Green and Resilient Mixed Supply Chain Network Design to Reduce Environmental Impacts and Deal With Disruptions

6 7 Disruption risks may halt or adversely affect supply chain operations and can lead to deviations in its 8 objectives. One of the most important objectives of the supply chain which can be adversely affected by 9 disruptions, is environmental objective. Therefore, considering supply chain resilience and environmental 10 aspects simultaneously is of great importance. In this paper, the problem of designing a green and resilient 11 mixed open and closed-loop supply chain network under operational and disruption risks has been studied. 12 A bi-objective mixed integer linear programming model is proposed to formulate the problem. Some 13 resilience strategies are applied to deal with disruption risks and enhance supply chain resilience. In order 14 to overcome the complexity of the problem and solve the problems with medium and large sizes, a new 15 meta-heuristic algorithm called multi-objective hybrid Ant-colony optimization and teaching and learning 16 based optimization (ACO-TLBO) has been proposed and compared with two hybrid metaheuristics and the 17 augmented ε -constraint method through various test problems. The outcomes showed that the ACO-TLBO 18 algorithm is very efficient in obtaining high-quality Pareto solutions and is the best method among the 19 proposed methods. Also, in order to show the applicability of the problem and validate the model and 20 solution methods, a real case study in the tire industry has been presented and analyzed. The results of 21 analyses demonstrate the high effectiveness of resilience strategies and the necessity of joint consideration 22 of resilience and greenness in the supply chain design. resilience of


Introduction 27
Today's organizations seek to exploit the advantages of right supply chain (SC) management in order to maintain 28 their position in the market, create competitive advantages, decrease costs, and generally speaking, manage their 29 supply chain efficiently. The design of the supply chain network plays a crucial role in the supply chain management 30 since it determines the physical structure of the supply chain and makes decisions on issues such as choosing location, 31 number, and capacity of facilities, selecting suppliers, and the like .

32
Climate change poses one of the greatest threats to human life (Meng et al., 2020). Global population growth has led 33 to an increase in energy consumption due to the necessity of responding to growing demands (Ramezanian et al., (Salema et al., 2007). Reverse logistics can improve the environmental aspect of sustainability because they are very 43 effective in reducing energy consumption, material consumption, and environmental pollution.

44
Supply chains are exposed to different risks that can be categorized as operational risks and disruptions. Operational 45 risks are rooted in the inherent uncertainty of supply chains, which include uncertainties in supply, demand, delivery 46 times, shipping times, and costs. Disruptions may occur in parts of the supply chain due to natural disasters (such as 47 floods and earthquakes), intentional or unintentional human actions (such as strikes, wars, and terrorist attacks), or 48 1 Resilient supply chain network design is a growing research field. Especially in recent months, due to the outbreak 105 of Covid-19 pandemic disease, its importance has become more and more realized.

106
The vast majority of studies in the field of resilient SCND have used resilience strategies to deal with disruptions.

107
Multiple sourcing as the most well-known strategy has been sued by Peng

153
(2021b) used multiple sourcing and backup suppliers for designing green and resilient meat SCND.

154
The above-mentioned studies did not consider reverse logistics. Yavari

163
In Table 1, the related papers are summarized and their main characteristics are specified. According to the table, it   164 can be seen that the issue of resilient mixed SCND has not been studied in the articles except for the previous work 165 of the authors. Secondly, the problem of SCND with joint consideration of resilience and environmental issues in 166 networks with reverse logistics has been only addressed in one paper whose network structure is closed-loop. Also, 167 the simultaneous consideration of resilience, greenness and responsiveness has not been studied in previous studies.

168
Moreover, only two papers have studied responsive resilient SCND and more research should be done in this field.

169
As can be seen from Table 1, the problem of resilient supply chain network redesign has been discussed in only one 170 paper, and there is a research gap in this area as well. Overall, to the best of our knowledge, no work has been done 171 on integrated green, resilient and responsive mixed SCND. Based on these descriptions the main contributions of this

Problem definition and mathematical modeling 193
In this section the addressed problem is described and the developed mathematical model is presented.

194
Problem definition

195
In this paper a multi-echelon, and multi-period and multi-product supply chain network design/redesign problem 196 is studied. The network structure is mixed open and closed-loop. The supply chain network consists of suppliers, 197 production centers, distribution centers (DCs), customers, collection/inspection centers, recycling centers and 198 remanufacturing centers. Production centers produce products using raw materials supplied by suppliers. These 199 products can be transferred to distribution centers and from those centers to customer zones or sent directly from 200 production centers to customer zones (two-channel distribution). The end-of-life (EOL) products are collected by 201 collection centers and after inspection, a portion of returned products are transferred to recycling centers and a 202 portion to remanufacturing centers. Also, a percentage of products that cannot be used for recycling or 203 remanufacturing are transferred to disposal centers. In remanufacturing centers, products are converted into usable 204 products of the same type as before, which have a relatively lower price and quality compared to main new 205 products. These products are also sent to distribution centers and distribution centers send them to customer zones.

