Author: Joseph George Kocheemoolayil

Publisher:

Published: 2016

Total Pages:

ISBN-13:

Sustaining the continued growth of aviation is critically dependent on managing its noise emission. Developing tools to predict airframe noise from first principles is a pacing item in this regard. Within this context, noise generated by flow past airfoils constitutes an important canonical problem that is also relevant to a wide variety of other applications such as wind turbine noise, cooling fan noise, turbofan noise, propeller noise and helicopter blade noise. The noise generated by a turbulent flow that encounters the trailing edge of an airfoil is the fundamental component of all these problems. Over the past 15 years, significant strides have been made towards using large eddy simulations (LES) for predicting airfoil noise from first-principles. However, they have largely been restricted to canonical configurations at low Reynolds numbers. Perhaps the restriction to low Reynolds numbers is the most serious limitation since majority of the experiments target full-scale Reynolds numbers making one-to-one comparisons impossible. This thesis focuses on extending the scope of LES based predictions to full-scale Reynolds numbers and non-canonical configurations such as the near-stall and post-stall regimes which have received very limited attention owing to their complexity. Wall-modeled large eddy simulations (WMLES) that combine LES with a model for unresolved near-wall turbulence are used to predict airfoil noise at high Reynolds numbers. The Benchmark Problems for Airframe Noise Computations (BANC) workshop is held every year as part of the AIAA/CEAS Aeroacoustics conference. Category 1 of the workshop targets airfoil trailing edge noise prediction at high Reynolds numbers relevant to engineering applications. No first-principles based approach free of empiricism and tunable coefficients has had success in this category to date. Independently validated far-field noise measurements are available for four configurations in the category. Our simulations predict trailing edge noise accurately for all four configurations. Detailed comparisons are made with dedicated experiments. Insensitivity of the simulation results to important aleatory and epistemic uncertainties is established. Resolution requirements for making accurate noise predictions using WMLES are identified through a systematic grid-refinement study. Developing the capability to predict airfoil noise for near-stall and post-stall configurations is necessary to investigate their suspected responsibility for a phenomenon known as Other Amplitude Modulation (OAM) of wind turbine noise. Predicting the flow past a wind turbine airfoil in the post-stall regime is a formidable challenge in itself. In particular, there is a school of thought that large scale three-dimensionality and extreme sensitivity to the experimental facility are inevitable and preclude the possibility of a fair comparison between simulations and measurements in this regime. However, in agreement with a recent theoretical study our simulation results indicate that the lower lift due to large scale three-dimensionality can be reproduced even in span-periodic simulations if the domain size is sufficiently large. The large span simulation predicts the pressure distribution around the airfoil with unprecedented accuracy. Successful prediction of pressure fluctuations on the airfoil surface beneath the suction side boundary layer is demonstrated in the near-stall and post-stall regimes. Previously unavailable two-point statistics of surface pressure fluctuations are documented.