Validation
How far can you trust Propforge's performance numbers? We run the model on propellers measured in the wind tunnel at the University of Illinois and publish the result — including where it fails.
Mean absolute error in the normal operating range (measured thrust at least 30% of static thrust). Target: within 15.0 %.
Method
- Blade-element momentum theory (Drela's QPROP formulation) with Prandtl tip and hub losses, solved at 40 blade stations.
- Airfoil lift and drag from a table of XFOIL-class polars (NeuralFoil) over camber, thickness, Reynolds number and angle of attack; flat-plate model beyond stall; Prandtl–Glauert compressibility.
- Blade geometry (chord and blade angle) exactly as measured by UIUC. The airfoil sections are not published, so each propeller gets one of two section classes (generic RC or thin slow-flyer), chosen from the product type — nothing is tuned per propeller.
- The only global setting, the transition parameter n_crit = 7, was chosen on half of the propellers; the other half served as a check.
- This page is generated by the test suite. If a change makes the model worse than the thresholds, the build fails.
Measured vs predicted
Show data table
| J | Measured (UIUC) | Propforge |
|---|---|---|
| 0.49 | 0.064 | 0.065 |
| 0.51 | 0.058 | 0.060 |
| 0.55 | 0.050 | 0.054 |
| 0.57 | 0.045 | 0.050 |
| 0.60 | 0.039 | 0.044 |
| 0.63 | 0.034 | 0.038 |
| 0.66 | 0.029 | 0.032 |
| 0.69 | 0.024 | 0.026 |
| 0.73 | 0.018 | 0.018 |
| 0.75 | 0.013 | 0.012 |
| 0.79 | 0.007 | 0.004 |
| 0.81 | 0.003 | -0.002 |
| 0.84 | -0.003 | -0.010 |
Show data table
| J | Measured (UIUC) | Propforge |
|---|---|---|
| 0.49 | 0.046 | 0.047 |
| 0.51 | 0.044 | 0.045 |
| 0.55 | 0.040 | 0.042 |
| 0.57 | 0.037 | 0.040 |
| 0.60 | 0.034 | 0.037 |
| 0.63 | 0.031 | 0.034 |
| 0.66 | 0.029 | 0.030 |
| 0.69 | 0.026 | 0.027 |
| 0.73 | 0.022 | 0.022 |
| 0.75 | 0.019 | 0.018 |
| 0.79 | 0.015 | 0.013 |
| 0.81 | 0.012 | 0.008 |
| 0.84 | 0.009 | 0.003 |
Show data table
| RPM | Measured (UIUC) | Propforge |
|---|---|---|
| 1975 | 0.098 | 0.094 |
| 2292 | 0.100 | 0.098 |
| 2632 | 0.102 | 0.101 |
| 2908 | 0.102 | 0.103 |
| 3211 | 0.103 | 0.105 |
| 3504 | 0.104 | 0.105 |
| 3815 | 0.104 | 0.106 |
| 4144 | 0.105 | 0.107 |
| 4409 | 0.105 | 0.108 |
| 4711 | 0.106 | 0.108 |
| 5011 | 0.107 | 0.108 |
| 5325 | 0.108 | 0.109 |
| 5624 | 0.112 | 0.109 |
| 5938 | 0.113 | 0.109 |
| 6231 | 0.113 | 0.110 |
| 6542 | 0.114 | 0.110 |
Show data table
| RPM | Measured (UIUC) | Propforge |
|---|---|---|
| 1975 | 0.055 | 0.051 |
| 2292 | 0.054 | 0.049 |
| 2632 | 0.054 | 0.049 |
| 2908 | 0.054 | 0.047 |
| 3211 | 0.054 | 0.046 |
| 3504 | 0.054 | 0.046 |
| 3815 | 0.054 | 0.045 |
| 4144 | 0.054 | 0.045 |
| 4409 | 0.053 | 0.044 |
| 4711 | 0.053 | 0.044 |
| 5011 | 0.053 | 0.044 |
| 5325 | 0.053 | 0.044 |
| 5624 | 0.051 | 0.044 |
| 5938 | 0.051 | 0.043 |
| 6231 | 0.051 | 0.043 |
| 6542 | 0.051 | 0.043 |
Every propeller
| Propeller | Assumed section | CT error (bias) | CP error (bias) | Points |
|---|---|---|---|---|
| APC Thin Electric 10×7 | Generic RC | 3.9 % (+1 %) | 4.1 % (−1 %) | 109 |
| APC Thin Electric 11×5.5 | Generic RC | 10.3 % (+9 %) | 8.3 % (+3 %) | 69 |
| APC Thin Electric 14×12 | Generic RC | 7.9 % (−4 %) | 10.5 % (−9 %) | 81 |
| APC Slow Flyer 10×4.7 | Slow flyer | 16.1 % (−16 %) | 9.9 % (−10 %) | 87 |
| APC Slow Flyer 11×7 | Slow flyer | 3.9 % (−3 %) | 6.7 % (−4 %) | 84 |
| APC Slow Flyer 9×6 | Slow flyer | 16.5 % (−17 %) | 15.5 % (−15 %) | 109 |
| APC Sport 11×6 | Generic RC | 4.3 % (−2 %) | 10.3 % (−10 %) | 90 |
| Graupner CAM Prop 10×6 | Generic RC | 11.2 % (+11 %) | 7.6 % (+3 %) | 95 |
| Graupner Super Nylon 11×8 | Generic RC | 16.7 % (+17 %) | 16.7 % (+14 %) | 91 |
| GWS Slow Flyer 10×4.7 | Slow flyer | 5.1 % (+1 %) | 6.5 % (+2 %) | 46 |
| Kyosho 10×6 | Generic RC | 5.6 % (+5 %) | 6.0 % (−3 %) | 107 |
| Master Airscrew 11×7 | Generic RC | 3.7 % (+0 %) | 6.1 % (−6 %) | 82 |
| Master Airscrew Scimitar 10×5 | Generic RC | 16.6 % (+17 %) | 9.6 % (−10 %) | 93 |
Mean absolute error in the normal operating range; bias in brackets (positive = Propforge predicts too high).
Known limits
- Static power is under-predicted (bias −16 %, 90th percentile 28.5 %); the root flow of a static propeller is highly three-dimensional. Static power bands are widened accordingly.
- Near zero thrust (high advance ratio) relative errors are meaningless; the absolute CT error there is up to 0.025 and CP is off by 32.1 % on average. The app switches to absolute bands in this region.
- At Reynolds numbers below 40,000 errors grow (CT 6.4 %, CP 11.7 %); small, slow propellers are the hardest case.
- The unknown airfoil sections of commercial propellers limit this comparison: for your own designs the section is known exactly, but printed surfaces are rougher than moulded ones.
- Not modelled: wake contraction, blade rake/skew aerodynamics, 3-D stall delay, blade deflection, ducts, cavitation, and anything CFD would capture.
Impellers and pumps
The pump-curve model has not been compared with measured pump curves yet, so its bands are deliberately wide (±25% head, ±30% power). Treat impeller numbers as a first guide, and measure your pump.
Data
Wind-tunnel data for 13 propellers (APC, Graupner, Kyosho, Master Airscrew, GWS) from Volume 1 of the UIUC Propeller Data Site, used with attribution.
Brandt, J.B., Deters, R.W., Ananda, G.K., Dantsker, O.D., Selig, M.S., UIUC Propeller Database, Vols 1–4, University of Illinois at Urbana-Champaign. m-selig.ae.illinois.edu/props/propDB.html. Airfoil polars: NeuralFoil (P. Sharpe, MIT licence), an XFOIL surrogate.