Download files
Abstract
Underwater radiated sound from marine vessels is a significant problem for research, fishing and military vessels, and is a major source of pollution in the marine environment. For an underwater vehicle such as a submarine, the major sound sources contributing to underwater noise are the on-board machinery, hull flow noise and the marine propeller. A detailed understanding on the sound generation, propagation and radiation into the acoustic far field is required before effective control and reduction of the radiated noise can be successfully achieved.
Flow induced noise generated by a marine vessel is the motivation for this thesis. However, the focus of this thesis is on developing numerical techniques that are suited to solving this particular application. As such, the flow induced noise of a marine vessel is not explicitly studied. Instead, a more general framework is presented that can predict low Mach number flow induced noise generated by any body of arbitrary size and shape travelling through a fluid medium. The turbulent structures in the flow are the sources of flow induced noise. Sound waves propagate from these acoustic sources and either travel directly to the far field or are first diffracted by the body. To numerically predict the far-field radiated sound produced by the body travelling through a fluid, the acoustic sources must be accurately captured. Also, the propagation of the resulting sound waves and their diffraction around the body has to be resolved.
In this thesis, a particle accelerated hybrid computational fluid dynamics (CFD) and boundary element method (BEM) technique is developed to predict low Mach number flow induced noise, the scattering of this noise by a body immersed in the flow and the resulting far-field sound pressure. A CFD model is used to predict the fluctuating hydrodynamic flow field. The acoustic sources are extracted from this data. A near-field formulation for the acoustic pressure and pressure gradient based on Lighthill's acoustic analogy is derived. A singularity regularisation technique has been adapted and extended to regularise the strongly singular and hypersingular integrals present in these near-field formulations. The incident sound field on the body is calculated using the near-field formulations and applied to a BEM model of the body to predict the scattered sound fields.
To address the significant computational cost and data storage requirement in extracting volumetric acoustic sources from CFD data, a multipole particle condensation technique has been developed. This particle condensation both reduces the amount of data that must be stored during the CFD simulation and accelerates the calculation of the acoustic wave propagation. The method uses a particle approximation of the acoustic source distribution and employs a Taylor series expansion of the harmonic Green's function to spatially condense the underlying acoustic sources and preserve their multipole moments. A technique to interface the near-field formulation with the multipole particle condensation technique has also been developed in this thesis. Using this interface technique, the incident acoustic field on the body is predicted with a reduced data storage requirement. The near-field formulation is applied to the acoustic sources immediately adjacent to the body to ensure the singularities present in the harmonic Green's function and its derivatives are regularised in a mathematically robust manner. The propagation of the acoustic waves from sources further away from the body is calculated using the multipole particle condensation technique. The particle accelerated hybrid CFD-BEM technique is demonstrated by calculating the flow induced noise generated by laminar flow past a cylinder.