A learning-guided hybrid genetic algorithm and multi-neighborhood search for the integrated process planning and scheduling problem with reconfigurable manufacturing cells

Abstract

Integrated process planning and scheduling (IPPS) is a crucial component of an intelligent manufacturing system. While most existing studies have focused on the manufacturing workshop, less attention has been given to the assembly and test workshops, which typically include reconfigurable manufacturing cells (RMCs). Therefore, this paper focuses on IPPS with reconfigurable manufacturing cells (IPPS_RMCs) in the context of assembly and test workshops. The objective of IPPS_RMCs is to minimize the makespan and total weighted tardiness, taking into account priority constraints and capability conversion limits of RMCs. To address and optimize this problem, a learning-guided hybrid genetic algorithm (LG_HGA) is proposed, which utilizes chromosome encoding to solve the process planning and scheduling problem synchronously. The LG_HGA incorporates NSGA-II as the global search and employs a learning-guided multi-neighborhood search (LG_MNS) to achieve a better balance between exploration and exploitation. In the global search phase, a problem-based methodology for gene operation is introduced. The LG_MNS consists of four neighborhood structures, based on critical paths and heuristic rules. Additionally, the learning-guided mechanism involves using a decision tree regression model to learn data from the knowledge base and determine how to perform local search. Through case tests of various sizes, the experimental results demonstrate that LG_HGA outperforms several advanced multi-objective evolutionary algorithms due to the proposed improved genetic operations, neighborhood structure, and learning mechanism.