发布时间:2021-09-13
报告人:雍稳安(清华大学)
邀请人:施保昌
报告时间:2021年9月16日(星期四)15:00-17:00
报告地点:腾讯会议ID:566 972 139
报告题目:Learning hyperbolic moment-closure models for the radiative transfer equation
报告摘要:In this talk, I will introduce our machine learning method to derive moment-closure models for the radiative transfer equation in slab geometry. To ensure good properties of the learned models, our main idea is to find a symmetrizer (a symmetric positive definite matrix) for the closure system. The resultant model can be shown to be globally hyperbolic and inherits the dissipativesness of the original kinetic equation. Several benchmark tests including the Gaussian source problem and two-material problem show good accuracy, correct diffusion limit and long-time stability of our machine learning closure model. This is a joint work with Juntao Huang, Yingda Cheng, Andrew J. Christlieb and Luke F. Roberts.
报告人简介:雍稳安,1992年于德国海德堡大学取得博士学位,2005年于海德堡大学取得德国教授资格(Habilitation),同年加盟清华大学,现为清华大学数学科学系长聘教授。研究兴趣包括应用偏微分方程、数值方法、非平衡态热力学、数学建模以及机器学习。主要学术贡献有:系统地创立了双曲松弛偏微分方程的数学理论、建立了格子Boltzmann方法的稳定收敛性理论、提出了正确描述可压缩粘弹性流体流动的数学模型、提出了非平衡态热力学的守恒耗散理论(CDF)等。是英国介观过程科学联盟(UKCOMES)的国际学术顾问,其在控制方面的工作曾被国际科技媒体Advances in Engineering遴选报道。