1. INTRODUCTION
Risk management is one of the most important topics in the supply chain. In fact, the academic communities and beneficiaries in the field have widely recognized the importance of risk management and the integrated method in the supply chain to deal with complexities and uncertainties. Enterprises seek to manage risks or classify unexpected disruptions and change their function in an unclear business environment (Munir et al., 2020). Supply chain risk management primarily aims to decrease the impact of these risks through the development of methods and models to identify, evaluate and decrease supply chain risks (Kirilmaz and Erol, 2017). In general, supply chain risk is a potential phenomenon that prevents the natural flow of materials and information in the chain, thereby causing interruptions in the chain (Wu et al., 2006). A wide range of risks in the supply chain may have adverse impacts on the supply chain performance. The close relationship among the supply chain members affects the entire supply chain, disrupting its performance (Christopher et al., 2003). As such, an organization must use proper strategies to manage and control risks in the supply chain. On the other hand, the supplier selection issue is an important and strategic decision in the supply chain, and choosing a proper set of suppliers is extremely important and crucial for the success of different industries.
The risk management process has three levels of risk identification, risk assessment, and risk reduction (Wagner and Bode, 2009;Tang, 2006;Kleindorfer and Saad, 2005). However, all assessments are based on the anticipation of an unknown future and must always be reviewed and updated since the process is a dynamic one. Therefore, the risk control and monitoring stage has also been added to the supply chain risk management process. In some studies, supply chain risk management has been carried out in five stages of risk identification, measurement, assessment, reduction, and control and monitoring (Kirilmaz and Erol, 2017). Risk assessment is the process of comparing the results of risk analysis with risk criteria to determine whether the risk is acceptable / tolerable or not. However, risk criteria are based on organizational objectives and can be extracted from standards, laws and policies, and other requirements (Hosseini and Ivanov, 2020). Risk reduction strategies can be divided into two reactive and preventive groups. In a reactive approach, no action is performed before a risky event. However, some measures are taken afterwards to decrease the effect and possibility of risk. On the other hand, risk reduction is planned before occurrence in the preventive approach. In addition, this method may include the implementation of programs to decrease the possibility and effect of risky events in the future (Kleindorfer and Saad, 2005).
In a recent study, Wang and Jie (2020) presented a framework for managing pharmaceutical supply chain uncertainty and risk in the petrochemical industry. Munir et al. (2020) combined supply chain risk management and supply chain performance improvement. These researchers used empirical instruments to decrease supply chain risks. In addition, they assessed supply chain risks, for which expenditure goals (e.g., minimization of storage costs, minimization of supplier selection costs, minimization of cost (penalty) of returned goods, and vehicle capacity), which have been overlooked until now, were considered in the present study. In addition, studies related to supply chain risk management were reviewed and selected based on risk identification, measurement, assessment, and reduction. In this regard, given the dispersion of supply risks in the research literature, which often does not conform to the environmental conditions of the supply chain studied in this study (due to domestic conditions), attempts were made to identify and select risks related to the supply of domestic suppliers. Afterwards, the supply risks were measured by supply chain network experts at Sapco and the issue of reducing the risk of the entire supply chain was put on the agenda. In the next step, the present study aimed to propose a multiobjective integrated model due to a lack of a previous simultaneous assessment of objectives such as cost reduction, risk reduction, and supply from suppliers. It is worth noting that the current research includes a preventive approach since, contrary to the reactive approach, this method accepts risk probability and properly manages risk before its occurrence by analyzing the events.
2. PROBLEM DEFINITION
In this section, the proposed model is described. Fig. 3 shows the framework of our approach. First, the manufacturer identifies potential suppliers and defines appropriate criteria. Then, decision makers evaluate suppliers by proposed fuzzy model. The results of this phase are the weights (importance) of suppliers based on qualitative metrics. In the next phase, the closed loop supply chain (CLSC) network is formulated as multiobjective mixedinteger linear programming model. In this stage, the related variables (strategic and tactical decision variables) are calculated.
The supply chain studied in this research was a single product threelevel supply chain with rings consisting of seven suppliers (Modern, New Union Auto Parts Co., Jamsaz, Seraj Noor Toos, Maadco, Fanavaran Parto Alvand, and Niknam), five manufacturers (Crouse, Sarv Sanat Sepahan, Mehrkam Pars, Mehrkhah, and Mehrsaz Gostar Pars) and three customers (Iran Khodro in Tehran, Khorasan and Tabriz sites), where car bumper manufacturers purchase fog lamps from suppliers and produce the end product (i.e., the set of bumper) and present it to its customers (Iran Khodro in Tehran, Khorasan and Tabriz sites).
2.1 The First Model
Indexes

