
Astani Civil and Environmental Engineering Seminar
Tue, Nov 12, 2019 @ 04:00 PM  05:00 PM
Sonny Astani Department of Civil and Environmental Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Qizhi He, Pacific Northwest National Laboratory (PNNL)
Talk Title: Machine Learning Enhanced Computational Mechanics: ReducedOrder Modeling and PhysicsInformed DataDriven Computing
Abstract:
Advances in the field of machine learning and the increasing availability of data from laboratory/field observations and highfidelity simulation are quickly changing how the world solves important scientific problems. In this talk, I will present our research on developing integrative computational methods that leverage both physicsbased models and datacentric techniques to address various challenges in computational mechanics related to civil, mechanical, and biological applications.
First, I will introduce physicspreserving reduced order schemes that employ simulation data and subspace projection to reduce computational cost while preserving critical physical properties in modeling fracture mechanics and thermal fatigue of electronic packages. Second, I will describe a manifold learningbased datadriven computing approach for modeling complex materials, where physicsbased simulation proceeds interactively with the local data manifold reconstructed from experimental data,circumventing the limitations of using phenomenological constitutive models.
Finally, I will discuss our recent work in PNNL on developing machine learning framework based on deep neural networks for discovering hidden physics and modeling subsurface flow and transport in heterogenous porous media. The proposed approach allows the seamless fusion and integration of measurements from multiphysics systems and the information provided by the physical conservation laws, which regularize the neural networks as informative priors. The effectiveness of meshfree methods for these datadriven computing will also be demonstrated.
This work is supported by NSF, US Army Engineer Research, and the DOE ASCR.
Biography: Dr. Qizhi He is currently a Postdoctoral Research Associate in the Advanced Computing, Mathematics and Data Division at Pacific Northwest National Laboratory (PNNL). He obtained his B.S. in Engineering Mechanics from Wuhan University, M.S. in Computational Mechanics from Dalian University of Technology, and M.A. in Applied Mathematics and Ph.D. in Structural Engineering from the University of California, San Diego (UCSD). His research mainly focuses on the development of machine learning enhanced computational methods to model complex biological and civil systems and provide solutions to protect reallife structures from extreme hazardous events. His work combines the concepts from physicsbased modeling, machine learning, and meshfree type approximation and discretization
Host: Dr. Roger Ghanem
Location: Kaprielian Hall (KAP)  209
Audiences: Everyone Is Invited
Contact: Evangeline Reyes