Adaptive Mesh Refinement for Computational Aeroacoustics

Abstract

UNIVERSITY OF SOUTHAMPTON ABSTRACT FACULTY OF ENGINEERING, SCIENCE & MATHEMATICS SCHOOL OF ENGINEERING SCIENCES Doctor of Philosophy ADAPTIVE MESH REFINEMENT FOR COMPUTATIONAL AEROACOUSTICS by Xun Huang This thesis describes a parallel block-structured adaptive mesh refinement (AMR) method that is employed to solve some computational aeroacoustic problems with the aim of improving the computational efficiency. AMR adaptively refines and coarsens a computational mesh along with sound propagation to increase grid resolution only in the area of interest. While sharing many of the same features, there is a marked difference between the current and the established AMR approaches. Rather than low-order schemes generally used in the previous approaches, a high-order spatial difference scheme is employed to improve numerical dispersion and dissipation qualities. To use a high-order scheme with AMR, a number of numerical issues associated with fine-coarse block interfaces on an adaptively refined mesh, such as interpolations, filter and artificial selective damping techniques and accuracy are addressed. In addition, the asymptotic stability and the transient behaviour of a high-order spatial scheme on an adaptively refined mesh are also studied with eigenvalue analysis and pseudospectra analysis respectively. In addition, the fundamental AMR algorithm is simplified in order to make the work of implementation more manageable. Particular emphasis has been placed on solving sound radiation from generic aero-engine bypass geometry with mean flow. The approach of AMR is extended to support a body-fitted multi-block mesh. The radiation from an intake duct is modelled by the linearised Euler equations, while the radiation from an exhaust duct is modelled by the extended acoustic perturbation equations to suppress hydrodynamic instabilities generated in a sheared mean flow. After solving the near-field sound solution, the associated far-field sound directivity is estimated by solving the Ffowcs Williams-Hawkings equation. The overall results demonstrate the accuracy and the efficiency of the presented AMR method, but also reveal some limitations. The possible methods to avoid these limitations are given at the end of this thesis.