Aircraft sensor fault diagnosis based on Kalman filter innovation sequence

Abstract

Methods of testing the agreement between the innovation sequence and white noise, and the detection of any change of its mathematical expectation have been discussed by Willsky (1976) and Himmelblau (1978). In this paper a novel approach based on the innovation sequence of Kalman filter is introduced for fault detection and isolation, and the method is applied to an aircraft model. The faults are assumed as changes in the covariance matrix of the sensor measurements of the aircraft. There is no real-time method for detecting faults affecting the covariance matrix. The method presented in this paper can quickly detect that type of faults.