Defect Deconvolution using 4th Order Statistics for Ultrasonic Nondestructive Testing

Abstract

Classification of defects using ultrasonic nondestructive testing (NDT) is primarily done in the field of industrial materials to provide useful information in order to assist in making administrative decisions in terms of maintenance and replacement. The technique presented in this paper utilizes the concept of defect induction as a convolution process between the clean sample and the defect signature. Hence, to identify the type of defect a deconvolution approach can be useful. Due to several similarities between the ultrasonic echoes and the usual modulated sinusoids, a motivation is present to use 4th order statistics for completely defining the waveform. Such a definition, when compared with standard defects, will provide useful insight in terms of defect classifications and understanding.