Proceedings of the ASME 2009 Fluids Engineering Division Summer Meeting

FEDSM2009
August 2-6, 2009, Vail, Colorado USA
Proceedings of FEDSM2009
ASME 2009 Fluids Engineering Division Summer Meeting
August 2-6, 2009, Vail, Colorado USA

FEDSM2009-78454
FEDSM2009-78454

DESIGN OF A PROPELLER FAN USING 3-D INVERSE DESIGN METHOD AND CFD
FOR HIGH EFFICIENCY AND LOW AERODYNAMIC NOISE

Hidenobu OKAMOTO Akira GOTO

Fluids Machinery & Systems Company Fluids Machinery & Systems Company
Ebara Corporation Ebara Corporation
Tokyo 144-8510, JAPAN Tokyo 144-8510, JAPAN
okamoto.hidenobu@ebara.com goto.akira@ebara.com

Masato FURUKAWA
Department of Mechanical Engineering
Kyushu University
Fukuoka 819-0395, JAPAN

ABSTRACT 2.6dB(A) reduction in aerodynamic noise. The flow

Three-Dimensional Inverse Design Method, where the 3- mechanisms leading to the higher efficiency and lower
D blade profile is designed for a specified blade loading aerodynamic noise were discussed based on experiments and
distribution, has been applied for designing a propeller fan the RANS steady flow simulations. Based on these
rotor with high efficiency and low noise. A variety of the investigations, design guidelines for the inverse design of
blade loading distributions (pressure jump across the blade), propeller fan rotors with higher efficiency and lower
vortex pattern (forced vortex, free vortex, and compound aerodynamic noise were proposed.
vortex) and the stacking conditions (sweep angles) were
specified and the corresponding 3-D blade configurations were INTRODUCTION
obtained. Among the 22 different designs, 14 propeller fan Propeller fans are widely used in a variety of equipments
rotors including the reproduced baseline fan were such as electric appliances (refrigerators, microwaves, etc.), air
manufactured by a rapid prototyping based on a selective laser conditioners, automobile radiators, and cooling devices for
sintering system (SLS) and tested. It was confirmed PCs/servers. Due to the strong need for environment-friendly
experimentally that the best design achieved about 5.7 points (low energy consumption, low noise) and lower manufacturing
improvement in the peak total-to-static efficiency and the costs of such equipments, a new design technology enabling
higher efficiency, smaller size, and lower noise are demanded.
However, this is not an easy task and inevitably requires detail
control of the three-dimensional (3-D) flow structures inside
propeller fans.
Recent advancements in Computational Fluids Dynamics
(CFD) have made it possible to envisage detail flow
mechanism leading to performance degradation and/or noise
generation. Regarding turbomachinery blade shape
optimization, a 3-D inverse design method has been applied
successfully to a variety of turbomachineries such as industrial
pumps and compressors, see Zangeneh et al. (1998), Goto and
Zangeneh (2002), Ashihara et al. (2004), and Bonaiuti and
Zanheneh (2004). In the inverse design method, the 3-D
Figure 1 Development process. blade profile is designed for specified aerodynamic parameters