i: index of the potential supplier

j: index of the manufacturer (assembly center) with potential disassembly line

k: index of customer

l: index of the piece

m: index of product
Parameters

FO_{i} : cost of supplier selection

FD_{mj} : fixed cost of launching a disassembly line for the mth product in the jth production center

TC_{lig} : cost of transportation of each l piece from the ith supplier to the jth manufacturer

TC_{mjk} : cost of transportation of each mth product unit from the jth manufacturer to the kth customer

$T{C}_{mki}^{\text{'}}$ : cost of transportation of each mth product unit from the kth customer to the disassembly center of the jth manufacturer

P_{lij} : cost of purchasing each lth piece of product from the ith supplier to the jth manufacturer

CP_{mj} : cost of production (assembly) of each mth product unit in the jth production center

CD_{mj} : cost of disassembly of each mth product unit in the disassembly line of the jth production center

CapL_{li} : supply capacity of the lth piece by the ith supplier

CapM_{mj} : production capacity of the mth product in the jth production center

CapD_{mj} : disassembly capacity of the mth product in the disassembly line of the jth production center

N_{lm} : number of the lth piece in each mth product unit

α_{l} : percentage of returned lth piece (faulty)

β_{m} : rate of returned mth product (faulty)

η_{mj} : rate of returned mth product (faulty) recyclable at the disassembly line of the jth production center

F_{lj} : the price of penalty for disposal of each returned lth piece (faulty) in the jth production center

D_{lj} : number of demands for the lth piece by the jth manufacturer

D_{mk} : number of demands for the mth product by the kth customer

S_{mjk} : sales price per each mth product unit by the jth manufacturer to the kth customer

H_{lj} : cost of maintaining each unit of intact lth piece in the jth production center

CapV_{ij} : capacity of the transportation vehicle from the ith supplier to the jth producer

CapV_{jk} : capacity of the transportation vehicle from the jth manufacturer to the kth customer

$Cap{V}_{kj}^{\text{'}}$ : capacity of the transportation vehicle from the kth customer to the disassembly line of the jth production center

R_{lj} : safety stock (SS) of the lth piece in the jth production center
Decision Variables

Q_{lij} : number of the Lth piece delivered from the ith supplier to the jth manufacturer

Q_{mj} : number of the mth product manufactured in the jth production center

Q_{mjk} : number of the mth product delivered from the jth manufacturer to the kth customer

QR_{lj} : total number of returned lth piece (faulty) in the jth production center

QR_{mkj} : number of returned the mth product from the kth customer to the disassembly line of the jth production center

Q_{lj} : total number of intact lth piece stored in the jth production center

NV_{ij} : number of transportation vehicles from the ith supplier to the jth manufacturer

NV_{jk} : number of transportation vehicles from the jth manufacturer to the kth customer

$N{V}_{kj}^{\text{'}}$ : number of transportation vehicles from the kth customer to the disassembly line of the jth manufacturer

