Date of Award
Doctor of Philosophy
Engineering Science and Mechanics
Lester W. Schmerr
In the field of ultrasonic Nondestructive Evaluation (NDE), it is important to be able to classify a flaw as cracklike or volumetric, based on features contained in an ultrasonic scattering response. In the present work, it demonstrated, based on model studies, that three fundamental features, obtained from the early time response of a flaw are in principle useful for classification purposes. These features are based on the Kirchhoff approximation for cracks and a new subtracted surface integral formulation of the Born approximation for volumetric flaws. To reliably extract such features from an experimental response, however, it is important to perform both deconvolution and low frequency extrapolation procedures. Here, we show that a new flaw-derived reference deconvolution method and a low frequency extrapolation technique, using the Gerchberg-Papoulis algorithm, can be effective for these purposes. Once these fundamental features are extracted, a simple adaptive learning procedure has been successfully applied for performing the classification.
Digital Repository @ Iowa State University, http://lib.dr.iastate.edu/
Lat Sang Koo
Koo, Lat Sang, "Ultrasonic flaw classification: an approach using modeling, signal processing, and adaptive learning " (1987). Retrospective Theses and Dissertations. 8667.