Hybrid heuristic for the location‐inventory‐routing problem in closed‐loop supply chain

• Published date: 01 May 2021
• Author: Zhengwen He,Haoxun Chen,Qunli Yuchi,Nengmin Wang
• DOI: http://doi.org/10.1111/itor.12621
• Date: 01 May 2021

Intl. Trans. in Op. Res. 28 (2021) 1265–1295

DOI: 10.1111/itor.12621

INTERNATIONAL

TRANSACTIONS

IN OPERATIONAL

RESEARCH

Hybrid heuristic for the location-inventory-routing problem in

closed-loop supply chain

Qunli Yuchia,b, Nengmin Wanga,b,∗, Zhengwen Heaand Haoxun Chenc

aSchool of Management, Xi’an Jiaotong University, Xi’an, Shaanxi, China

bERC for Process Mining of Manufacturing Servicesin Shaanxi Province, Xi’an, Shaanxi, China

cIndustrial Systems Optimization Laboratory, Charles Delaunay Institute and UMR CNRS 6281, Universityof

Technologyof Troyes, Troyes, France

E-mail: yuchiqunli@stu.xjtu.edu.cn [Yuchi]; wangnm@mail.xjtu.edu.cn[Wang]; zhengwenhe@mail.xjtu.edu.cn [He];

haoxun.chen@utt.fr [Chen]

Received 6 August2017; received in revised form 29 October 2018; accepted 15 November 2018

Abstract

This study examines a location-inventory-routing problem in a closed-loop supply chain (LIRP-CL) that

considers random demands, as well as random returns, from customers. In reality, some of the returned

products can be remanufactured as new products, and they share the same channel with new products. This

practice renders previous models no longer appropriate. We consider this practice in this paper and build

a new LIRP-CL system. To minimize the total costs of the system, this paper determines the locations of

distribution centers and remanufacturing centers, the inventory level of the system, and the delivery routes

from distribution centers to customers and from customers to remanufacturing centers. A mixed integer

nonlinear model was developed to solve this problem. To solve the model, a novel hybrid heuristic algorithm

based on tabu search and simulated annealing is proposed. The computational results and sensitivity analysis

are presented. In addition, some managerial insights are proposed.

Keywords:reverse logistics; remanufacturing; location-inventory-routing problem; heuristic

1. Introduction

Many industries recognize that globally optimizing supply chains offers substantial cost reductions

(Archetti et al., 2007). Facility locations, vehicle routing decisions and inventory control are im-

portant problems related to the global optimization of supply chain networks. In addition, these

three problems are interdependent. As noted by Salhi and Rand (1989), when ignoring routing de-

cisions, the location decision often results in suboptimal solutions. In addition, the facility location

problem of ignoring inventory control can lead to erroneous system cost estimation (Chen et al.,

∗Corresponding author.

C

2018 The Authors.

International Transactionsin Operational Research C

2018 International Federation ofOperational Research Societies

Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA02148,

USA.

1266 Q. Yuchi et al. / Intl. Trans. in Op. Res.28 (2021) 1265–1295

2011). Javid and Azad (2010) showed that significant savings were achieved when they considered a

combination of these three decisions in a single model. Thus, it is necessary to study the combined

location-inventory-routing problem (LIRP).

Considering different types of facility locations, vehicle routing models, and inventory strategies,

LIRP in forward supply chain networks has been studied in much of the literature. Shen and Qi

(2007) studied a LIRP that applied a vehicle routing model to approximate the shipment from

a warehouse to customers. A nonlinear integer programming model formulation was introduced.

Different from Shen and Qi (2007), Javid and Azad (2010) presented a mixed integer convex model

for LIRP without approximating shipment costs.Then, they introduced a hybrid algorithm based on

tabu search and simulated annealing (TS-SA). Unlike the two studies above, Guerrero et al. (2015)

considered a multiperiod LIRP and used a hybrid algorithm combining a column generation,

Lagrangian relaxation, and local search (LS) to solve the LIRP. Chen et al. (2017) considered LIRP

with time windows in a food distribution network and proposed improved ant colony optimization

with the mutation operation method.

However, different from forward logistics, reverse logistics has emerged, bringing returned prod-

ucts to remanufacturing/refurbishing locations. Reverse logistics not only benefits the environment

but also has the potential to improve cost effectiveness (Ma et al., 2016). Moreover, integrating

traditional forward supply chains and reverse supply chains has arisen due to increasing concerns

about supply chain sustainability. The integrated supply chain is called a closed-loop supply chain.

The differences between LIRP in a closed-loop supply chain (LIRP-CL) and LIRP are as follows:

(1) LIRP only considers the distribution problem of newproducts. LIRP-CL not only considers the

distribution problem of new products, but also considers the return problem of used products; (2)

due to considered used products, LIRP-CL addedreverse logistics on the basis of LIRP; (3) consid-

ering reverse logistic, LIRP-CL needs to consider the establishment of the remanufacturing centers

and the route of the used products. LIRP-CL has been implemented by many companies, such as

Dell, HP, and GM. These companies not only transport products from factories to customers but

also collect used products from customers. Thus, studying LIRP-CL is necessary.

Many researchers have studied facility locations, inventory control, and vehicle routing problems

in closed-loop supply chains. A facility location problem in a closed-loop supply chain network

wassolvedbyEaswaranand ¨

Uster (2010). They adapted Benders’ decomposition method and

used strengthened Benders’ cuts to improve computational efficiencies. Wang and Chen’s (2012)

researched a vehicle routing problem with time windows in a closed-loop supply chain network

and employed a co-evolution genetic algorithm to solve it. Moreover, Mitra (2012) addressed the

inventory management issue in closed-loop supply chains and obtained an optimal return policy.

As mentioned above, facility location, inventory control, and the vehicle routing problem connect

with each other, and considering them separately can result in suboptimal solutions (Prins et al.,

2006). Combinations of these three problems arecreated in closed-loop supply chains, and different

combinations result in different research problems. For example, Yu and Lin (2016) considered

location-routing problems together with simultaneous pickup and delivery, which were solved by a

two-phase simulated annealing (SA) approach. Diabat et al. (2015) used an exact algorithm based

on two-phase Lagrangian relaxationto solve a closed-loop location-inventory problem. To improve

the customer service level, Asl-Najafi et al. (2015) studied a dynamic closed-loop problem and

presented a multi-objective particle swarm optimization (PSO) algorithm to solve the problem.

Zhang and Unnikrishnan (2016) considered six different coordination strategies in a closed-loop

C

2018 The Authors.

International Transactionsin Operational Research C

2018 International Federation ofOperational Research Societies