The research team in acoustics at the École Centrale de Lyon was founded 50 years ago by Prof. Geneviève Comte-Bellot. In this paper we describe the growth of this team, now known as the Acoustic center of the École Centrale de Lyon (or Centre Acoustique in French), from the early 1970s to the present day. We highlight the evolution of the research interests and experimental facilities, and provide a selection of “historical” references at the end of the paper. The main current research topics are listed and illustrated, and are complemented by a partial list of recent references.

This paper presents the modeling, numerical simulations and experimental results of a differential microphone array and its associated post-processing methods. This study combines the development of an array geometry and the definition of a post-processing method adapted to applications within an aircraft. The research originality lies in the considered array and post-processing method, based on the approximation of the pressure gradient by a differential microphone calculation and linear combination of the signals. As the targeted frequency range of interest is from 200 Hz up to 5000 Hz, a fourth-order linear array is chosen, in order to handle at the same time the space constraints due to the aircraft environment and the inter-microphone spacing linked to the frequency range. The frequency band target is used to drive the antenna geometry definition. An experimental proof of concept is tested in an anechoic room. Simulations and experimental results are compared to an analytical model. The comparison highlights the method advantages as well as the experimental limitations of the designed linear array. Then, a first localization test is performed and finally, a modification of the post-processing method is suggested, to improve the directivity of the array low frequency.

In the first part of this paper, the authors offer, based on their experience, a short review of the main difficulties encountered in measuring flow-induced wall-pressure fluctuations. Some recent advances are presented and illustrated with a focus on 2-point statistical quantities, spatial cross-correlations and wavenumber-frequency spectra. The second part describes three experiments conducted at Ecole centrale de Lyon on wall-pressure measurements and their use. The SONOBL experiment is devoted to the study of the influence of mean external pressure gradients on point-spectra and cross-spectra with the objective of identifying the acoustic contribution of a turbulent boundary layer (TBL). The CANOBLE experiment is focused on measurements of TBL wall-pressure fluctuations, induced vibrations of a representative side panel and acoustic transmission into a cavity performed on a full-scale mock-up of a business jet. The final objective is to predict flow-induced noise into an aircraft cockpit or cabin in cruise conditions. Finally measurements made inside the duct of a small turbofan used in aircraft ventilating systems are described, where array techniques involving MEMS microphones are used to extract the noise emitted by the fan from contaminating hydrodynamic wall-pressure fluctuations.

Upon firing, most weapons emit very loud sounds. These sounds propagate over the battlefield and are distorted by the atmospheric effects, absorbed by the ground, reflected on or diffracted around buildings or mountains. It is of obvious operational interest to develop sensing systems to localize these sound sources, with arrays of distributed sensors. This study develops an original sensing approach. It uses the time-matching method, based on finding the best match between pre-calculated times of arrivals (TOAs) of the shot sounds and measured TOAs from a set of synchronous, distributed sensors. Predicting the TOAs requires a physical model able to factor in the impact of complex 3-D environments (wind and sound speed gradients, obstacles), and of complex sound sources (e.g., combination of muzzle blast and supersonic projectile wave). A very fast interface-tracking model is used, based on Sethian's Fast-Marching method, for pre-calculating the TOAs in a general and comprehensive framework. Applications to localization of shots in urban environments and to localization of long range artillery gun are presented. They demonstrate that, compared to standard methods, the above matching-and-marching approach can work without classification, with less sensors, or with a much smaller baseline array.

