Analysis and Synthesis of Aicraft Engine Fan Noise for Use in Psychoacoustic Studies

Abstract

Community noise impact is an important factor in design of current generation aircraft, especially when considering projected trends in flight volume and urbanization. Simulation is a useful tool to evaluate the human annoyance response due to both current and proposed aircraft, and it has some advantages over field studies or playback of recordings. However, current simulation methods which are based on time-averaged prediction methods do not include short term fluctuations observed in recordings of real aircraft engines. Those fluctuations in both tonal and broadband sources provide psychoacoustic clues to listeners when evaluating flyover noise realism. When those short-term fluctuations are not included, simulation realism may suffer and evaluation results might not be applicable to real aircraft.

This thesis presents work to analyze and model fluctuations in aircraft engine fan noise, using an existing set of static turbofan engine recordings. The inclusion of the observed fluctuations, which are unaccounted for in many current prediction and simulation routines, was expected to increase the perceived realism of simulated flyover events. The analysis of tonal fluctuations was performed by utilizing the complex-valued analytic signal to extract instantaneous amplitude and frequency. A simple parametric model was developed to represent each measured fluctuation using its spectral bandwidth and variance. The model was then used to generate new fluctuations which were perceptually similar to the original. Tonal synthesis was performed as the sum of many amplitude- and frequency-modulated tones. Analysis was also performed on the broadband fan noise component, which used output from the Short-Time Fourier Transform was used to characterize fluctuations in third-octave band SPL. Those fluctuations were not modeled as in the case of tonal fluctuations and were directly reproduced using an overlap-add synthesis tool.

A subjective listening test was then conducted to evaluate the perceptual similarity between synthesized and recorded fan noise. That test concluded that synthesized tonal noise which included short-term fluctuations was perceived as more realistic than noise without. It also concluded that the addition of broadband fan noise components tended to mask tonal fluctuations.