Data-driven analysis of geometric factors that impact LPBF part quality

Laser powder bed fusion (LPBF) enables engineers and designers to manufacture a much wider range of part geometries, often more quickly than would be possible with traditional manufacturing processes. However, LPBF parts are prone to part deformation, strain, and residual stress from the manufacturing process due to temperature gradients during fabrication. Practitioners must rely on costly experimentation or time-consuming simulations to predict such outcomes and incorporate design changes into their geometry to minimize such negative effects. In this research, we use a dataset of more than 1,000 topology optimized CAD parts with corresponding LPBF process finite element analysis simulations using Ti-6Al-4V material to derive data-driven design guidance. This dataset is unique in its high geometric variability, enabling unique insights for a wide range of manufacturable parts. We explore the impact of geometric factors (e.g. base contact area, overhanging surface area, bounding box size) on part deformation to quantify the relative influence of each variable on final part quality. Guidelines for how to design parts to minimize process-induced distortion are provided, allowing practitioners and design tool developers to incorporate these guiding principles into their work.