Any recommendations on detail level on 3D model of bicycle wheel?

I am designing a velomobile for my master thesis and would like to do a CFD simulation of the wheels with and without wheel covers based on your "Flow Over a Rotating Wheel with Moving Ground" example.

I would like to simulate wheel with somewhat realistic spokes but I have read some recommendations that complex geometry does not work. I suspect that modelling the wheel correctly with crossed and touching spokes would not work so well.

So I plan on making a simple model using:
- radial spoke pattern (no crossing or touching)
- simple spoke cross section with no elbows or change of cross section (butted)
- no spoke nipples

Would that be a suitable level of complexity for Caedium, or could I easily stretch this by adding spoke nipples, butting or even crossed (real) spokes?

I have used Kt-Kl-omega on the velomobile body. Would that be a reasonable turbulence model for the wheel at 10 m/s velocity?

No votes yet

Rotating wheel with spokes

As a first pass I suggest you model the wheel without spokes, i.e., focus on the hub and rim. For this type of simulation (axi-symmetric components) you can just specify an angular velocity on the rotating components as in the rotating wheel tutorial. This keeps the geometry and physics setup relatively simple.

If you choose to model the spokes then there is no limit on the geometry complexity, but you need to ensure that:

  • There are definite intersections at junctions. Having spokes just touch in a tangent condition will cause the meshing to fail. This constraint may make your geometry modeling difficult.
  • The rotating spoke region is confined within a cylinder within your outer flow volume. You will then need to construct (connect all faces) a multi-volume flow-domain, so that you can apply a Moving Reference Frame (MRF) condition to the cylinderical volume,
  • There is adequate resolution (using the Accuracy tool) of the spokes, likely requiring a massive number of cells, meaning your simulation runtimes will be long

If you are looking to simulate laminar to turbulent transition using the kt-kl-omega turbulence model (as opposed to a wall function-based turbulence model) then you will need to have an even finer mesh on the spokes to make sure you resolve y+ down to ~1.

In summary I think modeling spokes would be a complex process and be time consuming both in interactive model setup and in simulation runtime. I would not recommend this approach unless you are confident that retaining (as opposed to a disk or enclosure) and optimizing spokes is likely to significantly impact your overall design.

Trending Now