Weakness in semi-empirical broadband noise prediction: Investigation with RANS CFD for small scale moderate camber airfoil propellers

The understanding and characterization of aerial systems acoustic properties for noise reduction, once afterthought save for specialized fields or situations, in recent decades has become of critical importance as a design criterion to capture completely. Exploding unmanned aerial system (UAS) usage is leading to increased importance due to growing complaints about aircraft noise from populations due to increased UAS encroachment into the public airspace. Aerial system noise can be thought of as tonal or broadband in character with tonal mechanism being well understood and predicted. Broadband mechanisms are much more difficult to predict, requiring either computationally intensive methods to capture or using simplified lower fidelity models that are limited in applicability and/or accuracy. This work explores the applicability of the lower fidelity broadband airfoil self-noise prediction method of Brooks, Pope, and Marcolini to the analysis of moderate camber propellers operating in a lower Reynolds number regime. Experimental flow and boundary layer (BL) data was collected for a series of computational fluid dynamics (CFD) studies examining differing airfoils over a range of Reynolds numbers and angles of attack. Comparisons were conducted examining the results for Brooks, Pope, and Marcolini BL modelling against those from CFD. Also compared were Brooks, Pope, and Marcolini acoustic predictions using modelled BL data against those using CFD BL data. To investigate methods to address weaknesses found in the model, proof of concept CFD investigations using LES turbulence modeling to provide unsteady loading data for prediction of self-noise using Ffowcs-Williams Hawkings (FW-H) were conducted. Preliminary findings for the NACA 1412 for average flow condition CFD TE BL data and modeled TE BL data driven Brook’s self-noise predictions against FW-H unsteady loading predictions are presented. Finally, as an investigation of a Reynolds Averaged Navier Stokes trailing edge BL correction method, noise predictions for an OSU custom 18 x 18 3 blade propeller were investigated using a blade element method implementation of Brooks, Pope, and Marcolini. The results of these studies showed weakness in the method when predicting cambered airfoil self-noise, demonstrating weakness in BL modelling for both Reynolds number and angle of attack. Further it is shown that accurate BL capture is important to both frequency content and OASPL for model total self-noise and individual mechanism noise predictions. Suggestions are provided for further work to refine the identification of weaknesses in the model and develop improved prediction models and methodology.