Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297 www.springerlink.com/content/1738-494x
DOI 10.1007/s12206-013-0851-y
Experimental investigation on aero-acoustic characteristics of a centrifugal compressor for the fuel-cell vehicle† Kyoung-Ku Ha1,*, Tae-Bin Jeong1, Shin-Hyoung Kang1, Hyoung-Jin Kim2 Kwang-Min Won2, Chi-Yong Park3, Woo-Youl Jung3 and Kyung-Seok Cho3 1
School of Mechanical and Aerospace Engineering, Seoul National University, Seoul, 151-742, Korea 2 Power Train NVH Team, Hyundai Motor Company, Hwaseong, Gyeonggi, 445-706, Korea 3 Advanced Engineering Team II, Halla Climate Corporation, Daejeon, 306-230, Korea (Manuscript Received May 10, 2012; Revised February 18, 2013; Accepted May 31, 2013)
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Abstract A variety of centrifugal compressors are used in various fields of industry such as aircraft, home appliances, and vehicles. Comfort and quietness are important in these uses. As a result, noise has become an important consideration in compressor design besides the conventional performance parameters such as efficiency and pressure ratio. However, compressor noise has been difficult to understand because of the lack of information. The aim of this paper is to investigate the aero-acoustic characteristics of a centrifugal compressor for the fuelcell vehicle by experiments. The existing compressor system is modified to measure the internal pressure fluctuation at the impeller inlet, the impeller outlet and the diffuser outlet. Four microphone probes are also installed to determine the external noise levels and spectra of the compressor in an airtight room according to the RPM and mass flow rate. The test results show the possibility to tell the relative noise level of a centrifugal compressor with the internal pressure data. The external microphone signals have relation to the internal pressure signals. They have similar patterns and spectra. It is a noteworthy phenomenon because it is easier and inexpensive to predict pressure behaviors than noise characteristics of centrifugal compressors. The dominant noise source is the tonal noise during normal operation. But the broadband noise component due to the turbulent flow in the compressor increases during low flow rate operation. Computational simulations are carried out to describe these phenomena and to identify noise indicators. The turbulence kinetic energy and the pressure distribution obtained from CFD results may be indicative of the relative noise intensity of the compressor. The experimental facility, instrumentation and simulation conditions are described, and the results are presented in this paper. Keywords: Blade passing frequency; Centrifugal compressor; Fuel-cell vehicles; Impeller; Noise; Noise indicator; Tonal noise; Vaneless diffuser ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1. Introduction A centrifugal compressor is composed of an impeller, a diffuser, and a volute in general. It applies the centrifugal effect because of the change of radius between the inlet and the outlet. Therefore it has the advantage of a high pressure ratio in spite of its small size, and because of these properties, it is widely used in new domains such as aircrafts, home appliances and vehicles. Comfort and quietness are important in the daily uses of these applications. The subject of this study is a centrifugal compressor for the fuel-cell vehicle. It produces compressed air and sends it to the fuel-cell stack. It consumes the most power and it is the only rotating part with a high angular velocity in the system. As a result, in addition to efficiency and pressure ratio, noise has become an important consideration in the compressor design. But it is difficult to un*
Corresponding author. Tel.: +82 2 880 7118, Fax.: +82 2 883 1215 E-mail address:
[email protected] † Recommended by Associate Editor Simon Song © KSME & Springer 2013
derstand compressor noise because of the lack of information. The aim of this paper is to investigate the aero-acoustic characteristics of a centrifugal compressor in detail and to find noise indicators from computational simulations. If indicators were available, they would be helpful in determining the design point of a quieter compressor in the early stage of design of a development project. Many researches on understanding the noise characteristics of turbo machinery have been performed in the past. But their subjects have been mainly axial machinery [1-4] or fans [5-7] because of the demand for quieter aircrafts or air-conditioning. Neise showed the noise characteristics of various turbomachines such as axial compressors and centrifugal fans, and the mechanisms by which they are generated [2, 3, 5]. According to these studies, blade tonal noise is the main type of noise during normal operation, induced by the pressure difference between the pressure surface and the suction surface at the trailing edge. But the broad band noise caused by turbulent flow is more pronounced during off-design operations. Subse-
3288
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297