206
In recycling centers, raw materials are extracted from recycled products and also recycled products used in other 207 supply chains are produced. Extracted raw materials are sent to production centers to be utilized in producing new

218
It is assumed that supply chain facilities and vehicles are exposed to disruptions. When disruption occurs for 219 facilities, their capacity would be loosed partially or completely. Some resilient strategies are applied to cope with 220 disruption risks and increase the resilience of the supply chain, including multiple sourcing, facility fortification, capacity expansion, dual-channel distribution (mentioned in the previous paragraph), dynamic pricing, lateral 222 transshipment, and considering backup vehicles supplied by third-party logistics companies. Dynamic pricing 223 can be utilized as a risk mitigation strategy. In dynamic pricing, different prices can be determined in each period 224 for each product and each customer zone (Yavari et al., 2020). Given that the vehicles are exposed to disruptions, 225 backup vehicles provided by third-party logistics companies are intended to help compensate for the lost capacity 226 of the vehicles of the supply chain in the event of a disruption. Other strategies were explained in the last section.

227
As mentioned in the introduction, disruptions can adversely affect the SC responsiveness. Hence, responsiveness 228 should be considered in the SCND problem. In order to consider responsiveness, it can be included in the objective 229 function or controlled by imposing constraints (Sabouhi et al., 2020). In this article, the second approach is used.

230
In addition to the disruption risks, it is assumed that there is operational uncertainty in the production cost of main 231 products and the cost of purchasing raw materials. All the mentioned uncertainties, whether of the disruption type

