MOOSE can't converge when biaxial loading the anisotropic domain with tunnel in middle

[Yulongggggg]

Here's what i am simulating, biaxial loading was implemented to the domain, and each subdomain has different material properties. And there's also a pressure acting on the tunnel wall. When i am doing this, MOOSE can't converge, and here's my code: https://github.com/Yulongggggg/MOOSE_converge

[Yulongggggg]

@GiudGiud

[GiudGiud]

Hello

You need to add a dirichlet boundary condition on the displacements in one or more nodes. Maybe the corner nodes?

It is not converging because there are solutions for displacement that are part of the null space. Like extra rotation and translation solutions. Pinning nodes will remove these

Guillaume

[Yulongggggg]

Dear Giud,

Long time no see, i understand your points and I am trying to simulating the fields of tunnel under geostress. If I add Dirichlet BC it well be unrealistic, and the displacement field is different from the analytical solutions. Only conduct biaxial loading and don't fix the boundary condition can be align with the real situation underground. Is there any way to solve this?

[GiudGiud]

If I add Dirichlet BC it well be unrealistic,

it's the easiest solution though.
There are petsc options you can use to remove the null space, but might not work

Are you sure you cannot set dirichlet BCs?

I think by symmetry, the center of the top and bottom surfaces can be blocked in the X direction , which removes a translation mode and a rotation too
Then same on the center of the left and right surfaces, which should move in Y ? Which would remove 2 more modes