ODK begins using AI to evaluate quality of powders for 3D printing

The Additive Technologies Center (part of the United Engine Corporation, ODK) has introduced artificial intelligence for evaluating metal powders used in 3D printers. Specialists created a neural network model that automatically generates data required for admitting metal powders to production of parts using additive technologies. The AI not only distinguishes between 'powder' and 'background' but also isolates individual particles in images obtained with an electron microscope, helping to verify material quality. During tests, the architecture showed high accuracy with an error of less than 3%. The key advantage is the transition from manual processing of microscopic images to automated measurements, automating the most labor-intensive stage: extracting individual particles and calculating their characteristics. Parameters such as number of particles, area, perimeter, and equivalent diameter are determined. The accuracy of computer vision data was confirmed by comparison with classical laser diffraction results. Currently, manual analysis requires a specialist to sequentially process images, while AI performs all calculations automatically, leaving only image preparation and result verification to the operator. This significantly reduces labor costs, accelerates analysis, and decreases the influence of the human factor. In the near future, the technology will be tested on various powder batches, the analysis methodology will be formalized, and a convenient software interface will be developed, followed by pilot testing in production.