Fault diagnosis of push-pull circuits using least squares wavelet support vector machines

Author

Luo, Zhiyong ; Shi, Zhongke

Author_Institution

Coll. of Autom., Northwestern Polytech. Univ., Xi´´an

fYear

2006

fDate

19-21 Jan. 2006

Lastpage

1030

Abstract

An approach for fault diagnosis of push-pull circuits based on least squares wavelet support vector machines (LS-WSVM) is presented. Output voltage signals of push-pull circuits under faulty conditions are obtained with simulation. Then wavelet coefficients of output voltage signals are gained by wavelet decomposition, and faulty feature vectors are extracted from coefficients. After training multi-class LS-WSVM by faulty feature vectors, the LS-WSVM classifiers of the circuit fault diagnosis system are built. The simulation result shows the fault diagnosis method of the push-pull circuits with multi-class LS-WSVM is effective

Keywords

fault diagnosis; feature extraction; learning (artificial intelligence); least squares approximations; pattern classification; signal processing; support vector machines; wavelet transforms; circuit fault diagnosis; faulty feature vectors; least squares wavelet support vector machines; multiclass LS-WSVM; output voltage signals; push-pull circuits; wavelet decomposition; Circuit faults; Circuit simulation; Fault diagnosis; Kernel; Least squares methods; Pulse circuits; Pulse width modulation; Support vector machine classification; Support vector machines; Voltage;

fLanguage

English

Publisher

ieee

Conference_Titel

Systems and Control in Aerospace and Astronautics, 2006. ISSCAA 2006. 1st International Symposium on

Conference_Location

Harbin

Print_ISBN

0-7803-9395-3

Type

conf

DOI

10.1109/ISSCAA.2006.1627498

Filename

1627498