Professional Documents
Culture Documents
Wang2014 PDF
Wang2014 PDF
GT2014
June 16 – 20, 2014, Düsseldorf, Germany
GT2014-26813
ABSTRACT NOMENCLATURE
The performance of transonic compressors can be charac-
terized aerodynamically and aeroacoustically. In this paper, Roman symbols
the DLR SRV2 compressor without vaned diffusers and its re- c speed of sound
designed version are studied. The redesign strategy (Zangeneh f integral surface
et al. 2011 [1]) utilized the 3D inverse design and CFD analysis. fn shaft rotational frequency
Both compressors were analyzed in ANSYS CFX 11, and the FFT fast Fourier transform
computational results show that the predicted pressure-ratio and H( f ) Heaviside function
efficiency of the original compressor have good agreement with g = τ − t + r/c
experimental results. The simulations have also revealed that the G Green’s function
redesigned one is superior at both design and off-design points at M0 convective Mach number
different rotating speeds. n normal vector
This work applies a convective FW-H method to further in- p0 acoustic pressure
vestigate the noise radiation from these two compressors. As Pi j compressible stress tensor
the blade tip speed is supersonic, the permeable integral surface p0T thickness noise
scheme must be adopted. The flow quantities needed as the in- p0L loading noise
puts to the FW-H solver were extracted from the CFD solutions. r source-observer distance, |xx − y|
The numerical predictions of the noise SPLs at blade passing S p permeable surface
frequency and its harmonics match the experimental measure- t observer time
ments reasonably well. It is found that the original compressor Ti j Lighthill stress tensor
has significant variations of SPLs as the operating mass flow rate u fluid velocity
changes whereas the redesigned one has much slighter variations. v solid surface velocity
At peak efficiency the redesigned compressor has a lower noise x observer position
level. y source position
This study provides insights for the optimal design of a tran-
sonic compressor when good aerodynamic and aeroacoustic per- Greekq symbols
formance are both required. β 1 − M02
δ ( f ) Dirac delta function
∗ Address all correspondence to this author.
sure, and
2 2 2 2 i
¯ 2 = 1 ∂ − ∂ + 2 U0 j ∂ + U0iU0 j ∂
h
, (2)
c2 ∂t 2 ∂ x2j c2 ∂t∂ x j c2 ∂ xi ∂ x j
et al. (2010) [10] by including the convective flow, and their Qi = ρ(ui +U0i − vi ) + ρ0 (vi −U0i ),
extended formulation of FW-H equation was employed in the Li j = ρui (u j +U0 j − v j ) + (p − p0 )δi j − σi j ,
present work. The convective FW-H equation is shown below.
Ti j = ρui u j − σi j + [(p − p0 ) − c2 (ρ − ρ0 )]δi j , (3)
¯ 2 [H( f )p0 ] = ∂ ∂ ∂
+U0 j [Qi ni δ ( f )] − [Li j n j δ ( f )] where p0 is the ambient pressure, u and v are the fluid and solid
