Aeroacoutics of Moving Surfaces (AMS)

 

Aerodynamic Noise Generation - Basic Principles

Any inertial modification of a vortical motion in a compressible fluid generates sound. This general process holds for all types of deformation, say vortex pairing, vortex stretching, vortex breakdown, ... It is considerably more efficient when the modification is due to the interaction of vortical structures with solid surfaces. Since most flows of practical interest are turbulent, they carry and generate a large range of vortical structures resulting in broadband noise radiation. Moreover the same mechanism is also responsible for tonal noise when the periodic part of a flow (e.g. the wake from a blade row in turbomachines) interacts with a body. Furthermore, self-sustained oscillations in some unstable flow systems (wind instruments, cavities in a flow, ...) make initial flow disturbances more coherent, resulting in the emission of high-intensity tones.

Generally speaking, noise radiation from a mechanical system can be considered as a process dissipating a part of the mechanical energy of the system. However, the acoustic dissipation is so small that in most cases, it can be neglected to describe the dynamics of the system; the counterpart is that the energy dissipated as sound, although it is a quantity of interest for people working in acoustics and has a great impact on the ear, is very hard to extract accurately from a description of the system. When applied to the case of the aerodynamic noise radiated by a flow, this principle makes the sound waves hard to simulate directly using Computational Fluid Dynamics (CFD). The acoustic analogy developed by Lighthill (1952), then Ffowcs Williams & Hawkings (1969) overcomes this difficulty (except for self-sustained oscillations) by providing a way to relate the main features of the flow field, ignoring the very details, to the acoustic dissipation. This approach has been applied successfully to statistical predictions in aeroacoustics for a number of generic flow configurations and has led to simple but comprehensive models.

The main research activity of the team AMS is primarily focused on the understanding and simple modelling of aerodynamic noise generating mechanisms, for intended applications in an industrial context. The prediction methods are based on the acoustic analogy and can be used to post-process flow data that are assumed already available (for instance, from CFD). Other purely numerical strategies to directly predict aerodynamic sound generation and propagation, are currently developed elsewhere. An overview of these methods that are known as Computational Aero-Acoustics (CAA), can be found at turbulence and noise generation.

Besides analytical modelling of aerodynamic sources, that is also supported by through experimental investigations, the team also explores new propagation techniques (GFD) in steady mean flows. The goal is to provide an alternative to the classical use of the acoustic analogy, suitable for shear flows and swirling flows and easily adaptable in a complex geometry.

Research Topics

A - Isolated Airfoil Broadband Noise Studies

B - Aeroacoustics of the Bullroarer

C - Tonal Noise of an Airfoil

D - Inlet and Outlet Duct Radiation

E - High-Lift Device Noise (HLD)

F - Fan Rotational Noise Modelling

G - Rod-Airfoil Interaction and Turbomachinery Broadband Noise

H - Jet-Ring Interaction Noise

I - Aeroacoustics of Radial Compressors

J - Subsonic Fan Broadband Noise

 

A - Isolated-Airfoil Broadband Noise Studies

Collaboration with Valeo and CETIAT

Fundamental studies on the noise generated by isolated airfoils are necessary for the basic understanding of the noise generating mechanisms in complex systems, such as fans, turbomachines or high-lift devices. This is the reason why a deep attention is paid by the AMS team to this research field. A common activity is developed with Valeo and CETIAT as external partners, for applications to low-Mach number fans in free field. Two major mechanisms are investigated (Fig. A1): (1) turbulence-interaction noise, (2) trailing-edge noise. Blade-tip effects, tip vortices and leakage flows, even though also recognised as efficient sound generators, are still to be assessed.

Fig. A1 : Turbulence-interaction noise (left) and trailing-edge noise (right)

Trailing-edge noise is due to the scattering of upstream boundary layer disturbances as acoustic waves at the trailing edge, as a result of the unsteady Kutta condition; the equivalent sources concentrate at the trailing edge. It is recognised as a major source of sound for wind turbines and fans operating in clean flow conditions.

Turbulence-interaction noise is due to the impingement of oncoming disturbances; the equivalent sources concentrate at the leading edge. This mechanism is involved in most fans or turbomachines, the turbulence being already in the inflow or produced in the wakes of upstream obstacles or casing boundary layers.

