Fluid–structure interaction (FSI) phenomena are widely present in various industrial fields. In the nuclear sector, such phenomena occur within the reactor core and play a key role in the vibratory response of fuel assemblies subjected to coolant flow and external mechanical excitation, particularly during earthquakes. The long term goal is to perform best-estimate calculations of a reduced-size fuel assembly in order to extend the experimental results beyond the available measurements and improve the understanding of FSI mechanisms. To achieve this, we adopt a partitioned approach, which consists of treating the fluid and structural equations as two distinct subsystems, solved at each time step while exchanging fields at the interface through the corresponding boundary conditions. This approach allows the use of two separate solvers: the TrioCFD code is used to model incompressible unsteady flow, while Europlexus is employed to solve the nonlinear dynamic equations of the structure. The fluid–structure interface is tracked using the Arbitrary Lagrangian–Eulerian method. Two partitioned coupling algorithms explicit and implicit have been implemented between the two solvers. The explicit coupling is known to exhibit numerical instabilities in the presence of the so called added-mass effect. These instabilities can be overcome by the implicit coupling scheme. A Gauss–Seidel-type algorithm enables sub-iterations between the fluid and structural solvers until a prescribed convergence threshold is reached. However, this approach may require a large number of implicit iterations. To improve convergence performance, a dynamic relaxation technique is implemented in this work. In addition, we address the performance loss caused by the disparity between the time scales of the two solvers by enhancing existing sub-cycling techniques within the algorithms, which proved essential for improving computational efficiency. Finally, a study of the stability domains of the different coupling algorithms, as well as an analysis of the influence of sub-cycling on computational performance, is carried out in the present work, and a numerical validation is performed through comparison with results available in the literature.