Phenomenological modelling based on the use of coupled wake-cylinder oscillators has been implemented for several decades for vortex-induced vibration (VIV) response predictions of rigid circular cylinders and long flexible risers. Although such models can capture several VIV nonlinear phenomena, they still rely on empirical coefficients and parameters with some levels of uncertainties associated with fluid-structure interactions. Most of these are due to the variation in experimental test conditions giving rise to different benchmarking test data in the literature. A very few studies have quantified such model uncertainties.
To gain an insight into the relative contributions of these coefficients, this paper presents a sensitivity study based on a multivariate Sobol-Monte Carlo approach for a two-dimensional nonlinear coupled wake-cylinder oscillator model simulating combined cross-flow/in-line VIV. A preliminary investigation is carried out to identify the relative contribution to the model uncertainty of the geometrical, empirical wake and wake-cylinder coupling coefficients. The effect of Reynolds number (Re) in the subcritical flow regime is also taken into account through a lift, drag, Strouhal and wake coefficient depending on calibration with experimental data. A key challenge is the identification of suitable probability distribution function to model the scattering experimental datasets. A combination of Gaussian and uniform distributions are calibrated.
Parametric studies based on the combined Sobol-Monte Carlo simulations reveal the uncertainties due to the coupling of empirical coefficients and system parameters governing the two-dimensional lock-in and combined cross-flow/in-line VIV responses. The relative importance of the selected wake coefficients and geometric parameters appears to be strongly dependent on the system mass ratio as well as Re influencing the hydrodynamic properties including the lift and drag coefficients, the Strouhal number and stall parameter. For the cylinder with a lower mass ratio, greater uncertainties are found within the considered sub-critical Re range for both cross-flow/in-line VIV. This remark should also be recognized in the phenomenological VIV modelling of long flexible risers in offshore applications.