open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi everyone
To avoid the error (SOURCE POINT N°... OUT OF DOMAIN), please can you explain me how telemac2d find the mesh node to be assimilated to a geven source point defined by its coordinates (X,Y).
I think there are something like a tolerance used by telemac 2d (a search area defined by a circle...thus a given radius !). If yes, which subroutine telemac2d uses to do that and can I modify the distance used to identify the closed mesh node?
Any suggestion is welcome
thank you in advance

Hi

I think it's done in proxim.f

Hope this helps

Regards

Thank you first Mr Coulet to the support you give to everyone of us.
I work on a large river with many sources (tributaries). Telemac2D finds some sources out of domain which is not actually true. Hervouet indicates that is due to the tolerance used to identify the closest node of the mesh to the sources point and explain that small tolerance is higly necessary in prallel mode. This tolerance is about 10^-8.
In my case, I have a scilab code which calculate my sources (hydrology code : X,Y,Q(t) ) which is impossible to get value of X and Y with 10^-8 precision. In one of your post you explain that is possible to use mesh node rather coordinate of sources, I mean for each source point we can indicate its associted node from the mesh. This will be great. Shoud I modify proxim.f ? If yes, how to introduce sources in the steering file (as coordinates are mandatory)?
Regards

Hi
There is no already available solution I think.
But as the purpose of proxim subroutine is to convert the coordinates given in the steering file to the node number of the mesh, you probably could adapt this subroutine to your own need.
You should find a solution to fill the array IP with the right node_number of each source

Hope this helps

Regards

Génial,
Now lets assume that think is done, What should I do with de sources data I already introduced in the steering file (see below). Should I let them without any change and only modify proxim.f?

Abscisses des sources :
x;x2.........;xn
débits des sources
0;0;...........;0;
ordonnées des sources :
y1,y2,y3;.....yn
thank you in advance

yes, you should let the data as is.
The data will be use for the array allocation.
With this modification, you will just change the way telemac associate the coordinates with the node number.
Nevertheless, if you plan to run your model in parallel, you probably should take care of the decomposition ...
This will probably add some complexity to this treatment ...

Regards

Actually, now I'm only testing the model (scalar mode) before to run it in parallel with 200 processors. Hope the modifications do not raise other problems.
Thank you a lot Mr Coulet for your availability and support.

Try to well think about the parallelism because this will probably generate some problems in the future.
Node number will change locally, according to domain decomposition in parallel so it's really important to use the function local_to_global.
If a source node belongs to more than one domain, it could be taken in account in all the domain but with the right discharge!

Good luck

If I specify the node number for each source point (rather than coordinates), how do a point source can be shared by 2 or more sub-domains (in parallel mode)?

Hi
You don't have any control on the domain decomposition.
When you run a computation on 200 processors as you wrote, there is a split of the domain in 200 sub-domains. Each domain is a small mesh and the interface nodes belongs to at least 2 domain.