Fault modeling using Contact Module

[Alec-YT]

Hi,

I am trying to model a vertical fault in an elastic medium for a circular cavity that is submitted to a biaxial stress.

The rock mass on the left is called left rock, the rock mass on the right is called rightrock and the the line in between is called 'fault' (see the mesh below). I would like to model the fault as a discontinuity with no penetrability and with friction tangential contact.

In moose, I am using the Contact module. However, even if the script converges, I am not sure if it is the right way to proceed as perhaps I should create two lines to model the fault. Could you please let me know? Here is my code: https://github.com/NeohopeNY/Fault.git

Thank you!

[GiudGiud]

Hello

So the fault, is that the equivalent to a fracture problem?
If so we have XFEM capabilities for that purpose

Guilllaume

[GiudGiud]

ok good!
So in contact, can you try the other formulations? Notably with less penalty, frictionless (this might be an issue as system should move?), etc

btw are you doing quasi-static simulations? If so, and the forces are such that 1/2 goes low, and 1/2 of the mesh goes high, what is supposed to stop the evolution of a quasi-static step?

[Alec-YT]

I just tried different formulations/penalties but it doesn't converge.

Yes, computations are quasi-static (using the SolidMechanics module). How to stop the evolution since you're right one block moves up and the other moves down!

[GiudGiud]

that could be the problem. There needs to be a real steady state if you are using quasi static.

change the boundary conditions maybe? To have dirichlet instead of a force/pressure

[Alec-YT]

I have changed the Executioner to Transient (see below). Now it converges (but towards what who knows?? haha). However, it seems that there is a rotation at some point because in Paraview, the orientation of the slip have moved of 90 degrees, which should not be the case. I can also see non linear residuals steps. Maybe I should fix one node or two?
Thank you for your help!

[Executioner]
type = Transient
solve_type = NEWTON # 'PJNFK'
petsc_options = '-snes_ksp_ew'
petsc_options_iname = '-pc_type -snes_linesearch_type -pc_factor_shift_type -pc_factor_shift_amount'
petsc_options_value = 'lu basic NONZERO 1e-15'
line_search = 'none'
automatic_scaling = true
nl_abs_tol = 1.5e-3
nl_rel_tol = 1.5e-3
l_max_its = 40
start_time = 0.0
dt = 0.1
end_time = 1.0
[]

[GiudGiud]

Now it converges (but towards what who knows?? haha

Yeah I dont think that is the solution? Switching to transient for quasi static problems is used for load stepping.

Maybe I should fix one node or two?

yes if it can help fix the system