Soft Fault Diagnosis Of Analog Circuits

With the rapid development of electronical industry, whether safeguard the normal operation of electronic devices has been the bottleneck of the whole industry.Although now most of electronic devices are integrated digital IC, and the proportion is more and more, but analog circuit couldn't be substituted. In particular, 80% of the faults are analog circuit faults.Although in 1960s, scholars had begun to research the fault diagnosis of analog circuits, but decades later, research progress is very slow, most of works still concentrated in the analog circuit hard-fault diagnosis. The last decade,the Neural Network has been used on the diagnosis of the analog circuit soft-fault. The main purpose of this article is to explore a new method that can solve problems of analog circuit diagnosis based on the BP neural network soft-fault diagnosis of analog circuits,combined with the wavelet theory, Empirical Mode Decomposition(EMD) and so on.1. Artificial Neural NetworksRereach the theory of soft fault dignosis in analog circuits based on the BP nerural networks.Simulation results illustrate that BP neural networks are feasible to soft-fault diagnosis in analog circuits.2. Wavelet theoryRereach the throry of the wavelet. Because the wavelet packet decomposition has a great advantage on the high frequency signal decomposition processing, it is suitable for soft fault diagnosis of analog circuits, and then verified through actual examples.3.Wavelet DenoisingCurrently there are a lot of noise signal in the analog circuit, it need denoising before fault dignosis. A new threshold function Combined Dononho soft and hard threshold denoising method advantages and disadvantages By studying and comparing the traditional method of wavelet threshold denoising.then through the genetic algorithm optimization, The simulation results show that the algorithm meet the required precision.4. Empirical Mode DecompositionAs the current there have signal mixing phenomenon by using wavelet packet decomposition,and can not adapt to change,and there is a problem on selecting wavelet. By studying the EMD theory, propose a new method on fault data acquisition on the EMD theory. The results show that the diagnosis method based on EMD has more advantages.In thort,this article uses a different circuit modeling and simulation by Orcad,and through the simple Orcad via Matlab data communication,to import the data into Matlab,The conclusion is that the method is effective to achinve the desired results.