Fault Diagnosis for Analogy Circuits Based on Support Vector Machines

Abstract

When it is hard to obtain training samples, the fault classifier based on support vector machine (SVM) can diagnose faults with high accuracy. It can easily be generalized and put to practical use. In this paper, a fault classifier based on support vector machine (SVM) is proposed for analog circuits. It can classify the faults in the target circuit effectively and accurately. In order to test the algorithm, an analog circuit fault diagnosis system based on SVM is designed for the measurement circuit that approximates the square curve with a broken line. After being trained with practical measurement data, the system is shown to be capable of diagnosing faults hidden in real measurement data accurately. Therefore, the effectiveness of the algorithm is verified.