Research On Analog Circuits Fault Diagnosis By Wavelet And Fuzzy Neural Network

For forty years, research of analog circuits fault diagnosis has developed greatly, and the forefront of modern circuit theory has a series of diagnosis theory and methods. However, complicated theory and poor practicability of these methods make these methods' application foreground far from expectation of people. Neural Network (NN), wavelet theory and fuzzy theory have gained quick development and have been widely applied in fault diagnosis of analog circuits with tolerances recently.This paper mainly researches the analog circuits fault diagnosis with tolerances by Wavelet Neural Network (WNN) and Fuzzy Neural Network (FNN). The methods which based on Back Propagation Neural Network (BPNN) can decrease fault diagnosis time and enhance network of average diagnosis accurate ratio.First of all the paper discusses the principle of BPNN algorithm and the way to improve it, and applies the BPNN based on principle component analysis preconditioning in fault diagnosis of analog circuits with tolerances. Secondly, on the basis of classical WNN, the paper puts forward an improvement of WNN with many inputs and outputs, training the network with the back propagation arithmetic which adopt momentum and self-adaptive learning arithmetic. This paper applies the WNN in fault diagnosis of analog circuits with tolerances and researches the effect of its application. We also apply the BPNN to diagnose the same swatch, and compare the effect of the two methods. Afterwards, the paper combines fuzzy theory with BPNN. In the method, an integrated self-adaptive FNN is developed based on the Takagi-Sugeno fuzzy system. The training of network weights and optimization of membership functions are conducted employing hybrid algorithms. Comparing with the BPNN'method, thefeasibility and validity of the method are validated by simulation testing.