Repository navigation
This release lets the blade-element model run on real propeller geometry, with airfoil polars computed by NeuralFoil, and validates it against wind-tunnel measurements of 75 APC propellers.
Propeller geometry files
pythrust.io.read_apc_geometryreads APC's*-PERF.PE0geometry files: chord and blade angle per station, with the airfoils blended across APC's transition.pythrust.io.read_uiuc_geometryreads UIUC Propeller Database geometry files.pythrust.io.read_uiuc_performancereads UIUC static and forward-flight measurements into aTablePropeller.
Airfoil polars from NeuralFoil
AirfoilPolar.from_neuralfoiltabulates lift and drag over angle of attack and Reynolds number from a NACA 4- or 5-digit designation, a.datfile or coordinates. Beyond NeuralFoil's range the section is a flat plate.apc_airfoil_polars()gives the airfoils of APC's files: E63, APC12 (NACA 4412) and CLARK-Y.BladeElementPropellertakes these polars as sections. TheBladeSection(QPROP) path is unchanged.n_critdefaults to 7, the value that fits the measurements best.- New dependency:
setuav-neuralfoil, a NumPy-only build of NeuralFoil (MIT).
Validation
A new docs page, Reference → Validation, compares the models with UIUC measurements. Mean absolute error per propeller, on the 49 propellers not used to choose n_crit:
| Model | Static C_T | Static C_P | Forward C_T | Forward C_P |
|---|---|---|---|---|
| Blade element + NeuralFoil | 9.2% | 13.5% | 12.3% | 11.5% |
| APC's own tables | 14.5% | 12.5% | 16.0% | 11.3% |
validation/ holds the scripts to reproduce it.
Full changelog: v0.6.0...v0.7.0