Skip to content

cfd-openfoam.md

Latest commit

 

History

History
86 lines (59 loc) · 3.99 KB

cfd-openfoam.md

File metadata and controls

86 lines (59 loc) · 3.99 KB
Raw

OpenFOAM Case Generation and Bridge

Generate complete, runnable OpenFOAM case directories from a Kerf geometry description — blockMesh, boundary conditions, turbulence model, and solver settings in one call.

Module: packages/kerf-cfd/src/kerf_cfd/openfoam_bridge.py Shipped: Wave 9 LLM tools: cfd_run (analysis_type "cfd_turbulent", "cfd_multiphase")

What it is

OpenFOAM is the industry-standard open-source CFD solver used across automotive, aerospace, marine, and HVAC engineering. Setting up an OpenFOAM case correctly — with consistent boundary conditions, turbulence model initial fields, numerical scheme selections, and solver relaxation factors — requires deep OpenFOAM knowledge and takes hours by hand.

This module generates complete, runnable OpenFOAM case directories from a structured Python description: 0/ initial field files, constant/ transport properties and turbulence model coefficients, and system/ blockMeshDict, controlDict, fvSchemes, and fvSolution files. It supports simpleFoam (steady incompressible RANS), buoyantSimpleFoam (thermal RANS), and interFoam (VoF free-surface). Results can be parsed back by read_results.

How to use it

From chat (natural language)

"Set up a k-ω SST OpenFOAM case for flow over a bluff body at Re=50,000, inlet velocity 10 m/s"

The LLM calls cfd_run with analysis_type='cfd_turbulent'.

From Python

from kerf_cfd.openfoam_bridge import (
    OpenFOAMCaseSpec, export_to_openfoam, read_results,
)

spec = OpenFOAMCaseSpec(
    solver="simpleFoam",
    turbulence_model="kOmegaSST",
    nu=1.5e-5,
    inlet_velocity=(10.0, 0.0, 0.0),
    mesh_vertices=[...],  # blockMeshDict vertex list
    mesh_blocks=[...],
)
result = export_to_openfoam(spec, output_dir="/tmp/foam_case")
# Returns OpenFOAMExportResult with case_dir path and mesh quality warnings

From an LLM tool spec

{"tool": "cfd_run", "analysis_type": "cfd_turbulent",
 "solver": "simpleFoam", "turbulence_model": "kOmegaSST",
 "Re": 50000, "inlet_U": [10, 0, 0]}

How it works

The bridge writes OpenFOAM's native dictionary format using Python string templates. fvSchemes is set to second-order schemes (linearUpwind for convection, Gauss linear for diffusion) for production runs. fvSolution uses GAMG for pressure and smoothSolver for velocity. Turbulence initial fields (k, ε or ω) are estimated from turbulence intensity and mixing length: k = 1.5(U·I)², ε = Cμ^0.75 k^1.5/L.

read_results parses the OpenFOAM field files and returns pressure, velocity, and turbulence fields as NumPy arrays.

API reference

Function Returns Purpose
export_to_openfoam(spec, output_dir) OpenFOAMExportResult Write case directory
read_results(case_dir, time_step) ResultBundle Parse field files
write_polymesh(mesh, path) None Write polyMesh directory

OpenFOAMExportResult fields: case_dir, mesh_quality_warnings, estimated_cell_count.

Example

spec = OpenFOAMCaseSpec(solver="simpleFoam", turbulence_model="kEpsilon",
                         nu=1e-5, inlet_velocity=(5,0,0),
                         mesh_vertices=verts, mesh_blocks=blocks)
res = export_to_openfoam(spec, "/tmp/test_case")
print(f"Case written to: {res.case_dir}")
print(f"Mesh warnings: {res.mesh_quality_warnings}")

Honest caveats

The bridge generates valid OpenFOAM cases for common configurations, but mesh quality and boundary condition correctness are the user's responsibility. blockMesh generates structured hexahedral meshes only — complex geometries need snappyHexMesh (cfd-snappy-mesh). OpenFOAM must be installed locally (or on a trusted node offering compute — see docs/node-architecture.md) for execution — the bridge does not run the solver. Y⁺ wall treatment must be set manually based on mesh resolution.

References

  • Patankar (1980). Numerical Heat Transfer and Fluid Flow. McGraw-Hill.
  • OpenFOAM Foundation (2023). OpenFOAM User Guide, v11. openfoam.org.