Small UAV sensor fault detection and signal reconstruction

The method using least squares support vector machine (LS_SVM) and principal component analysis (PCA) for small UAV angular rate sensor failure detection, isolation and reconstruction was proposed. LS_SVM was used to establish the prediction model and sensor fault was diagnosed by generating residuals. PCA was used to isolate the sensor fault signal. According to the detection results, the fault signal could be replaced by the LS_SVM estimation value. The simulation results show that, the signal reconstruction accuracy, can guarantee the UAV flight performance in a safe range, the method was proved to have high reliability and stability.