|
The Imaging methods applied to Eddy Current (EC) testing [1] should allow, in the near future, the computational 3D reconstruction of the conductivity in every points of the inspected media [2]. Thus, in the context of the steam generator tube testing, a direct image of the defects will be provided. However, the in situ implementation of imaging methods requires both a specific probe and an efficient defect detection method to select the suspicious areas in which the imaging process will be performed. Moreover, in these areas, a preliminary defect classification can be profitably implemented in order to increase the a priori knowledge used to achieve the final imaging process. In this paper, we present a multi-segment probe answering the imaging specific requirements. We also present a model computation of the probe, based on the dyadic Green's function method using volume integrals. This model is used to develop both a defect and a classification method. The defect detection method is based on a wavelet/Bayes approach, designed for small notch detection. This method is applied to the EC signals provided by the probe during the inspection of notched tube samples. It allows an efficient detection (and localization) of the defects for inner and outer notches as small as 100µm wide, 15mm long and 10% deep. Finally, the detection scheme is extended to a notch classification, thanks to a EC signal database elaborated with the computed model.
|