Empirical sensitivity of two-dimensional nonlinear wake-cylinder oscillators in cross-flow/in-line vortex-induced vibrations

Phenomenological wake-cylinder oscillators have been extensively implemented for vortex-induced vibration (VIV) predictions. Although such models capture fundamental VIV phenomena, the maximum response estimations and comparisons with different experimental data reveal some quantitative discrepancies due to the model empiricism embedding some uncertainties through system variables. This vital issue has not been well addressed in the literature of VIV modelling. This paper presents a new comprehensive investigation into the sensitivity to empirical input variables of nonlinear wake-cylinder oscillators simulating the two-dimensionally coupled cross-flow/in-line VIV and amplified mean displacements of a flexibly mounted circular cylinder in uniform flows. The fluid-structure coupling terms are advanced by accounting for the higher-order nonlinear effects of fluctuating lift-drag forces and steady-drag dynamic magnifications, depending on the relative flow-cylinder velocities. A random sampling and variance-based sensitivity studies are carried out using Monte Carlo simulations which are computationally efficient based on the reduced-order model. This enables a large series of parametric examinations. Individual contribution, relative importance, coupling and interdependence of multiple input variables affecting output uncertainties are qualitatively and quantitatively evaluated. The Reynolds number dependence is also captured by correlating the wake and hydrodynamic coefficients with experimental data. Parametric studies highlight greater variations in the predicted amplitudes and mean displacements of the cylinder two-degree-of-freedom VIV with a lower mass ratio. Numerical findings allow for the identification of a few most influential variables to be treated as the empirically tuned coefficients. The improved understanding of model versatility and sensitivity enhances the calibration confidence and the response predictability with a reduced computational effort.