Spline-PINN: Approaching PDEs without Data using Fast, Physics-Informed Hermite-Spline CNNs

Partial Differential Equations (PDEs) are notoriously difficult to solve. In general, closed-form solutions are not available and numerical approximation schemes are computationally expensive. In this paper, we propose to approach the solution of PDEs based on a novel technique that combines the advantages of two recently emerging machine learning based approaches. First, physics-informed neural networks (PINNs) learn continuous solutions of PDEs and can be trained with little to no ground truth data. However, PINNs do not generalize well to unseen domains. Second, convolutional neural networks provide fast inference and generalize but either require large amounts of training data or a physics-constrained loss based on finite differences that can lead to inaccuracies and discretization artifacts. We leverage the advantages of both of these approaches by using Hermite spline kernels in order to continuously interpolate a grid-based state representation that can be handled by a CNN. This allows for training without any precomputed training data using a physics-informed loss function only and provides fast, continuous solutions that generalize to unseen domains. We demonstrate the potential of our method at the examples of the incompressible Navier-Stokes equation and the damped wave equation. Our models are able to learn several intriguing phenomena such as Karman vortex streets, the Magnus effect, Doppler effect, interference patterns and wave reflections. Our quantitative assessment and an interactive real-time demo show that we are narrowing the gap in accuracy of unsupervised ML based methods to industrial CFD solvers while being orders of magnitude faster.

[1]  Barbara Solenthaler,et al.  Data-driven fluid simulations using regression forests , 2015, ACM Trans. Graph..

[2]  Nicholas Geneva,et al.  Modeling the Dynamics of PDE Systems with Physics-Constrained Deep Auto-Regressive Networks , 2019, J. Comput. Phys..

[3]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[4]  George E. Karniadakis,et al.  Hidden Fluid Mechanics: A Navier-Stokes Informed Deep Learning Framework for Assimilating Flow Visualization Data , 2018, ArXiv.

[5]  Gordon Wetzstein,et al.  Implicit Neural Representations with Periodic Activation Functions , 2020, NeurIPS.

[6]  Jos Stam,et al.  Stable fluids , 1999, SIGGRAPH.

[7]  J. Monaghan,et al.  Smoothed particle hydrodynamics: Theory and application to non-spherical stars , 1977 .

[8]  Paris Perdikaris,et al.  Physics-Constrained Deep Learning for High-dimensional Surrogate Modeling and Uncertainty Quantification without Labeled Data , 2019, J. Comput. Phys..

[9]  Nils Thuerey,et al.  Deep Learning Methods for Reynolds-Averaged Navier–Stokes Simulations of Airfoil Flows , 2018, AIAA Journal.

[10]  J. Templeton,et al.  Reynolds averaged turbulence modelling using deep neural networks with embedded invariance , 2016, Journal of Fluid Mechanics.

[11]  Matthew J. Zahr,et al.  Physics-informed graph neural Galerkin networks: A unified framework for solving PDE-governed forward and inverse problems , 2021, Computer Methods in Applied Mechanics and Engineering.

[12]  Meire Fortunato,et al.  Learning Mesh-Based Simulation with Graph Networks , 2020, ArXiv.

[13]  Daniel L. K. Yamins,et al.  Flexible Neural Representation for Physics Prediction , 2018, NeurIPS.

[14]  Nicholas Zabaras,et al.  Bayesian Deep Convolutional Encoder-Decoder Networks for Surrogate Modeling and Uncertainty Quantification , 2018, J. Comput. Phys..

[15]  G. Karniadakis,et al.  Physics‐Informed Neural Networks (PINNs) for Wave Propagation and Full Waveform Inversions , 2021, Journal of Geophysical Research: Solid Earth.

[16]  Changhoon Lee,et al.  Deep unsupervised learning of turbulence for inflow generation at various Reynolds numbers , 2020, J. Comput. Phys..

[17]  Shiyi Chen,et al.  LATTICE BOLTZMANN METHOD FOR FLUID FLOWS , 2001 .

[18]  Ilias Bilionis,et al.  Deep UQ: Learning deep neural network surrogate models for high dimensional uncertainty quantification , 2018, J. Comput. Phys..

[19]  Ken Perlin,et al.  Accelerating Eulerian Fluid Simulation With Convolutional Networks , 2016, ICML.

[20]  YangCheng,et al.  Data-driven projection method in fluid simulation , 2016 .

[21]  Jiajun Wu,et al.  Learning Particle Dynamics for Manipulating Rigid Bodies, Deformable Objects, and Fluids , 2018, ICLR.

[22]  George Em Karniadakis,et al.  NSFnets (Navier-Stokes flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations , 2020, J. Comput. Phys..

[23]  Arnulf Jentzen,et al.  A proof that artificial neural networks overcome the curse of dimensionality in the numerical approximation of Black-Scholes partial differential equations , 2018, Memoirs of the American Mathematical Society.

[24]  P. Murdin MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY , 2005 .

[25]  Justin A. Sirignano,et al.  DGM: A deep learning algorithm for solving partial differential equations , 2017, J. Comput. Phys..

[26]  Ramin Bostanabad,et al.  Train Once and Use Forever: Solving Boundary Value Problems in Unseen Domains with Pre-trained Deep Learning Models , 2021, ArXiv.

[27]  Markus Schöberl,et al.  Predictive Collective Variable Discovery with Deep Bayesian Models , 2018, The Journal of chemical physics.

[28]  Paris Perdikaris,et al.  Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , 2019, J. Comput. Phys..

[29]  Stefan Riedelbauch,et al.  Direct Prediction of Steady-State Flow Fields in Meshed Domain with Graph Networks , 2021, ArXiv.

[30]  Michael Chertkov,et al.  Embedding Hard Physical Constraints in Neural Network Coarse-Graining of 3D Turbulence , 2020, 2002.00021.

[31]  Nicholas Geneva,et al.  Quantifying model form uncertainty in Reynolds-averaged turbulence models with Bayesian deep neural networks , 2018, J. Comput. Phys..

[32]  Heinrich Müller,et al.  SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernels , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[33]  F. Harlow,et al.  Numerical Calculation of Time‐Dependent Viscous Incompressible Flow of Fluid with Free Surface , 1965 .

[34]  Connor Schenck,et al.  SPNets: Differentiable Fluid Dynamics for Deep Neural Networks , 2018, CoRL.

[35]  Mark Ainsworth,et al.  Galerkin Neural Networks: A Framework for Approximating Variational Equations with Error Control , 2021, SIAM J. Sci. Comput..

[36]  Reinhard Klein,et al.  Learning Incompressible Fluid Dynamics from Scratch - Towards Fast, Differentiable Fluid Models that Generalize , 2020, ICLR.

[37]  Dimitris N. Metaxas,et al.  Realistic Animation of Liquids , 1996, Graphics Interface.

[38]  Vladlen Koltun,et al.  Lagrangian Fluid Simulation with Continuous Convolutions , 2020, ICLR.

[39]  Markus H. Gross,et al.  Deep Fluids: A Generative Network for Parameterized Fluid Simulations , 2018, Comput. Graph. Forum.

[40]  Lexing Ying,et al.  Solving for high-dimensional committor functions using artificial neural networks , 2018, Research in the Mathematical Sciences.

[41]  Nils Thuerey,et al.  Solver-in-the-Loop: Learning from Differentiable Physics to Interact with Iterative PDE-Solvers , 2020, NeurIPS.