SIGNATURE TECHNIQUES FOR DEFECT CHARACTERIZATION.

Several integral transformation techniques are presented that produce feature extraction magnification for various ultrasonic defect signals. A class of fast tensor transformations is studied. Such integral transform techniques as Fourier, Mellin, Hilbert, and Hadamard are also used to examine selected ultrasonic response functions from model defects. Ultrasonic response transforms are examined along with such modeled flaw problems as wall thinning and delamination detection, determining dispersive effects due to density variations or geometries, and attenuation resulting from absorption or grain boundary scattering. Several ultrasonic response functions are presented along with the various transform functions. The structural aspects of various transforms are also studied and approaches presented for examining larger classes of transforms than those that are normally considered in signal processing work today.