∂t ∂xj ∂ xi surface velocity, respectively, σi j is the viscous stress tensor, δi j
∂2 is Kronecker delta tensor and Ti j is the well-known Lighthill
+ [H( f )Ti j ], (1)
∂ xi ∂ x j stress tensor. If there is no convective flow, i.e., U 0 = 0, then
Eq. (2) reduces to the usual FW-H equation. This equation can
be solved in a similar matter as Farassat’s formulations. A brief
in which f is integral surface function, δ ( f ) is the Dirac delta derivation will be given below and more details can be found in
function, H( f ) is the Heaviside function, p0 is the acoustic pres- Ref. [10]
r
g = τ −t + , (5) Z h Q̇ n + Q ṅ i
i i i i
c 4π p0T (x,t) = ∗ 2
dS p
f =0 r (1 − MR ) τ
Z h ∂ r∗ Qi ni i
and x is the observer position, y is the source position, r is the rel- + − dS p
ative distance between source and observer, t is signal reception f =0 ∂ τ r∗2 (1 − MR )2 τ
time and τ is the signal emission time. However, a little modifica- h∂M Qi ni
Z i
R
+ dS p
tion must be introduced here to take account of the presence of a f =0 ∂ τ r∗ (1 − MR )3 τ
subsonic mean flow. It is assumed without losing generality that Z h r̃ Q̇ n + r̃ Q ṅ + r̃˙ Q n i
3 i i 3 i i 3 i i
the mean flow is along the positive x3 -direction. A 3-dimensional −M0 dS p
f =0 r∗ (1 − MR )2 τ
free-space Green’s function for the convective equation given by h ∂ r∗ r̃3 Qi ni
Z i
Blokhintzev (1956) [11] is +M0 dS p
∗2 2
f =0 ∂ τ r (1 − MR ) τ
Z h∂M r̃3 Qi ni i
δ (τ − t + r/c) R
G(xx,t; y , τ) = , (6) −M0 dS p
4πr∗ f =0 ∂ τ r∗ (1 − MR )3 τ
Z h r̃∗ Q n i
3 i i
−U0 ∗2
dS p , (12)
where f =0 r (1 − MR ) τ
−M0 (x3 − y3 ) + r∗
r= , (7)
β2 h r̃ L̇ n + r̃ L ṅ + r̃˙ L n i
1
Z
i ij j i ij j i ij j
4π p0L (x,t) = ∗ 2
dS p
cf =0 r (1 − MR ) τ
1
Z h ∂ r∗ Li j n j r̃i i
− dS p
c f =0 ∂ τ r∗2 (1 − MR )2 τ
q
r∗ = β 2 [(x1 − y1 )2 + (x2 − y2 )2 ] + (x3 − y3 )2 , (8)
1 h∂M Li j n j r̃i i
Z
R
+ dS p
c f =0 ∂ τ r∗ (1 − MR )3 τ
1
Z h L n r̃∗ i
ij j i
+ dS p , (13)
U0 c f =0 r∗2 (1 − MR ) τ
M0 = 1 − , (9)
c
where [ ]τ indicates the integrand is evaluated at the retarded time,
MR = vi r̂i∗ /c, and
q
β= 1 − M02 . (10)
x1 − y1 x1 − y1 x3 − y3
r̃1 = , r̃2 = β 2 , r̃3 = β 2 . (14)
r∗ r∗ r∗
The solution to the convective FW-H equation (Eq. (1)) is then
found as, The dS p in Eqs. (12) and (13 ) is the elementary area on the inte-
gral data surface, thus the quadrupole source term is legitimately
t
Z
δ (g) 3 neglected. The algorithm of implementing this formulation will
Z
0 ∂ ∂
p (xx,t) = +U0 Qi ni δ ( f ) d y dτ
∂t ∂ x3 −∞ V 4πr∗ be discussed in the next subsection.
I(yy, τ)
Z h i
4π p0 (x,t) = dS pi , (15)
f =0 r∗α1 (1 − MR )α2 τ
where I(yy, τ) is the re-organized source strength function de- FIGURE 4. Schematics of the nosie testing rig
pending on source position and time, and α1 , α2 are integers
which depend on the type of source, such as, monopole, dipole.
