Powder bed fusion with laser beam describes a popular additive manufacturing technique that allows for the creation of complex three-dimensional shapes for lightweight construction. However, the current melting and solidification processes may introduce defects that lead to printed components that do not meet the desired product quality requirements and standards. Automated process monitoring may aid in exhausting the full potential of powder bed fusion by reducing rejects, saving resources at the same time, and subsequently ensuring high product quality. We therefore propose to utilize machine learning algorithms with training data obtained directly from in situ measurements using acoustic emissions sensors as well as numerically from supplementary acoustics simulations. Here we outline the project and give the strategic roadmap for developing reliable methods that are capable of recognizing deviations from common system operations in the printing process due to defects and other artifacts. This work includes a preview of intermediate results from first machine learning experiments. Additionally, an early comparison of measurement and simulation data is given.
ICS file for iCal / Outlook