Simulating Vortex-Induced Vibrations in Sheared Current by Using an Empirical Time-Domain Model with Adaptive Parameters

Slender marine structures, such as risers and power cables are subject to various loads, where Vortex-Induced Vibrations (VIV) is known to have a significant impact on accumulation of fatigue damage in the materials. The stochastic nature of VIV makes it challenging to do accurate fatigue predictions even when the underlying numerical model is deterministic. The current state-of-the-art is to model VIV response in the time-domain, where semi-empirical models have shown promising results. However, there are significant uncertainties in the fatigue prediction associated with assuming the values of the empirical model parameters. In the present paper, an efficient gradient-free optimization method is proposed to adapt the empirical parameters based on curvature measurements from model tests. Prior to the optimization problem, a global sensitivity analysis was applied to determine which parameters that have the largest influence on relevant quantities of interest. A variance-based sensitivity analysis method using Sobol' indices was used together with a Polynomial Chaos Expansion to increase the computational efficiency of the method. The yearly fatigue damage was computed for model tests with a riser in sheared current and simulated using the optimal, adaptive parameters. Using adaptive parameters improved the prediction of curvatures, including both the maximum curvature and identification of the dominating frequency related to the given curvature. The predicted maximum fatigue damage was also improved, especially for the in-line direction.