Real-time order acceptance and scheduling for data-enabled permutation flow shops: Bilevel interactive optimization with nonlinear integer programming

Abstract

With the fourth-generation industrial revolution, manufacturing industries are focusing on dynamic, fully autonomous, and more customer-oriented production systems. This customer-oriented change converts classically static customer demand into that which is dynamic and real-time, as no prior information regarding customer demand is known in advance. This paper focuses on real-time order acceptance and scheduling (r-OAS) for a data-enabled permutation flow shop. To compensate for the shortage in prevailing approaches that make bottleneck-based decisions or assume that the intermediate buffers among workstations are infinite, an r-OAS scheme is generated based on a data-driven representation, which can concisely predict the dynamic production status of flow shops and the corresponding makespan of a job with finite intermediate buffer constraints. Using this representation, real-time job release planning (r-JRP) can be coupled with r-OAS to minimize various operational costs of flow shops (i.e., the costs of the work-in-process, earliness, and tardiness). In terms of the inherent interactive mechanism between r-OAS and r-JRP, in which r-OAS generates a decision space for r-JRP and r-JRP then feeds the lowest operational costs back for use in r-OAS decision-making, a bilevel interactive optimization (BIO) is formulated to simultaneously address the two subproblems based on the Stackelberg game. The r-OAS acts as the leader, while r-JRP acts as the follower. The BIO is a type of nonlinear integer programming, and a bilevel tabu-enumeration heuristic algorithm is developed to solve it. The efficiency of the BIO is verified through a practical case study. The results show that the BIO can increase the net revenue of flow shops by 2.97%, compared to the bottleneck-based approach, and by 2.45% and 0.92%, respectively, compared to step-by-step methodologies.