Analog Circuit Fault Diagnosis Using Lifting Wavelet Transform And Support Vector Machine

Analog circuit fault diagnosis plays a key role in many fields such as circuit design, equipment manufacture and instrument maintenance. The technology of diagnosis is still a challenging topic for academic researchers and engineers in the field of circuits and systems. It is difficult for traditional fault diagnosis theories and methods to solve the problems of fault diversity and complicacy, which caused by tolerance on component parameters, continuous and non-continuity response, and environmental factors, and other factors as well.Based on the modern diagnostic technology, this paper constructs a framework of analog circuit diagnosis which is combining with Lifting Wavelet Transform and support vector machine.The fault feature extraction and effective fault identification methods are two critical parts for intelligent fault diagnosis. So the optimization of feature extraction and design for classifiers for analog circuit fault diagnosis system are discussed thoroughly in this paper. The main research contents and achievements are summarized as following:On the basis of summing up the existing analog circuit fault feature extraction method, this paper comes up with the idea that applying the lifting db5 wavelet to the analog circuit faults diagnosis. This is because lifting db5 wavelet has a high degree of similarity with response of circuit, which contributes to extract the signal characteristic and decomposition faster. Compared with the non-lifting method, this method could reflect the characteristics of the original signal more accurately. Data which are obtained from simulation examples show the superiority of this method.For the problems encountered in analog circuit fault identification currently,this paper constructs least squares support vector machine which is based on mahalanobis distance.By using least squares linear system, it simplifies the complex problem solving; by introducing mahalanobis distance, it improves the sparsity of least squares support vector machine. The simulation shows that this method can be used effectively among the analog circuit fault diagnosis.In order to improve the generalization learning ability of support vector machine, this paper uses PSO algorithm to optimize the structure parameters of least squares support vector machine. The standard of PSO algorithm should be improved because it is easy to fall into local optimal and premature convergence problem. Attractive and repulsive control particle swarm optimization is proposed at the same time, by controlling the particle attraction and repulsion movement to increase the diversity of particle motion, avoiding particle fall into a local optimum and premature convergence during evolution.