A real-time rescheduling heuristic using decentralized knowledge-based decisions for flexible flow shops with unrelated parallel machines

Abstract

In a manufacturing planning and control system, a change of system environment or of the production requirements may invalidate the current production schedule. In that case, rescheduling as a self-adaption function of the system is necessary for generating a new schedule, regarding the current state of the production system. This rescheduling process is time critical and normally requires real time solutions. In this paper we present a rescheduling approach with offline self-learning and online self-decision-making abilities. It solves the rescheduling problem of flexible flow shops (FFS) with unrelated parallel machines. The optimality criterion is the makespan. The approach uses a centralized heuristic to guarantee the generation of active schedules. In addition, it integrates a decentralized knowledge-based decision making system in the heuristic. This decision making system can learn from previous scheduling problems and their schedules. Consequently, it uses the obtained knowledge to dynamically select the most appropriate dispatching rule for scheduling the production, depending on the current system state. Computational results show that the proposed approach is superior to only using one single dispatching rule constantly. Furthermore, due to its efficient runtime the approach is suitable for real time applications.