Sensor fault diagnosis algorithm based on adaptive UKF

To the questions of the abnormal values and the instruments failures for the nonlinear sensor measurement systems, a robust fault diagnosis algorithm based on adaptive unscented Kalman filtering (UKF) was proposed. An adaptive matrix was produced according to the innovation of UKF, then a systems' detector and a parts' detector were built which were made use of restraining the abnormal values and diagnosing the instruments' faults respectively. The innovation of UKF was introduced to status prediction by the adaptive function for modifying the error efforts between the abnormal values and status predictive values and achieving the systems' robust, which was called systems' detector. The innovation was separated into the different sensors' parameters to produce adaptive matrix, which was formed the parts' detector. The trace of adaptive matrix was made use of detecting whether a fault or not and isolating the faults. The simulation results show that the algorithm is robust to the abnormal values, and is accurate to detect the faults from the sensor instruments and compute the range of faults at the same time. The structure of the algorithm is simple and less computational load, and is a good reference for engineering application.