For controlling the noise radiated from vibrating structures excited by turbulent boundary layer (TBL) it is relevant to develop numerical tools for understanding how the structure reacts to TBL excitation. Usually, the wall pressure fluctuations of the TBL are described through statistical quantities (i.e. space-frequency or wavenumber-frequency spectra) which depend on the TBL parameters. On the other hand, the vibro-acoustic models (i.e. Finite Elements, Boundary Elements, Transfer Matrix Methods, Analytical models, etc.) evaluate deterministic transfer functions which characterise the response of the considered structures. The first part of this paper focuses on the coupling between the stochastic TBL and the deterministic vibro-acoustic models. Five techniques are presented. Numerical applications on an academic marine test case are proposed in order to discuss the calculation parameters and the interests/drawbacks of each technique. In the second part of the paper, the high frequency modelling with the Statistical Energy Analysis (SEA) method is considered. The focus is placed on the estimation of an important input of this method: the injected power by the TBL into the structure for each third octave band.

Acoustic recordings of artillery shots feature the signatures of the shot's muzzle, projectile, and impact waves modulated by the environment. This study aims at improving the sensing of such shots using a set of synchronous acoustic sensors distributed over a 1 km2 area. It uses the time matching approach, which is based on finding the best match between the observed and pre-calculated times of arrivals of the various waves at each sensor. The pre-calculations introduced here account for the complex acoustic source with a 6-degrees-of-freedom ballistic trajectory model, and for the propagation channel with a wavefront-tracking acoustic model including meteorological and terrain effects. The approach is demonstrated using three recordings of artillery shots measured by sensors which are more than 10 km from the point of fire and distributed at several hundred meters away from and around the target points. Using only the impact wave, it locates the impact point with an error of a few meters. Processing the muzzle and impact and projectile waves enables the estimation of the weapon's position with a 1 km error. Sensitivities of the localization method to various factors such as the number of sensors, atmospheric data, and the number of processed waves are discussed.

An experimental study of the near wake of a simplified truck model with an aspect ratio between the height and the width greater than one is presented. The influence of the underbody velocity at a constant ground clearance height is considered. The evolution of the model base pressure and of the near wake as a function of the underbody velocity permits to identify four classes of flow. For large values of underbody velocity, typically above 60% of the free stream velocity, the near-wake structure is similar to what obtained in bluff-body characterizations where the underbody flow momentum is sufficient to prevent its detachment from the ground. For smaller values of underbody velocity, of particular interest when considering real truck applications, base pressure mean value and near wake characteristics strongly depend on the underbody velocity; three different classes are defined through the values of time-averaged rear pressure as well as a qualitative analysis of the near-wake structure. Quantitatively, a momentum budget in the near-wake together with the characterization of the curvature of the underbody flow near the model end provide ad-hoc indicators for discriminating between the different flow classes.

Sound characteristics close to the ground strongly depend on the atmospheric and ground properties in terms of amplitude, shape, and time of arrival. Time-domain numerical modeling is able to accurately account for outdoor sound propagation and is valuable to decipher the interactions between the refractive, scattering, and ground effects. It remains computationally challenging for three-dimensional long range simulations. A high-order parallel Finite-Difference Time-Domain solver with the moving frame approach is presented, featuring accurate time-domain impedance boundary conditions and very efficient convolutional perfectly matched layers to artificially truncate the computational domain. The design of the absorbing boundaries is based on a stability analysis of the time integration scheme and is shown to optimize the absorption properties for grazing waves. The model allows for accurate propagation of impulse sounds in 3D over several hundreds of meters and up to 1000 Hz using a personal laptop with a simulation duration of a few hours. The numerical predictions are compared to experimental measurements under different weather conditions, with a special focus on the sensitivity to the mean vertical wind profile and the ground properties.

Wall-pressure fluctuations beneath a turbulent boundary layer have been measured in a closed wind tunnel with an inclined upper wall, in order to impose an external mean pressure gradient. Three flow configurations have been studied : the classical zero pressure gradient case (ZPG), and the case of favorable and adverse pressure gradients (noted FPG and APG, respectively). The input mean velocity was varied from 10 to 70m/s (10 to 100m/s for ZPG). A specially designed rotating linear array of 63 remote microphones and a reference pinhole microphone have been used for unsteady wall-pressure measurements. Point-spectra and cross-spectra between the microphones of the array have been measured. 2D wavenumber-frequency spectra have then been deduced from a spatial Fourier transform of the cross-spectra. Typical results for frequency spectra and 1D and 2D wavenumber spectra are shown, illustrating the effect of an external mean pressure gradient.

