Application of Model Data for Training the Classifier of Defects in Rail Bolt Holes in Ultrasonic Diagnostics

Abstract

The task of searching for defects on defectograms of ultrasonic inspection of rails using machine learning methods is considered difficult due to the lack of a sufficient representative training data set. In this study, on the example of defects in bolt holes of rails, the possibility of synthesizing an effective classifier trained on artificial data obtained by mathematical modeling is shown. Experiments have shown a high accuracy of predicting samples of real (not model) data exceeding 99%.