Multi-Objective Evolutionary Algorithm With Machine Learning and Local Search for an Energy-Efficient Disassembly Line Balancing Problem in Remanufacturing

Abstract

Product disassembly is a vital element of recycling and remanufacturing processes. The disassembly line balancing problem (DLBP), i.e., how to assign a set of tasks to a disassembly workstation, is crucial for a product disassembly process. Based on the importance of energy efficiency in product disassembly and the trend toward green remanufacturing, this study proposes an optimization model for a multi-objective disassembly line balancing problem that aims to minimize the idle rate, smoothness, cost, and energy consumption during the disassembly operation. Due to the complex nature of the optimization problem, a discrete whale optimization algorithm is proposed in this study, which is developed as an extension of the whale optimization algorithm. To enable the algorithm to solve discrete optimization problems, we propose coding and decoding methods that combine the features of DLBP. First of all, the initial disassembly solution is obtained by using K-means clustering to speed up the exchange of individual information. After that, new methods for updating disassembly sequences are developed, in which a local search strategy is introduced to increase the accuracy of the algorithm. Finally, the algorithm is used to solve the disassembly problem of a worm reducer and the first 12 feasible task allocation options in the Pareto frontier are shown. A comparison with typically existing algorithms confirms the high performance of the proposed whale optimization algorithm, which has a good balance of solution quality and efficiency.