Reasearh On Analog Circuit Fault Diagnosis Based On Support Vector Machine

Analog circuit fault diagnosis has been a research focus in the circuit testing domain, after nearly decades of development, has formed a set of more systematic theory and method. With the development of intelligence and automation technology, electronic equipment has been widely used in various industries and people's daily life, reliability requirements for equipment operation is self-evident, therefore doing research on analog circuits fault diagnosis is of great significance and practical value. The lack of fault samples is one of the difficulties of analog circuit fault diagnosis, however, support vector machine (SVM) has outstanding advantages in dealing with the small sample problem compared to the traditional identification method, therefore Fault diagnosis of analog circuits based on SVM has been studied in this paper.This paper has studied the efficiency of the fault feature extraction method based on wavelet decomposing, and has given the implementation of support vector machine multi-classification and the application of SVM in analogue circuit fault diagnosis. Fault diagnosis is difficulty for the complex analog circuit which has many fault modes, in order to solve this problem, a new fault diagnosis method based on multiple error correcting support vector machines (EC-SVM) is proposed in this paper, the whole diagnostic process is:Firstly, the circuit under test is divided into several smaller units according to its characteristics, and performing circuit simulation and analysis to determine the fault sets and test nodes; secondly, the fault features are extracted by joint time-frequency domain feature extraction method and composing training and testing samples; third, designing an EC-SVM classifier to identify fault unit and designing EC-SVM classifiers for each unit to identify faults within the region; finally, training each EC-SVM by training samples, and then the testing samples are input to the trained multiple EC-SVM classifiers to achieve the circuit fault diagnosis. At last, a complicated actual circuit is selected as fault diagnosis experiment to verify the effectivity of the method proposed in this paper, the experimental results show that this method can find fault quickly, its correct recognition rate is higher than that of the fault diagnosis method which use only one classifier.