This work studies differences in sound generation and radiation between incompressible and compressible flowfields; at the corresponding Mach number of 0.15 by evaluating the two dominant dipole terms in a temporal form of Curie's equation. The present work adds incompressible simulation results to a previously reported direct simulation of an open two-dimensional cavity oscillating in wake mode. The length-to-depth ratio of the cavity is L/D = 4, and the flow is considered to be laminar. The Reynolds number based on the cavity, depth is Re(D) = 1500. At these conditions, the flow is characterized by large variations in the streamwise force coefficient, showing similarities to a bluff-body wake flow, which makes the present case an attractive candidate for an incompressible approach. The root-mean-square levels of the two acoustic source terms show good agreement in the vicinity of the cavity, and the agreement in the radiated sound is almost perfect when isolating the cavity walls in the surface integration. The agreement in the radiated sound for the two flowfields is equally good when isolating the downstream wall extending from the cavity trailing edge and 10D downstream. However, when the surface integration comprises both the cavity walls and the downstream wall, a major mismatch in the directivity arises, and it is shown that this mismatch is primarily caused by an almost complete cancellation of the sources at the cavity bottom and the downstream wall in the incompressible simulations.

The case investigated is the flow past a generic side mirror mounted on a flat plate at the Reynolds number of Re(D)=5.2x10(5) based on the mirror diameter. The present work studies both flow and acoustic sources by evaluating two second-order advection schemes, different levels of turbulence modeling, and three different grids. The advection schemes discussed in the present study are a second-order upwind scheme and a monotonic central scheme. The turbulence models investigated cover three levels of modeling. These are the original formulation of the detached eddy simulation (DES) model, the Smagorinsky-Lilly sub-grid scale (SGS) model with near-wall damping, and a dynamic Smagorinsky model. The different grids are as follows: a primary grid where all parameter studies are conducted and a second grid with significantly higher wake resolution and to some extent also increased plate resolution, while maintaining the resolution at the front side of the mirror. The final grid uses a significantly higher plate resolution and a wake resolution similar to that of grid two, but a comparably lower mirror front side resolution as compared with the two other grids. The general outcome of this work is that the estimation of the grid cutoff frequency through a relation of the velocity fluctuation and the grid size matches both the experimental results and trend lines perfectly. Findings from the flow field show that the horseshoe vortex in front of the mirror causes pressure fluctuations with a magnitude exceeding the maximum levels at the rear side of the mirror. Its location and unsteady properties are perfectly captured in the final simulation as compared with the experiments conducted by Daimler-Chrysler. A laminar separation at the front side of the mirror is more or less found for all wall resolved cases except the DES simulation. The third grid fails to predict this flow feature, but it is shown that this effect has no significant effect on either the static pressure sensors at the mirror surface or at the dynamic sensors located downstream of the mirror. The simulation also supports the fundamental frequency based on the eddy convection in the mirror shear layer, which is shown to be twice as high as the frequency peak found in the lateral force spectra.

Accurately predicting both average flow quantities and acoustic sources at the front window of today's ground vehicles are still a considerable challenge to automotive companies worldwide. One of the most important aspects in terms of obtaining not only trustworthy results but also the most tedious one and therefore perhaps overlooked, is the control and outcome of the mesh generation process. Generating unstructured volume meshes suitable for large eddy simulations with high level representation of geometrical details is both a time consuming and an extremely computer demanding activity. This work investigates two different mesh generation processes with its main aim to evaluate their outcome with respect to the prediction of the two dominating dipole sources in a temporal form of the Curle's equation. Only a handful of papers exists that report a high level representation of the vehicle geometry and the aim of predicting the fluctuating exterior noise sources. To the author's knowledge no studies have been conducted in which both these source terms are evaluated quantitatively against measurements. The current paper investigates the degree to which the amplitude of these two source terms can be predicted by using the traditional law-of-the-wall and hex-dominant meshes with isotropic resolution boxes for a detailed ground vehicle geometry. For this purpose, the unstructured segregated commercial FLUENT finite volume method code is used. The flow field is treated as incompressible and the Smagorinsky-Lilly model is used to compute the subgrid stresses. Mean flow quantities are measured with a 14 hole probe for 14 rakes downstream of the side mirror. The dynamic pressure sensors are distributed at 16 different positions over the side window to capture the fluctuating pressure signals. All measurements in this work were conducted at Ford's acoustic wind tunnel in Cologne. All three simulations accurately predict the velocity magnitude closest to the window and downstream of the mirror head recirculation zone. Some variations in the size and shape of this recirculation zone are found between the different meshes, most probably caused by differences in the detachment of the mirror head boundary layer. The Strouhal number of the shortest simulation was computed from the fundamental frequency of the drag force coefficient. The computed Strouhal number agrees well with the corresponding results from similar objects and gives an indication of an acceptable simulation time. The dynamic pressure sensors at 16 different locations at the vehicle side window were also used to capture the levels of the two dipole source terms. These results are compared with the three simulations. With the exception of three positions, at least one of the three simulations accurately captures the levels of both source terms up to about 1000 Hz. The three positions with less agreement as compared with measurements were found to be in regions sensitive to small changes in the local flow direction. [DOI: 10.1115/1.4001340]

The investigated case is the o w past a generic side mirror mounted on a at plate at the Reynolds number of ReD = 5:2 10 5 based on the mirror diameter. The o w eld over the mirror is identied as sub-critical which makes LES an attractive candidate to the DES model used by some of the previous researchers. The present paper is a rst attempt to investigate the o w and acoustics of a generic side mirror at the Reynolds number of ReD = 5:2 10 5 by comparing DES and LES results using identical grid, second order schemes and the incompressible assumption. Two dieren t cases are investigated for this purpose. The rst case uses the hybrid DES-modelling technique combined with a "Trip-Less" approach proposed by Shur et al. and the second case is a partially resolved LES-simulation using the dynamic Smagorinsky model. The sound radiation for the present simulations is predicted from the Ffowkes-Williams and Hawkings analogy where the wall pressure uctuations and the pressure temporal derivative are the two source terms treated. The result shows good agreement for both cases in wall source levels even with the rather coarse mesh used in the mirror wake. The trends are however in slightly better agreement for the LES case than the DES case compared to the corresponding measured levels. The radiated sound for the DES case agrees signican tly better with the measured results compared to the LES results. The overprediction in radiated sound observed in the LES case is argued to primarily be caused by unresolved uctuations included in the surface integration. The DES model on the other hand is far more dissipative and reduces pollution of numerical noise in the signals. For an automotive application the wall pressure level results is by far the most important due to the interest in the driver and passengers subjective experience. Thus, the conclusion is that the LES approach is a better candidate compared to DES even though the mesh resolution over the plate by no means fullls the traditional restrictions in LES.

In this work a modified version of the Lighthill-Curle's analogy is applied to study the near field acoustics of an upstream laminar flow past an open cavity. Three incompressible cases have been computed and are compared against the corresponding compressible results. The three incompressible cases are carried out with different time-step sizes, distances from the cavity trailing edge to the outlet and spatial resolution in the streamwise direction. The aim of the work is to study the differences in compressible and incompressible sources in Lighthill-Curle's equation and their influence on the sound radiated. To cite this article: J. Ask, L. Davidson, C. R. Mecanique 333 (2005). (c) 2005 Academic des sciences. Published by Elsevier SAS. All rights reserved.