"tighten the mesh tolerance" ?

[Nadeem Malik]

Hi Neks,

I have come across a problem in generating a thin mesh with genbox and genmap: "tighten the mesh tolerance".

Previously, at created a straight forward 3D periodic box [-pi,pi]^3  of 24x24x24 elements with genbox, and genmap with tolerance 0.2. This created the *.re2 and *.ma2 files required for compiling. No problem, and good compilation and good results from simulations.

Now, I want to create a very thin domain of size [-1.5pi,1.5pi]x[-1.5pi,1.5pi]x[-pi/24,pi/24] -- essentially 2D and homogeneous in z. I now have 36x36x1 elements. genbox is fine. But when I run genmap I get the above error (the last few lines of output from genmap are given below). Even when I enter tolerances as small as 0.001 I get the same message:

        ..................

        1295 6  5  3  0.00000000E+00     2732 shift

        1296 2  5  3  0.00000000E+00     2733 shift

        1296 4  5  3  0.00000000E+00     2734 shift

        1296 5  5  3  0.00000000E+00     2735 shift

        1296 6  5  3  0.00000000E+00     2736 shift

     1     2     4  Matrix:  SELF!!

       1 SELF!!        1        2        1        2

       2 SELF!!       37       38       37       38

 ABORT: SELF-CHK            1           5           1           6

 Try to tighten the mesh tolerance!

 exit status           6

Please advise what to do.

Thanks

Nadeem

[YuHsiang Lan]

Hi Nadeem,

Do you have periodicity in Z direction?

genmap actually needs 3 layers of elements to build the connectivity for periodic.

Since you only have 1 layer, that might be the issue here.

   https://github.com/Nek5000/Nek5000/issues/777

If you insist to have only 1 layer +periodic, you can manually overwrite the connectivity in uservert.

Here is an example.

   https://github.com/Nek5000/nekRS/blob/master/examples/eddyPeriodic/eddy.usr

Basically, setting

Hope this helps,

Yu-Hsiang

--

[Nadeem Malik]

Thanks for the details. The domain is triply periodic. I will keep all this in mind for future.

But I have changed tack, and decided do a 2D simulation instead  -- that is another thread.

Regards, and happy new year (tomorrow)!

Nadeem

[Rajesh Ramesh]

Hi,

I am trying to simulate a 2D pseudo case (i.e. 2D in XY plane + small thickness in Z direction with only one element) where I have 1 element with periodic boundary conditions in Z direction. The link provided for an example to overwrite the connectivity in uservert doesn't exist anymore. Is there another example available and can someone please explain the workaround process i.e. how to modify the connectivity in the .map file to make 1 layer with periodic boundary conditions work? Looking forward to your responses.

Best,

Rajesh

[YuHsiang Lan]

Hi Rajesh,

Please find the uservert in this example

  https://github.com/Nek5000/nekRS/blob/v22.0/examples/eddyPeriodic/eddy.usr

Alternatively, you can modify the map file by manually copying the 2-5 columns to 6-9 columns with the same format (excluding the first line header).

Hope this helps,

Yu-Hsiang

--