Research On Soft Fault Diagnosis Of Nonlinear Analog Circuits Based On Fractal Theory

In the field of circuit testing,the research on soft fault diagnosis of nonlinear analog circuits is still a popular subject in the current mainstream.As the nonlinear analog circuit soft fault is complex and diverse,and the circuit also has the characteristics of response to nonlinear and component parameters tolerance.It is difficult to use traditional methods to diagnose the fault in practical engineering,and the efficiency is not high.The Volterra series method developed in recent years only has a good effect on the soft fault diagnosis of weak nonlinear analog circuits.But when the nonlinearity of the analog circuit system is serious and the order term of the characterization system is higher.The computational complexity becomes very large.Aiming at these problems,it is very necessary to develop a fault diagnosis method which is suitable for the soft fault of nonlinear analog circuit,which is simple and accurate.As one of the important branches of nonlinear disciplines.The fractal theory can effectively describe the local fractal characteristics existing in nonlinear processes.The fractal theory is applied to the soft fault diagnosis of nonlinear analog circuits,and the fault characteristics of the circuit can be better characterized by fractal dimension.This paper focuses on the theme of soft fault diagnosis of nonlinear analog circuits,based on the fractal theory.The nonlinear analog circuits are analyzed by generalized multifractal,empirical mode decomposition and multifractal spectrum,extract fault feature,and by using the method of correlation recognition with support vector machine to diagnose circuit.The main contents of the study are as follows:Application Research on fractal theory.Understand the formation and characteristics of fractal theory,the single-fractal and generalized multifractal dimension calculation method analysis.Taking the fault signal as an example,the fractal dimension of the fault signal is calculated by using the single fractal box dimension and the generalized multifractal dimension.The influence of different frequency input signals on the fault case signal characteristics is analyzed.A soft fault diagnosis method for nonlinear analog circuits based on fractal dimension is proposed.Aiming at the problem of low diagnosis and high computational cost,the fractal dimension method is used to diagnose the fault.Firstly;the mechanism and the process of single-fractal fault diagnosis are studied.Taking the nonlinear analog circuit as an example,the single-fractal theory is used to divide the fault state interval,calculate the range of the singlefractal dimension of the original fault signal,by comparing the fractal dimension of the fault signal to be measured to reach the purpose of fault diagnosis.Then,the fault diagnosis mechanism and dimension correlation of generalized multifractal are studied.The generalized multifractal dimension of the original fault state of the circuit is taken as the fault feature;the purpose of fault diagnosis is achieved by calculating the correlation between the generalized fractal dimension sample of the original fault and the generalized fractal dimension sample of the fault signal to be measured.A generalized multifractal soft fault diagnosis method based on empirical mode decomposition is proposed.For the generalized multifractal fault diagnosis method,the sample sequence has less priori data and sometimes can not accurately describe the problem of circuit fault characteristics.The generalized multifractal fault diagnosis method is extended to decompose the original multicomponent fault signal by empirical mode.The fractal dimension of each single component signal is calculated by generalized multifractal dimension method based on GP correlation dimension,and the original fault signal of the dimension of the sample sequence.Then,the fault diagnosis is realized by calculating the correlation between the original failure sample and the generalized multifractal dimension sample after the empirical mode decomposition of the fault signal to be measured.A soft fault diagnosis method based on multifractal spectrum and support vector machine is proposed.There are many disturbing information trends in the fault signal of nonlinear analog circuits.Fractal analysis of trend fluctuation is introduced into the multifractal spectrum of fault signal.Can effectively remove these interference trends,retain the original fault signal complete fractal characteristics.The physical meaning of multifractal spectral parameters is studied,and the parameters of the multifractal spectrum of the original fault state signal are calculated.The original fault feature vector is calculated as the original sample,the fault feature vector of the measured signal as test samples,which are input to the support vector machine classification,so as to realize the circuit fault diagnosis,results show that the diagnosis result is very high.