Nonlinear fault diagnosis and localization of dual rotor-bearing-casing system based on feature parameter identification

Considering complex structure of aero-engine rotor systems and the wide variety of faults, fault localization and identification face enormous challenges. According to previous studies, traditional feature extraction fault diagnosis methods are difficult to effectively diagnose and locate faults, and diagnosis methods based on data and artificial intelligence rely on a large amount of fault data, which is difficult to collect considering the actual test conditions of aeroengine. By combining the dynamical model and feature parameters of fault, parameter identification using limited measured sensor data is an effective and practical technique for fault diagnosis. This study proposes a fault diagnosis technique based on fault feature parameters identification using augmented extended Kalman filter (AEKF). A reduced order dynamical finite element model of an aeroengine dual-rotor-casing system is built to improve computational efficiency and a weighted overall iteration technique is employed to accelerate parameter stabilization and convergence. The AEKF is then employed to approach the complexities associated with the nonlinearity of the system model. The accuracy of the finite element model and fault identification method was verified by experiment. A numerical test is furthermore adopted for the aero-engine dual-rotor-casing system and the multiple fault feature parameters are identified, and the fault and location corresponding to each feature parameter are hence identified and localized. The results clearly demonstrate that the method can quickly identify linear and nonlinear faults of the rotor system and maintain the accuracy of fault parameter identification when the fault parameters change.