Mining scheduling knowledge for job shop scheduling problem

The optimal or near-optimal schedules generated by traditional optimization or approximation methods for job shop scheduling problems (JSSP) contain valuable scheduling patterns about this kind of scheduling problems. These patterns could be used to improve the dispatching performance and provide insights into the corresponding scheduling problems. This paper uses timed Petri nets to describe the dispatching processes of the job shop scheduling scenarios. On this basis, a data mining based scheduling knowledge extraction framework is developed to mine the expected scheduling knowledge from the solutions generated by traditional optimization or approximation method for JSSP. Based on this, we show how to use the extracted knowledge as a new dispatching rule to generate complete schedules. A novel method is further developed to combine the extracted knowledge with traditional heuristics to construct new composite dispatching rules which could gain better performance. Besides, we propose a novel approach to utilize the extracted knowledge to improve a Petri net based branch and bound algorithm used in this paper. A series of experiments is carried out to evaluate the performance of the proposed methods.