Discussion Forum

I've got a large model which can be divided into an FSI section, and a pure fluid flow section.

To keep the RAM requirements down, I am trying to use a segregated approach, wherein all of the FSI physics are solved in one step (Laminar 1, Solid Mechanics, FSI), and then the remaining pure flow section (Laminar 2) is solved in a second step.

I have been able to do things like using 2 different order Solid Mechanics nodes (Quad for beams, Linear for a big paddle), and can get them to work together by naming the dependent variables the same in both nodes. (u_solid in both) And some luck using multiple components to transfer fluid velocity using a nojac(genext<...>) command in the inlet to the non-FSI component, but I am not sure if that is one way coupling, like a simple open boundary for the FSI side.

Anyone have some prior experience with things like this? The end goal is to have a large (like 5-7M equation) model of a MEMS instrument, which will have a supporting FSI section, a solid mechanics/electrical driving section, and a second smaller FSI section near the driver, where the driving section's solid mechanics are linked to the supporting FSI section's solid mechanics... So a very large problem and resulting matrix.

The approach I am trying to implement is to break the model up into the 3 sections described above, either into discrete components that are solved in an internally-fully-coupled manner, in a larger segregated solver loop, or by having one model where different physics nodes are assigned to the different sections, and are somehow linked together. Is this doable in Comsol? Should this be done as a union with many physics nodes, an assembly with many nodes, or components with extrusion coupling?

Thank you in advance, I know this is a rather loaded and detailed question.