[long.1059]

I am working on a 3-D simulation of a wave tank for my senior capstone. We are testing how effectively swim lane dividers dampen waves. My setup is a 6 ft x 2 ft x 2 ft tank with a water depth of 1 foot. There are 4 small cylinders to represent the lane divider discs position in the center of the tank, half submerged. The waves are generated by a moving wall at one end with a dynamic mesh zone governed by a sinusoidal UDF.

I am using a transient, pressure based, multiphase, implicit VOF scheme.
My dynamic mesh is using spring based smoothing and remeshing.
I am using Non-Iterative Time Advancement
The volume fraction scheme is Modified HRIC (worked in 2-D trials)
BCGSTAB Stabilization Method
Time step size = 0.001s, for maximum of 4000 iterations (4 seconds)

I created the mesh using the patch conforming tetrahedron method with a refinement around the cylinders to help resolve the turbulence. I don't believe the mesh is exactly ideal. It has around 7 million elements. Even with a 16 core Linux server to run the simulation on, it is almost prohibitive.

Anyways, when I run the simulation, the interface between the water and air quickly diverges, after about 30-40 iterations, even the areas far away from the moving wall, where I would expect to see divergence. I have tried a few different values for the relaxation factors and the smoothing and remeshing options. Can anyone offer suggestions on how to keep the solution from diverging? Is it a meshing issue? If so, how can I make a higher quality mesh that is not too resource-intensive?

This is my first time posting. Please let me know if I can provide more details.

The first picture is the tank at the start: I patched half the volume to be water:
vof-0005.jpg

The second picture is as the divergence begins to escalate:
vof-0055.jpg

The third picture is the last one which saved before the calculation stopped as a result of divergence of the AMG solver:
vof-0060.jpg

The last picture is what it should look like after all 4000 iterations. This is with a structured, hexahedral mesh. There is no cylinder in the tank, which is why the mesh was so easy:
vof-4000.jpg

Here is a picture of the mesh, with a section plane at water level:
mesh.jpg