Diagnosis System Based On Neural Network And Fault Tree Launch Vehicle Failure

With the development of artificial intelligence, it could be found there are so many diagnosis techniques in current phase, which is called intelligent diagnosis phase. However, the current situation of launch vehicle fault diagnosis technology still depend on the expert experience and drawing the faulty tree at the scene of the accident. It could not demand to requirement of current the fault diagnosis for launch vehicle. To solve this question and meet the high density of satellite launching missions. A new method which depends on neural network and faulty tree is proposed.In the thesis, it analyses the method of faulty tree diagnosis and how to build a faulty tree. By using low-temperature control system of launch vehicle as a practical example, it designs a faulty tree. Simultaneously, it points out the faulty tree has the limitation of self-studying and self-adapting. Then, the thesis uses the neural network, Back-Propagation Network algorithm and also by using cryogenic power system of launch vehicle as a practical example to simulate in the MATLAB. It could come to a conclusion that neural network diagnosis could cover the shortage of faulty tree diagnosis. Then, the paper design a new diagnosis system based on launch vehicle. The system adopts Model-View-Controller design pattern and use Unified Modeling Language to design the key system module. The system is under the Visual Studio 2008 platform using C# development language and database of SQL Server 2005.At last, by using a typical fault of launch vehicle as a practical example, the results of diagnose shows that the system meet the functional requirement, simplicity of operator, good stability, strong fault tolerance and flexibility.