题目：Accelerated High Fidelity Simulations of Reacting Flows Using Analytic and Data-Based Reduced Order Models
时间：2026年11月12日 10:00-11:00
地点：机械与动力工程学院新工程馆242会议室
邀请人：张家博 副教授（新能源动力研究所）
Biography
Hong G. Im is a Professor of Mechanical Engineering at KAUST and currently serves as Deputy Chair of the Clean Energy Research Platform. His research and teaching interests are primarily fundamental and practical aspects of combustion and power generation devices using high-fidelity computational modeling. He is a recipient of the NSF CAREER Award and SAE Ralph R. Teetor Educational Award, and has been inducted as an International Member of the National Academy of Engineering of Korea, a Fellow of the Combustion Institute and American Society of Mechanical Engineers (ASME) and an Associate Fellow of the American Institute of Aeronautics and Astronautics (AIAA). He has also served as an Associate Editor for the Proceedings of the Combustion Institute, and currently on the Editorial Board for Energy and AI.
Abstract
Combustion of conventional and renewable fuels continues to play a vital role in zero-carbon sustainable energy. Advances in computing power have enabled predictive simulations of complex phenomena involving turbulent transport and reactions in great detail, but complete description of the practically relevant problems remains a challenge due to the wide spectrum of physical and temporal scales. This presentation will provide an overview of historical developments in mathematical and computational approaches to reduced order models for accelerated high fidelity reacting flow simulations in modern computing hardware. First, mathematical theories of computational singular perturbation (CSP) will be discussed and its application to dimensionality reduction by skeletal mechanisms and accelerated solver algorithms will be described, including the recent breakthrough in hash-map reuse of chemical basis vectors. Next, the principal component analysis (PCA) will be presented as a classical data-based dimensionality reduction technique, followed by the successful implementation of the combined PCA-CSP approach. Finally, data-based construction of latent space time integration using the autoencoder and neural ODE (AE-NODE) will be discussed as a new paradigm of deep neural network (DNN) application to accelerated computing, with theoretical insights obtained by information theory.
学院OA系统
jAccount登录
ENGLISH
地址：上海市东川路800号上海交通大学闵行校区机械与动力工程学院
邮编：200240
Email：sjtume@sjtu.edu.cn