[aspect-devel] Morency-Dorin with free surface

[John Naliboff]

Hi Payman,

Intrinsically, any highly non-linear rheology can easily produce solver 
issues and unexpectedly large velocities.  However, maybe there is 
something about Morency-Doin that is particularly unstable?

Perhaps give the visco-plastic model a shot and if the issues still 
persist then a closer look at the initial conditions is warranted. As a 
guide, the continental extension cookbook I sent around via email uses 
the visco-plastic material model.

Also, for the iterated Stokes solver, what is your convergence limit and 
how many non-linear iterations does it take to reach this value?

Cheers,
John

*************************************************
John Naliboff
Assistant Project Scientist, CIG
Earth & Planetary Sciences Dept., UC Davis

On 10/23/2016 11:06 AM, Mohamed Gouiza wrote:
> Hi Payman
>
> The visco plastic material model is no the same as the Druker model.  
> The visco plastic material model does allow for multiple compositional 
> fields.
> You should look it up in the ASPECT manual.
>
> Cheers
> Mohamed
>
> ---- Payman Janbakhsh wrote ----
>
> If you are referring to druker-prager material model, yes. But I would 
> need multiple compositions. My understanding is druker model only 
> takes one composition.
>
> payman
>
> -----Original Message-----
> From: Aspect-devel [mailto:aspect-devel-bounces at geodynamics.org] On 
> Behalf Of Jonathan Perry-Houts
> Sent: October 22, 2016 10:31 PM
> To: aspect-devel at geodynamics.org
> Subject: Re: [aspect-devel] Morency-Dorin with free surface
>
> Have you looked at the "visco plastic" material model?
>
> -JPH
>
> On 10/22/2016 6:27 AM, Payman Janbakhsh wrote:
> > Thanks Jonathan,
> > I have tried no-slip as well as tangential side boundary velocities 
> but not much difference. I always keep surface vertical projection set 
> to vertical when choosing top free surface. I even increased 
> temperature polynomial degree to 3 to increase the accuracy.
> > Then I set thermal expansivity to zero to simplify the problem a 
> bit. so it kept going for few time steps then froze again trying to 
> solve Stokes systems.
> > I have noticed this happens when RMS velocities starts going up 
> drastically . I have to figure out why velocities become so unphysical.
> > The whole purpose of using this material model was to incorporate a 
> visco-plastic rheology to 3D problems. So need to find my way around 
> to get there.
> > Any help is much appreciated from anyone.
> >
> > Cheers
> >
> > Payman
> >
> > -----Original Message-----
> > From: Aspect-devel [mailto:aspect-devel-bounces at geodynamics.org] On
> > Behalf Of Jonathan Perry-Houts
> > Sent: October 21, 2016 8:33 PM
> > To: aspect-devel at geodynamics.org
> > Subject: Re: [aspect-devel] Morency-Dorin with free surface
> >
> > Hi Payman,
> >
> > I haven't tried the Morency-Doin with a free surface, but it's 
> caused me trouble in the past in relatively simple models with similar 
> setups to the original Morency & Doin paper on which it's based. -- 
> Point being that the material model is finnicky, and I'm not surprised 
> it's having trouble.
> >
> > A few thoughts:
> >
> > Maybe no-slip on the sides is not a good choice, since the top is 
> unconstrained? This shouldn't be a problem, but who knows. Maybe worth 
> playing with?
> >
> > Using:
> >>  subsection Free surface
> >>    set Surface velocity projection = vertical  end
> > sometimes helps stabilize free surfaces. Failure on the first time 
> step makes this an unlikely cause for your problem, but maybe worth a try.
> >
> > You mentioned that your model is 2D, so this isn't a problem for now,
> > but keep in mind that the Morency-Doin model shouldn't be applied to
> > 3D models. The formation of the second invariant of the strain rate
> > tensor in the paper has never made sense to me, and so it's not
> > generalized to three dimensions in the code (which reminds me, I
> > should add an assertion that makes sure this is the case. Not sure how
> > I missed that
> > before!) -- Any insight from someone else to what the authors meant by
> > equation (4) in the paper (Morency and Doin, “Numerical Simulations of
> > the Mantle Lithosphere Delamination.”) would be helpful! (e.g. is it a
> > weird index notation I'm not familiar with?)
> >
> > I think that equation, as implemented in the code, is almost 
> certainly wrong. I included it for consistency with the paper after 
> discussions with Timo and Wolfgang like two years ago. Perhaps try 
> replacing it with the actual second invariant and try again?
> > (
> > https://github.com/geodynamics/aspect/blob/master/source/material_mode
> > l/morency_doin.cc#L104
> > )
> >
> > Beyond that I don't know what to tell you. Your parameters sound 
> reasonable from what you described. Keep us updated on what works! I'd 
> love to get that material model working for more than just a cookbook!
> >
> > Cheers,
> > JPH
> >
> > On 10/21/2016 01:48 PM, Payman Janbakhsh wrote:
> >> Hi John
> >>
> >> 2D , delamination, geometry aspect ration of 2x1 , free top surface,
> >> other boundaries zero velocity.
> >>
> >> Linear temperature profile within the upper 100km and lower 100km and
> >> a constant T in between.
> >>
> >> 25km upper crust, 30km lower crust, a weak zone within the lower
> >> crust( with a lower activation energy compare to the enclosing lower
> >> crust, so can set off the delamination).
> >>
> >> Have tried different reasonable densities and activation energies for
> >> every layer except mantle.
> >>
> >> Also have put Boundary composition model as : Initial composition  .
> >> not sure if this is correct.
> >>
> >> Running with IMPES or iterated Stokes made no difference
> >>
> >> Also velocity polynomial set to 2 and other polynomials set to 1
> >>
> >> also have turned off Local conservative discretization
> >>
> >>
> >>
> >> cheers,
> >>
> >>
> >>
> >> payman
> >>
> >>
> >>
> >> *From:*Aspect-devel [mailto:aspect-devel-bounces at geodynamics.org] *On
> >> Behalf Of *John Naliboff
> >> *Sent:* October 21, 2016 3:36 PM
> >> *To:* aspect-devel at geodynamics.org
> >> *Subject:* Re: [aspect-devel] Morency-Dorin with free surface
> >>
> >>
> >>
> >> Hi Payman,
> >>
> >> The Morency-Dorin material model should not have intrinsic problem
> >> with handling a free surface.  The issue is likely related to some
> >> other aspect of your model design.  What type of processes are you 
> modeling?
> >>
> >> Cheers,
> >> John
> >>
> >>
> >> *************************************************
> >>
> >> John Naliboff
> >>
> >> Assistant Project Scientist, CIG
> >>
> >> Earth & Planetary Sciences Dept., UC Davis
> >>
> >> On 10/21/2016 12:17 PM, Payman Janbakhsh wrote:
> >>
> >>     Hi
> >>
> >>     Does Morency-Dorin material model have difficulty with free 
> surface?
> >>
> >>     My models get stuck at solving Stokes system even at step zero when
> >>     I switch top boundary to free surface.
> >>
> >>     I have set CFL to 0.1 and theta to 0.7 even chanh3d max time 
> step to
> >>     1000 but still same issue. I have all other boundaries set to no
> >>     slip boundaries
> >>
> >>
> >>
> >>     Thanks
> >>
> >>     Payman
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>     Sent from my Samsung Galaxy smartphone.
> >>
> >>
> >>
> >>
> >>     _______________________________________________
> >>
> >>     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
> > _______________________________________________
> > 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
> _______________________________________________
> 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/20161024/16e1fcde/attachment-0001.html>