A systematic-theoretic analysis of data-driven throughput bottleneck detection of production systems

Throughput is one of the most critical performance indices for design, control, and operation management of production systems. Throughput bottleneck greatly impedes the overall performance of modern production systems. However, detecting throughput bottlenecks of production systems is a complicated task due to the complexity of production system dynamics. In this paper, a new data-driven bottleneck detection method is proposed based on rigorous mathematical proof for general serial production systems, which uses the routinely available industrial data on the plant floor to identify the throughput bottleneck location within production systems in both the short-term (transient) and long-term (steady-state) periods. Case studies are conducted to illustrate the effectiveness of the proposed method. The research outcomes will enhance the intelligent decision-making and real-time operation management capabilities in the modern manufacturing enterprises.