The problem of determining the area of defects in the surface layer of bearing parts according to eddy current non-destructive testing is considered. Methods of processing eddy current control data are given. The possibility of using a robust median polishing method to increase the information content of eddy current data is substantiated. It is proposed to use a sliding window, a standard deviation calculation, and a production rule formed by the Shannon information entropy criterion as tools for localizing defect patterns in the eddy current image of the control object. The results of the application of the developed localization algorithm based on eddy current control data of bearing parts obtained in real production conditions are presented.

Keywords: eddy current control, localization, defect, data analysis, recognition, surface layer, intelligent technologies, Shannon entropy, median polishing, classification problem