Intelligent Fault Diagnosis of Hypersonic Vehicle Based on ResCNN-LSTM-ATT

Accurate fault diagnosis is critical because it has a great impact on operational stability of hypersonic vehicles. Recent trends on various literatures shows that deep learning is a promising methodology to tackle many challenging tasks. In this study, a intelligent fault diagnosis method based on network is proposed for fault diagnosis. The network is constructed by a serial coupling of the one-dimensional Residual Convolution neural networks with Attention mechanism (ResCNN-ATT) and the Long short-term memory networks with Attention mechanism (LSTM-ATT), which is referred to as deep Residual Convolution LSTM attention network (ResCNN-LSTM-ATT). Experiments show that the proposed ResCNN-LSTM-ATT network endows a better ability to capture spatiotemporal correlations, which thus leads to the better accuracy comparing with the fault diagnosis algorithms based on FC-LSTM and ConvLSTM. According to the comparisons, effective improvements are guaranteed by the proposed ResCNN-LSTM-ATT based data-driven fault diagnosis method.