Investigation of beamforming techniques for rotating noise sources.

Over the last few decades, acoustic beamforming methods with phased microphone arrays have gained increasing attention from both industries and research institutes, providing a practical tool for acoustic-field investigations in wind-tunnel tests and achieving accurate results. The purpose of this Master Thesis is to provide a detailed overview of the main beamforming algorithms for the localization and quantification of sound sources generated by rotating machines, such as wind turbines, low-pressure fans, and low-speed propellers, as well as a thorough understanding of their potential and inner limits. The time-domain ROtating Source Identifier (ROSI) method and the frequency-domain Virtual Rotating Array Method (VRAM) method are currently the most widely used beamforming techniques in the acoustic community. The first one provides satisfactory results regardless of the geometry and spatial position of the microphone array, but it requires significant processing time and does not allow for the application of advanced beamforming methods. The second one supports advanced beamforming methods and decreases the computational cost considerably but is constrained by the microphone array shape and its position in relation to the inspected item. A mesh-based VRAM extension for irregularly shaped microphone arrays is developed and optimized to merge the benefits of both techniques. Its overall performance has been evaluated using two different synthetic benchmarks that feature monopolar rotating sources emitting white noise, yielding blurred beamforming maps with lower sound power levels and discouraging further investment in this direction. Furthermore, the CleanT deconvolution approach has been implemented and tested to meet the necessity of overcoming conventional beamforming limits, achieving good performances in finding the exact position of sound sources in rotating frames. Finally, all these methods have been employed to characterize the sound sources in an un-skewed low-pressure axial fan, highlighting the prevalence of leading-edge noise at lower frequencies and self-noise at higher frequencies. This latter investigation demonstrates how effective these tools are in real-world applications, resulting in a better understanding of the noise generation mechanisms underlying rotating devices.