240
Mathematical model

241
In order to formulate the problem, a bi-objective mixed-integer linear programming model has been developed.

242
The first objective is to maximize the total profit of the supply chain and the second one seeks to minimize the

246
Sets Set of suppliers, indexed by Set of existing and potential production centers, indexed by . The set of existing production centers is denoted by 0 , and the set of potential locations for production centers is denoted by . = ∪ 0 and ∩ 0 = Set of existing and potential DCs, indexed by and ′. The set of existing DCs is denoted by 0 , and the set of new potential locations for DCs is denoted by . = ∪ 0 and ∩ 0 = Set of customers, indexed by Set of potential locations for collection/ inspection centers, indexed by Set of potential locations for recycling centers, indexed by ℎ Set of potential locations for remanufacturing centers, indexed by Set of supply chains, indexed by Set of vehicle types related to SC and third-party logistics company, indexed by . The set of SC vehicle types is denoted by 0 , and the set of third-party logistics vehicle types is denoted by . Unit transportation cost of raw and recycled materials using vehicle type per unit distance Unit transportation cost of product type using vehicle type per unit distance Maximum number of existing distribution centers that can be closed The distance between supplier and production center The distance between production center and distribution center The distance between production center and customer The distance between distribution centers and ′ ( ≠ ′) ′ The distance between distribution center and customer The distance between customer and collection/ inspection center The distance between collection/ inspection center and remanufacturing center The distance between collection/ inspection center and recycling center ℎ ℎ ℎ The distance between remanufacturing center and distribution center The distance between recycling center ℎ and production center ℎ The distance between recycling center ℎ and production centers of supply chain ℎ Cost of producing one unit product type in production center with technology 1 under scenario 1 Cost per unit of adding extra production capacity for product type in production center with technology 1 under scenario 1 Cost of distributing one unit of product in distribution center Cost of collecting and inspecting one unit of product in collection/ inspection center Cost of recycling one unit of product in recycling center ℎ with technology 2 ℎ 2 Cost of remanufacturing one unit of product type in remanufacturing center with technology 3 3 Cost of disposing of one unit of product in disposal center Cost of purchasing one unit of raw material from supplier under scenario Cost of not meeting one unit of demand related to product type for customer cost of not meeting one unit of demand related to remanufactured product type for customer Cost of not meeting one unit of demand related to recycled products for supply chain Environmental impact of establishing production center with fortification level and manufacturing technology 1 1 Environmental impact of establishing distribution center Environmental impact of establishing collection/inspection center Environmental impact of establishing recycling center ℎ with fortification level and manufacturing technology 2 ℎ 2 Environmental impact of establishing remanufacturing center with fortification level and manufacturing technology 3 3 Environmental impact of producing a unit of product type in production center by using manufacturing technology 1 1 Environmental impact of handling a unit of product in distribution center Environmental impact of handling a unit of product in collection/inspection center Environmental impact of disposing a unit of product type or releasing in environment Environmental impact of processing a unit of product in recycling center ℎ by using manufacturing technology 2 ℎ 2 Environmental impact of processing a unit of product type in remanufacturing center by using manufacturing technology 3 3 Environmental impact of transporting a unit of raw material using vehicle type per unit distance Environmental impact of transporting a unit of product using vehicle type per unit distance The percentage of wasted raw material for producing one unit of product type Quantity of recycled materials obtained from recycling one unit of product Demand of customer related to price level for product type in period , under scenario Demand of customer elated to price level for remanufactured product type in period , under scenario Demand of supply chain elated to price level for recycled products in period , under scenario Offered price level for product type for selling to customer le Offered price level for remanufactured product type for selling to customer ′ le Offered price level for recycled materials for selling to SC lg Amount of determined value for SC responsiveness level related to the demand of customer Amount of determined value for SC responsiveness level related to the demand of SC ̅ Capacity of supplier Production capacity of production center Maximum addable capacity related to production center Distribution Capacity of production center Capacity of distribution center Capacity level for distribution center j Capacity of collection/ inspection Capacity level for collection/ inspection Capacity of recycling center ℎ ℎ Capacity of remanufacturing center Capacity of vehicle type Total number of vehicle type ( ∈ 0 ) for transporting raw materials from supplier in time period Total number of vehicle type ( ∈ 0 ) for transporting products from production center in in time period Total number of vehicle type ( ∈ 0 ) for transporting products from distribution center in in time period Total number of vehicle type ( ∈ 0 ) for transporting products from customer in in time period Total number of vehicle type ( ∈ 0 ) for transporting products from collection/inspection in in time period Total number of vehicle type ( ∈ 0 ) for transporting products from recycling center ℎ in in time period ℎ Total number of vehicle type ( ∈ 0 ) for transporting products from remanufacturing center in in time period Total number of vehicle type ( ∈ ) for transporting products in supply chain supplied from third -party logistics company in period Percentage of returned product type from customer Percentage of returned product type sent from collection/ inspection centers to remanufacturing centers Percentage of returned product type sent from collection/ inspection centers to recycling centers Available (non-disrupted) fraction of production capacity of production center with fortification level in period under scenario Available (non-disrupted) fraction of distribution capacity of production center with fortification level in period under scenario ′ Available (non-disrupted) fraction of capacity of supplier in period under scenario Available (non-disrupted) fraction of capacity of distribution center in period under scenario Available (non-disrupted) fraction of capacity of collection/ inspection center in period under scenario Available (non-disrupted) fraction of capacity of recycling center ℎ with fortification level in period under scenario ℎ Available (non-disrupted) fraction of capacity of remanufacturing center with fortification level in period under scenario Available (non-disrupted) fraction of total number of vehicle type ( ∈ 0 ) for transporting raw materials from supplier in period under scenario 1 Available (non-disrupted) fraction of total number of vehicle type ( ∈ 0 ) for transporting products from production center in period under scenario 2 Available (non-disrupted) fraction of total number of vehicle type ( ∈ 0 ) for transporting products from distribution center in period under scenario 3 Available (non-disrupted) fraction of total number of vehicle type ( ∈ 0 ) for transporting products from customer in period under scenario 4 Available (non-disrupted) fraction of total number of vehicle type ( ∈ 0 ) for transporting products from collection/inspection in period under scenario 5 Available (non-disrupted) fraction of total number of vehicle type ( ∈ 0 ) for transporting products from recycling center ℎ in period under scenario 6 ℎ Available (non-disrupted) fraction of total number of vehicle type ( ∈ 0 ) for transporting products from remanufacturing center in period under scenario 7 Probability of scenario occurrence 248

