This study developed a machine learning approach to detect carbide precipitation in the microstructure of austenitic stainless steel, specifically grade 316, using a convolutional neural network (CNN). Microstructural images were prepared and classified into three categories: as-received, heat-treated without carbide precipitation, and heat-treated with carbide precipitation. A CNN was trained and validated using two separate datasets of varying sizes to assess the impact of data quantity on classification performance. This automated microstructure recognition system offers potential benefits for additive manufacturing (AM) by enabling real-time quality assessment and feedback control, particularly for avoiding undesirable carbide formation during metal 3D printing. By linking microstructural analysis to processing conditions, this approach supports the development of defect-free, corrosion-resistant components and contributes to the integration of intelligent monitoring within digital manufacturing workflows.
Khoshnaw et al. (Tue,) studied this question.