Performance-Based Fault Detection for Model-Free Automatic Control Systems

This paper is dedicated to the performance-based fault detection problem for model-free automatic control systems, where the infinite-horizon quadratic index is adopted to describe the concerned system performance. Different from most of the observed-based fault detection methods, the designed performance-based approach mainly focuses on the performance degradations induced by faults. To be specific, the performance index embedded with the information of system dynamics is first reformulated. The performance residual, which is adopted as the evaluation function, is derived based on the Bellman equation. In order to remove the requirements for accurate system model and full state measurements, the performance residual is further represented by the input and output sequences. The relevant unknown matrix is identified based on the measurable process data. Finally, the validity of the theoretical results is demonstrated by the simulation studies on a ship propulsion system.