PD characterization using wavelet decomposition of acoustic emission signals

Abstract

The identification of partial discharge sources using acoustic emission measurements and artificial neural networks has been investigated. Measurement data was processed using the wavelet transform, which decomposed the acoustic signal into approximation and detail components at different levels. Two different arrangements of artificial neural networks were implemented: a feed forward network using the back propagation algorithm and a Kohonen self-organising map network using the learning vector quantization algorithm. They were used to characterize AE signals produced from different shapes of void within a polyethylene dielectric. The factors that influence the artificial neural network performance have been investigated.