Experimental acoustic identification of flow noise sources in expansion chambers

Transcription

1 Exerimental acoustic identification of flow noise sources in exansion chambers W. De Roeck, W. Desmet K.U.Leuven, Deartment of Mechanical Engineering, Celestijnenlaan 300 B, B-3001, Leuven, Belgium Abstract For the aeroacoustic design of exansion chambers, commonly installed in e.g. HVAC ducts or automotive exhaust systems, the aerodynamic noise generation mechanisms as well as the transmission characteristics of acoustic waves roagating in a non-quiescent medium have to be taken into account. The attenuation of downstream roagating acoustic waves should be maximized with a minimum of additional flow noise generation and often a comromise between both erformance arameters has to be made. In this aer, these two aeroacoustic roerties are exerimentally analyzed for a simle exansion chamber carrying a mean flow which is uniform in time. This is achieved using an active two-ort reresentation which is validated for a straight duct with a loudseaker laced at its center. The different exeriments are carried out on an oen circuit aeroacoustic wind tunnel where the test objects are located inside a semi-anechoic room. This allows to determine the noise emission and radiation in free-field environments and correlate the far-field ressure levels with the aerodynamic noise generation mechanisms described by the active two-ort comonents. 1 Introduction In exhaust ducts, exansion chambers are commonly installed to attenuate the noise emitted by e.g. IC engines, comressors and other rotating machinery. The energy of this engine noise is mostly concentrated around the harmonics of the engine firing frequency. Their contribution is dominant for low to medium engine seeds since aerodynamically generated noise sources are rather inefficient at low Mach numbers. When the engine seed increases, flow noise effects become more imortant and can even become the dominant source of exhaust noise [1]. In this framework, exansion chambers can become flow-excited noise generators rather than silencers. The emitted flow noise sectrum contains both broadband, generated by the turbulent structures in the confined jet and tailie, and tonal comonents. Tonal comonents can either be generated by a flow-acoustic feedback-loo inside the exansion chamber [2] or by the broadband excitation of the acoustic modes of the exansion chamber and tailie [3]. At resent, the research towards aerodynamic noise generating mechanisms in exansion chambers is rather scarce but of growing interest due to the decreasing noise generation achieved by most engine, comressor and fan manufacturers. Besides these undesirable noise generation mechanisms, exansion chambers are traditionally designed to dam out engine noise, fan noise and flow noise sources, generated at flow discontinuities or irregularities inside the exhaust system ustream to the exansion chamber. In contrast to the internally generated flow noise sources, the ustream and downstream noise attenuation rinciles are quite well understood [4, 5]. Since both the noise generation and roagation characteristics are of utmost imortance for the aeroacoustic design of exansion chambers, an active two-ort formulation, describing both henomena by a matrix reresentation, is determined exerimentally in this aer. The exeriments are carried out on an oen-circuit aeroacoustic wind tunnel in which a rootsblower is used to generate a time-uniform mean flow. Additionally, 455

2 456 PROCEEDINGS OF ISMA2008 the free-field radiation of the aerodynamically generated noise is investigated by acoustic ressure measurements erformed inside a semi-anechoic room. The reliminary results, resented in this aer, are merely used to validate the exerimental rocedure and the aeroacoustic test set-u. Based on these findings modifications of the aeroacoustic wind tunnel design and the active two-ort characterization are roosed which will be carried out in future research. The outline of this aer is the following. The next section discusses in more detail the lay-out of the aeroacoustic test set-u, secifically designed to analyze both the noise generation and roagation mechanisms in confined subsonic flows. Subsequently, the measurement rocedure to obtain the active and assive bi-ort characteristics is discussed and validated for a straight duct with a center loudseaker excitation. In the next section this active two-ort measurement rocedure is used for the for the aeroacoustic characterization of a simle exansion chamber carrying a uniform mean flow at different Mach numbers (M < 0.3). Finally, the major conclusions, drawn from this analysis are summarized. 2 Descrition of the test set-u A schematical overview of the exerimental test set-u is shown on the left of figure 1. The general urose of the test rig is the determination of the aeroacoustic noise generation and roagation mechanisms as well as the far field acoustic radiation for confined subsonic flow alications with inflow conditions that are reresentative for HVAC and automotive muffler alications. For this goal, an oen-circuit test set-u is used which can be subdivided into three major comonents: the flow generation, the flow rearation and the measurement sections. Figure 1: Schematical overview of aeroacoustic test set-u (left) and the flow generators (right). 2.1 Flow generation Two flow generators, shown on the right of figure 1, can be used: a rootsblower to generate a time-uniform flow field or a cold engine simulator for exeriments in time-ulsating flow fields. The rootsblower has three lobes to minimize both noise generation and small time-ulsations of the flow. A frequency regulator with PID controller, couled with downstream ressure and flow rate sensors, is attached to the rootsblower in order to ensure identical inlet conditions between different measurement camaigns. The maximum Mach number that can be achieved equals 0.3 for duct diameters equal to 0.05 m. Both the duct diameter and maximum Mach number are reresentative for automotive exhaust system alications.

