CFD Tutorial no. 18 — Experimental CFD validation

A benchmark, and what it reveals

This experimental CFD validation tutorial confronts your calculation chain with a documented reference propeller.

Experimental CFD validation: tutorial for qualifying your chain

CFD Tutorial no. 18 of 20. The actions described here are those of the HELICIEL CFD tool window, opened from Heliciel: they assume the PRO CFD/BEM extension is active. The preset values quoted are those of the catalogue shipped with the software; the numerical results depend on your geometry and your operating point, and should be read in the software.

Comparing one calculation to another calculation tells you nothing about the truth: it tells you that two models agree, or do not. Experimental CFD validation introduces a third judge, external to both: measurement.

The exercise: take a propeller whose performance has been documented for decades, calculate it with your chain, and look at the gap. What you are qualifying then is not the propeller — it is already known — but your own calculation chain: your presets, your hardware, your habits.

Prerequisites: Heliciel version 12.1.2 or later, which embeds the Wageningen B-series;
Duration: several calculations — it is an investment, done once;
What you get out of it: a gap figure you can quote in all your subsequent reports.

Tutorial outline:

Why an external reference
Generating the reference propeller
The coefficients to compare
Mesh convergence first
Interpreting the gap
Turning that gap into an asset

1: Why an external reference

A CFD calculation carries a long chain of assumptions: the mesh discretises, the turbulence model approximates, the numerical schemes diffuse, convergence stops on a threshold. Each one is reasonable, and their sum produces a gap that nobody can predict in absolute terms.

The only way to know this gap is to measure it on a case whose answer is known by other means. This is an operation to do once, seriously, rather than to hope for each time a result is delivered.

Three families of references can be used: wind-tunnel measurements on airfoils, rotor tests in a cavitation tunnel or a towing tank, and systematic series of marine propellers. This last family has a decisive advantage for this tutorial: Heliciel can generate the exact geometry of the most documented series.

diagram of the validation process with the three sources measurement BEM and CFD
The process: the measurement on one side, your two methods on the other, and two gaps to know

2: Generating the reference propeller

Since version 12.1.2, Heliciel embeds the Wageningen B-series, the marine propeller reference for some sixty years. Heliciel's File menu, entry B-Screw Series: classic reference marine propellers.

The complete geometry of the series is available: from three to seven blades, every area ratio, and a pitch-to-diameter ratio covering the usual range. Choose a configuration for which you have the published curves — that is the only criterion that matters here.

Once the geometry is applied to the project, you have in Heliciel a propeller whose performance is tabulated. Create the CFD case as usual: File > New CFD case (from current model).

Wageningen B-series propeller generated in Heliciel and displayed in the CFD bench
The reference propeller, generated by Heliciel and loaded into the CFD bench: geometry beyond dispute

3: The coefficients to compare

A marine propeller's performance is traditionally presented as dimensionless coefficients, plotted against the advance coefficient J: a thrust coefficient, a torque coefficient, and the open-water efficiency derived from them.

This presentation has the advantage of being independent of scale and regime: a model propeller and its full-size counterpart compare on the same curves, scale effects aside.

Heliciel produces these coefficients in two ways. Its multi-point analysis can plot the curves against the advance coefficient for marine propellers: that is your BEM curve. And the CFD bench's Matrix Me! table gives, for each calculated point, the forces from which you will build the corresponding CFD point.

Watch the definition used: the coefficients depend on the reference area and speed chosen. Check that yours match those of the source you are comparing against, or you will be measuring a convention.

coefficient curves against the advance coefficient with the CFD points overlaid
The reference curves, Heliciel's BEM curve, and your CFD points: three sources on one graph

4: Mesh convergence first

A gap measured on a single mesh is worthless. As long as the result changes when you refine, what you are measuring is the discretisation, not the method.

The approach: calculate the same point at three increasing preset levels, for example Trend, Study and Fine study. Plot the quantity against the number of cells. Two possible outcomes:

the curve flattens out: you have reached mesh convergence. The gap with the reference is then attributable to the method, and it can be quoted;
the curve keeps rising or falling: further refinement is needed before concluding anything.

This is the step most readily skipped, and it is the one that distinguishes a validation from a coincidence.

5: Interpreting the gap

Once mesh convergence is secured, the residual gap with the reference is explained by identifiable causes:

the absence of laminar-turbulent transition in the turbulence model, which makes drag pessimistic, all the more so at low Reynolds number;
numerical diffusion from the schemes, particularly with a first-order upwind scheme;
the Reynolds number gap between your calculation and the original test, often conducted on a model;
differences in conditions: an open-water test has no hull upstream, no nearby free surface, no support; neither does your calculation, but not necessarily in the same way;
the geometry itself: the discretisation of the exported model, the fineness of the trailing edge, the root fillet.

Run the same comparison with your BEM result. You then get two gaps — BEM against measurement, CFD against measurement — and it is their pairing that is instructive. It happens that BEM is closer to the measurement than CFD on global forces: that is not paradoxical. BEM relies on polars derived from measurements and on corrections calibrated on experience; it has experience built in where CFD has only equations.

graph of BEM-against-measurement and CFD-against-measurement gaps for several operating points
The two gaps, point by point: it is their comparison that qualifies your chain

6: Turning that gap into an asset

The result of this tutorial is not a propeller: it is a figure, the typical gap of your chain on a given family of machines. Keep it, and document it: the preset used, the mesh convergence level reached, the reference used, the range of advance coefficient covered.

This figure has immediate practical value. It lets you state, in a calculation report, what uncertainty a result carries — instead of delivering it bare. It also lets you recognise an outlier result: a thirty-percent gap on a chain you know typically runs at five is not a discovery, it is an error to chase down.

Repeat this exercise once per family of machines, if you handle several: a chain qualified on marine propellers is not qualified on ducted fans, whose configuration and numerical difficulties differ.

Building the calculation report itself is covered in tutorial no. 20.

The collection of twenty CFD tutorials

« Previous: The optimisation loop | Next: From the hull to the propeller »

This series complements the Heliciel design tutorials, which cover the BEM side: blade design, choosing the operating regime, performance curves. CFD comes after them, to verify and to see.