ECE Seminar Series - Joel B. Harley

Artificial Intelligence with Interdisciplinary Physical Knowledge

Joel B. Harley
Associate Professor & Kent and Linda Fuchs Faculty Fellow, Department of Electrical and Computer Engineering

Abstract:
Artificial intelligence (AI) is increasingly pervasive in our daily lives. However, its widespread adoption in many physically grounded disciplines remains limited. Three critical challenges face the integration of AI into these fields that must be addressed in the coming decade:

collecting sufficient amounts of data is difficult for practical and financial reasons, capturing the data diversity for informed decision-making can be impossible due to insufficient laboratory conditions or lack of observations (i.e., rare events), data-driven analytics are not interpretable or explainable as with physics-driven techniques or expert-driven systems.

This talk presents examples addressing these challenges for two physical domains, focusing on microstructure mechanics in materials science and human biomechanics. Specifically, we investigator strategies for building deep learning surrogate models within these domains. We will examine why traditional shallow and deep learning methods struggle with these challenges and explore how designing architectures with physical processes in mind can address complexities in the underlying data. We also discuss how the integration of physics and machine learning can improve the interpretability of our models to aid in basic science.

Biosketch:
Dr. Joel B. Harley is an Associate Professor and Kent and Linda Fuchs Faculty Fellow in the Department of Electrical and Computer Engineering at the University of Florida. His research integrates physical knowledge with machine learning to advance fundamental science and engineering. Dr. Harley has received multiple departmental and college-level teaching awards and is the recipient of the 2021 Achenbach Medal from the International Workshop on Structural Health Monitoring, the 2020 IEEE Ultrasonics, Ferroelectrics, and Frequency Control Society Star Ambassador Award, and the Air Force Summer Faculty Fellowship in both 2020 and 2018.

Dr. Harley has authored over 90 peer-reviewed journal papers and more than 90 conference proceedings. He currently serves as an Associate Editor for Structural Health Monitoring and as the Subject Editor for Artificial Intelligence and Machine Learning at NDT&E International. He has previously taught short courses on artificial intelligence at the Review of Quantitative Nondestructive Evaluation Conference as well as the Minerals, Metals & Materials Society (TMS) Annual Meeting.