3 AEROACOUSTICS AND FLOW NOISE 457 As second flow generator, a cold engine simulator can be used to obtain ulsating inlet conditions. This electronically driven simulator roduces a cold flow by using comressed air with similar mean flow velocities and velocity fluctuations as comared to a real automotive combustion engine. The cold engine simulator is secifically develoed to carry out exeriments of new concets of exhaust systems in an environment that is less violent for measurement equiment [6]. This tye of time-ulsating flow fields are not further analyzed in this aer. The inlet flow rofile for the time-uniform inflow, generated by the rootsblower, is measured using hot wire anemometry. The hot wire robe is traversed along the duct diameter and the mean flow rofile and turbulent intensity of the velocity fluctuations are shown in figure 2 for a rotational seed of the rootsblower of 25 Hz, which corresonds to an average flow velocity along the inlet duct diameter of the the test object of 32.7 m/s. A good agreement with the emirical 1/7th ower law equation is obtained, known to give a good agreement with exerimental data in a smooth ie at Re The turbulent intensity at the centerline is lower than 2%, indicating low turbulence inlet conditions. Figure 2: Mean rofile (left) and turbulent intensity (right) along the inlet duct diameter with a rootsblower rotational seed of 25Hz. 2.2 Flow rearation The rootsblower is resonsible for an increase in temerature of more than 60 o C. These relatively high temerature differences change the roagation seed of the acoustic waves and can modify the aerodynamic noise generation mechanisms. All results, resented in this aer, are carried out at these relatively high temeratures and, based on these measurements, it is concluded that temerature effects largely dominate the convective effects. For this reason, an aftercooler is being installed immediately after the rootsblower. The heat exchanger generates an outlet temerature of the comressed air of 10 o C above the ambient temerature. Together with the thermal inertia of the ducts of the acoustic labyrinth, the additional iing and the flow exansion inside the flow conditioner, a flow temerature at the outlet of the test object of 2 o C above the ambient temerature is achieved with temerature fluctuations less than 5% between different measurements. After the heat exchanger, the flow asses an acoustic labyrinth, consisting of aroximately 10 m of circular silencers. These ducts have a dual urose: on one hand, they reduce ustream generated noise sources, resulting in anechoic inflow conditions at the entrance the test object, which is referable to validate numerical aeroacoustic simulations. On the other hand, small ulsations onto the mean velocity, are filtered out since the acoustic labyrinth behaves as a long exansion chamber. A total transmission loss of around 50 db is achieved by the acoustic labyrinth in both the ustream and downstream direction. Subsequently, the flow is guided through a flexible iing inside a semi-anechoic room. A flow conditioner (left of figure 3) containing a divergent, a erforated straight duct with acoustic daming material, screens,

4 458 PROCEEDINGS OF ISMA2008 honeycombs and convergent nozzle is installed to obtain low-noise and low-turbulence inlet conditions. Due to the scarce sace, a 130 o bend with guiding vanes is used to direct the measurement section into the diagonal of the room, as indicated on figure 3. The flow temerature, flow rate and ressure entering the semi-anechoic room are continuously monitored, as well as the temerature and ressure difference at the convergent of the flow conditioner. The sensors signals can be couled with a PID controller and frequency convertor to determine the rootsblower s rotational seed in order to obtain identical inlet conditions during different measurement camaigns. Figure 3: Flow conditioner (left) and measurement section directed in the diagonal of the semi-anechoic room (right). 2.3 Measurement section Two different measurement sections can be used: A first one, consisting of two straight ducts (L = 5 m) (with attached loudseakers and flush-mounted dynamic ressure transducers) between which the test object can be laced. In this way, the assive two-ort characteristics are straightforward to obtain using a combined multile-load, multile-source method (section 3.1). This measurement section can also be used to determine the erformance of lining material and erforated lates with a grazing mean flow. The second measurement section (right of figure 3) is located inside a semi-anechoic room to avoid the influence of external noise sources, such as the flow generators or other laboratory equiment. This art of the set-u is used to determine the actual aerodynamic noise generating mechanisms with flush-mounted ressure transducers, hot wire anemometry and PIV measurements as well as the farfield radiation of the flow noise sources using microhone and acoustic intensity measurements. 3 Descrition of the exerimental rocedure 3.1 Active bi-ort characteristics Acoustic systems, located between two straight ducts in which acoustic lane wave roagation is occurring, are often referred to as acoustic two-orts (fig.4). Every acoustic two-ort, in which no acoustic generation or dissiation is occurring, the so-called assive two-ort, is uniquely characterized by an acoustic transfer matrix T. This transfer, transmission or scattering matrix links the one-dimensional acoustics variables at the inlet ( 1, u 1 ) (with and u resectively the acoustic ressure and velocity fluctuations in the x- direction) with those at the outlet ( 2, u 2 ). The major advantage of using such a linear network reresentation