This experimental study deals with wake-flow fluidic control behind a two-dimensional square back geometry positioned close to the ground. The fluidic control system is made of pulsed jets positioned at the upper edge of the model base. The objective of the fluidic action is to modify the wake-flow development, and as a consequence the static pressure distribution over the model base and hence the pressure drag. The main concern of this study is to determine to what extent the presence of a flow confined between the model and the floor influences the effectiveness of the control. Static pressure measurements at the model base and wake-flow characteristics derived from Ply measurements at a high acquisition frequency indicate global similarities between a case where an underbody flow exists and a case where this underbody flow is absent. For low actuation frequencies, discrepancies in the way the coherent structures due to the control develop in the shear layer appear. (C) 2014 Academie des sciences. Published by Elsevier Masson SAS. All rights reserved.

The effect of active flow control combining synthetic jets and inclined flaps on the flow behind a 1:8 scale simplified truck model is experimentally studied. Aerodynamic drag and base pressure measurements show that forcing the flow within a given range of actuation frequencies allows reducing the drag. However, results also show that such drag reductions greatly depend on the underside flow velocity.

A rigorous theory of interference of direct and ground-reflected waves in a turbulent atmosphere was developed by Clifford and Lataitis (JASA, 73, 1545-1550, 1983). The primary goal of the present paper is to generalize Clifford and Lataitis’s theory by taking into account developments in atmospheric acoustics since 1983. In particular, we reformulate the geometry of the problem that allows us to use the parabolic equation method. Furthermore, we take into account that the direct and ground reflected waves are scattered by both temperature and wind velocity fluctuations, and use Kolmogorov, Gaussian, and von Karman spectra to model spectra of these fluctuations. Finally, we obtain a formula for the mean squared sound pressure and relate it to that of the coherence function of a spherical sound wave for line-of-sight propagation. 1 INTRODUCTION In many cases of sound propagation in the atmosphere, a source and receiver are located close the ground. If the horizontal distance L between the source and receiver is not too large, the effects of atmospheric refraction can be ignored and the sound field p at the receiver is a sum of the direct wave and that reflected from impedance ground. In the absence of atmospheric turbulence, the interference between these two waves could lead to very deep minima in the amplitude of the resulting sound field. However, temperature and wind velocity fluctuations always exist in the atmosphere. These fluctuations lead to partial or complete loss of coherence between the direct and ground reflected waves. As a result, the mean squared sound pressure 〈 |p| 〉 at the interference minima can be increased by several dozen dB in comparison to that in a nonturbulent atmosphere. This phenomenon has been studied since the 1960s. A first rigorous approach for calculating the interference of the direct and ground reflected waves in a turbulent atmosphere was developed by Clifford and Lataitis [1]. In this reference, it was assumed that atmospheric turbulence is caused by temperature fluctuations. Furthermore, log-amplitude and phase fluctuations of the direct and ground reflected waves were calculated by using the Rytov method. Finally, an analytical formula for 〈 |p| 〉 was derived. Using this formula, 〈 |p| 〉 was calculated for a Gaussian spectrum of temperature fluctuations. These results have been widely used in subsequent research. The main goal of the present paper is to generalize the theory developed by Clifford and Lataitis [1] in the following directions. First, we reformulate the geometry of the problem. A new geometry allows us to use the parabolic equation method, which has a wider range of applicability than the Rytov method. Secondly, we assume that atmospheric turbulence is caused by both temperature and wind velocity fluctuations since in most cases the effects of velocity fluctuations on 〈 |p| 〉 are greater than the effects of temperature fluctuations (e.g. [2]). Finally, we derive analytical formulas for 〈 |p| 〉 for Kolmogorov, Copyright SFA InterNoise 2000 2 Gaussian, and von Karman spectra of temperature and wind velocity fluctuations. Using these formulas, the effects of atmospheric turbulence on 〈 |p| 〉 are studied numerically. 