quently tonal noise has been predicted by the use of the linearized method or the Ffowcs Williams and Howkings equation with the pressure field data [4, 6]. Rulik et al. [10] carried out the detached eddy simulation (DES) with the hybrid uRANS/LES methods to investigate the aerodynamic noise of an airfoil. Simple experimental equations to predict the noise level using the relative velocity [7] or the volume flow rate, the head and the efficiency of the compressor [8] have been suggested recently. However, it is difficult to apply the above methods to centrifugal compressors because of the intricate structure and the high speed operation of these compressors. Moreover, the noise produced by air compressors is radiated from several parts such as the filtered inlet, the compressor casing and the interconnected piping. Since the year 2000, thanks to the growth in demand and advancements in technology, the instability of centrifugal compressors and its related noise have been studied. Acoustic phenomena at the inlet and the outlet of centrifugal compressors were measured in detail [11, 12]. Vaneless and vaned diffusers have been studied [11, 14]. According to these authors, the blade passing frequency (BPF) leads to the outlet duct noise and the intensity of the instability is linked to the pressure difference between the pressure and suction surfaces of the blade during normal operation. Rafael et al. [15] examined the relationship between the pressure fluctuations and the noise characteristics of a centrifugal fan. They carried out a full annulus unsteady CFD simulation and an experiment. Lee et al. also investigated numerically as well as experimentally the characteristics of the inlet and the outlet duct noise of an impeller [16]. These methods, however, are too time-consuming to apply in the early stage of design. A good design point should be first found. But it is difficult to do because of the lack of information. So, this experimental study is investigates the noise characteristics of a centrifugal compressor. The existing system is modified to measure the pressure fluctuation at the impeller inlet, the impeller outlet and the diffuser outlet. The results show the relation between the internal pressure signals and the outer microphone signals; they have a similar pattern and spectra. The dominant noise source is the tonal noise caused by the blade passing during the normal operation. But the broadband noise component due to the turbulent flow in the compressor becomes stronger during the low flow rate operation especially in the near stall region. The steady CFD simulations of the compressor including a section of the passage are also carried out to describe these phenomena. It is definitely impossible to tell the noise itself with the steady status. But it would be more practical to find indicators and the noise generation mechanism. Such finds would save time and expenses in the early stage of design of a development project. Moreover, the turbulence kinetic energy and the pressure distribution from the CFD results can indicate the relative noise intensity of a compressor.
Fig. 1. Configuration of the centrifugal compressor system for the fuelcell vehicle.
Fig. 2. Test rig and setup for the acoustic measures of the compressor.
2. Experiment and numerical simulation 2.1 Experimental facility and instrumentation The test centrifugal compressor is typical for the fuel-cell vehicle. It is composed of an inlet duct with a filter and a silencer, the compressor assembly and the humidifier. The configuration of the system is shown in Fig. 1. The entrance of the inlet filter was exposed to the atmosphere and the outlet duct was linked directly to the exhaust pipe without the humidifier. The silencer was eliminated to enhance the responsiveness of the external microphones in this study. Four microphones in an isolated were located near the inlet duct, near the outlet duct, 1 meter above and beside the compressor. HEAD acoustics equipment was used to estimate the sound pressure level according to Eq. (1) below and the frequency spectra. Their names are indicated in Fig. 2, which shows the test rig for the centrifugal compressor and the setup for the acoustic measures. SPL ( LP ) = 20log10
p , pref = 2 ´ 10-5 [Pa] pref
(1)
3289
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297
Table 1. Specifications of the test compressor.
Table 2. CFD conditions.
Design RPM
40,000
Grid / Solver
No. of blades
18 (9 main, 9 splitter)
y+
40~80
b2/r2
0.0643
Turbulence model
k-ε model
r5/r2
1.286 Inlet B.C.
Total pressure, Total temperature, Flow vector
Outlet B.C.
Mass flow rate
Interface
Frozen rotor
β’1t
60°
β’2mean
55°
CFX turbogrid 11 / CFX 11
Fig. 4. Cross-section of a test compressor and locations of three pressure transducers.
Fig. 3. Schematic view of the compressor test facility.
A schematic view of the compressor test facility is shown in Fig. 3. Two static pressure sensors, ONDI dTRANS P30, and T-type thermocouples were installed at the inlet and the outlet duct, and they were used to estimate the aero-dynamic performance of the compressor. The flow control valve using a step motor and the rotational speed of the impeller were automatically controlled by the computer system and monitored by the mounted meters. The impeller was driven by a DC motor directly. Table 1 shows specifications of the centrifugal compressor. The design RPM was 40,000 and the number of blades was 18. So the blade passing frequency (BPF) was 12,000 Hz at the design speed. The r5 ratio to the impeller exit radius r2 is about 1.286. It had a short vaneless diffuser because of size restriction. The test compressor is sketched in Fig. 4. It shows the cross-section of the stage and the locations of the internal pressure transducers. Three static pressure transducers, Kulite XCQ-093 25D, were installed at the impeller inlet, the impeller outlet and the diffuser outlet, respectively, to measure pressure fluctuations. The circumferentially the transducers were near the volute tongue because of the violent pressure condition. They could measure a static pressure difference up to 25 psi and their maximum sampling rate was 240 kHz. The signals could be used to investigate the internal noise level and the spectra as well as the time-averaged static pressure.
