Abstract:

In order to solve the problems of inherent complexity of analog circuits,long detecting time and low correct recognition rate in traditional fault diagnosis methods, we propose a combinational method for fault diagnosis in analog circuits, which combines improved Least Squares Support Vector Machine (LSSVM) by Hidden Markov Model（HMM）with Volterra series.Firstly,we use frequencydomain core of Volterra series to extract circuit fault features.Then we adopt improved LSSVM by HMM for modal classification,and the fault diagnosis is completed.Simulation results show that compared with traditional BP neural network and LSSVM method,the proposed method is more efficient in system fault identification and faster in fault diagnosis.

Key words: fault diagnosis;LS-SVM;Volterra series;HMM;analog circuit