A discrete learning fruit fly algorithm based on knowledge for the distributed no-wait flow shop scheduling with due windows

Abstract

The distributed no-wait flow shop scheduling problem with due windows (DNWFSPDW) is a novel and considerable model for modern production chain and large manufacturing industry. The object of total weighted earliness and tardiness (TWETdw) is a common cost indicator in application. A discrete knowledge-guided learning fruit fly optimization algorithm (DKLFOA) is proposed in this study to minimize TWET in DNWFSPDW. A knowledge-based structural initialization method (KNEHdw) is proposed to construct an effective initial solution. In the KNEHdw, the property that the job has no waiting time between processing machines in the no-wait flow shop scheduling problem is abstracted as knowledge to instruct jobs to be placed in possible positions. The swarm center expands from a single individual to an elitist swarm in the vision search stage. A probability knowledge model is established based on the sequence relationship of jobs in the elite population. The feedback information in the iterative process using the probabilistic knowledge model leads the population to search in the direction with a high success rate. The inferior individuals are allocated to the corresponding elite individuals for the local search in the olfactory search stage. The knowledge of weight in due windows is utilized to avoid invalid search during the iteration process. The variable neighborhood descent (VND) strategy is adopted in the local search to enhance the accuracy of the proposed algorithm and jump out of the local optimal. The design of experimental method (DOE) is introduced to calibrate the parameters in the algorithm. The simulation results show that DKLFOA has advantages for solving DNWFSPDW problems comparing with the state-of-the-art algorithms.