Research on discrete intelligent workshop lot-streaming scheduling with variable sublots under engineer to order

Abstract

Engineer to order is a highly customized production pattern. A flexible job shop scheduling problem with lot streaming is introduced in the discrete intelligent workshop, which considers setup time, changeover time, and single-piece flowing workpieces under engineer to order. For lot streaming, the advantages of variable sublots are analyzed by comparing different lot splitting strategies. Upon variable sublots, a mixed-integer linear programming model is formulated, but gets restricted in solving large-size problems. Then a hyper-heuristic improved genetic algorithm consisting of high and low levels is proposed. An improved niche genetic algorithm is employed for the high level due to its excellent global search performance, while the low level encapsulates the particle swarm optimization as its perturbation operator. Furthermore, several problem-oriented components are designed to enhance its performance, for example, a high-low level coordination mechanism with clearly global–local division, a decoding method based on greedy and saving rules, a self-adaptive earliest start time, and a population initialization and resetting method. Finally, experiments are organized, and the result shows the proposed algorithm performs up to 24.46% better than the reference group in solving benchmark instances, for which several components are tested to distinguish their main contributions. The general computational performance is also examined through the CEC’17 test suite. For enterprise examples with medium to large sizes, an average 99.52% operation rate demonstrates its high practicality.