Artificial Intelligence for Materials Design and Process Optimization: AI for Materials Design and Process Optimization
Sponsored by: ACerS Glass & Optical Materials Division
Program Organizers: Adama Tandia, Corning; Venkatesh Botu, Corning Inc.

Monday 8:00 AM
November 2, 2020
Room: Virtual Meeting Room 41
Location: MS&T Virtual

Session Chair: Adama Tandia, Corning Incorporated


8:00 AM  
Physics-Informed Machine Learning for Predicting Glass Properties: Kai Yang1; Han Liu1; Mathieu Bauchy1; 1University of California, Los Angeles
    Data-driven modeling based on machine learning (ML) offers a promising route to develop robust composition-property models in glasses. However, traditional ML shows several limitations: (i) it requires a large amount of consistent data, (ii) it has a poor ability for extrapolation far from the training set, and (iii) it can potentially violate physics laws. To address these limitations, we present a new physics-informed ML framework that simultaneously leverages experimental measurements, simulation data, and physical knowledge. We show informing ML with physics-based knowledge greatly enhances the ability of ML models to extrapolate predictions from their training set—which is key to discover new glasses featuring properties that are very different from present glasses.

8:30 AM  
Stacking Fault Energy Prediction for Austenitic Steel: A Machine Learning Approach Aided by Thermodynamic Model: Xin Wang1; Wei Xiong1; 1University of Pittsburgh
    Stacking fault energy (SFE) plays an important role in the secondary deformation mechanism and mechanical properties of austenitic steels. An appropriate SFE will lead to the Transformation-induced plasticity/Twinning-induced plasticity and overcome the trade-off between strength and ductility. However, due to the complexity in the relationship between composition and SFE, there are no accurate and simple computational tools for modeling it. To solve this problem, we evaluate the CALPHAD-based thermodynamic models (CALPHAD: calculations of phase diagrams), and generate key attributes based on thermodynamic model to aid the machine learning (ML) algorithms get higher accuracy in predicting SFE. The results show that the thermodynamic-ML jointed model is more accurate and flexible than the existing models.