Aeroacoustic 3D computations for a NACA 0012 airfoil using LES and acoustic analogy
Title: Aeroacoustic 3D computations for a NACA 0012 airfoil using LES and acoustic analogy
SNIC Project: SNIC 2019/3-362
Project Type: SNIC Medium Compute
Principal Investigator: Marcus Berg <>
Affiliation: Uppsala universitet
Duration: 2019-06-27 – 2020-07-01
Classification: 20306


Although Computational aeroacoustics (CAA) is a useful technique for noise prediction and reduction in many applications, it generally costs much more than standard computational fluid dynamic (CFD) simulations. One reason is that acoustic waves have very small amplitudes in many cases. Since the noise is generated by vortices’ convection and diffusion in a flow or by their interaction with structure surfaces, these dynamic motions have to be captured with high accuracy, both temporally and spatially. Additionally, a large number of time steps is needed to analyze the sound spectrum over a wide range of frequencies. Thus, the computational time tends to become very long. While CAA for airfoils are a relatively new research topic, some studies have attempted to compute the acoustic field and sound propagation generated from an airfoil. Since the direct noise computation requires relatively large computational resources, the hybrid method is more popular and has been tested for prediction of noise from airfoils. In the hybrid method, only the flow in the near field around the airfoil is solved in CFD. The sound propagation is obtained via the wave equation, although the necessary resolutions are as high as in the direct noise calculation method. In some studies using the hybrid method, the sound levels are predicted well for dominant frequencies with a discrepancy of a few decibels, while it is still challenging to reproduce the acoustic waves for high frequencies. One issue in CAA for airfoils is that the data of sound sources obtained from CFD is very limited due to the reasons mentioned above. This can be demanding if one tries to simulate blades with long span. The spanwise computational domain size is generally around 5 to 20 % of chord length in previous works. If the span length is larger than the domain size, the sound source outside of the domain has to be extrapolated to know the sound level emitted from the entire blade. Most of the previous studies assume that the sound source occurs independently along the span. So, the sound pressure generated from the whole blade is obtained by simply scaling up the sound energy. We consider that this assumption is too crude in some flow conditions, so this project will propose a better way for correction to improve the noise prediction in a wider range of frequencies. Aeroacoustic computations are conducted in this project for a flow field around a cylindrical blade with a NACA 0012 airfoil section. The hybrid method is used for acoustic calculations where the 3D incompressible Navier-Stokes equations are solved using LES; then Curle’s acoustic analogy is applied to calculate the sound propagation. Our acoustic calculations will be validated by comparing with sound pressure measured by a group from NASA. As explained above, the spanwise correction for sound pressure will be applied. It will be tested whether this approach is more realistic than the scaling method which has previously been used in other studies to predict total noise.