Low Noise Design of Regenerative Blower by Combining the
FANDAS-Regen Code, Optimization Technique and
Phase-Shift Cancellation Concept
Chan Lee, Hyun Gwon Kil and Jun Kon Kim
Department of Mechanical Engineering, University of Suwon, Hwaseong, Korea
Keywords: Regenerative Blower, FANDAS-Regen Code, Optimization Algorithm, Low Noise.
Abstract: Low noise regenerative blower is designed by combining the FANDAS-Regen code and optimization
technique. The FANDAS-Regen code used in the present study defines blower design variables on rotating
impellers and fixed side channels, and then constructs the 3-D blower design geometry. Based on the
designed blower geometry, the FANDAS-Regen code also analyzes the blower performance as well as noise
characteristics by using momentum exchange theory coupled with pressure loss and leakage models and by
incorporating the performance prediction results into discrete and broadband noise models. With the
FANDAS-Regen code as a simulation engine, design optimization is conducted for impeller and side
channel design variables to minimize overall sound pressure level of blower under the constraints of
aerodynamic design requirements on pressure rise, efficiency and power consumption. Furthermore, for
more noise reduction of blower, a staggered impeller blade arrangement as a phase-shift cancellation design
concept is also applied to the optimized impeller design. The optimized blower model is manufactured and
tested by a chamber-type performance tester and narrow-band noise measurement apparatus. The
performance measurement results agree well with the FANDAS-Regen prediction, and the noise
measurement results show a remarkable noise reduction of 26 dBA through the present design optimization.
1 INTRODUCTION
Regenerative blowers are usually operated with high
pressure rise at low flow capacity and are widely
used as the gas supply equipment of fuel cell
automotive vehicles and distributed fuel cell power
systems (Badami and Mura, 2011; 2012). However,
because the most of these fuel cell systems are
located very close to human users, when
regenerative blower is used in the fuel cell
applications, its high noise characteristics would be
main shortcoming and hurdle to be applied in the
fuel cell applications. For this reason, there are
growing industrial needs of low noise regenerative
blower design.
Through the previous research by authors(Lee et
al., 2013), the FANDAS-Regen code as design-
analysis program for regenerative blower has been
developed and showed its good prediction
accuracies on performances and noise levels of
designed regenerative blower. In the FANDAS-
Regen code, 3-D blower geometry on impeller
blades and side channel is designed, and then blower
performances are predicted by the momentum
exchange theory between the rotating impeller
blades and the fixed side channel. After the predicted
performance results are obtained and then
incorporated with the noise models of the FANDAS-
Regen code, discrete frequency noise at blade
passing frequency and its harmonics are predicted by
acoustic mode analysis, and broadband noise is also
predicted by the combination of several correlation
models for inflow turbulence, impeller turbulence
and exhaust jet mixing. The performance and noise
prediction accuracies of the FANDAS-Regen code
are verified by comparing the prediction with the
measurement results of actual regenerative blowers.
Furthermore, with the use of the present analysis
method of the FANDAS-Regen code as simulation
engine, design optimization is conducted for two
impeller and side channel design variables to
minimize overall noise level of blower, and then a
phase-shift cancellation concept of staggered
impeller blade arrangement is also applied for more
noise reduction of optimized blower. The optimized
469
Lee C., Kil H. and Kim J..
Low Noise Design of Regenerative Blower by Combining the FANDAS-Regen Code, Optimization Technique and Phase-Shift Cancellation Concept.
DOI: 10.5220/0005514304690475
In Proceedings of the 5th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2015),
pages 469-475
ISBN: 978-989-758-120-5
Copyright
c
 2015 SCITEPRESS (Science and Technology Publications, Lda.)

blower by the present study is manufactured and
tested by using chamber-type test facility and
narrow-band noise measurement apparatus and its
measurement results are favorably compared with
the prediction by the FANDAS-Regen code. The
comparison results also show the overall noise level
of optimized blower is remarkably reduced by 26
dB(A) when compared with that of initial design.
2 DESIGN, PERFORMANCE AND
NOISE ANALYSIS METHODS
OF THE FANDAS-REGEN CODE
2.1 Blower Design Method
In general, regenerative blower is composed of
impellers equipped on double sides of rotating plate
and fixed side channel covering the impellers.
Blower’s design variables and geometry are shown
in Fig.1.
Figure 1: Geometry and design variables of regenerative
blower.
The main design variables of rotating impellers and fixed
side channel are defined as follows:
- Rotation speed( N )
- Tip diameter( D
2
=2r )
- Channel height( h )
- Channel width( W )
- Impeller blade inlet angle( β
1
)
- Impeller blade outlet angle( β
2
)
- No. of impeller blades( Z )
- Impeller blade thickness( d )
- Axial clearance( c )
- Extension angle( θ
c
)
Once the blower design variables are determined by
designer, 3-D blower shape and geometry are easily
obtained and then can be used for the input data of
performance analysis and CFD simulation.
2.2 Blower Performance Analysis
Models
In the FANDAS-Regen code( Lee et al., 2013), the
performance of blower is analysed by combining the
mean line analysis method for fluid flow and the
momentum exchange theory between impellers and
side channel. As shown in Fig. 2, the gas flow inside
regenerative blower is assumed to be typically three
dimensional and helical-toroidal motion where fluid
rotates in and passes along the space between
rotating impeller blades and fixed side channel. The
present study assumes mean streamline as the
representative one of the three dimensional fluid
flow phenomena.
(a) Flow behavior inside regenerative blower
(b) Cross section view on helical-toroidal flow
Figure 2: Flow pattern of regenerative blower.
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
470

