Recognizing defects in surface layer of bearings by using fractal analysis of eddy current signal

Bearing manufacturing requires strict adherence to the technological standards and quality control of products at all stages of processing. One of the most accurate and environmentally friendly quality monitoring methods in bearing production is the eddy current method. When using this method, a large array of numerical data is generated that reflects information on the physical and mechanical conditions of the surface. Thus, defect detection process is reduced to defining the task of data segmentation, that is, selection of parts of an array with different dynamics, which is rather difficult for an operator. To solve this problem, a pattern recognition system is proposed, the elements of which are an alphabet of classes (defects of the surface layer), a dictionary of features (fractal dimension of a data array), rules for decision making on recognition (method of standards). Fractal analysis is applied to a two-dimensional array consisting of several data vectors, which will increase the degree of automation of the process of searching and identifying sensor readings indicating a defect, since the defect affects the surface layer of the product not only in the area of its greatest prevalence. Three methods of calculating the fractal dimension of a two-dimensional array are presented. Application of the reference method of the theory of pattern recognition is substantiated, an analytical expression that reflects the method is presented. According to the proposed solutions, it is possible to classify the defect of the rolling surface of the bearing rings by preliminary calculating the fractal dimension for each defect. Implementation of the software developed by using this method can improve the efficiency of quality control in bearing manufacturing and update the technological process.