MODAL-BASED IDENTIFICATION OF ACOUSTIC EMISSION SOURCES IN THE PRESENCE OF ELECTRONIC NOISE

Based on the results of an earlier study, a specific scheme was demonstrated to identify acoustic emission (AE) source types using ratios of Lamb-wave modal amplitudes from two radiation directions in an aluminum plate 4.7 mm thick. The modal amplitudes were obtained from certain peak magnitudes of...

Full description

Saved in:
Bibliographic Details
Published in:Journal of acoustic emission 2004-01, Vol.22, p.1-21
Main Authors: Hamstad, M A, O'Gallagher, A
Format: Article
Language:English
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Based on the results of an earlier study, a specific scheme was demonstrated to identify acoustic emission (AE) source types using ratios of Lamb-wave modal amplitudes from two radiation directions in an aluminum plate 4.7 mm thick. The modal amplitudes were obtained from certain peak magnitudes of wavelet transforms (WT) of the AE signals at a propagation distance of 180 mm. The peak magnitudes were taken from energetic frequency-mode combinations. These combinations were the fundamental modes A0 at 60 kHz, S0 at 270 kHz and S0 at 522 kHz. The AE signal database was obtained using a validated finite element model (FEM) for three different source types, each located at six or seven depths in the plate, which had large lateral dimensions. The technique of source identification was demonstrated using ratios of the WT-determined modal magnitudes of the signals at a radiation angle of 45DG divided by those from a 0DG angle. Then, the effect on source identification of adding electronic noise (sensor/preamplifier) to the FEM signals was studied. Due to the random nature of the experimental wideband noise, a statistical study was necessary. At each signal-to-noise (S/N) ratio, 50 cases were examined for certain depths of each source type. The percentage of correct source identifications was determined as a function of the S/N ratio. To obtain a high percentage of correct identifications, a S/N ratio of at least six to one was found to be necessary.
ISSN:0730-0050