Designing and implementation of reverse logistics (RL) network which meets the sustainability targets have been a matter of emerging concern for the electronics companies in India.

The present study developed a two-phase model for configuration of sustainable RL network design for an Indian manufacturing company to manage its end-of-life and end-of-use electronic products. The notable feature of the model was the evaluation of facilities under financial, environmental and social considerations and integration of the facility selection decisions with the network design.

In the first phase, an integrated Analytical Hierarchical Process Complex Proportional Assessment methodology was used for the evaluation of the alternative locations in terms of their degree of utility, which in turn was based on the three dimensions of sustainability. In the second phase, the RL network was configured as a bi-objective programming problem, and fuzzy optimisation approach was utilised for obtaining a properly efficient solution to the problem.

The compromised solution attained by the proposed fuzzy model demonstrated that the cost differential for choosing recovery facilities with better environmental and social performance was not significant; therefore, Indian manufacturers must not compromise on the sustainability aspects for facility location decisions.

The results reaffirmed that the bi-objective fuzzy decision-making model can serve as a decision tool for the Indian manufacturers in designing a sustainable RL network. The multi-objective optimisation model captured a reasonable trade-off between the fuzzy goals of minimising the cost of the RL network and maximising the sustainable performance of the facilities chosen.

Resource depletion and e-waste generation through electronics use have reached an alarming stage in developing countries, posing a serious threat to the environment and human health (Wath, Dutt & Chakrabarti

Recent years have witnessed the growing interest of OEMs in redesigning their logistics network for managing the RL activities sustainably. Although deriving maximum economic benefits from the returns is the primary objective of RL, it inherently helps in creating a positive impact on the environment and society. Incorporating strategic recovery decisions within their logistics network can also help manufacturers enhance their bottom line. To begin with the process of implementation, it is fundamental for the OEM to focus on establishing new recovery facilities (RFs) and expanding existing facilities. The selection of location for setting up of facilities across the supply chain (SC) is an important decision, as it entails long-term cost obligations on the firms (Ertuğrul

The rest of the article includes the relevant literature review, the problem definition, and the proposed RL network followed by the proposed methodology. A brief description of the fuzzy programming approach is presented, which is applied to validate the proposed RL model using the data set of a real case study. Furthermore, the results are discussed, and finally, the concluding remarks are provided in the end.

In the last few years, the growing ‘take-back’ laws are challenging the manufacturing firms to redesign their logistics network to incorporate RL into their network. Although there is a plethora of literature on RL network designing (Darbari et al.

Deciding on the locations of warehouses, CCs and repair facilities are key aspects of the strategic plan of reverse SC configuration. However, the decisions regarding the facility locations must not be made in isolation but must be integrated within the RL network design to achieve an efficient recovery system. Many researchers have focussed on developing mathematical models for configuring logistics network design integrating facility location decisions and RL functions (Achillas et al.

The present RL network of the company consists of retail zones each having a designated CC, which is the collection point for the entire zone. All the returns consolidated at the CCs are collected by a 3PRLP, which pays a fixed price for all the returns irrespective of the quality. The 3PRLP then takes responsibility of the returns and the company’s legal duty of handling its returns is fulfilled. Because of customer pressure, market competence and environmental consciousness, the company aims to act more responsibly and design its RL network so as to take control of the processes independently or jointly with service providers. To realise the potential of huge economic benefits from the sale of secondary products, the company plans to perform the processes of collection, inspection, repairing and redistribution to secondary markets (SM) on its own. Strategic decisions are to be made regarding how many new RFs are required and where they should be located. Official dismantlers and recyclers must manage the process of dismantling, recycling and proper disposal as per the laws mandated by the government (Borthakur & Singh

Quick collection and redistribution to ensure a steady flow of returns across every facility.

Choosing suitable RFs, with adequate infrastructure, to carry out the repairing, refurbishing and repacking processes.

The location, number and capacity of RFs are determined within budgetary limits.

The RL network has positive environmental as well as social significance.

The proposed RL network designed to address the above concerns consists of CCs, an integrated dismantling centre (DMC), a SM, a scrap yard and an NGO as illustrated in

Reverse logistics network design.

