This boat propeller CFD tutorial covers the switch to a liquid fluid: immersion depth, absolute pressure, cavitation.

Start with tutorial no. 1

Moving from an air propeller to a marine propeller does not change the calculation chain: the same presets, the same two phases, the same solver. What changes is the fluid — and it changes everything: a thousand times denser, it produces forces a thousand times greater, and it can vaporise under the suction side. This tutorial follows a boat propeller from the BEM reference through to the wake, and pauses on the two points specific to water: absolute pressure and cavitation.

• Prerequisites: a marine propeller in Heliciel — here the boat propeller model supplied with the software;
• Duration of the case shown here: 49 s to 5 min 20 of meshing depending on the preset, 41 s to 1 min 58 for Phase 1, 4 min 18 for Phase 2;
• What you get out of it: a verified order of magnitude for thrust and torque, and an absolute pressure map on which to read the cavitation risk.

• Tutorial outline:

1. The propeller studied, and what the BEM says about it
2. Choosing the fluid, and what it entails
3. Immersion depth, and why it must be set before meshing
4. Creating the case and meshing
5. Three meshes on the same propeller
6. Phase 1 MRF: the first result
7. Phase 2: the sliding mesh
8. Viewing the flow
9. Reading absolute pressure and assessing cavitation
10. What this calculation does not model

1: The propeller studied, and what the BEM says about it

The boat propeller model supplied with Heliciel serves as the example. A methodological point first: a supplied model opens with its design point at zero. Run a calculation — Analyse operating point — before switching to CFD, otherwise the BEM reference written into the case folder will be zero and the whole comparison will be empty.

The results bar gives the BEM verdict at the operating point:

The designer's results bar: pressure jump, axial force, torque, power. These are the numbers the CFD will have to confirm.

2: Choosing the fluid, and what it entails

The project's fluid is carried over as-is by the CFD test bench: density, viscosity, saturated vapour pressure. Three immediate consequences, best kept in mind before being surprised by the results:

• The forces change scale. At 1 028 kg/m³ against 1.2 for air, a 1.14 m propeller turning at 105 rpm — a trawler's rotation speed — already produces more than a tonne of thrust. The colour scales, residuals and convergence thresholds adapt accordingly: nothing to set, but don't expect the orders of magnitude of the air case.
• Reynolds rises, Mach collapses. 1.3 million at the blade tip for 6,46 m/s: the flow is fully turbulent, and perfectly incompressible (Mach 0,004). This is the most comfortable regime for an incompressible solver.
• The reference pressure is no longer the atmosphere. It depends on immersion depth — that is the subject of the next section.

3: Immersion depth, and why it must be set before meshing

The depth of the propeller shaft sets the domain's reference static pressure. In our case, a shaft immersed at 1 m gives a reference pressure of 110 085 Pa, almost 9 000 Pa above atmospheric. This value is locked in at the first meshing: it goes into the case's boundary conditions, and changing it afterwards in the project no longer changes anything in the CFD folder already written.

So set the immersion depth before creating the case. This is not a presentation detail: it is what decides whether the depression calculated on the suction side crosses the vapour pressure or not.

4: Creating the case and meshing

File > New CFD case (from the current model). Heliciel exports the STL of the 3D prototype, detects the case type — here free propeller, no duct — and prepares the OpenFOAM folder. A non-ducted marine propeller uses the same presets as a free air propeller, from Level 1 Express to Level 8 Reference; tutorial no. 3 details the catalogue.

The results control panel, in the Results tab: phase displayed, surface pressures, flow lines, cut planes and iso-surfaces.

5: Three meshes on the same propeller

The same propeller was meshed and run through Phase 1 at three levels. Same geometry, same operating point, same solver: only the mesh changes.

Level 1, 15 246 cells: the seven blades are reduced to a patchwork of facets. At this stage, the geometry the solver sees is no longer really your propeller.

Level 3, 45 761 cells: the blades regain their shape, and the pressure gradient becomes readable from the leading edge to the trailing edge.

Level 5, 85 086 cells: smooth surfaces, continuous gradient. This is the tutorial's reference level.

