[CIG-SHORT] Fault strike not parallel with any axis gives small Jacobian error

[Carl W. Gable]

The attached image shows where the high aspect ratio elements occur.

Part of the problem occurs because the points used to distribute points 
start as a regular X,Y grid of points,

createpts / xyz / nx,ny,1 / &
   domain_xm  domain_ym  domain_zp / &
   domain_xp  domain_yp  domain_zp / &
   1,1,0 /

However, fault2 is no aligned with the coordinate axes:
surface / surf_fault2 / intrface / plane / &
  0.0 -3500.0 0.0 /&
  3000.0 -3510.0 0.0 /&
  0.0 -3600.0 -300.0/

The code is adding points on the interface to insure that elements do 
not cross the material interface you defined, but as Brad points out, it 
results in high aspect ratio elements.

Suggestions would be to:
a) Distribute points on the fault2 surface
or
b) Distribute points aligned with fault2 (regnpts distributed along a 
vector that is aligned with fault2)
or
c) Shoot points in the Y direction to put them on the fault2 surface.

Carl

On 10/10/2010 05:46 PM, Brad Aagaard wrote:
> Birendra-
>
> You have a bunch of nearly zero volume cells in your mesh. Note the
> number of cells with nearly zero aspect ratios listed as a result of the
> quality command when running LaGriT. This type of issue won't show up
> when you align the faults with the edges of the domain have have nice
> clean boxes with perpendicular sides. When you deviate from box type
> geometry, some cleanup of the mesh is usually necessary using LaGrit's
> mesh quality improvement routines (such as massage and reconnect). See
> the LaGriT documentation for details.
>
> For nice simple geometry like you have, you should be able to get a mesh
> with a minimum aspect ratio of close to 0.4. Anything less than about
> 0.2 or 0.3 will result in slow convergence in quasi-static problems.
>
> Brad
>
>
> On 10/9/10 9:02 PM, Birendra jha wrote:
>> Dear Brad,
>>
>> I am trying to run the attached case and get RuntimeError:
>> Determinant of Jacobian (3.73755e-07) for cell 11540 is smaller than
>> minimum permissible value (1e-06)!
>>
>> When the faults strike parallel to y axis it runs fine.
>>
>> Another question: I am trying to find the source code that computes
>> Jacobian matrix, for examples, for quasi-static problem (Sec 2.3 in
>> Pylith manual). In particular I want to modify the linear elasticity
>> Eq. (2.70) to include fluid pressure. Can you please help me in
>> locating the file(s) that implements this?
>>
>> Thanks&   regards Birendra
>>
>>
>>
>
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short


-- 

\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
Dr. Carl W. Gable, Staff Scientist

Team Leader, Subsurface Flow and Transport Team
EES-16, Computational Earth Science
MS T003, Los Alamos National Laboratory, Los Alamos NM 87545
Email:gable at lanl.gov   Voice:505-665-3533   Fax:505-665-8737
Schedule B
Homepage:                         http://www.ees.lanl.gov/staff/gable
LaGriT Mesh Generation Software:  http://lagrit.lanl.gov
Mesh Generation, Geological Apps: http://meshing.lanl.gov

Focus:Geophysics, Hydrology, Mesh Generation, Computational Geometry

/\/\/\/\/\/\/\/\/\/\/Correspondence / TSPA\/\/\/\/\/\/\/\/\/\/\/\/\/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: aspect.png
Type: image/png
Size: 76214 bytes
Desc: not available
Url : http://geodynamics.org/pipermail/cig-short/attachments/20101012/49a24df2/attachment-0001.png