Batch processing machine scheduling problems using a self-adaptive approach based on dynamic programming

Abstract

With the increasing trend of smart electronic devices, interlinked industries face various challenges in meeting market demand. The demand for customized small-batch and multi-variety products with agility in customer expectations makes the scheduling problem more complex. Batch-processing machine (BPM) scheduling refers to managing and organizing the execution of a group of tasks or jobs on a machine. BPM scheduling is a complex optimization problem critical in semiconductor production systems industries. A single BPM scheduling problem, considering multiple jobs with different sizes, release times, processing times, and due dates to minimize total earliness and tardiness, is studied in this paper. A mixed integer programming model is formulated to express the problem, including the related constraints. The self-adaptive hybrid differential evolution and tabu search (SDETS) algorithm with dynamic programming is proposed to solve the BPM-scheduling problem. The novel SDETS algorithm is embedded with four additional features: a) dynamic programming-based batch formation; b) right-left-shifting rules to identify the starting time of each batch; c) DE-self-adaptive mutation strategy to determine the job sequence and trade-off between exploration and exploitation; d) introduction of tabu-search to enhance the convergence rate. A comprehensive parametric tuning of the algorithms is conducted to optimize the performance and enhance the suitability for the specific problem set case instances. The findings suggest that the proposed algorithm surpasses the performance of the compared algorithms. Moreover, the SDETS method exhibits high convergence to find more precise and globally optimal solutions for large-scale problem instances, further emphasizing its practical applicability.