Welcome, Guest
Username: Password: Remember me

TOPIC: Vertical Mesh Transformation and Non-Convergance

Vertical Mesh Transformation and Non-Convergance 2 years 3 weeks ago #41404

  • toby.jhnsn
  • toby.jhnsn's Avatar
  • OFFLINE
  • Expert Boarder
  • Posts: 161
  • Thank you received: 8
Hi,

I've been trying to simulate an outfall discharge which requires a vertical planeto be fixed at the invert level of the outfall (around 2.2 m below MSL, LAT ~= -1.2 m MSL).

I have the model up and running fine with standard sigma transformation (MESH TRANSFORMATION = 1), however immediately see non-convergence on the first timestep when trying to implement the fixed plane at -2.2 with sigma transform everywhere else.

I have a total of 7 layers, and my modifications to the USER_MESH_TRANSF are as follows:

      DO IPLAN = 1, 7
        TRANSF_PLANE%I(IPLAN)=1
      ENDDO

      TRANSF_PLANE%I(5)=3
      ZPLANE%R(5)=-2.1D0

I assume these is something strange happening with the spacing of the layers somewhere in the grid, although I'm not sure how to debug the issue. Has anybody experienced this issue?

Thanks,
Toby
The administrator has disabled public write access.

Vertical Mesh Transformation and Non-Convergance 2 years 3 weeks ago #41406

  • toby.jhnsn
  • toby.jhnsn's Avatar
  • OFFLINE
  • Expert Boarder
  • Posts: 161
  • Thank you received: 8
Following up on this thread - its worth nothing that my minimum bed level in the model is -1.5 m MSL.

If the fixed level is specified below this level how does the model handle the mesh in these areas? I suspect there is something strange happening in these locations.
The administrator has disabled public write access.

Vertical Mesh Transformation and Non-Convergance 2 years 3 weeks ago #41410

  • pham
  • pham's Avatar
  • OFFLINE
  • Administrator
  • Posts: 1559
  • Thank you received: 602
Hello Toby,

Do you mean that every node of your mesh has its elevation above -1.5 m? With TELEMAC-3D, the code generates its own 3D mesh based on the 2D mesh of the geometry file, but the mesh is for the whole volume of water.

For every mesh transformation, plane 1 follows the bottom and plane number NUMBER OF HORIZONTAL LEVELS follows the free surface.

If you prescribe an intermediate horizontal plane below the bottom and if the elevation of the node of the mesh is below the bottom, TELEMAC-3D will generate a stack of planes following the bottom but with a minimum spacing between every plane until the node is above the bottom. This minimum spacing can be changed with the keyword MINIMUM DISTANCE BETWEEN PLANES CLOSE TO THE BOTTOM (default = 0.2 m). If you are not satisfied with this little spacing (after decreasing it), you have the opportunity to consider crushed elements (see the keyword THRESHOLD HEIGHT BEFORE CRUSHED ELEMENTS in the TELEMAC-3D user manual). In that case, you may have prisms with volume = 0 as they disappear from the 3D mesh.

But if you say that the minimum elevation of the bottom is -1.5m, with your implementation of user_mesh_transf, that will mean that you use 2 planes (number 6 and 7) to model the 3D volume of your 3D model.
You may have to change your previous choices.

Hope this helps,

Chi-Tuan
The administrator has disabled public write access.

Vertical Mesh Transformation and Non-Convergance 2 years 3 weeks ago #41412

  • toby.jhnsn
  • toby.jhnsn's Avatar
  • OFFLINE
  • Expert Boarder
  • Posts: 161
  • Thank you received: 8
Hi Chi Tuan,

Thanks again for your invaluable assistance.

What I meant is that the highest level in my model mesh is -1.5 m MSL - there are levels deeper this. At lowest tide level I'm left with around 0.5 m of water depth at this shallowest location. With sigma transform my 7 layers would be evenly spaced at these locations (element thickness of ~0.08 m)

In the case of fixing a level at -2.1 m MSL, the layer will intersect with the bottom and be forced towards the -1.5 m MSL bed level. I was concerned that some kind of very thin elements would be generated which could be leading to these instabilities.

I have just carried out a test case with MINIMUM DISTANCE BETWEEN PLANES CLOSE TO THE BOTTOM = 0.2 m (default). I set the layer above the bottom to be -10 m MSL. However, at the location of -1.5 m MSL bed levels the thickness of the first bottom element is only ~0.03 m). This is far to thin to likely statisfy the courant criteria so could be the reason for my issues and nowhere close to the specified 0.2 m so not sure whats going on here.

Also in the case of tidal flats are the elements simply crushed? Does Scheme 5, Option 4 work with these crushed elements or am I back to using 14?

Regards,
The administrator has disabled public write access.

Vertical Mesh Transformation and Non-Convergance 2 years 3 weeks ago #41416

  • pham
  • pham's Avatar
  • OFFLINE
  • Administrator
  • Posts: 1559
  • Thank you received: 602
Hello again Toby,

Sorry, my explanation was correct if there is enough water above the bottom (and fixed planes). If not enough, the minimum distance between fixed planes and above can be less (you can read the exact implementation in the calcot subroutine of the sources/telemac3d folder where the elevation of every 3D node is computed), it depends on the free surface elevation at that time (and bottom elevation).

With crushed elements, the good thing is that nothing is computed inside (contrary to small prisms) and every advection scheme compatible with tidal flats should work, in particular LIPS.

Chi-Tuan
The administrator has disabled public write access.
The following user(s) said Thank You: toby.jhnsn

Vertical Mesh Transformation and Non-Convergance 2 years 3 weeks ago #41418

  • toby.jhnsn
  • toby.jhnsn's Avatar
  • OFFLINE
  • Expert Boarder
  • Posts: 161
  • Thank you received: 8
Ok thanks - I've spent most of the day trying various combinations of options for:
  • THRESHOLD HEIGHT BEFORE CRUSHED ELEMENTS
  • MINIMUM DISTANCE BETWEEN PLANES CLOSE TO THE FREE SURFACE
  • MINIMUM DISTANCE BETWEEN PLANES CLOSE TO THE BOTTOM

Unfortuantely I haven't been able to get any successful runs with the exception of sigma transform only. I have tried adjusting the level of the layer and the layer number but the issue persists. I assume this would be a common problem for coastal models where the layers intersect with a beach for instance.

I would like the layer spacing to be as large as possible between the elements and I guess if they're below say 0.1 m then to crush them to avoid courant number issues.
The administrator has disabled public write access.

Vertical Mesh Transformation and Non-Convergance 2 years 3 weeks ago #41420

  • toby.jhnsn
  • toby.jhnsn's Avatar
  • OFFLINE
  • Expert Boarder
  • Posts: 161
  • Thank you received: 8
Hi Chi-Tuan,

The non-convergence was associated to the DYNAMIC PRESSURE STEP and the PPE solver. Will try to see what I can change to improve the stability of it, but for now am using the hydrostatic version to get around it.

Regards,
Toby
The administrator has disabled public write access.

Vertical Mesh Transformation and Non-Convergance 2 years 2 weeks ago #41480

  • pham
  • pham's Avatar
  • OFFLINE
  • Administrator
  • Posts: 1559
  • Thank you received: 602
Hello Toby,

Can you upload the steering file + output file when running with non-hydrostatic version please?
In particular to see what you use to solve the PPE + how it behave.

Chi-Tuan
The administrator has disabled public write access.
Moderators: pham

The open TELEMAC-MASCARET template for Joomla!2.5, the HTML 4 version.