Broadband Noise Prediction of an Axial Compressor Operating at Low Reynolds Number
This thesis is an investigation into the production and prediction of broadband noise in axial compressors operating at low Reynolds number. The low Reynolds number (chord based
The research was conducted by designing and constructing a new single stage rotor-stator compressor on which the subsequent research was based. The work utilized a combination of experimental, computational and analytical methods. Aerodynamic and acoustic properties were measured on the compressor installed in a duct with clean inflow over a wide range of operating points. These results were then used to validate a set of incompressible, wall resolved Large Eddy Simulations (LES) on operating points representing on and off-design behavior. In the existing literature, there have been few if any studies utilizing wall resolved LES of a 3D rotor blade geometry for aeroacoustic source evaluation. A mesh dependency study was conducted in order to assess the level of refinement needed for convergence of turbulent statistics. It was found that although a fine resolution was needed for convergence of velocity and pressure spectra describing acoustic sources, a relatively coarse LES could still provide similar estimates of second order statistics such as Reynolds stresses. When compared to experiment, the LES results show a pleasing ability to accurately capture both first and second order statistics as well as turbulent velocity spectra, which cannot be said about steady Reynolds Averaged Navier Stokes (RANS) simulations. The simulations reveal that a significant portion of the midspan region of the blade remains laminar for the entire blade chord up until the trailing edge. Spanwise migration of the flow due to the rotational forces allow the boundary layer to stay laminar and attached. Generation of turbulence is then mainly limited to the secondary flows at the hub and tip gap which account for a large proportion of the wake turbulence. The secondary flows also feed turbulence into the blade profile boundary layers in these regions.
A hybrid aeroacoustic prediction framework, based on the turbulent source description provided by the LES, and analytical acoustic propagation models are used to predict the overall sound power spectrum and compared favorably with the duct measurements. It is found that trailing edge noise and rotor-stator interaction noise are equally important contributors to the overall noise spectrum. However key narrowband humps which are not explained by these noise models exist, particularly off design.