Study On Fault Diagnosis Method Of Analog Circuit Based On Matrix Model Optimization

With the continuous progress of the era,China's independent research and development of chips is becoming a top priority,chip heat is also a key word in recent years,but there will be a lot of problems when the independent research and development of chips,because now the chip is more and more integrated;So when once appear problem,the chip will cause a great loss,and in many places,the cost of input circuit test is gradually higher than the actual development cost of the circuit;In particular,the analog circuit integrated chip is used in medical treatment,aerospace,communication and other fields.With the deepening of the scholars of analog circuit fault diagnosis research and discussion,need to face more and more challenges,but so far,there is still no a mature and effective method can be widely applied to the actual circuit,analog circuit itself exists because of the many restriction conditions,such as tolerance,resistance is poor,nonlinear,active and passive components.All these conditions will affect the analog circuit itself.When the analog circuit fails,there is no effective method to solve all the problems.At present,there are many papers published on the analog circuit fault diagnosis,most of which are artificial intelligence methods,which mainly rely on large amounts of data to complete the analog circuit fault diagnosis.This article mainly from the output voltage is measured,the measured voltage value,as to the circuit fault diagnosis matrix model,the two international standards Sallen?key,CTSV circuit and logarithmic amplifier circuit as experimental research object,through the different measurement output voltage value of time,through the different voltage value,the output of matrix,based on the model of matrix,the optimization of matrix did a series of research,from the characteristics of matrix and matrix dimension reduction,classification,will be better able to fault diagnosis of analog circuit,and failure to precise positioning.The work and methodological innovations involved in this paper are as follows:(1)In view of the existing analog circuit fault diagnosis method of artificial neural network,support vector machine(SVM),and other artificial intelligence algorithm,the artificial intelligent algorithm needs a lot of training samples,if not enough training samples or training time is not enough,is to cause a decline in fault diagnosis,so this paper presents a matrix characteristics analysis for analog circuit fault diagnosis method.The method establishes an output response matrix,in which the elements will change when the circuit fails.According to the matrix theory,found in the deep research in matrix theory,the inside of the matrix element changes,spectral radius of matrices and the largest singular value also subsequently and change,but may also occur two matrices with the same spectral radius,so this article use spectral radius and the largest singular value to find the difference between the matrix and reproduction radius and the largest singular value under the action of the common,Describe the properties of a matrix.The experimental results of Sallen?Key circuit,logarithmic amplifier circuit and CTSV circuit show that this method can judge the fault of analog circuit and locate the fault well.The effectiveness of the method in this paper has been verified in the Sallen?Key circuit,logarithmic amplifier circuit and CTSV circuit,and the fault diagnosis rate is as high as 100% in these three circuits.(2)In this paper,a method combining matrix model and machine learning is proposed.Compared with traditional machine learning,the diagnosis rate is improved,and the feasibility and effectiveness of matrix model for analog circuit fault diagnosis method are also verified.In order to solve the problem of feature extraction and feature classification of analog circuit fault diagnosis,a method of analog circuit fault diagnosis based on the Random Forest Algorithm(RF)was proposed.The method was based on the optimization matrix,and then a special optimization matrix model was established by three excitation methods.In addition,in the simulation software,the voltage values of the output nodes are measured at different times,and an output voltage value matrix is constructed jointly by the measured voltage values.When a circuit failure occurs,the elements in the output response matrix will change with the input of excitation.A new optimization matrix is generated by using local mean decomposition,and then the optimization matrix is input into random forest algorithm(RF),and the multi-dimensional vector can have different effective features.Through the combined action of bagging and decision tree,the optimized matrix model can be used to accurately study the fault diagnosis of single fault and multiple fault of analog circuit.Compared with other types of artificial intelligence algorithms,the optimized matrix random forest(RF)algorithm can not only satisfy the effect of feature extraction and feature classification at the same time,but also achieve99.5% fault diagnosis rate.