Pattern Recognition Analysis of Magnetomechanical Acoustic Emission Signals,
CALIFORNIA UNIV LOS ANGELES DEPT OF MATERIALS SCIENCE AND ENGINEERING
Pagination or Media Count:
In order to quantify detected signals of magnetomechanical acoustic emission MAE for nondestructive testing applications, we employed pattern recognition analysis. MAE waveforms were determined for variously heat-treated A533B steel specimens subjected to several levels of magnetic field and tensile stress. The autoregressive coefficients and envelopes of the MAE waveforms were extracted and utilized for pattern classification on the basis of the weighted nearest neighbor decision rule. The autoregressive coefficients improved the percentage of successful classification to 80. When the envelopes of MAE waveforms were employed as the features of pattern recognition analysis, almost perfect classification of MAE waveform was obtained. This method allowed the classification according to heat treatment, magnetic field and tensile stress. The results suggest that real-time recognition of underlying parameters of MAE waveforms is feasible since the envelopes of MAE waveforms are easily detected by hardwares and subsequent digital signal processing is relatively simple. Additional keyword reprints. Author.
- Test Facilities, Equipment and Methods