The analytical models developed are based on a statistical approach. Typically, the acoustic spectrum in the far field is related to the spectrum and spanwise coherence of the aerodynamic wall-pressure close to the trailing edge for mechanism (2) [A2] and to the oncoming velocity spectrum for mechanism (1). The models apply to thin airfoils with moderate thickness and camber. Their validity is assessed by comparison with dedicated experiments in the open-jet wind tunnels of the ECL-LMFA (Fig. A2). Aerodynamic data are measured to be introduced into the analytical models. Finally, the predicted sound field is compared to the measured one; the model is considered as validated if both the spectral content and the directivity are reproduced correctly. This can be checked globally on a frequency-angle colour plot. Sample experimental results are given in Fig. A3 for a flat plate with an non-zero angle of attack to the flow; measured and calculated noise spectra are compared in Fig. A4.

Fig. A2 : Typical experimental set-up in anechoic wind tunnel, showing a Valeo airfoil equipped with wall-pressure probes (capillary tubes and remote, small-size microphones at the bottom of the picture). The angle of attack is varied by rotating the top and bottom discs in the plates. The far-field noise is measured in the mid-span plane by means of an ordinary microphone on a rotating support.

Fig. A3 : Frequency-angle radiation plot for the self-noise of a flat plate inclined in a flow, measurements made in the mid-span plane. The horizontal red trace in the first third upper plot is attributed to the von Kármán vortex shedding and the extended red region at smaller frequencies is the true trailing-edge noise due to the scattering of boundary layer turbulence. Levels in dB. The vortex shedding noise only occurs for a thick or blunted trailing edge; it is not modelled in the plots of Fig. A.4 below.

Fig. A4 : Measured trailing-edge noise spectra of a flat plate for two angles of attack and two flow velocities (blue) and calculated spectra from wall-pressure measurements (red) and the analytical model based on the acoustic analogy (ECL experiment, courtesy of CETIAT). The peak noise due the vortex shedding at 5° angle of attack is not reproduced by the model.

[A1] Roger M. & Moreau S., Broadband Self-Noise from Loaded Fan Blades, AIAA Journal, 42(3), 536-544, 2004

[A2] Roger M. & Moreau S., Back-Scattering Correction and Further Extensions of Amiet's Trailing Edge Noise Model. Part 1: Theory, J. Sound Vib., in press

[A3] Moreau S. & Roger M., Effect of Airfoil Aerodynamic Loading on Trailing-Edge Noise, 9th AIAA/CEAS Aeroacoustics Conference, AIAA Paper 2003-, 2003

[A4] Moreau S. & Roger M., Competing Broadband Noise Mechanisms in Low Speed Axial Fans, 10th AIAA/CEAS Aeroacoustics Conference, AIAA Paper 2004-3039, 2004

 

B - Aeroacoustics of the Bullroarer

In-House Study

The Bullroarer is an aero-mechanical instrument used by aboriginal tribesmen (Fig.B.1). It is made of a flat, roughly airfoil-shaped piece of wood tied to a string and whirled through the air. In normal conditions, the wooden airfoil spins around its principal axis of inertia, this motion being sustained by the autorotation phenomenon. The vortex shedding induced by the relative lateral flow is responsible for the intense sound radiation.

Fig. B1 Typical bullroarers, from Atherton "Australian made... australian played", NSW University Press, 1990, and Sadie "The New Grove Dictionary of Musical Instruments", McMillan 1984

The Bullroarer is the only rotating blade device for which the motion is not prescribed but rather imposed by the instantaneous aerodynamic forces. As a consequence, the mechanical equations of motion and the equations of fluid dynamics are coupled and must be solved as such prior to any noise prediction based on the analogy. A full simulation code has been written using time-domain finite differences and a set of dedicated two-dimensional CFD results for a spinning airfoil of flat elliptical cross-section [B1]. A typical time-frequency analysis of both measured and computed acoustic signals is shown in Fig.B.2. All modulations involved in the Bullroarer dynamics and determinant for the subjective features of the sound are realistically reproduced. The bullroarer is understood as a test-case for methods coupling CFD and the acoustic analogy.

