Nonlinear manifold reduced order model

RESOURCE

Abstract

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 lowdimensional 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 software 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.
Developers:
Choi, Youngsoo [1] Kim, Youngkyu [1] Widemann, David [1]
  1. Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
Release Date:
2024-07-02
Project Type:
Open Source, Publicly Available Repository
Software Type:
Scientific
Version:
1.0
Licenses:
MIT License
Sponsoring Org.:
Code ID:
142366
Site Accession Number:
LLNL-CODE-867904
Research Org.:
Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
Country of Origin:
United States

RESOURCE

Citation Formats

Choi, Youngsoo, Kim, Youngkyu, and Widemann, David P. Nonlinear manifold reduced order model. Computer Software. https://github.com/LLNL/NM-ROM. USDOE National Nuclear Security Administration (NNSA). 02 Jul. 2024. Web. doi:10.11578/dc.20240904.2.
Choi, Youngsoo, Kim, Youngkyu, & Widemann, David P. (2024, July 02). Nonlinear manifold reduced order model. [Computer software]. https://github.com/LLNL/NM-ROM. https://doi.org/10.11578/dc.20240904.2.
Choi, Youngsoo, Kim, Youngkyu, and Widemann, David P. "Nonlinear manifold reduced order model." Computer software. July 02, 2024. https://github.com/LLNL/NM-ROM. https://doi.org/10.11578/dc.20240904.2.
@misc{ doecode_142366,
title = {Nonlinear manifold reduced order model},
author = {Choi, Youngsoo and Kim, Youngkyu and Widemann, David P.},
abstractNote = {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 lowdimensional 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 software 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.},
doi = {10.11578/dc.20240904.2},
url = {https://doi.org/10.11578/dc.20240904.2},
howpublished = {[Computer Software] \url{https://doi.org/10.11578/dc.20240904.2}},
year = {2024},
month = {jul}
}