ⇔ r⩾(2 − 2ℓ)f3 + (2ℓ − 1)f4 , ∀ℓ⩾0.5 (60) • λ = 1 shows that the decision-maker is choosing an optimistic
2(f4 − f3 ) approach for which Me = Pos holds.
{ } 2f2 − f1 − r
Cr f̃ ⩾r ⩾ℓ ⇔ ⩾ℓ ⇔ r⩽(2ℓ − 1)f1 + (2 − 2ℓ)f2, ∀ℓ⩾0.5 (61) The membership functions of the Me measure can be formulated as
2(f2 − f1 ) follows:
The uncertain model stated in Eq. (50) is transformed into its ⎧
⎪ 0, r⩽f1
equivalent crisp form using the Cr measure and EV operator as: ⎪
⎪
⎪
⎪
⎪
⎪ r − f
[ ] ⎪ λ
⎪ 1
, f1 ⩽r⩽f2
Minimize E[k] = E[ã]q + E b̃ r ⎪
⎪ f2 − f1
{ } ⎪⎨
Subject to Me f̃ ⩽r = λ, if f2 ⩽r⩽f3 (65)
{ } ⎪
⎪
⎪
⎪ r − f
Cr Ar⩽B̃ ⩾ℓ1 , ⎪
⎪ λ + (1 − λ)
3
, f3 ⩽r⩽f4
⎪
⎪
⎪
⎪ f4 − f3
⎪
⎪
Dr = 0, ⎩ 1, r⩾f4
{ }
Cr Fr⩽G̃q ⩾ℓ2 ,
⎧
Jr⩾1, ⎪
⎪ 1, r⩽f1
⎪
⎪
⎪
⎪
r⩾1, q ∈ {0, 1}, (62) ⎪
⎪ f
⎪ λ + (1 − λ) 2
− r
, f1 ⩽r⩽f2
⎪
⎪ f2 − f1
{ } ⎪⎨
where, ℓ1 andℓ2 represent the minimum satisfaction levels of chance Me f̃ ⩾r = λ, if f2 ⩽r⩽f3 (66)
constraints. The decision-maker decides the values of ℓ1 andℓ2 based on ⎪
⎪
⎪ f − r
⎪
⎪
available information. The equivalent crisp formulation of a PCCP
4
⎪
⎪ λ f4 − f3 ,
⎪
⎪
f3 ⩽r⩽f4
model provided in Eq. (62) is as follows: ⎪
⎪
⎪
⎩ 0, r⩾f4
The EV of an uncertain number f using the Me measure approach is

obtained as follows:
9
Table 2 ℓ2 , and λare decided by managers. This technique not only handles the
Generic form of payoff table for WAF-SCND model objectives. epistemic uncertainty but also allows decision-makers to consider any
Total supply chain Environmental Social impact point of uncertainty between the range of Nec and Pos measures. How
(TSC) cost impact ever, this technique cannot control the deviation of the objective func
TSC cost minimize TSC s.t. ⌣ ⌣ ⌣ ⌣ ⌣ ⌣ tions from their average/planned performance, which may pose serious
f21 = EI( r , v , w) f31 = SI( r , v , w)
(B20)–(B22) and risks in real-life situations. To overcome this limitation, a robust version
(23)–(49), of the MPCCP model with some modifications is developed based on the
methodology proposed by Pishvaee et al. [45].
⌣ ⌣
calculate:( r , v ,
⌣ ⌣ ⌣
w)f11 = TSC ( r , v ,
4.3.1. MRPCCp
⌣
w)
Environmental ⌢ ⌢ ⌢
f12 = TSC( r , v , w) minimize EI s.t. ⌢ ⌢ ⌢
f32 = SI( r , v , w) The robust PCCP is a risk-averse technique that incorporates opti
impact (B20)–(B22) and mality and feasibility robustness in optimization problems, simulta
(23)–(49),
⌢ ⌢ neously. The optimality robustness minimizes the deviation from the
calculate:( r , v ,
⌢ ⌢ ⌢ ⌢ objective optimal value, whereas the feasibility robustness ensures that
w)f22 = EI ( r , v , w)
the decisions of the considered optimization model are always feasible
Social impact f13 = TSC(r̈, v̈, ẅ) f23 = EI(r̈, v̈, ẅ) maximize SI s.t.
for all realizations of fuzzy parameters. In this research, the RPP-II
(B20)–(B22) and
(23)–(49), formulation of robust programming developed by Pishvaee et al. [45]
calculate:(r̈, v̈, is employed. The counterpart form of the MRPCCP model in Eq. (70) is
ẅ)f33 = SI (r̈, v̈, ẅ) provided below:
[{ } ]
(ℓ1 − λ)B1 + (1 − ℓ1 )B2
Minimize E [k] + κ(kmax − E[k]) + ω1 − B1 +
1− λ
∫+∞ ∫0 [{( ) } ]
(ℓ2 − λ)G1 + (1 − ℓ2 )G2
EMe [f ] = Me{f ⩾r} dr − Me{f ⩽r}. dr, ω2 − G1 c
1− λ
0 − ∞ (67) [ ]
(ℓ1 − λ)B1 + (1 − ℓ1 )B2
1− λ λ Subject to Ar⩽
= (f1 + f2 ) + (f3 + f4 ) 1− λ
2 2
Dr = 0,
Therefore, the Me measures of a trapezoidal fuzzy number f̃for f̃⩽r [ ]
(ℓ2 − λ)G1 + (1 − ℓ2 )G2
and f̃⩾r are provided in Eqs. (68) and (69), respectively [55,56]. Fr⩽ q,
1− λ
{ } r − f3 (ℓ − λ)f4 + (1 − ℓ)f3
Me f̃ ⩽r ⩾ℓ ⇔ λ + (1 − λ) ⩾ℓ ⇔ r⩾ , (68) Jr⩾1,
f4 − f3 1− λ
r⩾1, q ∈ {0,1}, 0.5⩽ℓi ⩽1.
{ } f2 − r (ℓ − λ)f1 + (1 − ℓ)f2 (71)
Me f̃ ⩾r ⩾ℓ ⇔ λ + (1 − λ) ⩾ℓ ⇔ r⩽ , (69)
f2 − f1 1− λ In Eq. (71), the first term E [k]is the EV that curtails the average value
Using the abovementioned derivations of the Me measure and EV of the system, whereas the second term κ(kmax − E[k]) decreases the
operator, the PCCP model provided in Eq. (50) is transformed into a maximum deviation around the EV of the objective function. The third
crisp counterpart of the MPCCP model as: and fourth terms minimize gap between the worst-case value and the
value of fuzzy parameter used in the constraint. Briefly, the proposed
[ ] [ ]
1− λ λ 1− λ λ
Minimize E [k] = (a1 + a2 ) + (a3 + a4 ) q + (b1 + b2 ) + (b3 + b4 ) r,
2 2 2 2
[ ]
(ℓ1 − λ)B1 + (1 − ℓ1 )B2
Subject to Ar⩽
1− λ
Dr = 0,
[ ] (70)
(ℓ2 − λ)G1 + (1 − ℓ2 )G2
Fr⩽ q,
1− λ
Jr⩾1,
r⩾1, q ∈ {0, 1},
0.5⩽ℓi ⩽1.
MRPCCP formulation seeks a tradeoff among three components of the

objective function: (1) EV, (2) feasibility robustness, and (3) optimality
Eq. (70) expresses an MPCCP model. This model is interactive in robustness of the system.
nature for which depending upon the given situation, the values of ℓ1 ,
Table 3
Scale to translate the linguistic variables into fuzzy numbers.
Linguistic variables Very low Low Medium High Very high Excellent
Fuzzy numbers (0,0,0.2) (0,0.2,0.4) (0.2,0.4,0.6) (0.4,0.6,0.8) (0.6,0.8,1.0) (0.8,1.0,1.0)
10
4.4. Solution methodology to obtain POSs of the WAF-SCND model ismaxf3 (r̈, v̈, ẅ) while (r̈, v̈, ẅ)are its optimal solutions. Afterward, the
values of TSC (f13 ) and environmental impact (f23 ) are obtained at the
After converting the uncertain WAF-SCND model into its equivalent value of f33 . Table 2 provides the methodology to obtain the payoff table.
MRPCCP formulation (refer to Eqs. (B-20)–(B-25) and (23)–(49)), the
next task is to obtain a set of POSs, which contains the most efficient, b) Develop a fuzzy membership function for the model objectives
⎧ ⌢ ⌢ ⌢
⎪ 0, If fTSC ⩾Max{TSC( r , v , w), TSC(r̈, v̈, ẅ)}
⎪
⎪
⎪
⎨ ⌢ ⌢ ⌢
Max{TSC( r , v , w), TSC(r̈, v̈, ẅ)} − fTSC
(72)
⌣ ⌣ ⌣ ⌢ ⌢ ⌢
μTSC (x) = ⌢ ⌢ ⌢ ⌣ ⌣ ⌣ , If TSC ( r , v , w) < fTSC < Max{TSC( r , v , w), TSC(r̈, v̈, ẅ)}
⎪
⎪ Max{TSC( r , v , w), TSC(r̈, v̈, ẅ)} − TSC ( r , v , w)
⎪
⎪
⎩ ⌣ ⌣ ⌣
1, If fTSC ⩽ TSC ( r , v , w)
⎧ ⌣ ⌣ ⌣
⎪ 0, If fEI ⩾Max{EI( r , v , w), EI(r̈, v̈, ẅ)}
⎪
⎪
⎪
⎨ ⌣ ⌣ ⌣
Max{EI( r , v , w), EI(r̈, v̈, ẅ)} − fEI
(73)
⌢ ⌢ ⌢ ⌣ ⌣ ⌣
μEI (x) = ⌣ ⌣ ⌣ ⌢ ⌢ ⌢ , If EI ( r , v , w) < fEI < Max{EI( r , v , w), EI(r̈, v̈, ẅ)}
⎪
⎪ Max{EI( r , v , w), EI(r̈, v̈, ẅ)} − EI ( r , v , w)
⎪
⎪
⎩ ⌢ ⌢ ⌢
1, If fEI ⩽EI ( r , v , w)
⎧ ⌢ ⌢ ⌢ ⌣ ⌣ ⌣
⎪ 0, If fSI ⩽Min{SI( r , v , w), SI( r , v , w)}
⎪
⎪ ⌢ ⌢ ⌢ ⌣ ⌣ ⌣
⎪
⎨ fSI − Min{SI( r , v , w), SI( r , v , w)} ⌢ ⌢ ⌢ ⌣ ⌣ ⌣
If Min{SI( r , v , w), SI( r , v , w)} < fSI < SI(r̈, v̈, ẅ)
μSI (x) = SI(r̈, v̈, ẅ) − Min{SI( r , v , w), SI(⌣r , ⌣
⌢ ⌢ ⌢ ⌣
v , w)}, (74)
⎪
⎪
⎪
⎩ 1,
⎪ If fSI ⩾SI(r̈, v̈, ẅ)
non-dominated and non-inferior solution of the considered model. The

procedure is provided below.
a) Obtain a payoff table for the model objectives

where μTSC, μEI, and μSI are the satisfaction levels for the WAF-SCND
Consider an MRPCCP formulation of a generic multi-objective model optimization model objectives.
g having I objectives as follows:
c) Transform the multi-objective WAF-SCND model into a single-
max/min (f1 (r, v, w), f2 (r, v, w), ... fI (r, v, w)) , ( i = 1, 2, 3...I)
objective model
Such that
(r, v, w) ∈ M
To transform the MRPCCP model presented in Eqs. (B-20)–(B-25)
Here, (r,v,w) are the decision variables and M is the feasible range. To and (23)–(49) into a single-objective model, a modified Torabi and
obtain the Pareto optimal solutions (POS) for the optimization model g, Hassini (TH) technique [57] is used. Procedure of the modified TH
first, a payoff table is developed. For this purpose, the first objective in technique is as follows:
which the TSC minimization is optimized as minf1 (r, v, w), subject to (B-
20)–(B-22) and (23)–(49). The optimized value of the TSC objective is i. Formulate a team of experts, determine the experience years of
f11 = TSC ( r , v , w), while( r , v , w)are its optimal solutions. Later on, the each member, and obtain the experience weight for every indi
⌣ ⌣ ⌣ ⌣ ⌣ ⌣
solution of f11 is used to obtain the value of environmental impact (EI) vidual of the team.
(f21 ) and social impact (SI) (f31 ) functions. In the second stage, the total ii. Make a scale to translate the linguistic variables into fuzzy
environmental impact objective is optimized as minf2 (r, v, w), subject to numbers.
iii. Obtain linguistic variables based evaluation for the preference of
(B-20)–(B-22) and (23)–(49). The value of the optimized objective is min
objectives, and convert these evaluations into fuzzy numbers
f22 ( r , v , w), while ( r , v , w)are its optimal solutions. Subsequently, the
⌢ ⌢ ⌢ ⌢ ⌢ ⌢
using the scale provided in Table 3.
solution of f22 is utilized to estimate the value of TSC (f12 ) and SI (f32 ). iv. Multiply each fuzzy number with its respective experience
Finally, in the last stage, the SI objective is optimized in combination weight.
with the system constraints. The optimized value of SI objective v. Acquire aggregate fuzzy number for objective as below:
11
( )
Γ1 φpes pes pes
1 + Γ2 φ2 + ... Γe φe Γ1 φmost + Γ2 φmost + ... Γe φmost Γ1 φopt opt opt
1 + Γ2 φ2 + ... Γe φe
Rh = , 1 2 e
, (75)
e e e
⎫
Maximize ℘Θ + {(1 − ℘)(φTSC μTSC (x) + φEI μEIe (x) + φSI μSI (x)) } ⎪
⎪
⎪
⎪
such that μTCS (x)⩾Θ ⎪
where, ⎪
⎬
μEI (x)⩾Θ
Rh = (φpes most
h , φh , φopt
h )and Rh is the aggregate fuzzy number for hth μSI (x)⩾Θ ⎪
⎪
⎪
objective Θ ∈ [0, 1] ⎪
⎪
⎪
⎭
Γe is the experience of eth expert. System constraints (B20)--(B22) and (23)--(49)
(78)
vi. Obtain objective priority weight using the Eq (76) as follows:
where ℘ and Θ are the compensation coefficient and minimum satis
φpes most
e + 2φe + φopt
φe = e
(76) faction level, respectively, whereas φTSC , φEI , and φSI are the objective
4
priority weights.
Based on the abovementioned description, the solution algorithm for
the WAF-SCND model is summarized as follows:
vii. Finally, calculate objective normalized weight using Eq (77) as Proposed MRPCCP-based solution algorithm for the WAF-SCND model
Step 1: Transfer the WAF-SCND
follows:
(uls , xora , ypb , IFlqt , IHBort , IGort , IDpt , Φlqt , Ψort , Πort , πilqt , πloqt , πoprt , πpdt ,πogrt ) to
φ MPRCCP (uls , xora , ypb , IFlqt , IHBort , IGort , IDpt , Φlqt , Ψort , Πort , πilqt , πloqt , πoprt , πpdt , πogrt ).
φe = ∑ e (77)
φe Step 2: Determine the value of the Me measure.
e Step 2.1: If the decision-maker has an optimistic approach, then, set 0 < λ⩽0.5.
Step 2.2: If the decision-maker has a pessimistic approach, then, set 0.5 < λ⩽1.
⌣ ⌣ ⌣
Step 3: Determine αPIS
i as the max/min of OFi( r , v , w) s.t. Eqs.
(B20)--(B22) and (23) − (49)for objective i. (PIS = positive ideal solution)
viii. Employ normalized objective preferences in TH method formu Step 4: Determine fi i+1 ( r , v , w)at the cost of OFi+1( r , v , w).
⌢ ⌢ ⌢ ⌢ ⌢ ⌢
Step 5: Determine fi i+2 (r̈, v̈, ẅ)at the cost of OFi+2(r̈, v̈, ẅ).
Step 6: Compare the values of fi i+1 ( r , v , w)and fi i+2 (r̈, v̈, ẅ)obtained from step 4 and
⌢ ⌢ ⌢
Table 4 ( )
step 5. If min of OFi holds, then, max fi i+1 ( r , v , w), fi i+2 (r̈, v̈, ẅ) = αNIS
⌢ ⌢ ⌢
PIS and NIS for WAF-SCND model objectives. i ;
otherwise, (NIS = negative ideal solution)

