Controlling local melt pool dimensions in laser powder bed fusion through multiscale physics-driven optimization

Laser Powder Bed Fusion is an additive manufacturing technique used to produce solid metal parts from metallic powder. It excels at producing complex geometries in small batches. However, processing conditions are being optimized using time and labor intensive trial and error approaches. Even after optimization, the contrast between the complex geometry and the limited set of optimized processing conditions yields a compromise between printability and melt pool homogeneity at best. We propose a multiscale modeling approach to obtain a physics-driven prediction of the space and time evolution of processing conditions tailored to a specific geometry. The workflow consists of two modeling techniques utilized to capture the physics of the process, both at the part scale and at the melt-pool-scale. The optimization strategy aims to minimize melt pool variations in the vicinity of geometric features. The numerical methods and optimization algorithm are detailed, convergence studies are presented, and the effectiveness of the optimization is verified by high speed thermal imaging and X-ray tomography. This physics based approach circumvents any a priori data acquisition and modeling campaigns and provides efficient means for geometry tailored process optimization mitigating up to 88% of melt pool variations in the vicinity of challenging geometric features.