• Corpus ID: 233481835

Learning fluid physics from highly turbulent data using sparse physics-informed discovery of empirical relations (SPIDER)

@inproceedings{Gurevich2021LearningFP,
  title={Learning fluid physics from highly turbulent data using sparse physics-informed discovery of empirical relations (SPIDER)},
  author={Daniel Gurevich and Patrick A. K. Reinbold and Roman O. Grigoriev},
  year={2021}
}
We show how a complete mathematical description of a complicated physical phenomenon can be learned from observational data via a hybrid approach combining three simple and general ingredients: physical assumptions of smoothness, locality, and symmetry, a weak formulation of differential equations, and sparse regression. To illustrate this, we extract a system of governing equations describing flows of incompressible Newtonian fluids – the Navier-Stokes equation, the continuity equation, and the… 

Figures and Tables from this paper

Physically-informed data-driven modeling of active nematics
A continuum description is essential for understanding a variety of collective phenomena in active matter. However, building quantitative continuum models of active matter from first principles can be
Ensemble-SINDy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and control
TLDR
This work leverages the statistical approach of bootstrap aggregating (bagging) to robustify the sparse identification of the nonlinear dynamics (SINDy) algorithm and shows that ensemble statistics from E-Sindy can be exploited for active learning and improved model predictive control.
Data-driven detection of drifting system parameters
TLDR
This paper demonstrates how machine learning can be leveraged to detect changes in global parameters from variations in an identified model using only observational data, and how this capability, when paired with first principles analysis, can effectively distinguish the effects of changing parameters from the intrinsic complexity of the system.

References

SHOWING 1-10 OF 21 REFERENCES
Robust and optimal sparse regression for nonlinear PDE models.
TLDR
This paper investigates how models of spatiotemporal dynamics in the form of nonlinear partial differential equations can be identified directly from noisy data using a combination of sparse regression and weak formulation, achieving accuracy that is orders of magnitude better than what existing techniques allow.
Robust learning from noisy, incomplete, high-dimensional experimental data via physically constrained symbolic regression
TLDR
It is demonstrated that combining a data-driven methodology with some general physical principles enables discovery of a quantitatively accurate model of a non-equilibrium spatially extended system from high-dimensional data that is both noisy and incomplete.
Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations
TLDR
Hidden fluid mechanics (HFM), a physics-informed deep-learning framework capable of encoding the Navier-Stokes equations into the neural networks while being agnostic to the geometry or the initial and boundary conditions, is developed.
Turbulence forecasting via Neural ODE.
Fluid turbulence is characterized by strong coupling across a broad range of scales. Furthermore, besides the usual local cascades, such coupling may extend to interactions that are non-local in
Using noisy or incomplete data to discover models of spatiotemporal dynamics.
TLDR
This work presents an alternative and rather general approach that addresses sparse regression difficulty by using a weak formulation of the problem, which allows accurate reconstruction of PDEs involving high-order derivatives, such as the Kuramoto-Sivashinsky equation, from data with a considerable amount of noise.
Data-driven discovery of partial differential equation models with latent variables
TLDR
It is shown that local polynomial interpolation combined with sparse regression can handle data on spatiotemporal grids that are representative of typical experimental measurement techniques such as particle image velocimetry, but it is found that the reconstructed model is sensitive to measurement noise and trace this sensitivity to the presence of high-order spatial and/or temporal derivatives.
Nonlinear multiobjective optimization
  • K. Miettinen
  • Chemistry
    International series in operations research and management science
  • 1998
TLDR
This paper is concerned with the development of methods for dealing with the role of symbols in the interpretation of semantics.
Extracting Sparse High-Dimensional Dynamics from Limited Data
TLDR
This work details three sampling strategies that lead to the exact recovery of first-order dynamical systems when the authors are given fewer samples than unknowns, which is stable to the sparse structure of the governing equations and robust to noise in the estimation of the velocity.
Spatiotemporal system reconstruction using Fourier spectral operators and structure selection techniques.
TLDR
A technique based on trigonometric spectral methods and structure selection for the reconstruction of spatiotemporal systems governed by nonlinear partial differential equations of polynomial type with terms of arbitrary derivative order and nonlinearity degree leads to a parsimonious model of the original system.
...
...