Paper

{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.15992v1</id>\n    <title>Physics-Constrained Neural Closure for Lattice Boltzmann Large-Eddy Simulation</title>\n    <updated>2026-03-16T23:04:17Z</updated>\n    <link href='https://arxiv.org/abs/2603.15992v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.15992v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>We present a physics-constrained, data-driven subgrid-scale (SGS) stress closure for large-eddy simulation (LES) in the lattice Boltzmann method (LBM). Trained on filtered-downsampled (FD) data from LBM direct numerical simulation (DNS) of forced homogeneous isotropic turbulence (FHIT) spanning multiple filter widths, a compact neural network maps nine macroscopic derivative inputs - six strain-rate and three vorticity components - to the six independent components of the SGS stress tensor; a deviatoric projection is applied post-inference to obtain the traceless stress used in the solver. Training combines a stress data loss with physics terms for SGS energy-transfer (Pi) matching, rotational equivariance under cube rotations, and compatibility of the implied SGS forcing with the divergence-based coupling.\n  The predicted stress is coupled to the solver through a split strategy: a dissipative, strain-aligned contribution is represented through an effective-viscosity projection, while the remaining anisotropic residual is applied through a forcing term. This construction is intended to retain both backscatter (via the effective viscosity) and non-dissipative anisotropic effects (via the residual forcing), while remaining compatible with LBM deployment. In the cases considered here, a priori results show good agreement with FD references across stress components and SGS-transfer statistics, and a posteriori rollouts improve several energetic and statistical measures relative to static and dynamic Smagorinsky baselines. A preliminary transfer test in turbulent channel flow is also reported without retraining. Finally, we demonstrate production deployment via ONNX Runtime, with throughput comparable to a dynamic Smagorinsky baseline in the tested configuration.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='physics.flu-dyn'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='physics.comp-ph'/>\n    <published>2026-03-16T23:04:17Z</published>\n    <arxiv:comment>32 pages, 16 figures</arxiv:comment>\n    <arxiv:primary_category term='physics.flu-dyn'/>\n    <author>\n      <name>Muhammad Idrees Khan</name>\n      <arxiv:affiliation>University of Rome Tor Vergata, Rome, Italy</arxiv:affiliation>\n    </author>\n    <author>\n      <name>Sauro Succi</name>\n      <arxiv:affiliation>Italian Institute of Technology, Rome, Italy</arxiv:affiliation>\n      <arxiv:affiliation>Harvard University, Cambridge, USA</arxiv:affiliation>\n    </author>\n    <author>\n      <name>Hua-Dong Yao</name>\n      <arxiv:affiliation>Chalmers University of Technology, Gothenburg, Sweden</arxiv:affiliation>\n    </author>\n    <author>\n      <name>Giacomo Falcucci</name>\n      <arxiv:affiliation>University of Rome Tor Vergata, Rome, Italy</arxiv:affiliation>\n      <arxiv:affiliation>Harvard University, Cambridge, USA</arxiv:affiliation>\n    </author>\n  </entry>"
}