Fig. 5. Computational grid.
2.2 Numerical method The compressor was examined by numerical analysis using the commercial CFD code, CFX 11 together with the k-ε turbulence model. ANSYS Turbogrid was used to generate the grid. Fig. 5 shows the computational mesh. An impeller passage with a main blade and a splitter blade and one passage for the vaneless diffuser were simulated as the computational domains. The periodic boundary condition was applied at the sides. The total number of elements was about 450 k including the inlet and outlet blocks. In the numerical simulations, the flow was assumed to be at a steady state with reference to the relative frame. Then the circular pressure distributions were assumed as a rotating perturbation in this study. The frozenrotor condition was adopted at the interface between the two stationary frames and the rotating frame. Numerical conditions are listed in Table 2.
3290
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297 1.8
0.002
0.0016
Total to static pressure ratio
Exp1, 40k RPM Exp1, 30k RPM Exp2, 40k RPM Exp2, 30k RPM
y
0.0012
0.0008
1.6
EXP, PRTS,diff. EXP, PRTS,imp. CFD, PRTS,diff. CFD, PRTS,imp.
near stall
1.4
1.2
0.0004
40k RPM 30k RPM
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
1
0
0.2
0.4
0.6
f
0.8
1
1.2
1.4
1.6
1.8
m/mref
1
0.8
Efficiency
Fig. 7. Performance map of the compressor (Exp. Vs CFD).
Exp1, 40k RPM Exp1, 30k RPM Exp2, 40k RPM Exp2, 30k RPM
0.6
0.4
0.2
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
f
Fig. 8. One dimensional volute loss model and equations.
Fig. 6. Dimensionless performance map and the stage efficiency.
3. Performance of the compressor stage The performance map of the compressor is shown in Fig. 6. It shows the dimensionless pressure coefficient and the stage efficiency. Eqs. (2) and (3) are used to define them. Q DP , y= ND 3 r N 2D2 & P (T7 s - T0 ) mC h= Power ( Motor Input )
f=
(2) (3)
where ΔP is the static pressure difference between the duct inlet and the volute outlet. T7s denotes the isentropic static temperature at the volute exit. The results show good agreement in normal operation. But there is a slight difference between the results in low flow rate operation according to rotational speed, because the trend of the intake losses had departed from the rule. Generally, the total pressure loss of an intake system is proportional to the inlet flow velocity. The coefficient of the present system was 4.36, but this was not correct in the low flow rate region. So, a repeat test was carried out to verify the first test. In Fig. 6, the results of the first test are represented in dotted lines. They are in keeping with the solid lines of the second test. The robustness of the experiment was proven by these figures.
The experiment and the simulation results are compared in Fig. 7, which also shows the near stall line based on the experimental result of ∂p/∂m at the impeller outlet (∂p/∂m≒0) in this study. The mass flow rate at the best efficiency point of the 40,000 RPM was selected as the reference mass flow (mref). In this figure, the dotted and the solid lines with the gradient symbol show good agreement. These lines indicate the pressure ratio at the impeller outlet. But the results at the diffuser outlet show about 3% difference from each other, seemingly due to the simple modeling. The one dimensional loss model in Fig. 8 was used to calculate volute exit conditions instead of the real feature simulations. And there was a notch and a gap between the impeller and the diffuser. These features were not reflected in the simulations, so the simulations predicted less loss in the passage. However, the simulations were carried out to describe the flows around the impeller to find noise indicators. The region of interest considered to be a tonal noise source during design operation was the impeller outlet. From this point of view, the impeller outlet shows sufficiently good results as shown in Fig. 7. A l5 = tan a 5 , AR = 7 A 5 CP =
2(l52 - 1/ AR ) for l5 AR > 1 AR (1 + l52 )
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297
CP =
l52 - 1/ AR for l5 AR £ 1 1 + l52
Km =
1 1 + l52
Kq =
(l5 - 1/ AR ) 2 for C7 < Cq 5 . 1 + l52
portional to the stress tensor pij which is given by æ ¶u ¶u pij = pd ij + meff ç i + j ç ¶x j ¶xi è
4. Aero-acoustic characteristics 4.1 Noise of the centrifugal compressor The sound spectrum of the centrifugal compressor included a tonal noise and a broadband noise. The tonal noise component is associated with the pressure fluctuation caused by blade passing. The pressure difference between the suction and the pressure surface is the main source of this noise. Broadband noise component is often a major part of noise, especially during low flow rate operation. It is due to the inflow turbulence, the tip vortex, the trailing edge and the rotating stall. The intense pressure fluctuations of turbulent flow are also cause broadband noise. To make a connection between fluid dynamics and acoustics, Lighthill rearranged the Navier-Stokes equations into an inhomogeneous wave equation. It presents a model for the acoustic field. ¶ 2Tij ¶2r ' - c02Ñ 2 r ' = 2 ¶t ¶xi ¶x j
where Tij = r uiu j - pij + ( p - p0 - c ( r - r 0 ))d ij .
