Title: 

Digital Twin: Synthesize Data and Domain Science to Accelerate Additive Manufacturing Adoption

Abstract:

While additive manufacturing sees accelerated applications in volume production, higher production yield competing with mainstream manufacturing demands more sophisticated process insights, foresights and control to deliver final products in both geometry and function with precision. We develop Digital Twins that replicate virtually the dynamics of additive manufacturing processes and product evolution.  These Digital Twins rely on both domain sciences (for example, integrated computational material engineering) and data sciences (for example, deep learning) to deliver timely prediction and optimization. They exploit processing and sensing dataset collected to ensure prediction specificity to overcome process variations. We apply these Digital Twins to improve and automate product design correction and additive manufacturing process tuning.  

In this seminar, I would like to highlight two Digital Twin efforts anchored on graph neural network architecture: one generates surrogate models rooted in computational physics delivering simulations with order of magnitude speedup; the other is data driven, using metrology dataset to predict and then correct product defects.  We are in collaboration with Nvidia to open-source these physics-informed AI projects. My team is very excited to collaborate with professor Hui Wang to further research, develop, and expand these graphnet oriented Physics Informed AI efforts,  in particular, supply industrial usecases and datasets, facilitate validations, and champion for adoption of successful research outcome in production environment. 

I will conclude this talk by providing an overview of research efforts in my team and highlighting additional areas that we could potentially develop collaboration with broader FAMU/FSU research community. 

Speaker bio: 

Dr. Jun Zeng is HP’s Distinguished Technologist and Head of the Digital Twin group.  Jun has 20+ years of industrial experiences in creating and commercializing software for improving cyber-physical systems. ​His publications include a co-edited book on computer-aided Design and a co-authored book on digital factory, and 50+ peer-reviewed papers.​ He is inventor of 71 U.S. granted patents. His academic training includes Ph.D. in mechanical engineering and M.S. in computer science, both from Johns Hopkins University.  ​ He is ACM member, and IEEE senior member. ​