Aeroacoustic source localization using bayesian approaches
Context
Global economic and demographic expansion is leading to an increase in noise pollution associated with land and air transportation as well as new energy developments (wind power). Noise pollution regulations are becoming increasingly strict and the reduction of aerodynamic noise sources, known as aeroacoustics, is therefore a major challenge. The characterization of aeroacoustic sources in the pre-project phase by means of wind tunnel tests is a key step to help understand noise generation mechanisms and to be able to develop effective noise reduction strategies.
For this purpose, various source localization techniques have been developed. The most well known is called Beamforming and was developed by Billinglsey and Kinns in 1974. This technique is based on the assumption that the sound field radiated by the sources under study follows a certain source model (usually monopole). It is then possible to localize the acoustic sources from farfield microphone measurements by interpreting the propagation delays measured between each microphone of the antenna and by knowing the source-antenna distance. However, the use of inverse methods is required for the evaluation of the sound level of the studied sources. Different methods based on deconvolution algorithms have been developed for that purpose: CLEAN, DAMAS.
More recently, source localization techniques based on bayesian statistics have also been developed. These techniques offer an original solution to the inverse problem of acoustic source localization. Work has been done in recent years to evaluate the potential of these methods on simple test cases. This approach seems particularly promising for the localization of aeroacoustic sources in wind tunnels and has already shown its potential in other fields of aeroacoustics, such as the characterization of porous materials or the impedance reduction of liners under grazing flow.