Study on signal recognition and diagnosis for spacecraft based on deep learning method

According to the large variety of data generated during the spacecraft test and fault diagnosis, this paper designs a multi class classification algorithm based on deep learning method. The algorithm uses the stack auto-Encoder to initialize the initial weights and offsets of the multi-layer neural network, and then monitor the parameters after the initialization with the gradient descent method. The algorithm can overcome many weaknesses of SVM, for example, it is too complex and occupied more space when data is large or the categories are huge. By studying the measured data, the expert knowledge can be provided for the spacecraft health management platform. Experimental data show that this depth learning algorithm can get a high accuracy in the classification of multi class signals of spacecraft.