Fault Diagnosis Of Electronic Circuit Based On Manifold Feature Extraction And Optimization

With increasing scale and complexity of integrated circuits in electronic equipment,the system function has been continuously perfect.However,such achievements are accompanied by several non-ideal problems,such as less measurable node,difficulty of accessing the potential fault information,low accuracy of fault diagnosis,which will results in higher requirements with the reliability consideration of fault diagnosis technology for electronic circuit.Study of the circuit fault diagnosis technology is the key to achieve effective fault feature extraction and accurate fault pattern recognition.In this paper,we carry out a research into the circuit fault diagnosis technology from the aspects of the two core.Respectively a kind of fault feature extraction technology based on LLE and MDS,and a novel SVM classifier model based on DCQGA are proposed.The simulation result of the circuit diagnosis examples show the validity of this proposed modeling method.The main contents of this paper are as follows:Firstly,we study the modeling principle and the fault characteristics of typical electronic circuit.Build the circuit fault model based on the analysis of soft and hard faults of electronic circuit failure.Then take the output response of the measurable node as multi-dimensional fault feature data of the circuit.Secondly,the paper put forward a feature extraction technology based on the fusion of linear and nonlinear manifold learning algorithm,which is the fault feature extraction based on LLE and MDS.The fusion algorithm can excavate the potential lower-dimensional manifold structure from fault data,and keep the distanced-based similarity unchanged.The paper extracted the fault features by applying the fault diagnosis algorithm based on multi-dimensional fault feature data.The results show that the distribution of fault feature that after dimension reduction has obvious distinguish.Finally,a classifier model based on optimization algorithm called DCQGA-SVM is used for circuit fault diagnosis.DCQGA-SVM has the characteristics of double chain optimization which can accelerate the optimization process and increase probability of acquiring the global optimal parameters.After feasibility analysis of this classifier based on two typical classification data set from UCI,the performance test about convergence and fault diagnosis accuracy based on electronic circuit fault diagnosis is proposed.The results not only verify the good classification performance of the classifier,but also confirm its validity and practicability in the circuit fault diagnosis.