open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

Here is a first answer, after examining your case :

The reason for loss of monotonicity is the keyword :

NUMBER OF SUB ITERATIONS FOR NON LINEARITIES : 2

This is a problem for me, there is probably somme loss of compatibility between the continuity equation and the advection equation in that case. In the while (the solution may take time and appear only in version 6.2) just suppress this keyword (and take a smaller time step if necessary).

I have noticed other problems:

In the initial conditions, there is a water depth equal to 0. on a dyke in an harbour. I wonder if this is normal (you would expect that dykes are not near to being flooded in normal situations). It seems to raise problems and triggers high velocities after a few time steps. It would be better to have a lower bottom at that place, so that the dyke boundary has never depth = 0. Generally speaking it is not recommended to have a slope going from 12 m of water to 0 in a single element, better put a wall in this case.

In your boundary conditions file you specify no slip conditions on boundaries (lines starting with 2 0 0). This contradicts the fact that you give a friction law on boundaries in your steering file. I changed the 2 0 0 into 2 2 2 (which may have triggered the problem mentioned above on the dyke...)

I would start first with a mixing-length model, rather than a k-epsilon model, which may not work properly given the mesh size (the velocity gradients will be underestimated, not to speak of k-epsilon on tidal flats).

In case of k-epsilon model, the keywords:

COEFFICIENT FOR HORIZONTAL DIFFUSION OF VELOCITIES and
COEFFICIENT FOR HORIZONTAL DIFFUSION OF VELOCITIES

should be equal to 1.E-6, as they are understood as being the laminar viscosity. Same problem with tracers.

Otherwise here are a few advice on keywords:

You can remove the keywords startin with : "TREATMENT ON TIDAL FLATS..."
You can remove CONSISTENT PTOJECTION : YES
You can remove MINIMAL VALUE FOR DEPTH = -1000.
You can remove RATIO OF 3D AND 2D TME STEPS : 1

Choose:

SOLVER FOR PROPAGATION : 1
SOLVER FOR PROJECTION : 1

I hope this helps,

With best regards,

Jean-Michel Hervouet

Thanks for your advice!

I've changed my steering file like you said. However, I got strange output on the screen at some time steps like

...
PROCESS BOUNDARY DATA AT 6710.00000000000
PROCESS BOUNDARY DATA AT 6715.00000000000
PROCESS BOUNDARY DATA AT 6720.00000000000
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN
PROCESS BOUNDARY DATA AT 6725.00000000000
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN
PROCESS BOUNDARY DATA AT 6730.00000000000
PROCESS BOUNDARY DATA AT 6735.00000000000
PROCESS BOUNDARY DATA AT 6740.00000000000
...

I guess the improper values of salinity may cause this error because I had the same situation when I didn't initialize salinity correctly.

Hello,

No, this is a crash. You should look where these infinite values occur, I'm sure it is due to large velocities on tidal flats on steep slopes in the harbour. You should avoid uncovering on the dykes in this harbour, it is not normal that water reaches at some locations the top of the dyke in your initial conditions.

JMH

You are right. The dykes caused some problems in this coarse mesh and I will fix that area and try it again.
Thanks for your help!

Hello, qilong

I'm using TELEMAC-3D to study the mixing process in a large estuary(Yangtze) and the turbulence model I use is also the k-epsilon model. Having read some of your posts, I find that our study cases are quite similar: tidal estuary, tracer, tidal flats(dikes),k-epsilon model and it seems that you do it well with a relatively large time step(5s) and I've followed your settings, unluckily my model cannot run even with a rather small time step of 1 or 2 second (mesh size: 50~5000m, 20 levels), maybe it is a matter of the k-epsilon model and I have tried to modify Kmin&Emin, nothing improved ...
Could you please be so kind to share with your ideas how to modify the k-epsilon model to allow for a relatively large time step...
Thanks in advance.

With best regard.

Hello,

After studying Robinson's case, I came to the conclusion that the problem is indeed the tidal flats with k-epsilon. I got the case running with modifications in subroutine kepscl3.f, where all divisions by DIST (distance to the bottom, should be replaced by MAX(DIST,1.D-6) for example. This was the main problem, however the linear systems for diffusion remain not very well conditioned (this is normal on tidal flats) and it takes large numbers of iterations, this is why I am still working on the case.

With best regards,

Jean-Michel Hervouet

Hi,

I really want to express my special thanks to Mr. Hervouet for your concern and kind help:). Modeling with TELEMAC is a wonderful learning experience though some issues remain there.

With best regards.
Robinson.

Hello Robinson,

I checked your steering file butI didn't find some keywords for dealing with tidal flats such as:

TIDAL FLATS = YES
OPTION FOR THE TREATMENT OF TIDAL FLATS = 1

Besides, the schemes for advection should also be consistant with way of dealing with the tidal flats. For example:

ADVECTION STEP = YES
SCHEME FOR ADVECTION OF VELOCITIES = 14
SCHEME FOR ADVECTION OF DEPTH = 5
SCHEME FOR ADVECTION OF K-EPSILON = 14
SCHEME FOR ADVECTION OF TRACERS = 14

I also have some problems in my 3D cases with k-epsilon model and I haven't solved it yet. But those keywods may help you to improve stability.

Best regards,
Qilong

Hi Qilong,

Thanks for your kind help. :)
Several key words on tidal flats like 'TIDAL FLATS = YES
OPTION FOR THE TREATMENT OF TIDAL FLATS = 1' are the default value of TELEMAC system so I didn't add it. About the advection schemes, after reading Mr. Hervouet's book I find that characteristic method is unconditionally stable(my dream is to achieve a relatively large time step) so I also use the default value, maybe it's not suitable for tidal flats. I'll try your suggesions. Thanks again.

with best regards.
Robinson

Hello,

Yes, the method of characteristics is unconditionally stable, with monotonicity, it can deal with tidal flats without stability problems, but it is not mass-conservative, so I use it only for the advection of velocities, and also k and epsilon.

JMH