Fast random wave generation in numerical tanks

Abstract

Generating and absorbing random waves in numerical models is a challenging problem, in particular when meaningful wave statistics should be generated to meet design sea-state requirements. The methodology presented herein allows for the generation of random wave fields (free surface elevation and velocities) to be reconstructed in time and in space by using window processing from a reference time series. It is demonstrated that the methodology is efficient in reproducing long non-repeating wave sequences by using only O(101)–O(102) wave components, rather than O(103)–O(104) required by a direct reconstruction from a single spectrum. This reduces the computational times required for the development of wave-train time-series elements by 40 times. Errors in instantaneous surface elevation and particle velocity between windowed and non-windowed reconstruction techniques were less than 0·4 and 0·2%, respectively, in the cases considered. The technique was combined with the relaxation zone method typically used in numerical wave tanks for generating waves. The simulations were performed using Proteus, a rapidly developing CFD−FEM toolkit for modelling fluid−structure interaction cases. The use of windowed reconstruction reduced the overall computational time associated with the simulation of waves in a numerical wave tank by ∼40 and ∼70% for serial and parallel execution. Results of the study show windowed reconstruction to be suitable for the representation of long-duration wave trains. Wave height and peak period are conserved within 2 and 1%, respectively.