Utilizing Sophisticated CFD Techniques to Study the Melt Pool Dynamics in Additive Manufacturing Processes

Abstract

The employment of synchrotron-based experimental validation, high-resolution X-ray imaging, and detailed microstructural investigations provides substantial confirmation of the proposed design’s accuracy and reliability. These advanced diagnostic techniques enable precise visualization of internal features, thereby facilitating a direct correlation between simulated and experimental outcomes. To further enhance predictive capability under conditions of limited labeled data, the approach incorporates Physics-Informed Neural Networks (PINNs), which integrate physical laws into the learning process. This framework enables the optimization of critical processing parameters to minimize keyhole porosity and lack-of-fusion defects, two predominant issues that compromise the integrity of metal additively manufactured components. The integration of such experimental and computational methodologies represents a holistic, cross-sectional approach that strengthens the understanding of defect formation mechanisms and supports the refinement of process–structure–property relationships. Furthermore, this strategy contributes to the development of robust, closed-loop control systems in metal additive manufacturing, thereby advancing the field toward more reliable, defect-free, and quality-assured production.