Deep Learning Based Active Fault-Tolerant Control for Missile Actuators

For a missile controlled by fork rudder, if the surface fails in the course of guidance, it is necessary to diagnose and deal with the failure in time to ensure the missile hits the target accurately. In this paper, a fault diagnosis and active fault-tolerance control scheme for missile surface based on deep learning is proposed. Firstly, the missile's speed, attitude and other parameters are taken as the training data, and the corresponding surface deflection angle is taken as the label. Then, the prediction model of surface is obtained through deep learning training. In the actual operation of the missile, the corresponding parameters are obtained from the sensor, and the predicted surface angle is obtained from the training model, which is compared with the actual surface angle, as the trigger condition of active fault-tolerant control. The simulation results show that the scheme can correctly identify the faults and reconstruct the deflection instructions of the rest surfaces, so as to ensure that missile can strike the target even if one of the surfaces malfunctions.