Collaborative Research: Process-Informed Latent Space Representation, Learning, and Monitoring for Smart Personalized Manufacturing

This collaborative project will contribute to the advancement of national prosperity and economic welfare by enabling process monitoring in personalized manufacturing, specifically one of a kind parts produced using additive manufacturing processes with complex geometries and novel materials. Monitoring personalized manufacturing processes for fabrication of high variety low volume products is a daunting task, requiring novel dimension-reduction and change detection methods beyond existing frameworks for mass production. This project will provide methodologies to reduce complexity in data representation, learning, and quality control for personalized manufacturing, facilitating the adoption of cost-effective personalized manufacturing technologies through improved product quality and simplified supply chains. The PIs will develop interdisciplinary curricular materials to train the next generation of manufacturing analytics workforce. This project will establish a new latent space monitoring methodology based on process-informed dimension reduction of the shape space for geometric quality control in smart personalized manufacturing. The methodology exploits the concept of manufacturing primitives to construct a process-informed latent space representation of 3D shapes. This representation enables the development of efficient domain-informed algorithms for (1) learning of the in-control shape quality distribution, (2) transfer learning of shape quality distributions between different process settings, and (3) latent space monitoring of part-to-part shape quality. The developed methodology will be validated in both metal-based wire-arc and polymer-based additive manufacturing processes. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.