Two-sided Disassembly Line Balancing Problem with Sequence-Dependent Setup Time: A Constraint Programming Model and Artificial Bee Colony Algorithm

• TDLBP-SDST is handled for the first time in the literature. • The proposed CP model for the TDLBP obtains the optimal results. • The MILP models and CP approach are proposed to solve the TDLBP-SDST. • Effective four metaheuristic approaches are also developed for the TDLBP-SDST. • TDLBP-SDST aims to minimize mated workstations and workstations, respectively. The large-size products can allow workers to perform tasks on both sides of the line. Hence, a two-sided disassembly line is preferred to ensure several advantages, such as a shorter line. The two-sided disassembly line balancing problem (TDLBP) is relatively new in the literature. This study considers the two-sided disassembly line balancing problem with sequence-dependent setup time (TDLBP-SDST) to reflect the real practice better, as sequence-dependent setup times may exist between tasks in many real-life applications. To the authors’ best knowledge, sequence-dependent setup time has not been considered for the TDLBP in the current literature. The proposed problem creates a more complicated problem. Therefore, proposing effective solution techniques is more critical for obtaining better results. This study proposes two new mixed-integer linear programming models and a novel constraint programming (CP) model to define and solve the TDLBP-SDST. A genetic algorithm, an artificial bee colony algorithm, and the improved versions of these two algorithms are also developed to solve the large-size problems due to the NP-hardness of the TDLBP-SDST. Furthermore, a novel CP model is proposed for the standard TDLBP without considering sequence-dependent setup times. Initially, we compare the performance of the proposed CP model to those of the previous state-of-the-art methods in the literature for the TDLBP without sequence-dependent setup time. The computational results show that the proposed CP model outperforms all the other solution methods and reports the best-known results for all existing benchmark instances for the TDLBP. Then, we present the computational results of the proposed models and algorithms for the TDLBP-SDST. Computational study on a comprehensive set of generated instances indicates that the proposed solution methods effectively solve the TDLBP-SDST.