[CIG-SHORT] comment on resolution of benchmark problems

[Charles Williams]

Hi Carl,

I had already done versions of this for the uniform resolution case.   
My mesh sizes are:

1000m:		82944 elements, 15625 nodes
500m:		663552 elements, 117649 nodes
250m:		5308416 elements, 912673 nodes

This seems to correlate exactly with what you have.  I was able to  
run the 1000m and 500m meshes on my Mac (2 GB RAM) using LithoMop (we  
still need to work a few memory kinks out of PyLith), but the final  
one bombed.  I just ordered another 2 GB RAM, so I'll see what  
happens when it arrives.  As far as the 125m option (adaptive or  
uniform), that seems like it's going to be a bit much for people to  
do on a desktop machine.  As far as I'm concerned, I think 125m  
resolution is more than we need.   Instead, I think a more useful  
option would be 250m near the fault that then coarsens away from the  
fault.  We could then see if we can approach the accuracy of the 250m  
uniform mesh, but using far fewer elements.  I'm assuming that's our  
goal, correct?

Let me know what you think.

Thanks,
Charles


On Jul 18, 2006, at 4:04 PM, Carl W. Gable wrote:

> I'd like some feedback on what type of mesh
> size is going to be realistic for the benchmark
> calculations. The definition talks about using
> 125m resolution if possible. Uniform elements
> will require about 42,000,000 elements for that
> resolution.
>
> A back of the envelope calculation shows that
> for the strike slip and reverse fault problem
> you have a fault that is a plane apx. 16km x 16km.
> If you line just the face of the fault with 125m
> spaced nodes, and then also put a plane of nodes
> of the same resolution above and below the fault
> you have:
>
> 16*8*16*8 = 16,384 nodes for a single plane.
>
> For three planes of nodes you have:
> 49152 nodes.
>
> If you connect those up to form tetrahedra
> you will get about 2*6*49152 = 589,826 tets
> just to mesh the fault plane.
>
> So even with an adaptive strategy to only
> refine the fault the number of elements is
> going to hit 1,000,000 very quickly. Is
> this going to be realistic for computations?
>
> Even the regular mesh calculations will have:
> (the factor for number of elements is between 5-6)
>
> 1000m resolution =>    24*24*24*6 =    82944 elements
>  500m resolution =>    48*48*48*6 =   663552 elements
>  250m resolution =>    96*96*96*6 =  5308416 elements
>  125m resolution => 192*192*192*6 = 42467328 elements
>
> Comments?
>
> Carl
>
>
> ----------------------------------
> Dr. Carl W. Gable, Staff Scientist
>
> Voice 505-665-3533 Fax   505-665-8737
> Email gable at lanl.gov
> http://www.ees.lanl.gov/staff/gable
>
> EES-6, MS T003
> Los Alamos National Laboratory
> Los Alamos NM 87545
>
> Hydrology, Geochemistry & Geology Group (EES-6)
> Focus: Geophysics, Hydrology, Mesh Generation
> Correspondence / TSPA
> ----------------------------------
>
>
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>

Charles A. Williams
Dept. of Earth & Environmental Sciences
Science Center, 2C01B
Rensselaer Polytechnic Institute
Troy, NY  12180
Phone:    (518) 276-3369
FAX:        (518) 276-2012
e-mail:    willic3 at rpi.edu


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://geodynamics.org/pipermail/cig-short/attachments/20060718/28bbe7b4/attachment.htm