Intermittent Fault Diagnosis of Dynamic Systems with Model Uncertainty and Disturbance: An Adaptive Nondeterministic Observer Approach

Abstract

Intermittent fault (IF) threatens the safety for modern dynamic systems severely. The target of this work is to swiftly detect the occurrence and relievement of IFs and accurately gauge their severity characterized by unknown values, enabling effective intervention, and rectification if necessary. An innovative adaptive nondeterministic observer-based technique for diagnosing IFs for a class of dynamic systems subject to model uncertainty and external disturbances is proposed. The term “nondeterministic” underscores the dynamic nature of threshold function and gain matrices for IF detection, as well as regressor for IF identification. For the IF detection task, a time-varying threshold function and a pattern-matched gain technique, ensuring both timely and precise IF occurrence and relievement signals, are proposed. For the IF identification task, an event-rectified regressor-based algorithm is proposed to deliver unbiased estimations of IFs. Moreover, a compelling illustrative example applying the proposed whole diagnosis scheme to an analog circuit system is given, showcasing the efficacy in addressing the challenges of diagnosing IFs in dynamic systems affected by uncertainty and disturbance.