Objective SC cost ($) Environmental Social ( )
min fi i+1 ( r , v , w), fi i+2 (r̈, v̈, ẅ) = αNIS
⌢ ⌢ ⌢
impact (Pt) impact (%) i .
SC cost ($) 29,893,180,000 699,740,000 0.165 Step 7: Developμi (x) ∀ i as in Eqs. (72)–(74).
(*PIS) (*NIS) Step 8: Transfer the multi-objective MRPCCP formulation of WAF-SCND into a single-
Environmental 89,269,310,000 635,347,100 (*PIS) 5.599 objective model using Eq. (78).
impact (Pt) Step 9: Put numeric values of compensation coefficient (℘) and objective priority
Social impact (%) 184,170,400,000 703,332,200 (*NIS) 69.166 weights (φI ) and obtain the POS set ϴ.
(*NIS) (*PIS) Step 10: if a suitable solution from set, ϴ is found, then
End
* ϖ, ϑ, θ = 0.85, and λ = 0.5. Else
Go to step 2.
lation as follows:
x 100000
x 10000
Supply chain cost Environmental impact

73,000
1,810,000
71,000
Total environmental impact (pt)
1,610,000
69,000
1,410,000
Total supply chain cost ($)
1,210,000 67,000
1,010,000 65,000
810,000
63,000
610,000
61,000
410,000
210,000 59,000
0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0%
Social impact (percentage)
Fig. 2. Supply chain cost and environmental impact objective tradeoff to social impact objective.
12
Fig. 3. Optimum solutions of sustainable WAF-SCND using the MRPCCP approach.
5. Implementation and evaluation Pakistan is utilizing very few biomass resources, which are mostly first-
generation biomass. However, animal fat, which is currently being used
To evaluate the efficiency and performance of the WAF-SCND model, for low-cost purposes, can be utilized as a feedstock to obtain biodiesel.
a case study based on the scenario in Pakistan is investigated. Currently, According to the Livestock and Dairy Development Board (LDDB) of
Table 5
Results of the WAF-SCND model under different combinations of objective priority weights.
Priority Objective value Objective satisfaction Location of biodiesel processing facilities
weights level
(w1, w2, w3) OF1 OF2 OF3 µtce µtei µsi Preprocessing centers Biorefineries Distribution centers
(1,0,0) 47,829,290,000 664,276,900 0.48 90.3 70.4 70.4 Lahore (1), Bannu (8), Mardan Skeikupura (1), Lahore (2), Rawalpindi,
(9), Makran (13) Makran (04), Karak (3), Karachi (5),
(5) Peshawar (7), Quetta
(11)
(0.8,0.1,0.1) 63,660,590,000 656,355,200 0.43 79.8 78.5 63.1 Lahore (1), Bannu (8), Makran DI Khan (2), Makran Lahore (2), Rawalpindi,
(13), M.Abad (15) (4), Karak (5) (3), Karachi (5), Sukkur
(6), Peshawar (7)
(0.7,0.2,0.1) 45,388,400,000 664,193,500 0.49 91.9 70.5 71.1 Lahore (3), Sargodha (4), Bannu Skeikupura (1), DI Lahore (2), Rawalpindi
(8), Mardan (9), Makran (12) Khan (2), Makran (4), (3), Karachi (5), Sukkur
Karak (5) (6), Peshawar (7), Quetta
(11)
(0.7,0.1,0.2) 63,661,550,000 656,355,200 0.43 79.8 78.5 63.1 Lahore (1), Bannu (8), Diamer D. I Khan (2), Makran Lahore (2), Rawalpindi
(11), Makran (13), M.Abad (15) (4), Karak (5) (3), Karachi (5), Sukkur
(6), Peshawar (7)
(0.7,0.15,0.15) 48,377,570,000 677,259,100 0.37 89.9 57.2 54.7 Sargodha (4), , Hyderabad (5), D. I Khan (2), Karachi Lahore (2), Rawalpindi
Larkana (6), Bannu (8), Kalat(12), (3), Karak (5) (3), Karachi (5),
M.Abad (15) Peshawar (7)
(0.65,0.25,0.1) 46,331,530,000 664,208,000 0.49 91.3 70.5 70.9 Lahore (1), Multan (2), Sargodha Sheikhupura (1), D. I Lahore (2), Rawalpindi
(4), Bannu (8), Makran (13), M. Khan (2), Makran (4), (3), Karachi (5),
Abad (15) Karak (5) Peshawar (7), Quetta
(11)
(0.65,0.1,0.25) 50,657,260,000 672,462,400 0.52 88.4 62.1 75.2 Lahore (1), Rawalpindi (3), Sheikhupura (1), D. I Lahore (2), Rawalpindi
Sargodha (4), Hyderabad (5), Khan (2), Makran (4), (3), Karachi (5),
Bannu (8), Mardan (9), Kalat (12), Karak (5), Zhob (6) Peshawar (7), Quetta
Makran (13), M.Abad (15) (11)
13
Fig. 4. Optimum solutions for sustainable WAF-SCND at φTSC = 0.8 , φEI = 0.1, and φSI = 0.1.
Pakistan, the annual meat consumption of the country is 3.873 million 15,000, and 10,000 tons; biorefineries: 36, 27, and 18 million gallons;
tons, which indicates that WAF-based biodiesel has great potential to biodiesel storage and distribution center: 5 and 3 million gallons. All the
significantly decrease the dependency of the country on conventional considered potential locations and their capacity levels in WAF-SCND
fossil fuels. Furthermore, to reduce environmental pollution, the gov model are decided by fulfilling the industrial, environmental, and
ernment is encouraging the use of renewable energy resources, which regional laws. Three planning horizons of one year each is considered.
will also lead to achieving sustainable development goals. In this aspect,
this study provides an effective decision-making tool that will enable 5.1. Data gathering of WAF-SCND model input parameters
authorities to achieve the goal of sustainable biodiesel production. This
case study considers 36 WAF supply points, 18 possible locations of fat Data for the supply potential of chicken, mutton, and beef fat at each
preprocessing centers, 7 possible locations of biorefineries, 12 possible supply point are obtained from LDDB of Pakistan (www.lddb.org.pk).
locations of biodiesel distribution and storage hubs, 20 biodiesel de The most pragmatic values of biodiesel demand in each demand zone
mand zones, and 10 glycerol demand hubs. Following capacity cate are obtained using past three years data of fossil diesel demand. Based
gories are considered for each facility: fat preprocessing center: 25,000, on this estimation, the demand for B10 type blend (90% diesel, 10%
Fig. 5. Optimum solutions for sustainable WAF-SCND at φTSC = 0.65 , φEI = 0.1, and φSI = 0.25.
14
Table 6
Cost of protecting the environmental and social objectives under different combinations of priority weights.
Objective priority weights Priority weights of WAF-SCND model objective functions
(1,0,0) (0.8,0.1,0.1) (0.7,0.2,0.1) (0.7,0.1,0.2) (0.7,0.15,0.15) (0.65,0.25,0.1) (0.65,0.1,0.25)
Cost of environmental and social objectives (million $) 0 63,660 45,388 63,661 48,377 46,331 58,657
Increase in cost (%) 0 24.9 − 5.4 24.8 1.13 − 3.23 18.46
biodiesel) is derived (see Table C15) for each planning horizon. Facility and its final distribution to the customers. The activities in the given
installation, production, handling, feedstock purchase, and storage costs supply chain include transporting raw animal fat between fat collection
are decided based on the information obtained from local companies, points and preprocessing plants, handling of the feedstock at pre
previous feasibility studies, and published case studies. Transportation processing centers, extracting fat oil from raw animal fat, holding of fat
cost among biodiesel processing facilities is estimated by multiplying the oil, transporting fat oil between preprocessing plants and biorefineries,
unit transportation cost by their respective distances, which is decided, processing of fat oil to biodiesel, holding of biodiesel at biorefinery,
based on quotations obtained from regional logistics companies. To transporting biodiesel between biorefineries and distribution centers,
convert animal fat into fat oil, the densities of chicken, mutton, and beef transporting glycerol to demand zones, holding of biodiesel at distri
fat (φq) are assumed as 0.9171 g/cm3, 0.9188 g/cm3, and 0.9190 g/cm3, bution centers, and transporting biodiesel between distribution centers
respectively [22]. Conversion factors for chicken, mutton, and beef fat and demand zones. The mission of this WAF-SCND model is to produce
oil to biodiesel (ηq) are considered to be 0.952, 0.783, and 0.964, biodiesel so that the demand of each zone can be fulfilled. Therefore, the
respectively [6]. functional unit selected for this LCA study is to estimate the environ
To calculate the environmental impact of all operations involved in mental impact of one gallon of biodiesel delivered to the demand zone.
the WAF-SCND model, LCA is performed. The scope of this assessment is Further, in this research, to assess the environmental impact of activates
to estimate the environmental impact of all the associated activates associated with WAF-SCND model, EI99 methodology using ecoinvent
starting from the animal fat collection points to the biodiesel production version 3.6 database (www.pre-sustainability.com) of SimaPro 9.0, is
Fig 6. Required number of biodiesel processing facilities with their capacity levels for each considered scenario.
15
μ_TSC μ_EI μ_SI Minimum satisfaction level (ϴ)
Objective satisfaction level 95.0%
75.0%
55.0%
35.0%
15.0%
-5.0%
φEI 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 0 0.25 0.5 0 0.25
φTSC 0 0.25 0.5 0.75
WAF-SCND Objective priority weight
(a)

100.0%
Objective satisfaction level
80.0%
60.0%
40.0%
20.0%
φEI 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 0 0.25 0.5 0 0.25
φTSC 0 0.25 0.5 0.75
(b)
Fig. 7. Sensitivity analysis on priority weights of the WAF-SCND objectives for (a)℘ = 0, (b)℘ = 0.25, (c)℘ = 0.5, (d)℘ = 0.75, and (e)℘ = 1.0.
employed. Environmental impacts of some parameters such as handling development level (RDL) are obtained for each division from the
of WAF, animal fat oil production, and some others are not available in Pakistan Bureau of Statistics (www.pbs.gov.pk), whereas data for job
the database; therefore, such parameters are defined separately as a new opportunities potential for each type of biodiesel processing facility is
process. Further, it is assumed that to transport the WAF from fat supply decided after analyzing the relevant literature [10,47]. Due to space
points to the preprocessing centers, a homogeneous fleet of diesel trucks limitations, transportation costs among biodiesel processing facilities
(42 tons capacity each) is considered, while for the transportation of are not provided; the rest of the data with most likely values are pro
animal fat oil and biodiesel to the demand zones, a homogeneous fleet of vided in Appendix C. The ill-known input parameters of the WAF-SCND
trucks (1500 gallons capacity each) is used. The other input parameters model are considered fuzzy. Using the trapezoidal possibility distribu
of the WAF-SCND model do not have substantial differences to the tion, the fuzzy parameters are converted into trapezoidal fuzzy numbers.
conditions defined in SimaPro; hence, they are directly obtained from Each trapezoidal fuzzy number consists of four values that are obtained
the software database. Notably, the values of environmental impact using Eqs. (79)–(82).
obtained from the software may be tainted with human measurement
∂(1) = (1 − h1 )∂most (79)
errors and uncertainties. Although the software is also capable of dealing
with such uncertainties, the effect of uncertainty associated with envi
∂(2) = {1 − (h1 × h2 ) }∂most (80)
ronmental impact values is further minimized by considering them as
uncertain. The most pragmatic values of the environmental impact of
∂(3) = {1 + (h3 × h4 ) }∂most (81)
various processes are provided in Appendix C.
In this research, to assess the social dimension of the WAF-SCND
∂(4) = (1 + h4 )∂most (82)
model, the SLCAP is used. The proposed social impact assessment
objective mainly consists of two components: employment indicator and
economic development indicator. Equal importance is considered for
both components. Data on unemployment rate and regional
16

