学术活动

博鱼在线注册(中国)有限公司官网 > 博鱼在线注册(中国)有限公司官网 > 学术活动 > 正文

博鱼在线注册(中国)有限公司官网

何启志教授学术报告

发布日期:2023-07-09 作者: 来源: 点击数:

主题:MachineLearning Enhanced Computational Mechanics

主讲人:何启志,美国明尼苏达大学教授

邀请人:唐旭海 教授

时间:2023年07月14日下午3:00-5:00

地点:博鱼在线注册(中国)有限公司官网十教南414会议室

主讲人简介:

Dr.Qizhi(“KaiChi") He joined the Department of Civil, Environmental, and Geo- Engineering at the University of Minnesota as an Assistant Professor in January 2022. He received his M.A. in Applied Mathematics (2016) and Ph.D. in Structural Engineering and Computational Science (2018) from UC San Diego. Before joining UMN, he worked as a postdoctoral research associate in the Scientific Machine Learning Group at Pacific Northwest National Laboratory (PNNL). Dr. He's research interests lie at the intersection of computational mechanics, scientific machine learning, and multiscale andmultiphysicsmodeling, with a focus on advancing the fundamental understanding and predictive capability for natural and engineered systems involving porous and composite materials.

Abstract:

Recentadvancements in machine learning and data-driven approaches provide new possibilities for the modeling and simulation of complex problems in applied science and engineering. In this talk, we will review our group’s work on developing physics-informed machine learning approaches for various forward and inverse problems in computational mechanics and beyond, with a focus on applications that involve inherent material complexities such as nonlinearities and heterogeneities. First, we discuss the effectiveness of the physics-informed neural network (PINN) method on various subsurface and mechanics problems, including steady and time-dependent forward modeling and parameter estimation. We will also introduce a hybrid computational framework based on integrating finite element schemes and data-driven learning models for solving different coupled systems, e.g., ice sheet dynamics. We demonstrate that the proposed network models can learn various explicit operators and underlying physics, enabling reduced-order modeling of nonlinear high-dimensional problems.