open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

this actually not a question, it is just a notice. I have noticed that when modelling systems with large water depths, one must refine the vertical mesh near the bottom, at least adding two layers in the lowest 1% of the water column, otherwise Telemac will not capture the impact of the friction which is limited to a very small layer near the bottom. And that means that without such refinement the velocity profile will not be correct.

Hello,

There is a difference between friction laws. The Strickler-based formulas, like Strickler or Manning, and Chézy, use the depth-averaged velocity and should not be too dependent on the refinement near the bed. On the contrary the Nikuradse law assumes a logarithmic profile near the bottom, thus some refinement is needed.

Regards,

JMH

Actually I used Manning however I noticed that the velocity profile is not typical as the literature, the velocity at the surface was 0.09 m/s and at the bottom was 0.08 m/s, this with version 6.1. with version 6.3 the behavior is already improved, so I found a zero velocity at the bottom, but what was strange is the maximum velocity was close to the bottom not at the surface especially if too many layers are close to the bottom.

attached are the velocity profile for version 6.3 and version 6.1. I have to say that the hydrodynamics with 6.3 is totally different from that of 6.1 for the same case

Hello,

Yes, this is strange. I checked on the case "canal", which is not refined on the bottom, and we do not see this behaviour, so this difference with previous versions was not noticed. It probably depends on the refinement itself, which must have some regularity I imagine... this is to investigate, but for example we assume that the first plane above the bottom is already in the logarithmic profile, which precludes refining too much.

What you can try is the new option:

VERTICAL VELOCITY DERIVATIVES : 2

1 is the default, with 2 the computation of vertical derivatives on velocity is based on logarithmic derivatives, which is better if you have a logarithmic profile.

With best regards,

Jean-Michel Hervouet

P.S. subroutine with an example of head loss attached again here. I do not know where it has gone in my previous post.

thank you very much indeed for your continuous support and help