1 8 0
It can be approximated as,
h I (y ,t − r∗ /c) i
N 1 6 0
i i
4π p0 (x,t) ≈ ∑ ∗α1
i
δ S pi , (16)
r
i=1 i (1 − M Ri ) 2 τ
α
1 4 0
S P L (d B )
The surface S p is divided into N panels and the integrand Ii is
evaluated at the centre of every panel, yi at the retarded times 1 2 0
in v e r s e
1 4 8
o r ig in a l haviours. Figure 7 compares the aerodynamic and aeroacoustic
1 4 6 performances of the original and redesigned compressor work-
1 4 4
ing from surging to choking conditions under full speed. The
2 .5 2 .6 2 .7 2 .8 2 .9 3 .0
SPLs are the tonal magnitudes at the BPF. It is interesting to note
M a s s flo w ( k g /s )
that the sound level of the redesigned compressor has slight vari-
FIGURE 7. Aerodynamic and aeroacoustic performance comparisons ations at the entire operating ranges, while the original compres-
between the original and redesigned compressor sor produces sound with significant changes as the mass flow
varies. More importantly, it can be seen that sound levels gener-
ated by the redesigned are lower than those by the original when
saw noise. The dominant source of these buzz-saw components the working mass flow rate is in range from 2.72 to 2.88 kg/s. In
is the rotor-alone pressure of shocks attached to the supersoni- fact, the computational work presented in Ref [1] confirmed that
cally rotating blade tips, and these shocks can propagate against the peak efficiency with the vaneless stage occurs around 2.73
the incoming flow to the upstream in the inlet duct. kg/s. Thus, the redesign compressor shows lower noise emission
By comparing Figs. 5 and 6 it can be seen that the tones at around its peak efficiency point. However, the baseline impeller
BPF and its harmonics are captured in prediction without large has lower level of noise at lower mass flow rates and close to the
discrepancies, but there are no buzz-saw components on the pre- stall point.
dicted spectrum. This is because the buzz-saw tones are due to In the following section, the CFD predictions of the flow
the amplitudes variations and nonlinear propagations of shocks physics in the blade channel at close to the peak efficiency point
near the blade tips, which would induce the energy redistribu- will be presented so as to explain the reason why the inversely
tion of the frequency spectrum, i.e., the initial energy at BPF redesigned compressor has a better aeroacoustic performance.
and its harmonics is redistributed at the harmonics of rotor shaft
frequency (engine orders). A more detailed explanation of buzz-
saw noise generation mechanism can be found in the work by 4 CFD analysis
McAlpine, Fisher and Tester (2006) [17]. In a similar fan noise The flow in the original SRV2 impeller was analyzed by us-
prediction study, Gliebe et al. (2000) [18] reported that steady ing commercial CFD CFX11 by Zangeneh et al. [1]. For this
0 .8
in v e r s e
0 .7
o r ig in a l
0 .6
2 .4 2 .5 2 .6 2 .7 2 .8 2 .9 3 .0
M a s s flo w ( k g /s )
6
( to ta l- to ta l)
5
in v e r s e
o r ig in a l
4
P R
3
2 .4 2 .5 2 .6 2 .7 2 .8 2 .9 3 .0
M a s s flo w ( k g /s )
FIGURE 10. Mach number contour comparison between the original
FIGURE 8. Comparison of the characteristics of the original and re- (LHS) and redesign (RHS) at 25% of span
design compressor
P re s s u re (k P a )
1 0 0
-1 0 0
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
S tr e a m w is e lo c a tio n
3 0 0
in v e r s e
2 0 0
o r ig in a l
P re s s u re (k P a )
1 0 0
-1 0 0
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
S tr e a m w is e lo c a tio n
FIGURE 12. Mach number contour comparison between the original FIGURE 14. Streamwise blade loading distributions on the main
(LHS) and redesign (RHS) at 75% of span blade (lower) and splitter (upper) at 15% of span
3 0 0
in v e r s e
2 0 0 o r ig in a l
P re s s u re (k P a )
1 0 0
-1 0 0
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
S tr e a m w is e lo c a tio n
3 0 0
in v e r s e
2 0 0 o r ig in a l
P re s s u re (k P a )
1 0 0
-1 0 0
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
S tr e a m w is e lo c a tio n
FIGURE 13. Mach number contour comparison between the original FIGURE 15. Streamwise blade loading distributions on the main
(LHS) and redesign (RHS) at 95% of span blade (lower) and splitter (upper) at 25% of span
that in the original design. The reduction of flow separation from 0.2 to 0.4 streamwise position.
