The Semaphore Identification and Fault Troubleshooting Modus for Spacecraft Originating from Deep Learning and RF Method

Abstract

The spacecraft environment and reliability test brings the huge type of data, traditional algorithm has many shortcomings when handling big data especially huge storage space and massive compute time. According to the deep learning network, the writer design a algorithm with the ability that gain the initial parameters of the multi-layer neutral network. After initialize the parameters with gradient descent method, we will select the parameters to help to classifying information more easily. By researching generated data, there is expert information which could help to build a platform to manage the health of spacecraft. Experimental data show that the deep neural network algorithm could be proved to achieve a desired accuracy through classification with multi-kind of spacecraft data environment and reliability experiment.