Deep Transfer Learning Based on Convolutional Neural Networks for Intelligent Fault Diagnosis of Spacecraft

The on-orbit operation and all-weather duty of spacecraft lead to the rapid increase of test data, which brings severe challenges to fault diagnosis. In recent years, deep learning has shown excellent performance in feature extraction and pattern recognition. More and more attentions have been paid to the application of deep learning in spacecraft fault diagnosis. However, the success of deep learning largely relies on sufficient labeled data. Due to the high reliability of spacecraft, the test data usually contains lots of normal sample points, while the faulty sample points are extremely scarce. Firstly, this paper designs a fault diagnosis model based on 1-D convolutional neural network, which directly processes the 1-D raw data and extracts features; then, for the first time, the transfer learning technology is introduced into the field of spacecraft fault diagnosis, and a domain adaptive deep transfer model is proposed. MMD is used to reduce the discrepancy of data probability distribution between the source and target domain. The results of the experiments show that the proposed model could accurately diagnose and identify the faults of spacecraft in different application scenarios.