Instruments and Systems: Monitoring, Control, and Diagnostics Annotation << Back Development of an Algorithm for the Thermograms Processing Based on the U-Net Artificial Neural Network A.V. FYODOROV, A.V. KOZLOVSKIY, I.O. KOTOVSHCHIKOV This paper is devoted to the application of a neural network based on the U-Net type to the problem of automatic defect recognition on thermograms of composite helicopter blades. A training dataset generator has been created that simulates a characteristic surface thermal response dynamics during to testing part with internal defects. A set of 15 thermograms was fed to the input of the network, and as a result of the network operation we had one image with a binary mask highlighting the detected defects. According to the results of approbation, the probability of automatic detection of defects using the developed algorithm was 71,5 %. In order to increase this indicator, it is planned to develop a methodology of the nondestructive testing by active thermography method, add to the algorithm a stage for identifying different sections of the composite blade and change the dataset generator. Keywords: neural network, active thermography, composite products, automatic defect search. DOI: 10.25791/pribor.5.2022.1335 Pp. 01-09.