Variables
Quantity of raw material shipped from supplier to production center using vehicle in period under scenario Quantity of product type produced at production center in period with technology under scenario 1 Quantity of added capacity to production center for producing product type with technology 1 in period under scenario 1 Quantity of product type shipped from production center to distribution center using vehicle in period under scenario Quantity of product type shipped from production center to customer using vehicle in period under scenario Quantity of product type shipped from distribution center to distribution center ′ using vehicle in period under scenario ′ Quantity of product type shipped from distribution center to customer using vehicle in period under scenario Quantity of remanufactured product type shipped from distribution center to customer using vehicle in period under scenario Quantity of product type shipped from customer to collection/ inspection center using vehicle in period under scenario Quantity of product type shipped from collection/ inspection center to remanufacturing center using vehicle in period under scenario Quantity of product type shipped from collection/ inspection center to recycling center ℎ using vehicle in period under scenario ℎ Quantity of remanufactured product type remanufactured with technology 3 shipped from remanufacturing center to distribution center using vehicle in period under scenario 3 Quantity of recycled materials recycled with technology 2 shipped from recycling center ℎ to production center using vehicle in period under scenario 2 ℎ Quantity of recycled materials produced with recycling technology 2 shipped from recycling center ℎ to supply chain using vehicle in period under scenario

265
( (5) and (6) state that at most one fortification level and one technology should be selected for 266 production centers, recycling centers and remanufacturing centers, respectively that are to be established.

283
Constraints (37) to (44) indicate the limitation on the capacity of supply chain vehicles. The capacity limit related 284 to third-party logistics company is assured by constraint (46).

287
(48)  is NP-hard (Soleimani and Govindan, 2015). Furthermore, the problem presented in this paper has a more complex 293 structure than the closed-loop supply chain network design problem. Therefore, exact optimization methods are 294 not applicable for solving medium and large-sized problems.

295
In this paper, three hybrid metaheuristics are proposed to cope with problem complexity and find high-quality 296 solutions, including improved hybrid genetic and particle swarm optimization algorithm (hybrid GA-PSO), improved hybrid genetic and simulated annealing ( hybrid GA-SA) a novel hybrid algorithm titled hybrid ant 298 colony optimization and teaching and learning-based optimization algorithm (hybrid ACO-TLBO). The 299 augmented ε-constraint method proposed by Mavrotas (2009) is also applied to verify these algorithms.

302
There are different methods for representing and encoding the solutions, among which two main approaches can 303 be mentioned, including the matrix representation approach (Michalewicz et al, 1991) and the priority-based 304 representation approach (Gen et al, 2006). In this paper, our representation method is similar to priority-based 305 representation. An example is provided to illustrate how solutions are represented. It is assumed that the number 306 of suppliers, production centers, distribution centers, customer zones, collection centers, recycling centers, 307 remanufacturing centers and other supply chains are 2, 2, 2, 3, 2, 2, 2 and 2, respectively. The number of time   corresponding to the establishment of distribution centers is determined based on this segment. In order to be concise, a pseudo-code is presented in Figure 4 which shows the general decoding procedure. For all segments, 330 such a procedure should be implemented with differences in details and considering the related constraints.

331
The second sub-chromosome consists of three sectors. The first sector determines the price level of the main 332 products and the second and third sectors determine the price level of the remanufactured and recycled products,

379
is a positive parameter and has a similar effect compared to pheromone evaporation rate of ACO algorithm.

380
The second phase of the ACOR algorithm is Pheromone Update. In this algorithm, the solution archive contains The new solution will replace the previous one if it has a better fitness value.

401
To the best of the authors' knowledge, the hybridization of ACOR and TLBO algorithm has not been studied 402 before. To develop the structure of the hybrid algorithm, it should be noted that the problem studied in this paper 403 is multi-objective, and an appropriate approach should be considered in designing the structure of the algorithm 404 for handling this issue. Figure

444
The pseudo-code of hybrid improved GA-SA algorithm is shown in Figure 7. In the presented pseudo-code Δ =

458
Case study and test problems

459
In this paper, a tire supply chain in Iran is presented as a real case study to show the applicability of the

479
The SC considered for the case study currently consists of one factory, two distribution centers and four suppliers.

490
Given that problem under study is multi-objective, it is not possible to assess the performance of the solution 491 methods using a simple criterion; therefore, in this paper, some performance metrics are applied in order to

520
In order to achieve the best performance of metaheuristics, in this section, the parameters of the metaheuristic 521 algorithms are tuned using the Taguchi method. This method is utilized in order to avoid plenty number of 522 experiments of full factorial experimental design. In this method, factors are classified into two categories: 523 controllable and noise. The desired value is represented by signal and the undesirable value is denoted by noise.

524
In the Taguchi method, the concept of signal to noise ratio (S/N), which represents the variation of response value, 525 is used. Taguchi method attempts to reduce the effect of noise factors (Kumar, 2017). There are three types of 526 responses, including "smaller is better", "nominal is best" and "larger is better" (Roy, 2010). In this article the 527 "smaller is better" is applied for tuning the parameters of algorithms.

