My 2D subduction static stress does not converge

PyLith
1

Hi,
I am using Pylith 2.2.1 and was built from source. I built my 2D subduction model with 6 blocks that have the same material properties because I want a homogeneous medium. The size of the mesh is 500m close to the fault and the surface, and becomes 8000 m in the deeper part. The mesh was created using Cubit 2022.11 and is composed of triangular elements. I impose 2.9 m of slip for a reverse fault and want to visualize the static stress change. My problem is that my model does not converge. I tested this using a different software (defmod) using the preconditioner pc_asm_overlap=2. I am trying to implement the same in Pylith but I have not been able to do it. I know my mesh is dense but I think it is still a simple model that should converge with the asm preconditioner. I attache the files I am using. Can you help me to define the proper solver type, and preconditioner that suits my problem? Thank you very much in advance.
Mesh available here:
https://livememphis-my.sharepoint.com/:u:/r/personal/lbznflrs_memphis_edu/Documents/3_1_SDZ_Homogeneous_6_blocks_cubit_Mw_8_0_500m.exo?csf=1&web=1&e=MYf9RW

SDZ_Ivan.zip (7.5 KB)

2

I can’t access the mesh at the Sharepoint link. Can you post the Cubit mesh in a zip file?

We do not recommend using the ASM preconditioner except for small problems. Table 4.4 In the PyLith v2.2.2 manual provides solver settings that will scale much better with problem size.

3

Dear Brad,

Thank you for your response. I send a new link to the mesh file:
3_1_SDZ_Homogeneous_6_blocks_cubit_Mw_8_0_500m.exo

I compressed the file but it was still to big to upload.
I also tried to use other solver and precondition but it was not clear to me how to define them in the script. I hope you can download the mesh and suggest a setting for my problem. Thank you for your support!

4

I do not understand why you want such a fine resolution mesh over the entire crust and slab. The mesh should be fine where you will have gradients in the strain field (for example, close to the fault). If you need a cell size of about 500 m along the fault to resolve the static stress changes, then coarsen it with distance from the fault at a rate of about 1.05 to 1.1. That is each cell gets 5-10% larger with distance from the fault. This is the bias in CUBIT cell sizing functions.

The solver settings for the custom preconditioner corresponding to those in Table 4.4 in the PyLith 2.2.2 manual are in
share/settings/solver_fault_fieldsplit.cfg.

5

Thank you for your response.

I am interested in the stress changes in the crust. I agree that my mesh is too dense right now. I am new creating meshes but my project focuses in the crust. I actually might decrease the size of elements a little more but I need to implement the biases toward the boundaries. I will also implement other solver and preconditioners in my models. Thank you for your help.