Soft Fault Diagnosis Of Analog Circuits Based On RCCA-SVM

With the improvement of analog circuit integration,the traditional methods and theories of circuit fault diagnosis are no longer suitable for the increasing scale of modern circuits,which makes nonlinear analog circuit fault diagnosis urgently need a new solution.At present,the two key parts of analog circuit diagnosis include feature data extraction and diagnosis.Therefore,the key technologies of data feature extraction are innovated and data preparation is made for the classification process.In the soft fault diagnosis of nonlinear analog circuits,single feature can not reveal fault information in many directions.Therefore,this paper takes the weighted fusion of fault characteristics of nonlinear analog circuits as the main innovation,and takes the fourth-order operational filter as the experimental research object.The main contents and innovations of this paper are as follows:1.Multi-modal feature data in frequency domain,time domain,and statistical characteristics based on fault features is used as the original feature set.In this paper,the characteristics of three modes are extracted from the output data of the nonlinear analog circuit:energy value,kurtosis value,and frequency domain characteristics(gain,cutoff frequency,center frequency).In this method,wavelet packet is used to extract part of the multi-modal features to provide multi-directional description of the fault,and then more fault information is obtained.Experiments on different dimensionality reduction methods and classifiers show that the accuracy of multi-modal fault feature diagnosis can be increased by 38%compared with single fault feature diagnosis.This experiment shows the advantages of multi-modal feature fusion in analog circuit fault diagnosis,also provides ideas for the selection of dimensionality reduction methods and classifiers.2.Based on the fusion reduction based on relief and CCA(RCCA).The feature weighted fusion algorithm reduces the dimension of fault features.In this paper,the weight and correlation of fault features are considered.The relifF algorithm is used to solve and filter the weight of all features,leaving the feature data with larger influence factors for fault diagnosis,filtering out the feature data with less influence on fault diagnosis;CCA algorithm is used to strengthen the correlation between features after the selection,and then the fusion features are obtained.Then principal component analysis(PCA)is used to reduce the dimension of fusion features,and the final feature data is obtained.Finally,the final feature data is used as the input data of support vector machine(SVM),and SVM is trained to get the diagnosis model and results.Two circuits are used as experimental cases to get the diagnosis results and accuracy.The experimental results show that the overall diagnostic accuracy of the method proposed in this paper is as high as 99.36%,which is 4 percentage points higher than that of the same kind of diagnostic method;and the accuracy of each specific fault circuit has been improved,which has advantages in solving the circuit types of individual diagnosis difficulty.3.Finally,in order to strengthen the practicality of the model and realize the automation of analog circuit fault diagnosis,this study based on the MATLAB GUI interface design for nonlinear analog circuit fault diagnosis.Carry out circuit testing on the designed interface,the diagnostic accuracy rate reaches 99.33%,and the type of test sample can be judged.