A Network Decomposition Approach of Analog Circuit Based on Alternating Voltage

Abstract

In order to solve the difficult of fault diagnosis in large-scale analog circuits, a soft-fault diagnosis method is proposed, which is based on network decomposition, sensitivity analysis and support vector machine (SVM). Firstly, tear and isolate the target network with the use of AC and DC excitation. Then, prefer test node according to the concept of sub-network sensitivity and extract its information as fault feature after the Monte Carlo analysis and normalized processing. Finally, achieve fault network identification with the use of support vector machine classifier. The diagnosis principles and steps are described. Finally, the efficiency of the method is shown by practical examples comparing with the existing method, this approach which is suitable for tolerance and non-linear circuits, has smaller less computation, higher accuracy, and stronger engineering practicality.