A data driven fault detection and isolation scheme for UAV flight control system

Fault detection and isolation (FDI) is critical for the safe flight of unmanned aerial vehicle (UAV). As it is difficult to build accurate physical model for UAV flight control systems, data driven models such as principal component analysis (PCA) have been applied in FDI of UAV. However, PCA based methods cannot take dynamic relations of inputs and outputs into consideration. In this paper, a dynamic PCA model combined with moving average technique is proposed for FDI in UAV, which is more suitable for dynamic systems. Furthermore, contribution analysis is firstly utilized to implement fault isolation of UAV. The case study on a simulated fixed-wing UAV control system under steady flight demonstrates the effectiveness of the proposed method.