報告時間：2023年6月13日9:30

報告地點：中以樓二層會議室

報告題目：Physics-Based and Data-Driven Modeling in Environmental Hydraulics

報告人：劉小峰 美國賓夕法尼亞州立大學長聘副教授（Tenured Associate Professor）

劉小峰，美國賓夕法尼亞州立大學土木與環境工程系長聘副教授，賓州州立大學計算與數據科學研究所、能源與環境研究所客座研究員。2000年本科畢業于清華大學水利工程專業，2003年碩士畢業于北京大學環境科學專業，2008年博士畢業于美國伊利諾伊大學厄巴納-香檳分校（UIUC）土木工程專業。2009年至今分別在UIUC、德克薩斯大學安東尼奧分校以及賓州州立大學擔任博士后、助理教授、長聘副教授。長期從事計算流體力學、泥沙輸移、環境流體力學相關研究，尤其專注于環境和水資源工程問題中的數值計算模型開發和應用。從美國NSF、美國墾務局、NCHRP、戰略環境研究與發展計劃等機構獲批項目經費累計超過430萬美元，在《Geophysical Research Letters》、《Water Resources Research》、《Journal of Computational Physics》、《Coastal Engineering》、《Journal of Hydraulic Engineering》等領域內頂級期刊發表論文100余篇。現任《Journal of Hydraulic Engineering》期刊副主編。榮獲2020年美國土木工程師協會（ASCE）State-Of-The-Art of Civil Engineering Award，以及2020年賓州州立大學Harry West Teaching Award。

報告摘要：

Physics-based and data-driven models are two different, yet closely related, approaches for capturing processes in real world. This talk will provide a broad overview of both approaches and then showcase some of our example projects in environmental hydraulics. Physics-based models (PBMs) solve governing equations using different numerical schemes. PBMs are founded on our knowledge of physical processes and their mathematical representations (usually in the form of partial differential equations and constitutive relationships). I will show examples models for turbulent flow and sediment transport around structures/objects. Data-driven models rely on the amount, representativeness, and quality of data to implicitly describe the physical processes. In high-dimensional space, data should be “big” enough to fully capture the input-output dynamics of a physical system. I will show examples of a deep-learning based surrogate for a 2D computational hydraulics model and its use for parameter inversion. Over time, the boundary between physics-based and data-driven models is blurred thanks to the rapid advancement in Machine Learning/AI. Hybrid approaches, such as physics-informed machine learning, have emerged. Some future research topics for environmental hydraulics will be discussed.

相關論文：

https://arxiv.org/abs/2112.10889

https://arxiv.org/abs/2203.02821

聯系人：徐云成    ycxu@cau.edu.cn