Abstract <p>The introduction of additive technology at manufacturing enterprises depends on monitoring of the internal defects in the additive layers during the production process (real time online monitoring) for expensive and highly reliable components, where no defects are permitted. To that end, an additive system for the production of metal components with X-ray tomographic monitoring of layer quality was developed in [1]. The tomographic images were analyzed by means of artificial intelligence (neural networks), with feedback to adjust the additive technology as necessary. Specialized X-ray tomographic systems for additive technology and specialized tomogram reconstruction software for a conical radiation beam, an incomplete set of radiation angles, and spiral scanning were discussed in [2–7]. Experiments regarding the production of additive layers (with and without defects) of Sv-08G2S-О alloy steel and the slots in the layers were described in [8]. Tomographic images of the defect-free and defective additive layers from a specialized high-resolution X-ray tomographic instrument were analyzed in [9]; the potential for neural network tomogram analysis and defect detection was considered. In the present work, we develop a convolutional neural network for the analysis of X-ray tomographic images (tomograms) showing the internal structure of additive layers, with a view to detecting defects of cavity type. Such defects are the most significant in terms of the strength of metal components produced by additive technology.</p>