The flow of returns of the above network begins with the collection and inspection process at CCs. The initial level of inspection determines the buyback value based on the working state, type and age of the model. The returns are then segregated for repair, donation to an NGO, dismantling or sent to the scrap yard. The returns not fit to be reused or donated are to be disassembled for recovery of parts and materials and sent to the DMC for part recovery or to the scrap yard for material recovery or safe disposal. The DMC is third-party owned, and the returns to be dismantled are collected by the 3PRLP from the CCs. Legal recyclers who take care of the discarded returns and parts for valuable material recovery and safe disposal manage the scrap yard.

The methodology adopted for configuring the proposed network can be outlined as follows:

For selection of CCs as RFs, the CCs are evaluated under a comprehensive list of sustainable criteria using the AHP-COPRAS method. The output of the AHP-COPRAS method is the utility of each CC that represents the relative performance of one CC over the other.

A bi-objective fuzzy mathematical model is formulated, which selects the CC to function as RF, determines the level of capacity expansion and allocates each CC to exactly one RF for sending the returns to be repaired. The key decisions are made while capturing a trade-off between the fuzzy objectives of minimising cost of the network and maximising the sustainable performance of the RFs.

The decision regarding the selection of CCs to function as RFs is of strategic importance and has major environmental and social significance because of the expansion of space, job creation and the community development initiatives undertaken.

For this reason, the following sustainable quantitative and qualitative criteria are chosen after an extensive literature survey and intensive interaction with the decision makers: Distance from DMC and centre of gravity (C1), Running cost (C2), Rent Cost (C3), Customer Service Rating (C4), Effectiveness of RL (C5), Environmental Considerations (C6), Technical Qualifications (C7) and Scope for Community Developments (C8) (Ertuğrul

The evaluation process based on the above tangible and intangible criteria is a complex and time-consuming group decision-making problem and an integrated MCDM is proposed, which combines the efficiencies of AHP and COPRAS method. Firstly, the weights of the criteria are determined through AHP (Saaty

The procedure for applying the AHP-COPRAS method is as follows: Suppose there are _{j}^{th} element representing the relative importance of criteria _{max} is the largest eigenvalue of A.

Because there are eight criteria of evaluation, _{1},_{2},…,_{m}

The priority vector is the weighted vector, which represents the importance weights of the criteria.

We now proceed on to evaluate the performance of the alternatives subject to the criteria by applying the following steps of the COPRAS method (Gadakh

_{ij}

_{ij} = y_{ij} w_{j}_{j}

_{ij}_{ij}

_{i}

Greater values of _{i}

_{i}_{max} = max{_{i}

A fuzzy optimisation model is formulated for configuring the proposed RL network using the assumptions and notations given below.

Locations and capacities of the CCs and DMC are known.

The demand of the SM is high.

The cost parameters are deterministic.

Activities outsourced to 3PRLP do not incur any extra cost to the company.

Sets:

Parameters:

^{trans}

^{rep}

^{penalty}

^{budget}

_{ij}

_{i}

_{i}

_{min} minimum threshold of utility

α_{j} percentage of products to be repaired

_{j}

Decision variables:

_{ij}

_{i}

The first objective of the fuzzy multi-objective programming (FMOP) problem minimises the total cost of the network encompassing the cost of transportation (from CCs to RF and from RFs to DMC), cost of repair, fixed and variable cost of expansion and penalty cost for under-utilisation of capacities of selected RFs. As the focus is largely on minimising costs, the profit from selling of repaired products is not considered in the objective. The second objective maximises the overall utility of the CCs selected for expansion, which ensures that CCs with better performance value are chosen.

Constraints:

Equation (

Fuzzy optimisation approach permits adequate solutions of real-world RL network problems having objectives that are normally fuzzy or imprecise in nature and cannot be quantified by crisp mathematical programming approaches. In the FMOP problem (P1) defined above, we have two conflicting objectives to be optimised simultaneously, and clearly, a trade-off is required between the objectives leading to a compromised solution. The fuzziness in the objectives is considered to provide flexibility to the DMs for choosing a preferred efficient solution.

