This thesis provides a study on volumetric acoustic source mapping with microphone array measurements. The topic has rarely been addressed in literature, despite its importance in some applications such as aeroacoustics. The aim is to remove the hypothesis of acoustic sources confined on surfaces, as it happens in common acoustic imaging approaches. This assumption may not be true and produce misleading results. After the identification of additional issues to deal with, inverse methods have been chosen to cope them. Two methods have been proposed and both are based on the method of Iteratively Re-weighted Least Squares (IRLS) to obtain sparse solutions. This approach is applied to Equivalent Source Method and Covariance Matrix Fitting, thus leading to ESM-IRLS and CMF-IRLS techniques. A tailored version of IRLS for acoustic problems has been developed in this work and is strictly linked to Bayesian Approach to inverse acoustic problems. An improved regularization strategy, rooted on Bayesian Regularization, has been developed to fulfil the needs of IRLS. Indeed, Bayesian Iterative Regularization makes IRLS able to produce accurate and reliable results. A novel use of CLEAN-SC as decomposition tool of Cross-Spectral Matrix is proposed and compared with the standard Eigenmode Decomposition, when combined with inverse methods. Methods proposed have been validated on simulated test cases that represent the conditions of standard and volumetric mapping. Also validation on experimental data is provided. The first is an airfoil in open jet, which is mapped with single planar array. The second is an aircraft model in wind tunnel, where a comparison between the use of one or two planar arrays is performed. This work aims at showing the feasibility of volumetric acoustic mapping with inverse methods, despite its intrinsic difficulty. A detailed discussion on theoretical hypothesis and algorithmic tricks necessary to achieve this result is provided.
Questa tesi fornisce uno studio riguardante mappature volumetriche di sorgenti acustiche, mediante misure con array di microfoni. Questo argomento non è approfondito in letteratura, nonostante la sua importanza in alcune applicazioni (e.g. aeroacustiche). L’obiettivo è quello di rimuovere l’ipotesi di sorgenti confinate su una superficie, tipica dei comuni metodi di acoustic imaging. Infatti, questa ipotesi può non essere vera e produrre risultati fuorvianti. Dopo aver identificato le problematiche ulteriori che comporta la rimozione di questa ipotesi, i metodi inversi sono stati scelti per far fronte ad esse. Le tecniche proposte si basano sull’Iteratively Re-weighted Least Squares per ottenere soluzioni sparse. Quest’ultimo è combinato con l’Equivalent Source Method ed il Covariance Matrix Fitting, generando così le tecniche denominate ESM-IRLS e CMF-IRLS. Una versione dell’IRLS specifica per i problemi acustici è stata sviluppata ed è strettamente legata all’Approccio Bayesiano per problemi acustici inversi. Inoltre, è stata sviluppata una strategia di regolarizzazione che soddisfa le necessità dell’IRLS. La Regolarizzazione Bayesiana Iterativa permette all’IRLS di ottenere risultati accurati ed affidabili. Viene proposto un nuovo utilizzo del CLEAN-SC come strumento di decomposizione della matrice dei cross-spettri. Tale metodo è comparato con la tipica Eigenmode Decomposition, una volta combinate con i problemi inversi. Le tecniche proposte sono state validate su test simulati rappresentativi delle condizioni di mappatura standard e volumetrica. Anche una validazione su dati sperimentali viene fornita. Il primo caso sperimentale è un profilo alare in un getto d’aria mappato con un singolo array piano. Il secondo è un modello di aereo in galleria del vento. Quest’ultimo viene utilizzato per valutare l’utilizzo di uno o due array combinati. L’obiettivo di questo lavoro è di dimostrare la fattibilità di mappe acustiche volumetriche, nonostante l’intrinseca difficoltà del problema. Viene fornita una discussione dettagliata sulle ipotesi teoriche e accorgimenti necessari per raggiungere questo risultato.
Inverse Methods for Three-Dimensional Volumetric Acoustic Mapping / Battista, Gianmarco. - (2019 Feb 25).
Inverse Methods for Three-Dimensional Volumetric Acoustic Mapping
BATTISTA, GIANMARCO
2019-02-25
Abstract
This thesis provides a study on volumetric acoustic source mapping with microphone array measurements. The topic has rarely been addressed in literature, despite its importance in some applications such as aeroacoustics. The aim is to remove the hypothesis of acoustic sources confined on surfaces, as it happens in common acoustic imaging approaches. This assumption may not be true and produce misleading results. After the identification of additional issues to deal with, inverse methods have been chosen to cope them. Two methods have been proposed and both are based on the method of Iteratively Re-weighted Least Squares (IRLS) to obtain sparse solutions. This approach is applied to Equivalent Source Method and Covariance Matrix Fitting, thus leading to ESM-IRLS and CMF-IRLS techniques. A tailored version of IRLS for acoustic problems has been developed in this work and is strictly linked to Bayesian Approach to inverse acoustic problems. An improved regularization strategy, rooted on Bayesian Regularization, has been developed to fulfil the needs of IRLS. Indeed, Bayesian Iterative Regularization makes IRLS able to produce accurate and reliable results. A novel use of CLEAN-SC as decomposition tool of Cross-Spectral Matrix is proposed and compared with the standard Eigenmode Decomposition, when combined with inverse methods. Methods proposed have been validated on simulated test cases that represent the conditions of standard and volumetric mapping. Also validation on experimental data is provided. The first is an airfoil in open jet, which is mapped with single planar array. The second is an aircraft model in wind tunnel, where a comparison between the use of one or two planar arrays is performed. This work aims at showing the feasibility of volumetric acoustic mapping with inverse methods, despite its intrinsic difficulty. A detailed discussion on theoretical hypothesis and algorithmic tricks necessary to achieve this result is provided.File | Dimensione | Formato | |
---|---|---|---|
Tesi_Battista.pdf
accesso aperto
Descrizione: Tesi_Battista.pdf
Tipologia:
Tesi di dottorato
Licenza d'uso:
Creative commons
Dimensione
71.99 MB
Formato
Adobe PDF
|
71.99 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.