Fault Diagnosis Of Analog Circuits Based On Optimal Measuring Points

The breakthrough of modern science and technology has led to the progress of the whole industrial system.After the large-scale popularization and used for these years,complex electronic technology equipment has been accepted by almost all industries.Convenient,intelligent and stable working state has become an indispensable guarantee for the stable progress of this era.Therefore,it is important for the high efficiency and reliability of diagnostic technology of electronic equipment.It has become more and more urgent.Many scholars have done a lot of research in the field of analog circuit fault diagnosis,and some effective methods have been put forward.But in the practical application of fault diagnosis,the following two problems will arise:(1)The problem of selecting the optimal measuring point.When collecting the original circuit signal,it is necessary to explore which location of the measuring point in the acquisition data to ensure that the collected original signal can contain all the fault feature information of the circuit,and make the number and dimension of the data the lowest;(2)The problem that the fault information is disturbed by noise.Usually,the fault information in the original signal collected will be disturbed by a lot of noise,and it is not easy to get enough fault samples.How to diagnose these disturbed fault information is a difficult problem.To solve the above two problems,a fault diagnosis method for analog circuits based on sparse denoising autoencoder neural network and optimal measurement points is proposed in this paper.In order to solve the problem of selecting optimal measuring points,this paper introduces a method of selecting measuring points based on sensitivity factor and fault isolation group/measure.This method can obtain the optimal combination of measuring points for fault feature information of the whole analog circuit,avoid a large number of invalid operations of the selected measuring points,and the original signal data collected from the optimal measuring points has less information,low dimension and circuit fault.Aiming at the second problem,this paper proposes a deep learning method of sparse denoising autoencoder neural network.This method needs to add a certain probability distribution of noise to the original signal in the process of using,because sparse denoising autoencoder neural network needs to add a noise signal to the original signal before processing the original data.Because of its own characteristics,the problem that the second fault information mentioned above is disturbed by noise can also be solved.In order to verify the effect of the optimal selection method and sparse denoising autoencoder neural network in fault diagnosis of analog circuits,an experimental simulation of analog circuits is carried out in the fourth chapter.After the sparse denoising autoencoder neural network iteratively calculates the original data collected from the optimal measurement points,the classification results are finally obtained,and the square of the experiment is obtained.The accuracy and validity of this method in fault diagnosis of analog circuits are fully illustrated by the error description curve and the analysis of classification and recognition results.