Model-based diagnosis of chaotic vibration signals

This paper presents a model-based approach to online monitoring and fault diagnosis of rotating machinery. Fault (e.g., rub, imbalance) modes of rotating machines are classified using nonlinear dynamic models with quasi-periodic and chaotic behavior. The paper identifies a class of fault scenario under which the well-accepted nonlinear state filters (e.g., EKF) cannot be used to monitor or diagnose the machinery. An effective on-line model-based monitoring and diagnosis algorithm is proposed. The algorithm is based on computationally efficient algorithms for signal processing and parameter identification.