Method for searching defects in gas turbine engine blades under visible light using the U-NET model

In the production of aircraft engine parts, methods that allow detecting surface discontinuities in the material are widely used in testing operations. One of these methods is the capillary method of non-destructive testing. To solve one of the specific tasks – detection of contamination on the inspected surface, a description of the method for searching for defects on the surfaces of gas turbine engine blades under visible light is presented. The solution to the problem of searching for contamination during inspection of blade surfaces is based on image segmentation using the U-NET convolutional neural network. The results of using the trained model on blades in the production units of PJSC “UEC-Saturn” are presented.