Fig. B2 : Typical time-frequency analysis of Bullroarer sound, as measured (left) and simulated (right). Time signals made non-dimensional by the maximum amplitude and frequencies scaled by the dominant pitch f0

[B1] Roger M., Aubert S., Autorotation and the Theory of Bullroarer Sound, 5th AIAA/CEAS Aeroacoustics Conference, AIAA Paper 99-1827, Seattle, 1999.

 

C - Tonal Noise of an Airfoil

In-House Study

In some conditions, an airfoil embedded in a homogeneous, uniform flow radiates high-intensity and high-pitch tones. This spectacular sound can be heard on glider wings in some flow conditions, on fan blades or isolated airfoils, generally at moderate Reynolds numbers. The tones involve instabilities in the laminar boundary layers (the so-called Tollmien-Schlichting waves) and acoustic scattering at the trailing edge. At the frequencies for which the acoustic waves propagating upstream are in phase with growing instabilities at their starting point on the airfoil surface, a self-sustained oscillation takes place and the unsteady motion is highly amplified. The mechanism requires a full coherence along the feedback loop between the instability and acoustic waves. Thus the tones are simply suppressed if the transition to turbulence in the boundary layers is forced.
The generation of tones due to TS waves is a highly coherent process used here to validate analytical models for airfoil trailing-edge noise directivity. It is not to be confused with the peak noise due to a periodic (von Kármán) vortex shedding (Fig. A.3 of section A - Isolated Airfoil Broadband Noise Studies). Frequency increases proportionally to the flow speed for the latter and to the power 1.5 of the flow speed for the former. (to be continued ...)

D - Inlet and Outlet Duct Radiation

Collaboration with Airbus-Toulouse and D. Casalino

Most subsonic fans operate in a duct rather than in free field. This is specially determinant for turbofan engines. The first effect is that the sound field radiated by the sources is re-structured by the successive reflections on the walls, leading to the well-known cut-on or cut-off duct modes. The second effect is that the propagated modes are partially transmitted outside and partially reflected inside, at the duct ends. The study of sound transmission along the duct is the natural continuation of the study of noise-generating mechanisms and, as such, is also included in the activities of the AMS team. Analytical diffraction theories based on the Wiener-Hopf technique are applied, typically assimilating the inlet of a turbofan to the end of a thin cylinder with uniform flow. Numerical techniques based on boundary element or finite element methods are necessary to account for precise inlet lip geometry (eg. scarfed inlet) or the presence of non-uniform flow. Some of them are assessed by Airbus-Toulouse within the scope of a collaboration (French CIFRE PhD Programs) [D1,D3]. An appealing alternative approach is the Green's Function Discretization (GFD) technique developed more recently [D2] (Fig. D2).

Fig. D1 Typical directivity diagrams of the noise radiated at inlet by a turbofan. Numerical computations using an integral method (continuous line) are compared to analytical predictions (dotted line). From [D1].		Fig. D2 Instantaneous pressure field in a duct with varying cross-section, as computed using GFD [D2]. The incident mode propagating from the left with 5 azimuthal lobes is continuously reflected back in the annular duct due to cut-on to cut-off transition in the cross-section constriction.

[D1] Lidoine S., Batard H., Troyes S., Delnevo A. & Roger M., Acoustic Radiation Modelling of Aeroengine Intake, Comparison Between Analytical and Numerical Methods, 7th AIAA/CEAS Aeroacoustics Conference, paper 2001-2140, Maastricht, 2001

[D2] Casalino D., Roger M. & Jacob M. C., Prediction of Sound Propagation in Ducted Potential Flows Using Green's Function Discretization, 9th AIAA/CEAS Aeroacoustics Conference, paper 2003-3246, Hilton Head, 2003

[D3] Druon Y., Lidoine S. & Roger M., Acoustic Radiation Modeling of Engine Exhaust Comparisons Between Analytical and Numerical Methods, 10th AIAA/CEAS Aeroacoustics Conference, paper 2004-, Manchester, 2004

 

E - High-Lift Device Noise (HLD)