2 PARABOLIC EQUATION Let the source and receiver be located at the heights hs and hr above the ground. We will use the Cartesian coordinate system ~ R = (x, y, z) with the center at the source, z -axis directed upward, and z -axis in the direction from source to receiver. In this case, the plane z=−hs coincides with the surface of the ground. We assume that the mean wind velocity is zero and the mean temperature T 0 does not depend on z. Temperature and wind velocity fluctuations in the atmosphere are denoted as T ′ ( ~ R ) and ~ν (R′). For this geometry, calculations of the statistical moments of a sound field are complicated by the presence of the surface of the ground. Therefore, we reformulate this geometry. In a new geometry, sound propagates in an unbounded turbulent medium which is symmetrical with respect to the plane z=−hs. Furthermore, the sound field at the receiver is a sum of the sound field due to the source and that due to an image source of the strength Q located at the point ~ Ri = (0, 0,−hs). (The image source is symmetrical to the real source with respect to the plane z=−hs.) Here, Q is the spherical-wave reflection coefficient given by [3]: Q = 2β [ 1 + i √ πde−d 2 erfc (−id)− β + (hs + hr) /R2 ] β + (hs + hr) /R2 , (1) where the distance from the image source to the receiver is R2 = √ (hs + hr) 2 + L2, the ground admittance is β, the complementary error function is erfc, and the numerical distance is d = [β + (hs + hr) /R2] √ ikR2/2 where k is the sound wavenumber. We assume that fluctuations in temperature and wind velocity are relatively weak so that sound backscattering can be ignored. Then, for a new geometry, the sound pressure p at the receiver satisfies the following parabolic equation [2]: [ 2ik ∂ ∂x + ( ∂ ∂y2 + ∂ ∂z2 ) + 2k (1 + ε/2) ] p = 0. (2) Here, ε ( ~ R ) = −T ′ ( ~ R ) /T0 − 2νx ( ~ R ) /c0 is a linear combination of temperature and wind velocity fluctuations. In this formula, c0 is the reference value of the sound speed and vx is the x -component of ~ν. Note that the random field ε ( ~ R ) is symmetrical with respect to the plane x=−hs. In the plane x = 0, the sound field p satisfies the following initial condition: p (~r) = (2iπ/k) [δ (~r − ~r1) + Qδ (~r − ~r2)] , (3) where ~r = (y, z), and δ is the delta function. Furthermore, ~r1 = (0, 0) and ~r2 = (0,−2hs) are the coordinates of the source and image source in the plane x = 0. It can be shown that Eqs. (2) and (3) are equivalent to initial equations describing sound propagation over the ground and boundary conditions for two cases: sound propagation over hard boundary in a turbulent atmosphere and sound propagation over impedance boundary in a nonturbulent atmosphere. Using physical reasoning, one would expect that Eqs. (2) and (3) approximately describe sound propagation over impedance boundary in a turbulent atmosphere if the turbulence is relatively weak. A parabolic equation (2) with the initial condition (3) are the starting equations in our approach for calculating the mean squared sound pressure 〈 |p| 〉 . Equations (2) and (3) can also be used for calculating other statistical moments of a sound field propagating over impedance ground, for example, the coherence function. Note that parabolic equations have been found to be very convenient for studies of wave propagation in random media (e.g. [4]). Many classical methods for calculating the statistical moments of a sound or electromagnetic field are based on parabolic equations. Therefore, we expect Eqs. (2) and (3) to be very useful for studies of sound propagation over impedance ground in a turbulent atmosphere. 3 FORMULA FOR 〈 |p| 〉 The solution of Eq. (2) can be sought in the following form Copyright SFA InterNoise 2000 3 p = p1exp (χ1 + iφ1) + p2exp (χ2 + iφ2) . (4) Here, p1 and p2 are the sound fields due to the source and image source in a nonturbulent atmosphere. Furthermore, χ1 and φ1 are the log-amplitude and phase fluctuations of a sound field emitted by a source in a turbulent atmosphere, while χ2 and φ2 are the log-amplitude and phase fluctuations of a sound field emitted by an image source. We assume that χ1, φ1, χ2 and φ2 have Gaussian distributions. With the use of Eq. (4), we calculate the mean squared sound pressure: 〈 |p| 〉 = 1 R2 1 + |Q| R2 2 + 2 |Q|C R1R2 cos [(R2 −R1) k + Ω + 〈χ1φ2〉 − 〈χ2φ1〉] . (5) Here, the brackets 〈〉 denote the ensemble average, Ω is the phase of Q so that Q = |Q| e, and C is given by C = exp [ 〈χ1χ2〉+ 〈φ1φ2〉 − 1 2 (〈 χ 1 〉 + 〈 χ2 〉 + 〈 φ1 〉 + 〈 φ2 〉)] . (6) Equation (5) describes interference between the sound waves emitted by a source and an image source. The factor C in Eq. (5) characterizes the coherence between these waves and hereafter will be called the ”coherence” factor. In a nonturbulent atmosphere C = 1, while in a turbulent atmosphere C is less than 1. Equations (5) and (6) contain second statistical moments of phase and log-amplitude fluctuations, e.g. 〈 |p| 〉 . In order to calculate these statistical moments, we need to know expressions for χ1, φ1, χ2 and φ2 in terms of the random field ε. These expressions are found by substituting Eq. (4) into Eq. (2) and using the first Rytov approximation. Then, the statistical moments of phase and log-amplitude fluctuations in Eqs. (5) and (6) are calculated. As a result, we obtain that in Eq. (5) the difference 〈χ1φ2〉 − 〈χ2φ1〉 is less than π/2 and can be omitted. Furthermore, the following formula is obtained for the coherence factor C : C (h) = exp [ − L 4h ∫ h 0 [b (0, 0)− b (0, z)] dz ]

