A Novel Strategy of Analog Circuit Fault Diagnosis Based on Support Vector Machine Ensemble

In order to solve the problem of improving the generalization performance and diagnosis precision when support vector machine is applied to multi-class analog fault diagnosis,a novel method of analog circuit fault diagnosis based on support vector machine ensemble was proposed.First,the unclassifiable samples of the iteration was added to training sample set to constitute the training sample set of next iteration,and the diversified classifiers were constructed.And then,the classifiers whose accuracy are not less than the average accuracy were selected to ensemble,and the overall diagnostic accuracy was improved.The experiment results show that the proposed strategy is applicable to both linear and nonlinear analog circuits fault diagnosis.