5 AEROACOUSTICS AND FLOW NOISE 459 is its indeendency of the number or tye of acoustic elements lying before or after the element under consideration. Figure 4: Acoustic bi-ort reresentation for an element in a duct system. One of the most commonly used reresentations of the scattering matrix is using the ressure of the rightand left-running acoustic waves ( +, ) (instead of the ressure and velocity fluctuations) at both ends of the bi-ort. In this way, the four scattering matrix elements T +, T and R +, R can be directly interreted as the transmission and reflection coefficients in, resectively, the downstream and ustream direction. If active noise generation and/or dissiation rocesses occur inside the filter, an active bi-ort reresentation can be used by adding two additional source terms in the downstream and ustream direction ( +s 2, s 1 ), the active two-ort arameters [7]. This leads to following matrix formulation: { } = [ T + R R + T ] { } + { +s 2 s 1 The assive two-ort comonents (four-ole arameters) T +, T, R + and R are indeendent of the internal or external noise generation mechanisms and the ustream and downstream imedance. As such, they describe the filter roer. Oosite to these comonents, the active noise generation comonents in the downstream +s 2 and ustream direction s 1 can be deendent on the ustream and downstream imedance [8]. } (1) 3.2 Exerimental determination of the active two-ort arameters The active and assive two-ort arameters can be determined analytically [4], numerically [9] or exerimentally. In this aer an exerimental three-ste rocedure is adoted, consisting of: 1. The determination of the amlitudes of the ustream and and downstream roagating waves at both ends of the acoustic two-ort. 2. Characterization of the scattering matrix (assive two-ort comonents). 3. Identification of the active two-ort arameters Ustream and downstream roagating waves For the determination of the ustream and downstream roagating waves ( +, ) at each end of the twoort, the dynamic ressure sensor signals at a minimum of two ositions on both ends are needed. The leftand right running amlitudes are directly linked to the acoustic velocity and ressure fluctuations, since at any location x in the duct following relations hold for lane acoustic wave roagation: (x, f) = + (x, f) + (x, f) = + (0, f)e jk+x + (0, f)e jk x (2) with x = 0 the osition of the reference lane and k + k 0 /(1 M) and k k 0 /(1 + M) the acoustic wavenumbers of, resectively the downstream and ustream roagating waves and k 0 = ω/c 0. For non

6 460 PROCEEDINGS OF ISMA2008 sace-uniform mean flows a more accurate formulation for k + and k can be obtained using a quasi 2D aroach [10]. Equation (2) can be written for all n microhone ositions and solving following overdetermined system of n equations with two unknowns, resulting in a solution for + and at both ends of the bi-ort: { + (f) (f) } = e jk+ x 1 e jk x 1 e jk+ x 2 e jk x 2. e jk+ x n. e jk x n (x 1, f) (x 2, f). (x n, f) (3) with the Moore Penrose seudo matrix inverse and x k the location of the k th microhone with resect to the reference lane. The exact locations of the microhone ositions are determined for each measurement section with a white noise excitation and a rigid wall termination. Using the transfer functions between the different microhones, the exact location are easily obtained since the ressure signals show a first minimum for each microhone when the distance between the rigid termination equals 1/4 of the acoustic wavelength. The overdetermination of this system of equations is desired to minimize the influence of aerodynamic ressure fluctuations, noisy signals and small measurement errors. Since the aerodynamic ressure fluctuations, resent in the microhone signals, do not satisfy equation (3) they are largely filtered out by these equations, leading to an increase in the signal-to-noise ratio. Further elimination of noise from the microhone signals caused by aerodynamic ressure fluctuations, internal aerodynamic sound sources and non-suressed noise from the source is achieved by using a large number of averages and transfer functions in equation (3) instead of the measured ressure fluctuations. These transfer functions are taken between the microhone signals and a reference signal correlated with the imosed sound field in the duct but uncorrelated with these noise sources. In this sense the electric signals e which drive the external sources of the m different exeriments are a convenient choice. For clarity reasons, the notation, used in this aer, still uses the ressure fluctuations of the k th measurement: k (f) instead of the transfer function H k (f) = k (f)/e k (f) Passive two-ort characteristics In order to determine the four-ole arameters, at least two measurements need to be erformed varying from each other by the osition of an external sound source (two-source technique) [11, 12], by the outlet imedance (two-load technique) [13] or a combination of both techniques. Equation (4) can then be written for m exeriments without active comonents, since with a strong enough loudseaker excitation the aerodynamic noise generation can be neglected ( +s 2 = s 1 0). As a result a combined overdetermined system of m equations with four unknowns T +, T, R + and R is obtained, which can be solved following a least square strategy: [ T + (f) R (f) R + (f) T (f) ] = [ + 2,1 (f) + 2,2 (f) ,m (f) 1,1 (f) 1,2 (f)... 1,m (f) ] [ + 1,1 (f) + 1,2 (f) ,m (f) 2,1 (f) 2,2 (f)... 2,m (f) ] (4) Solving this system of equations directly results in the assive two-ort comonents. The different scattering matrix comonents can easily be used to determine the acoustic roerties such as recirocity, imedance and transmission loss of the acoustic filter [4]. The transmission erformance of acoustic filters e.g. is often characterized by the transmission loss (TL), which is directly linked to the downstream transmission coefficient T + : ( ) 1 T L = 20 log T + (5)

