http://s.uconn.edu/meseminar9/17/21

Abstract: Developing polymers with desirable properties has historically relied on trial-and-error, which can take long time, and there is no guarantee of success. Machine learning has become an integral part of materials design, and it can potentially impact polymer development in a positive way. However, the lack of open-source data has impeded the development of machine learning-guided polymer development, i.e., polymer informatics. In this talk, I will discuss our effort in generating polymer structures using machine learning models trained on existing polymer database. I will describe the process we used to generate the PI1M database (PI1M refers to 1 million polymers for Polymer Informatics) . I will also talk about the challenges ahead of the continued research in this direction. In addition, I will introduce our work in generating polymer properties using molecular dynamics simulations and discuss the challenges to be addressed. Machine learning models for several polymer properties (e.g., thermal properties, gas separation performance) will be presented. I will also highlight the need of a greater community effort to advance the polymer informatics field.

 

Biographical Sketch: Dr. Tengfei Luo is a Professor in the Department of Aerospace and Mechanical Engineering (AME) with a concurrent appointment in the Department of Chemical and Biomolecular Engineering (CBE) at the University of Notre Dame (UND). Before joining UND, he was a postdoctoral associate at MIT (2009-2011) after obtaining his PhD from Michigan State University (2009). Dr. Luo’s research focuses on exploring the chemistry-conformation-property relationships of polymers using molecular simulations, machine learning and experiments. He is an ASME Fellow (2019), JSPS Invitational Fellow (2019), DuPont Young Professor Awardee (2016), DARPA Young Faculty Awardee (2015), and Air Force Summer Faculty Fellow (2015). 

---