Fundamental to multi-objective optimisation is the concept of efficient and properly efficient solutions. However, if the goals are fuzzy, the concept is extended to fuzzy efficient and fuzzy, properly efficient solutions, defined in terms of membership functions, instead of objective functions (Jiménez & Bilbao

We utilise fuzzy set theory (Bellman & Zadeh _{1}_{2}

Where

A fuzzy efficient solution

Using Zimmermann’s max–min operator approach (Zimmermann

The auxiliary variable α represents the degree of satisfaction to which the objective is satisfied.

The max–min approach used in (P2) ensures that if it has a unique optimal solution, then it is a fuzzy efficient solution of (P1). However, in the case of multiple optimal solutions, not every optimal solution of (P2) is a fuzzy efficient solution of (P1). To overcome the issue, the following weighted problem (P3) can be formulated in an endeavour to finding a fuzzy efficient solution (Tiwari, Dharmar & Rao _{1} and _{2} are weights assigned to _{1} and _{2}, while α_{1} and α_{2}represent their achievement levels, respectively.

An optimal solution of (P3) is a fuzzy, properly efficient solution of (P1) (Refer Lemma 1 in

Although the objectives with more importance are achieved at higher levels using the additive model, the ratio of the levels is not close to the ratio of that of the weights. This is desirable to truly preserve the relative importance of the objectives, defined by the DMs in terms of the weights.

Hence, in this article, the following weighted max–min model (P4) proposed by Lin (_{1}α/_{2}α)’ is the same as the ratio of the weights (_{1}/_{2})’:

An optimal solution of (P4) yields a fuzzy, properly efficient solution of the original FMOP problem (P1).

The fuzzy solution procedure discussed above for solving the proposed model is integrated into the following solution algorithm:

The following numerical example illustrates how the proposed model can be solved using the above solution algorithm.

The case study considered in here is of an electronics manufacturing company located in Delhi National Capital Region, India. The company has 8 CCs located in the region that carry out the initial collection as well as inspection of the returned products. The DMC is situated at Mayapuri, and the CCs are located at East of Kailash (A1), Vaishalli (A2), Noida (A3), Dwarka (A4), Vasant Kunj (A5), Sushant Lok (A6), Karol Bagh (A7) and Model Town (A8) as shown in

Location of collection centres.

For the purpose of evaluation, the relative importance of the criteria is derived using the AHP as shown in

Weights of criteria.

Goal | C1^{−} |
C2^{−} |
C3^{−} |
C4^{+} |
C5^{+} |
C6^{+} |
C7^{+} |
C8^{+} |
Eigenvector |
---|---|---|---|---|---|---|---|---|---|

C1 | 1.00 | 1.00 | 0.50 | 3.00 | 1.00 | 1.00 | 0.50 | 2.00 | 0.120 |

C2 | 1.00 | 1.00 | 0.50 | 2.00 | 0.50 | 0.33 | 0.50 | 2.00 | 0.090 |

C3 | 2.00 | 2.00 | 1.00 | 3.00 | 1.00 | 2.00 | 2.00 | 2.00 | 0.200 |

C4 | 0.33 | 0.50 | 0.33 | 1.00 | 0.50 | 0.50 | 0.50 | 1.00 | 0.060 |

C5 | 1.00 | 2.00 | 1.00 | 2.00 | 1.00 | 1.00 | 1.00 | 2.00 | 0.150 |

C6 | 1.00 | 3.00 | 0.50 | 2.00 | 1.00 | 1.00 | 2.00 | 3.00 | 0.160 |

C7 | 2.00 | 2.00 | 0.50 | 2.00 | 1.00 | 0.50 | 1.00 | 2.00 | 0.130 |

C8 | 0.50 | 0.50 | 0.50 | 1.00 | 0.50 | 0.33 | 0.50 | 1.00 | 0.065 |

These weights are utilised in the initial decision matrix of the COPRAS method as shown by

Initial decision matrix.

