top
请输入关键字
力学系学术报告【4月7日】:What can Machine Learning help with Materials Modelling?



SEMINAR

 

SERIES

澳门太阳娱乐网站官网

 

力学与工程科学系

湍流与复杂系统国家重点实验室

 

What can Machine Learning help with Materials Modelling?


报告人:胥柏香德国达姆施塔特工业大学材料系教授
  间:2023年4月7日(周五)上午9:30
  点:澳门太阳娱乐网站官网1号楼210会议室


内容简介

Mechanical and functional properties of engineering solid materials rely essentially on their microstructure, which further depend on the process history of the materials. It is generally a challenging task of materials modelling to recapture these correlations. The classical physics-law driven measures, e.g. constitutive modelling and multiscale techniques like homogenization, have been the focus of materials modelling in the last centuries. They have extended and will remain to push the research front of materials modelling greatly. However, due to the complexity of the microstructure and the advance of manufacturing, materials modelling and the related optimization remain a challenging task, which goes far beyond the limit of the current methodologies alone. It has to be assisted by other methodologies. With the advent of novel data science methods and Machine Learning (ML) approaches, which are particularly promising to recapture intricate correlations and for data processing. In combination of the classical and modern measures, there are new ground-breaking opportunities for materials modelling and design.

In this work I will discuss chances, capabilities and issues of ML and data science approach to assist materials modelling and simulations. After introducing ML in a nutshell and recap of materials modelling, I will demonstrate through case studies, how ML can be used in constitutive modelling, computational mechanics, and multiscale simulations, as well as microstructure characterization and reconstruction. Moreover, I will also show a ML surrogate model, trained on large thermal simulation data, for predication of melt pool size of powder bed fusion additive manufacturing from all kinds of materials and process parameters.


报告人简介

胥柏香教授于2002年获得南京河海大学学士学位,2008年获得2138cn太阳集团古天乐博士学位,同年获得德国洪堡奖学金。2016年成为德国达姆斯塔特工学大学终身教授。近期研究课题着重于功能及能源材料多场耦合问题的理论建模及数值模拟,以及相关的数据驱动的多尺度模拟和机器学习。近五年在Science, Nature Materials, Materials Horizons, Nano Energy, JMPS, Acta Materialia,Int. J. Plasticity, IJSS等重要国际期刊上发表SCI检索论文100余篇。近五年主持或参与欧洲,德国及洲政府科研项目约15项。