1 Copyright © 2009 by ASME

(blade loading). Because of this, more systematical/logical distribution of rV" at the rotor exit, i.e. vortex pattern (or
optimization for the aerodynamic performances and noises can Euler work distribution). Another important design input is
be expected. In addition to these, thanks to the development stacking conditions, which specify the blade shape (and the
of rapid prototyping technologies such as Selective Laser sweep angle) at a specified location. For incompressible
Sintering (SLS), a number of models can be manufactured and potential flows, )rV" )m is directly connected to the blade
tested experimentally in a short period.
In the present paper, a new design approach based on 3-D loading (pressure difference across the blade p+ - p-) as follows.
2+ )rV"
inverse design method and CFD has been developed for p+ * p* = %W (1)
B mbl )m
propeller fans. The development process consists of three
sub-processes of “Inverse Design”, “Fine mesh CFD,” and The major advantage of using the inverse design method
“Experimental Validation,” see Fig.1. First, the 3-D inverse is the fact that the complex 3-D geometry of the blade with
design method and coarse mesh CFD were briefly described. controlled 3-D pressure fields can be designed very easily by
Secondly, important inverse design parameters were proposed manipulating )rV" )m . On the contrary, the major
and some representative rotors have been designed, drawback of this method is that the optimum )rV" )m
manufactured, and tested experimentally. Then, the detail distribution is not known beforehand for a specific design
investigation of the flow fields has been introduced based on target. For the optimization of performance characteristics
the detail CFD work performed for the same rotors by Jeong, (efficiency, pressure rise etc.), the effects of the specified
J., et al. (2007). Finally, the guidelines for designing
)rV" )m distribution can be evaluated numerically including
propeller fans with high efficiency and low noise have been
proposed. 3-D viscous flow effects with reasonable reliability. However,
we still need to rely on experiments for finding good loading
NOMENCLATURE distribution in terms of noise control. So, the major purpose
! : number of blade of this study is to find the design inputs for the inverse design
Hn: normalized helicity method suitable for high efficiency and low noise design
through numerical and experimental investigations.
m : non-dimensional meridional distance
The 3-D inverse design method adopted in this study is
rV" : averaged angular momentum
based on potential flow theory, so “Coarse mesh CFD” was
Wmbl : meridional velocity on the blade used to confirm the performance characteristics and the overall
#" : wrap angle difference between the hub and the tip flow feature of each design including viscous effects. For this
$ : flow coefficient purpose, CFX-TASCflow (ver. 2.11.0) was used while adopting
% : density standard k-, turbulence model. In order to save time for
& : Euler head coefficient specified in the inverse design mesh modeling and CFD calculations, parallel ducts were
&s : static pressure rise coefficient assumed both for upstream and downstream channels and only
' : total-to-static efficiency one blade-to-blade pitch was modeled, see Fig.2. The
( : total pressure loss coefficient circumferential periodic condition, zero swirl and uniform
static pressure at inlet, and flow rate at outlet were specified as
INVERSE DESIGN PROCESS the boundary conditions. Although the upstream and
A commercial design code “TURBOdesign-1” from downstream flow passage configurations were simplified, the
Advanced Design Technology Ltd. was used in this study, tip clearance of 1 mm between the rotor tip and the casing was
where the 3-D inverse design method proposed by Zangeneh modeled properly. The total number of the mesh grids was
(1991) was adopted. In this method, the main inputs are the 191,184.
distributions of the blade loading parameter )rV" )m along
INVERSE DESIGN OF A BASELINE FAN
the non-dimensional meridional distance m and the spanwise
To start with, the inverse design was applied to reproduce
commercially available propeller fan (called “baseline fan”).
Mass flow rate The leading edge outer diameter was 182mm with 5 blades,
Inflow the design revolution of 3200rpm, and the design flow
coefficient of $ = 0.280 (10m3/min). The meridional
configuration and the blade thickness distribution were set
identical to the baseline fan, and a linear stacking condition
Ps=const. was adopted at the trailing edge location simulating the sweep
angle of 27 degrees (between the tip and the hub) for the
Figure 2 Model and boundary conditions for Coarse baseline fan.
mesh CFD.