Collaboration with Airbus-Toulouse

During take-off and landing operations, an aircraft must have its global lift preserved with a relatively low velocity, which is made possible by means of defected high-lift devices, namely leading-edge slats and trailing-edge flaps. The gaps and cavities formed when the HLD are deflected generate instabilities and turbulent disturbances that radiate sound when interacting with the surface. This happens both around the slat cavity and wing leading edge, as well as in the rear wing cavity and flap leading edge. The resulting sound sources are distributed along the span. Additional, localised, and highly three-dimensional sources take place at the flap side-edges, slat tips or any other geometrical singularity (holes in the wing for refuelling, for instance). Though less noisy than the landing gears, the HLD on an aircraft contribute to the noise exposure around airports, essentially during approach and landing. This will become especially crucial for the future large aircraft.
The aerodynamic noise from HLD is a research topic investigated in collaboration with Airbus-Toulouse, on the basis of experiments on model wings or typical cross-sections in the ECL anechoic wind tunnel. Ad hoc analytical prediction methods have been proposed for the spanwise distributed sources as well as for some aspects of the 3D flap-edge and slat-edge sources [E1, E2, E3]. The agreement of analytical predictions with in-flight measurements is shown in Fig. E.1.
In some flight conditions, where the trailing-edge extends perpendicularly to the oncoming flow (unswept wing configuration), the slat or flap cavity in the model scale experiments has been shown to radiate discrete frequencies similar to the tones of cavities in a grazing flow [E4].

Fig. E1 : Ability of dedicated analytical models to reproduce flight-test measurements on an aircraft with retracted landing gear (courtesy of Airbus-Toulouse, from [E3]).

[E1] Molin N. & Roger M. The use of Amiet's Methods in Predicting the Noise From 2D High-Lift Devices, 6th AIAA/CEAS Aeroacoustics Conference, Lahaina, AIAA paper 2000-2064, 2000

[E2] Roger M. Diffraction par un bord de fuite; application au bruit d'un volet hypersustentateur, Troisièmes Journées sur l'Acoustique de Galbrun, La Rochelle, 2003

[E3] Molin N., Roger M. & Barré S., Prediction of Aircraft High-Lift Device Noise Using Dedicated Analytical Models, 9th AIAA/CEAS Aeroacoustics Conference, paper 2003-3225, Hilton Head, 2003

[E4] Roger M. & Pérennès S., Low-Frequency Noise Sources in Two-Dimensional High-Lift Devices, 6th AIAA/CEAS Aeroacoustics Conference, paper 2000-1972, Lahaina, 2000

 

F - Fan Rotational Noise Modelling

Collaboration with Valeo and Technofan (SNECMA Group)

Most fans are made of rotating blades (rotor) and downstream, non-rotating outlet guide vanes (stator). Besides the broadband noise due to the interaction of turbulent flows with the blades and vanes, a sharp tonal noise can be generated due to the periodic rotor-stator interaction: the blade wakes induce periodic dipole-like sources on the stator vanes, and the sound is radiated at the blade passing frequency and harmonics. Additional rotor-alone tones at the same frequencies are generated due to the interaction with a distorted, stationary mean flow. Wake-interaction noise and distortion noise are the main contributions to what is known as the rotational noise. Distortion noise generally depends on the installation of the fan and is therefore difficult to investigate. Wake-interaction noise is easier to predict: when the wake impingement on stator vanes cannot be accurately simulated by CFD, simplified analytical approaches based on similarity laws for wakes and the linearised theories of unsteady aerodynamics can be used to provide a reasonable approximation of the source terms.
For fans operating in free field or equivalent conditions (cooling fans for cars, for instance), the rotational noise is often a nuisance. The results obtained with analytical models in free field as shown in Fig. F.1 can help, for instance, to optimise blade (B) and vane (V) counts for minimum tonal noise in a rotor-stator stage. The optimum depends on the observer location and the expected installation (the arrangement B=V must be avoided in a duct).

Fig. F1 Free-field directivity patterns of rotor-stator interaction tones at twice the blade passing frequency (m=2) for different blade and vane counts (B,V) in the case of a non-compact circular distribution of vane segments (typical configuration of air-conditioning application in aircraft). The rotation axis is the vertical direction. Source radius, rotational frequency, ... The arrangement (9,9) essentially radiates on axis. The arrangement (9,13) essentially radiates at 90° from the axis.