7 AEROACOUSTICS AND FLOW NOISE Active two-ort characteristics Oosite to the assive two-ort comonents, the active noise generation comonents in the downstream +s 2 and ustream direction s 1 are deendent on the ustream and downstream imedance [8]. As a result, an overdetermined system of equations, using a different imedance at the inlet or outlet section, cannot be used for the determination of the active characteristics. Also multile sources techniques cannot be used since they can masque the actual aerodynamically generated noise, resulting in a low signal-to-noise ratio. Thus, the active two-ort arameters have to be obtained exerimentally (for one secific combination of inlet and outlet imedance) with one additional measurement without loudseaker excitation and solving the system of two equations (eq. 4) for the two unknowns ( +s 2, s 1 ) using the reviously determined assive two-ort arameters. { +s 2 s 1 } = { } [ T + R R + T ] { + 1 In order to further increase the signal-to-noise ratio a large number of averages is needed and transfer functions are used in equation (6). As reference signal two additional microhones are used which are located further away from the inlet and outlet reference lanes. The aerodynamic ressure fluctuations is this region are uncorrelated with those measured by the other microhones and, as a result, a further suression of these non-acoustical disturbances is achieved. 2 } (6) 3.3 Validation of the active two-ort characterization This exerimental rocedure is validated for a straight steel duct (fig.5) with a length and diameter of resectively 0.5 m and 0.04 m. In the middle of the duct a loudseaker is laced which is excited with different excitation signals in order to validate whether the inut signal is correctly retrieved by the active two-ort estimation. The validation is carried out without the resence of a flow field for six different configurations, varying the osition of the external seaker and the imedance at both ends of the test set-u. For the exeriments, two flush-mounted ressure transducers are used at the ustream end and three sensors at the downstream end. Three different excitation signals of the central loudseaker are used: A white noise excitation to determine the loudseaker characteristic. A single frequency excitation at 220 Hz, 440 Hz and 1000 Hz to determine the caability of the exerimental rocedure to cature narrow-band tonal henomena. A broadband sectrum which is reresentative for the aerodynamic noise generation in exansion chambers. Figure 5: Test set-u for the validation of the exerimental rocedure Passive two-ort characteristics The real and imaginary arts of the scattering matrix coefficients are shown in figure 6. As exected for a straight duct a recirocal behavior (T + = T and R + = R ) is observed with a sinusoidal attern with a

8 462 PROCEEDINGS OF ISMA2008 unity amlitude for the real and imaginary arts of the transmission coefficient and a value close to zero for the reflection coefficients [14]. Small discreancies are caused by a jum of the wall imedance at the lace where the central loudseaker is laced and the connections between the lexiglass measurement ducts and the test duct made of steel. Between 1500 Hz and 2000 Hz additional errors are observed due to resonances between the different microhone ositions, resulting in a singular system of equations for the determination of the left- and right running acoustic waves (eq. (3)). The use of more microhones would eliminate these errors. For the goal of this aer, the validation of the exerimental rocedure and of the aeroacoustic test set-u, this is not further investigated. In the remaining art of this aer results are thus only shown u to a frequency of 1250 Hz. Figure 6: Real and imaginary art of the scattering matrix coefficients for a straight duct with center-laced loudseaker Active two-ort characteristics In order to validate the active two-ort characterization, detailed knowledge about the loudseaker characteristic, laced in the center of the test duct, is needed. This characteristic is determined using a white noise excitation and measuring the acoustic resonse of the system. The result, shown in figure 7, shows a relatively flat resonse starting from 400 Hz. Below this frequency a reduced resonse of the excitation signal by the loudseaker is exected. Figure 7: Characteristic of the central loudseaker, determined with a white noise excitation. The reviously determined scattering matrix is now used to redict the noise sources which are generated

