Signature extraction from acoustic signals and its application for ANN based engine fault diagnosis. (1st January 2012)

• Record Type: Journal Article
• Authors: Prakash, Om; Singh, Vrijendra; Kalra, Prem Kumar
• Is Part Of: International journal of signal and imaging systems engineering. Volume 5:Number 3(2012)
• Page Start: 220
• Page End: 226
• Publication Date: 2012-01-01
• Subjects: acoustic -- ANN -- artificial neural network -- LOOCV -- leave one out cross validation -- signature -- STFT -- short time Fourier transform; Signal processing -- Periodicals; Imaging systems -- Periodicals; Information theory -- Periodicals; 621.3822
• Languages: English
• ISSNs: 1748-0698

Abstract

In present study, the approach of signature extraction from acoustic signals based on Short Time Fourier Transform (STFT) and its use for Artificial Neural Network (ANN) based fault diagnosis of internal combustion engine is explored. STFT can provide a time-frequency resolution data for signal signature extraction. It is suitable for extracting mechanical fault information form acoustic signals. In present work, a protocol of time dependent frequency information for development of signature and its application in engine fault diagnosis is proposed. The results of the protocol application show that the extracted signatures of seven classes of acoustic signals as engine fault information are effective for the development of classification model.