[CIG-SHORT] general mesh optimization

[Brad Aagaard]

Rebecca-

Creating a good quality finite-element mesh for elastodynamics problems 
and complex geometry is still one of the big challenges in doing these 
types of problems. Both LaGriT and CUBIT are sophisticated, complex, 
general mesh generators, so there are many ways to create a mesh for any 
given problem. As a result, there is no way to "force" either of these 
mesh generators to output a mesh that works with PyLith.

Understanding the examples and slowly adding complexity is the method we 
recommend for progressing towards realistic problems. I would start by 
changing the tet4 example, which uses a strike-slip fault, to more 
closely match your geometry (while keeping the fault vertical). Run a 
simulation without any fault interface and then add slip on the fault 
interface. Once you are satisfied that both of these cases are working 
properly, then try changing the dip of the fault. Again, run without the 
fault and then add slip on the fault.

As for an exmaple with a dipping fault, the benchmark problem located in 
the CIG SVN repository at 
http://geodynamics.org/svn/cig/short/3D/PyLith/benchmarks/trunk/quasistatic/reversenog 
contains a reverse fault. The LaGriT scripts are in the meshes 
directory. You need to have subversion to check something out from the 
repository (I have attached the LaGriT scripts for the 1 km resolution 
mesh). To generate the mesh run "lagrit < mesh_tet4_1000m.lagrit".

Some additional tips:

* Always do a quality check before trying to use a mesh and do not run a 
simulation until the element quality is within a reasonable range (see 
my previous email).

* Examine the mesh before running a simulation using any of the various 
tools that can display the meshes from LaGriT. These can be a huge help 
in identifying the regions that have poor quality cells.

Brad

Rebecca Bendick wrote:
> Hi, there.  I've successfully run all the 3d examples provided in the  
> PyLith distribution, and I'm on to building my own problems, but  
> without success.  I took the advice offered in the <embedded fault>  
> thread and, rather than use a complicated input from LaGRIT, tried to  
> modify the mesh from the tet4 example for a thrust problem.  I can  
> build meshes that look fine, but they don't seem to meet the  
> requirements of PyLith.  I've gotten a few different errors,  
> especially ALE:Exception and the one below.  Is there some way to  
> force LaGRIT or whatever mesher is being used to output a mesh that  
> will work in PyLith?  Or, is there a thrust-fault example that I could  
> use as a template to prevent geometry problems?
> 
> Thanks,
> Becks
> 
> 
> screen dump:
> 
>   File "/Users/buildbot/install/powerpc_darwin_single_nosched_binbot/ 
> lib/python2.3/site-packages/pylith/feassemble/feassemble.py", line  
> 416, in initialize
> RuntimeError: Determinant of Jacobian (2.08321e-08) for cell 74 is  
> smaller than minimum permissible value (1e-06)!
> 
> /Users/rbk/pylith/pylith-1.4.2-darwin-powerpc/bin/nemesis: mpirun:  
> exit 1
> /Users/rbk/pylith/pylith-1.4.2-darwin-powerpc/bin/pylith: /Users/rbk/ 
> pylith/pylith-1.4.2-darwin-powerpc/bin/nemesis: exit 1
> 
> 
> 
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
> 

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: mesh_tet4.lagrit
Url: http://geodynamics.org/pipermail/cig-short/attachments/20100120/98945b15/attachment.txt 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: mesh_tet4_1000m.lagrit
Url: http://geodynamics.org/pipermail/cig-short/attachments/20100120/98945b15/attachment-0001.txt