Fault Diagnosis Method for An Aircraft Attitude Control System based on Deep Learning

This article proposes a deep-learning-based fault diagnosis method for an aircraft attitude control system. The core idea is to construct two deep neural networks to approximate the mapping between actuator fault information and the real-time attitude generated by the dynamics. The proposed fault-diagnosis network consists of an input layer, a convolutional neural network (CNN) layer, a two-layer long and short-term memory network (LSTM), and a fully connected layer in series. The attitude quaternion and the attitude angular velocity are taken as the inputs of the networks, and the estimations of the bias fault size of each actuator are the outputs. Finally, numerical simulations are carried out to verify the proposed method and to make a comparison with the existing diagnose method. The simulation results show the effectiveness of the proposed method and the superiority over the compared method.