open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all,

My TKE profile is odd: at the bottom I'm experiencing high TKE values when I expect them to be lower. In the attached image, you can see the profile with the modelled (black) vs. my observed (blue) data. When I expect them to curve back there is instead a high value.

I'm assuming this is something to do with the boundary conditions?

I'm using a k-e model on both horizontal and vertical. My case file is attached; if anyone could give it a glance over or generally give some words of wisdom I'd highly appreciate it!

Many thanks!

Looking at the diagram, the high value at the bottom could be the result of taking a boundary value.

Where are the boundary values calculated? Is it at the first point at the bed, at the node, or just above the bottom at a specified point? I'm thinking I may need a smaller cell size at the bottom.

I've been looking into this further to no avail:

- I though it looked like a problem with the no-slip condition. I changed from Manning's law to Nikuradse (ks), as the bulk coefficient approach might be averaging near the bottom. However, results with Nikuradse have the same issue i.e. very high TKE values at the bottom, instead of the curve back to lower values as expected.

- I also considered that I was sampling from the boundary: the numerical bottom being 1/10 of the height of first mesh layer. However, with higher Ks I was getting larger TKE at elevations higher than 1/10th (around 0.22m).

- Looking at the Z+ (Z+ = U* * Z/v) and ks limit (KS < 33 * Z), I found that the ks limit was too small when using small meshes but the Z+ was too small when using larger mesh sizes. This makes an awkward balancing act which is difficult to resolve. This might explain the issue with Nikuradse, however I'm unsure given that the issue also effects Manning's n.

The only other thing I can think of is that there's something going on with the wall function, and else I need a model to resolve the near-bed boundary layer process. Such a one that's not wall-function based. Does Telemac have such an option?

Hi,

I think the blue dots in your plot is from measured data?

Near the bottom, it seems that Reynolds number decrease and it comes to a transition between turbulence and laminar flow. The k drops due to the turbulence damping.

This is not easy to capture. Normally, you will need a low-Reynolds k-epsilon model with certain kind of damping functions towards the bottom. The standard k-epsilon model is only valid for the fully developed turbulent flow.

Or, maybe you could try to use finer grid in vertical and smaller time step? But I'm not sure if it could solve your problem.

Regards,
Qilong