Alternatives | C1^{−} |
C2^{−} |
C3^{−} |
C4^{+} |
C5^{+} |
C6^{+} |
C7^{+} |
C8^{+} |
---|---|---|---|---|---|---|---|---|

A1 | 64.8 | 4000 | 110 | 9 | 0.202117 | 0.058518 | 9 | 0.048296 |

A2 | 57.0 | 3000 | 60 | 4 | 0.058200 | 0.203224 | 4 | 0.245266 |

A3 | 63.8 | 3000 | 60 | 7 | 0.074075 | 0.081324 | 6 | 0.138629 |

A4 | 66.5 | 2500 | 35 | 5 | 0.083738 | 0.221617 | 5 | 0.186362 |

A5 | 63.5 | 4000 | 100 | 8 | 0.161560 | 0.120989 | 9 | 0.131778 |

A6 | 85.5 | 3000 | 65 | 7 | 0.104758 | 0.061716 | 8 | 0.096593 |

A7 | 51.0 | 3500 | 70 | 9 | 0.195223 | 0.120837 | 9 | 0.081224 |

A8 | 50.5 | 3000 | 60 | 8 | 0.120330 | 0.131774 | 8 | 0.071852 |

Distance between collection centres.

d_{ij} |
i1 | i2 | i3 | i4 | i5 | i6 | i7 | i8 |
---|---|---|---|---|---|---|---|---|

i1 | 0.0 | 8.0 | 4.0 | 11.0 | 6.0 | 9.7 | 7.0 | 9.0 |

i2 | 8.0 | 0.0 | 4.5 | 16.9 | 12.8 | 17.7 | 9.5 | 9.0 |

i3 | 4.0 | 4.5 | 0.0 | 15.0 | 10.0 | 13.9 | 9.0 | 10.1 |

i4 | 11.0 | 16.9 | 15.0 | 0.0 | 5.3 | 8.3 | 7.9 | 10.4 |

i5 | 6.0 | 12.8 | 10.0 | 5.3 | 0.0 | 6.1 | 6.0 | 9.2 |

i6 | 9.7 | 17.7 | 13.9 | 8.3 | 6.1 | 0.0 | 12.1 | 15.3 |

i7 | 7.2 | 9.5 | 9.0 | 7.9 | 6.0 | 12.1 | 0.0 | 3.1 |

i8 | 9.5 | 9.0 | 10.1 | 10.4 | 9.2 | 15.3 | 3.1 | 0.0 |

Following steps 1–7 of the AHP-COPRAS method, the weighted normalised matrix is derived and the priority value

Normalised weighted decision matrix.

Weights | 0.122 | 0.092 | 0.205 | 0.062 | 0.150 | 0.166 | 0.138 | 0.065 | Sum of beneficiary attributes | Sum of non-beneficiary attributes | Relative significance value | Utility |
---|---|---|---|---|---|---|---|---|---|---|---|---|

-Alternatives | C1^{−} |
C2^{−} |
C3^{−} |
C4^{+} |
C5^{+} |
C6^{+} |
C7^{+} |
C8^{+} |
Sj^{+} |
Sj^{-} |
Qj | Ui |

