Spectral decomposition of a turbulence-excited vibroacoustic system

The applicability of the spectral decomposition method to noise spectra generated inside a cavity enclosed by turbulence-excited elastic structures is investigated. Based on previous theoretical and experimental findings, we show that the noise spectra may be spectrally decomposable if the convection speed of the boundary layer is relatively slow, and thus all resonant structural modes are hydrodynamically fast. As an application, spectral decomposition is attempted for data obtained from two ground vehicles tested in an aeroacoustically treated wind tunnel. We show that the synthesized noise spectra generated from the decomposed source and filter functions are in fairly good agreement with the measured spectra.This study is also concerned with the spectral decomposition algorithm. Similar to an algorithm recently proposed by other researchers, the proposed algorithm formulates the spectral decomposition as simultaneous equations. Input data generated from numerical simulations are used to compare the various algorithms in terms of computational accuracy and cost. Unlike conventional methods that calculate the decomposed function incrementally, we show that algorithms based on simultaneous equations yield far better results in terms of accuracy at the expense of increased computational memory. However, the proposed algorithm requires significantly fewer equations and unknowns compared to the recently proposed algorithm that is based on a formulation of simultaneous equations.