Surrogate model based uncertainty quantification of CFD simulations of the viscous flow around a ship advancing in shallow water

Abstract

Usually, Computational Fluid Dynamics (CFD) simulations are carried out under deterministic conditions. However, there are actually many uncertain factors such as fluid properties, boundary conditions and working parameters that may have effects on the system response. It is critical to study the uncertainty propagation in the stochastic CFD model. In this paper, one of surrogate models for uncertainty quantification (UQ) problems, non-intrusive polynomial chaos (NIPC), is introduced. Stochastic two-dimensional cavity flow with uncertain lid velocity is taken as a study case to verify the effectiveness of NIPC method in comparison with Monte Carlo (MC) method. The results show that NIPC method can greatly improve the efficiency of UQ; the influence of velocity uncertainty has obvious directivity in the flow field and the fluid viscosity has retardation effect on uncertainty propagation. Then the NIPC method is applied to quantify the uncertainty of resistance, sinkage and trim of the Japan Bulk Carrier (JBC) model advancing with an uncertain speed in shallow water. Numerical simulations are carried out using the RANS-based CFD method. The UQ results indicate that the uncertainty of relevant variables can be evaluated accurately with small samples, and under the same sampling points, the higher the order of polynomial chaos expansion (PCE), the more accurate the UQ results.