Fans used in air conditioning units for aircraft are made of a rotor-stator arrangement in a duct. Due to the relatively large number of blades, small diameter and high rotation speed, such fans radiate noise over an extended frequency range. High-pitch tones can be a nuisance for the crew as well as for the passengers and must be modelled for noise control studies. The corresponding acoustic waves propagating in the duct are calculated by means of the classical modal approach (see section D - Inlet and Outlet Duct Radiation). A prototype code has been written in collaboration with Technofan (SNECMA Group) for a fast use integrated in a pre-design cycle, assessing the effect of compared blade and vane numbers on the interaction tone levels. Typical, indicative in-duct results are shown in Fig. F.1 for the same parameters as in Fig. F.1. The counts around (9,9) are worse, as expected for in-duct installation, whereas the count (9,13) is around the optimum.

Fig. F2 - Typical study of the effect of vane count on wake-interaction noise in a rigid duct. From blue to red: multiples 1 to 5 of the blade-passing frequency. The bars indicate power levels in relative decibels. Top figure: upstream transmission, bottom figure: downstream transmission. Number of blades B=9, number of vanes V between 7 and 18.

 

G - Rod-Airfoil Interaction and Turbomachinery Broadband Noise

European Research Programs

Subsonic turbulence-interaction noise was investigated in the European project TurboNoiseCFD on a rod-airfoil configuration: an airfoil was located in the wake of a cylinder and radiated a broadband spectrum around the preferred frequencies associated to the rod vortex shedding. The broadband part of the noise emission is due to two factors. First, the shedding is not purely periodic because of its limited spanwise correlation. Secondly, at sub-critical Reynolds numbers of the cylinder flow that was considered here, the wake immediately transitions to turbulence. Thus a turbulent wake dominated by large quasi periodic remainders of the von Kármán vortices impinges onto the airfoil resulting in both narrow-band and broadband interaction noise [G1]. The advantages of this configuration are twofold:

• its tonal part mimics the interaction of a gust with and airfoil, as well as the interaction of a rotor wake with stator vanes in a simple non-rotating flow;
• its broadband part models the turbulence-airfoil interaction noise.

Moreover, due to its simplicity, the configuration can be used as a benchmark to test a code capability to model broadband noise. This was actually the main goal of the contribution of the team to TurboNoiseCFD. An experiment was carried out in the ECL anechoic wind tunnel (Fig. G.1) and the experimental results were compared to numerical ones obtained with various unsteady CFD techniques: unsteady RANS computations and LES. The far field was predicted using the Ffows Williams-Hawkings analogy. Fig. G.2 shows the comparison between the spanwise vorticity components obtained by two CFD approaches and PIV measurements. Aerodynamic wall pressures on the airfoil are then compared on Fig. G.3 a) and the acoustic far field pressures on Fig. G.3 b). This study confirmed the deterministic nature of RANS computations, and also the fact that natural frequencies such as the rod shedding frequency arising from absolute flow instabilities are ill-estimated by U-RANS. In order to reproduce the broadband spectrum using U-RANS data, a stochastic variation satisfying the rod spanwise coherence statistics was added to the retarded times during the far field computation.

New Investigation on Fan Broadband Noise Generation

In the European 6th Framework project PROBAND, a new assault on the assessment, prediction and eventually reduction of broadband noise generation in Fan/OGV stages of Ultra High By-pass Ratio engines will be attempted. Major broadband noise source mechanisms will be modelled and suitable prediction tools will be developed for turbo-engine design. These tools will be able to predict broadband sound generation using standard industrial CFD codes (RANS, U-RANS). In parallel, more advanced CFD techniques, such as LES will be applied to ever more complex flow configurations of practical interest for engine manufacturers. A series of experimental campaigns ranging from a small scale single airfoil experiment in ECL wind tunnel to a large scale experiment at the Anecom test rig will be carried out to provide validation data for the models and the computations.

Two major source mechanisms will be investigated by the AMS team, fan self noise and fan tip / turbulent boundary layer interaction. The first one has already been extensively studied, as described in section A - Isolated Airfoil Broadband Noise Studies. The second one is due to the interaction of the blade tip vortices of a fan with the turbulent structures advected by the duct boundary layer. The importance of this noise source is still unclear since it has been impossible so far to isolate from others on test rigs. The team plans therefore to carry out a specific single-airfoil experiment in the ECL anechoic wind tunnel and to develop a suitable analytic model.