9 AEROACOUSTICS AND FLOW NOISE 463 by the loudseaker in the middle of the acoustic element. In order to simulate the influence from ossible ustream noise sources, generated in the aeroacoustic wind tunnel by e.g. flow noise sources inside the flow conditioner, the test object is excited at its front end by an additional (low-amlitude) broadband signal. The active two-ort characteristics, obtained with a sine excitation of the central loudseaker are shown in figure 8. It is noticed that the tonal behavior at all frequencies (220 Hz, 440 Hz and 1000 Hz) is retrieved. As can be exected, the noise is radiated equally in the ustream and downstream direction. Thus, the exerimental rocedure is able to retrieve narrow-band tonal henomena, even with noise levels, generated by the frontal loudseaker, which are of the same order of magnitude as those of the sine waves generated by the center loudseaker. Figure 8: Active two-ort comonents with a tonal excitation at 220Hz (left), 440Hz (center) en 1000Hz (right). Similar results hold for the broadband excitation signal, shown in figure 9. The whole frequency content of the original signal is retrieved with some small differences at low frequencies which follow the same trend as the loudseaker characteristic (fig.7). Similar as for the tonal excitation, no significant difference between the ustream and downstream radiation is observed. It can be concluded that the active two-ort characterization is able to redict both tonal and broadband noise generation inside an acoustic element. Figure 9: Active two-ort comonents with a broadband excitation. Bottom line: excitation signal; To lines: active two-ort comonents 3.4 Noise radiation The free-field radiation of the aerodynamic noise sources, generated inside the exansion chamber is determined using 9 microhones, laced in a quarter circle (radius=1.00 m, angular resolution 10 o ), around the exit of the test object. In this way, the radiated directivity attern and the free-field ressure field is determined. A comarison between the free-field ressure obtained with the exansion chamber installed as test

10 464 PROCEEDINGS OF ISMA2008 object and with a straight duct of an equal length as the exansion chamber gives a first idea about the aerodynamically generated noise sources since, due to the acoustic labyrinth and the absorbent material of the flow conditioner, the inflow inside the measurements section can be assumed to be anechoic. The insertion difference (ID) is defined as: ) ID = 20log ( with without with with and without the ressure measured inside the semi-anechoic, resectively with and without the exansion chamber installed. Tailie resonance are known to be the dominant tonal aeroacoustic source in the radiated noise sectrum of exansion chamber tye of alications [3]. These resonance cannot be redicted with the active twoort technique since only the exansion chamber is treated as a two-ort element. As a result, it is difficult to directly comare the frequency content of the active two-ort comonents with the radiated noise sectrum. Therefore, the overall sound ressure levels (OSPL), containing also a contribution from the internally generated broadband flow noise sources, are comared for various flow seeds with those of the active twoort arameters. In this way, ossible noise generating mechanisms inside the exansion chamber, which influence the final acoustic radiation can be identified. (7) 4 Discussion of the results 4.1 Descrition of the exansion chamber The geometry of the exansion chamber, studied in this aer, is shown in figure 10. The inlet and outlet ies of the exansion chamber are laced with an offset of 25 mm from the central axis of the exansion chamber. The diameters of the inlet and outlet ies equal 40 mm and the exansion chamber has an exansion ratio equal to The lexiglass muffler is connected, at its inlet, to the flow conditioner and at its outlet by two 0.5 m ie with the same inner diameter as the muffler inlet and outlet ies. The acoustic ressure signals are measured with two and three flush-mounted microhones at resectively the ustream and downstream end of the exansion chamber. Plexiglass test object are used because of otical access and its low wall roughness, which makes it ossible for future research to carry out flow visualization techniques and to easily validate numerical aeroacoustic simulations. Figure 10: Geometry of the exansion chamber. 4.2 No flow Four different measurements are erformed to determine the scattering matrix elements of the exansion chamber by varying the loudseaker excitation (ustream and downstream end of the test object) and the outlet imedance. Since the inlet and outlet diameters equal 0.04 m, the lane wave assumtion, on which the two-ort formulation is based, is valid for frequencies below the cut-on frequency of the traversal modes, equal to aroximately 1.84c 0 /πd 5000 Hz. Figure 11 shows a comarison between the transmission

