Learning Forum

[yu272]

Hi,

I am doing a system coupling simulation with Transient Structural and Fluent. I always use the fixed time stepping for system coupling (for example, dt = 0.001). Therefore, the system coupling setting overrides both mechanical and fluent settings. 

I often met cases where fluent met negative cell volume, where I needed to decrease dt in system coupling (dt_sc) thereby fluent dt to resolve. I already tried finer mesh resolution in interested region but negative cell volume may happen once FSI interface undergoes quick deformation. 

However, in simulation with fixed dt for system coupling, the mechanical solver can adaptively integrate with smaller dt till dt_sc in system coupling is reached, while fluent can only do dt_fluent = dt_sc. I noticed Fluent has an adaptive time stepping option, but my trial always results in a mismatch between fluent time and system coupling time. 

Can you tell me if I can do a fixed dt for system coupling, while fluent can do an adaptive dt to avoid negative cell volume issues?

Thank you!

Eric

[Rob]

The Fluent timestep is fixed on the System Coupling side, so unfortunately you'll need to reduce the value there. Depending on how complex the bendy bit is you may find the Fluent Intrinsic FSI to be a good option. 

[yu272]

Hi Rob,

Thank you for the information. Unfortunately, my problem requires two-way coupling so I have to do system coupling simulations. 

I have tried to reduce the time step to a very small value and this negative cell volume still happens, while last step all solutions converge well. Other methods I tried included refining mesh in the corresponding area for the initial mesh. Is there any other method you recommend?

Sincerely,

Eric

[Rob]

The Intrinsic FSI is fully coupled, it's just not as an advanced structural solver. 

Refining the mesh tends to need a smaller time step to avoid negative elements as one mechanism for their creation is a point/surface moving more than one cell size in one step. Also check for near wall-wall contact as that will give skew cells. 

[yu272]

Thanks Rob, and yes, my problem involves contact problem as well. A side question relates to smaller time step is that, when I use smaller time step (10^-4 vs. 10^-3), the system coupling becomes harder to converge, even the displacement and forces between source and target looks almost the same. On the other hand, sometimes pretty evident difference between source and target results in convergence. How do you usually avoid convergence issue? I have tried to increase iterations for system coupling and try to use similar mesh size for both FSI interface.

Sincerely,

Eric