Study Of Spacecrafts Multi-Fault Diagnosis Based On Neural Network

In order to make deep space exploration better, spacecraft fault diagnosis system must be able to deal with multi-faults. This has a big relationship with the safety of spacecrafts and the success of deep space exploration tasks. Neural network not only has the ability of dealing with complex pattern, associating, extrapolating and memorizing, but also has strong ability of self learning. It can catch up on the fault that heuristic rule can't make diagnosis conclusions because of its noncompleteness. So neural network is appropriate to the fault diagnosis systems.This dissertation makes a research on the technique of the spacecraft fault diagnosis with neural network. First, it researches the category and design process of the fault diagnosis model based on neural network. The model includes 6 categories and 5 design processes. And then it holds forth a three layers diagnosis model based on BP network. We design the model according to the 5 processes above, and give out the input, output, important parameters. At last, it trains and simulates the three layers diagnosis model with MATLAB, and analysis the result of simulation. The result is satisfying.The result of the experiment reaches the goal, which declares that the model is valuable in actual use. But because the experiment condition and time have much limitness, also, the knowledge of spacecraft fault diagnosis is board, this model has a lot of problems to solve. It's evident that it has an actual connotation to do researches on this model. It's success will make good contribution to the space industry of our country.