Microstructure, Property and Process Optimization for Powder-Bed Fusion Additive Manufacturing Using Small Dataset Deep Learning

The dataset size in machine learning may directly influence the model's accuracy, robustness, and ability to uncover nuanced patterns critical for enhancing efficiency and reducing costs. This work presents recent advancements in optimization of metal additive manufacturing, particularly powder-bed fusion 3D printing process, through the application of deep learning techniques with limited datasets. Focusing on microstructure, property, and process optimization, we address challenges in achieving full-spectrum yet precise control. By leveraging small dataset deep learning techniques, the research explores efficient strategies for predicting material behavior and optimizing 3D printing parameters by making use of titanium alloy and stainless steel as model materials. Key findings demonstrate significant improvements in understanding microstructural evolution and mechanical properties, enhancing the feasibility and scalability of additive manufacturing processes. This work underscores the potential of tailored deep learning approaches to revolutionize manufacturing paradigms, fostering innovation in materials science and engineering.