(4)
The above equation is the celebrated Lighthill equation of aero-acoustics with a source term on the right-hand side. Tij is the Lighthill tensor and the first term is the unsteady convection of flow, the second is nonlinear acoustic generation processes, and the last term describes the transport of momentum due to viscous stresses. Lighthill’s theory allows to compute the sound field by considering fluctuating stress field but it is valid only for an unbounded fluid. To describe the aerodynamic noise radiated from blades the influence of boundaries cannot be neglected. So Ffowcs Williams and Hawkings included the influence of moving surface. 2 ¶2r ' 2 ¶ r' c 0 ¶t 2 ¶xi2
¶ 2Tij ¶xi ¶x j
-
¶ ¶xi
æ ¶f ç p d( f ) ç ij xj ¶ è
ö 2 ¶u ÷ - meff k d ij . ÷ 3 ¶xk ø
(6)
It contains the viscous stresses and the aerodynamic pressure. The third term of FW-H, Eq. (5) is a monopole term proportional to the acceleration of the surface in the normal direction. This is a source that occurs as a result of the fluid displaced by the swept volume of a moving boundary and the strength is a function of the normal velocity of the surface. FW-H equation is useful in case where the aerodynamic forces of a moving body are known. But in practice, it is difficult to figure out the pressure field of a centrifugal compressor because of the high rotational speed and the intricate structure. However, during the design point operation when the effects of turbulence on noise are of minor importance and the viscous component is very small compare to the normal force component, it may turn into a simple problem. This is the start point of my argument. If the compressor operate at the design point only, so the noise source acted like monopole, we could expect and evaluate the relative noise strength of the compressor with pressure data from a simple steady simulation. Verifications of this hypothesis were carried out and the results are described in the next paragraph in detail. 4.2 Impeller inlet
2 0
=
3291
ö ¶æ ¶f ö r u d(f ) ÷+ ÷ ÷ ¶t çç 0 n xi ÷ø ¶ è ø
(5)
where δ (f) is the delta function and f is positive in the fluid region. This equation is commonly labeled the FW-H equation. The first term on the righthand side is identical to Lighthill’s term. Its strength is given by the Lighthill tensor and it represents the quadrupole source. The second is the dipole sources arising from the fluctuation of forces on the surface. It is pro-
The A-weighted sound pressure levels from the impeller inlet transducer are presented in Fig. 9. It shows a dramatic change at point I. The noise levels are similar before this point. But the value at point I is about 20 dB higher than that at the point II. This phenomenon can be explained by Fig. 10. This figure shows the sound pressure spectra at the impeller inlet. The spectra of II and III look the same. The blade passing frequency (12,000 Hz) and its harmonics are dominant. And the difference is shown only in the low frequency section (1.5 to 6 kHz). But the spectra of the operating point I is quite different from the other spectra. The noise levels increase in all ranges. So, the noise can be considered as broadband noise without prominent discrete tones. This is a property of broadband noise caused by turbulent flows. Fig. 11 shows the circular averaged turbulence kinetic energy (TKE) contours on the meridional surface, which were obtained from CFD simulations. TKE is characterized by measured root mean square velocity fluctuations like Eq. (7). k=
1 (u1¢ ) 2 + (u2¢ ) 2 + (u3¢ ) 2 . 2
(
)
(7)
There is a weak tip clearance flow in the impeller shroud at the best efficiency point III. But the clearance flow and the secondary flow become stronger as the mass flow decreases. The disturbance caused by the blades seems to affect the up-
3292
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297
180
Sound prssure level [dBA]
40,000 RPM 30,000 RPM
I
160
II
III
140
(a) Point I 120
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
(b) Point II
(c) Point III
Fig. 11. Turbulence kinetic energy contours on the meridional surface.
m/mref
Fig. 9. SPL of the impeller inlet transducer.
(a) Point I
(b) Point II
(c) Point III
Fig. 12. Circular averaged streamlines on the meridional surface (CFD).
