On Implicit Coupling in Fluent Adapters

[tiandaoAmor]

I am currently developing an adapter for fluent, and have successfully completed some flow field and flow field coupling examples and a Fluid–structure interaction example. Currently, it is possible to transfer data effectively, but I am currently facing a problem. As I am not very familiar with the implementation of CFD, I am not sure which physical field information needs to be reset to achieve implicit coupling in the flow field? I roughly understand that there are at least velocity and pressure fields, but I am not very familiar with other physical fields. For example, when opening the energy equation, do we still need to add temperature and density fields?

If possible, could you please let me know which physical quantities of the fields I should recover in the implicit iteration.

[mtree22]

First, to be overly explicit, the fields with which the coupling is concerned are only at the interface between the fluid solver and the structural solver when you're considering an FSI analysis. The fluid solver and the structural solver being coupled together should properly handle the propagation of any changes in information at their shared boundary throughout the remaining respective volumetric domains. The question of what information needs to be coupled at the interface depends on what information you are trying to model within the volumetric fluid and / or volumetric solid, so it's a bit nuanced. The most basic difference will be whether you are modeling the fluid as an incompressible or a compressible flow. If incompressible, you do not consider any temporal or spatial changes in density or internal energy (temperature). This leaves requirements for the forces, velocity, and position of the interface. Position and velocity are obviously related, so one may not need to explicitly couple both values at the interface, but both values should be reconciled somehow. The fluid velocity must match the structures velocity at the interface, but the structure may have deformed because of the fluid forces applied to it and this deformation should be manifest in a change in the fluid domain boundary location as it will have an effect on the surrounding fluid velocities and pressures.

If you're using a compressible fluid solver it may make sense to also couple temperature at the boundary. An additional physical restraint definitely exists that states the temperatures should be the same at the interface, but the structural solver would have to model thermal effects in order for this to make sense anyway. Density does not make sense to couple as the density of the fluid and the structure need to be matched at the interface. In fact, the closer the fluid and structural density are to one another the more difficult an implicit coupling becomes, due to something called the "added mass effect".

I'm excited to hear that you're developing a fluent adapter, as the customer I had driving my development of one has pivoted their request from us, so I've stopped working on it. Please take a look at https://github.com/precice/fluent-adapter/issues/1 and https://github.com/precice/fluent-adapter/tree/fluent_21_2_0 to see if its best for you to pick up from that point or if it makes the most sense for you to continue your own path. Any contribution you can make is much appreciated.