100.0%
80.0%
60.0%
40.0%
20.0%
φEI 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 0 0.25 0.5 0 0.25
φTSC 0 0.25 0.5 0.75
(c)

100.0%
80.0%
60.0%
40.0%
20.0%
φEI 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 0 0.25 0.5 0 0.25
φTSC 0 0.25 0.5 0.75
(d)

80.0%
70.0%
60.0%
50.0%
40.0%
30.0%
20.0%
φEI 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 0 0.25 0.5 0 0.25
φTSC 0 0.25 0.5 0.75
(e)
Fig. 7. (continued).
5.2. Results and discussion The extreme solutions presented in Table 4 indicate that all objec
tives, namely SC minimization, environmental impact minimization,
After coding the uncertain WAF-SCND into an equivalent crisp and social impact maximization, are contradictory in nature. For
formulation, the PIS and NIS are obtained. This WAF-SCND model has instance, at the positive ideal solution (PIS $29,893,180,000) of the SC
been coded in LINGO 18.0 on a PC with Intel Core i7-4790 and a 3.6 GHz cost minimization objective, the social impact objective holds a negative
processor. The model has 1285 constraints, 8219 continuous variables, ideal solution values of (NIS 0.165%). This is because the goal of min
and 53 binary variables. After solving the proposed model at ϖ, ϑ, θ = imum total SC system is achieved by opening least possible number of
0.85, and λ = 0.5, the extreme solutions are computed as listed in facilities with minimum capacity at economically suitable locations in
Table 4. each echelon. However, social aspect installs facilities with maximum
17
Table 7
Performance of the WAF-SCND optimization model under different scenarios.
Data set MRPCCP PCCP Cost of robustness
Cost Environmental Social Cost Environmental Social Economic Environmental Social

($) (pt) (%) ($) (pt) (%) ($) (pt) (%)
1 29,893,180,000 635,347,100 88.4 28,321,887,898 605,975,242 95.4 1,571,292,102 29,371,858 7.00

2 32,187,704,093 734,735,618 89.9 30,601,589,544 700,679,280 95.4 1,586,114,548 34,056,338 5.50
3 30,249,107,334 662,390,362 88.4 28,677,805,486 632,244,673 95.4 1,571,301,847 30,145,690 7.00
4 32,426,136,052 739,156,102 90 30,679,676,450 703,398,504 95.5 1,746,459,601 35,757,599 5.50
5 30,798,715,171 682,336,100 88.4 29,222,173,971 649,861,131 95.4 1,576,541,200 32,474,970 7.00
6 30,258,425,580 664,811,274 88.4 28,527,929,710 632,008,865 95.4 1,730,495,870 32,802,408 7.00
7 30,890,820,032 693,508,713 89 29,272,437,342 659,247,168 95.5 1,618,382,690 34,261,545 6.50
8 32,626,136,052 745,958,220 90 31,082,241,502 709,221,457 95.5 1,543,894,549 36,736,763 5.50
9 31,987,704,093 732,825,312 89.8 30,383,766,587 698,637,477 95.4 1,603,937,505 34,187,835 5.60
10 31,101,054,289 707,669,973 89.8 29,477,616,592 674,843,747 95.5 1,623,437,697 32,826,226 5.7
Mean 31,241,898,269 699,873,877 89.21 29,624,712,508 666,611,754 95.44 1,617,185,761 33,262,123 6.23
Standard deviation 942,374,973 36,297,343 0.71 941,938,489 34,369,934 0.05 64,782,333 2,159,962 0.69
number and capacity level in areas with low development, which is ⎧

