An improved multi-objective multifactorial evolutionary algorithm for assembly line balancing problem considering regular production and preventive maintenance scenarios

Abstract

• ALBP considering PM scenarios promotes the productivity and smoothness. • Multi-task optimization benefits the simultaneous optimization of all scenarios. • The matrix reduction method is proposed to reduce the decoding time. • The neighborhood search operator is designed to expedite the convergence speed. • Statistical results indicate IMO-MFEA with two improvements outperforms others. Assembly line balancing considering regular production and preventive maintenance scenarios (ALBP_RP/PM) is a new and effective way to simultaneously handle the assembly production and equipment maintenance. It generates multiple allocation plans in advance. If a workstation is to be maintained, the corresponding allocation plan is applied to bypass this workstation and enable production continuity within the assembly line. However, the ALBP_RP/PM was generally solved as a single-task optimization problem, implying the coupling relationship of allocation plans under different scenarios was essentially ignored. For this reason, it is hereafter treated as a multi-task optimization problem under the premise that the assembly line balancing under each scenario is an independent task. On this basis, an improved multi-objective multifactorial evolutionary algorithm with two extra improvements is proposed to minimize cycle times and operation alterations. Specifically, the native evolutionary operators are reformulated to encourage the inter-scenario sharing of effective knowledge fragments among allocation plans to accelerate the simultaneous optimization of all the tasks; a matrix reduction method is proposed to reduce the decoding time in generating operation sequence; a neighborhood search operator is designed to strengthen the convergence performance of the algorithm. Comprehensive computational results demonstrate that the inter-scenario sharing mechanism in the multi-task optimization expedites the convergence speed and improves the operation allocation plans effectively. Besides, the proposed algorithm clearly outperforms the other five single-task optimization algorithms and four multi-task ones under the same computational limits.