Two lessons, and the second is unexpected:

• Agreement on a coarse mesh is not validation. At Level 1, the CFD torque falls to within 2.5% of the BEM torque — remarkable agreement… that unravels as soon as you refine: −18% at Level 3, then −24% at Level 5. The two errors of the coarse mesh — unresolved blade and missing boundary layer — were cancelling each other out. Had you only computed Level 1, you would have concluded the validation was successful.
• Here, refining degrades mesh quality. Non-orthogonality stays capped at 70°, but the maximum skewness rises from 6.05 to 10.36 and the number of concave cells doubles, from 4 222 to 8 423. The reliability index drops from 50 to 44% while the cell count is multiplied fivefold. Seven wide, heavily twisted blades that overlap in projection and have thin trailing edges: this is a geometry significantly harder to mesh than a slender three-blade rotor. On this type of blade, the limit is not the cell count but snappyHexMesh's ability to place them cleanly.

Practical consequence: on a marine propeller with heavy blade overlap, expect a residual gap that does not close as you refine. Here it settles around −24% on thrust and −18% on torque. This is not a verdict on the propeller, it is a verdict on the mesh — and the right course of action is to check the boundary layer (tutorial no. 9) and mesh quality before adding cells.

6: Phase 1 MRF: the first result

Phase 1 solves the flow in MRF — Multiple Reference Frame — a fixed mesh, with rotation introduced as a term in the equations inside a cylindrical zone. Steady-state calculation: it looks for equilibrium, not the history of the motion (tutorial no. 4). On Level 5, it converged in 1 min 58 and 129 iterations.

After Phase 1: 818,6 N of thrust and 192,9 N·m of torque for the blades, 2,12 kW at the shaft, propulsive efficiency 0,608 — to be compared with the BEM's 1 156,4 N, 253,19 N·m and 0,654 recalled above.

Efficiency is the quantity that holds up best: 0,608 against 0,654, a 7% gap, where thrust and torque show 29 and 24. This makes sense — both fall short together, and their ratio suffers less than each taken alone. It is also a trap: an efficiency that lands close does not guarantee the forces are correct.

7: Phase 2: the sliding mesh

Phase 2 actually turns the rotor: the propeller zone slides against the fixed domain, and the two meshes communicate through a non-conformal interface recomputed at every time step (tutorial no. 5). On this case it is fast — 4 min 18, and it stops on its own once the forces converge.

This time the calculation stops on its own: the monitor declares the forces stabilised after 0,9 revolution (a variation of 0,96 % on thrust and 0,60 % on torque, against a 1 % threshold). The averages above are therefore consolidated, and they confirm Phase 1 to within half a point. Above all, note the fluctuation: 21 N peak on 834, i.e. 2,5 % of thrust, at 12,25 Hz. This is the pulsation that fatigues a shaft line and that can be heard in a hull — and it is the one figure that Phase 1, steady-state by construction, could not give.

After Phase 2, the displayed source becomes "Sliding Mesh (real rotation)". Watch how you read it: the panel shows the final instant, while the comparison against the BEM is made on the average of the last blade passages, recorded in the case's benchmark file. In an unsteady calculation, a snapshot is not a result.

8: Viewing the flow

One rule first: uncheck the 3D model to look at surface pressures. The grey STL draws itself over the coloured field and produces a speckled, unreadable map. Without it, the propeller's coloured shape is the result.

Surface pressures in Phase 2, 3D model hidden: overpressure on the pressure side, depression on the suction side, and the trailing edges marked in red.

The flow lines over the pressures tell the story of the fluid's path: the acceleration as it crosses the disk, the wake being set into rotation, the contraction of the stream tube downstream — the exact opposite of what a wind turbine does.

Flow lines coloured by axial velocity: the fluid enters slowly (blue-green) and exits accelerated (red-orange). The stream tube narrows: this is the signature of a propulsive machine.

The cut planes each answer a different question: the vertical one shows the accelerated jet downstream, the normal one cuts across it, and the cylindrical one follows the blade at a given radius and is read mainly in turbulence.

Vertical cut plane of axial velocities: the propulsive jet, faster than the upstream flow.

