Image processing for an automatic detection of defect signals from electromagnetic cartographies

As the population of nuclear power plants ages, new defects are appearing in steam generator tubes (stress corrosion, corrosion pitting and intergranular corrosion). For more sophisticated expert appraisal of these defects, tubes can be examined by multifrequency eddy-current testing with an absolute coil (diameter value of 1 mm). A device, consisting of a push-puller mechanism and a motor-driven probe carrying this absolute coil, gives a helical movement to scan the inner surface of the tube. The signals obtained can be represented in the form of cartographies (3D representation in which the coordinates are the circumference, the length and amplitude of the X or Y component at a given frequency). The detection of defect signals by visual examination of these eddy-current cartographies is not always reproducible. The article describes an image processing procedure for the detection of defect signals which leads to a better reproductibility for more safety