completely against the economic perspective, which in turns increase ⎪
⎪
⎪
0, If fsi ⩽0.165
⎨
the total supply chain cost several times, thus, the priority of economic μsi (x) =
fsi − 0.165
(85)
, If 0.165 < fsi < 69.1662
objective and social objective are totally against each other. Further, it is ⎪
⎪
⎪ 69.1662 − 0.165
⎩
observed that at the PIS of environmental impact objective (PIS 1, If fsi ⩾69.1662
635,347,100), the value of social impact is quite low (5.59%) and total
The solution methodology of the WAF-SCND model is interactive in
SC cost ($89,269,310,000) is increased almost three times. This is due to
nature, which not only adds flexibility and robustness in the decision-
the fact that minimum environmental impact objective is achieved by
making process but also provides learning about the system to the SC
opening minimum number of facilities with least capacity level in major
managers. Hence, multiple efficient solutions based on the given cir
cities where biodiesel demands are higher so that transportation envi
cumstances can be obtained by varying the values of the confidence
ronmental impact can be minimized. Contrary to this, social aspect in
level of the uncertain constraints (ϖ, θ, and ϑ), uncertainty measure (λ),
stalls maximum number of facilities with maximum capacity level and
and coefficient of compensation (℘). The model results are illustrated in
tries to increase economic development indicator in low developed areas
Fig. 3, which are calculated at ϖ, θ, ϑ = 0.85, ν = 0.5, and℘ = 0.75. For
that are far away from the high biodiesel demand cities. As far as the
the above provided scenario, three objectives of WAF-SCND model, i.e.,
relation of total SC cost and environmental impact objective is con
TSC, total environmental impact, and social impact, are achieved as
cerned, initially, an inverse relation is seen between them. This inverse
85.58%, 69.03%, and 71.9%, respectively. Fig. 3 shows that among 15
relation is because of conflict in using biodiesel production technology
potential locations of animal fat preprocessing centers, five locations,
usage, where cheaper technology has higher environmental impact. On
namely Lahore, Sargodha, Bannu, Mardan, and Makran each with a
the other side, both the objectives share harmony (direct relation) in the
capacity level of 25,000 tons are chosen as optimal locations. Based on
aspect that both try to minimize the transportation distance among
the obtained results, the preprocessing center located in Lahore and
biodiesel processing facilities to achieve minimum transportation cost
Bannu has the highest amount of animal fat assigned in all planning
and environmental impact. Tradeoffs among the WAF-SCND model
horizons.
objectives are graphically depicted in Fig. 2.
After analyzing the results, it is found that owing to lower per ton fat
Then, using the PIS and NIS, the membership functions for each
preprocessing cost (economies of scale factor); the optimization model
objective are developed. Eqs. (83)–(85) provide the membership func
preferentially picks the centers with processing capacity level of 25,000
tions for the TSC, environmental impact, and social impact objectives,
over the centers having capacity level of 15,000 and 10,000 tons.
respectively.
Further, it is also observed that in each region, the fat preprocessing
centers near the highest feedstock supply potential are chosen so that the
total SC cost and total environmental impact during transportation can
be minimized. The model selects four biorefineries, which are estab
lished in Sheikhupura, DI Khan, Karak and Makran, among seven
⎧
⎪
⎪ 0, If ftsc ⩾1.84170 × 1011
⎪
⎪
⎨ 11
1.84170 × 10 − ftsc
μtsc (x) = , If 2.98931 × 1010 < ftsc < 1.84170 × 1011 (83)
⎪
⎪ 1.84170 × 10 11
− 2.98931 × 10 10
⎪
⎪
⎩
1, If ftsc ⩽2.98931 × 1010
⎧
⎪
⎪ 0, If fei ⩾7.0332×1008
⎪
⎪
⎨ 08
7.0332×10 − fei
μei (x)= , If 6.3534×1008 <fei <7.0332×1008 (84)
⎪
⎪ 7.0332×10 08
− 6.3534×1008
⎪
⎪
⎩
1, If fei ⩽6.3534×1008
18
potential locations. The required animal fat oil for Sheikhupura bio scale, the TSC is minimized. In contrast, if the combination of priority
refinery is mainly supplied from the preprocessing centers located in weights (0.65,0.1,0.25) is considered, the WAF-SCND model prefers to
Lahore and Sargodha. DI Khan and Karak biorefineries obtains their raw choose the maximum possible processing facilities with the highest job
fat oil from the Bannu preprocessing center. While raw animal fat oil for opportunity potential in regions with low RDL values at each stage so
Makran biorefinery is provided by its own fat preprocessing center. The that the social indicator of the SC can be maximized. A graphical
model selects the lowest biorefinery capacity level of 18 million gallons demonstration of the optimal decisions for the WAF-SCND model for the
for each operational location. This is due to the fact that among the three combination of priority weights (0.65,0.1,0.25) is displayed in Fig. 5.
types of facility installations: preprocessing centers, biorefinery, and
distribution center, the biorefinery installation have the highest eco 5.2.2. Impact of the biodiesel processing facilities capacity levels on WAF-
nomic value. Therefore, by choosing the lowest capacity level and SCND decisions
installing biorefinery in multiple regions having low RDL level will This section explores the impact of biodiesel processing facilities
alleviate the social aspect to a large extent. On the other side, biodiesel capacity levels (economies of scale), on the performance of WAF-SCND
produced in Sheikhupura is supplied to Lahore and Karachi distribution model. In the provided supply chain model, a compromise among eco
centers. Karak oil refinery sends its biodiesel to Rawalpindi and Pesha nomic, environmental, and social aspects is achieved by the strategic
war distribution centers. DI Khan biorefinery delivers its biodiesel to choices between SC configuration and production capacity levels of
Sukkur distribution center and Makran biorefinery provides biodiesel to biodiesel processing facilities. A crucial factor in a cost and environ
Quetta distribution center. In the detailed analysis, it is observed that mentally efficient SCND is to attain a balance between transportation
the results regarding biodiesel production technology indicate that all cost and economies of scale: whereas higher capacity levels of process
biorefineries prefer to use the second type of biodiesel production ing facilities (preprocessing plant, biorefinery, and distribution centers)
technology. Although this production technology imposes a little higher enable mangers to attain lower costs by the virtue of economies of scale,
per gallon biodiesel production cost, its low environmental impact and it increases the upstream transportation cost and environmental impact
higher social impact have made it more desirable. The model selects six by transporting the WAF-SCND resources (animal fat, fat oil, and bio
biodiesel storage and distribution centers that are built in Lahore, diesel) to large distances. In this context, the proposed optimization
Rawalpindi, Karachi, Sukkur, Peshawar, and Quetta, among 12 candi model provides an efficient trade-off between distributed SC and econ
date sites. From the obtained result, it is found that the optimization omies of scale aspect by simultaneously considering the large array of
model chooses the storage and distribution centers with higher capac possible SC configurations, feedstock supply, biodiesel demand, pro
ities for all the selected sites due to economies of scale. Moreover, the duction locations, and production scales. To possibly cover the impact of
storage and distribution centers are established near the biorefineries capacity levels on the WAF-SCND decision following five scenarios are
and demand zones, which reveals that the WAF-SCND model prefers to considered: (i) centralized only, (ii) decentralized only, (iii) equal
locate distribution and storage centers so that the environmental impact objective weightage, (iv) low feedstock supply, and (v) high biodiesel
and cost owing to long-distance transportation are diminished. demand. Fig. 6 provides a summary of required number of biodiesel
processing facilities with their capacity levels for each considered sce
5.2.1. Impact of objective priority weights on the WAF-SCND model nario. In the centralized supply chain scenario the major goal of the
decisions optimization model is to take the maximum advantage of economies of
To analyze the impact of objective priority weights on the decisions, scale. For this purpose, the model has selected the maximum possible
the proposed model is solved with different weight combinations. The capacity levels of biodiesel processing facilities wherever possible;
ranges of 0.65–1, 0–0.25, and 0–0.25 are selected as priority weights of however, smaller capacity levels are also selected for the locations
the SC cost, environmental impact, and social impact objectives. The where transportation costs among biodiesel processing facilities are
results are presented in Table 5, where it is evident that all three ob higher. Contrary to this, decentralized supply chain seeks to establish
jectives conflict with each other. biodiesel processing facilities with smaller capacity levels at multiple
Consequently, decision-makers have to pay higher costs to achieve locations and tries to minimize system cost by minimizing the trans
higher levels of social and environmental dimensions. Economic and portation cost. Decentralized SCs become more suitable at higher bio
environmental objectives share harmony up to some extent, but the diesel demands, as higher biodiesel production levels will gradually
maximization criteria for social objective is totally different. This trend diminish the difference between centralized and decentralized SCs.
can also be verified by comparing Figs. 4 and 5 graphically, where it is For the third scenario, where equal weightage for objectives (eco
revealed that to attain higher satisfaction levels of the social objective, nomic, environment, and social) is considered, WAF-SCND model tries
managers have to pay the cost of lower satisfaction levels for the envi to achieve a tradeoff by minimizing the total SC cost and setting up some
ronmental and economic objectives by opening many biodiesel pro of the processing facilities at most economically and environmentally
cessing facilities in underdeveloped areas. suitable locations with maximum capacity levels and other facilities in
The corresponding cost to attain the social and environmental goals the low RDLs. In this scenario, it is observed that to maximize the social
under different combinations of objective function priority weights are aspect in a balanced way the optimization model selects biodiesel pro
listed in Table 6. The protection cost varies from − 5.4% to 24.9% of the cessing facilities with lowest capacity levels in several regions having
optimum cost solution within the selected range. Further, the cost of low RDLs while to maximize the economic aspect it chooses facilities
achieving the environmental and social goals increases exponentially with highest capacity levels in most economically suitable locations. For
when higher priority weights are given to social impact aspects. the low feedstock supply scenario, where there is a high demand for
Furthermore, it is also seen that economic and environmental ob feedstock, optimization model builds minimum possible number of
jectives also share coordination to a large extent. This is because the processing facilities as a result the total cost of the system increases
transportation cost is directly proportional to the environmental impact marginally due to expensive feedstock and its higher transportation cost
of transportation. Moreover, a highly centralized SC network is devel among less number of biodiesel processing facilities with respect to the
oped when higher weights are assigned to the SC cost and environmental base scenario. Finally, for the last scenario, to fulfill the high biodiesel
impact objectives because a centralized network can achieve higher demands the optimization model installs the maximum number of bio
satisfaction levels for both objectives. For example, if the combination of diesel processing facilities with economically suitable capacity levels.
priority weights (1,0,0) is considered for the economic, environmental,
and social objectives, respectively, the WAF-SCND model selects the 5.2.3. Impact of the coefficient of compensation on WAF-SCND decisions
minimum possible number of biodiesel processing facilities with the and objective satisfaction levels
maximum capacity at each stage, hence, as a result of economies of Here, a sensitivity analysis is performed by changing the values of ℘,
19
φTSC , φEI , and φSI jointly, to reveal the influence of the aforementioned compromises.
parameters on the WAF-SCND model objective satisfaction levels.
Because the solution algorithm described in Section 4.3 is interactive in 5.2.5. Managerial insights
nature, several POSs can be obtained by varying the values of ℘, φTSC , This section provides the insights based on the key observations for
φEI , and φSI . Further, the values of these parameters are decided by the policy makers and managers dealing with the animal fat based biodiesel
managers considering real-time circumstances. Fig. 7a–e illustrate the SC within the paradigm of sustainability. Firstly, the managers must
graphical results of the sensitivity analysis. prefer to choose largest possible capacity level of biodiesel processing
Since the sum of objective priority weights φTSC , φEI , and φSI is 1, the facility (preprocessing plant, biorefinery, and distribution center) to
values of φTSC and φEI are provided in the graph. From the figures, it is benefit from economies of scale and enhanced economic value and
evident that the dispersion of the values of the objective satisfaction employment indicator, wherever possible. Secondly, while selecting the
levels and minimum satisfaction level (ɵ) is increasing from Fig. 7a–e. As potential locations for biorefineries manager must consider both the
the value of the coefficient of compensation increases, it starts to downstream and upstream of the supply chain. This consideration will
generate a balanced form of solution for the model objectives. This is help in achieving sustainability goals by minimizing both cost and
because at higher values of ℘ in the TH method, the priority of the environmental impact. Thirdly, if in the initial phase equal priority for
model is to augment the minimum satisfaction level for all objectives. sustainability dimensions is not feasible due to budget constraints, the
Consequently, the satisfaction levels of the WAF-SCND model have the tax credit policy (for tax credit policy see: [34]) can also be adopted by
highest dispersion in Fig. 7a and the least dispersion among μTSC, μEI, the decision makers to swiftly transform from lower levels of sustain
and μSI in Fig. 7e. ability to higher one. Results of the WAF-SCND model also provide in
sights to gain a better understanding of complex sustainable SC network.
5.2.4. Validation of the WAF-SCND optimization model and solution
methodology 6. Conclusions
To verify the robustness and effectiveness of the decisions obtained
using the WAF-SCND optimization model and MRPCCP solution meth This research introduced a multi-period WAF-SCND model for the
odology, nine random datasets are produced. For example, if f̃ = sustainable design of a biodiesel SC under epistemic uncertainty. The
(f1, f2, f3, f4 ) is a fuzzy parameter following a trapezoidal possibility dis proposed optimization model is among the first ones in the area of
tribution, then four points of the trapezoidal fuzzy number are obtained biofuel SC that explored the potential of WAF to produce biodiesel and
using Eqs. (79)–(82). Using the same methodology, trapezoidal fuzzy integrated related tactical and strategic level decisions in WAF-SCND so
numbers for all uncertain parameters of the optimization model are that an optimized tradeoff among the sustainability dimensions can be
generated and datasets are developed. Subsequently, each dataset is made. The latest environmental and social LCA-based techniques were
solved for the WAF-SCND model objectives using the proposed MRPCCP employed for obtaining the social and environmental impacts of the
solution methodology and the performance of the solutions is analyzed. WAF-SCND decisions. Because the business nature of the WAF-based
It is noteworthy that the values of the WAF-SCND model objectives are biodiesel SC is highly dynamic and particularly, the collected data is
obtained independently so that while solving the PCCP model and contaminated with epistemic uncertainty, an interactive solution
MRPCCP model, the penalty for the robustness can be estimated. For methodology named MRPCCP was presented and validated by imple
each of the realization, the results using the PCCP and MRPCCP solution menting it in a real case study.
methodologies are presented in Table 7. By analyzing the results, it is The results of the computational analysis exhibited that decisions of
evident that by bearing almost 5% increase in total SC cost, 6% increase WAF-SCND model change significantly when different aspects of sus
in environmental impact objectives and by compromising almost 7% for tainability are given priority. Major conclusions drawn from this
the social impact objective, the decisions for the proposed model can be research are:
secured against the associated epistemic uncertainty.
Furthermore, for each dataset, the standard deviation and average of 1. Total SC cost increases exponentially as higher goals for environ
the model objectives for both the PCCP and MRPCCP solution method mental and social objectives are set, that depicts a significant tradeoff
ologies are calculated. It is observed that the average and standard de among WAF-SCND model objectives.
viation values of the WAF-SCND objectives are in the acceptable range, 2. Facility installation cost and biodiesel production cost contributed
which also demonstrates the usability of the developed optimization the lion share to the total SC cost objective. Therefore, it is indis
model and solution methodology under operational uncertainty. pensable to search for more cost efficient technologies that would
In the obtained results, a constant penalty level for all datasets is substantially contribute to the commercial feasibility of biodiesel
considered; however, if the penalty level of the model chance constraint production.
violation is increased, the difference between the results of PCCP and 3. In the entire animal fat-based biodiesel SC, resource transportation
MRPCCP is increased exponentially. Additionally, with the increase in among processing facilities was the largest segment that was 56% to
penalty levels, the standard deviation of the results of both methodol the total environmental impact objective in which up-stream and
ogies also increases. As mentioned in Section 4.2, the MRPCCP solution down-stream of SC contributed 55% and 45%, respectively.
methodology uses the RPP-II formulation proposed by Pishvaee et al. 4. With a 5% increase in total SC cost objective, 6% increase in envi
[45]. The RPP-II variant decreases the deviation over the EV of the ronmental impact objective, and 7% increase in social impact
model objective function. According to Pishvaee et al. [45], the RPP-II objective, the decisions of the WAF-SCND model can be protected
formulation is extremely sensitive to the values of chance constraint from uncertain environment.
penalty violations. Hence, a higher penalty cost will increase the stan 5. With just an increase of 1.13% in total SC objective, the desirable
dard deviation of the results, which will make the RPP-II formulation- goal of social and environmental protection can be achieved.
based MRPCCP methodology unsuitable for real-time implementation. 6. Sensitivity analysis of the objective priority weights demonstrated
For such situations, decision-makers must adopt risk-averse approaches, that higher priority of economic and environmental objectives makes
and the MRPCCP solution methodology must be developed based on the SC decisions more centralized, whereas higher priority of the
more conservative formulations such as Hard Worst RPP and RPP-III. social objective makes them more decentralized.
Thus, it is concluded that the MRPCCP approach is highly suitable for
uncertain environments and allows biodiesel SC managers to make Being a preliminary study in the area of WAF-based biodiesel pro
robust decisions with minor economic, environmental, and social duction modelling, extension of this research is possible in several di
rections. For example, the proposed WAF-SCND model can be extended
20
using the coordination concept, where a subsidized price can be offered & editing. Biswajit Sarkar: Validation, Supervision, Writing - review &
to the suppliers of WAF. Moreover, optimization of the regional logistics editing, Funding acquisition.
of raw animal fat from slaughterhouses to collection centers will make
this study more realistic. In this study, operational risk was efficiently
incorporated by using RPP techniques; however, the aspect of disruption Declaration of Competing Interest
risk was not considered. Consideration of operational risk and disruption
risk simultaneously will make this research more practical. (see Jab The authors declare that they have no known competing financial
barzadeh et al. [58] and Samani et al. [59]). interests or personal relationships that could have appeared to influence
the work reported in this paper.
CRediT authorship contribution statement
Acknowledgement
Muhammad Salman Habib: Conceptualization, Methodology,
Software, Writing - original draft, Writing - review & editing. Muham The work is supported by the National Research Foundation of Korea
mad Tayyab: Resources, Formal analysis, Investigation, Writing - re (NRF) grant funded by the Korea Government (MSIT)
view & editing. Sadaf Zahoor: Software, Data curation, Writing - review (NRF2020R1F1A1064460).
Appendix A
Notation
Indices
i index of fat supply points
l index of fat collection and preprocessing centers
q index of animal fat type
o index of biorefineries for biodiesel production
p index of biodiesel distribution and storage centers
d index of biodiesel demand zone
g index of glycerol demand zone
r index of available biodiesel production technologies
s index of preprocessing centers capacity level
a index of biorefinery capacity level
b index of biodiesel distribution and storage center capacity level
t index of planning periods
Parameters Economic objective parameters
feconomic total biodiesel SC cost
Costopen total cost of installing biodiesel processing facilities
Costfeed total cost of animal fat feedstock purchase
Costhand total cost of handling animal fat
Costinv total cost of inventory holding during various stages of the biodiesel SC
Costprod total cost of producing animal fat oil, biodiesel, and glycerol in the biodiesel SC
Costtrans total cost of transporting animal fat, fat oil, biodiesel, and glycerol among biodiesel processing facilities
PP̃Cls cost of opening a fat preprocessing plant in region l having capacity level s
BĨRora cost of opening a biorefinery for biodiesel production in region o having technology r having capacity level a
BD̃Dpb cost of opening biodiesel distribution and storage centers in region p having capacity level b
FP̃Ciqt animal fat purchasing cost at supply point i for animal fat type q in period t
HP̃PClq handling cost of animal fat at preprocessing plant l for animal fat type q
IH̃Plqt cost of holding inventory of animal fat type q at preprocessing plant l in period t
IH̃Bort cost of holding biodiesel inventory in biorefinery o having technology r in period t
IH̃Gort cost of holding inventory of glycerol in biorefinery o having technology r in period t
IH̃Dpt cost of holding biodiesel inventory at distribution and storage center p in period t
CP̃PClqt cost of fat oil production at collection and preprocessing plant l for fat type q in period t
CB̃Port cost of biodiesel production using technology r at biorefinery o in period t
CG̃Port cost of glycerol production using technology r at biorefinery o in period t
CC̃Tilq cost of transporting fat type q from animal fat supply point i to preprocessing center l
CC̃OTloq cost of transporting fat oil type q from preprocessing center l to biorefinery o
CB̃DTop cost of transporting biodiesel from biorefinery o to distribution and storage center p
CB̃DTpd cost of transporting biodiesel from distribution and storage center p to demand zone d
CG̃Tog cost of transporting glycerol from biorefinery o to demand zone g
Environmental impact objective parameters
fenvI total environmental effect of biodiesel SC
EnvIopen total environmental effect of establishing biodiesel SC processing facilities
EnvIhand total environmental impact of handling feedstock at biodiesel processing facilities
EnvIinv total environmental impact of feedstock, fat oil, and biodiesel inventory holding at biodiesel processing centers
EnvIprod total environmental impact of fat oil, biodiesel, and glycerol production at biodiesel processing centers
EnvItrans total environmental impact of transporting feedstock, fat oil, biodiesel, and glycerol among processing facilities
EPC̃Hls environmental impact of establishing an animal fat collection and preprocessing center at location l with capacity level s
EB̃IRora environmental impact of establishing a biorefinery at location o using technology r with capacity level a
EB̃DDpb environmental impact of establishing a biodiesel distribution and storage center at location p with capacity level b
EHP̃PClq environmental impact of handling animal fat type q at fat collection and preprocessing center l
(continued on next page)
21
(continued )
EĨHPlqt environmental impact of holding animal fat type q at preprocessing center l in period t
EĨHBort environmental effect of holding biodiesel at biorefinery o produced using technology r in period t
EĨHGort environmental impact of holding glycerol at biorefinery o produced using technology r in period t
EĨHDpt environmental impact of holding biodiesel at distribution and storage center p in period t
EP̃PClqt environmental impact of producing animal fat oil at preprocessing center l for fat type q in period t
EB̃Port environmental impact of producing biodiesel at biorefinery o using technology r in period t
EG̃Port environmental impact of producing glycerol at biorefinery o using technology r in period t
EC̃Tilq environmental effect of transporting animal fat type q from fat supply point i to preprocessing plant in region l
EC̃OTloq environmental effect of transporting animal fat oil type q from a preprocessing plant in region l to biorefinery o
EB̃DTop environmental effect of transferring biodiesel from biorefinery o to distribution and storage center p
EB̃DTpd environmental effect of transferring biodiesel from distribution and storage center p to biodiesel demand zone d
EG̃Tog environmental effect of transferring glycerol from biorefinery o to glycerol demand zone g
Social objective parameters
fsocial total social responsibility of the biodiesel SC
ŨEl percentage unemployment rate in region l
ŨEO percentage unemployment rate in region o
ŨEp percentage unemployment rate in region p
J̃ls jobs created by opening fat collection and preprocessing center l having capacity level s
J̃ora jobs created by opening biorefinery o using technology r having capacity level a
J̃pb jobs created by opening distribution and storage center p having capacity level b
EṼls economic value of establishing fat collection and preprocessing center l with capacity level s
EṼora economic value of establishing biorefinery l using technology r with capacity level a
EṼpb economic value of establishing distribution and storage center p with capacity level b
RDLl development level of region l
RDLo development level of region o
RDLp development level of region p
ED economic development indicator
EDmax maximum value of economic development indicator
EDmin minimum value of economic development indicator
EI employment indicator
EImin minimum value of employment indicator
EImax maximum value of employment indicator
ϒ importance weight of economic development indicator
ζ importance weight of employment indicator
Constraint parameters
μ̃dt demand for biodiesel at demand zone d in period t
μ̃gt demand for glycerol at demand zone g in period t
σ̃iqt quantity available for fat type q at supply point i in period t
φq conversion factor of animal fat type q
ηq yield factor of animal fat oil type q
Capls maximum processing capacity of fat collection and preprocessing center l with level s
Capoa production capacity at biorefinery o with level a
Cappb storage capacity of biodiesel distribution center p at capacity level b
Decision variables
IFOlqt inventory level of animal fat oil type q at fat preprocessing plant l in period t
IBort inventory of biodiesel produced in period t at biorefinery o using technology r
IGort Inventory level of glycerol produced in period t at biorefinery o using technology r
IDpt inventory of biodiesel in period t at distribution and storage center p
πilqt amount of fat type q transported from animal fat supply point i to preprocessing center l during period t
Φlqt amount of animal fat oil of type q produced at preprocessing plant l in period t
Ψort volume of biodiesel produced using technology r at biorefinery o in period t
Πort volume of glycerol produced using technology r at biorefinery o in period t
πloqt quantity of animal fat oil type q transferred from preprocessing plant l to biorefinery o during period t
πoprt volume of biodiesel produced using technology r and transferred from biorefinery o to distribution and storage center p during period t
πpdt volume of biodiesel transferred from distribution and storage center p to biodiesel demand zone d during period t
πogrt volume of glycerol produced using technology r transferred from biorefinery o to glycerol demand zone d during period t
1 if fat preprocessing center l with capacity level s is operational; 0 otherwise
uls =
{
0
1
{
0 1 if biorefinery o with biodiesel produciton technology r having capacity level a is operational; 0 otherwise
xora =
1
1 if biodiesel distribution and storage center p with capacity level b; 0 otherwise
ypb =
{
0
1
22
Appendix B
Formulation of the animal fat-based biodiesel SC model into an MRPCCP model
The solution methodology for the proposed WAF-SCND model consists of the following two phases.
Phase 1: Conversion of an uncertain animal fat-based-biodiesel SC model into a PCCP model
Cost objective function crisp formulation
pccp
E[feconomic pccp
] = Costopen pccp
+ Costfeed pccp
+ Costhand pccp
+ Costinv pccp
+ Costprod pccp
+ Costtrans (B-1)
Fixed facility opening costs

