[4C2] Investigation into distinguishing defects from corroded surfaces in ultrasonic imaging
Y Cai, A Mulholland and J Zhang
University of Bristol, UK
When a defect resides on a corroded and uneven surface, the situation becomes significantly more complex. This complexity arises from several additional challenges: reduced inspection angular coverage to avoid specular reflections from the surface, possible wave paths through the weld region, scattering from the uneven surface with unknown topography and the presence of corrosion zones near the surface that may contain porosity. Such porous regions can behave like highly reflective point scatterers, the ultrasonic scattering characteristics of which resemble those of crack tips. In this study, we investigate the use of statistical classifiers, including the Kolmogorov-Smirnov (K-S) test, skewness and kurtosis calculations, applied to the amplitudes of specific pixel groups in ultrasonic images formed from a single full matrix capture (FMC) array dataset. Segmentation images are generated to identify three regions: no defect, defect and corrosion. The performance of the classifiers is evaluated using array datasets obtained from both simulations and experimental measurements. Results show that the skewness-kurtosis approach delineates defects clearly but loses reliability when defects are adjacent to corrosion. The K-S test effectively detects both defects and corrosion, but it struggles to distinguish between defects and the edges of corroded areas.
