Title :
Improving neural network methods for time domain fault analysis of nonlinear analog circuits by feature selection
Author_Institution :
Inst. of Circuit Theor., Meas. Sci. & Mater. Sci., Tech. Univ. of Lodz, Lodz, Poland
Abstract :
The strategy of feature extraction and selection enabling to improve efficiency of fault detection methods for analog nonlinear circuits is presented in the paper. Simple algorithm for data selection, ensuring the proper diagnosis of faulty circuits having limited number of testing points, under assumption, that complex signal processing tools are not available, is proposed and tested.
Keywords :
analogue circuits; circuit testing; electronic engineering computing; fault diagnosis; feature extraction; network analysis; neural nets; signal processing equipment; time-domain analysis; complex signal processing tool; data selection; fault detection method; feature extraction; feature selection; neural network method; nonlinear analog circuit; time domain fault analysis; Analog circuits; Artificial neural networks; Circuit faults; Classification algorithms; Feature extraction; Signal processing algorithms; Testing;
Conference_Titel :
Signals and Electronic Systems (ICSES), 2010 International Conference on
Conference_Location :
Gliwice
Print_ISBN :
978-1-4244-5307-8
Electronic_ISBN :
978-83-9047-4-2
