A modified method for fault detection and isolation of redundant inerial measurement unit in dynamic environment

Abstract

According to the reliability and safety requirement for aircraft, fault detection and isolation (FDI) is essential to improve the fault tolerant performance of the navigation system. However, conventional principal component analysis (PCA) method does not consider the fault detection of the dynamic system and a high rate of false alarm rate. This paper proposes a modified method combined both superiority of generalized likelihood test (GLT)and principal component analysis (PCA). Firstly, this method through comparing the distribution of the projected points in the feature plane of the sensors data and historical normal navigation data detects the fault. Secondly, determines the time interval of failure by analyzing the length of each projection. Finally we compute the contribution of the variables to the statistic Q to carry out fault isolation. Monte Carlo simulation is carried out in order to verify the validity of the algorithm. Results show that this method can isolate the influence of the vehicle's own vibration on the sensors data effectively and can reduce the false alarm rate. Furthermore, this method can correctly detect and isolate the 2.4e — 4rad/ S amplitude of the fault while the noise level is 1e — 6. Consequently, this method can provide a new theoretical reference for the fault detection of aircraft redundant inertial measurement unit (RIMU).