Expert Decision Support System for Aeroacoustic Classification from Deconvolved Beamforming Maps

Abstract

We present an expert decision support system for time-invariant aeroacoustic source classification from deconvolved beamforming maps and results based on scaled airframe half-model wind tunnel measurements. The system consists of three steps: the identification of acoustic sources from the deconvolved maps, the calculation of their acoustic properties, and the clustering of the sources based on their properties. In this paper, we present and compare two methods to identify and extract source spectra from the beamforming maps. The first relies on the spatial normal distribution of CLEAN-SC results for aeroacoustic broadband sources. The second uses hierarchical clustering methods. We propose a variety of aeroacoustic features that capture the characteristics of the spectra while being independent of absolute parameters such as the Mach number. Based on these features the aeroacoustic sources are clustered using unsupervised machine learning methods to determine similar or atypical behavior. This expert support system helps aeroacoustic specialists in classifying the identified sources to support them in analyzing their typical behavior and identifying spurious sources.