open TELEMAC-MASCARET The mathematically superior suite of solvers

I want some clarification on incorporating sub-meshes.

I'm refining the banks of a channel, and using sub-meshes to do so. My main mesh encompasses the entire channel, with the sub-meshes within the main mesh.
I notice when I use the T3 Mesh Generator it runs on the lowest default edge length, which includes the edge length for the bank sub-meshes. My main channel then ends up being more refined than necessary. If I put both the bank sub-meshes and the main channel in the sub-mesh child object then the all the meshes are as refined as wanted. However, looking at the resulting mesh in the 2D window, the main mesh elements are visible beneath the bank sub-meshes, despite being in the same mesh object.

Am I doing this correctly? Do I need to create a hard outline for the sub-meshes to fill?

Clarification would be really appreciate.
Many thanks.

I think my confusion comes from that, in the boundary conditions output, it treats all the sub-meshes (both banks and channel) as having their own boundaries and nodes (see image attached) when I was expecting the boundary to occur solely on the perimeter of the object.

Looking at the Telemac 2D tutorial for BlueKeneue the merged meshes only have boundaries on the perimeter, but not so for mine. Why would this be?

Hi
I think you didn't understand how the submesh works on the meshing process of BK.
A submesh is a part of the final mesh without any modification. So you cannot place the bank and the main channel in the submesh area of the final mesh generator.
As your submesh is a strong constraint for the meshing process, it takes time for the global mesh to grow from the submesh edge lengh to the right edge length.

In the global mesh generator, you should include the submesh of banks and give de edge length value you want (default value) but according to the configuration, this is not necessarily the edge length you will have.
Maybe you could post all the BK file in order to try were you make a wrong manipulation

Regards

Thankyou, I see where I went wrong. Primarily, my approach using the global mesh generator was wrong. Using the channel generator for the bed, then using that as a submesh and then letting the global mesh generator fill in the banks seems to have worked.

My last question comes from how the channel mesher approaches cell resolution. For example, I can assign the banks to have a default edge length of 0.2m however the channel generator has different options (cross channel node count/along channel interval.) How does the channel mesher approach assigning edge lengths? I couldn't find anything in my documentation.

Hi
you could only define the along channel interval as a kind of edge length.
On the cross direction, you fix the number of node in the cross section which remain the same all along the channel so if the distance between the bank change, the edge length change...

regards

Hi,

So to find the number of nodes needed along the channel for a certain resolution (say, 0.2m) it would just be a case of dividing the required resolution by the length of the channel?
Is there a way to ensure a certain mesh resolution? I realise that there will be some variation given how the edge length is produced, so it it's more of a case as approximating the channel as close to the desired 0.2m resolution as possible.

Many thanks.

Hi,

The AlongChannelInterval should really be called AlongChannelEdgeLength. It explicitly defines the
element resolution along the channel. (Except for
the very last row of elements all edge lengths are
constant along the channel).

The edge lengths across the channel are computed from
the local channel width divided by the CrossChannelNodeCount
if you only supply the two banks.

If a thalweg is supplied the cross channel edge
lengths are computed separately on each
"half"/side of the channel.

The algorithm works (more or less) as follows.
1. The 2 banks (or banks and thalweg) are resampled
to the AlongChannelInterval value.
2. Cross channel edges are then defined as straight lines
between pairs of the points along the channel (from 1).
3. Intermediate nodes are computed along these cross
channel edges by dividing by CrossChannelNodeCount. This
results in a series of quadrilateral element chains
along the full length of the channel.
4. These quadrilaterals are then split by their shortest
diagonal.

Hope this helps... Martin

PS: Way back when, we investigated the possibility of having
the ability to define a variable "along channel" edge length
sort of like the mesher density but never found the funding to
implement this. Another possibility would be to define
an along channel aspect ratio target...
(lots of possibilities )

Thanks for explaining, that really helps. I had treated the AlongChannelNodeCount as similar to edge length, so it's good to know I'm on the right track!