Turbulence kinetic energy @ Imp. inlet
500 40,000 RPM 30,000 RPM
400
I
300
200
II
100
III
0
-100
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
m/mref
Fig. 13. Turbulence kinetic energy at the impeller inlet (CFD). Fig. 10. A-weighted noise spectra of the impeller inlet at I, II, III.
stream during low flow rate operation, according to (a) and (b). The black arrows indicate the location of the impeller inlet transducer. The TKE around it is increased rapidly at point I. This trend agrees with the result shown in Fig. 12. Circular averaged streamlines on the meridional surface are shown in these figures. A separation and backflows present in the impeller inlet as indicated by Fig. 12(a). But the separation and backflows are diminished at the best efficiency point (c). The averaged values of the TKE on the cross section where the inlet sensor is placed are shown in Fig. 13. The tendency of
the values is similar with that of the inlet sound pressure level in Fig. 9. The value increases greatly at point I. The flowdynamic similarity is also valid in the 30,000 RPM line. These findings suggest that the TKE can be an indicator of impeller inlet noise because the pressure fluctuation caused by the impeller blades is small. But the TKE is available during low flow rate operation as a noise indicator. No clear difference of the TKE exists under common operation. The noise level of a machine is a function of the impeller and the diffuser outlet condition in this case.
3293
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297
Sound pressure level [dBA]
180 40,000 RPM @ Imp. outlet 40,000 RPM @ Diff. outlet 30,000 RPM @ Imp. outlet 30,000 RPM @ Diff. outlet
a
160
b
c
d
140
(a) 120
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
m/mref
Fig. 14. SPL of the impeller outlet and the diffuser outlet transducers.
4.3 Impeller outlet and diffuser outlet (b)
(c) Fig. 15. A-weighted noise spectra of the impeller and the diffuser outlet. 130
Imp OUT Diff OUT
Lp [dBA]
120 110 100 90 80 0
1k
2k
3k
4k
5k
6k
Frequency [Hz] (a) 130
Imp OUT Diff OUT
120
Lp [dBA]
Fig. 14 shows the A-weighted sound pressure levels at the impeller exit and the diffuser exit. Filled symbols represent the level at the diffuser exit. The impeller has a minimum noise level at the best efficiency point (c). It shows a mild increase at the off-design points (b), (d). Then the level rises rapidly in the near stall operation. Sound pressure spectra at the impeller exit and the diffuser exit are also shown in Fig. 15. These figures show the changes of the sound pressure spectra according to mass flow rate. The BPF components are visible at every flow rate. Especially, the sound pressure spectra are subject to it and its harmonics at point (c). These are called the tonal noise. Fig. 15(b) has a similar pattern with (c) in the high frequency region. But the broad band noise in the low frequency region is increased in this case. The circle in the Fig. 15(b) clearly describes this phenomenon. As the mass flow decreases, the broad band noise becomes stronger. Finally, the tonal components are no longer dominant in the near stall range, as represented by point (a). The BPF (12,000 Hz) components are still shown under this operation. But its near harmonics appear distinctly, and new dominant components exist in the low frequency region. Fig. 16 shows the enlarged view of the rectangular area in Figs. 15(a), (c). This section is important because the human ears become sensitive to the noise in this frequency range. The noise level increases in the overall range by 2-5 dB; this is the broadband phenomenon caused by the turbulence and also observed at the impeller inlet. But unlike the broadband noise at the inlet, the impeller outlet noise spectrum has prominent discrete tones during the near stall operation. The dominant frequency is about 91.5% of the fn, 610 Hz. It is almost certain that this is the rotating stall frequency. The rotating stall has a speed in the range of 10 to 30% of fn in general. But a high frequency rotating stall was reported by Kämmer [18]. According to him, it developed when the stall occurred at the inlet and the outlet of the impeller at the same time. He showed that it was generated in the impeller by the periodic break down of energy transfer from the rotor to the flow. It was a kind of super-
110 100 90 80 0
1k
2k
3k
4k
5k
6k
Frequency [Hz] (b) Fig. 16. Enlarged view of the rectangular area in Figs. 15(a), (c).
3294
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297
(a)
Sound pressure level [dBA]
88 40,000 RPM 30,000 RPM 20,000 RPM
80
72
64
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
m/mref
Fig. 18. A-weighted sound pressure level of the 1 m Upper Mic. 104
(b)
Sound pressure level [dBA]
40,000 RPM 30,000 RPM
96
88
80
72
(c)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
m/mref
Fig. 17. Comparison of levels in divided frequency ranges.