∑L ∑ S [ ]
pccp 1− λ λ
Costopen = (PPCls(1) + PPCls(2) ) + (PPCls(3) + PPCls(4) ) uls +
l s
2 2
∑
O ∑ A [
R ∑ ]
1− λ λ
(BIRora(1) + BIRora(2) ) + (BIRora(3) + BIRora(4) ) xora + (B-2)
o r a
2 2
∑ B [
P ∑ ]
1 − λ( ) λ( )
BDDpb(1) + BDDpb(2) + BDDpb(3) + BDDpb(4) ypb
p b
2 2
Feedstock purchase costs and handling costs

I ∑
∑ L ∑Q ∑ T [ ]
pccp pccp 1 − λ( ) λ( )
Costfeed + Costhand = FPCiqt(1) + FPCiqt(2) + FPCiqt(3) + FPCiqt(4) π ilqt +
i l q t
2 2
(B-3)
I ∑
∑ L ∑ T [
Q ∑ ]
1 − λ( ) λ( )
HPPClq(1) + HPPClq(2) + HPPClq(3) + HPPClq(4) πilqt
i l q t
2 2
Inventory holding cost

L ∑
∑ Q ∑T [ ]
pccp 1 − λ( ) λ( )
Costinv = IHPlqt(1) + IHPlqt(2) + IHPlqt(3) + IHPlqt(4) IFlqt
l q t
2 2
∑
O ∑ T [
R ∑ ]
1− λ λ
(IHBort(1) + IHBort(2) ) + (IHBort(3) + IHBort(4) ) IBort +
o r t
2 2
(B-4)
∑
O ∑ T [
R ∑ ]
1− λ λ
(IHGort(1) + IHGort(2) ) + (IHGort(3) + IHGort(4) ) IGort +
o r t
2 2
∑ T [
P ∑ ]
1 − λ( ) λ( )
IHDpt(1) + IHDpt(2) + IHDpt(3) + IHDpt(4) IDpt
p t
2 2
Chicken oil, mutton oil, beef oil, biodiesel, and glycerol production cost
∑
L ∑Q ∑ T [ ]
pccp 1 − λ( ) λ( )
Costprod = CPPClqt(1) + CPPClqt(2) + CPPClqt(3) + CPPClqt(4) Φlqt +
l q t
2 2
∑
O ∑ T [
R ∑ ]
1− λ λ
(CBPort(1) + CBPort(2) ) + (CBPort(3) + CBPort(4) ) Ψort + (B-5)
o r t
2 2
∑
O ∑ T [
R ∑ ]
1− λ λ
(CGPort(1) + CGPort(2) ) + (CGPort(3) + CGPort(4) ) Πort
o r t
2 2
Transportation cost
23
I ∑
∑ L ∑ T [
Q ∑ ]
pccp 1 − λ( ) λ( )
Costtrans = CCTilq(1) + CCTilq(2) + CCTilq(3) + CCTilq(4) πilqt +
i l q t
2 2
∑
L ∑
O ∑ T [
Q ∑ ]
1 − λ( ) λ( )
CCOTloq(1) + CCOTloq(2) + CCOTloq(3) + CCOTloq(4) πloqt +
l o q t
2 2
∑
O ∑
P ∑ T [
R ∑ ]
1 − λ( ) λ( )
CBDTop(1) + CBDTop(2) + CBDTop(3) + CBDTop(4) πoprt + (B-6)
o p r t
2 2
∑
P ∑ T [
D ∑ ]
1 − λ( ) λ( )
CBDTpd(1) + CBDTpd(2) + CBDTpd(3) + CBDTpd(4) πpdt +
p d t
2 2
O ∑
∑ G ∑ T [
R ∑ ]
1 − λ( ) λ( )
CGTog(1) + CGTog(2) + CGTog(3) + CGTog(4) πogrt
o g r t
2 2
Environmental objective function
pccp
E[fenvI pccp
] = EnvIopen pccp
+ EnvIhand pccp
+ EnvIinv pccp
+ EnvIprod pccp
+ EnvItrans (B-7)
Environmental impact of establishing biodiesel SC facilities

S [
L ∑
∑ ]
pccp 1− λ λ
EnvIopen = (EPCHls(1) + EPCHls(2) ) + (EPCHls(3) + EPCHls(4) ) uls +
l s
2 2
O ∑
∑ A [
R ∑ ]
1− λ λ
(EBIRora(1) + EBIRora(2) ) + (EBIRora(3) + EBIRora(4) ) xora + (B-8)
o r a
2 2
∑ B [
P ∑ ]
1 − λ( ) λ( )
EBDDpb(1) + EBDDpb(2) + EBDDpb(3) + EBDDpb(4) ypb
p b
2 2
Environmental impact of feedstock handling

I ∑
∑ L ∑ Q ∑T [ ]
1 − λ( ) λ( )
pccp
EnvIhand = EHPPClq(1) + EHPPClq(2) + EHPPClq(3) + EHPPClq(4) πilqt (B-9)
i l q t
2 2
Environmental impact of holding inventory at different stages of the biodiesel SC

L ∑
∑ Q ∑ T [ ]
pccp 1 − λ( ) λ( )
EnvIinv = EIHPlqt(1) + EIHPlqt(2) + EIHPlqt(3) + EIHPlqt(4) IFlqt
l q t
2 2
∑
O ∑ T [
R ∑ ]
1− λ λ
(EIHBort(1) + EIHBort(2) ) + (EIHBort(3) + EIHBort(4) ) IBort +
o r t
2 2
(B-10)
∑
O ∑ T [
R ∑ ]
1− λ λ
(EIHGort(1) + EIHGort(2) ) + (EIHGort(3) + EIHGort(4) ) IGort +
o r t
2 2
∑ T [
P ∑ ]
1 − λ( ) λ( )
EIHDpt(1) + EIHDpt(2) + EIHDpt(3) + EIHDpt(4) IDpt
p t
2 2
Environmental impact of biodiesel and glycerol production

Q ∑
L ∑
∑ T [ ]
pccp 1 − λ( ) λ( )
EnvIprod = EPPClqt(1) + EPPClqt(2) + EPPClqt(3) + EPPClqt(4) Φlqt +
l q t
2 2
∑
O ∑ T [
R ∑ ]
1− λ λ
(EBPort(1) + EBPort(2) ) + (EBPort(3) + EBPort(4) ) Ψort + (B-11)
o r t
2 2
∑
O ∑ T [
R ∑ ]
1− λ λ
(EGPort(1) + EGPort(2) ) + (EGPort(3) + EGPort(4) ) Πort
o r t
2 2
Environmental impact of transporting material among SC facilities
24
I ∑
∑ L ∑ T [
Q ∑ ]
pccp 1 − λ( ) λ( )
EnvItrans = ECTilq(1) + ECTilq(2) + ECTilq(3) + ECTilq(4) πilqt +
i l q t
2 2
∑
L ∑
O ∑ T [
Q ∑ ]
1 − λ( ) λ( )
ECOTloq(1) + ECOTloq(2) + ECOTloq(3) + ECOTloq(4) πloqt +
l o q t
2 2
∑
O ∑
P ∑ R [
T ∑ ]
1 − λ( ) λ( )
EBDTop(1) + EBDTop(2) + EBDTop(3) + EBDTop(4) πoprt + (B-12)
o p t r
2 2
∑
P ∑ T [
D ∑ ]
1 − λ( ) λ( )
EBDTpd(1) + EBDTpd(2) + EBDTpd(3) + EBDTpd(4) πpdt +
p d t
2 2
O ∑
∑ G ∑ R [
T ∑ ]
1 − λ( ) λ( )
EGTog(1) + EGTog(2) + EGTog(3) + EGTog(4) π ogrt
o g t r
2 2
Social objective function
pccp
E[fsocial ] = ϒEDpccp pccp
norm + ζEInorm (B-13)
EDpccp − EDpccp
EDpccp
norm =
min
(B-14)
EDpccp pccp
max − EDmin
pccp
EI pccp − EImin
pccp
EInorm = pccp pccp (B-15)
EImax − EImin
S [(
L ∑
∑ )( )]
1− λ λ 1− λ λ
EI pccp = (UEl(1) + UEl(2) ) + (UEl(3) + UEl(4) ) (Jls(1) + Jls(2) ) + (Jls(3) + Jls(4) ) uls +
l s
2 2 2 2
∑
O ∑ A [(
R ∑ )( )]
1− λ λ 1− λ λ
(UEo(1) + UEo(2) ) + (UEo(3) + UEo(4) ) (Jora(1) + Jora(2) ) + (Jora(3) + Jora(4) ) xora + (B-16)
o r a
2 2 2 2
B [(
P ∑
∑ )( )]
1 − λ( ) λ( ) 1− λ( ) λ( )
UEp(1) + UEp(2) + UEp(3) + UEp(4) Jpb(1) + Jpb(2) + Jpb(3) + Jpb(4) ypb
p b
2 2 2 2
S [
L ∑
∑ ]
1− λ λ
EDpccp = (EVls(1) + EVls(2) ) + (EVls(3) + EVls(4) ) uls (1 − RDLl )+
l s
2 2
∑
O ∑ A [
R ∑ ]
1− λ λ
(EVora(1) + EVora(2) ) + (EVora(3) + EVora(4) ) xora (1 − RDLo )+ (B-17)
o r a
2 2
∑ B [
P ∑ ]
1 − λ( ) λ( )
EVpb(1) + EVpb(2) + EVpb(3) + EVpb(4) ypb (1 − RDLp )
p b
2 2
⎡⎧ ∑ L ∑ S [ ] ⎫ ⎤
⎪ 1− λ λ ⎪
⎪
⎪ (EV ls(1) + EV ls(2) ) + (EV ls(3) + EV ls(4) ) uls (1 − RDLl )+ ⎪
⎪
⎢⎪ 2 2 ⎪ ⎥
⎢⎪⎪
⎪ l s ⎪
⎪
⎪ ⎥
⎢⎪ ⎪ ⎥
⎢⎪⎪ O ∑ A [ ] ⎪
⎪ ⎥
⎢⎨ ∑ R ∑
1− λ λ ⎬ ⎥
pccp ⎥
⎢ (EVora(1) + EVora(2) ) + (EVora(3) + EVora(4) ) xora (1 − RDLo )+ − EDmin ⎥
⎢⎪ 2 2 ⎪
⎢⎪ ⎪ ⎥
⎢⎪ ⎪
o r a
⎪ ⎪ ⎥
⎢⎪⎪
⎪ [ ] ⎪
⎪
⎪ ⎥
⎣⎪⎪ ∑
P ∑ B
1 − λ( ) λ( ) ( ) ⎪
⎪ ⎦
⎪
⎩ EVpb(1) + EVpb(2) + EVpb(3) + EVpb(4) ypb 1 − RDLp ⎪
⎭
2 2
(B-18)
p b
EDpccp
norm = ϒ ×
EDpccp pccp
max − EDmin
25
S [(
L ∑
∑ )( )]
1− λ λ 1− λ λ
(UEl(1) + UEl(2) ) + (UEl(3) + UEl(4) ) (Jls(1) + Jls(2) ) + (Jls(3) + Jls(4) ) uls +
l s
2 2 2 2
∑
O ∑ A [(
R ∑ )( )]
1− λ λ 1− λ λ
(UEo(1) + UEo(2) ) + (UEo(3) + UEo(4) ) (Jora(1) + Jora(2) ) + (Jora(3) + Jora(4) ) xora +
o r a
2 2 2 2
∑ B [(
P ∑
1 − λ( ) λ(
⎫ ⎤
UEp(1) + UEp(2) + UE ⎛
2 2 ⎪
⎪
⎡⎧
p b ⎪
⎪ ⎥
⎜ ⎪
⎪ ⎥
⎪
⎪ ⎜ ⎪
⎪ ⎥
⎢⎪⎪ ⎜ ⎪
⎪ ⎥
⎢⎪⎪ ⎜ ⎪
⎪ ⎥
⎢⎪⎪ ⎜ )( ) ] ⎪
⎬ ⎥
⎢⎪⎪ ⎜ ) 1 − λ ( ) λ ( ) pccp ⎥
⎢⎨ ⎜
⎜3)+UEp(4) Jpb(1) + Jpb(2) + Jpb(3) + Jpb(4) ypb − EImin ⎥
pccp
EInorm ⎢
=ζ×⎢ ⎜ 2 2 ⎪
⎪
⎪
⎪
⎥
⎥ (B-19)
⎢⎪⎪ ⎜ ⎪
⎪ ⎥
⎢⎪⎪ ⎜ ⎪
⎪ ⎥
⎢⎪⎪ ⎜ ⎪
⎪ ⎥
⎣⎪⎪ ⎜ ⎪
⎪ ⎦
⎪
⎪ ⎜ ⎪
⎭
⎩ ⎜
⎜
p⎜ pccp pccp
⎜ EImax − EI
min
⎜
⎜
⎜
⎜
⎜
⎜
⎜
⎜
⎜
⎜
⎜
⎜
⎜
⎝
subject to
∑
L
(ϖ − λ)σ iqt(1) + (1 − ϖ)σiqt(2)
πilqt ⩽ ∀i, q, t (B-20)
l
1− λ
∑
P
(θ − λ)μdt(4) + (1 − θ)μdt(3)
πpdt ⩾ ∀ d, t (B-21)
p
1− λ
∑
O ∑
R
(ϑ − λ)μgt(4) + (1 − ϑ)μgt(3)
πogrt ⩾ ∀ g, t (B-22)
o r
1− λ
and constraints (23)–(49).

