MS003 - Physics-Informed Machine Learning for Surrogate Models in Continuum Mechanics
Keywords: CFD, engineering design, neural operators, PINNs
As digital twins become increasingly integral to engineering design, there is a growing
demand for computational models that are capable of predicting complex physical
systems both accurately and efficiently, at best, in real-time. This mini-symposium is
dedicated to exploring the role of Physics-Informed Machine Learning methods—such
as Physics-Informed Neural Networks (PINNs [1, 2]) and neural operators [3, 4]—in the
creation of surrogate models specifically designed for digital twins in continuum
mechanics.
The focus will be on how these machine learning techniques can address real-world engineering challenges in fluid and structural mechanics, and how they can be integrated with and enhance established numerical methods such as finite element and isogeometric analysis. Key topics include innovations in model training and network architecture, improvements in computational efficiency, and advancements in managing geometrically complex domains.
We invite contributions that demonstrate practical applications of these methods for developing and refining surrogate models. Additionally, we encourage submissions that provide solutions for reducing training times, improving computational efficiency, and handling complex geometries within digital twin environments.
REFERENCES
[1] Lagaris E., Likas A., Fotiadis D.I., Artificial neural networks for solving ordinary and partial differential equations, IEEE Transactions on Neural Networks, 9, pp. 987-1000, 1998. [2] Raissi, M., Perdikaris P., Karniadakis G.E., Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, Journal of Computational Physics, 378, pp. 686-707, 2019.
[3] Lu L., Pengzhan J., Guofei P., Zhang Z., Karniadakis G.E., Learning nonlinear operators via DeepOnet based on the universal approximation theorem of operators, Nature Machine Intelligence, 3, pp. 218-229, 2021.
[4] Li Z., Kovachki N.B., Azizzadenesheli K., Liu B., Bhattacharya K., Stuart A., Anandkumar A., Fourier neural operator for parametric partial differential equations. In International Conference on Learning Representations, 2021
The focus will be on how these machine learning techniques can address real-world engineering challenges in fluid and structural mechanics, and how they can be integrated with and enhance established numerical methods such as finite element and isogeometric analysis. Key topics include innovations in model training and network architecture, improvements in computational efficiency, and advancements in managing geometrically complex domains.
We invite contributions that demonstrate practical applications of these methods for developing and refining surrogate models. Additionally, we encourage submissions that provide solutions for reducing training times, improving computational efficiency, and handling complex geometries within digital twin environments.
REFERENCES
[1] Lagaris E., Likas A., Fotiadis D.I., Artificial neural networks for solving ordinary and partial differential equations, IEEE Transactions on Neural Networks, 9, pp. 987-1000, 1998. [2] Raissi, M., Perdikaris P., Karniadakis G.E., Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, Journal of Computational Physics, 378, pp. 686-707, 2019.
[3] Lu L., Pengzhan J., Guofei P., Zhang Z., Karniadakis G.E., Learning nonlinear operators via DeepOnet based on the universal approximation theorem of operators, Nature Machine Intelligence, 3, pp. 218-229, 2021.
[4] Li Z., Kovachki N.B., Azizzadenesheli K., Liu B., Bhattacharya K., Stuart A., Anandkumar A., Fourier neural operator for parametric partial differential equations. In International Conference on Learning Representations, 2021