[G1] Casalino D., Chiocchia G. & Roger M. Interaction Noise from an Airfoil in a K\'arm\'an Vortex Street, 7th AIAA/CEAS Aeroacoustics Conference, paper 2001-2213, Maastricht, 2001

[G2] Boudet J., Grosjean N. & Jacob M.C. Wake-airfoil interaction as broadband noise source, accepted for publication in Int. J. Aeroacoustics, . 2003.

[G3] Boudet J., Casalino D., Jacob M.C. & Ferrand P. - Prediction of broadband noise: airfoil in the wake of a rod, 42nd AIAA Aerospace Sciences Meeting and Exhibit, Reno, Nevada, 2004.

 

H - Jet-Ring Interaction Noise

In-House Study, Stella Serafini (Politecnico di Torino)

A flat ring in the shear layer of a circular jet (Fig.H.1) radiates noise due to the impingement of the jet shear layer oscillations or turbulence. This dipolar, interaction noise is much higher than the pure quadrupole-like, mixing noise from the jet. The ring can be considered as an annular airfoil in a turbulent flow with axisymmetric statistical properties, as well as a device to scan the turbulence in a jet. The configuration is studied both theoretically and experimentally as a basic one for fundamental mechanisms in aeroacoustics.

Existing analytical approaches developed for rectangular airfoils have been recently adapted to the ring geometry [H1] and provide theoretical expressions for the radiation of helical modes of oscillation of the jet with the ring. Two flow regimes can be obtained. For a simple circular nozzle lip and small separating distance between the nozzle and the ring, discrete tones corresponding to self-sustained oscillations are heard: the acoustic back-reaction from the ring reinforces the jet shear layer instability modes. For larger separating distance or for a corrugated nozzle, the acoustic feed-back is suppressed and the jet shear layer only provides an incident turbulent flow for ring broadband noise radiation.

Preliminary experimental results are given in Fig. H.2.

Fig.H.1 Ring in the turbulent shear flow from a circular nozzle		Fig.H.2 Angle-Frequency plot of jet-ring interaction noise in self-sustained oscillation configuration, showing high-level tones. Nozzle and ring diameter 10 cm, separation 4 cm, flow speed 37 m/s, far-field distance 2m. The tone multiple-lobed patterns in red are associated with privileged oscillating modes of the jet.

 

I - Aeroacoustics of Radial Compressors

Collaboration with Vibratec & Liebherr-Aerospace Toulouse (LTS)

Centrifugal or radial compressor rotors are encountered in rotorcraft engines and air-conditioning units for aircraft. The major noise contribution is the tonal noise at the rotor blade passing frequency and harmonics. In some subsonic centrifugal compressors, part of this noise, radiated at inlet, is due to the rotor wake impingement on the radial diffuser vanes. A dedicated analytical model has been developed to describe the transmission of spiral wave patterns generated on the vanes into the rotor inter-blade channels [I1]. This first step will lead to a global model chaining the transmission to sound propagation through the channels and radiation in the inlet duct.

Fig. I1 Typical pressure patterns illustrating the transmission of converging spiral waves with n = 5 lobes into the inter-blade channels of a 12-bladed centrifugal rotor, with no flow (left) and radial flow (right); kR0 : Helmholtz number based on the rotor outer radius. From [I1]

[I1] Roger M. Analytical Modelling of Wake-Interaction Noise in Centrifugal Compressors With Vaned Diffusers, 10th AIAA/CEAS Aeroacoustics Conference, paper 2004-2994, Manchester, 2004

 

J - Subsonic Fan Broadband Noise

Collaboration with Valeo, the SNECMA and the CETIAT

Ph.D. Yannick Rozenberg (CIRT)

The analytical methods for broadband noise prediction of isolated airfoils are presently extended and applied to subsonically rotating fan blades in free space or in a duct. The extension accounts for Doppler effect, and blade lean and sweep. Accompanying experiments are developed in free field (CETIAT set-up). (study in progress, to be completed)

haut/top