Principal investigator: Qian Wang
University: The Pennsylvania State University
Industry partners: Autodesk
Laser-based additive manufacturing (AM) processes involves many process parameters that affect the final geometry, mechanical properties, material microstructure, and surface roughness. Existing analytical models are often restricted by oversimplified assumptions and are thus not suitable for real applications, whereas high-fidelity numerical models such as finite element analysis-based models can be computationally expensive to be used in real-time build control. This project proposes a machine learning approach to model the relationship between process parameters and the build geometry. It will also include an online learning method where the process control can learn from newly available data sets to vary process parameters and improve build geometry accuracy and build quality. A suite of machine learning algorithms will be examined for their efficacy in predicting build geometry and tuning process parameters through learning to achieve the target accuracy of build geometry and target build quality. Project success will help reduce the amount of trial and error currently required in the AM industry and thus will help reduce the associated costs.
In this project, the PI will collaborate with Autodesk, whose State College Pennsylvania office is part of the development team for Autodesk’s pioneering Netfabb AM software. The PI and her graduate student funded under this project will have access to Netfabb to generate simulated training data sets on various builds with different process parameters. We expect that the results of this project will help advance the current capability of the Netfabb software, especially towards real-time build control. Such a partnership between Penn State and a world-leading AM software manufacturer will help bridge the gap between academic research and industrial use in the AM society and help position the Pennsylvania Commonwealth at the frontier of the field.