Conservativeness-Reduced Fault Diagnosis of Aeroengine Sensor Fault Considered Multi-Source Uncertainty

The aeroengine operating process is exposed to the multi-source uncertain environment consisting of aeroengine epistemic uncertainty and control system stochastic uncertainty. In order to extend the applicability of the diagnosis system in complex environments by quantifying the statistical properties of stochastic models based on polynomial chaos expansion (PCE), this paper extends the linear Kalman filter (LKF) to achieve optimal filtering under multi-source uncertainty. Next, with the introduction of the adaptive weighting matrix and the accompanying standardized threshold, the sensitivity of the diagnosis system to minor faults is greatly improved. Based on the dimension-reduced filter bank, the conservativeness-reduced fault detection and isolation (FDI) is finally achieved. Numerical simulations demonstrate that the proposed optimal filtering is efficient under multi-source uncertainty. Compared with the traditional constant weighting matrix-based FDI system, the proposed adaptive weighting matrix-based FDI system is less conservative, more sensitive to minor faults, and can better ensure the security and reliability of aeroengine under multi-source uncertainty.