openfoam-cfd-copilot

OpenFOAM CFD Co-Pilot (by iLab)

This skill turns a plain-language CFD goal into a correct OpenFOAM case and helps diagnose it — doing the tedious, error-prone bookkeeping (y+, patch matching, turbulence fields, residual reading) so the user focuses on the physics.

If the openfoam-mcp MCP server is connected, prefer its tools (they read and write real case files). If it is not, apply the reference formulas and templates in this document by hand.

Workflow

Follow these steps in order. Ask the user only for what you still need (characteristic velocity, characteristic length, fluid, steady vs transient, heat transfer yes/no, and which OpenFOAM version they run).

1. Classify the problem and pick the setup.

   • Internal (pipe/duct/channel), external (plate/airfoil/bluff body), backward-facing step, or lid-driven cavity.
   • Compute the Reynolds number and decide laminar vs turbulent.
   • Choose the solver and turbulence model (rules below).
   • MCP: recommend_setup(goal, velocity, length, fluid, transient, heat_transfer).

2. Size the near-wall mesh.

   • Pick a wall-treatment strategy: wall-resolved (y+ ≈ 1) or wall functions (y+ 30–300). Avoid the buffer layer (5 < y+ < 30).
   • Compute the first-cell height for the target y+.
   • MCP: first_cell_height(y_plus, velocity, length) / y_plus_from_height(...).

3. Generate the mesh.

   • Build a graded blockMeshDict from a geometry preset; grade the wall-normal direction to hit the first-cell height from step 2.
   • MCP: generate_blockmesh(output_path, preset, first_cell_height=…).
   • For an imported geometry, check the STL is watertight before snappyHexMesh. MCP: analyze_stl(path).

4. Scaffold the case and set inlet turbulence.

   • Write system/, constant/, 0/.
   • Compute inlet k, ε, ω, νt from turbulence intensity (~5% internal) and a length scale (~0.07 × hydraulic diameter).
   • MCP: scaffold_case(case_dir, solver, velocity, openfoam_flavor) + inlet_turbulence(velocity, intensity, length_scale).