In this phase, the animal fat-based biodiesel SCND model provided in Eqs. (1)–(49) is converted into a PCCP model using the EV operator and Me
measure given by Eq. (67) and Eqs. (68)–(69), respectively. A crisp PCCP-based counterpart formulation of animal fat-based biodiesel SC model is
presented in Eqs. (B-1)–(B-22).
Phase 2: Conversion of the PCCP-based animal fat biodiesel SC model into an MRPCCP model
In the second phase, the PCCP formulation of the animal fat-based biodiesel SCND model is converted into an MRPCCP model using Eq. (71). For
this purpose, a new objective function containing optimality robustness and feasibility robustness terms is formulated. The MRPCCP formulation of the
studied model is indicated in Eqs. (B-23)–(B-25).
[ pccp ] ( [ pccp ] )
pccp
Minimize E feco + κtc fmax, eco − E feco +
∑
I ∑ T [
Q ∑ ]
(ϖ − λ)σiqt(1) + (1 − ϖ)σ iqt(2)
φfeed
sup,eco − σiqt(1) +
i q t
1− λ
T [ ] (B-23)
∑
D ∑
(θ − λ)μdt(4) + (1 − θ)μdt(3)
φbio
dem,eco − μdt(4) +
d t
1− λ
∑ T [
G ∑ ]
(ϑ − λ)μgt(4) + (1 − ϑ)μgt(3)
φgly
dem,eco − μgt(4)
g t
1− λ
26
pccp
(pccp pccp
)
Minimize E[fenvI ] + κenv fmax, envI − E[fenvI ] +
∑
I ∑ T [
Q ∑ ]
(ϖ − λ)σiqt(1) + (1 − ϖ)σ iqt(2)
φfeed
sup,envI − σ iqt(1) +
i q t
1− λ
∑ T [
D ∑
(θ − λ)μdt(4) + (1 − θ)μdt(3)
] (B-24)
φbio
dem,envI − μdt(4) +
d t
1− λ
∑ T [
G ∑ ]
(ϑ − λ)μgt(4) + (1 − ϑ)μgt(3)
φgly
dem,envI − μgt(4)
g t
1− λ
Table C1
Installation cost of animal fat preprocessing, biorefinery, and storage and distribution facilities ($).
Potential animal fat preprocessing facilities Potential biorefinery facilities Potential storage and distribution facilities
l Division Value o Division Value (r1) (million) Value (r2) (million) p Division Value (b1) Value (b2)
1 Lahore (LHE) 6,900,000 1 Sheikhupura 3200 5700 1 Bahawalpur 1,200,000 1,680,000

2 Multan (MLT) 4,000,000 2 D. I Khan 1500 2600 2 Lahore 4,500,000 6,300,000
3 Rawalpindi (RWP) 5,800,000 3 Karachi 2900 5200 3 Rawalpindi 2,700,000 3,780,000
4 Sargodha (SGD) 5,300,000 4 Makran 1300 2100 4 Sargodha 1,300,000 1,820,000
5 Hyderabad (HDD) 6,400,000 5 Karak 1400 2300 5 Karachi 4,500,000 6,300,000
6 Larkana (LK) 4,100,000 6 Zhob 1100 2000 6 Sukkur 2,400,000 3,360,000
7 Sukkur (SUK) 5,600,000 7 Muzzafarabad 2100 4000 7 Peshawar 2,700,000 3,780,000
8 Bannu (BNU) 3,900,000 8 Mansehra 2,600,000 3,640,000
9 Mardan (MD) 4,200,000 9 Sawat 1,000,000 1,400,000
10 Baltistan (BTT) 6,500,000 10 Gilgit 2,100,000 2,940,000
11 Diamer (DM) 4,100,000 11 Quetta 2,100,000 2,940,000
12 Kalat 4,300,000 12 Muzaffarabad 2,800,000 3,920,000
13 Makran 5,100,000
14 Sibi 5,500,000
15 Muzaffarabad (MFG) 5,700,000
Table C2
Waste animal fat purchase costs and handling costs.
q Feedstock handling cost ($/ton) Feedstock purchase cost ($/ton)
Chicken fat 200 300

Mutton fat 250 400
Beef fat 300 400
Table C3
Animal fat oil, biodiesel, and glycerol production cost.
q Animal fat oil production cost Biodiesel production cost at Glycerol production cost at
at preprocessing center ($/gallon) biorefinery ($/gallon) biorefinery ($/gallon)
Chicken fat 1.60 4.1 3.0

Mutton fat 1.70 4.4 3.5
Beef fat 1.85 4.5 3.5
Table C4
Other cost input parameters of the proposed model.
Parameter definition Value
Cost of transferring animal fat from fat supply point i to fat preprocessing center l ($/ton-km) 1.5
Cost of transferring animal fat oil from fat preprocessing plant l to biorefinery o ($/gallon-km) 0.5
Cost of transferring biodiesel from biorefinery o to storage and distribution center p ($/gallon-km)
Cost of transferring biodiesel from distribution and storage center p to biodiesel demand zone d ($/gallon-km)
Cost of transferring glycerol from biorefinery o to glycerol demand zone g ($/gallon-km) 0.7
27
Table C5
Environmental impact of installing animal fat preprocessing, biorefinery, and storage and distribution facilities (pt).
l Division Value o Division Value (r1) (thousands) Value (r2) (thousands) p Division Value (b1) Value (b2)
1 Lahore 20,000 1 Sheikhupura 26,110 17,407 1 Bahawalpur 5800 7500

2 Multan 17,000 2 D. I Khan 11,868 7912 2 Lahore 24,200 31,500
3 Rawalpindi 16,500 3 Karachi 23,737 15,824 3 Rawalpindi 14,500 18,900
4 Sargodha 14,500 4 Makran 10,681 7121 4 Sargodha 4800 6300
5 Hyderabad 20,000 5 Karak 11,868 7912 5 Karachi 24,200 31,500
6 Larkana 10,500 6 Zhob 9494 6329 6 Sukkur 12,100 15,700
7 Sukkur 16,500 7 Muzzafarabad 17,802 11,868 7 Peshawar 14,500 18,900
8 Bannu 10,500 8 Mansehra 14,500 18,900
9 Mardan 11,000 9 Sawat 4800 6300
10 Baltistan 19,000 10 Gilgit 9700 12,600
11 Diamer 11,000 11 Quetta 9700 12,600
12 Kalat 11,000 12 Muzaffarabad 14,500 18,900
13 Makran 13,500
14 Sibi 15,000
15 Muzaffarabad 16,500
Table C6
Environmental impact of animal fat oil, biodiesel, and glycerol production cost.
q Environmental effect of animal fat oil Environmental effect of biodiesel Environmental effect of glycerol
production at preprocessing center (pt/gallon) production at biorefinery (pt/gallon) production at biorefinery (pt/gallon)
Chicken fat 5.00 2.093 2.097

Mutton fat 6.44
Beef fat 6.44
Table C7
Environmental impact input parameters of the proposed model.
Parameter definition Value
Environmental impact of handling animal fat at preprocessing center l (for all q) (pt/ton) 57
Environmental impact of transporting animal fat from fat supply point i to fat preprocessing center l (pt/ton-km) 0.00469
Environmental impact of transporting animal fat oil from fat preprocessing center l to biorefinery o (pt/gallon-km) 0.000016
Environmental effect of moving biodiesel from biorefinery o to storage and distribution center p (pt/gallon-km)
Environmental impact of transporting animal fat oil from storage and distribution center p to biodiesel demand zone d (pt/gallon-km)
Environmental impact of transporting glycerol from biorefinery o to glycerol demand zone g (pt/gallon-km)
Table C8
Density of fat oil type q and conversion factor of fat oil to biodiesel.
q ηq (%) Density (g/cm3)
Chicken fat 95.2 0.917

Mutton fat 78.3 0.918
Beef fat 96.4 0.919
28
Table C9
Quantity of animal fat feedstocks available at fat supply points during each period (tons).
i q = 1 (chicken fat) q = 2 (mutton fat) q = 3 (beef fat)
t=1 t=2 t=3 t=1 t=2 t=3 t=1 t=2 t=3
1 525 450 675 795 1005 960 1425 1920 1575

2 450 390 540 1275 1500 1260 1125 1425 1455
3 795 630 840 2625 2925 2775 3225 4275 2925
4 975 960 1200 1125 2010 1260 3420 4455 3675
5 1125 1110 975 2250 3450 2783 4575 6825 4470
6 1050 885 900 1725 1950 1650 2775 3825 2925
7 675 660 555 1875 3030 2385 4350 5970 4440
8 600 615 450 2970 4125 3450 2640 3735 3030
9 750 795 600 1875 1980 2040 2850 3885 2325
10 945 795 930 2175 2220 1800 4020 4350 3735
11 300 330 375 1575 1875 1545 1920 2910 2535
12 915 758 960 1125 1800 1575 2865 3225 2595
13 1380 1200 1440 2850 3225 3075 5325 7275 5610
14 435 540 405 900 1290 1050 1800 2505 1920
15 570 480 525 1065 1215 1125 2010 2355 2250
16 390 398 375 1110 1290 1125 1920 2250 1545
17 562 390 615 900 1095 675 1740 2100 1500
18 585 510 600 1350 1440 1425 2160 2790 2250
19 450 555 405 960 1110 810 1920 2355 1725
20 375 405 390 1875 2175 1875 1440 2040 2190
21 630 675 660 1725 2325 2175 2700 2940 1740
22 390 375 450 825 1275 1095 1350 2250 2490
23 705 555 720 1425 1800 1575 2955 3525 3720
24 1020 757 1050 3075 4575 3225 4455 6000 3885
25 555 405 690 2175 2925 2025 1200 2250 1545
26 735 675 840 825 1275 975 900 1950 1545
27 420 375 450 825 1125 975 1500 1800 1560
28 270 360 255 750 1200 960 1410 2325 1875
29 315 300 285 810 1110 885 975 1800 1650
30 435 375 360 1125 1425 975 1545 2595 1995
31 930 885 975 2325 3075 2475 3825 5325 4425
32 712 690 660 675 945 690 960 1800 1200
33 450 405 465 795 1320 1050 1875 2550 1800
34 540 510 510 660 945 750 1350 1815 1530
35 420 390 495 975 1275 1125 1410 2475 2775
36 840 795 975 1725 2925 2325 3975 4545 3225
Table C10
Processing capacities of potential biodiesel processing facility locations (per period).
Potential preprocessing center facilities Potential biorefinery facilities Potential storage and distribution facilities
l Division Value (tons) o Division Value (thousand gallons) p Division Value (b1) (gallons) Value (b2) (gallons)
1 Lahore 26,500 1 Sheikhupura 39,600 1 Bahawalpur 1,200,000 1,680,000
2 Multan 12,000 2 D. I Khan 18,000 2 Lahore 5,000,000 7,000,000
3 Rawalpindi 21,000 3 Karachi 36,000 3 Rawalpindi 3,000,000 4,200,000
4 Sargodha 17,500 4 Makran 16,200 4 Sargodha 1,000,000 1,400,000
5 Hyderabad 26,500 5 Karak 18,000 5 Karachi 5,000,000 7,000,000
6 Larkana 12,500 6 Zhob 14,400 6 Sukkur 2,500,000 3,500,000
7 Sukkur 20,500 7 Muzzafarabad 27,000 7 Peshawar 3,000,000 4,200,000
8 Bannu 12,500 8 Mansehra 3,000,000 4,200,000
9 Mardan 13,000 9 Sawat 1,000,000 1,400,000
10 Baltistan 25,000 10 Gilgit 2,000,000 2,800,000
11 Diamer 13,000 11 Quetta 2,000,000 2,800,000
12 Kalat 13,000 12 Muzaffarabad 3,000,000 4,200,000
13 Makran 16,000
14 Sibi 18,500
15 Muzaffarabad 20,500
29
Table C11
Unemployment rate at potential biodiesel processing facility locations (%).
l Division Value o Division Value p Division Value

