The Research Of Soft Fault Diagnosis Of Analog Circuit In Space Electrical Instruments

With the development of global science and military technology research, space electronic instrument systems are more and more widely applied in aerospace field. In order to ensure the system’s efficiency continually, real-time circuit testing and fault diagnosis must be carried on. Until now,digital circuits have had its practical fault diagnosis methods,but the existed analog circuit fault diagnosis methods still can’t meet the requirement. Therefore, the research on functional diagnosing methods of analog fault in digital-analog hybrid circuit of space electronic instrument system has become the objective need.At present, there are two kinds of means that often used to solve the analog fault diagnosis problems: fault dictionary method and the diagnosis methods which based on NN(neural network). Hence, this article will base on them and set analog circuits of space electronic instrument systems as research object to study the soft fault diagnosis methods. The main research results are as follows:To the problem analog circuits of space electronic instruments are too complex to implementing fault diagnosis,this article, based on analyzing analogy circuit characteristics of typical space electronic instruments, puts forward the idea of classifying circuit according itsfunctional feature and proposes a circuit fault rapid detection method to the divided circuit.As to satisfy different motives of fault diagnosis, this paper, based on improving the existed fault dictionary methods,achieves a hierarchical fault dictionary soft fault diagnosis method. This method can respectively achieve circuit fault detection, fault element location and fault element parameter identification. Therefore it has the advantages of high speed and precision of fault diagnosis.The process of building fault dictionary is always complex by using hierarchical fault dictionary method in diagnosing analog circuit soft fault of space electronic instruments.Therefore, this article, based on analyzing of wavelet theory and neural network theory, proposes a modified relax-type wavelet neural network method to diagnose analog circuit soft fault of space electronic instrument system. This new method which makes full use of the wavelet analysis and principal component analysis’ s advantages in data feature extracting and BP neural network’s advantages in pattern recognizing, has the characteristics of network convergence speed fast, fault diagnosis accuracy high and‘dictionary’building automation.For directly using BP algorithm or its improved algorithm,relax-type wavelet neural network can easily trapped in local minimum and cannot easy to jump out. Therefore, this article which combined genetic algorithm with BP algorithm proposes an optimized soft fault diagnosis method-GA neural network method.Experiment shows that the method can reduce the length of neural network’s training steps, improve the ability to identify failure and reduce the risk of network convergence in local.