open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi

In my opinion, it's a mistake to come back to matisse! The problems you have comes probably from a bad use of BK.
I will have a look into your mesh and try to give you some hints but I'm sure of one things, if you used matisse by the past, you should forgot the matisse methodology and try to think with the BK methodology (for example, if you choose resample outline in the parameters, the outline is resample according the density criteria ...)

Regards

Hi again

I opened your project and generate the mesh.
I realized that you work in Lambert93 coordinates and i think that's the main problem.
I have a recent discussion on this subject with the Bk developper (Martin Serrer) and in your case you're facing to a problem of number truncature.
BK could compute mesh in any type of coordinates but when you export your mesh as a telemac geometry file, some rounding are mades with the mesh coordinates.
I've made the test in your case. I generated a first mesh with the BK result and onserved your problem with fudaa with a lot of flat triangle.
So, in BK I shifted the mesh from -6 000 000 on Y axis and generated a second mesh. It's remain only small triangles on the boundary but none in the middle of your mesh!
You could also see the difference in the BK windows (particularly on the outline of buildings where the shape is linear and doesn't look like a stairs) this is due also to the rounding of the OpenGL library used by BK.

Hope this helps

I supposed I have use too long time Matisse. I need to learn the operating mode of BK. But I am sure it is a good mesher!
I thought BK has a mapper so it knows how manage Lambert93 coordinates.
Thank a lot for your help it’s work!

Just to add a bit more info...

The problem stems from how floating point numbers are represented.

A 4 byte float has approximately 7 significant digits (decimal)
of precision in the mantissa.

If you are using Lambert/UTM/etc., the Y values are around 5,000,000.
So “Sub-meter” values cannot be represented.

In this case you have to use a local coordinate system.
(as Christophe suggested)
Use “Edit->Shift in X/Y” to shift all your data before attempting to
generate the mesh.

In addition, the OpenGL library/driver that we use for all
visualization only uses 4 byte floating point.
Hence the “jumpiness” you seen when zoomed in.

Blue Kenue uses double precision internally for most
operations until the end of the meshing process, because
coordinates in a Telemac SELAFIN file are stored as 4 byte floats
so you're snookered in any case.

The SELAFIN file definition does allow double precision coordinates
but I have never seen such a file so I think it's for "future reference".

Hope this helps… Martin