[CIG-SHORT] sensitivity of Drucker-Prager yielding to time step

Tue Oct 7 21:27:35 PDT 2014

Hi Eric,

It’s hard for me to remember your problem exactly, but if I’m remembering correctly there was localized deformation when you had no fault — is that correct?  If so, the bug I just fixed may have solved that issue.  When comparing the results of the old and new versions for your plastic-nofault problem, I get:

old version:  plastic yield at the first time step, which magnitude varying from 0 to 3.58e-3
new version:  plastic yield at the third time step, with uniform yield throughout and a magnitude of 2.9e-6

Looking briefly at your plastic fault model, we get plastic yield in the second time step, occurring above the left-hand end of the fault and below the right-hand end of the fault.

Do these results seem reasonable?

If so, the fixed version is in the master and maint branches in our git repository.  Alternatively, we should have a minor release within the next week or so.

Cheers,
Charles

On 8/10/2014, at 10:29 am, Eric Lindsey <elindsey at ucsd.edu> wrote:

> Hi Charles,
> 
> We talked briefly about this issue last month at SCEC, and I would still love to find a resolution despite having dropped it for a few months. I've attached the full set of input files so you can take a look yourself. The parameter I've been varying is "dt" in pylithapp.cfg, and earlier I reported the odd results when I vary this in combination with the input file "plastic_nofault_noshear.cfg". Some other input files are also included -- elastic or plastic, with or without a fault, and with or without an initial shear stress applied.
> 
> Let me know if you're able to reproduce my issue, and have any ideas how to fix it.
> 
> Thanks!
> Eric
> 
> 
> 
> On Tue, Jul 22, 2014 at 3:32 PM, Charles Williams <willic3 at gmail.com> wrote:
> Hi Eric,
> 
> I’ve been on vacation for the past month, but I was curious whether you had made any more progress on this problem.  If not, could you send us all of your input files, and we will have a look at it?
> 
> Cheers,
> Charles
> 
> 
> On 12/07/2014, at 1:15 pm, Eric Lindsey <elindsey at ucsd.edu> wrote:
> 
>> Hi Brad,
>> 
>> In all cases (except possibly the dt=1yr case) it's yielding significantly before the yield condition that I calculate. My initial stress state has no shear stress, so it should be plenty far from the yield condition. The initial isotropic compression is -100MPa, and the friction angle is 30 degrees -- for an inscribed Drucker-Prager yield condition, I found that this implies yielding when the shear stress reaches 69.3 MPa. None of the models yielded at that value (i.e. at the first time step that exceeded that level).
>> 
>> I do have one confusion about the initial stress state. I was setting sigma_zz_initial equal to the other two compressional stresses (isotropic compression), but this technically violates the plane strain assumption that sigma_zz = poisson_ratio*(sigma_xx + sigma_yy). Is this a problem? My poisson ratio is 0.25, so I re-ran a few of the models and the calculation of equation 5.101 for the case where sigma_zz_initial=-50MPa, in which case I predicted that yielding should occur at a shear stress of 111.8 MPa. But I get yielding at 40MPa for the dt=50years case, and 41MPa for dt=10years.
>> 
>> Sorry to send so much detail about this issue... your help is much appreciated!
>> 
>> Eric
>> 
>> 
>> On Fri, Jul 11, 2014 at 1:41 PM, Brad Aagaard <baagaard at usgs.gov> wrote:
>> Eric,
>> 
>> How close is your initial stress state to the yield condition? At what time do you expect yielding to occur? Have you computed the yield condition (equation 5.101 on page 88/252 in the PyLith manual) from the PyLith output to see if it matches your expectations of when yielding should occur?
>> 
>> My guess is that your load increment might be too big, so larger time steps push the response much further outside the yield surface. Intuitively I would think this would give a higher shear stress at yielding at larger time steps but that is the opposite of what you see. This is why I suggest going back to the yield criterion and starting from there.
>> 
>> Brad
>> 
>> 
>> 
>> 
>> On 07/10/2014 06:11 PM, Eric Lindsey wrote:
>> Hi all,
>> 
>> I'm getting some strange results with the DruckerPragerPlaneStrain
>> material. In short, I'm finding that the value of shear stress at which
>> yielding/failure of the material occurs depends on my timestep...
>> 
>> I have a 2D box and apply Dirichlet boundary conditions on all 4 sides, and
>> an initial isotropic compression -- including the zz component as required
>> for models with plasticity. There is zero initial displacement, then I
>> apply simple shear to the domain, such that the stresses remain uniform
>> everywhere as shear stress increases.
>> 
>> It turns out that the point at which yielding occurs depends significantly
>> on the time step. Initial compression is -100MPa. The shear modulus
>> (30GPa), displacement rate (1m/yr), and domain size (150km) implies a
>> stressing rate of 0.2MPa/year. Here is a table of some results:
>> 
>> timestep dt (yr)         shear stress at yielding (MPa)
>> 50                               30
>> 10                               36
>> 5                                 58
>> 1                                 76.6
>> 
>> The log file doesn't indicate convergence problems, but could this be a
>> tolerance issue? Or a more subtle solver problem? My .cfg files and a log
>> of the screen output are attached.
>> 
>> Thanks,
>> Eric
>> 
>> 
>> 
>> _______________________________________________
>> CIG-SHORT mailing list
>> CIG-SHORT at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>> 
>> 
>> _______________________________________________
>> CIG-SHORT mailing list
>> CIG-SHORT at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>> 
>> _______________________________________________
>> CIG-SHORT mailing list
>> CIG-SHORT at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-short
> 
> Charles A. Williams
> Scientist
> GNS Science
> 1 Fairway Drive, Avalon
> PO Box 30368
> Lower Hutt  5040
> New Zealand
> ph (office): 0064-4570-4566
> fax (office): 0064-4570-4600
> C.Williams at gns.cri.nz
> 
> 
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-short
> 
> <test_2d_plasticity.tar.gz>_______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-short

Charles A. Williams
Scientist
GNS Science
1 Fairway Drive, Avalon
PO Box 30368
Lower Hutt  5040
New Zealand
ph (office): 0064-4570-4566
fax (office): 0064-4570-4600
C.Williams at gns.cri.nz

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/cig-short/attachments/20141008/ca23739f/attachment.html>