RC Trainer Generator

A macOS-compatible Python tool that builds a flyable JSBSim/FlightGear aircraft from either a user-supplied aerodynamic polar (USER_TABLE) or an automated OpenFOAM sweep (AUTO_CFD). Control surfaces are analytical; no deflected STL files are needed.

Quick start

python3 -m venv .venv
source .venv/bin/activate
pip install -e '.[test]'
rc-trainer generate examples/user_table.yaml

Outputs are written below output/: a dense polar, propulsion speed sweep, plots, engineering report, JSBSim XML, and a FlightGear aircraft directory with a macOS launcher.

To inspect configuration without generating files:

rc-trainer validate examples/user_table.yaml

USER_TABLE requires no Docker. AUTO_CFD uses opencfd/openfoam-default, probes its OpenFOAM environment, creates a reusable case, runs the AoA sweep by changing freestream direction (not geometry), and accepts only converged points unless configured otherwise.

With STL geometry in place, run the complete workflow with:

rc-trainer generate examples/auto_cfd.yaml --output output/my-cfd-aircraft --run-cfd

Use --resume-cfd to reuse completed AoA rows in an existing cfd/aoa_results.csv; new rows are written as each AoA finishes. Incomplete or failed AoA points are rerun. Set cfd.include_questionable: true to use non-converged points while retaining their status in the CSV.

The supplied example uses eight MPI processes and a locally refined nominal mesh of approximately 185k cells. Each point writes an individual output/cfd/log.alpha_*.txt log and the mesh writes output/cfd/case/log.mesh.txt plus mesh_stats.json. Select preview, coarse, nominal, or fine with cfd.mesh_level.

See docs/CONFIGURATION.md for inputs and physical assumptions.

The complete mode examples are USER_TABLE and AUTO_CFD. The AUTO_CFD example uses the supplied component STL files and derives wing reference quantities automatically.