Metraipolis › archive › dc_d0de82dcc52e

Diff-NekRS: A Scalable Differentiable Framework for Multi-Timestep Solver-in-the-Loop Training — arXiv:2609.23208

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.23208
License (as archived)
CC0-1.0
Archived
2026-10-09T23:41:13.843Z
SHA-256 of the archived text
0831d6ca94e1b909a63b6aa85253c1711369d7e3c750b46beb81fb98eee9260b
Stable URL of this page
https://metraipolis.com/documento/dc_d0de82dcc52e

How to cite

Diff-NekRS: A Scalable Differentiable Framework for Multi-Timestep Solver-in-the-Loop Training — arXiv:2609.23208 — arXiv (en), https://arxiv.org/abs/2609.23208 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d0de82dcc52e (Metraipolis archive dc_d0de82dcc52e, sha256 0831d6ca94e1b909a63b6aa85253c1711369d7e3c750b46beb81fb98eee9260b)
BibTeX
@misc{dcd0de82dcc52e,
  title = {Diff-NekRS: A Scalable Differentiable Framework for Multi-Timestep Solver-in-the-Loop Training — arXiv:2609.23208},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.23208},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d0de82dcc52e; sha256 0831d6ca94e1b909a63b6aa85253c1711369d7e3c750b46beb81fb98eee9260b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.23208. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Hybrid physics-machine-learning solvers improve under-resolved simulations by embedding trainable corrections into the time integration. During autoregressive inference, repeated solver-model interactions can amplify small errors, motivating multi-timestep solver-in-the-loop training. However, production solvers rarely expose the derivatives needed to backpropagate through such rollouts. We introduce Diff-NekRS, a scalable differentiable framework that embeds neural corrections directly in the GPU-accelerated NekRS incompressible-flow solver. NekRS computes the authoritative forward trajectory, a manually implemented exact discrete adjoint differentiates the supported fully discrete timestep, and LibTorch supplies neural vector-Jacobian products and parameter gradients. End-to-end Taylor and centered finite-difference tests verify the assembled gradient for two-dimensional cylinder flow (2Dcyl) and the three-dimensional Taylor-Green vortex (3DTGV) across five horizons and 12-1,020 MPI ranks. At 1,020 ranks, optimizer-enabled post-setup training updates retain 54.5%-78.0% and 80.7%-81.9% weak-scaling efficiency for 2Dcyl and 3DTGV, respectively. In 200-step autoregressive inference, the M = 50 model reduces the three-seed median terminal relative L2 velocity error by 59.2% for 2Dcyl and 12.1% for 3DTGV relative to the uncorrected coarse-grid P = 2 baseline, and retains wall-clock speedups of 5.38x and 2.49x, respectively, relative to the corresponding P = 7 configurations for equal simulated-time intervals. These results establish a verified and scalable path for multi-timestep solver-in-the-loop training that improves coarse-grid trajectory accuracy while retaining a speed advantage over the high-order reference

Tema degli esploratori: fisica e astronomia (retro-catalogo arXiv: annunci del 2026-09-22 in physics.comp-ph).
Categorie arXiv: Computational Physics (physics.comp-ph) ; Fluid Dynamics (physics.flu-dyn).
Pagina del lavoro: https://arxiv.org/abs/2609.23208 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-09T22:46:47.356Z dalla pagina https://arxiv.org/catchup/physics.comp-ph/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).