High-Order Spectra-Based Deconvolution of Ultrasonic NDT Signals for Defect Identification

Pulse echo signals measured in ultrasonic NDT include the effect of the measuring systems, the propagation paths taken by the ultrasonic waves, and are corrupted by additive noise. For instance, the ultrasonic signals for a particular reflector recorded under the same conditions, but using different transducers can be quiet different. This leads to the difficulty of comparing and analyzing signals particularly in automated defect identification systems employing different transducers. Generally, it is assumed that the measured pulse echo is obtained by linearly convolving the defect impulse response with the measured system response. Deconvolution operation therefore, seeks to undo the effect of the convolution and extract the defect impulse response which is an essential step for the identification and characterization of defects. Conventional deconvolution techniques such as least square, Wiener filter, and minimum variance deconvolution [1] are based on a priori knowledge of second-order statistics (SOS) of the noise and the input signal. In practice however, ultrasonic pulse echo signals are found to be non-minimum phase systems and the acoustic noise due to scattering from the grains inside the propagation medium does not have a readily known statistic [2]. SOS-based deconvolution techniques, being phase-blind cannot therefore, accurately estimate the defect impulse response