통합형 공급망 네트워크 모델 설계와 유전알고리즘 접근법을 이용한 해법

Abstract

In general, supply chain (SC) considers two ways of its flow. First way is a forward logistics (FL) network and second way is called as a backward or reverse logistics (RL) network. In FL network. several components such as suppliers, manufactures, distribution centers, and retailers are considered for each stage of SC. In RL network. some components such as collection centers, recovery centers (remanufacturing centers or refurbishing centers), redistribution centers, waster disposal centers and secondary markets (= markets for used products) are taken into consideration for each stage of SC. Many studies for optimizing FL and RL networks have been performed. However, unfortunately, most of these studies mainly focused on each research direction of FL or RL networks separately. Therefore, the research considering both FL and RL networks is required since most of products is produced and distributed in FL network and the retuned (or used) products is collected,remanufactured, reused or disposed in RL network during their product life cycles. In this paper, a integrated supply chain network (ISCN) model is designed. The ISCN model is consisted of suppliers, manufactures, distribution centers, and retailers in FL network and customers, collection centers, recovery centers, secondary markets and waste disposal centers in RL network. The ISCN model is formulated in a single-objective, non-linear mixed integer programming (SNMIP) model. The SNMIP model is implemented using a hill climbing (HC), simulated annealing (SA), and genetic algorithm (GA) approaches. As a benchmark, LINGO approach is used and its performance is compared with those of HC, SA, and GA. Each approach is programmed on Matlab environment using various types of the ISCN model and compared their performances using various types of measure. The experimental result shows that the GA approach outperforms the HC and SA approaches. For our future study, larger sized ISCN models will be employed and more recent approaches such as cuckoo search, particle swarm optimization and hybrid GA are used for various comparisons.