Search Results

Inverse numerical acoustics is a method which reconstructs the source surface normal velocity from the sound measured in the near-field around the source. This is of particular interest when the source is rotating or moving, too light or too hot to be instrumented by accelerometers. The use of laser vibrometers is often of no remedy due to the complex shape of the source. The Inverse Numerical Acoustics technique is based on the inversion of transfer relations (Acoustic Transfer Vectors) using truncated Singular Value Decomposition (SVD). Most of the time the system is underdetermined which results in a non unique solution. The solution obtained by the truncated SVD is the minimal solution in the RMS sense. This paper is investigating the impact of non homogeneities in the mesh density (local mesh refinement) on the retrieved solution for underdetermined systems. It will be shown that if transfer quantities are inverted as such, big elements get a higher weight in the inversion.

Finite elements have been successfully used over the past decade to predict the vibro-acoustic behavior of complex large systems as encountered in the transportation industry. Nevertheless, some challenges are still not completely solved as for instance the modeling of multilayer porous materials used to reduce the noise in cavities. A simple model based on local impedance and added mass has been widely used in the past to model those acoustic trim materials at low frequency but shown limitations when the frequency range increases. To circumvent this limitation, approaches based on finite element formulations have been developed to model the poroelastic materials. They range from simple equivalent fluid models to complex models involving a solid phase and a fluid phase. However, those approaches require important modeling effort, computer memory and solution time. A unique approach to model acoustic trim material is presented in this paper.

Source identification applied to a truck engine and using inverse numerical acoustics is presented. The approach is based on acoustic transfer vectors (ATV) and truncated singular value decomposition (SVD). Acoustic transfer vectors are arrays of transfer functions between surface normal velocity and acoustic pressure at response points. They can be computed using boundary element methods (indirect, direct or multi-domain direct formulations) or finite element methods (in physical or modal coordinates). Regularization techniques such as the so-called L-curve approach are used to identify the optimum SVD truncation. To increase the reliability of the source identification, the approach can use velocity measurements on the boundary surface as well as the standard nearfield pressure measurements. It also allows for linear or spline interpolation of the acoustic transfer vectors in the frequency domain, to increase computational speed.

This paper presents the numerical modeling of noise radiated by an engine, using the so-called Acoustic Transfer Vectors and Modal Acoustic Transfer Vectors concept. Acoustic Transfer Vectors are input-output relations between the normal structural velocity of the radiating surface and the sound pressure level at a specific field point and can thus be interpreted as an ensemble of Acoustic Transfer Functions from the surface nodes to a single field point or microphone position. The modal counter part establishes the same acoustic transfer expressed in modal coordinates of the radiating structure. The method is used to evaluate the noise radiated during an engine run-up in the frequency domain. The dynamics of the engine is described using a finite element model loaded with a rpm-dependent excitation. The effectiveness of the method in terms of calculation speed, compared with classical boundary element methods, is illustrated.