increases the effective passage area, and resultantly weaken-
ing the shock strengths. Figure 19 shows that the hub-to-shroud Mach number con-
2. Shocks at the LE of the main blade of the redesigned com- tour details at the leading edges of two compressors. It confirms
pressor are significantly weaker than those of the original up that shocks near the hub of the redesign are weaker than the orig-
to 50% of the span (Figs. 9 - 11), and it can also be found inal one; however, this phenomenon did not continue all the way
in Figs. 14-16 that the original main blade sustains stronger to the shroud. The shocks caused by the redesign after 50% of
loading near the LE up to the mid of span. the span have larger strength than those of the original. This is
3. Shocks formed between the PS of main blade and SS of further validated by Fig. 20, which shows the efficiency varia-
splitter in the redesigned compressor are weaker those in the tions at the TE from hub to shroud. It can be clearly seen that
original one at all spanwise positions. This is also confirmed the inverse design has higher efficiencies up to 60% of the span.
by the blade loading distributions. It can be seen in Figs. 9 Figure 21 demonstrates the efficiency and static entropy varia-
- 13 that the pressure loading near the LE of the inverse de- tions of two compressors from the impeller inlet to outlet. It can
signed splitter is smaller than that of the original splitter, and be seen that the redesigned impeller generated less static entropy
this phenomenon is quite obviously shown in Figs. 16 to 18 than the original design, which consequently results in a higher
2 0 0 o r ig in a l
P re s s u re (k P a )
1 0 0
1 0 0
0
0
-1 0 0
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
-1 0 0
S tr e a m w is e lo c a tio n 0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
3 0 0
S tr e a m w is e lo c a tio n
in v e r s e 3 0 0
2 0 0 o r ig in a l in v e r s e
P re s s u re (k P a )
2 0 0 o r ig in a l
P re s s u re (k P a )
1 0 0
1 0 0
0
0
-1 0 0
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
-1 0 0
S tr e a m w is e lo c a tio n 0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
S tr e a m w is e lo c a tio n
FIGURE 16. Streamwise blade loading distributions on the main
blade (lower) and splitter (upper) at 50% of span FIGURE 18. Streamwise blade loading distributions on the main
blade (lower) and splitter (upper) at 95% of span
3 0 0
in v e r s e
2 0 0 o r ig in a l
P re s s u re (k P a )
1 0 0
-1 0 0
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
S tr e a m w is e lo c a tio n
3 0 0
in v e r s e
o r ig in a l
2 0 0
P re s s u re (k P a )
1 0 0
-1 0 0
FIGURE 19. Mach number contour comparison between the original
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
(LHS) and redesign (RHS) at LE
S tr e a m w is e lo c a tio n
0 .8
efficiency.
The CFD simulations show that the redesigned impeller gen-
S p a n n o r m a lis e d
0 .6
erates weaker shocks in the inducer and significantly reduces the in v e r s e
flow separation. The weaker shock structure noticeably reduces o r ig in a l
the generated noise. 0 .4
0 .2
5 Conclusion
In this paper, we examined the aerodynamic and aeroacous-
tic performance of a transonic centrifugal compressor and its re- 0 .0
0 .6 0 .7 0 .8 0 .9 1 .0
designed version. The employed 3D inverse design methodology Is e n tr o p ic c o m p r e s s io n e ffic ie n c y
has provided a better solution in terms of the aerodynamic and
partially aeroacoustic aspects. The aerodynamic performance is FIGURE 20. Comparison of spanwise variation of isentropic com-
achieved for the entire operating range. Noise radiations can also pression efficiency at the TE
be reduced when the redesigned compressor was operating near
0 .9 0
[8] Lighthill, M.J. 1952. On sound generated aerodynami-
in v e r s e
0 .8 5
o r ig in a l cally. I: general theory. Proc. R. Soc. London. A, 231,
0 .8 0 pp.564-587.
