Acoustic emission analysis using pattern recognition

The purpose of the work is to design a method of differentiating events, caused by a growing crack in a critical structure from innocuous noise sources, from the waveform in order to improve the accuracy of acoustic emission monitoring systems for metal structures. The development approach is to generate acoustic emissions from known (crack, fretting, electrical transient) sources in the laboratory, compute certain quantities (features) from the recorded signal, and design discriminant functions to differentiate among the sources. The selection of the features for the discriminant function is crucial. They are selected on the basis of physical as well as empirical grounds. The acoustic signal is influenced by subtle environment effects as well as the instrument system, the geometry of the structure being monitored and its interaction with the changing geometry of the growing crack. The challenge of the work is finding features that are applicable to a range of monitoring situations and are reasonably stab...