Probabilistic detection assessment based on one-dimensional signal distribution in eddy current testing under low signal-to-noise ratio

This study proposes a novel approach for evaluating the probability of detection (POD) by considering the spatial distribution characteristics of measured signals, with a particular focus on its application in eddy current testing (ECT) under a low signal-to-noise ratio (SNR) situation. To simulate such a situation, eighteen welded austenitic stainless-steel plates were prepared, and seventy-three artificial slits were introduced parallel to their weld beads to simulate weld cracks. ECT signals were collected from the samples using a uniform eddy current probe. Experimental results revealed that evaluating flaw presence solely based on the maximum amplitude of measured signals, as employed in the conventional â-a approach, was quite misleading. In contrast, the proposed approach, which quantifies deviations in the one-dimensional spatial distributions of signals inspired by the Gini coefficient used in economics to assess income inequality, provided fewer false positives and false negatives. A multi-parameter POD model was then developed, formulating POD as a function of both the length and depth of a slit using these deviations, namely the Gini values. Decision thresholds, which are critical for POD estimation, were determined from the maximum signal features of the seven slit-free samples. The results demonstrate that the proposed Gini-based approach enables a simpler POD analysis while reducing false detections and mitigating the challenge of setting a proper decision threshold under a low SNR situation, in contrast to the conventional amplitude-based method.