11 AEROACOUSTICS AND FLOW NOISE 465 and reflection coefficients in both directions of the exansion chamber obtained with the four different measurements and an analytical one dimensional calculation. The recirocal behavior of the system is accurately redicted together with the transmission and reflection coefficients, which resectively show a maximum and minimum value at a frequency of 577 Hz and its higher harmonics, which equals the longitudinal resonance frequencies of the exansion chamber. Threedimensional wave roagation inside the exansion chamber occurs starting from the first traversal resonance of the exansion chamber, occurring at aroximately 1050 Hz. This results in a larger amlitude of the transmission coefficients. Below this frequency only small differences between the 1D solution and the exeriments are observed which are mainly caused by the offset between the inlet and outlet ie, not taken into account in the analytical results. Due to the small signal-to-noise ratio at frequencies with a low transmission coefficient [15], the results show some surious noise on to of the signal in these regions. Figure 11: Comarison between the transmission and reflection coefficients (db) for the exansion in absence of a flow field and reference 1D solution. 4.3 Uniform mean flow The same measurements are taken with a time-uniform mean flow to evaluate the erformance of the current aeroacoustic wind tunnel. The measurements are carried out with different flow velocities by changing the rotational seed of the rootsblower. The mean flow velocities, obtained with hot wire anemometry, near the inlet of the test object that are considered, corresond to 21.7 m/s (18 Hz), 24.7 m/s (20 Hz), 32.7 m/s (25 Hz), 41.3 m/s (30 Hz), 50.7 m/s (35 Hz), 60.9 m/s (40 Hz) and 69.7 m/s (45 Hz). It should be mentioned that the results, discussed in this aer, are erformed in absence of the heat exchanger which makes it difficult to control the temerature of the air flow. Temeratures inside the flow conditioner between 20 o C and 50 o C are observed and measurements over long eriods of time, needed to take a large number of averages, could not be erformed in order not to destroy the flexible iing between the rootsblower and the semi-anechoic room. As already mentioned before, a heat exchanger is now included in the test set-u which allows to control the inlet temerature and to erform measurements over longer eriods of time. Future research is focused on the further imrovement of the reliminary results shown in the remaining art of this aer.

12 466 PROCEEDINGS OF ISMA Passive two-ort arameters Figure 12 shows the amlitude of the ustream and downstream transmission and reflection coefficients at different flow seeds. The recirocal behavior is not noticeable since the resence of a mean flow is resonsible for a different roagation attern in both directions. In comarison with the no flow measurements the resence of surious noise is more aarent and of growing imortance with increasing flow velocities. This is caused by the resence of non-acoustic ressure fluctuations, also known as seudo-sound. Furthermore, the aerodynamic noise generation becomes larger when the flow velocity increases. This can result in the aerodynamic noise generation inside the exansion chamber becoming not negligible anymore with resect to the external acoustic excitation on which the determination of the assive two-ort characteristics is based. A further increase of the sound levels generated by the external loudseaker can result in a nonlinear acoustic wave behavior, which tyically occurs at in-duct sound ressure levels above 140 db [16]. Future research will investigate the use of more microhones in both directions and alying a larger amount of frequency averages to minimize these noisy erturbations and the influence of non-acoustic ressure fluctuations. Figure 12: Comarison between the transmission and reflection coefficients (db) for the exansion chamber at various flow velocities and reference (quiescent) 1D solution. In general, the assive two-ort arameters follow the same trend as the 1D analytical results and the no flow measurements. Besides the increase in surious noise on to of the signal, not much difference is noticed between the exeriments at different flow velocities. Due to the high exansion ratio of the chamber, the Mach number becomes very small inside the chamber. Even at the maximum flow seed of 69.7 m/s, the surface averaged Mach number inside the exansion chamber is of the order of magnitude of 0.01 and, as a result, convection effects are not very ronounced. The maximum values of the transmission coefficients slightly move towards higher frequencies at increasing flow velocity in both the ustream and downstream direction. This is mainly caused by the higher temerature inside the exansion chamber, which results in an increase in sound seed at growing flow velocities. The transmission coefficient in the downstream direction increases with higher flows seeds. This is also noticed when looking at the reflection coefficient in the ustream direction which decreases with increasing

