Solver-in-the-Loop: Learning from Differentiable Physics to Interact with Iterative PDE-Solvers

Abstract

Finding accurate solutions to partial differential equations (PDEs) is a crucial task in all scientific and engineering disciplines. It has recently been shown that machine learning methods can improve the solution accuracy by correcting for effects not captured by the discretized PDE. We target the problem of reducing numerical errors of iterative PDE solvers and compare different learning approaches for finding complex correction functions. We find that previously used learning approaches are significantly outperformed by methods that integrate the solver into the training loop and thereby allow the model to interact with the PDE during training. This provides the model with realistic input distributions that take previous corrections into account, yielding improvements in accuracy with stable rollouts of several hundred recurrent evaluation steps and surpassing even tailored supervised variants. We highlight the performance of the differentiable physics networks for a wide variety of PDEs, from non-linear advection-diffusion systems to three-dimensional Navier-Stokes flows.

Published in NeurIPS 2020; Source-code and further details at: https://github.com/tum-pbs/Solver-in-the-Loop

References in corpus (22)

• Sequence to Sequence Learning with Neural Networks
• Discovering governing equations from data: Sparse identification of nonlinear dynamical systems
• DGM: A deep learning algorithm for solving partial differential equations
• Hidden Physics Models: Machine Learning of Nonlinear Partial Differential Equations
• Super-resolution reconstruction of turbulent flows with machine learning
• Deep Learning Methods for Reynolds-Averaged Navier-Stokes Simulations of Airfoil Flows
• Learning data driven discretizations for partial differential equations
• Hamiltonian Neural Networks
• Learning to Simulate Complex Physics with Graph Networks
• Deep Neural Networks for Data-Driven Turbulence Models
• Accelerating Eulerian Fluid Simulation With Convolutional Networks
• Learning Particle Dynamics for Manipulating Rigid Bodies, Deformable Objects, and Fluids
• Deep Dynamical Modeling and Control of Unsteady Fluid Flows
• Hidden Fluid Mechanics: A Navier-Stokes Informed Deep Learning Framework for Assimilating Flow Visualization Data
• Lagrangian Neural Networks
• Compressed Convolutional LSTM: An Efficient Deep Learning framework to Model High Fidelity 3D Turbulence
• Learning Neural PDE Solvers with Convergence Guarantees
• Learning Compositional Koopman Operators for Model-Based Control
• Neural Networks Trained to Solve Differential Equations Learn General Representations
• Learning to Control PDEs with Differentiable Physics
• Differentiable Molecular Simulations for Control and Learning
• AdvectiveNet: An Eulerian-Lagrangian Fluidic reservoir for Point Cloud Processing

Cited by in corpus (31)

• Machine learning accelerated computational fluid dynamics
• A Physics-informed Diffusion Model for High-fidelity Flow Field Reconstruction
• Physics-Informed Neural Operator for Learning Partial Differential Equations
• Learned discretizations for passive scalar advection in a 2-D turbulent flow
• Physics Informed Neural Fields for Smoke Reconstruction with Sparse Data
• A posteriori learning for quasi-geostrophic turbulence parametrization
• Teaching the Incompressible Navier-Stokes Equations to Fast Neural Surrogate Models in 3D
• Numerical investigation of minimum drag profiles in laminar flow using deep learning surrogates
• Known Operator Learning and Hybrid Machine Learning in Medical Imaging -- A Review of the Past, the Present, and the Future
• XLB: A differentiable massively parallel lattice Boltzmann library in Python
• Differentiable Programming for Earth System Modeling
• Lettuce: PyTorch-based Lattice Boltzmann Framework
• A Multifidelity deep operator network approach to closure for multiscale systems
• Physics-Aware Neural Implicit Solvers for multiscale, parametric PDEs with applications in heterogeneous media
• MPMNet: A Data-Driven MPM Framework for Dynamic Fluid-Solid Interaction
• Physics-Integrated Variational Autoencoders for Robust and Interpretable Generative Modeling
• Variational Data Assimilation with a Learned Inverse Observation Operator
• Deep Neural Networks to Correct Sub-Precision Errors in CFD
• Learning Incompressible Fluid Dynamics from Scratch -- Towards Fast, Differentiable Fluid Models that Generalize
• Neural-Parareal: Dynamically Training Neural Operators as Coarse Solvers for Time-Parallelisation of Fusion MHD Simulations
• Using neural networks to solve the 2D Poisson equation for electric field computation in plasma fluid simulations
• Differentiable Cosmological Hydrodynamics for Field-Level Inference and High Dimensional Parameter Constraints
• Physics-Guided Discovery of Highly Nonlinear Parametric Partial Differential Equations
• Learning to Assimilate in Chaotic Dynamical Systems
• Physics-based Deep Learning
• A TVD neural network closure and application to turbulent combustion
• Accelerated Gradient-based Design Optimization Via Differentiable Physics-Informed Neural Operator: A Composites Autoclave Processing Case Study
• PICT -- A Differentiable, GPU-Accelerated Multi-Block PISO Solver for Simulation-Coupled Learning Tasks in Fluid Dynamics
• A Differentiable Solver Approach to Operator Inference
• Learning invariance preserving moment closure model for Boltzmann-BGK equation
• Physics-Aware Downsampling with Deep Learning for Scalable Flood Modeling