(a) 104
4.4 External microphones and the noise indicator The A-weighted sound pressure levels of the upper mic. according to the operating conditions are plotted in Fig. 18. The mass flow rate at the best efficiency point of the 40,000 RPM was selected as the reference mass flow (mref). In this figure,
40,000 RPM 30,000 RPM
Sound pressure level [dBA]
positioning of the inducer and diffuser stalls and its frequency was 89% of the fn. Figs. 14 and 15 show the change of the SPL at the impeller and the diffuser outlet, too. The sound pressure level defined in equally divided frequency ranges are plotted in Fig. 17. It is more clearly visible that the noise is reduced by the vaneless diffuser. But the aspects of the noise reduction are different according to the mass flow rate. Fig. 17(a) shows the highest value of noise in the low frequency region and it shows a small decrease because it originated from the turbulence generated by the rotating stall, the separation and the tip clearance flow. But, in Fig. 17(c), the noise level in the BPF range becomes the dominant component. It shows a large decrease because of the mixing effect. The pressure fluctuation should be reduced at the diffuser outlet. Fig. 17(b) shows a neutral phenomenon. As a result, the overall sound pressure level at the diffuser exit shows behavior in the shape of Fig. 14, a small reduction at low flow rate and a large reduction at the high flow rate. These results are in agreement with Fig. 15.
96
88
80
72
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
m/mref
(b) Fig. 19. SPL of (a) the inlet duct and (b) the outlet duct microphones.
the result at the 20,000 RPM operation shows a flat pattern. It seems that the drive noise was stronger than the aero-acoustic noise. However, the aerodynamic characteristics of the compressor are described by the lines of rotation speed over the 20,000 RPM. The noise level is minimized at the best efficiency point and increased rapidly in the low flow rate region of the 40,000 RPM case. A similar behavior is observed on the line of the 30,000 RPM. Fig. 19 shows the A-weighted SPL at the inlet duct (a) and the outlet duct (b) microphones. Patterns are similar with the
3295
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297 140 100
Lp [dBA]
0.15
Imp OUT Diff OUT MIC_OUT MIC_U
120
Exp. Steady CFD
0.1
80 60
0.05
40
p' /pmean
20 0 0
2k
4k
6k
8k
10k
12k
14k
16k
18k
20k
0
Frequency [Hz] -0.05
(a) Best efficiency point 140 100
Lp [dBA]
-0.1
Imp OUT Diff OUT MIC_OUT MIC_U
120
-0.15
80
0
0.2
0.4
0.6 revolutions
60
0.8
1
40
Fig. 21. A-weighted noise spectra of the impeller, the diffuser outlet transducers and the upper mic., the outlet duct mic.
20 0 0
2k
4k
6k
8k
10k
12k
14k
16k
18k
20k
Frequency [Hz]
1200
(b) Near stall operation
P' rms (Exp.) fitting P' rms (steady CFD)
1000
Fig. 20. A-weighted noise spectra of the impeller, the diffuser outlet transducers and the upper mic., the outlet duct mic.
µ N2
patterns of the internal pressure transducers at the impeller and the diffuser outlet. The inlet duct noise levels follow the similar pattern of the impeller outlet noise level in Fig. 14. It shows that the external inlet duct mic. was pushed into the pressure signal at the impeller exit and the dominant noise source of a centrifugal compressor is a blade loading noise. Fig. 19(b) shows good agreement with the SPL of the diffuser outlet transducer in Fig. 14. The outlet duct seems to be affected by the diffuser outlet conditions directly. Patterns in Figs. 18 and 19 are similar with the patterns of the internal pressure signals in Fig. 14 along the RPM and the mass flow rates. They have the same aspects and they have minimum value at the same flow rate. The relevancy between the internal transducers and the external microphones is also shown in Fig. 20. A-weighted spectra of the internal transducers are compared with those of the external microphones. The signals take on a similar aspect each other. The BPF and its harmonics are dominant at the best efficiency point. But the broadband noise increase of the internal sensors and the rotating stall frequency component are found during low flow rate operation, as shown in Fig. 20(b). The external spectra reflect those components well. Fig. 21 shows the static pressure fluctuations during one revolution at the impeller exit. CFD results are compared with the experimental results. The circular pressure distributions are used to calculate the p'rms in this study. The distributions were assumed as a rotating perturbation. Each line has eighteen peaks. The light dashed line represents the measured data of pressure perturbation at the highest efficiency point in 40,000 RPM operation. The solid line indicates the results of the steady CFD simulation. Since the blade effect is principal and effects of turbulence in the flow on noise are minor at the
p' rms
800
600
400
200
0
0
10000
20000
30000
40000
50000
RPM
Fig. 22. p'rms values according to the RPM at the impeller exit (along the highest efficiency point, Exp. Vs. CFD).
