Toward high-fidelity Numerical Wave Tank development: Scale resolving Partially-Averaged Navier–Stokes simulations of dam-break flow

Abstract

Numerical Wave Tanks (NWTs) based on Navier–Stokes equations are emerging as valuable tools for simulating various ocean and coastal engineering flows. By capturing essential hydrodynamic non-linearities, including turbulent processes, these Navier–Stokes based NWTs can potentially provide high-accuracy data to complement traditional experimental wave basins for design and development applications. Many NWT approaches currently use the RANS (Reynolds Averaged Navier–Stokes) method. For many coastal resiliency and wave-energy converter applications, a higher fidelity turbulence treatment may be needed. In this work, we explore the applicability of the scale-resolving Partially-Averaged Navier–Stokes (PANS) technique for NWT computation. The PANS model is employed to study the dam-break benchmark case which is an important benchmark case for seawall and wave energy converter hydrodynamics analysis. Computations using RANS are also provided as a reference for accuracy and computational effort. The objectives are to: (i) investigate the effect of enhancing turbulence scale resolution on the pressure field, wave elevation and three-dimensional effects within the tank; and (ii) examine the influence of initial kinetic energy and grid resolution on the results. It is shown that, the early transient behavior is reasonably insensitive to scale resolution or initial turbulence level. Indeed, RANS and PANS of different scale resolutions yield similar results in the early stages as the flow physics is dominated by quasi-linear flow processes. The differences between RANS and PANS appear at later times. The so-called second pressure peak, standing wave amplitude and three-dimensional effects are better captured with increasing scale resolution for the same grid resolution. Overall, the results indicate that PANS is a viable model for NWT applications, with more detail being captured at a similar computational cost compared to RANS.