A Multi-Objective Memetic Algorithm for Complex Product Assembly Workshop Scheduling Problem with Reconfigurable Manufacturing Units

Abstract

The hybrid flow shop scheduling problem is a typical NP hard workshop scheduling problem that involves parallel machine scheduling and is also common in complex product assembly workshops. However, in complex product assembly workshops, the capabilities of units are dynamically changing, and few studies have simultaneously considered the reconfigurable capabilities of manufacturing units and parallel machine scheduling. Therefore, based on the hybrid flow shop scheduling problem, this article studies the complex product assembly workshop scheduling problem with reconfigurable manufacturing units. In order to adapt to production reality, consideration was also given to setup time, limited configuration resource allocation, energy consumption, and transportation restrictions. The goal is to minimize the makespan, maximum workload of the maximum load manufacturing unit, and total energy consumption . In addition, a multi-objective memetic algorithm (MOMA) was designed, which uses NSGA-II as the global search operator and Hill Climbing algorithm as the local search operator. We also adopted various computing resource allocation strategies to balance exploration and exploitation in MOMA, such as calling local search with fixed probability and controlling local search depth. Meanwhile, a new encoding and decoding method and neighborhood structure were proposed. Last but not least, compared to the six classic multi-objective evolutionary algorithms, the proposed MOMA algorithm is superior. Overall, compared to NSGA-II, the average optimization percentages for each objective were 11.3%, 13.5%, and 9.7%, respectively. The final average IGD and HV values of MOMA increased by 61.2% and 1081.5%, respectively. Case studies also demonstrate that MOMA has good performance in large-scale scheduling problems.

Keywords: Hybrid flow shop scheduling problem, Complex product assembly workshop, Reconfigurable manufacturing units, Energy consumption, Multi-objective memetic algorithm, Computing resource allocation