2 Copyright © 2009 by ASME

 The Coarse mesh CFD was applied for the baseline fan • Vortex pattern (spanwise distribution of the Euler work)
rotor to extract rV" and )rV" )m distributions. Then • Stacking condition at the trailing edge (sweep angle)
these distributions were smoothed out and used as the design The range of variations of each design parameter was
inputs for the 3-D inverse design. Thus designed fan was determined as follows by evaluating preliminary design results.
called “baseline_inverse,” and its ID number is M1T115S27. • Peak value location of )rV" )m ; m = 0.1 (baseline value),
The characteristic features of the baseline fan were the forced 0.2, and 0.3.
vortex swirl flow pattern (maximum loading at the tip) and the • LE loading (Leading edge value of )rV" )m ); baseline
fore-loading at all spanwise locations having the peak location
value + 0.5, ! 0.0, and – 0.5.
of the )rV" )m close to the leading edge (around 10%-chord
• Vortex pattern; forced vortex (maximu loading at tip =
location). The )rV" )m value at the leading edge was baseline), free vortex (uniform loading), compound vortex
finite, which means moderate flow incidence angle was (maximum loading at midspan). The vortex pattern was
applied at the design condition. Coarse mesh CFD confirmed expressed as the Euler work ratio between the tip value &tip
that both fan had identical performance characteristics. The and the average value &average.; &tip/&average = 1.15 (forced
baseline_inverse was used as a starting point for the vortex), 1.0 (free vortex), 0.92, 0.78, and 0.42 (compound
parametric survey of the design parameters. vortex).
• Stacking condition (Sweep angle); baseline value + 15, !
DESIGN PARAMETERS 0.0, and – 15 degrees.
For the systematic survey of the effects of the design In total, 22 different design cases were investigated in the
parameters, the following three parameters (Fig. 3) were study. Based on the Coarse mesh CFD results, and 11 of
selected as they would have large influence on the location,
strength and behavior of the tip leakage vortex, thus affects (8) (7) (2)

the aerodynamic performances and the noise level. All other (9)

design inputs were set equal to those for the baseline_inverse. (6)
(5) (4)
(3)
(1)
• Streamwise distribution of )rV" )m (location of the peak
blade loading and its amount at the leading edge) (6)

(1)Test fan, (2)Pressure tap (measurement of pressure rise), (3)Shaft&Bearing, (4)Torque meter,
(5)Motor, (6)Dissipative chamber, (7)Filter, (8)Nozzle (measurement of flow rate), (9)Booster fan

Figure 4 Schematic view of experimental apparatus for

fan performance and noise.
Table 1 Design conditions of fabricated models
Degree Location
Vortex type Sweep
ID Number of loading of peak
&tip / &midspan angle
at LE loading
L1T115S27 Low 10% Forced – 115% 27deg.
M1T115S27 Moderate 10% Forced – 115% 27deg.
H1T115S27 High 10% Forced – 115% 27deg.
M3T115S27 Moderate 30% Forced – 115% 27deg.
M1U100S27 Moderate 10% Free – 100% 27deg.
M1M092S27 Moderate 10% Comp. – 92% 27deg.
M3M092S27 Moderate 10% Comp. – 92% 27deg.
M1T115S12 Moderate 10% Forced – 115% 12deg.
M1T115S42 Moderate 10% Forced – 115% 42deg.
M4M046S27 Moderate 40% Comp. – 46% 27deg.
M4M092S27 Moderate 40% Comp. – 92% 27deg.
M2M092S27 Moderate 20% Comp. – 92% 27deg.
M3M078S27 Moderate 30% Comp. – 78% 27deg.
M3M092S12 Moderate 30% Comp. – 92% 12deg.
Figure 3 Inverse design parameters. Baseline_SLS Baseline fan reproduced by SLS

3 Copyright © 2009 by ASME

them were selected for experimental evaluations. After the • All of the fans design by the 3-D inverse design method
preliminary tests, three additional rotors were designed and showed higher efficiencies both at the BEP and DP than
tested. Table 1 summarizes the inverse designed fan rotors those for the baseline.
with their IDs and design parameters. • Most of the fans designed by the 3-D inverse design
method showed higher static pressure rise than that for the
EXPERIMENTAL RESULTS baseline.
Measurement Facilities and method • Efficiencies tend to be improved when the maximum blade
Figure 4 shows the equipments used for the loading parameter )rV" )m was moved to the
measurements. A model fan rotor was attached at the end of downstream locations.
the overhung shaft and placed in a mouth ring. There was no • Compound vortex type design cases, having maximum
stay vanes or struts around the fan rotor to eliminate the noise loading at midspan location, tend to have higher
generation due to the rotor-stator interactions. The efficiencies."
equipment was designed based on ISO 5801, but the cross
Noise Measurements
section of the chamber downstream from the rotor had larger
The results of the overall noise levels in SPL measured at
area ratio of 20.4 than the value of 16 specified in ISO 5801.
the best efficiency points (BEP) and the maximum flow rates
The static pressure rise was calculated using the wall static
were summarized in Fig. 6. The symbol " represents the
pressure of the chamber. The flow rate was measured using
values in dB(L), ! represents the values in dB(A) at BEP, !
nozzles with the nozzle diameter of 75mm. The shaft power
was measured with a torque meter. The equipment was !+$! !+' &
placed in an anechoic chamber with the background noise of !+#(
? 7:9
? 7 , 89
15.5dB(A) and 35 dB(C). The microphone was placed on the !+#% !+' !
centerline of the rotating shaft in front of the suction mouth !+#" A 7 :9
ring 600mm away from the faceplate. In order to attenuate !+") !+& &
the noises generated by the flow measurement nozzles and the
$,-