highest efficiency point, the two lines have similarities. p'rms values at the impeller exit are also presented in Fig. 22. Those are the results of the highest efficiency points along the operation RPM. The black rectangles represent p'rms values obtained from an experiment. They are proportional to the square of the RPM. The red triangles indicate p'rms from steady simulations. They have about 75-78% values of the experimental data at each point. But they are also proportional to the square of the RPM. This is a characteristic of a noise source acting like monopole. So, this result confirmed p'rms to be a noise indicator which makes the noise problem simple at the design point operation. It is almost certain that p'rms can tell the relative strength of a noise source at the highest efficiency point, to say the least of it. Fig. 23 indicates the p'rms values at the impeller exit calculated in the steady CFD simulations. These figures show their changes according to the flow coefficient and RPM. It is possible to split operating range into three regions in this figure. The central region B is the normal operation zone. This region is distinguished from the unstable region on the left side A, near stall operation. Section C is the high flow rate region. Fig.
3296
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297
1200 40,000 RPM 30,000 RPM
1000
P' rms/r
800
600
400
200
0
A 0
0.2
0.4
0.6
B 0.8
1
1.2
C 1.4
1.6
1.8
2
2.2
f
Fig. 23. p'rms values calculated in the steady CFD simulations along the dimensionless flow coefficient. 180 40,000 RPM @ Imp. outlet 40,000 RPM @ Diff. outlet 30,000 RPM @ Imp. outlet 30,000 RPM @ Diff. outlet
Sound pressure level [dBA]
170
160
150
140
130
120
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
f
Fig. 24. SPL of the impeller outlet and the diffuser outlet transducers along the dimensionless flow coefficient.
24 shows the internal noise level at the impeller outlet and the diffuser outlet obtained from the pressure transducers. It is the same figure with Fig. 14, but the x axis is replaced with the dimensionless flow coefficient. The p'rms in Fig. 23 and the measured sound pressure level in Fig. 24 have a similar trend in section B, the normal operation zone. They have minimum values at the same flow coefficient. Especially the diffuser outlet lines are quite similar to the p'rms lines. This shows that the p'rms obtained from CFD simulations can be an indicator of the noise level of the present compressor in normal operating conditions, too.
5. Conclusions In this study, aero-acoustic characteristics of a centrifugal compressor were described. The existing system was modified to measure the internal pressure fluctuations around the impeller. The near-field spectra and sound pressure levels were described with these data. They showed that aero-acoustic noise became important over 20,000 RPM and the dominant noise source was the tonal component in the design and the
high flow operation. But the broadband noise caused by the turbulence became stronger and led to the rapid increase of the noise level in low flow rate operation. Especially, there was prominent tone at the impeller and the diffuser outlet unlike the inlet at the near stall operation. It seems to be the high frequency rotating stall and its speed is about 91.5% of fn. It is also visible that the SPLs were reduced through the vaneless diffuser but, the aspects varied with the flow rate. There was a close relationship between internal pressure signals and external noise levels of microphones. They showed similar patterns and spectra. The steady CFD simulations including a single passage were also carried out to illustrate aero-acoustic characteristics and to find the SPL indicators. It is impossible to identify the noise itself with the steady simulations. But it is probably the best practical way to consider the noise in the early design stage of a development project. And it could be helpful to determine a good design point for a quieter compressor in the early stage of development. In conclusion, this paper showed that the p'rms obtained from CFD simulations have a similar trend with the SPL lines of external microphones and they have minimum values at the same flow coefficient. They are also proportional to the square of the RPM in normal operation. Therefore it is certain that p'rms obtained from CFD simulations can be an indicator of the noise level of the present compressor in normal operating conditions. It can indicate the relative intensity of noise and make the noise problem simple in this study. However, the conclusion about the noise indicator p'rms is restricted on the present compressor at this time. It is necessary to verify this hypothesis on another centrifugal compressor. In addition, each line has the similar pattern according to the flow coefficient in Figs. 23 and 24. It shows that noise characteristics of a centrifugal compressor are related with the aerodynamic conditions of the internal flow. It also shows the possibility of a dimensionless noise curve of the centrifugal compressors. So there is a necessity for doing additional research to check the dimensionless possibility.
Acknowledgment This work supported by Research Program supported by the Hyundai Motor Company, Republic of Korea.
Nomenclature-----------------------------------------------------------------------BPF f fn LP p'rms r SPL TKE
: Blade passing frequency : Frequency : Rotor shaft frequency : Sound pressure level (Pref=20 [μPa]) : Root mean square pressure : Radius [mm] : Sound pressure level : Turbulence kinetic energy
K.-K. Ha et al. / Journal of Mechanical Science and Technology 27 (11) (2013) 3287~3297
Greek
η ф ψ
: Efficiency : Dimensionless flow coefficient : Dimensionless pressure coefficient
Subscripts 0 1 2 5 7
: Intake duct inlet : Impeller inlet : Impeller outlet (=imp.) : Diffuser outlet (=diff.) : Volute outlet (=vol.)
