Multi-objective disassembly line balancing problem with parallel stations

In this study, we have developed solution methods for the multi-objective disassembly line balancing problem (DLBP) with parallel stations that is a common problem in many real life applications, but rarely address in DLBP literature. The existing concept of parallel stations in assembly line increase production rates and improve flexibility in designing lines. In this study, disassembly and assembly line balancing problems are considered complementary of each other and the disassembly tasks and precedence relations among them are defined using AND/OR Graph (AOG).First, a deterministic DLBP with parallel stations is considered and a mixed integer programming model of the problem is proposed. In addition, a goal programming and a fuzzy goal programming model based on the proposed mixed integer programming model of the problem are presented. Such multi-objective decision making approaches have not been studied in DLBP literature before. A small size sample problem is solved and a computational study is carried out on generated test problems to investigate the efficiency of these models. The proposed goal programming models enable decision makers to simultaneously consider many objectives in precise and fuzzy environments as well as increasing the flexibility of reproduction systems.In reality, disassembly task times may have a significant variation depending on the physical state of the returned products. Thus, in this study, the dissasemly task times are considered stochastic and a non-linear binary integer programming model for multi-objective optimization of stochastic DLBP is proposed. The objective of this model is both to minimize the line balance and design cost of the line. Since the most basic form of DLBP is NP-hard, the inclusion of stochastic task times, different line designs and simultaneously considered conflicting objectives increase its complexity. Thus, this study proposes a new solution procedure based on genetic algorithms (GA) to find the set of Pareto-optimal solutions for multi-objective DLBP. In order to reach more Pareto-optimal solutions and to improve the quality of such solutions, several properties such as different fitness evaluation approaches, repair algorithms and a diversification strategy are implemented in the proposed solution procedure and their effects on the performance of GA are investigated. To measure the effectiveness of the proposed GA, a goal programming model based on the proposed non-linear binary integer programming model for multi-objective optimization of stochastic DLBP is developed. GA and goal programmimg model are tested on different sized problems and the results are compared to each other. Computational results show that high quality solutions on test problems are obtained in a reasonably short time with the proposed GA.