open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

Recently I tested a telemac-2D model which simulated a river and its tidal flats. But I found something strange in the result. I got velocities at closed boundary. Does anyone know what happened at these boundaries? Thank you very much. Attached are the steering file and the part of result.

Hello,

This is a difficult point to understand:

The impermeability of closed boundary is ensured in a weak form (in the sense of finite elements). That is to say that the flux across boundaries taken into account for the continuity is really zero for every degree of freedom (so there is no problem of mass conservation), but that the observed velocity may go across the boundary if the finite element approximation is not very good locally. So velocities across a closed boundary are a sign that locally you have a rather bad resolution (but mass conservation is preserved).

I am not sure that the bottom of your figure is an entrance (in which case the mesh could be improved). When the boundary conditions force a velocity to be across a nearby solid segment, the next velocity on the segment will tend to be in an opposite direction so that the integral of fluxes be however zero, you see this tendancy at the bottom right of your figure. This is why I give the advice to prescribe a profile at entrances so that velocities touching a solid boundaries be zero. This is specially true for lateral boundaries with an inflow at 90° angle to the main flow.

With best regards,

Jean-Michel Hervouet

Thank you very much for helping me to understand this problem. I will try to prescribe a profile at the entrance. And indeed the bottom of my figure is the entrance. You said the mesh could be improved. Could you tell how to do improve the mesh? Will the result become better if I use smaller resolution? Thanks again!

Hello,

You can improve the mesh at the entrance with a boundary that has 90° angles on the left and right sides. The general idea is that no velocity must be prescribed across a solid segment, so I would suggest that at least the first row of elements at the entrance forms a rectangle, and the prescribed velocities must be normal to the upstream boundary, and tangent to the lateral sides (like in a rectangular channel).

With best regards,

Jean-Michel Hervouet