FAULT DETECTION METHOD COMBINED RESIDUAL CHI-SQUARE AND IMPROVED SPRT ALGORITHM FOR INTEGRATED NAVIGATION SYSTEM

摘要

The fault detection and isolation are very important to the spacecraft for improving the accuracy and reliability of the navigation system. An integrated fault detection algorithm for the integrated navigation system is proposed in this paper to identify multiple fault types, such as soft fault, hard fault and outliers. In this paper, chi square-improved sequential-probability-ratio-test (SPRT) algorithm with moving detection cycle is put forward to accomplish the integrated detection of these three kinds of faults. The residual chi-square algorithm is sensitive to fault which acts directly on the measurements. Traditional SPRT fault detection method reduces real time capabilities and can not determine the fault end time. Aiming at the defects of the two algorithms, this paper presents an integrated fault detection algorithm with higher detection sensitivity for soft fault. If the residual chi-square algorithm determins the fault ends, the improved SPRT algorithm detection value is reset in a timely manner. As a result, the improved SPRT algorithm can still detect the non-first faults. Besides, the moving detection cycle is set to the integrated fault detection algorithm. According to the logic of the moving detection cycle, if the majority of detection values exceed the detection threshold in a detection period, fault has occurred and the faulty subsystem will be isolated. If detection values exceed the threshold just a few times, outliers should be considered and the measurements will be reconstructed. The simulation results show that this integrated algorithm can detect outliers effectively and has good detecting effect for soft fault and hard fault. Therefore, this integrated algorithm can effectively improve the ability and sensitivity of the system fault detection and enhance the reliability of the integrated navigation system.