Nested Mesh Interaction

[klaw13]

Hi there,

I'm trying to simulate the flow of water inside a cavity flowing inside the red volume. As the water flows throughout the cavity, the pressure from the flow should be able to push on the mesh boundary to move it. If the force is large
enough, both the outer and inner volumes should become cylinders.

input file: input.i.txt )

I initially tried to just simulate the flow of the water inside the cavity without considering any deformations to the mesh. When I run the simulation, I can get the water to flow about half way through, but then the simulation fails to converge.

I thought it might be because I need to take into consideration the effects on the mesh, so I tried to base my approach on the 3D_struc_acoustic.i example, and added in kernels to handle the convected mesh.

However, I have not been able to converge at any timestep. Currently, I'm getting the following error:

*** ERROR *** MooseVariableFE: Old time derivative of solution (u_dot_old) is not stored. Please set uDotOldRequested() to true in FEProblemBase before requesting u_dot_old.

[GiudGiud]

Hello

The effect on the mesh should be a sufficiently different issue that you should try to get a resolved solution without modeling this effect first.

Once you get a good solution without the fluid structure interaction, @somu15 should be able to help with that error message, which I assume is related to a FSI kernel

Best,
Guillaume

[klaw13]

@somu15 Yes, I've used the mesh generator and specified the primary side as the fluid and solid as secondary. Thanks for looking into this!

[somu15]

Sure. Would it be possible to share your recent input file and cubit exodus file if any? Thanks.

[klaw13]

Sure I've attached my current input file and output of the mesh

[mesh_output.zip](https://github.com/idaholab/moose/files/7287168/mesh_output.zip
input.i.txt
)

[somu15]

@klaw13 I am able to run the transient simulation. The biggest change is using the NewmarkBeta time integrator instead of Heun. I don't know why we can't work with Heun as I have never used this before. I am attaching the input file that I have used. I am also attaching an image showing that the simulation is running. A couple of suggestions:

1. You can use the DynamicTensorMechanics instead of setting up the ADStressDivergenceTensors kernels manually. With this, you don't have to provide as much information to the InertialForce kernel as you are doing previously.
2. Try using ComputeSmallStrain in the Materials block if this is justifiable. Then, you can use Newton in the solve_type. Right now, the solves are not converging well but that is a different issue. :)

input copy.i.txt
.

[klaw13]

Thanks @somu15! Making those changes works. That is quite strange it doesn't work with the Heun time integrator.

[fdkong]

It is a fascinating application. We currently do not have a 3D example in FSI module, but we have a 2D model, "fsi/test/tests/fsi_2d", with which you could start.

The example is simple, but it is pretty general. Most of the components, internals, and kernels, should be ready to use for your applications.

Also, are there any specific requirements of how the mesh should be defined (more specifically at the interface) for FSI to work correctly?

We require solid and fluid meshes to match each other on the interface, even though that can be lifted in the future.

[fdkong]

I meant the exodus file that was generated by using "--mesh-only"

[klaw13]

mesh_exodus.zip

[fdkong]

Does gmsh has an equivalent command-line of "merge volume solid fluid" of cubit?

The interface is not merged between fluid and solid. That can be done via "merge volume solid fluid" in cubit. Could you do the same thing in gmsh?

[GiudGiud]

@klaw13 any updates on this?

[klaw13]

@fdkong Yes, fortunately I was able to figure out how to merge the volumes together directly in gmsh. The simulation runs now with the extended fsi_3d version. I'm still working on applying this setup to my custom application and will follow up if I continue getting errors. Thank you!