We address the problem of obtaining accurate predictions of the vortex-induced vibrations (VIV) of flexible structures placed in cross-flow by reconstructing fluid force databases in parametric form. The forces are expressed in terms of hydrodynamic coefficients, which depend on several variables and parameters, including the structural properties and geometry, frequency and amplitude of vibration, and the flow conditions, so that a comprehensive database obtained with systematic experiments would require an intractably large number of tests. Therefore, we develop a method for determining optimal parametric hydrodynamic databases directly from a variety of free- and forced-vibration experiments on both rigid and flexible cylinders, by minimizing the error between the predicted responses using a software driven by the force database, and the measured data. We first describe the process of selecting an appropriate parametrization of the fluid forces as function of reduced frequency and non-dimensional amplitude of vibration, informed by Gaussian-Process-Regression-guided automated experiments. Several examples are selected to demonstrate the method’s ability to acquire optimal hydrodynamic databases, including data sets highlighting the effect of Reynolds number and the addition of in-line motion to cross-flow rigid cylinder VIV. The proposed methodology is further demonstrated to be effective when used with the semi-empirical program VIVA for flexible cylinder VIV, yielding significant improvements in prediction accuracy compared to baseline results. It therefore serves as a potential solution for the development of a digital twin for marine risers, an in-situ, real-time riser response prediction and structural health monitoring system.
Read full abstract