ISBN 978-3-8439-2821-2

978-3-8439-2821-2, Reihe Thermodynamik

Carlo Sovardi
Identification of Sound Sources in Duct Singularities

240 Seiten, Dissertation Technische Universität München (2016), Softcover, A5

Zusammenfassung / Abstract

This dissertation deals with the acoustic characterization of ducted singularities such as orifices, valves or sudden area expansions. More precisely, the accurate assessment of the noise sources and of the acoustic scattering at those singularities in ducted flows is explored. Here, the strategy adopted is based on a system theory approach. Acoustic elements composing a generic set-up are modelled as Linear Time Invariant systems. This modelling is achieved by means of the so-called LES/SI method. At first an acoustically excited Large Eddy Simulation (LES) is performed. Subsequently, acoustic data series educed from the LES are analysed by means of System Identification (SI) methods to characterize the system dynamics. In this work, a new LES/SI approach based on prediction error methods and a Box-Jenkins (BJ) model structure is introduced. It affords a concurrent identification (and modelling) of both acoustic scattering matrix and noise sources by means of only one excited LES. The identified acoustic elements are subsequently used to build acoustic 1D network models. Hence, by combining two or more simpler acoustic elements, the possibility to analyse more complex acoustic systems is explored. Limits of applications of the network modelling as well as its capacity to reproduce noise sources for given acoustic feedbacks are analysed. The analysis is carried out for ducts in which one single orifice or two tandem orifices have been installed. These are only test cases to investigate the methods proposed here. Hence, the application of the numerical techniques here developed should not be limited to the cases considered. Numerical results from the LES/SI are validated against experiments carried out at the Laboratoire d’Acoustique de l’Université du Maine (LAUM). The accuracy of the estimations is evaluated by quantifying their uncertainties. Finally, low-order network models are used to perform parametric studies on the distance between two tandem orifices and on the acoustic reflections at the boundaries of the configurations analysed.