Probabilistic detection capability evaluation of eddy current testing to detect cracks on the divertor monoblock surface

Abstract

This study proposed a multi-parameter probability of detection (POD) model to probabilistically assess the capability of eddy current testing (ECT) for the inspection of divertor monoblock surfaces. The model assumes that a signal caused by the presence of a crack follows a probability density function with the depth, length, and position of the crack as variables. The parameters of the model are estimated by the combinational use of measured and experimental signals, and the probability that the signal exceeds a certain threshold is regarded as the probability of detection as the conventional â-a POD analysis. To demonstrate the model, sixty-three tungsten samples with an artificial slit were prepared to simulate the thermal fatigue cracking of the monoblock plasma-facing surface. The samples were measured using a differential plus-point ECT probe with a 1 mm lift-off and 500 kHz frequency. Correspondingly, numerical simulations of an eddy current simulator based on the $A-\varphi$ formulation were performed. The difference between the normalized slit signals and a flawless signal, which is obtained with the aid of image processing, was considered as the signal response to mitigate the influence of large noise caused at the edges of monoblocks. The results show that the proposed model can evaluate POD much more reasonably than conventional models that represent a flaw using one or two parameters.