Title:
Knowledge-Informed Machine Learning for Concrete Property Prediction
Author(s):
Zhanzhao Li
Publication:
Web Session
Volume:
Issue:
Appears on pages(s):
Keywords:
DOI:
Date:
10/29/2023
Abstract:
While machine learning has become a powerful complement to empirical analysis and physical modeling in predicting concrete properties, its capabilities are yet to be fully exploited due to the massive data requirements and generalizability challenges. In this presentation, we will introduce a knowledge-informed machine learning framework that enables one to integrate domain knowledge with experimental data and overcome these major challenges. This framework is shown (1) to drastically accelerate model convergence and increase model performance even with limited training data, (2) to guarantee generalizability to realistic scenarios, making it reliable under extereme real-world conditions, and (3) to flexibly incorporate knowledge from not only empirical formulas but also physics-based models for robust property prediction and materials discovery.