1 Lahore 0.050 1 Sheikhupura 0.055 1 Bahawalpur 0.080
2 Multan 0.062 2 D. I Khan 0.075 2 Lahore 0.050
3 Rawalpindi 0.055 3 Karachi 0.065 3 Rawalpindi 0.055
4 Sargodha 0.065 4 Makran 0.115 4 Sargodha 0.065
5 Hyderabad 0.085 5 Karak 0.080 5 Karachi 0.065
6 Larkana 0.100 6 Zhob 0.090 6 Sukkur 0.096
7 Sukkur 0.096 7 Muzzafarabad 0.070 7 Peshawar 0.075
8 Bannu 0.075 8 Mansehra 0.070
9 Mardan 0.064 9 Sawat 0.065
10 Baltistan 0.069 10 Gilgit 0.070
11 Diamer 0.070 11 Quetta 0.070
12 Kalat 0.105 12 Muzaffarabad 0.070
13 Makran 0.115
14 Sibi 0.080
15 Muzaffarabad 0.070
Table C12
Regional development level at potential biodiesel processing facility locations.
l Division Value o Division Value p Division Value

1 Lahore 0.877 1 Sheikhupura 0.738 1 Bahawalpur 0.645
2 Multan 0.718 2 D. I Khan 0.496 2 Lahore 0.877
3 Rawalpindi 0.871 3 Karachi 0.854 3 Rawalpindi 0.871
4 Sargodha 0.728 4 Makran 0.400 4 Sargodha 0.728
5 Hyderabad 0.715 5 Karak 0.615 5 Karachi 0.854
6 Larkana 0.618 6 Zhob 0.295 6 Sukkur 0.659
7 Sukkur 0.659 7 Muzzafarabad 0.754 7 Peshawar 0.756
8 Bannu 0.613 8 Mansehra 0.676
9 Mardan 0.703 9 Sawat 0.618
10 Baltistan 0.523 10 Gilgit 0.564
11 Diamer 0.523 11 Quetta 0.664
12 Kalat 0.405 12 Muzaffarabad 0.754
13 Makran 0.400
14 Sibi 0.441
15 Muzaffarabad 0.754
Table C13
Number of jobs created by establishing biodiesel processing facilities.
l Division Value o Division Value (r1) Value (r2) p Division Value (b1) Value (b2)
1 Lahore 264 1 Sheikhupura 490 588 1 Bahawalpur 91 118
2 Multan 120 2 D. I Khan 315 378 2 Lahore 379 492
3 Rawalpindi 210 3 Karachi 450 540 3 Rawalpindi 227 295
4 Sargodha 174 4 Makran 225 270 4 Sargodha 76 98
5 Hyderabad 264 5 Karak 360 432 5 Karachi 379 492
6 Larkana 126 6 Zhob 315 378 6 Sukkur 189 246
7 Sukkur 204 7 Muzzafarabad 410 492 7 Peshawar 227 295
8 Bannu 126 8 Mansehra 227 295
9 Mardan 132 9 Sawat 76 98
10 Baltistan 252 10 Gilgit 151 197
11 Diamer 132 11 Quetta 151 197
12 Kalat 132 12 Muzaffarabad 227 295
13 Makran 162
14 Sibi 186
15 Muzaffarabad 204
30
Table C14
Economic value of biodiesel processing facilities ($).
l Division Value o Division Value (r1) million $ Value (r2) million $ p Division Value (b1) Value (b2)
1 Lahore 6,900,000 1 Sheikhupura 3200 5700 1 Bahawalpur 1,200,000 1,680,000
2 Multan 4,000,000 2 D. I Khan 1500 2600 2 Lahore 4,500,000 6,300,000
3 Rawalpindi 5,800,000 3 Karachi 2900 5200 3 Rawalpindi 2,700,000 3,780,000
4 Sargodha 5,300,000 4 Makran 1300 2100 4 Sargodha 1,300,000 1,820,000
5 Hyderabad 6,400,000 5 Karak 1400 2300 5 Karachi 4,500,000 6,300,000
6 Larkana 4,100,000 6 Zhob 1100 2000 6 Sukkur 2,400,000 3,360,000
7 Sukkur 5,600,000 7 Muzzafarabad 2100 4000 7 Peshawar 2,700,000 3,780,000
8 Bannu 3,900,000 8 Mansehra 2,600,000 3,640,000
9 Mardan 4,200,000 9 Sawat 1,000,000 1,400,000
10 Baltistan 6,500,000 10 Gilgit 2,100,000 2,940,000
11 Diamer 4,100,000 11 Quetta 2,100,000 2,940,000
12 Kalat 4,300,000 12 Muzaffarabad 2,800,000 3,920,000
13 Makran 5,100,000
14 Sibi 5,500,000
15 Muzaffarabad 5,700,000
Table C15
Demand for biodiesel along the time horizon (gallons).
d Demand zones t=1 t=2 t=3
1 Vehari 200,000 400,000 450,000

2 Muzaffargarh 400,000 350,000 390,000
3 Kasur 1,500,000 2,000,000 1,900,000
4 Gujranwala 1,000,000 1,100,000 1,050,000
5 Rawalpindi 1,100,000 1,500,000 1,900,000
6 Chakwal 500,000 600,000 700,000
7 Jhang 100,000 700,000 800,000
8 Karachi 2,200,000 2,300,000 2,000,000
9 Hyderabad 1,200,000 1,100,000 1,000,000
10 Larkana 150,000 250,000 200,000
11 Badin 90,000 60,000 70,000
12 Benazirabad 70,000 90,000 100,000
13 Abbotabad 1,400,000 1,500,000 1,300,000
14 Dir 300,000 500,000 450,000
15 Kohat 1,200,000 1,000,000 1,100,000
16 Chitral 350,000 250,000 450,000
17 Skardu 300,000 250,000 280,000
18 Hunza 200,000 300,000 250,000
19 Quetta 100,000 90,000 110,000
20 Kalat 500,000 800,000 850,000
Table C16
Demand for glycerol along the time horizon (gallons).
g Demand zone/period t=1 t=2 t=3
1 Kasur 50,000 65,000 70,000

2 Gujranwala 140,000 130,000 125,000
3 Rawalpindi 200,000 220,000 210,000
4 Peshawar 150,000 190,000 180,000
5 Karachi 250,000 290,000 260,000
6 Hyderabad 80,000 70,000 85,000
7 Abbotabad 120,000 125,000 115,000
8 Kohat 70,000 60,000 80,000
9 Quetta 50,000 70,000 80,000
10 Lahore 280,000 380,000 290,000
[ pccp ] ( pccp [ pccp ] )

Maximize E fsoc − κsoc fmin, soc − E fsoc +
∑ T [
Q ∑
I ∑ ]
(ϖ − λ)σ iqt(1) + (1 − ϖ)σiqt(2)
φfeed
sup,soc − σ iqt(1) +
i q t
1− λ
∑ T [
D ∑
(θ − λ)μdt(4) + (1 − θ)μdt(3)
] (B-25)
φbio
dem,soc − μdt(4) +
d t
1− λ
∑ T [
G ∑ ]
(ϑ − λ)μgt(4) + (1 − ϑ)μgt(3)
φgly
dem,soc − μgt(4)
g t
1− λ
subject to system constraints (B-20)–(B-22) and (23)–(49).
31
Appendix C
Abbreviations of the WAF-SCND facility installation locations.
l Division Short name o Division Short name p Division Short name
1 Lahore LHE 1 Sheikhupura SWV 1 Bahawalpur BHV

