open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,
I have seen in the manual that the value of the velocity diffusivity could be adjusted to damp recirculation patterns.
I am not very familiar with turbulence physics, so I have a question: If I increase the value of this coefficient in order to remove some instabilities, would it affect the velocity field? Which type of values, for instance, would be realistic?

Thank you for any help on this subject,
Caroline

Hello,

First make sure that if you use the k-epsilon model the diffusivity should be at its real physical value of about 1.E-6.

Now the diffusivity has been measured in nature from nothing to 500 m2/s, so it does not help a lot for an educated guess. I always say : take 0.1 m2/s to start with and values up to 2 are usual in real situations (I mean not scale models).

Just keep in mind that the artificial diffusion of an upwind advection scheme is about U*DX/2 where U is the velocity and DX the mesh size, so you won't see any effect if your diffusivity is much smaller than this. Think of a flow of 1 m/s in a tidal computation with 1 km mesh size and you will understand that diffusion is useless in such cases, unless some stabilisation is needed in which case it will require high values of diffusivity, which is artificial.

With best regards,

Jean-Michel Hervouet

Hi,

I have a related question. I have a coastal scale ocean model and I would like to test the sensitivity of my model to velocity diffusivity using a constant viscosity. I would like to know what values of velocity diffusivity I should use in my steering file that correspond to realistic values in nature (up to 500 m2/s) as you mentioned in you post.

All I can find on this is that the default value is 10^-4 and that 10^-6 corresponds to the molecular viscosity of water.

Cheers

Danny

Hello

This is a question that could not be answered !
There is no mean to predict a priori the most realistic value of velocity diffusivity. The sensitivity of the model to this value is another issue. This can be simply done by varying the value from 1-4 to 1 for instance.

with my best regards

Riadh ATA

Hi Riadh,

Thanks for your reply, so are you suggesting that running my model with velocity diffusivity between 1e-4 and 1 will represent a realistic range of velocity diffusivity that you would expect to see in tidal flows?

How can I obtain the corresponding velocity diffusivity magnitudes from these values?

Cheers

Danny

yes, these values corresponds the values given with the keyword VELOCITY DIFFUSIVITY.

kind regards

Riadh

Hi,

I have tried running some simulations with different velocity diffusivity values, however I am getting an error:

GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 100 RELATIVE PRECISION:
NaN
INTERP: POINT 1 NOT IN ELEMENT 1
SHP(1,IP)= NaN
SHP(2,IP)= NaN
SHP(3,IP)= NaN
EPSILO= 1.000000000000000E-006 IPID= 0

PLANTE: PROGRAM STOPPED AFTER AN ERROR

I assume from reading other posts this is an issue with grid size. As I am modelling a large domain my mesh varies from 5 km to 100 m. I have tried varying the velocity diffusion (from 1E-6 to 10) however I get the same error each time. So my question is this; is there any way of testing my models sensitivity to velocity diffusivity using a constant viscosity with such a large range of element sizes? If no, is there any other way of apporaching the problem? Steering file attached.

Cheers

Danny

Hello

The mesh size does not cause any kind of trouble, the element size variation as well.
This could be rather a problem of mesh quality (bad topology, negative jacobian,...)
It could be also a probelm with simple precision of coordinates: if X and Y coordinates have very large values, the simple precision could not catch all the decimals and thus cause overlapping edges. In this case, you need to generate double precision mesh.
To start with, add the keyword CHECKING MESH = YES. Telemac will make a preliminary checkup.

With my best regards
Riadh

Hi Riadh,

Thanks for your reply,I think you're right, its a rounding issue with large x and y coordinates. Is there a way of converting my single precision geometry file to double precision? (preferably not using python as suggested in www.openmascaret.org/index.php/assistanc...ile-sel-in-bluekenue).

Cheers

Danny

Hello Danny

No Blue Kenue does not handle double precision. We are doing that using a commercial mesher called Janet.
An easier way to proceed is to translate your coordinates to smaller values. for instance if you have X values like 6 220 500.562358, you can remove 6 200 000 everywhere in your domain, and then run you simulations with simple precision.
You can do that manually or using BK (EDIT -> SHIFT IN X/Y)

with my best regards
Riadh