Feature engineering of CAM geometry for the Control of the L-PBF process

The additive manufacturing (AM) laser powder bed fusion (L-PBF) process involves a complex combination of multi-physics and multi-scale phenomena that usually generate unpredictable quality defects, undermining the process’s reliability and repeatability. These defects are often traced back to the process’s thermal history, defined by the spatiotemporal sequence of laser-scanned regions. Fortunately, using machine learning to analyze printing data offers a promising approach to understanding the complexity of heterogeneous data - in terms of nature, dimensionality, and quantity - within printing datasets. Thus, the proposed paper aims to extract and synthesize key features for improved process regularities and quality identification.