Distinguished Seminar in Computational Science and Engineering
March 23, 2023, 12 PM
Automated Model Discovery – A new paradigm in engineering science?
Ellen Kuhl
Stanford University
Abstract
Constitutive modeling and parameter identification are the cornerstones of engineering science. For decades, the gold standard in constitutive modeling has been to first select a model and then fit its parameters to data. However, the scientific criteria for model selection remain poorly understood, and the success of this approach depends largely on user experience and personal preference. In this seminar, I will propose a new method that simultaneously and fully autonomously discovers the model, parameters, and experiment that best explain a wide variety of soft matter systems. Mathematically, model discovery translates into a complex non-convex optimization problem. We solve this problem by formulating it as a neural network, and leverage the success, robustness, and stability of state-of-the-art optimization tools from deep learning. However, we do not use classical off-the-shelf neural networks, which are known to overfit sparse data, violate the fundamental laws of physics, and introduce parameters without a physical meaning. Instead, we design our own family of constitutive neural networks with activation functions that feature popular constitutive models and parameters that have a clear physical interpretation. I will illustrate how to train, benchmark, and validate these networks for a range of soft materials including rubber, skin, and the human brain. I will show that, out of thousands of possible models, our network robustly discovers a unique constitutive model that outperforms traditional models and, at the same time, identifies the best experiment to train itself. Constitutive artificial neural networks could initiate a paradigm shift in constitutive modeling, from user-defined model selection to automated model discovery, which could forever change how we simulate engineering systems.
Bio
Ellen Kuhl is the Walter B. Reinhold Professor in the School of Engineering and Robert Bosch Chair of Mechanical Engineering at Stanford University. She is a Professor of Mechanical Engineering and Bioengineering who received her PhD from the University of Stuttgart in 2000 and her Habilitation from the University of Kaiserslautern in 2004. Ellen’s research integrates physics-based modeling with machine learning and creates interactive simulation tools to understand, explore, and predict the dynamics of living systems. She has published more than 200 peer-reviewed journal articles and a textbook on Data-Driven Modeling of COVID-19. Ellen is a founding member of the Living Heart Project, a translational research initiative to revolutionize cardiovascular science through realistic simulation with 400 participants from research, industry, and medicine from 24 countries. She is the current Chair of the US National Committee on Biomechanics and a Member-Elect of the World Council of Biomechanics. Ellen is a Fellow of the American Society of Mechanical Engineers and of the American Institute for Mechanical and Biological Engineering. She received the National Science Foundation Career Award in 2010, was selected as Midwest Mechanics Seminar Speaker in 2014, and received the Humboldt Research Award in 2016 and the ASME Ted Belytschko Applied Mechanics Award in 2021. Ellen is an All American triathlete, a multiple Boston marathon runner, and a three-time Kona Ironman World Championship qualifier.
Automated Model Discovery – A new paradigm in engineering science?
Ellen Kuhl
Stanford University