>>11279465Cont.
The idea is basically to introduce fictional fluxes (compare with fictional forces in mechanics) that let you "perfectly" track interfaces, in this case the interface between the solid and the fluid. This lets you arbitrarily move the vertices of the spatial discretization, and therefore the elements. If this was not done, a huge loss of precision is incurred by the fact that any movement must be projected onto the discrete elements. As it does not make much sense to have "half of a beam" in one element, and indeed we have no way of calculating "half of the beam equation" in an element, this proves crucial for simulating these types of problems. This issue is not present to the same degree in Fluid-fluid simulations (where different species are present), where you are anywhere solving Navier Stokes, and can reasonably approximate the behaviour of one element by the Fractional quantity of species A and B etc - and advect the same fractional quantities by the velocity field obtained from solving the navier stokes equation.
I would wager that the simulation in OPs pic solves two different PDE:s, and those are the beam equation for the cyan solid, and navier stokes everywhere else. The concept fluid structure interaction is reserved for these types of _mixed_ computations, where different PDEs are solved on the same domain, but different regions.
The interaction portion comes from calculating the forces on the solid body, typically via evaluation of line integrals, conversely, the solid applies pressure on the fluid, and this need also be calculated. When both these things are done, we have a FSI (fluid struct. interact.) simulation.
