讲座时间：2025年12月29日星期一上午9:30
讲座地点：X30224
讲座题目：Pre-trained Surrogate Models in A Multiscale Computational Homogenization Framework
主 讲 人：Poh Leong Hien
个人简介：
Poh Leong Hien is an Associate Professor in the Department of Civil and Environmental Engineering (CEE) at the National University of Singapore (NUS). He is currently the Deputy Head (Graduate and CET Programmes) in the CEE department, as well as the Director of the NUS Centre for Protective Technology. Beyond NUS, he has served as the elected chair of the ASCE EMI Modeling Inelasticity and Multiscale Behavior committee (2023-2025), and is an elected council member of the Institution of Engineers, Singapore. He is interested in multi-scale methods and various aspects of materials modelling, with a focus on the damage and failure response of materials and structures. He has received the Asian-Pacific Association for Computational Mechanics Young Investigator Award in 2019, the Inaugural Young Researcher Award in the 4th International Conference on Damage Mechanics 2023, as well as several teaching awards in NUS over the years. He currently serves in the editorial board of Engineering Failure Analysis, Finite Element in Analysis and Design, International Journal of Damage Mechanics and the Early Career Editorial Board of International Journal for Numerical Methods in Engineering.
讲座内容：
A composite material typically exhibits complex behavior at the engineer scale, arising from the interactions between its underlying constituent phases. It is generally difficult to develop an engineering model that adequately captures the essential micro mechanisms that propagate onto the macro scale. To this end, the multiscale computational homogenization method enables a consistent coupling across length scales. However, the typical computational homogenization method is still computationally too expensive for most practical problems. In this presentation, we address this bottleneck with an offline development of a microscopic surrogate model for a given micro-structure, to be incorporated into a standard nonlinear FE framework, for rapid online implementations at the macro scale. For the offline training phase, we adopt the transformer-based architecture within a pre-training and fine-tuning framework. To reduce the data generation cost, a simplified source representative volume element (RVE) is utilized, to rapidly generate a huge source dataset for a pre-training process. The surrogate model is incorporated into a macro FE framework, and its predictive capabilities illustrated via different examples.
