Session: Ultrasonic Arrays I

Paper Number: 169461

169461 - A General Information Theoretic Framework for Multi-Path Phase Coherence Imaging 

Abstract: 

The concept of Phase Coherence has recently been adopted as a means of imaging small flaws in metallic components. Typically rendered from Full Matrix Capture (FMC) datasets, Phase Coherence Images (PCIs) are produced by converting elementary A-scans into a representation of instantaneous phase, most commonly by taking the sign of the digitized signal (sign coherence), prior to forming an image by applying the Total Focusing Method (TFM). The resulting PCI can be used as a multiplicative weight on a standard TFM image or, as is increasingly common, be analyzed directly. PCIs have been demonstrated to improve sensitivity to small omni-directional flaws such as crack tips, voids, fissures, etc., however, care must be taken to set appropriate dynamic range limits on PCIs so that inspectors do not consider incoherent noise as relevant or fail to consider scatterers which are highly coherent (compared with background noise). The approach taken by the authors in a previous work was to define PCIs in probabilistic terms, with the image representing the amount of Shannon Information against the hypothesis that no coherent scatter is present at each pixel, from which, together with the number of elements and total number of pixels can be used to define a minimum threshold. In this work, the authors extend the approach of defining PCIs in terms of the statistical difference with random noise to work with the signal sign approximation of phase coherence and Delay Multiply and Sum (DMaS) beamforming in a way which also incorporates multiple imaging paths. The general methodology involves using the multiplication between signal sign values for each transmission/reception/imaging path combination to estimate the distribution of signal sign for each pixel and then comparing the estimated distribution to the distribution of the signal sign for random noise (Rademacher distributed) by means of the Kullback-Leibler (KL) Divergence measure of statistical distance. The sampling distribution of the KL Divergence of random noise is then used to compute significance levels for PCIs rendered in this way, from which, conservative thresholds for PCIs can be set. The proposed method of generating and thresholding PCIs is demonstrated to produce high fidelity images of cracks, showing considerable improvement over the PCIs rendered by any individual imaging path or multiple imaging paths combined by summation or multiplication. The proposed algorithm can be viewed as a general information theoretic framework for phase (sign) coherence imaging supporting an arbitrary selection of imaging paths.

Presenting Author: Jonathan Lesage Acuren

Presenting Author Biography: Jonathan is an NDT Applications Engineer with Acuren's Applications Development Group. He began his career at Eclipse Scientific (since acquired by Acuren) as a signal/image processing specialist in 2016. Since that time, he has been working on novel techniques for processing Nondestructive Testing data and technique development for challenging inspections. He holds a Ph.D. from the University of Toronto and a CSWIP Phased Array Level II certification. His research interests are focused on applications of advanced ultrasonic array imaging techniques and statistical methods to improve detection, sizing and characterization of subtle damage mechanisms in an industrial context.

Authors:

Jonathan Lesage Acuren
Mohammad Marvasti Acuren
Oliver Farla Eclipse Scientific