Acquisition and validation of expert knowledge for high-mix and low-volume production scheduling problems

Abstract

The application of a cyber physical system (CPS) in a production system to improve the efficiency and effectiveness of its operation is called a cyber physical production system (CPPS). A digital twin (DT) is widely regarded as a cornerstone for the realization of CPPS, and many tools, methods and guidelines for developing DT have been proposed in recent years. However, DT alone is not sufficient for solving practical problems that occur in actual production systems (e.g., frequent rescheduling caused by unpredictable events, such as machine failure or engineering changes). Because DT is a digital representation of a physical object, it cannot individually provide a problem-solving procedure for unpredicted situations. Therefore, the integrated utilization of human experts’ knowledge with DT is indispensable for the realization of CPPS. This paper proposes a systematic method for acquiring experts’ knowledge that can be integrated with a DT to construct a production system that is robust against unpredicted changes in a production environment. In particular, the study focuses on the knowledge acquisition for the high-mix and low-volume production scheduling problem. This is because quick modification of the existing schedule considering complicated constraints among processes, resources, delivery time and so forth is regarded as a typical problem that only human experts can solve. Through a case study in the die machining industry, this study demonstrates the validity and limitations of the method and concludes that the proposed method is effective for acquiring experts’ knowledge.