Speaker:

Dr. John Hedengren

Professor

Chemical Engineering Department

Brigham Young University

Utah, USA

Abstract:

Physics-informed deep learning is a technique that combines machine learning with physics-based information to enhance predictive accuracy. This hybrid model takes advantage of advanced mathematical models and data-driven methods to improve predictions. However, traditional physics-based models can be too slow and complex, making them challenging to use. A known weakness of purely data-driven approaches is the extrapolation potential. This presentation explores how hybrid models can be applied in automation and optimization solutions, addressing the limitations of standard artificial neural networks that have poor extrapolation potential when used outside the training region. Physics-informed models offer new possibilities for combining physics-based and data-driven elements. Challenges and opportunities are discussed for applications in drilling automation, thermophysical property prediction, and model predictive control.

Speaker Bio :

Prof. John Hedengren leads the Process Research and Intelligent Systems Modeling (PRISM) group at Brigham Young University with a focus on physics-informed machine learning for optimization of energy systems. He received his doctorate degree from the University of Texas at Austin and worked as a consultant and senior engineer for 7 years on advanced control and optimization for polymers. His work includes the Python GEKKO optimization suite and the Arduino-based Temperature Control Lab that is used by 70 universities for process control education. His 85 publications span topics of data science, machine learning, smart grid optimization, unmanned aerial systems, and predictive control. He teaches online courses on machine learning, data science, advanced control, and introduction to engineering programming that are accessed by 30,000 students each week. He served as a Distinguished Lecturer for the Society of Petroleum Engineers in 2018-2019. He is the recipient of the 2014 David Himmelblau award and the 2018 Computing Practice award from the American Institute of Chemical Engineers CAST Division. He is the Communications Chair of the American Automatic Control Council and Chair of the IEEE CSS Control Education technical committee.