[CIG-LONG] Crash Error in TimeIntegrand

[María Helga Guðmundsdóttir]

Hi Walter,

The changes are geometrical, in the aspect ratio of the bend, and/or the ratio of bend width to crustal thickness. We have the bend width fixed to 40 km and are testing crustal thicknesses (D) of 5, 10, and 160 km and bend lengths (L) of 40 and ca. 148 km. The models have fared as follows:

D 160, L 148: completes 70 km of disp. in 100 timesteps.
D 10, L 148: completes about 60 km of disp. in 310 timesteps.

D 5, L 148: crashes after 17 km, 89 timesteps.
D 160, L 40: crashes after 37 km, 288 timesteps.
D 10, L 40: crashes after 10 km, 10 timesteps.
D 5, L 40: crashes after 4 km, 7 timesteps.

I suspect that if the two models that "succeeded" were run long enough, they would run into the same mesh distortion problem – the time that takes to crop up varies hugely among the models that fail before we stop them.

We also have three other sets of models with the same geometries, but different material properties. At least one of those has already crashed in a similar fashion.

Please let me know what files you would like to see, and I will get them to you ASAP.

Thanks for your help –
Maria



María Helga Guðmundsdóttir
MS Student, Stanford University
Geological and Environmental Sciences
mariahg at stanford.edu

On May 7, 2012, at 10:52 PM, Walter Landry wrote:

> María Helga Guðmundsdóttir <mariahg at stanford.edu> wrote:
>> Hi Walter,
>> 
>> I've been running the restraining bend models that George Hilley has
>> been troubleshooting with you lately, with mixed results. Two of
>> them have run fine upto the amount of displacement we want, but
>> another four models have crashed in an identical fashion at
>> completely different points during the run (timesteps 8, 11, 90, and
>> 289 respectively, with total displacements of 4, 10, 17, and
>> 37km).
> 
> The problem is the mesh distortion.  I am attaching a picture taken
> from the inside, and you can see that the height of the mesh almost
> goes to zero.  I am not sure what to do about this. What do you change
> between the runs that work and the runs that do not?
> 
> Cheers,
> Walter Landry
> 
> <Distorted.png>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://geodynamics.org/pipermail/cig-long/attachments/20120508/c21151e1/attachment.htm