Biography
Yupeng Zhang is currently with the Department of Mechanical and Aerospace Engineering at the University of California, Los Angeles (UCLA). He received his Ph.D. in Materials Science and Engineering from Texas A&M University, specializing in solid mechanics, Bayesian statics, and materials characterization, followed by postdoctoral positions at Northwestern University, and the California Institute of Technology. Zhang is a member of the ASCE - EMI Machine Learning in Mechanics Committee and the recipient of NAIRR grant as a single PI, the Future Faculty Symposium Travel Award from the Society of Engineering Science in 2023, the Clearfield Materials Fellowship at Texas A&M University, and the Mitacs Globalink Research internship from Mitacs Canada. Zhang has mentored over ten graduate and undergraduate students and serves as a reviewer for such as Journal of Applied Mechanics (ASME), Journal of Engineering Mechanics (ASCE), Mechanics of Materials (Elsevier), European Journal of Mechanics / A Solids (Elsevier), Journal of Materials Research (Springer), Journal of Engineering Materials and Technology (ASME), Experimental Mechanics (Society for Experimental Mechanics, SEM), Physics of Fluids(AIP). His research interests include solid mechanics, materials characterization Bayesian statistics, inverse problems, and AI/machine learning, focusing on multiscale modeling of complex system, and thermo-magneto-mechanical couplings.
SRP Project Title
Iterative learning for materials and structures
NAIRR Project
Geometry Effects on Iterative Learning for Multiscale Modeling of History-Dependent Metamaterials
Topical Areas
Applied Computer Science; Applied Mathematics; Artificial Intelligence and Intelligent Systems; Civil Engineering; Computer Science; Geology and Solid Earth Sciences; Informatics, Analytics and Information Science; Infrastructure and Instrumentation; Materials Engineering; Mechanical Engineering; Other Computer and Information Sciences; Other Engineering and Technologies; Statistics and Probability; Visualization and Human-Computer Systems
Abstract
A big challenge in advancing AI methods that enable scientific discovery is to understand what kinds of data are necessary to obtain accurate and transferable surrogate models. We explored this question in the context of the history-dependent behavior of materials and structures. We introduced an iterative approach where we used a rich arbitrary class of trajectories to train an initial model. We then iteratively updated the class of trajectories with those that arise in large-scale simulation and used transfer learning to update the model. We showed that such an approach converges to a highly accurate surrogate, and one that is transferable. In our current NAIRR Pilot project, we are investigating how geometry influences iterative learning in multiscale modeling of history-dependent metamaterials. Through the Sustainable Research Pathways (SRP) Program, we aim to develop AI/ML-based digital twins of physical materials and structures under various boundary conditions. The digital twins will serve as fast and reliable surrogates for multiscale modeling, optimization, and inverse design. This research connects AI/ML with mechanics, materials, structures, and design. Participants will gain experience with machine learning, numerical simulation, with applications spanning mechanical civil, and materials engineering.
Desired Skills
Participants with interests in artificial intelligence, machine learning, numerical modeling, or materials / structures / mechanics are encouraged to apply. A background in engineering, computer science, applied math, applied statistics, or related fields is helpful, but not required. Curiosity and enthusiasm are especially valuable.
Lightning Talk Title
AI/ML model for complex mechanical systems
Keywords
solid mechanics, materials characterization, Bayesian statistics, inverse problems, AI/ML, multiscale modeling of complex system, thermo-magneto-mechanical couplings.