Research On Analog Circuit Soft Fault Diagnosis Method Based On Morphological Fractal Dimension And Deep Learning

In daily life and production,analog circuit is widely used in electronic equipment.because of its nonlinear and high fault rate characteristics,analog circuit is always a hot issue in the field of circuit fault diagnosis.In the electronic circuit system,the vast majority of circuit faults originate from analog circuits,and the maintenance cost of analog circuit faults accounts for more than 90% of the total circuit maintenance cost,while the methods of analog circuit fault diagnosis are seriously lagging behind.Based on the above problems,it is of great significance to carry out theoretical and technical research on analog circuit fault diagnosis.In this paper,a method based on variational mode decomposition(VMD-MMFD)for soft fault feature extraction in analog circuits is proposed.First of all,the signal is subject to variational mode decomposition to suppress the influence of environmental noise,and a number of intrinsic mode components IMF with different center frequencies are obtained.The fractal dimension of the signal feature information component IMF is calculated to obtain high-dimensional feature sets.Then,the kernel principal component analysis(KPCA)is used to reduce the dimension of highdimensional feature sets,eliminate the overlapping and redundant parts,and complete the soft fault feature extraction of analog circuits based on mode decomposition and fractal dimension.Through the simulation circuit and software,the experimental results of different feature extraction models are analyzed,and the appropriate feature extraction model is selected,which provides a data basis for subsequent fault diagnosis.The attention residual neural network is used for in-depth learning.Although the feature information of fault samples is reduced to a certain extent,the number of network training parameters is reduced and the training time of the attention residual network model is greatly shortened through the grayscale processing of pictures.The experiment compares the diagnosis effects of different convolution kernels and different diagnostic models,and analyzes and selects the appropriate convolution kernels and diagnostic models to complete the research of analog circuit soft fault diagnosis.The experiment shows that the method based on mathematical morphology fractal dimension and attention residual neural network has better fault diagnosis effect.