Improving Additive Manufacturing Through Physics-Aware Deep Learning

Additive manufacturing (AM), more commonly known as 3D printing, is an exciting new set of technologies that shows great promise, since it allows greater design freedom and has less material waste than conventional manufacturing methods. A problem that needs to be solved in order to expand AM for large-scale manufacturing is that process-induced defects affect quality.

In a project called “Multiphysics Data Assimilation Framework Based on Process-Aware Neural Operator for Failure Prediction in Additive Manufacturing,” Assistant Professor Qizhi He (Civil, Environmental, and Geo- Engineering; MSI PI) and Assistant Professor Ju Sun (Computer Science and Engineering; MSI PI) are developing a novel knowledge-augmented machine learning tool that will quickly and reliably predict defects, using thermomechanical models and process monitoring data. The study focuses on laser powder bed fusion (LPBF), a type of AM technology that has been widely used in industries for a wide spectrum of materials including metals, polymers, and ceramics. This research will advance understanding of the process-structure-properties relation and hidden defect mechanisms in metal AM. It will also promote the application of AI technology and information science to real-time data assimilation for extreme manufacturing conditions. 

This project recently received a UMII Seed Grant. UMII Seed Grant funds are intended to promote, catalyze, accelerate and advance UMN-based informatics research in areas related to the MnDRIVE initiative, so that U of M faculty and staff are well prepared to compete for longer term external funding opportunities. This Seed Grant falls under the Robotics, Sensors, and Advanced Manufacturing research area of the MnDRIVE initiative.

Research Computing partners:

• University of Minnesota Informatics Institute
• Minnesota Supercomputing Institute