SIGNAL PROCESSING TECHNIQUES FOR EDDY-CURRENT DATA

摘要

The nonlinear image restoration software and the automatic flaw detection software have produced significant results. The advantages of the nonlinear method are that it produces high resolution images, excellent noise reduction, and is applicable to various flaw-coil configurations. Disadvantages of this method are, a longer execution time (a few minutes) relative to linear methods (a few seconds), higher complexity, and a rather extensive task of creating the training data set. However, once created, the training data set can be used for all restorations whose flaw geometry falls in the domain of the training geometries. The automatic flaw detection package had a successful detection rate of 82% without counting any of the so-called false calls that might be true flaws. The software is capable of finding flaws not initially found by human inspectors, although we feel that human inspectors could ultimately find every flaw identified by the software if they spent the appropriate time studying the data. The four authors worked as a team for several years on these projects. While John Moulder will be sorely missed, the remaining three authors are very thankful they had the opportunity to work with John and share in his pleasure and enthusiasm for science.