Fault Reconstruction Method of Neural Network Observer Group for High-Speed Vehicle

Aiming at the actuator failure of the attitude control system for high-speed vehicle, a fault reconstruction method of observer group based on neural network classifier is introduced. Firstly, the model of the vehicle and the failure model are established. Secondly, a dataset that represents the characteristics of step and sinusoidal fault information is established, which is used to train neural networks in order to classify fault information. Then, according to the classification results, appropriate observers related to distinct fault types for fault reconstruction are selected. For the purpose of solving the problem that different fault types have different performance requirements for observers, an observer group which contains a high-order sliding mode observer and an iterative learning observer is proposed. It can meet high accuracy requirements of step-form faults and fast response requirements of sinusoidal-form faults, so as to realize higher effective fault reconstruction. The classification fault reconstruction method can achieve excellent estimation of angular velocity and fault value. At last, the efficiency of the introduced method is verified by simulation.