open TELEMAC-MASCARET The mathematically superior suite of solvers

I'm looking for help in using a Hotstart file in Telemac2D. I'm running a large ocean model (750,000 nodes) and needed to get it going with a smallish timestep. Now that it is up and running, I'd like to increase the timestep. I've updated my .cas file to include a Hotstart, and a change in timestep:

/ INITIAL CONDITIONS
/

---

/
/COMPUTATION CONTINUED = NO
/INITIAL CONDITIONS = 'CONSTANT ELEVATION'
/INITIAL ELEVATION = 0
COMPUTATION CONTINUED = YES
PREVIOUS COMPUTATION FILE ='SalishSea_HS.slf'

SalishSea_HS.slf is the results file from the small timestep start-up run; it opens fine in BlueKenue, I can see all the variables and all the timsteps. However, when I try and run Telemac with a continued computation it crashes right after starting to read in the previous computation file:

....
CORFON (TELEMAC2D): 1 BOTTOM SMOOTHINGS

TITLE OF PREVIOUS COMPUTATION: Salish Sea_Calib_Feb16

NAME: VELOCITY U UNIT: M/S
NAME: VELOCITY V UNIT: M/S
NAME: WATER DEPTH UNIT: M
NAME: FREE SURFACE UNIT: M
NAME: BOTTOM UNIT: M
NAME: SCALAR FLOWRATE UNIT: M2/S

Any ideas gratefully received,

Tamsin

Hello,

We would need to know exactly where it crashes. For example you can add the line :

DEBUGGER : 1

in your parameter file, so that we shall get approximately in which subroutine it stops. It could be that :

* Your new time step is too large (just try with the previous time-step to be sure)
* You use the k-epsilon model and did not save k and epsilon, so the restart reinitialises k and epsilon from scratch.

You should also keep the lines :

INITIAL CONDITIONS = 'CONSTANT ELEVATION'
INITIAL ELEVATION = 0

because the subroutine CONDIN is called even in case of restart (e.g. Boussinesq equations need to know the initial depth considered as reference depth)

With best regards,

Jean-Michel Hervouet

C:\fakepath\SalishSea_v15_Fast.cas_2011-08-24-16h10min10s.sortie
Merci Jean-Michel.

I've tried your suggestions with the following outcomes:

1. I added the INITIAL ELEVATION and INITIAL CONDITIONS back into the parameter file --> No change

2. I changed the timestep back to the original smaller timestep --> No change

3. I checked for the K epsilon terms. I'm using the exact same parameter file for both the small timestep and large timestep - apart, of course, from the changes to to the initial conditions and timestep. The TYPE OF ADVECTION is currently set to 1;5 (in both cases).

4. I added the DEBUGGER codeword to the parameter file, but I'm not sure where to look for outputs from this subroutine. I've attached my .sortie file, which to me looks no different from previous efforts, in hopes that you can help some more.

Merci de ton aide,

Tamsin

Hi Tamsin

You're right when you said there's no interesting information in your .sortie file.
The program seems to be blocked during the analysis of previous computation file.
Is there a particular message on screen after the crash?
Is the previous computation file a very large file? if so, maybe you could try to extract only the last time step and try the hot start with this file.

Hope this helps

Merci Christophe.

There is no message on the screen after the crash. I just get the command prompt.

The previous .slf file I'm trying to run from is just over 400 MB - large, but I imagine not too unwieldy. I'll definitely try and extract only the last timestep - only I don't know how. I'm working with BlueKenue - can you give me some more directions?

Thanks

Hello,

OK, it seems that it crashes right after reading the file or when reading the file. Did you wait a sufficiently long time ? we see no crash message and maybe it is reading a very long file before getting to the end of it (but well it normally skips the records without reading all of them...). It could be linked however to the large size for some reason. Either you can send us your full case or give an address to download it if it is too big, or you can add in your Fortran file subroutines like LITENR in library bief and put flags in it to trace what is read and when it stops.

With best regards,

Jean-Michel Hervouet

Thanks.

I have waited days (several times) at this point - so I'm hoping that this is not the problem. I'm posting the files to an ftp site:

ftp.nhc-van.com/main.html
Login: TELEMAC
Password: Help

I'd like to remove these as quickly as possible from the ftp site - so if you can, please email (tlyleATnhc-van.com) when you've got them.

Thanks again,

Tamsin

Hi Tamsin

Try to remove variables of your restarting file one by one (You can use Fudaa to do it) and run telemac after that. I suggest to start with friction coefficients and bathymetries. Telemac will re-load them from your fortran file or from your geometry file.
I hope that after that Telemac will run normally

Good luck
Riadh

Hi
I downloaded your computation file but:

• the steering file is not a continuation sterring file
• there is 4 liquid boundaries but in the test_tides.prn, we could only find Q(1), SL(1), SL(2) and SL(3) (Note that SL(1) is not used because the first boundary is a discharge boundary). There is probably a problem here!
• if i try to run the simulation on my computer i have an error message

  *************************************
                       *    END OF MEMORY ORGANIZATION:    *
                       *************************************
  
   ELEBD: ERROR OF NUMBERING IN THE ELEMENT:******        POSSIBLE REASON:
  
          THE BOUNDARY CONDITION FILE IS NOT
          RELEVANT TO THE GEOMETRY FILE
  
  
  
   PLANTE: PROGRAM STOPPED AFTER AN ERROR

Please check those points and maybe you will find the origin of your problem...
Hope this helps

Christophe,

Thank you very much for taking the time to look at my files. And, apologies, because it seems I must have posted an incorrect liquid boundaries file. I hope I didn't waste too much of your time. The liquid file I'm using has correct inputs (Q(1), SL(2), SL(3), SL(4)) and is working just fine.

Regardless, I seem to have things working now. I took Riadh's advice (see above) and removed all but the velocity and free surface components from my selafin file and tried again - this time with success. So, it seems that the problem was probably related to the size of the hotstart file (now 2.5 GB before parsing the unnecesary information).

So, for the moment my problem is solved. However, I may run across this again if and when my velocity and free surface only selafin files become unwieldy. What I think I need to do is to create a single timestep selafin file. I've tried to do this in BlueKenue to no avail. I can't seem to find an obvious way to save or extract a single timestep from the selafin file. If I save/extract/calculate a single timestep for any of the variables and then add these to a new selfafin object - they lose their metadata related to time. If I update the time headers in the text file and try to add these to a Selafin file, BlueKenue crashes. I will mention this in a post I am working on for the pre/post-processor forum.

Thanks again for your time,

Tamsin