UIO-based sensor fault diagnosis for aero-engine control systems

The problem of sensor fault detection and isolation for aero-engine control systems was studied and an approach to overcome the impact of various disturbances on the performance in diagnosis of the control systems was proposed in this paper. This approach applied unknown input observer (UIO) theory to decouple the disturbance in the aero-engine dynamic model. Then, a bank of UIOs for the sensors in the control system were designed, and a series of residual features for the control system's sensors were extracted. By analyzing the characteristics of magnitudes in the queue of residuals, this paper accomplished the sensor-fault detection and isolation. Numerical simulation results with the disturbances of Gaussian white noise, a moderate change in the operating point of the engine, and non-Gaussian noise show that, the UIO-based method can effectively detect and isolate sensor faults irrespective of the class of the system disturbance, and outperform the Kalman filter-based algorithm in term of diagnosis robustness.