The various methods to obtain 1-point and 2-point statistical properties of wall-pressure fluctuations from CFD are described and discussed. If only averaged flow quantities are available through Reynolds Averaged Navier Stokes computations, empirical models or sophisticated statistical modeling have to be used to estimate wall-pressure spectra and spatial correlations. While very useful at design stage, their applicability to complex flows or geometries seems quite limited. Considering the rapid growth of computational power, it seems clear that the main pathway for the near future is to rely on time-dependent flow simulations, typically Large Eddy Simulations, and to estimate the pressure statistics through a posteriori signal processing. It seems also possible, at the moment only for relatively high Mach number flows, to estimate not only the hydrodynamic part but also the tiny acoustic contribution. Examples of computations of this acoustic contribution to wall-pressure are given together with related experiments.

The goal of this experimental study is to investigate the wall pressure wavenumberfrequency spectra induced by a turbulent boundary layer in the presence of a mean pressure gradient. The mean pressure gradient is achieved by changing the ceiling angle of a rectangular channel flow. Wall pressure spectra are measured for zero-, adverse- and favorablepressure-gradient boundary layers by using a pinhole microphone in conjunction with a high-frequency-calibration procedure. A linear antenna based on a non-uniform distribution of remote microphones mounted on a rotating disk is also developed to obtain a direct measurement of both aerodynamic and acoustic components of wavenumber-frequency spectra. First results, comparisons and analyses are then discussed.

