Artificial neural network for classifying fracture mechanisms of pure aluminum based on the wavelet transform parameters of AE signals

Authors:
Publisher website
Google Scholar
Abstract
Fracture of commercially pure aluminum in the as-supplied state and after annealing during quasi-static tension is investigated. The wavelet transform parameters identify the acoustic emission (AE) signals from dislocation processes and microcrack formation. The architecture of the multilayer perceptron is optimized. AE signals from different sources are classified with an accuracy of up to 92.5% for training data and 87.6% for test data. The energy parameter and the frequency of the local AE event are the most important information for the proposed neural network.

This publication has 22 references indexed in Scilit: