Acoustic emission analysis using pattern recognition

Three different pattern recognition techniques were applied to a data set consisting of acoustic emissions (AE) caused by crack growth and acoustic signals caused by extraneous noise sources. The purpose was to test the techniques for prescreening acoustic events and filtering out those that are produced by extraneous sources. The results were surprisingly good. Most of the noise waveforms used in this study cannot be distinguished from valid AE by visual examination; however, we were able to correctly classify 90% of the waveforms as either valid AE or noise using the least squares decision rule. Since the current application of AE data in nondestructive evaluation rely primarily on counting acoustic events, the waveforms caused by noise in the environment must be filtered out to avoid misinterpretation. The results of this investigation have convinced us that pattern recognition concepts can be used to design such a filter.