Research On The Analog Circuit Fault Diagnosis Method Based On Wavelet-Fuzzy Bp Network

With the development of modern electronic technology, analog circuit integration degree is higher and higher, and the circuits scale is bigger and bigger, moreover, the analog circuit itself has tolerance, nonlinearity etc., all those make the fault forms change constantly, and increase the difficulties of fault diagnosis, the traditional fault diagnosis methods are no longer applicable. Therefore, people try to seek a new method for analog circuit fault diagnosis constantly, and it also has become one of the research hotspots since the recent decades.First of all, on the basis of predecessors’ research, aiming at the problems of the BP neural network whose convergence speed is slow and easy to fall into the local minimum point, this paper puts forward an improved BP algorithm, then applies it to the fault diagnosis of an analog circuits, through the experimental analysis proving that it works well.Secondly, aiming at the difficulties of extracting the fault feature information during the process of analog circuit fault diagnosis, a method based on wavelet packet analysis to extract the fault feature is put forward. After a multi-scale wavelet packet decomposition to the fault feature signal, we reconstructs it and extract the energy of signals in different frequency band as the fault characteristic vectors, then construct the training samples and testing samples, and through an instance analysis, the superiority of the proposed method is verified.Thirdly, aiming at the problems of "black box mapping" and unclear knowledge expression when using neural network for analog circuit fault, this paper puts forward a method of combining the fuzzy logic with neural network as the fuzzy neural network, and applies it to the analog circuit fault diagnosis, uses fuzzy logic to explain the neural network learning process, and puts forward the method of constructing a fuzzy BP network.Finally, this paper puts forward a method of analog circuit fault diagnosis based on wavelet-fuzzy BP network, uses wavelet packet analysis to extract feature information, then fuzzifies the feature information with the fuzzy BP network fault dictionary method and inputs them to the improved BP network for fault diagnosis. After that, we apply the method to a QMT series of automatic block molding machine control system circuit, and realize each function module circuit fault diagnosis of the system successfully, the feasibility of the method is validated, and the method provides a theoretical basis for the enterprise to apply it to production.