528
(58) In the above equation, represents the response value and denotes the number of orthogonal arrays. In order to 529 tune the parameter of algorithms, at first, the level of parameters belonging to algorithms are determined. The 530 levels of parameters are presented in Table 4. In this table, Ψ = | | + | | + | | + | | + | | + | | + | | + | |.

531
The determined values are selected based on vast experiments and the related papers of the literature.

532
In the next step, a Taguchi design is created using Minitab software and finally the Taguchi design is analyzed to 533 choose the best levels of parameters.

534
Using Taguchi Table 4 show the proper levels selected.

552
The results are reported in Table 5, and also illustrated in Figures 13-18. In order to validate the algorithms, 553 augmented ε-constraint method is used, which based on the results it is not able to solve medium and large-sized 554 problems. 20 grid points was considered for the augmented ε-constraint (interested readers are referred to 555 Mavrotas, 2009 for studying the details of augmented ε-constraint method). The considered time limit for all 556 solution methods is 60000 seconds (NA means no answer could be found in the predetermined time limit). The 557 spent CPU time for solving small-sized problems indicate the NP-hardness of the problem.   Table 6. According 567 the obtained values, ACO-TLBO is selected as the best solution method, for it has the smallest value of direct 568 distance. The Pareto fronts of the three algorithms for the case study problem is given in Figure 19.

583
As mentioned before, in this paper it is assumed that SC network is under disruption risks, and some resilience 584 strategies are applied to increase the SC resilience and cope with disruptions. In the following, the effects of 585 resilience strategies on SC objectives are investigated. The problem related to the case study is chosen for doing 586 the analyses. Figure 21 represent Figure 22 depicts the objective functions under the considered conditions. In 607 Figure 22(a) the first objective function is optimized without considering the second one, and in Figure 22

610
The values of the recycled materials and released/disposed EOL products are represented in Figure 23. As can be 611 seen in resilient mode (condition 8), the amount of recycled materials is higher and accordingly less raw materials 612 are consumed than in non-resilient condition (condition 1). Also, in non-resilient mode, more EOL products 613 remain in the environment. Thus, the protection of natural resources in the non-resilient mode is much less and 614 the environmental pollution in this state is much more. All in all, it is concluded that resilience is necessary for a 615 supply chain to be green.  The outputs meet our expectations and are logical. Given that the first objective function is profit maximization, 629 in responsiveness rates between 0.3 to 0.7, since the supply chain is able to meet customer demand up to about 630 70% (Note the constraints of responsiveness), the amount of objective function is constant in this range. With 631 increasing responsiveness rate, the supply chain is not able to meet demand, and shortage costs increases and 632 consequently, supply chain profitability is reduced. Regarding the second objective function, which is to minimize 633 the negative environmental effects, it can be said that the problem seeks to reduce production and other activities 634 in order to reduce the objective function, but the constraints on responsiveness rate prevent the objective function 635 to reach near zero values. Furthermore, as the responsiveness rate increases, the amounts of production, 636 transportation, and other activities increase, and consequently the second objective function deteriorates.

639
The case study for the problem studied in this paper is in the tire industry. However, the presented model is general 640 and can be used in other industries with slight changes. Managers and engineers of the tire industry and other 641 industries can get insight from the problem under study to identify disruption and operational risks of their supply 642 chain and use the presented stochastic model and resilience strategies to deal with them. They should note that 643 when disruption occurs for SC facilities, the company has problems in the production of products and delivering 644 them to customers. Subsequently, Due to the company's inability to meet customers' demands, shortage costs 645 increase and sales revenues decrease, and the company may suffer losses. On the other hand, having weakness in 646 resilience of the supply chain network, the negative environmental effects increase. The first reason is that as the 647 capacity of the supply chain decreases, more new facilities must be established to respond to demands. Also, with 648 the reduction of facility capacity and the opening of new facilities, the amount of transportation will increase. On 649 the other hand, with the occurrence of disruptions, the reverse logistics activities are reduced or stopped, and 650 consequently more EOL products are released in environment or disposed, and more raw materials are consumed 651 to produce the products, so the the negative environmental effects increase. Resilience strategies can help in 652 mitigation of these negative impacts.

653
The proposed model will help the relevant managers and engineers in selecting suppliers, choosing the location  objectives of supply chains that are important to stakeholders can be degraded by disruptions. Therefore, paying 663 attention to supply chain resilience against disruptions is very important to protect the objectives. In this paper, 664 the issue of green and resilient supply chain network design was investigated. The structure of the studied supply 665 chain network was mixed open and closed-loop, and operational and disruption risks were taken into account.

666
Resilience strategies were applied to mitigate the disruption risks, and the uncertainty of the problem was handled