[CIG-SHORT] [EXTERNAL] Applying Gravitational Forces with Pylith

[Brad Aagaard]

On 07/31/2018 11:44 PM, Niloufar Abolfathian wrote:
> Hello,
> 
> I am intending to apply gravitational forces in my quasi-static model. 
> My model includes a fault and two materials, elastic and viscoelastic. 
> In order to get a steady model, I am running the model for more than 
> 20,000 years.Going through the examples I am following what you are 
> doing in the example hex8/step17 and subduction/step8a-b-c. At the 
> moment I am getting my initial results with infinitesimal strain, 
> gravitational forces, and no initial stress.

If the geometry is such that there are shear stresses in the elastic 
material at the boundary between the elastic and viscoelastic layers, 
there will be shear stresses in the viscoelastic layer that cannot 
completely relax. This will result in a steady-state solution with 
nonzero velocities. If the viscous strains are not well resolved, (see 
#3), you may not be converging to the steady-state solution. See 
examples/2d/gravity for examples and the manual for the accompanying 
discussion.

> Since I talked with Brad at SSA, and he was warning about implying 
> gravitational forces, I like to know your suggestion regarding my model:
> 
> 1. I am applying gravitational forces, from the manual you suggest 
> using finite strain rather than infinite strain. Is this always true? I 
> am having an error while running with lgImplicitDeform.

Without the entire error message, it is extremely difficult to help.

> 2. My understanding is that I have to run my model with lgImplicitDeform 
> and gravitational forces until I get steady state, and use the output 
> stress as input for the final round oflgImplicitDeform and gravitational 
> forces and initial stresses. What is your intuition on going through 
> this process?

Conceptually, what you describe is the reasonable approach. I am not 
sure what you are asking. Perhaps your question is answered in the 
manual in the section discussing examples/2d/gravity.

> 
> 3. What about adding few blocks around my model, with larger mesh size? 
> Can they reduce the probable effect (rotation) caused by the 
> gravitational forces?

By rotation, I assume you mean that there is a lateral variation in 
density that causes the model to want to rotate. If the lateral 
variations in density/geometry are near the boundaries, a larger model 
might help (so that the lateral variations do not occur near the 
boundary). A larger mesh size will introduce larger discretization 
errors and will likely make things worse. You should examine the viscous 
strains to analyze how they are decreasing and how well they are resolved.

You may want to start with a model without any horizontal variations in 
density/geometry and iterate on your workflow until you get something 
that works. Then introduce the lateral variations in density/geometry 
one piece at a time so that you can identify at what stage problems arise.

Regards,
Brad