!+"& 6 $ 67 , 89

'
6 - 67 :9
booster fans, splitter duct silencers (dissipative chambers) were !+"# !+& !
placed upstream of the nozzle and the downstream of the !+!* @ 7, 89
booster fan. !+!' !+% &
The model fan rotors were manufactured by using !+!$ 6 ' 67 , 89
@ 7:9

Selective Laser Sintering (SLS). In order to confirm the !+!!

6 ' 67 :9
!+% !
effects of fabrication method (and possible deformation of the ! " # $ % & ' ( ) * "! "" "# "$ "% "& "'
3"3!*#0#(

3$3!*#0#(

3%3!%'0#(

3%3!*#0#(

3#3!*#0#(

3$3!()0#(

3$3!*#0"#
3"5"!!0#(
3"2""&0#(

3$2""&0#(

3"2""&0"#

3"2""&0%#
4"2""&0#(
1"2""&0#(
,-./010

blade due to the centrifugal force), the baseline fan was

reproduced by SLS (named baseline_SLS). Preliminary noise B 6C6, - ./D;=/6E- = :/.;<=6>-./
measurements showed large difference of 5dB between
baseline fan and reproduced baseline_SLS. To focus the Figure 5 Comparison of overall performances
discussion into the broadband noises, the noises less than
150Hz and those with blade passing (NZ) frequency (and non- '&
6

aerodynamic noise peak) were removed from the overall

'%
values, which reduced the difference to 2dB in maximum
'$ ,-./D;=/6E-=
across the flow range. After this post-processing,
SPL [dB(L)] , SPL [dB(A)]

F,G1I67,89
repeatability and similarity of noise measurements had been '#

confirmed. ! '"
F,G1I673-J
'!
Performance Measurements F,GHI67,89
6

F,GHI673-J
The results of the aerodynamic performance measured at &*

the best efficiency point (BEP) and the design point (DP) were &)

summarized in Fig. 5. The symbol ! represents $ at BEP, ! &(

represents &s at DP, " represents ' at BEP, and " represents &' 6KHLF,G1IM6N/.O6/EE;P;/=PQ666 6KHLF,G1IM6R-J;RSR6EDTU6V-O/

' at DP. The lines represent the performances for the

6KHLF,GHIM6N/.O6/EE;P;/=PQ666 6KHLF,GHIM6R-J;RSR6EDTU6V-O/
&&
baseline fan and the broken line represents the design flow ! " # $ % & ' ( ) * "! "" "# "$ "% "& "'
3"3!*#0#(

3$3!*#0#(

3%3!%'0#(

3%3!*#0#(

3#3!*#0#(

3$3!()0#(

3$3!*#0"#
3"5"!!0#(
3"2""&0#(

3$2""&0#(

3"2""&0"#

3"2""&0%#
4"2""&0#(
1"2""&0#(
,-./010

rate. The measured data can be summarized as follows.

