A new approach for restoration of eddy current images

Abstract

Eddy current images of defects are blurred due to convolution of point spread function of eddy current probe with defects. Disturbing variables such as lift-off, surface roughness, and material property variations influence the eddy current images. In order to restore the length, width, depth, and orientation of surface-breaking defects in the presence of disturbing variables, a new and comprehensive approach has been developed. This approach uses artificial neural network and image processing methods. Studies on austenitic steel plates confirm that through this approach it is possible to restore the spatial information of surface-breaking defects of uniform or slowly varying depth and also to form their accurate three-dimensional pictures. This approach is fast as well as amenable for automation.