open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I am using telemac for dam break modelling. For my modelization, I used the method of EF and the méthod of VF in order to make a comparison between these two methods.

I have a question about the different numerical schemes whiche are proposed (with the key word "type of advection") : How can I know which scheme I have to use ? I don't know how to choose the scheme and what scheme will be the best for EF or for VF and why ?

A second question, I don't know too how to choose the "finite volume scheme". I don't understand what it is exactly ?

Is there someone who can help me please ?

Have a nice day !

Hello,

This is an answer to the first question : TYPE OF ADVECTION

For dam breaks you can use the method of characteristics : TYPE OF ADVECTION 1;5, but with a little slow down of the flood wave (the second number 5 is now mandatory).

It is better (but more expensive) to use the distributive schemes adapted to dry zones :

TYPE OF ADVECTION : 13;5 that will require also :

TREATMENT OF NEGATIVE DEPTHS : 2
CONTINUITY CORRECTION : YES

I would suggest that you start with 1;5 and when all is correct try 13;5 to see the sensitivity.

With best regards,

Jean-Michel Hervouet

Hi
This an answer for your second question.
To use Finite volume method you have to add the following keyword:

EQUATIONS = 'SAINT-VENANT VF'
To choose which scheme you need to add :
FINITE VOLUME SCHEME =
For this keyword you can choose
0: Roe scheme (1st order in space, good numerical properties but risk of negative water depth)
1: Kinetic scheme (1st orde in space, good numerical properties (shock capturing, positive depth, mass conservation))
2: Kinetic scheme (2nd order in space, same properties as for 1)
3: Tchamen scheme (1rs order in space, good numerical properties, well suited for wetting and drying but more diffusive)
4: Zokagoa scheme (1st order in space, same properties than 3)

It it important to notice that FV kernel is explicit in time, so it needs more CPU time than FE. Furthermore you need to add the following kew words
VARIABLE TIME-STEP =YES
DESIRED COURANT NUMBER = 0.8 (between 0 an 1)
DURATION = (in stead of number of time steps )

Moreover, you can obtain 2nd order in time by adding the following keyword
NEWMARK TIME INTEGRATION COEFFICIENT = 0.5

With my best regards
Riadh

Thank you both of you for your answers, they are very helpful !

Can I ask an other about the FINITE VOLUME SCHEME ?
When you use finite volume, there are two schemes (the one with the key word "finite volume scheme" and the one with "type of advection") ? Or in this case the one with "finite volume scheme" replaces the one with "type of advection" ?

And what is the difference between a first order in space and a second order ?

Hello

Yes there is no need for "type of advection" at least for the standard hydrodynamical part of the simulation.
The order in space gives the precision with whom Telemac reproduces the real solution. In other words, if you adopt a 1st order scheme, you have a discretization error of O(dx^2) where dx is the mean length of the elemements in the mesh you are using.

All my best regards

Riadh

Ok, thank you very much ! It is more understandable for me now !