Modelling of Extreme Ocean Waves Using High Performance Computing

Abstract

This thesis describes the development of a fully nonlinear numerical model for the simulation of surface water waves. The model has the ability to compute the evolution of both limiting and overturning waves arising from the focussing of wave components in realistic ocean spectra. To accomplish this task, a multiple-flux implementation of a boundary element method is used to describe the evolution of a free surface in the time domain over an arbitrary bed geometry. Unfortunately, boundary element methods are inherently computationally expensive and although approximations exist to reduce the complexity of the problem, the effects of their use in physical space is unclear. To overcome some of the computational intensity, the present work employs novel computational approaches to both reduce the run time of the simulations and make accessible predictions of wave fields that were previously unfeasible. The advances in computational aspects are made through the use of parallel algorithms running in a distributed computing environment. Further acceleration is gained by running parts of the algorithm on many-core co-processing devices in the form of the, habitually called, graphics processing unit. Once a reasonably efficient implementation of the boundary element method is achieved, attention is turned to further algorithmic optimisations, particularly in respect of computing the kinematics field underlying the extreme wave events. The flexibility of the model is demonstrated through the accurate simulation of extreme wave events, this includes near-breaking and overturning wave phenomena. Finally, by harnessing the power of high performance computing technologies, the model is applied to an engineering design problem concerning the wave-induced loading of an offshore jacket structure. The work presented is not merely a study of a single wave event and its interaction with a structure, but rather a whole multitude of wave-structure interaction events that could not have been computed within a realistic time frame were it not for the use of high performance computing. The outcome of this work is the harnessing of distributed and accelerated computing to enable the rapid calculation of numerous fully nonlinear wave loading events to provide a game changing outlook on structural design and