open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I am trying to perform a short simulation only to test the convergence of my mesh. (Created with Blue Kenue).

The problem is that even after more than 30 tests, all with different meshes, it keeps "Blowing up". The results show values of Free Surface oscillating quickly and then becoming Infinity or NaN.

What it looks like is that the number of iterations for the tracers (Salinity and Temperature) is exceeding the limit (5000) without reaching the minimum relative precision (1E-6).

Every time my simulation would blow up I looked at the results and modified the refinement of the mesh in the region where the problem started in order to make the elements more isosceles-like. But this just seems to move the starting point of the error to another location.

I'm attaching the latest version of my mesh including the necessary files to run the simulation.

My questions are:
What are the basic criteria that a triangular grids in general MUST fulfill in order to make TELEMAC3D run without major issues?

What is making TELEMAC3D return bad values for my mesh, specifically?

Thanks in advance ^-^

mega.nz/#!FtIA2B5B!doJUdycCD6zNKq3S2L56Jd2Bs1GMNfJ99q5QUJP5_PE

Hello

We can not give a full answer for your question. The best way to do is to proceed gradually.

You need to start by looking at the mesh convergence of hydrodynamics, maybe in 2D and then in 3D. Then you check if the obtained mesh is optimal for tracer in 3D. This depends on whether there is stratification or not and if there are other 3D effects or not.
The 3D mesh will be optimized rather by vertical resolution (number of planes and their distribution) and not by changing the 2D mesh.

I hope that this helps

with my bets regards

Riadh ATA

Hello, Riadh

If I understand correctly, you suggest to test the mesh in 2D and then, if it works, in 3D? Then later include salinity and temperature?

And how do I know if the mesh is optimal for tracers in 3D?

My study is on the Southern Brazilian Shelf and the Patos Lagoon, so it includes oceanic and estuarine regions, so what type of vertical distribution of layers do you suggest? Evenly spaced? More concentrated on the surface or on the bottom? Both?

Thanks for your attention.

Best regards,

Phelype

Hi Phelype,

I am also dealing with 3D model, so I am watching every post!

As Riad suggested, you should start from a simpler CAS file, don't make it more difficult.

For example in your CAS file, why did you put:

MAXIMUM NUMBER OF ITERATIONS FOR PPE = 5000
MAXIMUM NUMBER OF ITERATIONS FOR PROJECTION = 5000
MAXIMUM NUMBER OF ITERATIONS FOR PROPAGATION = 5000
MAXIMUM NUMBER OF ITERATIONS FOR VERTICAL VELOCITY = 5000
MAXIMUM NUMBER OF ITERATIONS FOR DIFFUSION OF TRACERS = 5000
MAXIMUM NUMBER OF ITERATIONS FOR DIFFUSION OF K-EPSILON = 5000
MAXIMUM NUMBER OF ITERATIONS FOR DIFFUSION OF VELOCITIES = 5000

If I was you, I would put these as default values, then adjust gradually or just change one by one to see what happens.

Are you dealing with "oil spill"?? why don't you look at examples, like this one: .....\examples\telemac2d\riv_art

Hope this helps

Regards
Huy

Hello, Huy

This CAS file has already been widely used in this type of simulation for my study area. I, for one, have already used it many times.

This large number of the MAXIMUM NUMBER OF ITERATIONS keywords is only because sometimes the solvers struggle a little to solve in the harbour area of Rio Grande, near the outlet of the Patos Lagoon, so the high number of MAXIMUM NUMBER OF ITERATIONS assures that Telemac3D reaches the required relative precision.

About the oil spill, No, I am currently not using it. I am first trying to make the simulation work for the hydrodynamics for then activate the Oil Spill module.

Thanks for your attention.

Best regards,

Phelype

Just guess from 20 files you attached in previous post (such as oil1, oil1). It must be hard to investigate your case.

Just final suggestion for you: check your fortran file as I see that:

OPEN(NFIC,FILE="../3dboundarybig.slf",
...
OPEN(NFIC,FILE="../temparbig.slf",
.......

but actually you don't have these files in attachment?

If you have to deal with "DIRECTION", please understand how to convert them correctly in your input files.

Regards
Huy

Hello, Huy

Oil1 is just the name of the mesh.

The files "3dboundarybig.slf" and "temparbig.slf" are used when restarting the simulation, which is not the case.

The attached files are:
- amplitude.dat - amplitude of the imposed tides.
- fortran_hidro.f95 - Fortran file.
- malha_nivel_2014i.dat - Initial condition of Free Surface Height.
- malhanivel_2014.dat - Boundary condition of Free Surface Height.
- malhapressao_2014.dat - Boundary condition of Atmospheric Pressure.
- malhasal_2014.dat - Boundary condition of Salinity.
- malhasal_2014i.dat - Initial condition of Salinity.
- malhatem_2014.dat - Boundary condition of Temperature.
- malhatem_2014i.dat - Initial condition of Temperature.
- malhatempar_2014.dat - Boundary condition of Air Temperature.
- malhau_2014.dat - Boundary condition of the U component of current.
- malhau_2014i.dat - Initial condition of the U component of current.
- malhav_2014.dat - Boundary condition of the V component of current.
- malhav_2014i.dat - Initial condition of the V component of current.
- malhavento_2014.dat - Boundary condition of U and V components of wind velocity.
- Oil1.cli - Boundary file of the mesh.
- Oil1.slf - Geometry file.
- phase.dat - phase of the imposed tides.
- steering.str - CAS file.
- tabsigma2.dat - Vertical distribution of the planes.

As you can see, there is not much complication to it. Just the Initial and boundary conditions.

Since the wind and current files are in U and V components, there is no need to convert directions.

Thanks for your attention.

Best regards,

Phelype

Hello Phelype

Yes, I mean you can optimize your 2D mesh and the number/positions of vertical planes in order to reproduce good hydrodynamics. This can be done by calibrating the model with some observations on velocities and/or water heights.
After that, you can add salinity and temperature and calibrate them with an additional calibration step.
If you do not have any data mentioning that there is persisting stratification effects, you can start with equidistant distribution of vertical planes. After that, you can adapt this distribution using general Sigma transform.

I hope that this helps

with my best regards

Riadh

Hello, Riadh

I'll try doing this, improving the mesh in 2D then adding the tracers.

Thank you very much for your advice.

Best regards,

Phelype