Transient SST initialization fails on refined mesh / higher Reynolds number

Please login with a confirmed email address before reporting spam

Hello,

I am modelling 3D flow around an inclined circular cylinder in water. The cylinder axis is aligned with the z-direction and periodic conditions are used in the spanwise direction.

Instead of rotating the geometry, I decompose the towing velocity into:

Un = Usin(alpha) Ua = Ucos(alpha)

where alpha is the angle between the cylinder axis and the towing direction.

I impose Un as the cross-flow velocity, and I account for the axial component using a reference-frame/Galilean transformation by applying a cylinder moving-wall velocity of approximately -Ua.

The final goal is to run sweeps over approximately:

U = 1-25 knots alpha = 1-40 deg D = 20-25 mm

The attached model contains a stationary k-epsilon study followed by a transient SST study, with the SST initialized from the k-epsilon solution.

On a coarse mesh this works well and vortex shedding develops correctly. However, when I refine the mesh or increase the Reynolds number, the transient SST often fails right at startup.

Typical errors are:

"Failed to find consistent initial values"

or the segregated solver repeatedly reduces the first timestep and is not able to converge.

The problem seems to be related to the initialization/nonlinear coupling rather than the linear solver. I have also tested PARDISO.

A stationary SST solution is not a good workaround in this case because the flow is strongly unsteady and the stationary SST problem itself does not reliably converge.

I would appreciate advice on the most robust way to initialize the transient SST solution, especially regarding:

  • Starting SST from a stationary k-epsilon solution.
  • Transferring a developed transient SST solution from a coarse mesh to a finer mesh.
  • Correctly handling k, omega, and the SST wall-distance variable during mesh transfer.
  • Whether consistent initialization should be modified or disabled in this situation.
  • Recommended startup settings for timestep, BDF order, segregated damping, Jacobian updates, and maximum segregated iterations.
  • A robust continuation strategy for future sweeps in U, alpha, and D.
  • My main objective is to find an initialization procedure that remains stable when the mesh is refined and when the operating conditions change.

Any suggestions would be greatly appreciated.



Reply

Please read the discussion forum rules before posting.

Please log in to post a reply.

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.