Technische Akustik


Alle News

Methods for analyzing and characterizing sound sources in rotating systems

Simon Jekosch has successfully defended his PhD thesis "Methods for analyzing and characterizing sound sources in rotating systems". 



Sound generation by rotating machinery is oftentimes perceived as noise. Microphone array methods that rely on rotating beamforming are a versatile method to identify and locate these noise sources with high accuracy. To accurately analyze acoustic sources moving in a circular path, specific algorithms must be employed to account for the movement of the rotating components which would otherwise be represented as a stationary ring.

​There are two approaches to compensate for the motion of the rotating parts in acoustic source analysis: moving the computational grid or the array sensors along the source path. While the latter method is faster to compute, it has two limitations: the microphone geometry must be arranged in a ring around the center of rotation, and no changes in the source emission can be resolved over time. 

​To overcome the limitations of the sensor placements in analyzing moving acoustic sources, this work presents several spatial interpolation techniques and a filtering approach based on angle intervals to account for periodically changing signals. Additionally, a method for computing the source directivity in the rotational reference frame is proposed, and an extended ray tracing algorithm is presented to account for arbitrary time-invariant flow fields.

​The capabilities of the proposed algorithms are tested with synthetic test cases of array measurements. The evaluation of those cases illustrates the general operating principles. To demonstrate their practical application three experimental campaigns were conducted. The first two campaigns involved a five-blades fan and utilized spiral and ring array geometries. The third campaign involved a rotating speaker array with steerable source directivity and a spiral array geometry.

The experimental evaluation demonstrates that the proposed methods are capable of localizing and separating sound sources with reasonable accuracy. Moreover, they are able to assign sources that periodically change their level during rotation according to their angle in the rotating reference system and to reconstruct their directional characteristics. The integration of the sound propagation taking into account the rotating sound bearing medium as well as the superposed flow helps to improve the source localization quality. Overall, these methods reduce the limitations of using frequency domain array methods for rotational noise detection in experimental settings.

© Simon Jekosch