[Zymon]

Hi,

I need your help. I am dealing with this problem for a couple of weeks.

I am simulating a membrane pump. A pressure plate presses a membrane into the fluid chamber. Due to the reduction in volume the blood is pressed out of the chamber. The membrane has a low Young's Modulus (about 5 MPa) and its density is almost the same as the one of blood. So the Membrane is very flexible, and here is the main problem of the FSI simulation, I guess.

With a hexahedral SOLID185 mesh, frictionless Augmented Lagrange Contact Conditions (FKN: 0.1, agressive Update each iteration, Auto Detection Value of Pinball Region) and a direct solver the uncoupled FEM simulation (ANSYS transient structural, dummy pressure at the FS-interface) converges well. Coupling it with CFX it diverges however in the first time step. Specifically in the FEM part of the second coupling iteration. I added a plot of the convergence history.

So, when I use a much higher Young's Modulus and density of the membrane the FSI simulation converges (untill the point the simulation crashs because of unphysical material behaviour). Using a tetrahedral FEM mesh (also SOLID185), pure penalty algorithm and lower FKN value the FSI simulation runs for several time steps even with not so high values (the standalone FEM case has problems after some time steps with this configuration). However, using a lower time step the simulation diverges again.

I expected an artificial added mass instability because of the density ratio and the general behaviour. So I tried several things:
- very low underrelaxation factors
- source term at the interface with a very high mass flux pressure coefficient (up to 1e37). This lowers the force at the interface (can be seen in a monitor) but has no effect on convergence
- double precision
- higher time steps
- and so on. All the tipps I could find in this forum.

But it does not help. The convergence history always looks almost the same and the simulation diverges in the first time step. Has anybody experience the same problem? Are there any other parameters I could apply?