[Gmsh] lower limit for characteristic length

Christophe Geuzaine cgeuzaine at ulg.ac.be
Tue Dec 16 09:41:01 CET 2008


Zenker, Dr. Matthias wrote:
> Of course. Here it is. 
> 
> BTW, I use gmsh 2.2.6 on a WIN XP SP 2, Pentium 4, 3.2 GHz, 1.5 GB RAM
> computer.
> 

The default random perturbation factor seems to be a bit small in your
case. The following (increased "-rand" from "1.e-9" to "1.e-7")

   gmsh small_cl.geo -2 -rand 1.e-7

seems to work just fine, but I had to stop it before the end due to 
insufficient memory on my laptop. Multiplying all the element sizes by 3 
and using the faster del2d algorithm

   gmsh small_cl.geo -2 -rand 1.e-7 -clscale 3 -algo del2d

produced a nice mesh with about 23 million triangles.


> If I could force gmsh not to ignore the cl of the spline control points,
> it would probably be faster for such a case (if it wouldn't crash). Is
> there a trick to do that? Would it help to put isolated points with the
> desired cl at the desired location?

You should probably use Fields to fine-tune the element sizes. There are 
some examples on the mailing list, as well as in demos/fields.geo.



> 
> Thank you and best regards,
> 
> Matthias
> 
>> -----Original Message-----
>> From: Christophe Geuzaine [mailto:cgeuzaine at ulg.ac.be] 
>> Sent: Monday, December 15, 2008 1:07 PM
>> To: Zenker, Dr. Matthias
>> Cc: gmsh at geuz.org
>> Subject: Re: [Gmsh] lower limit for characteristic length
>>
>> Zenker, Dr. Matthias wrote:
>>> Hi there,
>>>  
>>> the question remains: is there a lower limit for the characteristic 
>>> length cl, and why does gmsh crash when some points have a small cl?
>> It shouldn't crash, except when "cl" divided by the overall 
>> size of model is getting close to machine accuracy (or more 
>> accurately, close to the value of the random pertutbation 
>> used in the meshing algorithms, which is 1e-9 by default).
>>
>> Could you send a file that shows the problem?
>>
>>
>>> I would like to add another one: Is it possible that gmsh 
>> ignores the 
>>> cl of the control points of a spline except for the first and last 
>>> one? If
>> That's what Gmsh does: the cls of the control points are ignored.
>>
>>
>>
>>> so, is there a way to force gmsh to take into account the cl of the 
>>> intermediate points also?
>>>  
>>> Best regards,
>>>  
>>> Matthias
>>>
>>>     
>> --------------------------------------------------------------
>> ----------
>>>     *From:* gmsh-bounces at geuz.org [mailto:gmsh-bounces at geuz.org] *On
>>>     Behalf Of *Zenker, Dr. Matthias
>>>     *Sent:* Monday, November 17, 2008 10:49 AM
>>>     *To:* gmsh at geuz.org
>>>     *Subject:* Re: [Gmsh] lower limit for characteristic length
>>>
>>>     I was wrong, it crashes also for cl=0.0005 (but after a 
>> looong time,
>>>     and with a runtime error).
>>>
>>>         
>> --------------------------------------------------------------
>> ----------
>>>         *From:* gmsh-bounces at geuz.org 
>> [mailto:gmsh-bounces at geuz.org] *On
>>>         Behalf Of *Zenker, Dr. Matthias
>>>         *Sent:* Monday, November 17, 2008 10:16 AM
>>>         *To:* gmsh at geuz.org
>>>         *Subject:* [Gmsh] lower limit for characteristic length
>>>
>>>         Hi,
>>>
>>>         I wonder if there is a lower limit for the characteristic
>>>         length. I have a rather simple 2D geometry, and if 
>> I set the cl
>>>         for just two points to 0.0001, gmsh crashes during 
>> the meshing
>>>         after having given the error message "3 identical points in
>>>         Qtest". The .gmsh-errors file is not written. With 
>> cl=0.0005,
>>>         this does not happen (but it takes veeery long to 
>> compute the mesh).
>>>         Best regards,
>>>
>>>         Matthias
>>>
>>>         
>> ----------------------------------------------------------------
>>>         ERBE Elektromedizin GmbH
>>>         Dr.rer.nat. Matthias Zenker
>>>         Dipl. Phys.
>>>         Grundlagenentwicklung Technologie
>>>         Waldhoernlestrasse 17
>>>         D-72072 Tuebingen
>>>         Tel.: + 49 (0) 7071 - 755 226
>>>         Fax: + 49 (0) 7071 - 755 5226
>>>         E-Mail: <_mailto:Matthias.Zenker at ERBE-med.de_>
>>>         URL: _http://www.ERBE-med.de_
>>>         
>>> ----------------------------------------------------------------
>>>
>>>
>>>
>>>           	  	 
>>>
>>>         ERBE Elektromedizin GmbH
>>>         Firmensitz: 72072 Tuebingen
>>>         Geschaeftsfuehrer: Christian O. Erbe, Reiner Thede
>>>         Registergericht: Stuttgart HRB 380137
>>>
>>>
>>>
>>>
>>>       	  	 
>>>
>>>     ERBE Elektromedizin GmbH
>>>     Firmensitz: 72072 Tuebingen
>>>     Geschaeftsfuehrer: Christian O. Erbe, Reiner Thede
>>>     Registergericht: Stuttgart HRB 380137
>>>
>>>
>>>
>>>
>>>   	  	 
>>>
>>> ERBE Elektromedizin GmbH
>>> Firmensitz: 72072 Tuebingen
>>> Geschaeftsfuehrer: Christian O. Erbe, Reiner Thede
>>> Registergericht: Stuttgart HRB 380137
>>>
>>>
>>>
>>>
>> ----------------------------------------------------------------------
>>> --
>>>
>>> _______________________________________________
>>> gmsh mailing list
>>> gmsh at geuz.org
>>> http://www.geuz.org/mailman/listinfo/gmsh
>>
>> --
>> Prof. Christophe Geuzaine
>> University of Liege, Electrical Engineering and Computer 
>> Science http://www.montefiore.ulg.ac.be/~geuzaine
>>
> 
> _____________________________________________________________________
> ERBE Elektromedizin GmbH
> Firmensitz: 72072 Tuebingen
> Geschaeftsfuehrer: Christian O. Erbe, Reiner Thede
> Registergericht: Stuttgart HRB 380137
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> gmsh mailing list
> gmsh at geuz.org
> http://www.geuz.org/mailman/listinfo/gmsh


-- 
Prof. Christophe Geuzaine
University of Liege, Electrical Engineering and Computer Science
http://www.montefiore.ulg.ac.be/~geuzaine