open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi
I have been simulating a heated discharge into a lake using Telemac 2D and got good results. Now I am trying to use Telemac 3D and my simulations are taking very long time to run. To start I've created a test case where I am running just for an hour, time step=1s, mesh edge size(30m) and 5 vertical layers. I tried to use the scheme 1 for advection of velocities but I was having an error so after checking the forum I changed it to 14. I am using the non-hydrostatic version. Also I am using very simple turbulence model (constant horizontal and vertical viscosity).

I wanted to initialize the temperature varying with depth so I am using PARTICULAR initial conditions and I programed the tracer initialization. So, initially I made it to vary linearly with depth but then since it was taking such a long time to run, I though there might be something wrong in my code and I decided to just use two values one for the deeper layers and another for the upper layers. Also originally I wanted to use Mesh transformation =3 (so I can specify the depth of each layer) but then I though this could also slow the simulation and I end up changing it to 1 (constant layers).

For this test I didn't include the wind and air temperature.

I really don't know what the problem could be and in which ways I can improve it. To give an idea I am using 24 processors and it took 12 hours to run 30 min of simulation. I am attaching my case and Fortran files. In the fortran file, I made changes in lines: 64,310-311,320-321,473-496,673-680.

Also I included my output for the first 30min of simulation.

Thanks in advance for your help

Viviana

Hello,

I see nothing wrong in your simulation. The progresses made between 6.1 and 6.2 would make the method of characteristics work for the velocities and it is true that scheme 14 is rather expensive. The cost of 3D compared to 2D is a minimum of 4 to 5 more (case of one layer of elements) 10 times or more with 5 planes would be normal. Now you have very few iterations of solvers, so you can probably have a larger time step.

You can try the test case malpasset in 2D and compare with the benchmarks on this website to see if your parallel installation works properly.

With best regards,

Jean-Michel Hervouet

Hello

Thanks so much for your help and time. I will try to increase the time step and see what happens. The parallel installation is working fine, since I have using the same cluster to ran the 2D simulations.

Some questions:

1) in the CONDIM subroutine, I wanted to initialize the temperature for the bottom layer to vary linearly with depth. Is it correct to do it like this:

IF(NTRAC.GT.0) THEN
DO ITRAC=1,NTRAC
! CALL OS( 'X=C ', X=TA%ADR(I)%P, C=TRAC0(I))
IF( ITRAC.EQ.1 ) THEN ! ITRAC=1: is your tracer
DO IPLAN = 1,NPLAN ! IPLAN=1: bed, IPLAN=NPLAN: surface
IF( IPLAN.GE.3 ) THEN
DO JPOIN = 1,NPOIN2
TA%ADR(1)%P%R((IPLAN-1)*NPOIN2+JPOIN)=24.D0
ENDDO
ELSEIF ( IPLAN.EQ.1 ) THEN
DO JPOIN = 1,NPOIN2
TA%ADR(1)%P%R((IPLAN-1)*NPOIN2+JPOIN)=5.D0
ENDDO
ELSE
DO JPOIN = 1,NPOIN2
TA%ADR(1)%P%R((IPLAN-1)*NPOIN2+JPOIN)=5.D0+19.D-1*
&abs(MESH3D%Z%R((IPLAN-1)*NPOIN2+JPOIN)-MESH3D%Z%R(JPOIN))
ENDDO
ENDIF
ENDDO
ENDIF
ENDDO
ENDIF

2)If I want to read the initial temperature from a file, how should I go about it?
3)This might be a very elementary question but I am getting confuse. In order to use mesh transformation=3, if the maximum depth in the domain is lets say around 12m, then:
ZPLANE%R(01) = -12.D0. Then what will happen if in the domain there some small areas that are below this level (-12.3 or 12.6m)? will they not be included? or this will cause an error?

Sorry for all the questions

Thanks in advance

Viviana

Hello,

1) The Fortran is correct, the abs in the formula is not necessary. I hope it will give something between 5 and 24 but it depends on mesh.

2) It depends on what file it is, if it is a variation on the vertical, a cluster of points. You may have to interpolate every of your 3D mesh in view of a cluster of points, placed at random.

3) For transformation 3, the first plane will always follow the bottom, it cannot be fixed.

Regards,

JMH

Thanks so much, that clarified all my questions. I will try to use a linear variation for now and then if I need I will try to read from a file.
I really really appreciate your time and help.

Thanks again

Viviana