[aspect-devel] convergence failure

Tue Mar 15 13:42:35 PDT 2016

Hi Mohamed,

At the moment, the other non-linear material models we have are drucker 
prager and diffusion dislocation; and which model is best for you 
depends on what you want to do.
The drucker prager model implements plasticity (and there is the 
cookbook "Crustal deformation" in the manual where it is described in 
more detail), while the diffusion dislocation model implements a 
combination of diffusion and dislocation creep.

For the model setup you posted, I can imagine that one point that causes 
problems are the large viscosity contrasts. If you set the solver scheme 
to "IMPES" and the "Initial adaptive refinement" to zero, the Stokes 
solver will converge for this very first iteration, and you can see in 
the output that you have a viscosity contrast of ~13 orders of magnitude 
in the uppermost layer of your model (~10 cells). Such a strong 
viscosity contrast makes it difficult for the Stokes solver to converge. 
I never used the Morency & Doin material model, so I do not know which 
parameters lead to such a strong viscosity contrast, but no matter which 
model you use as a starting point, it will probably help you to find 
problems in your model if you start from a simpler setup with lower 
viscosity contrasts and weaker non-linearity and then increase the 
complexity of your model step by step.

Best,
Juliane

On 03/13/2016 11:31 PM, Jonathan Perry-Houts wrote:
> I haven't worked with nonlinear models much, so I can't recommend a good
> one. I wrote that plugin to reproduce their 2004 paper results, but
> never used it outside of that context. I'm sure someone else on this
> mailing list can be more helpful!
>
> Cheers,
> Jonathan
>
> On 03/12/2016 04:38 PM, Mohamed Gouiza wrote:
>> Hi Jonathan,
>>
>> Thank you for the recommandation.
>> I will try another material model. Do you have any advice on which one
>> to use instead of the Morency and Doin?
>>
>> Regards,
>> Mohamed
>>
>>> On Mar 11, 2016, at 7:03 PM, Jonathan Perry-Houts
>>> <jperryh2 at uoregon.edu <mailto:jperryh2 at uoregon.edu>> wrote:
>>>
>>> Hi Mohamed,
>>>
>>> The Morency Doin material model has always been problematic. (I can
>>> verify that your model also fails to converge on my computer, for what
>>> it's worth.)
>>>
>>> I'll look into it more later, but to be honest my recommendation is to
>>> avoid using that material model. It's a good starting point for building
>>> your own nonlinear material model, but for some reason it's always been
>>> trouble in practice.
>>>
>>> Sorry to not be more help. Maybe after looking more closely at your
>>> model I'll have more suggestions, but at first glance it doesn't look
>>> like you've changed anything too significant from the cookbook.
>>>
>>> -Jonathan
>>>
>>> On 03/11/2016 12:34 AM, Mohamed Gouiza wrote:
>>>> *Hi All,***
>>>>
>>>> * *
>>>>
>>>> *I am trying to run a modified version of the Morency & Doin example.*
>>>>
>>>> *The major modification I made is to the materiel model which now
>>>> consists of a crust (30km thick) and a mantle lithosphere with variable
>>>> thickness (170km thick for 0<x<1000km; 90km thick for 1000<x<2000km). I
>>>> also modified the temperature model accordingly (the temperature
>>>> increase linearly until the base of the lithosphere which is at 1330C;
>>>> temperature in the sublisthospheric mantle is constant at 1330C;
>>>> temperature increase linearly in the lower mantle)*
>>>>
>>>> *The model doesn’t converge on timestep 0 (t=0 years).*
>>>>
>>>> * *
>>>>
>>>> *I increased the Max nonlinear iterations to 100,000, it didn’t work.*
>>>>
>>>> *I changed the solver scheme from iterated IMPES to iterated Stokes, it
>>>> didn’t work.*
>>>>
>>>> * *
>>>>
>>>> *Am I doing something wrong?*
>>>>
>>>> * *
>>>>
>>>> *I’ve attached the parameter file as well (m-cvx-003.prm).*
>>>>
>>>> * *
>>>>
>>>> *Thank you for your help.*
>>>>
>>>> * *
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> ## Hosts assigned to job 404757.1:
>>>>
>>>> ##
>>>>
>>>> ## h3s3b12.arc2.leeds.ac.uk <http://h3s3b12.arc2.leeds.ac.uk> 16 slots
>>>>
>>>> ## h3s3b14.arc2.leeds.ac.uk <http://h3s3b14.arc2.leeds.ac.uk> 16 slots
>>>>
>>>> ## h3s3b15.arc2.leeds.ac.uk <http://h3s3b15.arc2.leeds.ac.uk> 16 slots
>>>>
>>>> ## h6s2b4.arc2.leeds.ac.uk <http://h6s2b4.arc2.leeds.ac.uk> 16 slots
>>>>
>>>> ##
>>>>
>>>> ## Resources granted:
>>>>
>>>> ##
>>>>
>>>> ## h_vmem = 2G (per slot)
>>>>
>>>> ## h_rt   = 48:00:00
>>>>
>>>>
>>>>
>>>> ## nodes/ppn parallel request:
>>>>
>>>> ##
>>>>
>>>> ## nodes = 4     np  = 64
>>>>
>>>> ## ppn   = 16    tpp = 1
>>>>
>>>>
>>>>
>>>> -----------------------------------------------------------------------------
>>>>
>>>> -- This is ASPECT, the Advanced Solver for Problems in Earth's
>>>> ConvecTion.
>>>>
>>>> --     . version 1.4.0-pre
>>>>
>>>> --     . running in DEBUG mode
>>>>
>>>> --     . running with 64 MPI processes
>>>>
>>>> --     . using Trilinos
>>>>
>>>> -----------------------------------------------------------------------------
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> -----------------------------------------------------------------------------
>>>>
>>>> The output directory <output/> provided in the input file appears not to
>>>> exist.
>>>>
>>>> ASPECT will create it for you.
>>>>
