Active Fault-Tolerant Strategy for Flight Vehicles: Transfer Learning-Based Fault Diagnosis and Fixed-Time Fault-Tolerant Control

In this article, we focus on the issue of active fault-tolerant strategy in the context of hypersonic vehicles. The proposed approach involves addressing the challenges of transfer learning-based fault diagnosis and implementing fixed-time fault-tolerant control. Based on a serial coupling of the 1-D residual convolution neural networks with attention mechanism (ResCNN-ATT) and the long short-term memory networks with attention mechanism (LSTM-ATT), a fault diagnosis deep residual convolution LSTM attention (ResCNN-LSTM-ATT) network is proposed. To deal with the insufficient data fault diagnosis problem, transfer learning technique is utilized based on the constructed ResCNN-LSTM-ATT network. Based on fault diagnosis information, a fixed-time nonsingular terminal sliding mode controller is designed to guarantee system tracking performance in the presence of actuator damage. Simulation results are performed to show the effectiveness of the proposed method based on the hypersonic vehicle model of NASA Langley Research Center.