open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello!
A question maybe to experienced river or flume (for example lab flume) Telemac modellers..

Telemac calculates velocities in the nodes of the solid boundaries. Now I want a no slip condition, so at the solid boundaries the velocities should be zero. Otherwise I have also problems with flow continuity which could be relevant especially when modelling narrow flumes etc.
Would you activate in such cases the parameter LIUBOR = 0 and LIVBOR = 0 in the boundary condition file? And also when modelling large rivers (width = 200-1000 m)?

I would be glad for an answer!
Regards,
Clemens

Yes, to have no slip conditions, you may have LIUBOR and LIVBOR = 0 for relevant nodes. However, depending on the mesh size, this is not very good. Think that near a wall you have a logarithmic velocity profile and that finite elements will do a linear interpolation between wall and nearby inner points, which is not very logarithmic. So unless you have very very small elements, some friction is better than a no slip condition that may slow down the flow.

Another point is that whatever the velocities that you see on solid boundaries, there is no continuity problem. The flux across solid boundaries injected in variational formulation is really zero, so no wall will contribute to a flux of water. This is a little bit difficult to understand when you see some velocities slightly across a solid wall, but this normal flow is discarded. This is the effect of a weak formulation as opposed to a strict "no normal flow" condition. Velocities across solid walls give you however an idea of local errors due to discretization.

With best regards,

Jean-Michel Hervouet

Thanky you very much for the answer.

I don`t understand the second part regarding the continuity problem.

Not clear to me is: if I consider for example a straight channel:
Do the velocities at the solid boundaries contribute to a portion of the flux in the downstream direction?
If so, then the calculatet "inner" velocities are lower compared for example to a real case (physical model test of this straight channel).

Best regards,
Clemens

Yes, the velocities on solid walls will not contribute to normal fluxes, but they will contribute to longitudinal fluxes, so your inner velocities may be lower than reality. However with no slip condition it is probable that the inner velocities will be larger than reality. This depends on your mesh size, but given the fact that boundary layers are generally very small, and smaller than mesh size, no slip condition is rather artificial (its influence propagates till the next interior points due to linear interpolation).

JMH

Thank you very much for the answer. I hope I can attend the Telemac User Club (to get further answers

So my assumption:
when modelling narrow channels or small rivers, use of a denser mesh near the solid boundary to minimize the influence of the velocities at the solid boundaries to the contribuiton of the longitudinal fluxes.

Best regards,
Clemens

I'm not very much familiar with Finite Element methods, which, as far as I know, use mostly a node-centered variable location.
Finite Volumes methods usually use a non-staggered variable location. From a physical point of view with this scheme it "sounds" more correct to me that at the solid boundary I have zero velocity and in the wall nearest center of the cell a velocity derived for example via the log wall law. The cell width doesn't matter much, provided that the y+ value is in a reasonable range.

For me it is not much clear, that with Finite Element methods I have to accept, that my core velocities are lower than reality. It is easier to implement a node-centered scheme in Finite Element methods?

Best regards,
Clemens

I don't think there is much difference between finite volumes and finite elements here. In finite volumes some boundary conditions (fluxes, etc.) are just in an imaginary extra cell out of the domain. In finite elements, saying that we have velocities on the boundary is admitting that we are outside the boundary layer, so that this tiny layer is not in the domain. The important thing is to have the correct wall friction.

JMH