College of Science & Engineering
Twin Cities
This group's current research focuses on developing novel mathematical and computational approaches that enable deep understanding of natural and engineered systems. This will be achieved by leveraging knowledge-augmented artificial intelligence (AI) to discover hidden mechanisms together with high-performance physics-based simulation to model complex physics processes. Current research interests include meshfree materials modeling, damage and fracture mechanics, data-driven mechanics, reduced-order modeling, structural inverse design, and physics-informed deep learning for geophysical inverse problems. They are also interested in the general topics about encoding mechanics models in machine learning methods as well as advancing machine learning in mechanics. Three projects using MSI include:
- Developing a novel neural-integrated meshfree solver, which will provide an efficient solution framework to advances understanding of the mechanical behaviors of solid materials. This can be applied to model metal additive manufacturing process as well as the fracture mechanism. This approach will also be applied to topology optimization of programmable materials.
- Developing a physics-informed surrogate model to simulate large scale geosystem, e.g., ice sheet melting process, by using a neural network operator learning approach together with physics-informed machine learning. This coupled framework will provide 100X the speed of conventional numerical solvers.
- Introducing novel machine learning (ML) tools to learn the InSAR images together with other available optical remote sensing data to enhance the capacity of continuously monitoring deformation rate across large spatial regions. The developed InSAR and ML model will be validated on a well selected synthetic dataset and the study areas to demonstrate its accuracy and robustness over a variety of geological and environmental conditions.