Skip to content

Tune XLB accuracy on 2D cylinder flow #38

Description

@dionhaefner

XLB's lattice Boltzmann scheme has a known O(Ma²) compressibility floor at high resolutions (documented in mosaic/benchmarks/problems/navier_stokes_grid/exclusions.py). On the forward/cylinder experiment, this manifests as elevated error relative to collocated peers.

Goal: Investigate whether adjusting the Mach number, relaxation parameters, or collision operator improves XLB's accuracy on this experiment.

How to get started:

  1. Run the baseline: mosaic run -p ns-grid -e forward/cylinder -s XLB --debug
  2. Snapshot: mosaic status --format json > before.json
  3. Adjust parameters in mosaic/tesseracts/navier-stokes-grid/xlb/tesseract_api.py or via input_overrides in the problem config
  4. Compare: mosaic status --format md --diff-against before.json

Good entry point for someone familiar with lattice Boltzmann methods. See CONTRIBUTING.md — Tuning an existing solver for the full workflow.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions