Research On Test And Soft Fault Diagnosis Approach Of Analog Circuit Based On Wavelet Packet

Analog circuit fault diagnosis is an indispensable part of the circuit system,with the increasing scale and complexity of scientific and technological circuits and the faster processing speed of the circuit system,new requirements are put forward for analog circuit fault diagnosis.Because the development of fault diagnosis technology in circuit system is much slower than that in other parts,it is difficult to meet the needs of electronic system development,which has become a bottleneck in the field of electronic test system.In analog circuit fault diagnosis,fault feature extraction and pattern recognition are two key parts of analog circuit fault diagnosis.Two new methods for fault diagnosis of analog circuits are studied in this paper.Firstly,on the basis of the feature extraction method based on wavelet packet,a new feature extraction method based on wavelet packet and multifractal is proposed in this paper.General fault signals of analog circuit have the characteristics of nonlinear,wavelet packet transform is a linear method,the extraction of its classification characteristics is not complete.Multifractal is an important means to study the nonlinear problem,therefore,wavelet packet and multifractal are combined to realize complementary advantages and feature extraction for analog circuit faults.Secondly,ICA technology is fused on the basis of wavelet packet,another feature extraction method based on the combination of wavelet packet and ICA is proposed,and the breakthrough point of fault diagnosis in analog circuit field using ICA technology is found.ICA technology is different from the previous feature extraction,which is based on non-gaussian and independence.The low-frequency and high-frequency components are obtained through the processing of wavelet packet,and then used as the ICA input to extract the independent components of each frequency layer,so as to obtain the fault characteristics.Finally,BP neural network is used as the fault classifier,and the feasibility of each method is verified by simulation.Although the diagnosis speed of wavelet packet combined with multifractal extraction method is faster than that of wavelet packet combined with ICA technology in terms of diagnosis time,the accuracy is a little less.In practice,different diagnostic methods can be selected according to different diagnostic requirements.