Do the FSI simulations require a convergence check? Some coworkers requested me to put a convergence check step in the FSI simulations. However, I am not sure why I have to put the convergence check in it. The FSI simulations require a communication between the fluid and the structure. The fluid pressure is put in the structural calculation and the structural deformation is used to adjust the fluid mesh. Where does the convergence check require? Why does it require? Could anyone please advice me?

In reality the fluid-structure system should be solved all at once in one big "matrix solve". By solving them separately you are introducing a lag in the solution which can cause numerical difficulties if the solution is not iterated within each timestep. The convergence check is to make sure that the structural and fluid solutions have converged within each timestep.

I am solving the FSI simulations separately. For example, solve the fluid solution first then the fluid pressure is put in the structural calculation. I am not using a sinlge matrix for the structure and the fluid. Based on your comments, I think I do not need to check the convergence since I am solving two systems separately. Is my statement right?

Hi Jinwon,

The reason why you need to resort to convergence checks is primarily because of that 'lag' you are introducing between the fluid and the structure. When the structural equations are solved, say for eg : Mx'' + Kx = F. All these quantities have to be at the same time level (say t = n+1). But you dont have the aerodynamic forces (F) at t = n+1. So you use the forces at t=n initially to obtain a 'predicted' displacement 'x'. Using this displacement, recompute your velocity, acceleration etc and solve for the pressure..hence F. You will continue to do this till your 'x' is converged.

Regs...Dominic

If there is no convergence check in the FSI simulations, what happen there?

Your understanding of my comment was exactly opposite..read it again.

It is possible that the simulation will become numerically unstable.

algorithms that do not check convergence within each time step are called loose or weak coupling algorithms. there is a time lag between fluid and structure domains which has influence on convergence and stability of the system but that does not mean that your solution must be incorrect, you can obtain correct or reasonable results by loose coupling. actually that depends of the problem type, non-linearity degree of the whole system, and the time step size you are applying. additional iterations between fluid and structure domains within each time step are implying better convergence and stability to the system and they are known as close or tight or strong coupling algorithms. you should have a compromise between using those different kinds of algorithms i think. additional iterations mean additional computation and additional work, if the problem does not need those additional iterations then it is unnecessary to have strong coupling. if you have access to the source codes making a switch for close and loose coupling would be a better choice i think. by that way, you can have initial calculations for the problem you are dealing with to understand if it needs close coupling or whether loose coulping would be sufficient. having the whole systems of equations in an unified matrix system is known as monolitic coupling, and it is not that similar to loose or close coupling. because there is no seperate solvers for fluid and structure domains, and needs more effort to write and to maintain an unified solver.