Through the momentum exchange between fluid
and impeller due to this flow motion, gas pressure is
gradually raised along tangential flow path and its
overall pressure rise( ∆p
s
) is calculated by
2
21
12
2
1
2(/4)
1
2
sm uu
pf
cc
pQ CC
rr
KK
Aur r u u
u
φ
ρ

Δ
=−−+


(1)
where
22 2
2
22
1cot
ucm
c
CAQ
ru
ur u AAu
β


Δ
=− +




(2)
22
2
2
12 2
1.5 1.1(2 2 / )
[1 ( / ) ] 1.5 1.1(2 2 / )
u
uZrr
βπ
βπ
Δ+−
=
−++−
(3)
c
Q
uA
ϕ =
1
1
u
Cr
ur
ϕ =
(4)
2
22
2
1
21
11
2
12 12
2
11
cot cot
2sin
cm c cm
m
cc
AQ u AQ
Ar
K
AAAuruAAu
ββ
β

 
 


++ +−
 
 



    

(5)
Here ρ, u, Q, Q
m
and φ are fluid density, impeller
rotation speed, flow capacity, circulating flow rate
and flow coefficient. More detailed description and
variable definition about momentum-exchange
theory are referred to Badami et al( Badami and
Mura, 2012 ).
As represented in equation (1), proper models
for pressure losses and the leakage flows should be
constructed for accurate performance prediction, so
the FANDAS-Regen code uses the correlations of
Table 1, which are expressed as the functions of
blower design variables( Lee et al., 2013 ).
Table 1: Pressure loss and leakage model coefficients.
2.3 Blower Noise Modelling and
Analysis Method
Based on the blower performance prediction results,
the FANDAS-Regen code considers two kinds of
noise components. One is the discrete frequency
noise at blade passing frequency( BPF ) and another
is the broadband noise distributed over wide
frequency range, which are produced due to inflow
turbulence at blower inlet, turbulence within
impeller and side channel and turbulent jet at blower
exit.
Discrete frequency noise is produced due to the
pressure difference between adjacent impeller blades
rotating at BPF. In the present study, the pressure
difference is defined by ∆p
s
/Z
e
or 2∆p
s
/Z
e
and its
pressure fluctuation can be modelled as shown in
Fig. 3:
Figure 3: Theoretical model for the pressure fluctuation
between adjacent impeller blades.
where ∆p
s
is overall blower pressure rise calculated
by the performance prediction method of section 2.2,
and Z
e
, d and r represent effective number of
impeller blades, impeller thickness and impeller tip
diameter respectively, and θ is angular coordinate in
the direction of tangential flow. As shown in Fig. 3,
fluid pressure rise is achieved through 1
regenration/1 pitch of fluid at low flow capacity( ϕ <
ϕ
lim
) while it being achieved through the 1
regeneration/ 2 pitches at high flow capacity(ϕ >
ϕ
lim
). Here ϕ is the flow coefficient as non-
dimensional flow capacity parameter, and ϕ
lim
is
assumed as 0.75 from the experiment of Badami and
Mura( Badami and Mura, 2013 ).
Under the assumption of dipole type noise
radiation, classical acoustic theory (Wright, 1975)
on the rectangular-shaped pressure fluctuation of Fig.
3 gives the following root mean square value of
acoustic pressure( p
a
’ ) as:
0
2
1
2
1
2
1
2
2
2
2
2
1
=
