X_{i} : binary variable for selection or lack of selection of the ith supplier

Y_{mi} : 1, if the disassembly line of the mth product is established in the jth production center; otherwise, 0.
The objective function (1) minimizes the total costs of the proposed network, which include costs of ordering parts from a potential supplier (potential supplier selection), cost of launching a potential disassembly line in the production center, cost of purchasing pieces from the supplier, cost of production of each product unit in the production center, cost of disassembly of faulty products in the potential disassembly line of the production center, cost of storage of each unit of intact pieces in the warehouse of the manufacturer, cost of disposal of faulty pieces by the manufacturer, cost of transportation of pieces between the supplier and manufacturer, cost of transportation of products between the manufacturer and the customer, and cost of transportation of faulty products between the customer and manufacturer. Furthermore, revenues from the sale of products were deducted from expenses. Constraints (2) determine the total number of faulty pieces in the production center while constraints (3) show the number of faulty products returned by customers and the total number of intact pieces in the manufacturer’s warehouse. Constraints (5) demonstrate that the number of intact pieces should not be less than safety stock (SS) so that ordering could be made before going out of stock. Constraints (6) ensure that the number of manufactured products should not exceed the total number of intact pieces based on the number of pieces used. Constraints (7) show that the number of manufactured products should be equal to the number of products delivered to the customer. Constraints (8) show the number of pieces delivered from the supplier to the manufacturer, and constraints (9) determine the number of products delivered from the production center to the demand point (customer). Constraints (1012) demonstrate the supplier capacity limit, production capacity limit, and disassembly line, respectively. In addition, constraints (1315) show the vehicle capacity limitation for transportation from the supplier to the manufacturer, from the manufacturer to the customer, and from the customer to the production center’s disassembly line. Finally, constraints (16) demonstrate binary variables, whereas constraints (17) show the condition for variables to be positive and correct.
2.2 The Second Model: Minimization of Risks of Suppliers under Different Scenarios
At this stage, the risks identified for this research, including quality problems, shortage of raw materials, rising prices of raw materials, supplier capacity risk, failure of machinery and equipment, IT system failure, accidental risks (e.g., fire), risk of industrial actions (e.g., strikes and sanctions), failure of the transportation system, failure of management policies, exchange restrictions (e.g., increased customs duties) and inflation rate were selected. In addition, they were inquired and measured by supply chain experts based on a questionnaire using the method the FMEA method. Afterwards, a scenariobased mathematical programming model was presented (under three scenarios of good, moderate and bad, with an occurrence possibility of 0.2, 0.6, and 0.2) to minimize the total level of supply risks.
Indexes
Parameters

p_{s}: the probability of occurrence of the sth scenario

λ_{iqs} : level of qtype risk for the ith supplier in the sth scenario

NS: minimum required number of suppliers

Risk_{is} : total risk related to the ith supplier in the sth scenario

XL_{ib}: number of transferable products between highrisk and lowrisk suppliers (output of the third model)

Q_{lij}: number of lth pieces delivered from the supplier to the jth manufacturer (output of the first model)
Decision Variable
The objective function (18) minimizes the total risk level, and constraints (19) express the minimum number of required suppliers. Constraints (20) calculate the total risk level related to each supplier in each scenario. Constraints (21) show the range of changes in the variable. Constraints (22) claims that the transfer of parts between two high and lowrisk suppliers occurs when that supplier is selected and the part is delivered to the manufacturer based on the risk supplier. Constraints (23) demonstrate binary variables while constraints (24) show the condition for variables to be positive and correct.
2.3 The Third Model: Maximization of Supply from the Lowerrisk Suppliers
In this section, the supply method from the revised suppliers was used, and suppliers with relatively lower costs and less risk were applied. The model primarily aims to maximize the flow of products from a highrisk supplier to a lowrisk supplier. Therefore, the parameters of the decision variables in the objective function are the positive differences between the normalized risk value of the suppliers. The objective function value presents no value. However, since the objective function attempts to maximize, it establishes the condition for transfer from a highrisk supplier to a lowrisk supplier.
Indexes

i: supplier

a: higherrisk suppliers compared to the ith supplier

b: lowerrisk suppliers compared to the ith supplier
Parameters

RN_{ij} : positive difference in the value of normalized risks between suppliers

Q_{i} : allowed transferable amount from the ith supplier

C_{i} : capacity of the ith supplier

Q_{lij} : number of the lth piece from the ith supplier to the jth manufacturer (output of the first model)
Decision Variables

