Prediction and Examination of Rotor Sound using Computational Fluid Dynamics

Abstract

Rotors are extensively used in aerodynamic applications, e.g., propellers, helicopter rotors, wind turbines, and fans. They are a large source of unwanted noise. The sound produced by low speed rotors has received less attention in the literature than sound produced by high speed rotors. These differences in speed and environment can be described by the Mach number, typically defined as the ratio of total tip speed of the rotor to the speed of sound. This dissertation discusses the implementation of a general numerical methodology for the prediction of sound using computational fluid dynamics and examines the production of noise by a low Mach number, marine propeller operating at design conditions. To simulate the fluid flow, the Navier–Stokes equations of fluid motion are solved using both large–eddy simulation and direct numerical simulation. Incompressible and compressible simulations based on the finite-volume methods of Mahesh et al. and Park and Mahesh are conducted. These methods have been used to successfully predict flowfields across a variety of problems in the past. The simulation results are used in conjunction with the Ffowcs-Williams and Hawkings acoustic analogy to predict the sound generated from the simulated flowfield. Within this dissertation, the general acoustic prediction tool is developed. Sound waves are inherently a compressible phenomena, and while the governing equations are well known, the prediction of sound (across Mach numbers) is still a very active field of research. Computationally, it becomes difficult to capture the range of spatial and temporal scales in many acoustic problems. Compressible simulations can require large domains that are highly resolved to capture sound waves. Incompressible simulations do not suffer from these adverse effects, however, it is important to ensure that the acoustic prediction remains accurate when the simulation does not capture the acoustic wave propagation. Hence the tool is applied to predict sound from the flowfield of the canonical pulsating sphere. Results predicted from both compressible and incompressible simulations are discussed. Noise generation is known to be exacerbated by the pressence of a geometric surface singularity, such as blade tips, corners, trailing edges. The case of noise generated by a trailing edge is examined directly using a two-dimensional compressible DNS of laminar flow around a NACA 0012 airfoil. The simulations are conducted at a Mach number of 0.4 and a Reynolds number of 50, 000 based on the airfoil chord length. Finally, the method is used to predict sound from an incompressible LES of propeller DTMB 4381 operating at design conditions. The simulations are conducted at a Reynolds number of 894, 000 and an advance ratio of 0.889 based on the propeller diameter. Three distinct source regions are found along the propeller blade: the hub-root region, the blade mid-span, and the blade tip region. Flow separation in the hub boundary layer interacts with the blade-root surfaces to generate a directionally independent, broadband sound field. This interaction decays with increasing radius along the propeller blades. In the blade mid-span, where the blade generates most of its thrust, high levels of sound are created by boundary layer turbulence convecting past the blade trailing edge. The sound source region is confined very close to the blade trailing edge. The highest levels of sound are found to be generated very near the blade tips. The radial truncation of the surface leads to the shedding of an unsteady tip vortex. The locally separated flow near the core of the vortex creates high levels of unsteady surface loading.