Abstracts Submitted: 464
Number of Users: 528
View Abstracts Submitted
Back to home Page
Acoustic holography is widely used in automobile, aerospace and defense applications for correctly pinpointing noise sources. It also helps in characterizing noise regions of interest in terms of frequency and amplitude level to understand its dominance. In recent years, Near-field Acoustic Holography (NAH) technique has gained more importance among other acoustic localization techniques since it captures rapidly decaying evanescent waves in near-field of source thus producing a high resolution source image. Hence high resolution localization of nearby sources is easily possible using NAH but quantification of the source is a concern due to inversion of evanescent waves which is focused upon in this work. The amplitude quantification of source is a issue in general when dealing with inverse problems. In NAH, the ill-conditioning of data matrix occurs leading to amplification of source levels because data on measurement plane is back propagated towards source plane. The noise associated smaller singular values tend to amplify and burst-off the solution when measurement plane data is inversed. Hence a k-space low pass filter called Tikhonov Regularization is used to suppress the high wavenumber noise components from data matrix to avoid amplification of source levels. The level of filtering to be done is also an important aspect since too much filtering would tend to affect localization by smoothening the solution and too less filtering would lead to amplification of source levels. For correct choice of regularization parameter that serves as an input to Tikhonov regularization, a new approach is devised. The new parameter estimation method is based on suitable choice of singular value, calculated from decomposing transfer function matrix that links measurement and source plane data, which serves a best trade-off for effective localization and amplitude quantification. This study aims to correctly localize and quantify a sound source of interest by solving the inverse problem associated with NAH. The work presents a novel approach to tackle ill-conditioning associated with data inversion by choosing a suitable regularization parameter for avoiding deterioration of solution matrix. This method is compared with existing and widely used parameter estimation methods like Generalized Cross Validation (GCV) and L-Curve and found to perform well if both localization and quantification is of interest.
© Copyright 2017 All Rights Reserved