A matheuristic-based self-learning evolutionary algorithm for lot streaming hybrid flow shop group scheduling with limited auxiliary modules

Group scheduling enhances production flexibility and efficiency in mass customization However, it overlooks differences of due dates in customized orders and functional/quantity constraints of molds. Therefore, lot streaming and module assignment strategies are needed. To address this, this paper investigates the lot streaming hybrid flow shop group scheduling with limited auxiliary module constraints(HFGSP_LSAM). To minimize the total weighted tardiness and makespan, a new mixed integer linear programming model and a matheuristic-based self-learning evolutionary algorithm(MSEA) are proposed. This algorithm develops a new matheuristic-based hybrid initialization to generate better initial solutions. A double layer self-learning evolution is developed to collaborate operators which include six knowledge-based local search operators and six global crossover operators. The experimental study, based on 360 small and 960 large instances, demonstrates that the matheuristic-based hybrid initialization and double layer self-learning evolution can enhance 84% and 13% performance of MSEA, as well as the proposed MSEA is superior to other well known algorithms in solving HFGSP_LSAM. An industrial case study is conducted to confirm the superiority of MSEA and provide two recommendations for managers to balance production efficiency and due dates.