Sensor fault diagnosis for flight control system based on Cubature Kalman filter

This paper aims to provide a feasible scheme to detect sensor faults and reconstruct signals for flight control system. An approach which combines Cubature Kalman Filter (CKF) with CKF-based nonlinear unknown input observer (NUIO-CKF) is proposed to generate residuals. And the method of Sequential Probability Ratio Test (SPRT) is introduced to detect sensor faults. This design can overcome the shortcomings of using single filtering method and increase the accuracy of fault detection. In order to reconstruct the right state signal under sensor failure conditions, a joint estimation method of state and fault based on CKF is proposed. Due to CKF's excellent nonlinear tracking performance, sensor fault can be estimated and the right signal can be reconstructed by taking fault signal as an extended state. The simulation results on the aircraft longitudinal model with typical sensor failure modes (jam fault and gain fault) demonstrate the effectiveness of the proposed methods.