Evaluating and correlating multimodal process dynamics, microstructure features, and mechanical properties in laser powder bed fusion

Laser powder bed fusion (LPBF) in additive manufacturing holds the potential for efficiently producing high- resolution components with intricate geometries. However, LPBF-printed parts often exhibit deformation, defects, and suboptimal mechanical performance, limiting their applications in critical industries. The melt pool characteristics, spatters, and in-process layer surface properties play a crucial role in determining the microstructure formation and defect generation during LPBF, consequently affecting the properties of printed components. This work aims to develop a framework for revealing the relationships between complex LPBF process dynamics, microstructure, and mechanical properties, utilizing the authors' unique in-situ multi-sensor monitoring big data. The study investigates the relationships between process signatures-such as melt pool geometry, temperature, spatter, and layer surface features-and outcomes like grain characteristics, hardness, and fatigue life, using support vector machine regression models. It reveals the importance of acquiring and combining physically meaningful quantities like absolute melt pool temperature, spatter count, and in-process layer surface roughness for accurate part property prediction. These approaches outperform traditional intensity-based monitoring methods. The demonstrated framework of multi-sensor in-situ monitoring and multimodal feature fusion promises to significantly enhance the understanding and optimization of LPBF processes for producing advanced materials and components with sophisticated designs.