A multi-objective memetic algorithm for mixed-model two-sided disassembly line balancing problem

Environmental impact is increasing day by day with the growing population, affluence level and technological developments. Today, as a result of this increasing, the level of Carbon Dioxide has exceeded five hundred parts per million (ppm). In order to overcome environmental impacts such as the increased level of Carbon Dioxide, it is necessary to change the classical production/consumption approach. For this, Life Cycle Engineering (LCE) is recommended as a new approach. LCE is an approach that aims to minimize the environmental impact caused by the product, production or consumption by being involved in every step of the product life cycle. An important step of LCE is to design an easily disassemble, recyclable product at the design stage of product life cycle in order to minimize the environmental impact it will create when the products are at the end of their life as End-of-Life (EOL) product. Considering that the designed products will become EOL products after a while, the design of efficient disassembly systems is of critical importance in order to minimize the environmental impact. The Disassembly Lines that emerged in this context allow the efficient disassembly of the EOL products in bulk. Maximizing the efficiency of installed disassembly lines is critical to minimizing environmental impact. For this reason, the Disassembly Line Balancing (DLB) problem emerged in 1999 in order to optimize various performance criteria. The DLB problem has found wide coverage in the literature, and a lot of research has been done. However, the number of studies for the Two-Sided Disassembly Line Balancing (TDLB) problem, which is a line design that minimizes costs such as the cost of rotating on the line for disassembly of large-size products, is very few. In particular, the Mixed Model Two-Sided Disassembly Line Balancing (MTDLB) problem, in which the number of models is more than one, was introduced for the first time within the scope of this thesis. In this thesis, the MTDLB problem has been investigated in order to minimize the sum of the design cost, the number of stations and sum of the selected disassembly task times of the disassembly line, which allows more than one similar or dissimilar products to be disassembled on the two-sided disassembly line. The Transformed AND/OR Graph (TAOG) precedence relations diagram is used for the MTDLB problem. The TAOG precedence relations diagram is a precedence diagram that takes into account all possible disassembly sequences in the event that only the physical conditions are considered due to the nature of the disassembly, and the functionality is not important. A Mixed Integer Linear Programming (MILP) based mathematical model has been developed to optimize the performance measures determined for the MTDLB problem. When the objective function of the developed mathematical model is hierarchical, it gives optimum results for small and medium sized cases with Gurobi solver. However, for the solution of large-size cases, the solution time is growing considerably. Therefore, a Multi-Objective Memetic Algorithm (MOMA) has been proposed for the solution of the MTDLB problem. The proposed approach is tested on a series of test problems and the results are compared with the highly preferred algorithms for multi-objective optimization in the literature. The results obtained show that the results of the MOMA algorithm, investigated for the MTDLB problem, give better results than the other algorithms.e