Short answer is yes. This is exactly what most automotive OEMs are doing with MSC Nastran acoustics.

If you have a cavity only mesh, the free faces of the fluid mesh are considered rigid. The physical interpretation is a perfectly reflective boundary condition, like a reverberation chamber (which is almost perfectly reflective).

To have a flexible boundary, you need to define the flexibility, and this is typically done by enclosing the fluid cavity in a structural mesh. Where the structure touches the fluid, a coupling matrix is generated automatically and couples the fluid free faces to the structural "wetted" surface. Be careful, the fluid cavity mesh is usually coarser than the structural mesh, and this can lead to regions where either the free faces of the fluid are not sufficiently close to the structure (so no coupling is generated for these regions), or if the coupling search tolerances are too large, you can end up coupling to structure which is physically too far away from the free faces. This may happen where the structure has a double panel, one behind the other, and coupling occurs to the second panel when it should not. You can output the coupling surface as a list of GRID points and elements that have been detected as "coupled" and look at them in a graphics system. Then you can either "tune" the tolerances of the search algorithm or even add to or remove from the GRID point/element list to "manually" define the regions where coupling is to be generated.

DG