Author(s): Sergio Croquer; Pilar Diaz-Carrasco; Sebastien Poncet; Jay Lacey; Ioan Nistor
Keywords: Computational fluid dynamics; Numerical modelling; Wave-structure hydrodynamics; Vertical walls
Abstract: World climate forecasts predict increasing sea levels and more extreme weather phenomena in the upcoming decades. This implies greater demands for existing and future coastal protection structures, since extreme events will be more violent and frequent. For instance, wave climate models for the Gulf of St. Lawrence predict a 5-10% increase in extreme wave heights by 2100 due to the disappearance of sea ice, which acts as a wave attenuator (Ruest, 2016). Numerical models, and in particular computational fluid dynamics, is a promising tool to tackle this problem as the hydrodynamics of wave-structure interactions can be studied in great detail. Reynolds-Averaged Navier-Stokes (RANS) models of wave impacts have gained popularity in recent years given the availability of specific frameworks such as olaFlow (Higuera, 2017) and OpenFOAM (OpenCFD, 2014) as well as the ever increasing available computational power. Yet, setting up a RANS model requires the adjustment of several parameters including: numerical schemes, turbulence model, interface capture methods, and grid type such that knowing a priori an optimal configuration in terms of accuracy vs. computational cost is rarely obvious. The main objective of this study is to provide insight on the effect of different numerical settings such as: discretization schemes, air thermodynamic model, interface capture methods, mesh type and turbulence modelling. To this end, a two-step numerical benchmark has been carried out. Firstly, the influence of the basic solver settings (time and advection numerical schemes, air thermodynamic models and interface capture method) has been studied using for validation several canonical liquid impacts: 1D Piston, 2D falling water column and sway tanks. Secondly, the best performing configuration has been chosen as baseline to assess the effects of different two-equation turbulence models and meshing approaches on the accuracy for predicting experimental wave impact data on a vertical structure (Kisacik, 2012).