A Genetic Programming-Based Iterative Approach for the Integrated Process Planning and Scheduling Problem

Abstract

The integrated process planning and scheduling (IPPS) problem is studied in this article, in which operation sequencing, process plan selection, and machine selection are decided simultaneously. For different scenarios, three mixed-integer linear programming (MILP) models are designed. Then, in view of the workload of machines and processing times of jobs, two machine selection techniques are introduced to simplify the optimization of these MILP models. By exploring the structural properties of the MILP models, we put forward a novel lower bound to act as a measurement for the performance of the related algorithms. Considering the real-time requirement and complexity of instances in practice, we design a hybrid greedy heuristic based on a new decision structure of the problem and dispatching rules. Furthermore, in order to create effective dispatching rules to improve the hybrid greedy heuristic, enhanced genetic programming (GP)-based iterative approach is proposed. Experimental results indicate that our approaches are better than other available approaches for the IPPS problem and can reduce the computational time while providing high-quality solutions. <i>Note to Practitioners</i>&#x2014;Due to the widespread use of computer numerical control machines and flexible manufacturing systems in the modern manufacturing industry, the IPPS problem has been widely recognized among both academia and industry over the years. So far, many scholars have put forward several approaches to solve this problem. However, for the instances in practical applications, good solutions are still difficult to get in a reasonable time. The rule-based heuristics, on the other hand, have the advantages of practicability and high efficiency, are much more preferred in practice. Nonetheless, it not only takes plenty of time but requires domain expertise to construct valid dispatching rules for complicated production problems. As an evolution-based machine learning approach, GP has been widely utilized to create dispatching rules for certain scheduling problems. Therefore, in this study, an enhanced GP-based iterative approach is proposed to create dispatching rules for the rule-based heuristic. The computational effectiveness and efficiency of the heuristic based on dispatching rules created by GP are verified in experimental results, and it opens the way for handling the IPPS problem in dynamic environments and multiobjective systems in practice.