A1 | 0.016 | 0.014 | 0.040 | 0.010 | 0.030 | 0.010 | 0.021 | 0.003 | 0.074 | 0.070 | 0.112 | 75.3 |

A2 | 0.014 | 0.011 | 0.022 | 0.004 | 0.009 | 0.034 | 0.009 | 0.016 | 0.072 | 0.046 | 0.129 | 86.9 |

A3 | 0.015 | 0.011 | 0.022 | 0.008 | 0.011 | 0.014 | 0.014 | 0.009 | 0.055 | 0.048 | 0.111 | 74.3 |

A4 | 0.016 | 0.009 | 0.013 | 0.005 | 0.013 | 0.037 | 0.012 | 0.012 | 0.078 | 0.037 | 0.149 | 100 |

A5 | 0.015 | 0.014 | 0.037 | 0.009 | 0.024 | 0.020 | 0.021 | 0.009 | 0.083 | 0.066 | 0.123 | 82.6 |

A6 | 0.021 | 0.011 | 0.024 | 0.008 | 0.016 | 0.010 | 0.019 | 0.006 | 0.058 | 0.055 | 0.107 | 71.8 |

A7 | 0.012 | 0.012 | 0.026 | 0.010 | 0.029 | 0.020 | 0.021 | 0.005 | 0.085 | 0.050 | 0.138 | 92.9 |

A8 | 0.012 | 0.011 | 0.022 | 0.009 | 0.018 | 0.022 | 0.019 | 0.005 | 0.072 | 0.044 | 0.131 | 88.2 |

The inference drawn from

The data set provided by the company is as follows: The fixed cost of expansion of each CC is Rs. 150 000, Rs. 100 000, Rs. 120 000, Rs. 100 000, Rs. 110 000, Rs. 130 000, Rs. 130 000 and Rs. 110 000, respectively, while the per square foot cost of expansion is Rs. 110, Rs. 60, Rs. 60, Rs. 35, Rs. 100, Rs. 65, Rs. 70 and Rs. 60, respectively, for each CC. The repair cost per unit product at each CC is Rs. 130. The per kilometre transportation cost is Rs. 10, the penalty cost per unit is Rs. 20 and the total budget for expansion is Rs. 450 000. The lower limits for the capacity of the CCs are 180, 250, 150, 250, 180, 170, 185 and 240, respectively, with the upper limits set as 350, 200, 400, 250, 250, 300 and 260, respectively. The distance of each CC from the dismantling unit at Mayapuri (in km) is 9.4, 13.5, 12.4, 4, 5, 10.5, 4 and 6.4, respectively.

The fuzzy bi-objective model proposed in the study aims at attaining RL network design in which strategic decisions of evaluation and selection of CCs as RFs and their capacity expansion and operational decision of determining the amount of returns to be repaired are integrated. While doing so, the model seeks for a trade-off between the overall cost of the network and the sustainable performance of the selected CCs. Firstly, to understand the nature of conflict between the objectives, the bi-objective problem is first solved as two separate problems using Lingo 11.0, utilising the above data set. The minimum threshold of utility is taken as 70% and at most, three RFs can be opened. The company wants to eliminate the option of selecting CCs with low environmental and social performance; therefore, the minimum threshold for environmental criteria as well as social criteria is taken as 7. Solving for objective 1, which is the cost objective, the model yields a cost of Rs. 1 665 790 and total utility value of 261.22. The CCs selected for expansion are A2, A3 and A4 and the expansion budget utilisation is Rs. 422 200 with a penalty cost of Rs. 21 000. To optimise the cost objective, the model has opted for CCs with lower expansion costs and compromised on their sustainable performance. Maximisation of the second objective leads to a different result. The total utility value of 281.165 is attained at the cost of Rs. 2 088 840 with A4, A7 and A8 selected as RFs. The selected CCs rank first, second and third in the evaluation process and incur a higher cost of expansion. The cost of Rs. 427 900 is required for expansion, bearing a penalty cost of Rs. 23 000. The two single-objective models clearly show the conflict between the goals. This clearly justifies the use of fuzzy programming approach for providing flexibility to the DMs in explicitly adjusting the target values and tolerance levels of the goals. A compromised solution is arrived by developing Lingo code for problem (P4) and by defining appropriate membership functions using the target and tolerance values for the goals. The weights assigned by the DMs are _{1} = 0.3 and _{2} = 0.7. The compromised solution obtained at the DM’s satisfaction level of 0.829 yields a cost of Rs. 1 829 847 and an overall utility value of 279.8565. Thus, the fuzzy model has effectively attained a compromised solution as per DM’s desirability level. To elaborate upon, we analyse the result findings carefully.

