open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi Martin,

There is no "one best way" to define density.
It all depends on what you are trying to accomplish.
Without seeing your t3m file I can't diagnose why the thin elements are being produced.

There is a lot of flexibility built into the BK mesher (perhaps too much) The point is to treat it as a "computer assistant tool" rather than a black box that you simply run and assume that it will generate an optimal mesh. In all cases you the user has to verify that the mesh is correct.

A couple or rules of thumb.

1. fixed geometry (hardlines and hard points) should not be placed close to each other or to the outline unless the local density is at least somewhat compatible.

2. "Less is More". That is to say you should supply the minimum number of constraints possible and let the mesher proceed automatically. I have seen people supply a set of hard points at approximately the resolution of the final mesh and wonder why there are "funny" triangles.

Cheers... Martin

Hi,

Here is my t3m. I am a "green" user of BK and for that matter, 2D modeling, but I dont think i've used that many constraints. But, I'll let you be the judge :)

Thanks,

Martin

Hi Riadh,

I've finished simulating with the changes you mentionned and although there is still quite a bit of variation in the flow at the downstream boundary, the level variation near the upstream boundary has decreased about tenfold, which is quite good. But, after 2 hours, it still hasn't totally stabilized. I will try to run it for much longer over the weekend to see what happens.

Martin,

I've tried what you suggested and the time step for the same Courant number has increased by about 12 %, which is nice. The simulation has finished yet, so I can't comment on the other impacts.

Thanks again to both of you for the provided help !

Martin

Hi Martin,

So your constraints are OK although I would have extended the density polygons in a couple of areas to cover small bits of the outline that are left hanging.

I've found the source of the very thin elements. It has to do with numerical precision (significant digits). (I covered this in a previous post somewhere in a Blue Kenue thread).
If you shift all of your source data (-350000 in X and -5200000 in Y) (use "Edit->Shift in X/Y") the resulting mesh will be much more reasonable. NOTE: The next version of the Blue Kenue (probable release in the fall) will be fully double precision and these problems will disappear.

Speaking of versions, I notice that you are using an older version (3.2.31) than is currently available on our WEB site (3.3.4). I recommend that you upgrade.

Cheers... Martin

Here are the release notes for 3.3.4

Sept 2012
Version 3.3.4

1. New feature: Added "Edit->Line Set->Invert Selected Line"
2. New feature: Added Lat/long graticle option to 2D View "Display" property page. NOTE: the view must be a projected view (ie. UTM, Albers, etc.) for this to have any affect.
3. New feature: Added support for saving FE mesh objects (scalar and vector) to Tecplot format.
4. New feature: Added "Load" and "Save" buttons to the colourscale property page. Allows the user to store and recall colour scale spec. to/from "cs1" files.
5. New feature: Added multi-frame mapping for r2s to r2s
6. BUGFIX: Mesher: Delaunay triangulation: detect and deal with machine precision problem (epsilon) Usually surfaced when a very regular mesh is generated.
7. BUGFIX: colour scale legend: ensure that legend changes properly when selecting a different attribute.
8. Various minor BUGFIXES and corrections

Hi,

I've just shifted all my data like you said and just by looking at the mesh I can immediately see the difference. I will run a case over the weekend and see what happens. Thanks for the tip !

As for the current version of BK, I will try to see how long our IT guys take to integrate it to our environnement.

Many thanks again,

Martin

Hi,

I ran the proposed modifications over the weekend and my model still does not converge. The level near the upstream boundary is stable within about 1 mm, which in itself might seem great, but the volume of water still varies by as much as 500 m³ between result printout (1000 time steps, about 3.5 minutes). I've shifted my mesh and can see the difference, added a softer initialization as suggested by Riadh and still doen't reach a steady state. I've joined my current files.

Could anyone provide help please ?

Regards,

Martin