La prediction du bruit rayonne par des structures excitees par un ecoulement turbulent represente un enjeu en aero et hydro-acoustique. Une methodologie en trois etapes est developpee afin de determiner la reponse vibro-acoustique d'une structure excitee par une couche limite turbulente : Dans un premier temps l'ecoulement turbulent est modelise avec une methode stationnaire RANS qui permet d'obtenir les parametres caracteristiques de la couche limite formee sur la structure. L'excitation dynamique generee par cette couche limite est ensuite calculee a partir de modeles semi-empiriques des fluctuations de pression parietale. Finalement, ces spectres de pression parietale sont introduits dans un modele vibro-acoustique de la structure pour en deduire sa reponse. Dans ce papier, on s'interesse plus particulierement aux deux premieres etapes. On etudie les modeles de fluctuations de pression parietale qui peuvent traduire au mieux l'excitation induite par la couche limite turbulente, et on cherche a relier les valeurs des differents parametres de ces modeles a des quantites que l'on peut estimer a partir d'un code RANS. Cette methode se distingue de la plupart des travaux precedents, ou la premiere etape consistait a effectuer un calcul potentiel de l'ecoulement et a injecter les resultats dans un modele de couche limite. Les cas de validation retenus pour cette phase correspondent a deux series d'experiences realisees a l'Ecole Centrale de Lyon dans le cadre du projet europeen ENABLE et du projet ANR SONOBL. Ces experiences mettent en jeu un ecoulement dans une veine fermee avec un gradient de pression statique nul, favorable ou adverse, induit par une hauteur de conduite variable (paroi superieure profilee dans le cas du projet ENABLE, et inclinable dans le cas du projet SONOBL). Des calculs RANS k-e et k-w ont ete realises et les grandeurs globales comparees aux experiences (epaisseurs caracteristiques de couche limite, pression statique, ...), validant les calculs stationnaires. Les parametres d'ecoulement sont ensuite injectes dans differents modeles du spectre de fluctuations de pression parietale : modele de Goody [1] en l'absence de gradient de pression moyenne, et modele recent de Rozenberg [2] en presence d'un gradient de pression defavorable. Les differents spectres parietaux sont ensuite compares aux resultats experimentaux. [1] M. Goody, Empirical spectral model of surface pressure fluctuations , AIAA journal, 2004, vol. 42, p. 1788-1794. [2] Y. Rozenberg, G. Robert, S. Moreau, Wall-pressure spectral model including the adverse pressure gradient effects , AIAA Journal, 2012, vol. 50, no 10, p. 2168-2179.

Active flow control around a two-dimensional geometry positioned near the ground is experimentally studied. Results for two model configurations are reported: one square-back and the other with a flap added at the rear-corner. The model rear end static pressure and the near-wake flow properties are derived from wall-pressure measurements, high-speed Particle Image Velocimetry acquisitions and hot-wire anemometry. The control flow is obtained from a pulsed jet system driven in open loop. The study aims at distinguishing the individual influence of the passive flap and the control system. The actuation frequency is shown to be a determining parameter regarding rear end pressure increase.

Sound propagation outdoors is strongly affected by atmospheric turbulence. Under strongly perturbed conditions or long propagation paths, the sound fluctuations reach their asymptotic behavior, e.g., the intensity variance progressively saturates. The present study evaluates the ability of a numerical propagation model based on the finite-difference time-domain solving of the linearized Euler equations in quantitatively reproducing the wave statistics under strong and saturated intensity fluctuations. It is the continuation of a previous study where weak intensity fluctuations were considered. The numerical propagation model is presented and tested with two-dimensional harmonic sound propagation over long paths and strong atmospheric perturbations. The results are compared to quantitative theoretical or numerical predictions available on the wave statistics, including the log-amplitude variance and the probability density functions of the complex acoustic pressure. The match is excellent for the evaluated source frequencies and all sound fluctuations strengths. Hence, this model captures these many aspects of strong atmospheric turbulence effects on sound propagation. Finally, the model results for the intensity probability density function are compared with a standard fit by a generalized gamma function.