Thermodynamics-informed latent space dynamics identification

RESOURCE

Abstract

This software showcases a latent space dynamics identification method, namely tLaSDI, that embeds the first and second principles of thermodynamics. The latent variables are learned through an autoencoder as a nonlinear dimension reduction model. The latent dynamics are constructed by a neural network-based model that precisely preserves certain structures for the thermodynamic laws through the GENERIC formalism. An abstract error estimate is established, which provides a new loss formulation involving the Jacobian computation of autoencoder. The autoencoder and the latent dynamics are simultaneously trained to minimize the new loss. Computational examples demonstrate the effectiveness of tLaSDI, which exhibits robust generalization ability, even in extrapolation. In addition, an intriguing correlation is empirically observed between a quantity from tLaSDI in the latent space and the behaviors of the full-state solution.
Developers:
Cheung, Siu Wun [1] Park, Jun Sur [1] Choi, Youngsoo [1] Shin, Yeonjong [2]
  1. Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
  2. North Carolina State University, Raleigh, NC (United States)
Release Date:
2024-07-31
Project Type:
Open Source, Publicly Available Repository
Software Type:
Scientific
Version:
0.1
Licenses:
MIT License
Sponsoring Org.:
Code ID:
142379
Site Accession Number:
LLNL-CODE-867909
Research Org.:
Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
Country of Origin:
United States

RESOURCE

Citation Formats

Cheung, Siu Wun, Park, Jun Sur, Choi, Youngsoo, and Shin, Yeonjong. Thermodynamics-informed latent space dynamics identification. Computer Software. https://github.com/pjss1223/tLaSDI. USDOE National Nuclear Security Administration (NNSA). 31 Jul. 2024. Web. doi:10.11578/dc.20240905.1.
Cheung, Siu Wun, Park, Jun Sur, Choi, Youngsoo, & Shin, Yeonjong. (2024, July 31). Thermodynamics-informed latent space dynamics identification. [Computer software]. https://github.com/pjss1223/tLaSDI. https://doi.org/10.11578/dc.20240905.1.
Cheung, Siu Wun, Park, Jun Sur, Choi, Youngsoo, and Shin, Yeonjong. "Thermodynamics-informed latent space dynamics identification." Computer software. July 31, 2024. https://github.com/pjss1223/tLaSDI. https://doi.org/10.11578/dc.20240905.1.
@misc{ doecode_142379,
title = {Thermodynamics-informed latent space dynamics identification},
author = {Cheung, Siu Wun and Park, Jun Sur and Choi, Youngsoo and Shin, Yeonjong},
abstractNote = {This software showcases a latent space dynamics identification method, namely tLaSDI, that embeds the first and second principles of thermodynamics. The latent variables are learned through an autoencoder as a nonlinear dimension reduction model. The latent dynamics are constructed by a neural network-based model that precisely preserves certain structures for the thermodynamic laws through the GENERIC formalism. An abstract error estimate is established, which provides a new loss formulation involving the Jacobian computation of autoencoder. The autoencoder and the latent dynamics are simultaneously trained to minimize the new loss. Computational examples demonstrate the effectiveness of tLaSDI, which exhibits robust generalization ability, even in extrapolation. In addition, an intriguing correlation is empirically observed between a quantity from tLaSDI in the latent space and the behaviors of the full-state solution.},
doi = {10.11578/dc.20240905.1},
url = {https://doi.org/10.11578/dc.20240905.1},
howpublished = {[Computer Software] \url{https://doi.org/10.11578/dc.20240905.1}},
year = {2024},
month = {jul}
}