AeroRotor-eVTOL Analysis Suite

A modular, multi-fidelity Blade Element Momentum Theory (BEMT) propeller analysis and verification framework for small-scale fixed-pitch propellers in the low-Reynolds regime ($30{,}000 \le Re \le 150{,}000$). It produces sub-second aerodynamic performance predictions and validates them end-to-end against UIUC wind-tunnel experiments and a full OpenFOAM 3D RANS CFD simulation.

Baseline target: APC 10x4.7 propeller · 2-blade · variable-RPM / fixed-pitch electric drive.

Status: implemented and validated. 52 tests passing, ruff clean. Portfolio project — analysis/design tool only, not flight-critical or certified software.

---

What it does

• Aerodynamic engine (Subsystem 1.0): BEMT solver with Standard (thin-airfoil) and Enhanced (XFOIL tabular) polar modes, Prandtl tip/hub losses, and an optional Du-Selig 3D rotational stall-delay correction — built on an exact velocity-form momentum balance that stays valid at high induction and is non-singular in hover (J=0).
• Trim solver (Subsystem 2.0): Newton–Raphson vertical force balance for hover and steady axial climb, including airframe parasitic drag (equivalent flat-plate area $f$).
• V&V pipeline (Subsystem 3.0): percentage-error scoring against the UIUC and OpenFOAM references, MET-01…06 acceptance gates, and a generated VALIDATION.md report + overlay plots.

The blade geometry and wind-tunnel data are real (UIUC Propeller Database); the E63 section polars are XFOIL-generated; the spanwise-loading reference is a genuine OpenFOAM MRF RANS run. See data/README.md for full provenance.

Validation results

Metric	Checks	Result	Gate
MET-01	Thrust coefficient at hover (J=0)	6.6% error	±5%
MET-02	Power coefficient over 0 ≤ J ≤ 0.5	42% error	±7%
MET-03	Peak-efficiency advance ratio	ΔJ 0.075	±0.03
MET-04	Spanwise dT/dr vs. CFD (Pearson r)	0.96 — PASS	≥0.95
MET-05	Trim force residual	<1e-4 N — PASS	<1e-4 N
MET-06	50-point sweep compute time	~0.1 s — PASS	<1 s

The CFD reference was independently anchored (total thrust within ~7% of the wind-tunnel value) before use. The model is validated where it is valid — hover thrust within ~7%, spanwise loading shape matching CFD at r = 0.96 — and quantitatively bounded where it is not: axial-climb power is under-predicted ~40%, traced (not guessed) to the fundamental limit of a 2D-polar BEMT method at these Reynolds numbers. Every inexpensive correction (Re-range, transition, tip-loss, Du-Selig stall-delay) was tested and ruled out. See log.md for the full investigation.

Getting started

Requires Python 3.11+. XFOIL and OpenFOAM (for regenerating the polars / CFD reference) are external and live in WSL — not needed to run the analysis with the committed data.

python -m venv .venv && source .venv/Scripts/activate   # Windows; use bin/activate on Linux/macOS
pip install -e ".[dev]"                                  # numpy, scipy, pyyaml, matplotlib, pytest, ruff

python main.py -c config/apc_10x47_config.yaml           # full run -> artifacts/VALIDATION.md
pytest -q                                                # 52 tests

Data generation and meshing are reproducible via scripts/ (UIUC CSVs, XFOIL polars, blade STL, CFD post-processing) and cfd/Allrun_{mesh,solve,post}.sh.

Documentation

Doc	Role
system_requirements.md	SRD — constraints, functional requirements, KPI gates
system_architecture.md	SAD — subsystem decomposition, ICD data contracts, control flow
conops.md	ConOps — operational phases, user modes, execution lifecycle
PROJECT_MAP.md	File & module directory map
data/README.md	Dataset provenance and generation
log.md	Change & decision log (D1…D21)
CLAUDE.md	Working agreements & coding conventions

Method & stack

Velocity-form BEMT (hover-stable), Prandtl tip/hub losses, Du-Selig 3D stall-delay; XFOIL-generated low-Reynolds E63 polars; OpenFOAM MRF RANS reference (k-ω SST, ~775k cells). Python (NumPy / SciPy), test-driven (52 tests), reproducible from committed scripts.

License

MIT