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
S tr e a m w is e lo c a tio n [9] Curle, N. 1955. The influence of solid boundaries upon
4 0
aerodynamic sound. Proc. R. Soc. London. A, 211, pp.505-
S ta tic e n tr o p y ( J k g ^ - 1 K ^ - 1 )
in v e r s e
2 0 o r ig in a l 514.
0
[10] Najafi-Yazdi, A., Bres, G.A., and Mongeau, L. 2010. An
acoustic analogy formulation for moving sources in uni-
-2 0
formly moving media. Philos Trans R Soc A, 467, pp.144-
-4 0 165.
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
S tr e a m w is e lo c a tio n
[11] Blokhinstev, D. 1956. Acoustics of a nonhomogeneous
moving medium. NACA TM-1399.
FIGURE 21. Comparison of isentropic compression efficiency in [12] Farassat, F. 1975. Theory of noise generation from moving
streamwise location from impeller inlet to outlet bodies with an application to helicopter rotors. tech. rep.,
NASA TR R-451, 1975.
[13] Farassat, F. 2007. Derivation of Formulations 1 and 1A of
peak efficiency. In the subsequent work, an optimization strategy Farassat. NASA TM-2007-214853, 2007
that couples inverse design method and the current aeroacoustic [14] Brentner, K.S. and Farassat, F. 2003. Modelling aerody-
model will be employed to redesign another transonic compres- namically generated sound of helicopter rotors. Progress
sor with vaned diffusers. in Aerospace Sciences, 36, pp.592-594.
[15] Wang, P. and Zangeneh, M. 2013. Numerical prediction of
aerodynamic sound generated in a supersonic centrifugal
ACKNOWLEDGMENT compressor. 19th AIAA conference, May 27 - 29, 2013,
We would thank Napier Turbochargers Ltd and EPSRC for Berlin, Germany.
their financial support of this project. [16] Casalino, D. 2003. An advanced time approach for acous-
tic analogy predictions. Journal of Sound and Vibration,
261, pp. 583-612.
REFERENCES [17] McAlpine, A., Fisher, M.J., and Tester, B.J. 2006. “Buzz-
[1] Zangeneh, M., Amarel, N., Daneshkhah, K., and Krain, saw” noise: A comparison of measurement with predic-
H. 2011 Optimization of 6.2:1 Pressure Ratio centrifugal tion. Journal of Sound Vibration, 290, pp.1202-1233.
compressor impeller. Proceedings of ASME Turbo Expo [18] Gliebe, P., Mani, R., Shin S., Mitechell, B., Ashford, G.,
2011: Power for Land, Sea and Air, June 6-11, 2011, Van- Salamah, S. and Connell, S. 2000. Aeroacoustic predic-
couver, Canada, GT-2011-46505. tion codes. NASA/CR-2000-210244.
[2] Krain, H. and Hoffman, B. 1998. Flow physics in high
pressure ratio centrifugal compressors. ASME-Summer
Meetings, FEDSM98-4853. 1998
[3] Eisenlohr, G., Krain, H., Richter, F. and Tiede, V. 2002. In-
vestigations of the flow through a high pressure ratio cen-
trifugal impeller. ASME Turbo Expo 2002, Amsterdam,
Netherlands, GT-2002-30394.
[4] Raitor, T. and Neise, W. 2008. Sound generation in cen-
trifugal compressor. Journal of Sound and Vibration, 314,
pp.491-509.
[5] Krain, H. and Hoffman, B. 2007. Improved High Pressure
Ratio Centrifugal Compressor. ASME Turbo Expo 2007,
Montreal, Canada, GT2007-27100.
[6] Krain, H. and Hoffman, B. 2008. Flow study of a
redesigned high-pressure-ratio centrifugal compressor.
Journal of Propulsion and Power, 24, pp.1117-1123.