Nonlinear convergence problem with static friction fault

Hi Brad,
I want to know the strike and dip of vertices on fault. And I obtain the *_info.h5 file about these informations. But I can’t figure out the values, for example, there is array with 339 rows and 3 columns in the ‘/vertex_fields/strike_dir’. The 339 rows are corresponding to the number of all vertices. Do the 3 columns represent the x, y, z directions? And these values in the range of -1~1 don’t looks like angles.

The strike_dir vector is the direction of the along-strike vector in global coordinates. The components will be in the range [-1, 1]. If the fault strike is aligned with the x-coordinate direction the strike_dir will be (1, 0, 0). If the fault strike is aligned with the y-coordinate direction, the strike_dir will be (0, 1, 0). The dip_dir is the direction of the dip vector in global coordinates. A vertical fault with have an dip_dir of either (0, 0, 1) or (0, 0, -1), depending on the direction of the fault normal. You can compute the fault strike and dip angles from these vectors.

Thanks,I’ll try it.

Hi Brad,
I add three fault in my 3d model using planar fault for simplicity (the first picture). And the simulation with the single gyf fault can run to completion when I set the friction coefficient to 0 and don’t specify the buried edge. Whereas, the simulation with single klf or lmsf fault and no specified buried edge can’t run, the nonlinear residual norm can’t converaged and become larger and larger. Why does the planar fault also have the kind of problem? I try to reduce the time step and smooth the fault surfaces according to your advice, but it doesn’t work.

I figure the slip rate of the gyf fault along strike. Then I find the location of fault nodes changes after run the simulation (the second picture). Then, I try the simulation with specified buried edge, the node location doesn’t change. Is this the problem with the buried edge? The slip on the fault will decrease if I specify buried edge, so I don’t want to specify it. How to fix it?


From your diagram, I am unable to determine the geometry for each fault. Do the faults klf and gyf cut through the entire domain (all edges are on the boundaries of the domain)? If so, then this is probably the source of the problem.

Are you sure this is a reasonable boundary value problem to solve to answer your science question? Normally, we solve boundary value problems with the fault embedded in the middle of the domain to avoid undesirable affects from truncating the boundary.

Hi Brad,
The klf and gyf do cut through the entire domain. and I don’t specify the buried edges in *.cfg file. The fault edges are on the the boundaries of the domain, and the nodes of faults are removed from the nodes of Dirichlet boundary conditions. I also increase the domain aera, but it doesn’t work.

What do you mean by “I also increase the domain but it doesn’t work”? Did you increase the domain while keeping the fault the same size and moving all boundaries several fault lengths away (you will need to add buried edges)? What specifically doesn’t work?

Sorry, I didn’t make myself clear. I just increase the distance between the faults and all boundaries about a lmsf length away, and extend the fault lengths to cut through the domain. It doesn’t work means that the same error occurs, namely that the nonlinear residual norm can’t converaged.

I don’t want to specify the buried edge. because it will reduce the slip on the fault. I tested the example in “3d/hex8/step12.cfg” using the 0 friction coefficient (figure a). As shown in figure (b), the velocity bc is imposed on -x and +x faces with 1 cm/a (along the Y direction). If the buried edges aren’t specified, the slip rate on fault is 2 cm/a along the Y direction, which is reliable. When the buried edges are specified, the slip rate will decrease shown in figure (c).

I slove my science question by referring to the Hergert and Heidbach (2010). They obtain the velocity bc from the GPS observation and use a small domain. If I increase the domain area, then physical properties of the increased area (e.g. elastic parameters ,viscosity or other faults in the increased area) between the boundaries and my focus area (near fault) may influence the slip on the fault and the modeled GPS velocity.

Thanks, Brad.