• The flow rates at the BEP were smaller than the design
Design Case
flow rate for most of the cases.
Figure 6 Comparison of overall noise

4 Copyright © 2009 by ASME

and " represent values at the maximum flows in dB(L) and FINE MESH CFD RESULTS
dB(A), respectively. The lines represent the noise levels for
Numerical Scheme and Computational Mesh System
the baseline. The results can be summarized as follows.
As for the “Fine mesh CFD”, the compressible 3-D RANS
• The baseline_inverse (M1T115S27) showed similar noise
equations were solved by an implicit unfactored upwind
with baseline_SLS, namely the noise level at the BEP was
relaxation scheme using the k-. two-equation turbulence
large while that at the maximum flow was low.
model. See Furukawa et al. (1992) for the detail. Different
• The difference between the baseline and the baseline_SLS
from the coarse mesh CFD, the actual fan configurations for
is large for dB(L) compared to that for dB(A).
experimental tests were modeled for Fine mesh CFD with H-
• At the BEP, the design cases of M1M092S27 and
type grid by combining five computational domains. The
M3M092S37 showed lower noise level by 2 to 3 dB. total number of mesh is about 980k cells. Note here, the
• At the maximum flow rate, the noise reduction by the detail internal flow measurements had been made as a part of
inverse design was minor. the project using hot-wire velocimeter and dynamic pressure
sensors (not shown in this paper), and reasonable agreements
between the measurements and the Fine mesh CFD had been
confirmed. "
Computational Results and Discussions
Based on the performance and noise measurements, the
0.35
6,-./D;=/
0.7
following 6 models were selected and the Fine mesh CFD was
0.30
63"2""&0#(
63"3!*#0#( 0.6 performed.
• baseline
63$3!*#0#(
61"2""&0#(
63"2""&0"#
• Model 1 - M1T115S27: baseline_inverse
0.25 0.5

0.20 0.4 • Model 2 - M1M092S27 : low noise but with low efficiency
• Model 3 - M3M092S27 : best efficiency and lowest noise
-

'

0.15 0.3
• Model 4 - L1T115S27 : high noise
0.10 0.2 • Model 5 - M1T115S12 : low noise
6,-./D;=/

0.05
63"2""&0#(
63"3!*#0#( 0.1
Jeong J. et al. (2007) reported the detail flow structure for
63$3!*#0#(
61"2""&0#( these 6 models. In this report, Model 1 and Model 3 were
63"2""&0"#
0.00
0.15 0.20 0.25 0.30 0.35
0.0
0.40
selected and their characteristic feature of the flow fields was
$ discussed at the design flow ($ =0.280). Figure 7 presents the
experimentally measured performance characteristics for these
Figure 7 Measured performance characteristics 6 models. Model 3 shows the best peak efficiency among the
models. The lightly loaded leading edge (Model 4) resulted
in poor performance characteristics at large flow range.
Vortex structure. To understand the complex vortex
structure, vortex identification method based on the critical-
point theory was used. The vortex structure was clarified by
showing color contours of the normalized helicity Hn on
identified vortex. Figure 8 shows the vortex structure
together with the limiting streamlines on the blade and the hub

Flow

Rotation

Rotation

(a) Model 1 (b) Model 3 (a) Model 1 (b) Model 3

Figure 8 Vortical flow structures Figure 9 Vortical flow structures near suction surface

