skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: A fast and accurate physics-informed neural network reduced order model with shallow masked autoencoder

Journal Article · · Journal of Computational Physics
 [1];  [2];  [2];  [1]
  1. Univ. of California, Berkeley, CA (United States)
  2. Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
+ Show Author Affiliations

Traditional linear subspace reduced order models (LS-ROMs) are able to accelerate physical simulations in which the intrinsic solution space falls into a subspace with a small dimension, i.e., the solution space has a small Kolmogorov n-width. However, for physical phenomena not of this type, e.g., any advection-dominated flow phenomena such as in traffic flow, atmospheric flows, and air flow over vehicles, a low-dimensional linear subspace poorly approximates the solution. To address cases such as these, we have developed a fast and accurate physics-informed neural network ROM, namely nonlinear manifold ROM (NM-ROM), which can better approximate high-fidelity model solutions with a smaller latent space dimension than the LS-ROMs. Our method takes advantage of the existing numerical methods that are used to solve the corresponding full order models. The efficiency is achieved by developing a hyper-reduction technique in the context of the NM-ROM. Numerical results show that neural networks can learn a more efficient latent space representation on advection-dominated data from 1D and 2D Burgers' equations. A speedup of up to 2.6 for 1D Burgers' and a speedup of 11.7 for 2D Burgers' equations are achieved with an appropriate treatment of the nonlinear terms through a hyper-reduction technique. Lastly, a posteriori error bounds for the NM-ROMs are derived that take account of the hyper-reduced operators.

Research Organization:
Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
Sponsoring Organization:
USDOE National Nuclear Security Administration (NNSA)
Grant/Contract Number:
AC52-07NA27344; LLNL-JRNL-814844; 20-FS-007
OSTI ID:
1843130
Alternate ID(s):
OSTI ID: 1868572
Report Number(s):
LLNL-JRNL-814844; 1023653; TRN: US2302597
Journal Information:
Journal of Computational Physics, Vol. 451; ISSN 0021-9991
Publisher:
ElsevierCopyright Statement
Country of Publication:
United States
Language:
English

References (45)

Robust model reduction by $$L^{1}$$-norm minimization and approximation via dictionaries: application to nonlinear hyperbolic problems journal January 2016
Physics-constrained deep learning for high-dimensional surrogate modeling and uncertainty quantification without labeled data journal October 2019
Nonlinear principal component analysis using autoassociative neural networks journal February 1991
Efficient estimation of cardiac conductivities via POD-DEIM model order reduction journal May 2017
Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations journal February 2019
POD-DEIM reduction of computational EMG models journal March 2017
Approximation by superpositions of a sigmoidal function journal December 1989
Solving high-dimensional partial differential equations using deep learning journal August 2018
Non-linear model reduction for the Navier–Stokes equations using residual DEIM method journal April 2014
Design optimization using hyper-reduced-order models journal November 2014
DGM: A deep learning algorithm for solving partial differential equations journal December 2018
Adaptive h -refinement for reduced-order models : ADAPTIVE journal November 2014
Reduced order modeling based shape optimization of surface acoustic wave driven microfluidic biochips journal June 2012
Transformed Snapshot Interpolation with High Resolution Transforms journal January 2020
Approximation theory of the MLP model in neural networks journal January 1999
POD/DEIM nonlinear model order reduction of an ADI implicit shallow water equations model journal March 2013
A Survey of Projection-Based Model Reduction Methods for Parametric Dynamical Systems journal January 2015
Galerkin Proper Orthogonal Decomposition Methods for a General Equation in Fluid Dynamics journal January 2002
fPINNs: Fractional Physics-Informed Neural Networks journal January 2019
Quantifying total uncertainty in physics-informed neural networks for solving forward and inverse stochastic problems journal November 2019
POD/DEIM Reduced-Order Modeling of Time-Fractional Partial Differential Equations with Applications in Parameter Identification journal April 2017
Galerkin v. least-squares Petrov–Galerkin projection in nonlinear model reduction journal February 2017
Transport Reversal for Model Reduction of Hyperbolic Partial Differential Equations journal January 2018
Nonlinear Model Reduction via Discrete Empirical Interpolation journal January 2010
The numerical solution of linear ordinary differential equations by feedforward neural networks journal June 1994
The GNAT method for nonlinear model reduction: Effective implementation and application to computational fluid dynamics and turbulent flows journal June 2013
Learning representations by back-propagating errors journal October 1986
SNS: A Solution-Based Nonlinear Subspace Method for Time-Dependent Model Order Reduction journal January 2020
Machine Learning Approximation Algorithms for High-Dimensional Fully Nonlinear Partial Differential Equations and Second-order Backward Stochastic Differential Equations journal January 2019
Implementation and detailed assessment of a GNAT reduced-order model for subsurface flow simulation journal February 2019
Artificial neural networks for solving ordinary and partial differential equations journal January 1998
Neural-network-based approximations for solving partial differential equations journal March 1994
Fast Multiscale Reservoir Simulations With POD-DEIM Model Reduction journal June 2016
Efficient non-linear model reduction via a least-squares Petrov-Galerkin projection and compressive tensor approximations
  • Carlberg, Kevin; Bou-Mosleh, Charbel; Farhat, Charbel
  • International Journal for Numerical Methods in Engineering, Vol. 86, Issue 2 https://doi.org/10.1002/nme.3050
journal October 2010
The Proper Orthogonal Decomposition in the Analysis of Turbulent Flows journal January 1993
Reconstructing phase space from PDE simulations journal November 1992
The Discrete Empirical Interpolation Method: Canonical Structure and Formulation in Weighted Inner Product Spaces journal January 2018
POD-DEIM Based Model Order Reduction for the Spherical Shallow Water Equations with Turkel-Zwas Finite Difference Discretization journal January 2014
Localized Model Reduction in Porous Media Flow∗1The authors would like to thanks the US DoD - Army Research Once and Foundation CMG for supporting this project. journal January 2015
Conservative model reduction for finite-volume models journal October 2018
A unified deep artificial neural network approach to partial differential equations in complex geometries journal November 2018
Analysis of a complex of statistical variables into principal components. journal January 1933
Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders journal November 2019
The Deep Ritz Method: A Deep Learning-Based Numerical Algorithm for Solving Variational Problems journal February 2018
Gradient-based constrained optimization using a database of linear reduced-order models journal December 2020

Similar Records

Nonlinear manifold-based reduced order model
Technical Report · Tue Sep 29 00:00:00 EDT 2020 · OSTI ID:1843130
Reduced-order modeling of advection-dominated systems with recurrent neural networks and convolutional autoencoders
Journal Article · Fri Mar 05 00:00:00 EST 2021 · Physics of Fluids · OSTI ID:1843130
Latent Space Dynamics Identification
Software · Mon Feb 07 00:00:00 EST 2022 · OSTI ID:1843130

Related Subjects

97 MATHEMATICS AND COMPUTING
Nonlinear manifold solution representation
Physics-informed neural network
Reduced order model
Nonlinear dynamical system
Hyper-reduction