NUMERICAL AND ANALYTICAL INVESTIGATION OF SUBCYCLING IN THE FLOW PROBLEM OF A STRONGLY-COUPLED PARTITIONED FLUID-STRUCTURE INTERACTION SIMULATION

Fluid-structure interaction (FSI) simulations can be used to quantify the frequency, damping constant and amplitude of the vibration of equipment such as piping and heat exchangers. Typically, the time step is the same in the flow and structural equations, but this causes long computational times when the time step is restricted due to stability requirements of only one solver. In that case, a more efficient approach is to use so-called subcycling with a different time step in the flow and structural solver. In this paper, only subcycling with a smaller time step in the flow solver compared to the structural solver is analyzed. The research presented here is split into two parts: an analytical study and a numerical computation of the one-dimensional flow in an elastic cylindrical tube. Firstly, a monolithic analytical FSI calculation is analyzed with a Fourier stability analysis. This allows to verify the stability of the solution by considering the eigenvalues of the problem as a function of the perturbation wavenumber. The conclusions drawn from the analytical study are subsequently verified in a partitioned numerical FSI simulation, coupling the flow solver Fluent with the structural solver Abaqus. The implicit coupling is achieved using an interface quasi-Newton method with an approximation of the inverse of the Jacobian (IQN-ILS), implemented in the in-house code Tango. The research shows that a stable solution is attained for significant subcycling in the flow problem: the results indicate that the solution remains temporally stable even if the time step in the flow solver is only one tenth of the structural time step. However, some (temporally stable) oscillations in the resulting pressure profile on the pipe wall arise when the time discretization schemes applied in the flow and structural solvers are different. These oscillations do not persist when the same time discretization scheme is applied.