5 Copyright © 2009 by ASME

surfaces. Even at the design point, the tip leakage vortex • When the leading edge loading is low (Model 4), the vortex
from the blade leading edge extended towards the adjacent breakdown is rather weak and the tip leakage vortex
blade and interacted with the pressure surface of the tip doesn’t diffuse quickly. This is good for suppressing high
leading edge of the adjacent blade for both cases. Especially loss generation. However, a strong interaction may occur
in the case of Model 1, the tip leakage vortex interacted between the tip leakage vortex and the adjacent blade
directly with the leading edge of the adjacent blade and pressure surface, which will result in high noise generation.
slightly overleaped the leading edge. The normalized helicity • Moderately loaded leading edge design (Model 1, 2, 3, and
contours showed abrupt change from the value 1 (red color) to 5) inducing strong tip leakage vortex breakdown makes the
–1 (blue color) just after the formation of the tip leakage vortex vortex core diffuse quickly, which results in low noise
near the leading edge. This suggests the existence of flow generation due to the weak vortex/blade interactions.
stagnation at the location of zero helicity and the following • Design parameters need to be tuned to avoid direct
vortex breakdown (Delery, 1994). It was confirmed that the interaction between the tip leakage vortex and the leading
tip leakage vortices had broken down for both design cases of edge of the adjacent blade. Especially when the vortex
Model 1 and Model 3. overleaps the leading edge and creates a sort of inlet gust
Limiting streamlines. In the case of Model 1 (Fig. (flow fluctuation), the noise level will be high (Model 1)
9(a)), the limiting streamlines around the tip leading edge despite of the diffusion of the vortex breakdown.
region shows strong secondary flows in the radial direction due • When the compound vortex design is adopted (Model 2 and
to the effects of the tip leakage vortex formation. The 3), the overall loading for the tip blade is reduced. The
attachment line can be clearly observed at the tip leading edge noise generation will be rather low if the tip leakage vortex
region, and the very strong effects of the tip leakage vortex are breakdown is well controlled to avoid the direct interaction
observed here. This was caused by the formation of very between the tip leakage vortex and the leading edge of the
strong tip leakage vortex due to the large blade loading of adjacent blade. The efficiency will increase due to the
Model 1 in this area. In the case of Model 3 (Fig. 9(b)), no smooth flow pattern on the blade suction surface.
strong radial secondary flows are observed in the tip leading • The design parameters need to be tuned to suppress the
edge area and the flow pattern on the blade suction surface was accumulation of high loss fluid in the casing region for
smooth and good compared to other design cases. high efficiency. The flow incidence need to be reasonably
Total pressure loss coefficient. On Fig. 10, total matched with the blade leading edge to avoid the formation
pressure loss contours are presented on four cross sectional of leading edge separation vortex merging with the tip
planes approximately normal to the tip leakage vortex. The leakage vortex, which results in very low fan efficiency.
vortex structure with color contours of the normalized helicity
is also presented in Fig.10. Both design cases of Model 1 and DESIGN GUIDELINE
Model 3 show high loss region covering the wide area along Figure 11 summarizes the design conditions for each fan
the casing. This high loss region was generated by the rotor cases and shows the cases with high efficiency and low
formation of the tip leakage vortices in the circumferential noise. The fan noises were low in the cases of compound
direction and their breakdown just after the vortex formation. vortex type (&tip /&midspan = 0.92), while the efficiencies were
Effects of design parameters. The relationship
between the design parameters and the flow structure (and & tip / & midspan
1.15 1.00 0.92 0.78 0.46
the resulting performance/noise characteristics) is M1T115S27 M1M092S27
0.1 M1U100S27
summarized as follows based on the CFD investigation on Blade loading
(Model 1) (Model 2)
0.2 M2M092S27
all six models. peak location M3M092S27 High
m 0.3 M3T115S27
(Model 3)
M3M078S27
Efficiency
0.4 M4M092S27 M4M046S27
Stacking = 27 deg. for each case

m = 0.1 for each case Low Noise

M1T115S12 Low
Stacking 12 (Model 5) Noise Similar
angle M1T115S27
27 (Model 1) Efficiency
(degree)
42 M1T115S42 Moderately loaded leading edge

Stacking = 27 deg. for each case

Rotation L1T115S27
Low (Model 4)
Flow Leading edge M1T115S27
Mod. (Model 1) Low Noise
loading
and High Efficiency
High H1T115S27

(a) Model 1 (b) Model 3

Figure 11 Summary of design parameters and their effects
Figure 10 Total pressure loss contours. on efficiency and noise.

