[CIG-LONG] Gale stress tensor verse pressure output in .dat and vtk files

[Louis Moresi]

I think that it would be better if Gale and Underworld internally dealt with
the total stress or deviatoric stress plus pressure and not the slightly odd
mixture of the two which is something we compute for numerical convenience.
I made this point to the developers and I hope that will filter through.

Stress boundary conditions can be slightly tricky in the case where the
viscosity is varying within the element, I believe, because there are some
assumptions in the way that the integrals get put together which may
introduce errors. However, that is something which we test through
benchmarking, so I am also surprised to hear that the code is producing
nonsense and would like to know more details.

L.



On 4 August 2010 08:33, John Naliboff <jbnaliboff at ucdavis.edu> wrote:

> Hi Walter,
>
> Yep, the stress values make a lot more sense now that I've added the
> pressure back in.
>
> So in general, would it be reasonable to say that the sum of the deviatoric
> stress values is a good indicator how large the compressibility term is
> (i.e. since the deviatoric stress tensor trace should sum to zero)?  I
> checked and in some models this non-zero value is quite small (>4 orders of
> magnitude smaller than non-hydrostatic stresses), while in others it is
> larger (1 or 2 orders of magnitude smaller than non-hydrostatic stresses).
>
> Also, what are some of the analytical solutions you tested for the side
> stress boundary conditions?  The few tests I've tried behaved reasonably
> well, but I'd definitely like to also try out some of the scenarios that
> don't give good matches to the analytical solutions.
>
> Thanks again,
> John
>
>
> On Jul 31, 2010, at 12:07 PM, Walter Landry wrote:
>
> > John Naliboff <jbnaliboff at ucdavis.edu> wrote:
> >> Hi Walter,
> >>
> >> I have a number of questions regarding the Gale stress tensor and
> pressure output values in the .dat and vtk files.
> >>
> >> Specifically, I've been doing tests with side stress boundary conditions
> >
> > I have never gotten side stress boundary conditions to work well.  I
> > always had problems with the numerical solution being far too
> > different from the analytic solution.  However, with a HydrostaticTerm
> > component, the hydrostatic term is already subtracted out, so that
> > might work.
> >
> >> and in doing so I noticed that there seems to be a discrepancy
> >> between the stress tensor values and the magnitude/value of the
> >> pressure.
> >
> >> My interpretation of the pressure is (sigma_xx + sigma_yy)/2 in 2D
> >> or (sigma_xx + sigma_yy + sigma_zz)/3 in 3D.
> >
> > The stress that is output is mostly the deviatoric stress.  Because of
> > the numerical compressibility term, it will not be completely
> > deviatoric.  To get the true pressure you would then add the trace of
> > the stress to the pressure-like variable.
> >
> > In practice, if the compressibility term gets significant, then you no
> > longer have something that is close to incompressible.  So you would
> > need to throw out the whole solution anyway.
> >
> > Cheers,
> > Walter Landry
> > walter at geodynamics.org
>
> _______________________________________________
> CIG-LONG mailing list
> CIG-LONG at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-long
>
------------------------------

<http://picasaweb.google.com.au/lh/photo/UsGfMSzQE-6tpUBaW4klK6O6j9vpN3ZPghWjdy2iDOg?feat=embedwebsite>
Professor
LouisMoresi
 louis.moresi at monash.edu
Google Profile <http://www.google.com/profiles/louis.moresi> | Google
Calendar
<http://www.google.com/calendar/embed?src=louis.moresi%40gmail.com>Underworld
Geodynamics Software <http://www.underworldproject.org>
Mobile phone: +61 4 2850 1907

  ------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://geodynamics.org/pipermail/cig-long/attachments/20100804/f424ec52/attachment.htm