−








+
−
+
u
u
u
u
r
rr
c
c
LowNoiseDesignofRegenerativeBlowerbyCombiningtheFANDAS-RegenCode,OptimizationTechniqueand
Phase-ShiftCancellationConcept
471

at ϕ < ϕ
lim
at ϕ > ϕ
lim
(6)
where m= 1 means fundamental mode, m=2,3, …
mean its harmonic modes, and θc, θ and R are side
channel extension angle, noise measuring angle and
radius.
Broadband noise is produced from three main
noise sources of inflow turbulence, impeller
turbulence and exhaust turbulent jet. The present
study employs well-verified correlation model
corresponding to each noise sources (Mugridge,
1976; Goldstein, 1976), and their noise prediction
results are superimposed over frequency range. It is
noted that all the present broadband noise models
are expressed in terms of blower design variables
and performance parameters.
Figure 4: Performance and noise measurements.
2.4 Verification of the FANDAS-Regen
Code Prediction Method
The FANDAS-Regen code is applied to two existing
blower models for verifying its prediction accuracies
of performance and noise. The performance and
noise measurements on two Hwang-Hae blower
models (Hwang-Hae, 2012) are made by chamber-
type test facility and by the precision sound level
meter or the PULSE, a FFT analyser, as shown in
Fig. 4.
Figs. 5 and 6 show the performance and the
noise prediction results of Mini H-200
model( Hwang-Hae, 2012 )by the FANDAS code.
The performance prediction results are well-agreed
with the measurement over entire flow capacity
range, and the noise spectrum analysis result derived
from the performance prediction is also verified
through the good comparison between the prediction
and the measurement at blade passing frequency and
over wide frequency range. It is noted the noise
measurement of Mini H-200 at maximum flow
capacity is made from octave band analysis at 1.5 m
from blower inlet.
Fig. 7 and 8 represent the comparison between
the FANDAS-Regen code and the measurement
results of Mini H-100 model( Hwang-hae, 2012 ).
As shown in Fig. 7, the predicted performances by
the FANDAS-Regen code are well-agreed with the
measurement. Fig. 8 shows the predicted noise
spectrum at maximum flow capacity is reasonably
agreed with the 1/3 octave band measurement.
Figure 5: Aero-acoustic performance map of Mini H-200.
Figure 6: Noise spectrum of Mini H-200.
R
mZ
r
mZd
mZ
rp
Rp
c
s
a
)cos(
2
sin22),(
'
θ
θ
θ
















Δ
=
R
mZ
r
mZd
mZ
rp
Rp
c
s
a
)cos(
2
sin24),(
'
θ
θ
θ
















Δ
=
Flow capacity[LPM]
0 50 100 150 200 250 300 350
Efficiency[%], Power[kW] and OASPL[dB]
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
Static pressure[mmAq]
0
100
200
300
400
500
600
700
800
900
1000
1100
1200
1300
: Static pressure[mmAq]-Pred.
: Efficiency[%]-Pred.
: Power[kW]-Pred.
: OASPL[dBA]-Pred.
: Static pressure[mmAq]-Exp.
: OASPL[dBA]-Exp.
Frequency[Hz]
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000
Noise level
0
10
20
30
40
50
60
70
80
90
100
: SPL[dB]-Pred.
: SPL[dBA]-Pred.
: SPL[dBA]-Exp.
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
472

Figure 7: Aero-acoustic performance map of Mini H-100.
Figure 8: Noise spectrum of Mini H-100.
From the comparison results of Figs. 5-8, the
performance and noise prediction method of the
FANDAS-Regen code is expected to be used as
reliable simulation engine suitable for blower design
optimization.
3 BLOWER NOISE REDUCTION
BY DESIGN OPTIMIZATION
The present study conducts design optimization for
noise minimization of the blower with the following
design parameters:
- Rotation speed = 7400 rpm
- Side channel height( h ) = 20 mm
- No. of impellers( Z ) = 39
- Axial clearance( c ) = 0.2 mm
- Impeller blade thickness( d ) = 1 mm
- Tip diameter( D
2
) = 100 ∼ 200 mm
- Side channel width( W ) = 5 ∼ 15 mm
As mentioned above, tip diameter( D
2
) and side
channel width (W) are treated as the design variables
of optimization study while the other parameters
being as fixed values.
Therefore, the present optimization problem can
be formulated with the constraints for design
requirements as follows:
Find D
2
and W to minimize OASPL[ dBA ]
Subject to design flow capacity( Q
d
) = 350[ LPM ]
design pressure( ΔP
s
) ≥ 13 [ kPa ]
design-pt. efficiency( η
s
) ≥ 25 [%]
design-pt. power ≤ 500 W
(a) Static pressure
(b) Overall sound pressure level
Figure 9: Solution-finding histories.
The above optimization problem is solved by the
STQDAO algorithm of gradient-based sequential
approximate optimization technique in the PIAno
code (PIDOTECH, 2013) and its solution-finding
histories are shown in 9. After 10 iterations from
initial design condition, with satisfying pressure
constraint, optimum design is obtained as D
2
= 144
mm, W = 12 mm and results in the noise reduction
by 2.78 dBA compared with the initial design.
In addition to the optimization mentioned before,
for more noise reduction, the present study also
Flow capacity[LPM]
0 50 100 150 200 250 300 350
Efficiency[%], Power[kW] and OASPL[dB]
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
Static pressure[mmAq]
0
100
200
300
400
500
600
700
800
900
1000
1100
1200
1300
: Static pressure[mmAq]-Pred.
: Efficiency[%]-Pred.
: Power[kW]-Pred.
: OASPL[dBA]-Pred.
: Static pressure[mmAq]-Exp.
: OASPL[dBA]-Exp.
Frequency[Hz]
0 100020003000400050006000700080009000100001100012000
Noise level
0
10
20
30
40
50
60
70
80
90
: SPL[dB]-Pred.
: SPL[dBA]-Pred.
: SPL[dBA]-Exp.
LowNoiseDesignofRegenerativeBlowerbyCombiningtheFANDAS-RegenCode,OptimizationTechniqueand
Phase-ShiftCancellationConcept
473

