[CIG-SHORT] Convergence

[Tabrez Ali]

Hello again

After playing with the mesh (to improve element quality) things seem  
to work but the solver takes close to 550 iterations to converge.

I also noticed that without the shallow fault, but with some other  
BCs the solution converges relatively quickly (~200 iterations).

But with the shallow fault, even with zero slip the solution takes  
~550 iterations. Is this because the matrix structure is different b/ 
w the two cases (i.e., with fault but with zero slip vs no fault at  
all) or is something still wrong in my setup?

Thanks for all the help
Tabrez


On Dec 10, 2008, at 12:09 PM, Brad Aagaard wrote:

> Tabrez-
>
> It sounds like there is either a problem in the fault parameters or  
> possibly a
> bug. I am guessing it is an error in a parameter since slip on the  
> deeper
> part seems to work. I would request more diagnostic output about  
> the fault
> parameters (see below) to check to make sure things look consistent  
> with what
> you are trying to do.
>
> You can request more diagnostic output (fields that do into the  
> fault info
> file), such as the fault orientation (normal_dir, strike_dir, dip_dir,
> final_slip, and slip_time) by setting the
> problem.interfaces.fault.output.vertex_info_fields value.
> The default is
> ['normal_dir', 'final_slip_rupture', 'slip_time_rupture']
> See the manual for details. Note that the fault slip is output in  
> the along
> strike, up dip, fault normal coordinate system not the global  
> coordinate
> system.
>
> Brad
>
>
>
> On Wednesday 10 December 2008 8:49:27 am Tabrez Ali wrote:
>> Brad/Charles
>>
>> Thanks for your reply. Yes the residuals do increase after each time
>> step. I have a low dipping fault which curves from very shallow  
>> dip to
>> somewhat steeper dip (as in shallow subduction).
>>
>> If I specify slip on the steeper part then things work fine.  
>> Things also
>> work fine when I dont specify slip on this particular fault but have
>> other faults/BC's in place (on the same mesh).
>>
>> Its slip on the shallow part which seems to mess it up.
>>
>> Also my mesh quality isnt that bad. I do have a few elements with  
>> aspect
>> ratios of 50+ (in cubit) but nothing greater than 100. And my  
>> material
>> property for the fault is the same as used in the examples/ 
>> benchmarks.
>>
>> Regards
>> Tabrez
>>
>> Brad Aagaard wrote:
>>> Tabrez-
>>>
>>> Does the solution blow up (residual increases) or just fail to  
>>> converge
>>> (residual approaches some value)? If the solution blows up,  
>>> something is
>>> wrong. Either the problem is not setup correctly or there is a bug.
>>>
>>> There are a number of things that can cause the solution to  
>>> converge very
>>> slowly. In addition to the mesh quality issue that Charles  
>>> pointed out,
>>> some other things you might look into include:
>>>
>>> (1) Do you get the same behavior with a coarser mesh?
>>>
>>> (2) Does the solution converge when you omit the fault in the  
>>> parameters
>>> (i.e., use the same mesh but don't have any fault interface  
>>> conditions)?
>>> If you have zero displacement BC, try simple compression without  
>>> a fault.
>>>
>>> (3) What is Poisson's ratio? Do any cells have a Poisson ratio  
>>> greater
>>> than 0.45 (only values greater than about 0.48 should cause  
>>> problems)?
>>>
>>> (4) What physical properties are you using for the fault (this  
>>> affects
>>> the conditioning of the system and can affect convergence if they  
>>> are not
>>> reasonable)? In a test problem, the solution converged slightly  
>>> faster
>>> for uniform physical properties for the fault (only used to  
>>> condition the
>>> system) compared with the actual 3-D variation.
>>>
>>> Brad
>>>
>>> On Wednesday 10 December 2008 12:05:31 am Charles Williams wrote:
>>>> I would look at the element quality in the mesh.  You can do  
>>>> this in
>>>> ParaView.  The low angles may be giving you poorly-formed elements,
>>>> and this could cause problems.  I doubt that incorrect fault
>>>> parameters is causing the problem, unless you're doing something
>>>> that's giving you very large strains.
>>>>
>>>> Charles
>>>>
>>>> On Dec 10, 2008, at 6:43 PM, Tabrez Ali wrote:
>>>>> Brad/PyLith Users
>>>>>
>>>>> It seems that whenever I use a low angle thrust fault (very low  
>>>>> dip)
>>>>> the
>>>>> solution (to the quasi-static problem) after a certain number  
>>>>> of time
>>>>> steps fails to converge and blows up. Can wrong fault parameters
>>>>> such as
>>>>> up_dir or normal_dir also cause this?
>>>>>
>>>>> Regards
>>>>> Tabrez
>>>>> _______________________________________________
>>>>> CIG-SHORT mailing list
>>>>> CIG-SHORT at geodynamics.org
>>>>> http://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
>>>> NOTE NEW E-MAIL ADDRESS
>
>