References [1] T. Fukano, Y. Takamatsu and Y. Kodama, The effect of the tip clearance on the noise of low pressure axial and mixed flow fans, Journal of Sound and Vibration, 105 (1986) 291308. [2] F. Kameier and W. Neise, Rotating blade flow instability as a source of noise in axial turbomachines, Journal of Sound and Vibration, 203 (5) (1997) 833-853. [3] J. März, C. Hah and W. Neise, An experimental and numerical investigation into the mechanisms of rotating instability, Journal of Turbomachinery, 124 (2002) 367-375. [4] L. Pinelli, F. Poli, M. Marconcini, A. Arnone, E. Spano and Da\. Torzo, Validation of a 3D linearized method for turbomachinery tone noise analysis, Proc.of ASME Turbo Expo, Vancouver, British Columbia, Canada (2011) GT201145886. [5] K. R. Fehse and W. Neise, Generation mechanisms of lowfrequency centrifugal fan noise, AIAA Journal, 37 (1999) 1173-1179. [6] S. Kaelladi, S. Kouidri, F. Bakir and R. Rey, Predicting tonal noise from a high rotational speed centrifugal fan, Journal of Sound and Vibration, 313 (2008) 113-133. [7] Y. Li, H. Quyang, J. Tian, Z. Zheng and Z. Du, Internal flow and noise investigations about the cross-flow fan with different blade angle, Proc. of ASME Turbo Expo, Glasgow, UK (2010) GT2010-23280. [8] M. Cadorin, M. Pinelli, E. Podeschi, F. Pompoli and A. Zanardi, Experimental characterization of the influence of auxiliary devices on the noise generated by industrial centrifugal fan and correlation to the geometrical and fluid dynamic parameters, Proc. of ASME Turbo Expo, Vancouver, British Columbia, Canada (2011) GT2011-46088.
3297
[9] A. Guedel, M. Robitu, N. Descharmes, D. Amor and J. Guillard, Prediction of the blade trailing-edge noise of an axial flow fan, Proc. of ASME Turbo Expo, Vancouver, British Columbia, Canada (2011) GT2011-45256. [10] S. Rulik, S. Dykas and W. Wroblewski, Modeling of aerodynamic noise using hybrid SAS and DES method, Proc. of ASME Turbo Expo, Glasgow, UK (2010) GT2010-22696. [11] H. J. Feld, S. Aschenbrenner and R. Girsberger, Investigation of acoustic phenomena at the inlet and the outlet of a centrifugal compressor for pressure ratio 4.5, Proc. of ASME Turbo Expo, New Orleans, Louisiana, USA (2001) GT-0314. [12] T. Raitor and W. Neise, Sound generation in centrifugal compressors, Journal of Sound and Vibration, 314 (2008) 738-756. [13] H. Tang, D. Qi and F. Lu, Identification of acoustic dipole sources in a propylene centrifugal compressor stage, Proc. of ASME Turbo Expo, Glasgow, UK (2010) GT2010-22408. [14] D. Sakaguchi, M. Ishida, H. Ueki and H. Hayami, Analysis of noise generated by low solidity cascade diffuser in a centrifugal blower, Proc. of ASME Turbo Expo, Berlin, Germany (2008) GT2008-50750. [15] R. Ballesteros-Tajadura, S. Velarde-Suarez, J. P. HurtadoCruz and C. Santolaria-Morros, Numerical calculation of pressure fluctuations in the volute of a centrifugal fan, Journal of Fluids Engineering, 128 (2006) 359-369. [16] H. S. Sun, H. G. Shin and S. G. Lee, Analysis and optimization of aerodynamic noise in a centrifugal compressor, Journal of Sound and Vibration, 289 (2006) 999-1018. [17] K. K. Ha and S. H. Kang, Optimization method for centrifugal compressor design using the surrogate management framework, Proc. of ASME-JSME-KSME Joint Fluids Engineering Conference, Hamamathu, Japan (2011) AJK201122036. [18] N. Kämmer and M. Rautenberg, A distinction between different types of stall in a centrifugal compressor stage, Journal of Engineering for Gas Turbines and Power, 108 (1) (1986) 83-92.
Kyoung Ku Ha is a Senior Research Engineer in Research & Development Division, Hyundai Motor Group since 2012. Dr. Ha received his B.S. from School of Mechanical Engineering, Seoul National University in 2000, and received his Ph.D. from School of Mechanical Engineering, Seoul National University in 2012.