A flexible integrated forward/reverse logistics model with random path

Abstract

This dissertation focuses on the structure of a particular logistics network design problem, one that is a major strategic issue for supply chain design and management. Nowadays, the design of the supply chain network must allow for operation at the lowest cost, while providing the best customer service and accounting for environmental protection. Due to business and environmental issues, industrial players are under pressure to take back used products. Moreover, the significance of transportation costs and customer satisfaction spurs an interest in developing a flexible network design model. To this end, in this study, we attempt to include this reverse flow through an integrated design of a forward/reverse supply chain network design, that avoids the sub-optimal solutions derived from separated designs. We formulate a cyclic, seven-stage, logistics network problem as an NP-hard mixed integer linear programming (MILP) model. This integrated, multi-stage model is enriched by using a complete delivery graph in forward flow, which makes the problem more complex. As these kinds of problems belong to the category of NP-hard problems, traditional approaches fail to find an optimal solution in sufficiently short time. Furthermore, considering an integrated design and flexibility at the same time makes the logistics network problem even more complex, and makes it even less likely, if not impossible, for a traditional approach to provide solution within an acceptable time frame. Hence, researchers develop efficient non-traditional techniques for the large-term operation of the whole supply chain. These techniques provide near optimal solutions particularly for large scale test problems. In our case within this thesis, to find a near optimal solution, we apply a Memetic Algorithm with a neighborhood search mechanism and a novel chromosome representation called "extended random path direct encoding method" which includes two segments. Chromosome representation is one of the main issues that can affect the performance of a Memetic Algorithm. To illustrate the performance of the proposed Memetic Algorithm, LINGO optimization software as commercial package serves as a comparison for small size problems. We show that the proposed algorithm is able to efficiently find a good solution for the flexible, integrated, logistics network. Each algorithm has some parameters that need to be investigated to provide the best performance. In this regard, the effect of different parameters on the behavior of the proposed meta-heuristic algorithm is surveyed first. Then, the Taguchi method is adapted to identify the most important parameters and rank the latter. Additionally, Taguchi method is applied to identify the optimum operating condition of the proposed Memetic Algorithm to improve the results. In this study, four factors that are defined inputs of the proposed Memetic Algorithm, namely: population size, cross over rate, local search iteration, and number of iterations are considered. The analysis of the parameters and the improvement in results are both illustrated by a numerical case studies. Finally, to show the performance of the Memetic Algorithm, a Genetic Algorithm - as a second meta-heuristic algorithm option - is considered as regards large size cases.