13 AEROACOUSTICS AND FLOW NOISE 467 flow seed, indeendent of the frequency. The oosite henomenon, although less ronounced, is noticed for the reflection in the downstream and transmission in the ustream direction. This effect can be caused by a convective amlification effect or acoustic refraction by the mean flow gradients. Further research is however needed to confirm these findings since the unstable temerature between different measurements and the resence of aerodynamic ressure fluctuations makes it difficult to draw final conclusions from these results Active two-ort arameters One additional measurement without external loudseaker excitation is carried out to determine, in combination with the reviously obtained scattering matrix coefficients, the active two-ort characteristics. These arameters are shown in figure 13 for the maximum flow velocity of 69.7 m/s. For the exeriments carried out with lower flow velocities, similar results are obtained. From this figure it is clear that the major noise radiation is occurring in the downstream direction ( + s > s ). The generated noise sectrum has a broadband nature and some tonal henomena, mainly in the ustream comonents can be notified. Figure 13: Active two-ort characteristics (db) in the downstream + s (left) and ustream s (right) direction in the exansion chamber with a flow seed of 69.7 m/s. This tonal behavior may be caused by a flow-acoustic feedback couling, also known as Rossiter modes [17], where the resence of acoustic resonances, in this case the longitudinal chamber resonances, trigger the shear-layer instabilities. The fact that this tye of tonal flow noise generation is not aarent in the downstream direction can be caused by two reasons. On one hand, the broadband noise generation in the downstream can be of larger amlitude, thus masking the tonal comonents. On the other hand, due to the exerimental rocedure, the noise resent in scattering matrix coefficients, esecially in the downstream direction, results in a large amount of surious noise on the active two-ort comonents. The active two-ort characterization technique is a design tool to identify aerodynamic noise radiation rather rather than to rovide a hysical exlanation of the noise generating henomena that are identified. The acoustic filter is modeled as a black box which only takes into account the noise roagation and radiation at the ustream and downstream end of the acoustic element. As a result, direct information about the noise generating mechanisms occurring inside the element is not resent. For this reason, flow visualization techniques using Particle Image Velocimetry (PIV) are going to be used to further characterize the actual aerodynamic noise generating mechanism inside the exansion chamber Noise radiation A last tye of measurements, erformed on the exansion chamber with a time-uniform mean flow excitation, are free-field measurements inside the semi-anechoic room with the exansion chamber or a straight duct with the same length installed on the test rig. In this way, the additional aerodynamic noise generation that

14 468 PROCEEDINGS OF ISMA2008 is obtained by relacing a straight duct with the exansion chamber (insertion difference) is evaluated. The insertion difference with two different lengths of the tailie (0.225 m and m) are shown for a mean flow velocity of 69.7 m/s in figure 14. A similar acoustic ressure field is obtained by measurements on a number of automotive muffler geometries [3]. Figure 14: Insertion difference (db) of the exansion chamber (flow velocity equal to 69.7m/s with short (0.225m) and long (0.975m) tailie. In contrast to the active bi-ort comonents, a strong tonal behavior is observed at a frequency of about 700 Hz for the short tailie. This eak occurs at the tailie resonance of the system, which could be theoretically estimated to be around 760 Hz (without flow correction) for a length of the tailie equal to m. To verify wether these resonances are generated by tailie resonances a longer duct was used and resonances at aroximately 175 Hz and its higher harmonics are observed, corresonding to the theoretical resonances of a tailie with a length of m. Tailie resonances are thus clearly identified to be the dominant source of tonal aeroacoustic noise generation for the exansion chamber geometry discussed in this aer. Figure 15: Influence of the flow velocity on the acoustic ressure radiation with and without exansion chamber, the active two-ort arameters and the tonal radiated noise. Figure 15 shows the influence of the flow velocity on the overall sound ressure measured in the semianechoic room with and without the exansion chamber. In this figure the influence of the velocity on the active two-ort arameters and the amlitude of the tonal eaks in the radiated noise is also shown. For confined subsonic flows, the radiation efficiency of a monoole, diole (interaction of turbulence with rigid walls) and quadruole (free-stream turbulence) source scales resectively linear, with the third ower and the

15 AEROACOUSTICS AND FLOW NOISE 469 fifth ower of the flow velocity [1]. A linear deendency with resect to the flow velocity is observed for the radiated sound ower without exansion chamber and the tonal tailie resonances, while the third ower scaling is observed for the sound ower radiation with exansion chamber and both active two-ort comonents. This evidences that the major broadband noise radiation is caused by the interaction of turbulence with the rigid walls of the exansion chamber while tonal noise sources are generated by tonal resonances. The relative imortance of broadband noise sources with resect to the tonal noise sources grows with increasing flow velocity due the the third ower scaling with the flow velocity. 5 Conclusions This aer resents and validates an active two-ort measurement technique to study the effect of a mean flow field on the roagation and generation of acoustic waves for confined subsonic flow alications, commonly encountered in HVAC ducts or automotive exhaust systems. For this urose a low-noise, low-turbulence aeroacoustic wind tunnel is used in which the flow is generated by a three lobe rootsblower. The three-ste rocedure to determine the active two-ort arameters consists in the determination of the left- and rightrunning acoustic waves based on multile microhone measurements; the characterization of the scattering matrix comonents based on a combined multile-load, multile-source technique; and the identification of the active noise generation comonents in the ustream and downstream directions based on the reviously obtained arameters. This measurement technique is successfully validated for the no flow case of a straight duct with center loudseaker. The reliminary results, obtained on a simle exansion chamber configuration, are used to further refine the aeroacoustic test set-u and the exerimental two-ort characterization methodology. A very good agreement of the scattering matrix comonents between the exeriments and a analytical 1D solution are obtained for a quiescent medium. The same behavior is observed in the time-uniform flow case, although a large amount of noise, caused by aerodynamic ressure fluctuations and large temerature fluctuations make it difficult to investigate the effect of a flow field on the transmission characteristics of the exansion chamber. The noise generation inside the exansion chamber has a broadband nature and mainly occurs in the downstream direction, scaling with the third ower of the velocity. This indicates a diole aerodynamic noise generation mechanism, caused by the interaction of turbulence with the exansion chamber walls. It is shown that this mechanism is resonsible for the final broadband acoustic radiation, measured inside a semi-anechoic room. Tonal noise radiation, scaling linearly with the flow velocity is also observed and is caused by the excitation of tailie resonances, not taken into account in the two-ort formulation. Future research will focus on an imrovement of the aeroacoustic test set-u including a heat exchanger to control the temerature variations. This also ermits to measure over longer eriods of time which makes it ossible to incororate a larger amount of averages in the data rocessing. Together with the inclusion of more ressure transducers the resence of non-acoustic noise on the measurement results can be reduced, which enables to study in more detail the mean flow effects on the acoustic roagation and the radiation of aerodynamically generated noise sources. The generation mechanisms of these flow noise sources will be investigated more thoroughly using flow visualization techniques. In this way, this work is the first ste towards the construction of an aeroacoustic test set-u which allows to study the aeroacoustic behavior of confined subsonic flow alications and to validate advanced numerical methodologies to simulate both noise generation and roagation mechanisms in the resence of non-uniform mean flows. Acknowledgements The authors want to acknowledge the financial suort given by the Research Foundation of Flanders (FWO G ) and the Institute for the Promotion of Innovation by Science and Technology in Flanders (SBO- IWT ).