2 Multan MLT 2 D. I Khan DIK 2 Lahore LHE
3 Rawalpindi RWP 3 Karachi KHI 3 Rawalpindi RWP
4 Sargodha SGD 4 Makran Makran 4 Sargodha SGD
5 Hyderabad HDD 5 Karak KRK 5 Karachi KHI
6 Larkana LK 6 Zhob ZHB 6 Sukkur SUK
7 Sukkur SUK 7 Muzzafarabad MFG 7 Peshawar PES
8 Bannu BNU – – – 8 Mansehra MSH
9 Mardan MD – – – 9 Sawat SWT
10 Baltistan BTT – – – 10 Gilgit GIL
11 Diamer DM – – – 11 Quetta QT
12 Kalat Kalat – – – 12 Muzaffarabad MFG
13 Makran Makran – – – – – –
14 Sibi Sibi – – – – – –
15 Muzaffarabad MFG – – – – – –
References sustainable supply chain management. Int J Prod Econ 2021;231:107867. https://
doi.org/10.1016/j.ijpe.2020.107867.
[19] Sarkar B, Tayyab M, Kim N, Habib MS. Optimal production delivery policies for
[1] Ambat I, Srivastava V, Sillanpää M. Recent advancement in biodiesel production
supplier and manufacturer in a constrained closed-loop supply chain for returnable
methodologies using various feedstock: a review. Renew Sustain Energy Rev 2018;
transport packaging through metaheuristic approach. Comput Ind Eng 2019;135:
90:356–69. https://doi.org/10.1016/j.rser.2018.03.069.
987–1003.
[2] Imran M, Salman Habib M, Hussain A, Ahmed N, Al-Ahmari MA. Inventory routing
[20] Wyatt VT, Hess MA, Dunn RO, Foglia TA, Haas MJ, Marmer WN. Fuel properties
problem in supply chain of perishable products under cost uncertainty.
and nitrogen oxide emission levels of biodiesel produced from animal fats. J Am Oil
Mathematics 2020;8:382.
Chem Soc 2005;82:585–91. https://doi.org/10.1007/s11746-005-1113-2.
[3] Habib MS, Sarkar B, Tayyab M, Saleem MW, Hussain A, Ullah M, et al. Large-scale
[21] Hoekman SK, Broch A, Robbins C, Ceniceros E, Natarajan M. Review of biodiesel
disaster waste management under uncertain environment. J Cleaner Prod 2019;
composition, properties, and specifications. Renew Sustain Energy Rev 2012;16:
212:200–22.
143–69. https://doi.org/10.1016/j.rser.2011.07.143.
[4] Habib MS, Lee YH, Memon MS. Mathematical models in humanitarian supply chain
[22] Sander A, Antonije Košćak M, Kosir D, Milosavljević N, Parlov Vuković J, Magić L.
management: a systematic literature review. Mathemat Probl Eng 2016;2016:
The influence of animal fat type and purification conditions on biodiesel quality.
3212095. https://doi.org/10.1155/2016/3212095.
Renew Energy 2018;118:752–60. https://doi.org/10.1016/j.renene.2017.11.068.
[5] Mishra U, Wu J-Z, Sarkar B. A sustainable production-inventory model for a
[23] Rajak U, Verma TN. Effect of emission from ethylic biodiesel of edible and non-
controllable carbon emissions rate under shortages. J Cleaner Prod 2020;256:
edible vegetable oil, animal fats, waste oil and alcohol in CI engine. Energy
120268. https://doi.org/10.1016/j.jclepro.2020.120268.
Convers Manage 2018;166:704–18. https://doi.org/10.1016/j.
[6] Banković-Ilić IB, Stojković IJ, Stamenković OS, Veljkovic VB, Hung Y-T. Waste
enconman.2018.04.070.
animal fats as feedstocks for biodiesel production. Renew Sustain Energy Rev 2014;
[24] Ren J, Tan S, Yang L, Goodsite ME, Pang C, Dong L. Optimization of emergy
32:238–54. https://doi.org/10.1016/j.rser.2014.01.038.
sustainability index for biodiesel supply network design. Energy Convers Manage
[7] Palak G, Ekşioğlu SD, Geunes J. Analyzing the impacts of carbon regulatory
2015;92:312–21. https://doi.org/10.1016/j.enconman.2014.12.066.
mechanisms on supplier and mode selection decisions: an application to a biofuel
[25] Gebremariam SN, Marchetti JM. Biodiesel production through sulfuric acid
supply chain. Int J Prod Econ 2014;154:198–216. https://doi.org/10.1016/j.
catalyzed transesterification of acidic oil: techno economic feasibility of different
ijpe.2014.04.019.
process alternatives. Energy Convers Manage 2018;174:639–48. https://doi.org/
[8] Ahmed W, Sarkar B. Management of next-generation energy using a triple bottom
10.1016/j.enconman.2018.08.078.
line approach under a supply chain framework. Resour Conserv Recycl 2019;150:
[26] Zuo Q, Zhu X, Liu Z, Zhang J, Wu G, Li Y. Prediction of the performance and
104431. https://doi.org/10.1016/j.resconrec.2019.104431.
emissions of a spark ignition engine fueled with butanol-gasoline blends based on
[9] Babazadeh R. Application of fuzzy optimization to bioenergy-supply-chain
support vector regression. Environ Prog Sustain Energy 2019;38:e13042.
planning under epistemic uncertainty: a new approach. Ind Eng Chem Res 2019;58:
[27] Zhang, Z.; E, J.; Chen, J.; Zhu, H.; Zhao, X.; Han, D.; Zuo, W.; Peng, Q.; Gong, J.;
6519–36. https://doi.org/10.1021/acs.iecr.8b05617.
Yin, Z. Effects of low-level water addition on spray, combustion and emission
[10] Ghaderi H, Moini A, Pishvaee MS. A multi-objective robust possibilistic
characteristics of a medium speed diesel engine fueled with biodiesel fuel. Fuel
programming approach to sustainable switchgrass-based bioethanol supply chain
2019, 239, 245-262, doi:https://doi.org/10.1016/j.fuel.2018.11.019.
network design. J Cleaner Prod 2018;179:368–406. https://doi.org/10.1016/j.
[28] Barrios CC, Domínguez-Sáez A, Martín C, Álvarez P. Effects of animal fat based
jclepro.2017.12.218.
biodiesel on a TDI diesel engine performance, combustion characteristics and
[11] Babazadeh R. Optimal design and planning of biodiesel supply chain considering
particle number and size distribution emissions. Fuel 2014;117:618–23. https://
non-edible feedstock. Renew Sustain Energy Rev 2017;75:1089–100. https://doi.
doi.org/10.1016/j.fuel.2013.09.037.
org/10.1016/j.rser.2016.11.088.
[29] Ahmed W, Sarkar B. Impact of carbon emissions in a sustainable supply chain
[12] Kirubakaran M, SelvanV. Arul Mozhi. A comprehensive review of low cost
management for a second generation biofuel. J Cleaner Prod 2018;186:807–20.
biodiesel production from waste chicken fat. Renew Sustain Energy Rev 2018;82:
https://doi.org/10.1016/j.jclepro.2018.02.289.
390–401. https://doi.org/10.1016/j.rser.2017.09.039.
[30] Mousavi Ahranjani P, Ghaderi SF, Azadeh A, Babazadeh R. Hybrid multiobjective
[13] Kerihuel A, Kumar MS, Bellettre J, Tazerout M. Use of animal fats as CI engine fuel
robust possibilistic programming approach to a sustainable bioethanol supply
by making stable emulsions with water and methanol. Fuel 2005;84:1713–6.
chain network design. Ind Eng Chem Res 2018;57:15066–83. https://doi.org/
[14] Kleinová A, Vailing I, Lábaj J, Mikulec J, Cvengroš J. Vegetable oils and animal fats
10.1021/acs.iecr.8b02869.
as alternative fuels for diesel engines with dual fuel operation. Fuel Process
[31] Bairamzadeh S, Pishvaee MS, Saidi-Mehrabad M. Multiobjective robust
Technol 2011;92:1980–6. https://doi.org/10.1016/j.fuproc.2011.05.018.
possibilistic programming approach to sustainable bioethanol supply chain design
[15] Lapuerta M, Rodríguez-Fernández J, Oliva F, Canoira L. Biodiesel from low-grade
under multiple uncertainties. Ind Eng Chem Res 2016;55:237–56. https://doi.org/
animal fats: diesel engine performance and emissions. Energy Fuels 2008;23:
10.1021/acs.iecr.5b02875.
121–9.
[32] You F, Wang B. Life cycle optimization of biomass-to-liquid supply chains with
[16] Alptekin E, Canakci M. Optimization of pretreatment reaction for methyl ester
distributed-centralized processing networks. Ind Eng Chem Res 2011;50:
production from chicken fat. Fuel 2010;89:4035–9. https://doi.org/10.1016/j.
10102–27. https://doi.org/10.1021/ie200850t.
fuel.2010.04.031.
[33] Ziolkowska JR. Evaluating sustainability of biofuels feedstocks: a multi-objective
[17] Babazadeh R, Razmi J, Pishvaee MS, Rabbani M. A sustainable second-generation
framework for supporting decision making. Biomass Bioenergy 2013;59:425–40.
biodiesel supply chain network design problem under risk. Omega 2017;66:
https://doi.org/10.1016/j.biombioe.2013.09.008.
258–77. https://doi.org/10.1016/j.omega.2015.12.010.
[18] Sarkar B, Sarkar M, Ganguly B, Cárdenas-Barrón LE. Combined effects of carbon
emission and production quality improvement for fixed lifetime products in a
32
[34] Gonela V, Zhang J, Osmani A, Onyeaghala R. Stochastic optimization of [47] Pishvaee MS, Razmi J, Torabi SA. An accelerated Benders decomposition algorithm
sustainable hybrid generation bioethanol supply chains. Trans Res Part E Logist for sustainable supply chain network design under uncertainty: a case study of
Transport Rev 2015;77:1–28. https://doi.org/10.1016/j.tre.2015.02.008. medical needle and syringe supply chain. Transport Res Part E Logist Transport
[35] de Jong S, Hoefnagels R, Wetterlund E, Pettersson K, Faaij A, Junginger M. Cost Rev 2014;67:14–38. https://doi.org/10.1016/j.tre.2014.04.001.
optimization of biofuel production – The impact of scale, integration, transport and [48] Rabbani M, Zhalechian M, Farshbaf-Geranmayeh A. A robust possibilistic
supply chain configurations. Appl Energy 2017;195:1055–70. https://doi.org/ programming approach to multiperiod hospital evacuation planning problem
10.1016/j.apenergy.2017.03.109. under uncertainty. Int Trans Operat Res 2018;25:157–89.
[36] Cambero C, Sowlati T. Incorporating social benefits in multi-objective optimization [49] Habib, M.S. Robust Optimization for Post-Disaster Debris Management in
of forest-based bioenergy and biofuel supply chains. Appl Energy 2016;178: Humanitarian Supply Chain: a Sustainable Recovery Approach., Hanyang
721–35. https://doi.org/10.1016/j.apenergy.2016.06.079. University, 2018.
[37] Santibañez-Aguilar JE, Morales-Rodriguez R, González-Campos JB, Ponce- [50] Habib MS, Sarkar B. An integrated location-allocation model for temporary disaster
Ortega JM. Stochastic design of biorefinery supply chains considering economic debris management under an uncertain environment. Sustainability 2017;9:716.
and environmental objectives. J Cleaner Prod 2016;136:224–45. https://doi.org/ [51] Mousazadeh, Torabi SA, Zahiri B. A robust possibilistic programming approach for
10.1016/j.jclepro.2016.03.168. pharmaceutical supply chain network design. Comput Chem Eng 2015;82:115–28.
[38] Sarkar M, Sarkar B. How does an industry reduce waste and consumed energy https://doi.org/10.1016/j.compchemeng.2015.06.008.
within a multi-stage smart sustainable biofuel production system? J Cleaner Prod [52] Habib MS, Asghar O, Hussain A, Imran M, Mughal MP, Sarkar B. A robust
2020;262:121200. https://doi.org/10.1016/j.jclepro.2020.121200. possibilistic programming approach toward animal fat-based biodiesel supply
[39] Azadeh A, Vafa Arani H. Biodiesel supply chain optimization via a hybrid system chain network design under uncertain environment. J Cleaner Prod 2020. https://
dynamics-mathematical programming approach. Renew Energy 2016;93:383–403. doi.org/10.1016/j.jclepro.2020.122403. 122403.
https://doi.org/10.1016/j.renene.2016.02.070. [53] Liu B, Liu Y-K. Expected value of fuzzy variable and fuzzy expected value models.
[40] Zhang Y, Jiang Y. Robust optimization on sustainable biodiesel supply chain IEEE Trans Fuzzy Syst 2002;10:445–50.
produced from waste cooking oil under price uncertainty. Waste Manage 2017;60: [54] Xu J, Zhou X. Approximation based fuzzy multi-objective models with expected
329–39. https://doi.org/10.1016/j.wasman.2016.11.004. objectives and chance constraints: application to earth-rock work allocation. Inf Sci
[41] Orjuela-Castro JA, Aranda-Pinilla JA, Moreno-Mantilla CE. Identifying trade-offs 2013;238:75–95. https://doi.org/10.1016/j.ins.2013.02.011.
between sustainability dimensions in the supply chain of biodiesel in Colombia. [55] Mousazadeh, Torabi SA, Pishvaee MS, Abolhassani F. Health service network
Comput Electron Agric 2019;161:162–9. https://doi.org/10.1016/j. design: a robust possibilistic approach. Int Trans Operat Res 2017. https://doi.org/
compag.2018.03.009. 10.1111/itor.12417.
[42] Hosseinalizadeh R, Arshadi Khamseh A, Akhlaghi MM. A multi-objective and [56] Sarkar B, Guchhait R, Sarkar M, Cárdenas-Barrón LE. How does an industry
multi-period model to design a strategic development program for biodiesel fuels. manage the optimum cash flow within a smart production system with the carbon
Sustain Energy Technol Assess 2019;36:100545. https://doi.org/10.1016/j. footprint and carbon emission under logistics framework? Int J Prod Econ 2019;
seta.2019.100545. 213:243–57. https://doi.org/10.1016/j.ijpe.2019.03.012.
[43] Mirhashemi MS, Mohseni S, Hasanzadeh M, Pishvaee MS. Moringa oleifera [57] Torabi SA, Hassini E. An interactive possibilistic programming approach for
biomass-to-biodiesel supply chain design: An opportunity to combat desertification multiple objective supply chain master planning. Fuzzy Sets Syst 2008;159:
in Iran. J Cleaner Prod 2018;203:313–27. https://doi.org/10.1016/j. 193–214. https://doi.org/10.1016/j.fss.2007.08.010.
jclepro.2018.08.257. [58] Jabbarzadeh A, Fahimnia B, Sheu J-B. An enhanced robustness approach for
[44] Goedkoop MJ. The Eco-indicator 99 a damage oriented method for life cycle managing supply and demand uncertainties. Int J Prod Econ 2017;183:620–31.
impact assessment methodology report. Pre Concultants 1999. https://doi.org/10.1016/j.ijpe.2015.06.009.
[45] Pishvaee, Razmi J, Torabi SA. Robust possibilistic programming for socially [59] Samani MRG, Hosseini-Motlagh S-M. An enhanced procedure for managing blood
responsible supply chain network design: a new approach. Fuzzy Sets Syst 2012; supply chain under disruptions and uncertainties. Ann Oper Res 2019;283:
206:1–20. https://doi.org/10.1016/j.fss.2012.04.010. 1413–62.
[46] Benoît C, Norris GA, Valdivia S, Ciroth A, Moberg A, Bos U, et al. The guidelines for
social life cycle assessment of products: just in time! Int J Life Cycle Assess 2010;
15:156–63. https://doi.org/10.1007/s11367-009-0147-8.
33

Biodiesel supply chain

Uploaded by

Copyright:

Available Formats

Biodiesel supply chain

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Biodiesel supply chain

Uploaded by

Copyright:

Available Formats

Energy Conversion and Management 225 (2020) 113345

Contents lists available at ScienceDirect

Energy Conversion and Management

Management of animal fat-based biodiesel supply chain under the

You et al. [32] Cellulosic Bioethanol Edible ✓ ✓ ✓ ✓ ✓ ✓

∑ Environmental impact of holding inventory at different stages of the

Animal fat oil, biodiesel, and glycerol production cost ∑

3.3.2. Environmental impact objective function O ∑

beef fat oil produced at the respective processing center in a specific

i IGort = IGo,r,t− 1 + − πogrt ∀o, t|t⩾2 (35)

Constraints (25) and (26) provide the total produced volume of O ∑

fat oil produced at preprocessing centers minus the quantity of animal ∑

To deal with the shortcomings of possibilistic programming, robust

life situations, the attitude of the decision-makers lies somewhere be­

The EV of an uncertain number f using the Me measure approach is

MRPCCP formulation seeks a tradeoff among three components of the

Fuzzy numbers (0,0,0.2) (0,0.2,0.4) (0.2,0.4,0.6) (0.4,0.6,0.8) (0.6,0.8,1.0) (0.8,1.0,1.0)

non-dominated and non-inferior solution of the considered model. The

a) Obtain a payoff table for the model objectives

otherwise, (NIS = negative ideal solution)

Supply chain cost Environmental impact

Fig. 3. Optimum solutions of sustainable WAF-SCND using the MRPCCP approach.

(1,0,0) (0.8,0.1,0.1) (0.7,0.2,0.1) (0.7,0.1,0.2) (0.7,0.15,0.15) (0.65,0.25,0.1) (0.65,0.1,0.25)

μ_TSC μ_EI μ_SI Minimum satisfaction level (ϴ)

Objective satisfaction level 95.0%

WAF-SCND Objective priority weight

μ_TSC μ_EI μ_SI Minimum satisfaction level (ϴ)

μ_TSC μ_EI μ_SI Minimum satisfaction level (ϴ)

μ_TSC μ_EI μ_SI Minimum satisfaction level (ϴ)

μ_TSC μ_EI μ_SI Minimum satisfaction level (ϴ)

Cost Environmental Social Cost Environmental Social Economic Environmental Social

1 29,893,180,000 635,347,100 88.4 28,321,887,898 605,975,242 95.4 1,571,292,102 29,371,858 7.00

number and capacity level in areas with low development, which is ⎧

(continued on next page)

Formulation of the animal fat-based biodiesel SC model into an MRPCCP model

Phase 1: Conversion of an uncertain animal fat-based-biodiesel SC model into a PCCP model

Cost objective function crisp formulation

Fixed facility opening costs

Feedstock purchase costs and handling costs

Inventory holding cost

Environmental objective function

Environmental impact of establishing biodiesel SC facilities

Environmental impact of feedstock handling

Environmental impact of holding inventory at different stages of the biodiesel SC

Environmental impact of biodiesel and glycerol production

Environmental impact of transporting material among SC facilities

Social objective function

and constraints (23)–(49).

1 Lahore (LHE) 6,900,000 1 Sheikhupura 3200 5700 1 Bahawalpur 1,200,000 1,680,000

Chicken fat 200 300

Chicken fat 1.60 4.1 3.0

1 Lahore 20,000 1 Sheikhupura 26,110 17,407 1 Bahawalpur 5800 7500

Chicken fat 5.00 2.093 2.097

Chicken fat 95.2 0.917

t=1 t=2 t=3 t=1 t=2 t=3 t=1 t=2 t=3

1 525 450 675 795 1005 960 1425 1920 1575

l Division Value o Division Value p Division Value

l Division Value o Division Value p Division Value

1 Vehari 200,000 400,000 450,000

1 Kasur 50,000 65,000 70,000

life situations, the attitude of the decision-makers lies somewhere be