open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I have a case study of an estuary with TELEMAC-3D and to begin with, I cut the river part of the whole area as my model domain(Model_T3D.zip). Thanks to the steering file of one study case 'Berre lagoon' provided by Jean-Michel Hervouet,I set up my model but find something puzzling: it seems that T3D is quite sensitive to the preconditioning rule and to speed up a little, I choose the key words below:

PRECONDITIONNEMENT POUR LA DIFFUSION DU K-EPSILON : 34
PRECONDITIONNEMENT POUR LA DIFFUSION DES VITESSES : 34

then it always stops after several steps with an error message(error_01.PNG), I really don't understand why this doesn't work for me(I did not use preconditioning 17). Then I make them the default value:2, it goes well to the end. But on the screen it sometimes shows 'EXCEEDING MAXIMUM ITERATIONS...(error_02.PNG)', and modeled time series elevations/velocities(elevation.PNG) is a little bit weird.
I have checked several times but really don't know where I'm doing wrong...
Can anyone experienced please give me a hand, thanks in advance!

with best regard.
Robinson

Hello Robinson,

The preconditioning option 34 is a combo of preconditionings 2 and 17. You can read an explanation in the dictionnary for the keyword PRECONDITIONING FOR PROPAGATION. That is why you got your error message.

Did you try to decrease the time step of your simulation? Then, does your simulation still stop?
To investigate on a bad case, you could try to look at your simulation just before the error message occurs.

Hope this helps,

Chi-Tuan

Hi, Chi-Tuan

Thanks for your quick reply.
I'm sorry I cannot find PRECONDITIONING:34 in TELEMAC-3D user manual(release 5.8), though using preconditioning 2 is ok.

The minimum size of mesh in my case is around 270 m, and the maximum velocity is around 5 m/s, which makes me think that a time step of 30 s is ok with safe Courant number.
I have tried a time step of 10 s and just like before it can go through the end but the message of 'EXCEEDING MAXIMUM ITERATIONS' still appears...
I will try more tests.

with best regard.
Robinson

Hello,

The problem probably comes from the k-epsilon model being used with tidal flats (the epsilon equation does not seem to work in your case, see NaN in listing). Maybe you should first try to set up your model with a mixing-length model to start with, and when it is about OK move to k-epsilon. In this case look at subroutines CSTKEP and SOUKEP, e.g. in CSTKEP KMIN and EMIN should perhaps be modified and EMIN set to 1.D-10 as in version 6.1, instead of 1.D-16. With large elements, time steps of 30 s and large aspect ratios we are certainly far beyond the normal range of use of the k-epsilon model... In SOUKEP you will see that there are two models, option 2 being hardcoded. In the next version it will be much more parameterised.

With best regards,

Jean-Michel Hervouet

Hi,Jean-Michel Hervouet

Thank you very much for your very helpful instructions.
Yes there're tidal flats both in my small and large models, and in my large model there are even dikes to be considered.

I have already tested the constant viscosity model, then mixing length model and it seems that both work quite well and efficient, but when it comes to the k-epsilon model, I've never succeeded till now.

I don't have any experience in turbulence modeling, but I can see in the validation manual 'Berre Lagoon', it also has large elements(~2000m), large time step(30s)and large aspect ratios with k-epsilon model, and it works quite well. The differences I can see are the tidal flats and velocity magnitude.

'...certainly far beyond the normal range of use of the k-epsilon model' then could you please give me some suggestions on how can I improve my model, can I just modify the default value of KMIN and EMIN?

Thanks again.

with best regard.
Robinson

Hello,

What you could just try is a value of KMIN = 1.D-10 (take subroutine CSTKEP in a FORTRAN FILE). Then I do not know, the k-epsilon model probably degenerates when the depth is 0, the Berre lagoon has water everywhere, but elements with no volume, and it still works. I'll try your case and see what happens, but cannot give you a deadline.

Regards,

JMH

Hi,

It's quite OK, thank you so much!
I'll try the value of KMIN = 1.D-10.

with best regard.
Robinson

Hello,

I've had some other test cases with K-Epsilon model and I find that if those tidal flats and dikes are removed or replaced by a relatively large water depth, then it works fine. So as you've told me, my problem is a matter of the tidal flats/dikes. And I've tried KMIN = 1.D-10 which does not help. also I've tried 14:N-SCHEME FOR TIDAL FLATS for both velocities and k-epsilon but Exceeding maximum iteration always show up after some steps. I'm really fruastrated...
Is it possible for a large model with tidal flats to use k-epsilon model? Is there any principals related to this kind of problems to follow ?
thanks very much.

with best regard.

Hi Robinson,

TELEMAC-3D user manual release 5.8 is not up-to-date... Anyway, the explanation about cumulated preconditings can be extended: 14 = 2*7 or 21 = 3*7 are described, but it is the same for 34 = 2*17.
When I talked about the dictionnary, I meant the very useful (and often mentionned) file describing every keyword in the lib directory of a software; in this case, it is telemac3dv6p2.dico

FYI, TELEMAC-3D user manual release 6.2 is to be available in a very near future in French. For the English version, you will have to wait a little bit more.

Regards,

Chi-Tuan

Hi Chi-Tuan,

Thanks for the reply. I think I understand what you mean, though I cannot find the preconditioning option of 34 in 'telemac3dv6p2.dico'.

I'm really looking forward to the next release of TELEMAC-3D user manual

with best regard.
Robinson