16 470 PROCEEDINGS OF ISMA2008 References [1] Desantes J.M., Torregrosa A.J., Broatch A., Exeriments on Flow Noise Generation in Simle Exhaust Geometries, Acta Acustica united with Acustica, Vol. 87, (2001). [2] Maurel A., Ern P., Zielinska B.J.A., Wesfreid J.A., Exerimental Study of Self-Sustained Oscillations in a Confined Jet, Physical Review E, Vol. 54, (1996). [3] Davies P.A.O.L, Holland K.R., The Observed Aeroacoustic Behaviour of Some Flow-Excited Exansion Chambers, Journal of Sound and Vibration, Vol. 239, (2001). [4] Munjal M.L., Acoustics of Ducts and Mufflers, John Wiley and Sons (1987). [5] Davies P.A.O.L., Practical Flow Duct Acoustics, Journal of Sound and Vibration, Vol. 124, (1988). [6] De Roeck W., Desmet W., Exerimental Analysis of the Aerodynamic Noise Generating Mechanisms in a Simle Exansion Chamber, AIAA-aer (2008). [7] Lavrentjev J., Åbom M., Boden H., A measurement Method for Determining the Source Data of Acoustic Two-Port Sources, Journal of Sound and Vibration, Vol. 183, (1995). [8] Morfey C.L., Sound Generation and Transmission Ducts with Flow, Journal of Sound and Vibration, Vol. 14, (1971). [9] W. De Roeck, V. Solntseva and W. Desmet, Numerical methodologies to redict the noise generation and roagation mechanisms in multile exansion chambers AIAA-aer, No , [10] Dokumaci E., On the Proagation of Plane Sound Waves in Ducts Carrying an Incomressible Axial Mean Flow Having an Arbitrary Velocity Profile, Journal of Sound and Vibration, Vol. 249, (2001). [11] Munjal M.L., Doige A.G., Theory of a Two-Source Location Method for Direct Exerimental Evaluation of the Four-Pole Parameters of an Aeroacoustic Element, Journal of Sound and Vibration, Vol. 141, (1990). [12] Åbom M., Measurment of the Scattering-Matrix of Acoustical Two-Ports, Mechanical Systems and Signal Processing, Vol. 5, (1991). [13] Lung T.Y., Doige A.G., A Time-Averaging Transient Testing Method for Acoustic Proerties of Piing Systems and Mufflers, Journal of the Acoustical Society of America, Vol. 73, (1983). [14] W. De Roeck, Hybrid Methodologies for the Comutational Aeroacoustic Analysis of Confined Subsonic Flows. Ph.D.-thesis, K.U.Leuven, Belgium, [15] Åbom M., Boden H., Error Analysis of Two-Microhone Measurements in Ducts with Flow, Journal of the Acoustical Society of America, Vol. 83, (1988). [16] Davies P.A.O.L, Holland K.R., The Measurement and Prediction of Sound Waves of Arbitrary Amlitude in Practical Flow Ducts, Journal of Sound and Vibration, Vol. 271, (2004). [17] Rossiter J.E., Wind Tunnel Exeriments on the Flow Over Rectangular Cavities at Subsonic and Transonic Seeds, Royal Aircraft Establishment, technical reort No (1964).