6 Copyright © 2009 by ASME

high in the cases of the peak loading location around m = 0.3 vortex are very weak and the flow patterns on the blade
(30%-chord location). The stacking condition did not show suction surface is very smooth, which results in high
clear tendency for efficiency, which suggests complex efficiency.
interaction with other design parameters. The following
design guidelines are proposed for designing propeller fan by ACKNOWLEDGMENTS
inverse design approach with similar design specifications. The present work had been conducted as a part of a multi-
• Leading edge loading: to be moderately loaded. Excessive sponsor consortium project in Japan between 2001 and 2004 in
loading forms the leading edge separation, while very weak collaboration with four universities and 16 sponsoring
loading may create high noise due to the strong interaction companies. The authors would like to express their
between the tip leakage vortex and the adjacent blade. appreciation to Prof. Y. Tsujimoto (Osaka University), Prof. C.
• Peak location of the blade loading parameter )rV" )m : to Kato (University of Tokyo), and Prof. K. Kaneko (Saga
be set as follows for high efficiency; University) for their contributions to the project and the fruitful
0.4 >= m >= 0.2 discussions throughout the projects.
• Vortex type: maximum loading at the midspan location to
REFERENCES
reduce the blade loading at the tip for low noise;
Zangeneh, M., Goto, A., and Harada, H., 1998, “On the
0.92 >= &tip / &midspan >= 0.46
Design Criteria for Suppression of Secondary Flows in
• Stacking conditions: stacked at the rotor trailing edge and
Centrifugal and Mixed-Flow Impellers,” ASME Journal of
the amount of the stacking (#", sweep angle) to be set at
Turbomachinery, Vol.120, pp723-735
moderate values as follows for low noise;
Goto, A. and Zangeneh, M., 2002, “Hydrodynamic
27 degrees > #" >= 12 degrees Design of Pump Diffuser Using Inverse Design Method and
CFD,” ASME Journal of Fluids Engineering, Vol. 124, pp.
CONCLUDING REMARKS
319-328
To establish a new design technology for propeller fans Ashihara, K. et.al., 2004, “Optimization of Microturbine
based on 3-D inverse design method for low noise and high Aerodynamics using CFD, Inverse Design and FEM Structural
efficiency, parametric design of fan rotors, steady flow CFD,
analysis (1st report: Compressor Aerodynamics)”, ASME
and model tests were performed. The systematic survey of Paper GT2003-53431.
the effects of the aerodynamic design parameters of the 3-D Bonaiuti, D. and Zangeneh, M., 2006, “On the Coupling
inverse design was made and clarified through experiments as
of Inverse Design and Optimization Techniques for
well as fine mesh RANS based steady flow CFD. The major Turbomachinery Blade Design,” ASME Turbo Expo
conclusions are as follows. (Barcelona), ASME Paper GT2006-90897
(1) The control of the tip leakage vortex behavior is the key to
Jeong, J., et al., 2007, “On the Relationship between Vortical
optimize the propeller fan efficiency and noise Flow Behavior and Aerodynamic noise in Propeller Fans,” The 9th
characteristics. Asian International Conference on Fluid Machinery (Jeju,
(2) Among the inverse design parameters, the peak location of
Korea), AICFM9-086
the blade loading parameter )rV" )m influenced the Zangeneh, M., 1991, “A Compressible Three
efficiency, while the effects of the vortex pattern (spanwise Dimensional Blade Design Method for Radial and Mixed Flow
Euler work distribution) and the stacking condition (sweep Turbomachinery Blades,” Int. J. Numerical Methods in Fluids,
angle) had large effects on the noise. Vol. 13, pp. 599-624
(3) The leading edge loading needs to be controlled properly Furukawa, M., Nakano, T., and Inoue, M., 1992,
as the excessive loading (mismatching) could lead to “Unsteady Navier-Stokes Simulation of Transonic Cascade
leading edge separation, which would merge with the tip Flow Using an Unfactored Implicit Upwind Relaxation
leakage vortex and generate large losses. Scheme With Inner Iterations,” ASME Journal of
(4) Moderate leading edge loading is effective to make early Turbomachinery, Vol. 114, No.3, pp. 599-606.
breakdown of the tip leakage vortices, which results in the Delery, J. M., 1994, “Aspects of Vortex Breakdown”,
strong diffusion of the vortex core and contribute to Progress in Aerospace Sciences, Vol. 30, No. 1, pp. 1-59
lowering noise due to its weak interaction with the
adjacent blade.
(5) When the tip leakage vortex overleaps the leading edge of
the adjacent blade, despite the diffusion of the vortex core,
the noise level increases due to the flow fluctuation
(similar to inlet gust).
(6) When the overall tip loading is reduced (midspan
loading), the secondary flows induced by the tip leakage

7 Copyright © 2009 by ASME