The primary focus of the fuzzy model was to optimally select CCs as RFs and determine the number products to be repaired at each RF, so that a trade-off can be attained between the cost and the sustainable performance of the CCs. Because A4 was the common choice in both cases, it has obviously been selected in the final compromised solution as well. The three CCs selected to function as RFs are Vaishalli (A2), Dwarka (A4) and Karol Bagh (A7), with ranks first, second and fourth. The total number of products to be repaired at these RFs are 334, 310 and 201, respectively; thus, a total of 845 products are repaired. A total budget of Rs. 412 600 is needed with a penalty cost of Rs. 4100. At Vaishalli (A2), a total of 334 units from East of Kailash (A1), Vaishalli (A2) and Noida (A3) are repaired; at Dwarka (A4), a total of 310 returns from Dwarka (A4), Vasant Kunj (A5) and Sushant Lok (A6) are repaired, whereas at Karol Bagh (A7), a total of 201 returns from Karol Bagh (A7) and Model Town (A8) are repaired. The total budget required for expansion is Rs. 431 800 with a penalty cost of Rs. 41 000. The results clearly validate the efficiency of the fuzzy multi-objective optimisation model proposed in the study. The model selects the CCs that can function as RFs, determines the number of products that can be repaired and accordingly the capacity of the selected CC is expanded so the burden of the penalty cost is not much. Although the penalty cost has increased, the budget for expansion has been optimally utilised. The main focus of the firm for designing the RL network is to gain economical value from the returns and enhance their sustainable image. A total of 845 products are repaired, which provide substantial economical gains to the firm. Furthermore, the setting up of RFs based on the sustainable criteria ensures that the firm can attain a sustainable recovery channel for handling its returns.

The implications drawn from the result findings can be summarised as follows:

The results reaffirm that electronic manufacturers must not consider adopting RL just as a legal burden and outsource it to 3PRLs, but as an opportunity for collaborating with other reverse SC actors for economic as well as socio-environmental gains.

The compromised solution attained by the proposed fuzzy model demonstrates that the cost differential for choosing RFs with better environmental and social performance is not significant; therefore, manufacturers must not compromise on the sustainability aspects for facility location decisions.

Because the proposed model has a general structure, it can be suitably adopted by other industries in modifying their approach towards RL and incorporating environmental and social considerations at the design phase of the RL network.

Government regulations, customer pressure and market competence are binding factors for electronics manufacturers to design a sustainable recovery model for consumer returns. The manufacturers have also started to realise the potential of RL and consequently are inclined to handle the returns themselves more effectively and responsibly. In this regard, a bi-objective fuzzy decision-making model is proposed in the study, which can serve as a decision tool for the Indian manufacturers in designing a sustainable RL network for managing end-of-life and end-of-use returns. It is a two-phase model where the first phase involves using a combined AHP-COPRAS methodology for evaluation of the CCs for carrying the repair and refurbishment process. The final selection is done in the second phase with the aid of a mixed integer fuzzy linear programming formulation, which effectively determines the number, location and capacity of RFs, the penalty cost and the number of returns to be repaired at each selected facility. Fuzzy programming approach is used effectively in obtaining a properly efficient solution that satisfies the DM’s desired aspiration level for the goals of minimising cost and maximising the sustainable performance of the RFs. The notable feature of the model is the evaluation of CCs under financial, environmental and social considerations and integration of the facility selection decisions with the network designing. This study indicates through the results discussed how an electronic firm in India can earn a sustainable system for their returns in the most cost-efficient way. The study can further elaborate on the nature of collaboration with third-party providers and association with NGOs to develop a RL network that clearly defines and determines the role of each party in the functioning of the RL network.

The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.

All authors have contributed equally to the manuscript.

Consider the following multiple objective programming (MOP) problem (Steuer

Because of the incompatibility of the objectives, a unique feasible solution which optimises all the objectives of the above MOP problem simultaneously does not exist. Generally in real scenario, the DM compromises on choosing an efficient solution to the MOP problem, which is defined as follows:

_{i}_{i}^{*}) ∀i = 1,2,…_{i}_{i}^{*}) for some

_{i}_{i}^{*}) ≤ M(_{r}^{*}) – _{r}_{r}^{*}) > _{r}_{i}_{i}^{*})

The efficient solution represents a compromised solution; however, a properly efficient solution represents a better compromised solution, which is definitely preferred by DMs for solving the MOP problem.

Geoffrion (