open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I'm trying to model shear generated flow recirculations in cavities lateral to the main stream. Typically, I'm concerned with upstream and downstream ports for navigation locks: flow is going through a weir, and a guard wall protect an almost still water area for vessels entering/leaving the lock. Due to the shearing at the opening of thoses ports, some vortices develop in the "still" area.

For the downstream port, I've got quite nice Telemac results (see picture below). These results are obtained with a probably quite classical parameterisation of Telemac (solving wave equation, with caracteristics for advection, constant eddy viscosity = 1E0). I could tune the eddy viscosity to get correct shape and intensity of the vortex, by comparison with scale model measurements.


For the upstream port, I used to get some satisfactory results in the past, with V5P6, when using the primary equation, with semi-implicit type of advection (2). I was indeed told long time ago that this type of advection was less diffusive... The picture below shows a typical vortex obtained this way, that is quite coherent with measurements...


With V6P2 however, I don't get this vortex anymore in the upstream port... Even with the primary equation and type of advection fixed at 2. I've also tested different eddy viscosity, mesh refinement (down to 1m in the shearing zone), and lateral wall friction values. Without any success.

Would anybody have a suggestion on how to improve these results?

Thanks in advance,

Didier

Hello Didier,

No idea ! Especially if it is the same with primitive equations and wave equation. We did recently a series of tests on asymmetric recirculations after sudden expansions, and it turned out to be correct, though you must be careful with turbulence and friction. One thing that was changed between 6.1 and 6.2 is the semi-implicitaion (one coefficient was hardcoded in the wave equation option and this has been released), see the feature logs at the bottom of the home page. Maybe we can have a look at your steering file to get more insight.

Regards,

Jean-Michel Hervouet

Hello Jean-Michel,

Results are indeed the same with primitive and wave equation.
Tests with the semi-implicitation coefficient either resulted in no difference in some cases, or in increased simulation instabily when decreasing the semi-impliciation. But no vortex.

I attach my steering file, together with the geometry, in case you would have any further idea having a look on it...

Regards,

Didier

Failed to upload the geometry files. Try again.

Didier

Hello,

General remarks on the parameters (I am not sure it will act on the recirculations) :

use :

NOMBRE DE SOUS-ITERATIONS POUR LES NON-LINEARITES = 1

it will go much faster (more than 1 is for waves or tsunamis)

use :

MASS-LUMPING SUR LA VITESSE = 1.
MASS-LUMPING SUR H = 1.

this is a full mass-lumping, and simplifies the job for solvers.

use :

OPTION DE SUPG : 0;0

options 1 is default, it is a hidden diffusion to avoid oscillations, that is replaced now by COMPATIBILITE DU GRADIENT DE SURFACE LIBRE, so it is better not mixing both.


You say you tried to change the lateral friction, this is a sensitive parameter for recirculations, try also without lateral friction. For diffusion I am not sure, you need to transfer momentum to trigger a recirculation in your geometry, so a low diffusion is maybe not the better.

This is all I can see so far,

Regards,

Jean-Michel