>>>> -----------------------------------------------------------------------------
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> Number of active cells: 4,096 (on 6 levels)
>>>>
>>>> Number of degrees of freedom: 87,782 (33,410+4,257+16,705+16,705+16,705)
>>>>
>>>>
>>>>
>>>> *** Timestep 0:  t=0 years
>>>>
>>>>    Solving temperature system... 0 iterations.
>>>>
>>>>    Solving crust system ... 0 iterations.
>>>>
>>>>    Solving lithosphere system ... 0 iterations.
>>>>
>>>>    Rebuilding Stokes preconditioner...
>>>>
>>>>    Solving Stokes system... 0+82 iterations.
>>>>
>>>>    Residual after nonlinear iteration 1: 1
>>>>
>>>>
>>>>
>>>>    Rebuilding Stokes preconditioner...
>>>>
>>>>    Solving Stokes system...
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> ----------------------------------------------------
>>>>
>>>> Exception on processing:
>>>>
>>>>
>>>>
>>>> --------------------------------------------------------
>>>>
>>>> An error occurred in line <760> of file
>>>> </home/ufaserv1_k/earmgo/safe/aspect/source/simulator/solver.cc
>>>> <http://solver.cc>> in function
>>>>
>>>>     double aspect::Simulator<dim>::solve_stokes() [with int dim = 2]
>>>>
>>>> The violated condition was:
>>>>
>>>>     false
>>>>
>>>> The name and call sequence of the exception was:
>>>>
>>>>     ExcMessage (std::string("The iterative Stokes solver " "did not
>>>> converge. It reported the following error:\n\n") + exc.what() + "\n See
>>>> " + parameters.output_directory+"solver_history.txt" + " for convergence
>>>> history.")
>>>>
>>>> Additional Information:
>>>>
>>>> The iterative Stokes solver did not converge. It reported the following
>>>> error:
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> --------------------------------------------------------
>>>>
>>>> An error occurred in line <1216> of file
>>>> </home/ufaserv1_k/earmgo/safe/dealii/include/deal.II/lac/solver_gmres.h>
>>>> in function
>>>>
>>>>     void dealii::SolverFGMRES<VectorType>::solve(const MatrixType&,
>>>> VectorType&, const VectorType&, const PreconditionerType&) [with
>>>> MatrixType = aspect::internal::StokesBlock; PreconditionerType =
>>>> aspect::internal::BlockSchurPreconditioner<dealii::TrilinosWrappers::PreconditionAMG,
>>>> dealii::TrilinosWrappers::PreconditionILU>; VectorType =
>>>> dealii::TrilinosWrappers::MPI::BlockVector]
>>>>
>>>> The violated condition was:
>>>>
>>>>     false
>>>>
>>>> The name and call sequence of the exception was:
>>>>
>>>>     SolverControl::NoConvergence (accumulated_iterations, res)
>>>>
>>>> Additional Information:
>>>>
>>>> Iterative method reported convergence failure in step 37667. The
>>>> residual in the last step was 1.50021e+12.
>>>>
>>>>
>>>>
>>>> This error message can indicate that you have simply not allowed a
>>>> sufficiently large number of iterations for your iterative solver to
>>>> converge. This often happens when you increase the size of your problem.
>>>> In such cases, the last residual will likely still be very small, and
>>>> you can make the error go away by increasing the allowed number of
>>>> iterations when setting up the SolverControl object that determines the
>>>> maximal number of iterations you allow.
>>>>
>>>>
>>>>
>>>> The other situation where this error may occur is when your matrix is
>>>> not invertible (e.g., your matrix has a null-space), or if you try to
>>>> apply the wrong solver to a matrix (e.g., using CG for a matrix that is
>>>> not symmetric or not positive definite). In these cases, the residual in
>>>> the last iteration is likely going to be large.
>>>>
>>>> --------------------------------------------------------
>>>>
>>>>
>>>>
>>>> See output/solver_history.txt for convergence history.
>>>>
>>>> --------------------------------------------------------
>>>>
>>>>
>>>>
>>>> Aborting!
>>>>
>>>> ----------------------------------------------------
>>>>
>>>> --------------------------------------------------------------------------
>>>>
>>>> mpirun noticed that the job aborted, but has no info as to the process
>>>>
>>>> that caused that situation.
>>>>
>>>> --------------------------------------------------------------------------
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> -------------------------------------------------
>>>> Mohamed Gouiza, Research Fellow
>>>> Basin Structure Group, Institute of Applied Geosciences
>>>> University of Leeds, School of Earth and Environment
>>>>
>>>> Leeds,  LS2 9JT, UK
>>>>
>>>> M.Gouiza at leeds.ac.uk <mailto:M.Gouiza at leeds.ac.uk>
>>>> <mailto:M.Gouiza at leeds.ac.uk>
>>>> +44 7985 782073
>>>> -------------------------------------------------
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Aspect-devel mailing list
>>>> Aspect-devel at geodynamics.org <mailto:Aspect-devel at geodynamics.org>
>>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>>>
>>> _______________________________________________
>>> Aspect-devel mailing list
>>> Aspect-devel at geodynamics.org <mailto:Aspect-devel at geodynamics.org>
>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>
>>
>> _______________________________________________
>> Aspect-devel mailing list
>> Aspect-devel at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>
>
>
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20160315/7aabb7e1/attachment-0001.html>