Learning Forum

[Marcin Nowak]

Hello, 

I am trying to solve FSI problem, using System Coupling, Fluent and Mechanical. I succesfully solved many cases like this, however, now I am encountering a problem, as I am trying to solve with the option "Large deflection: off" in Mechanical. The reason why I apply such setting in this case, is to compare the result with the analytical model considering small deflection.

The simulation is a simple flow through a flexible pipe. It runs succesfully for the option "Large deflection: on", however, when I set it as "off", I encounter convergence problem, shown at the pictures below. 

Simply speaking, the simulation runs fine and with convergence, up to a certain time point, where it diverges, no matter what settings I try to apply (as for dynamic mesh options, nr of fluent iterations, Scale factor of fluid-solid interface, under-relaxation factors for data transfers, time step size...). I can elaborate on the settings that I try to tune if needed.

The simulation diverges at the below deformation state (scaled 20 times):

 

To repeat - the large deflection FSI works fine, the separate solvers works fine. The mechanical-only solver for "large deflection=off" works fine.
At this point I am wondering, whether this non-convergence for deactivated large deflection, is done on purpose in Ansys software for FSI. In all the Ansys materials, guides, tutorials, I found only information "Large deflection - Should always be set to On" without elaboration.

Please give me some remarks, what could I do in this scenario

Kind regards

[Rahul]

Hello 
Large deflection should usually be set to On.

When set to Off the structural mesh position is not updated during the simulation, so System Coupling loads are applied to the original mesh.

If it is transient consider reducing the time step first for transient cases.

Consider the highest frequency, f, of interest on FEA side, and then dt = 1/20f and Consider what a suitable time-step is for good CFD convergence, or to capture shedding freq. Choose the minimum of the two timesteps.

Also, see if quasi newton stabilization helps. Quasi-Newton Solution Stabilization (ansys.com)

[Rahul]

Also refer Improving the Accuracy of Transient Analyses (ansys.com) and Improving Coupled Analysis Stability (ansys.com).