Cut plane normal to the flow: the active annulus seen across, and the slowed-down core behind the hub.

Cylindrical cut plane at r = 0,228 m (40 % of the radius), in turbulence: the boundary falls exactly at the propeller plane — but read the scale before concluding: it runs from red at zero up to blue at the maximum, and it is the upstream side, in blue, that carries the strongest turbulent viscosity. Probed at 40 %% of the radius: 0,0159 m²/s at 1.2 m upstream, against 0,0058 m²/s in the jet — that is 11 800 times the viscosity of water upstream, and 4 300 times downstream. This is the opposite of what is expected, and it should be read as a marker of the conditions imposed at the domain inlet, not as a map of the wake's real turbulence. This is the view that best shows the disorder the machine manufactures.

Two iso-pressure levels (109 062 and 109 317 Pa): the detached pockets at the blade tip are the depression cores. Only tick one or two: all four together form a continuous mass that no longer shows anything.

9: Reading absolute pressure and assessing cavitation

This is the point where a marine propeller truly stands apart. The test bench displays pressure in absolute pascals, not as a deviation from the reference: our calculation's scale runs from 107 867 to 111 255 Pa around a reference of 110 085 Pa. The cavitation question then comes down to one line:

does the calculated minimum pressure drop down to the fluid's saturated vapour pressure?

Here: a minimum of 107 867 Pa against a vapour pressure of 1 300 Pa at 10°C. The margin is two orders of magnitude: no cavitation risk at this operating point, and this was expected — 105 rpm for a 1.14 m diameter works out to 6.5 m/s at the blade tip, an idling tug's speed.

The reasoning itself remains valid whatever the case, and it is on fast propellers — launches, outboards, tidal turbines at the blade tip — that it gets tight. Three habits to take:

• read the minimum of the pressure scale, not the average;
• check that it does correspond to the suction side at the blade tip, where relative velocity is maximum — the cylindrical cut plane and the iso-pressures show it;
• remember that immersion depth shifts the whole scale: the same propeller at 30 cm below the surface instead of one metre loses 7 000 Pa of margin.

Heliciel also has a cavitation analysis on the BEM side, which works element by element; the two approaches complement each other — one gives the criterion, the other shows where.

10: What this calculation does not model

Three omissions, worth knowing so as not to over-interpret:

• No free surface. The calculation is single-phase and gravity-free: water fills the entire domain. No waves, no ventilation from the surface, no shallow-immersion effect other than through the reference pressure.
• No cavitation as such. The solver does not make the water change phase: it gives the pressure, and it is up to you to compare it with the vapour pressure. A map that would drop below the vapour pressure signals a risk, it does not simulate the vapour pocket or the accompanying loss of thrust.
• No hull ahead of the propeller. The upstream flow is uniform, whereas a boat propeller works in the hull's wake, with a velocity field that is strongly non-uniform over one revolution. This is precisely what Phase 2's RMS underestimates here: our 2,5 % fluctuation is that of a propeller in open water, not that of a propeller behind a sternpost.

Tutorial no. 19 covers the hull → propeller sequence, and no. 14 the energy-capture case, of which the tidal turbine is the marine counterpart.

The collection of twenty CFD tutorials

« Previous: Checking an airfoil polar  |  Next: Wind turbine and tidal turbine »

1. First CFD propeller simulation
2. Reading the OpenFOAM case folder
3. Choosing the mesh preset
4. The MRF zone and Phase 1
5. Phase 2 with sliding mesh
6. Comparing BEM and CFD
7. Reading the convergence of a run
8. Turbulence models
9. Boundary layer and y+
10. Reading a pressure map
11. Where the forces come from
12. Checking an airfoil polar
13. Marine propeller and cavitation (you are here)
14. Wind turbine and tidal turbine
15. Fan in a closed duct
16. Aircraft propeller, cruise and static thrust
17. The optimisation loop
18. Calibrating on a reference
19. From the hull to the propeller
20. The deliverable calculation file

This series accompanies the Heliciel design tutorials, which cover the BEM side: blade design, rotation speed selection, performance curves. CFD comes after them, to verify and to see.