XL_{ib} : the number of transferable products between highrisk and lowrisk suppliers

QN_{i} : remaining product in the ith supplier

Cap_{n} : remaining capacity in the ith supplier
The objective function (25) aims to maximize the transfer of products from a risky supplier to a lowrisk supplier. Constraints (26) show that the number of transferable products from the ith supplier to the jth supplier should not exceed the allowed number of transferable products.
Constraints (27) declare that the number of transferable products from the kth higherrisk supplier (compared to the ith supplier) to the jth lower risk supplier (compared to the ith supplier) should be smaller than the remaining capacity of the ith supplier. Formula (28) calculates the risk value, and constraints (29) show the type of X decision variable. Constraints (30) demonstrate the relative value of the total risk based on the lowrisk supplier. Constraints (31) show the remaining products in the ith supplier, and constraints (32) show the remaining capacity of the supplier.
2.4 Multiobjective Integrated Model
In this section, we integrated the multistage model of the previous stage and examined it as a multiobjective model. Therefore, the integrated model included three objective functions of stages 1, 2 and 3 with all the constraints of the three stages.
3. RESULTS
3.1 Report of Multistage Model Results
This section reports and analyzes the results according to the input values in Table 1.
3.2 Comparison of Results of Supply from Suppliers (RiskFree/Risky)
In this section, the two modes of without considering risk criterion and with considering risk criterion were compared in ordering mode.
After considering the risk criterion, the amount of supply from some suppliers was changed, which indicated the proper functioning of the program and modeling. In both cases, the amount of demand from manufacturers, which was a total of 1134,000 units, was met.
3.3 Multiobjective Integrated Model Results
In the proposed Epsilon constraint method, we first calculated the payoff table, which included the best and worst responses of each objective function, for all three objective functions. In the end, we reached a 3*3 payoff table for the model.
After obtaining the payoff table, one of the objective functions was selected as the main objective function, and the range of changes in the other two objective functions (i.e., the difference between the best and worst responses) was calculated in the mentioned table, and 21 Pareto fronts were obtained. Afterwards, the best Pareto front was selected by the decisionmaker, which is shown in Table 4. In addition, the responses were compared to the optimal response of the multiobjective model. Since the main purpose of this study was to be cautious about higherrisk suppliers, and supply occurred from lowerrisk suppliers, and the model simultaneously sought to minimize supply chain costs, the integrated multiobjective model reduced supply chain risk (second objective function) again while maintaining the values of objective functions one and three with little change compared to the multistage model. It is worth noting that it is very unlikely that an optimal solution was found that could simultaneously optimize all the objective functions defined in the problem. In most cases, the defined objective functions are in conflict with each other in the multiobjective optimization problem.
4. CONCLUSION
The approach presented in the current research can help senior managers of companies achieve competitive advantages. This is mainly due to the fact that, as observed, the amount of supply by the supplier was determined in addition to optimally minimizing the costs of purchasing, transportation, storage, supplier selection, and returned product (penalty). Afterwards, we identified supply chain risks and evaluated and measured them through the FMEA method following receiving experts’ opinions. Following that, risks were minimized in a scenario based model, and the amount of supply by the supplier was modeled by considering the risk criterion. In the end, the mentioned multistage model (three stages) was integrated and solved as a multiobjective model using the Epsilon constraint method. According to the results, the multiobjective model yielded better results, compared to the multistage model, which confirmed the proper performance of the former. Given the wide range of materials discussed and the tools used in this research, the following are proposed for future studies: developing a multi product model or a model with more levels (distributors ring, customers ring), considering periodic and timebased risks and risk of products, considering the cause of return, supply chain resilience and stability, applying the proposed method to different industries to evaluate their performance, and developing largescale problems using accurate, heuristic and metaheuristic methods.