Advanced Prognostics and Health Management for Distributed Plasma Control Systems: A Stochastic Timed Colored Petri Net and Machine Learning-Based Approach

The LingShu plasma control system (PCS) is a distributed, real-time control system designed for tokamak control, featuring a modular architecture and adaptability to heterogeneous hardware. To enhance its reliability and stability, a prognostics and health management (PHM) component was implemented to address health maintenance challenges in distributed environments, where unplanned downtimes can cause delays, data loss, and equipment damage. The PHM component spans the full PCS lifecycle, playing key roles in development, operation, and maintenance phases. During development, a stochastic timed colored Petri net (STCPN) model was introduced to simulate the behavior of complex real-time systems, identifying race conditions and deadlocks in the design phase. This marks the first use of STCPN in fusion control systems, strengthening system robustness and reliability. In operation, the PHM component integrates real-time monitoring, fault diagnosis, and prediction, leveraging LingShu’s modular and plugin-based design for flexible subsystem adaptation. It evaluates system health and reminds PCS to take actions, such as algorithm switching, to ensure stability and minimize downtime. During maintenance, detailed logs and visualization interfaces enable rapid fault localization and troubleshooting, reducing downtime and costs. The PHM component employs decision table-based fault diagnosis, analytic hierarchy process (AHP) inspired health evaluation, and long short-term memory (LSTM) models for resource usage prediction. This approach addresses challenges like resource fluctuations from algorithm switching and “closed box” algorithms, offering dynamic health evaluation standards to enhance monitoring accuracy and system adaptability. Deployed during the 2024 EAST campaign, the PHM component operated continuously, supporting PCS reliability and demonstrating its potential in fusion research.