employs a phase-shift cancellation concept on
impeller blade arrangement design. As known from
Fig. 3 of section 2.3, acoustic pressure is radiating
from each impeller blade in the form of sinusoidal
wave with the period of blade pitch. Thus, because
impeller blades are equipped and arranged along
angular direction on double sides of rotating plate, if
the impeller blades are arranged with staggered type
as shown in Fig. 10, the acoustic pressure radiating
from impeller on one side could be cancelled by that
from impeller on another side (Lee and Kil, 2014;
Kim et al., 2014 ).
Figure 10: Phase-shift cancellation concept.
Fig. 11 shows the pressure rise curves of the blower
model obtained through the applications of design
optimization and phase-shift cancellation concept.
The predicted pressure curves are well agreed with
the test ones at various RPM conditions, and the
optimized model satisfies its design requirement of
static pressure at design point.
Figure 11: Pressure rise curves of optimized blower.
Fig. 12 shows the comparison of the predicted noise
spectrum of initial blower model with the measured
one of the optimized blower model. As shown in Fig.
12, remarkable noise reductions of discrete
frequency noise components can be found at BPF
and its harmonics. These noise reduction effects
might be due to both the design optimization of 2.78
dBA and the phase-shift cancellation of 22.9 dBA.
The overall sound pressure level of the optimized
blower is significantly reduced by about 26 dBA
compared with the initial design case.
Figure 12: Noise spectrum of optimized blower.
4 CONCLUSIONS
The present study conducts noise reduction design of
regenerative blower by combining the FANDAS-
Regen code, optimization technique and phase-shift
cancellation concept. The FANDAS-Regen code, as
simulation engine of optimization problem, can
predict blower performances as well as noise levels
and its prediction results are well agreed with the
measurements. The optimization process by
STQDAO algorithm is carried out for low noise
blower design and then phase-shift cancellation
concept of staggered impeller arrangement is also
applied to the optimized blower. The present design
optimization gives the final design with 26 dBA
noise reduction compared with the initial design.
ACKNOWLEDGEMENTS
This work was supported by the Development of the
Regenerative Blower for fuel cell application of the
Korea Institute of Energy Technology Evaluation
and Planning (KETEP) grant funded by the Korea
government Ministry of Knowledge Economy.
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
474

REFERENCES
M. Badami, and M. Mura, Comparison between 3D and
1D Simulations of Regenerative Blower for Fuel Cell
Applications, 2012, Energy Conversion and
Management, 55, 93-100.
C. Lee, H.G. Kil, G.C. Kim, J.G. Kim, J.H. Ma, and K.H.
Chung, 2013, Aero-acoustic Performance Analysis
Method of Regenerative Blower, J. of Fluid
Machinery( in Korean ), 16(2).
M. Badami and M. Mura, 2011, Setup and Validation of a
Regenerative Compressor Model Applied to Different
Devices, Energy Conversion and Management, 52,
2157-2164.
S. E. Wright, 1976, The Acoustic Spectrum of Axial Flow
Machines, J. of Sound & Vibration 45(2), 165-223.
B. D. Mugridge, 1976, Noise Characteristics of Axial and
Centrifugal Fans as Used in Industry, Shock and
Vibration Digest, 45(3).
M. E. Goldstein, 1976, Aeroacoustics, McGraw-Hill.
Regenerative Blower Performance and Noise Test
Report, 2012, Hwang Hae Electric Co.
PIAnO(Process Integration, Automation and Optimization
user’s Manual, Ver.3.5, 2013, PIDOTECH, INC.
C. Lee and H.G. Kil, 2014, Impeller for reducing noise
and regenerative blower with the same impeller,
Korean Patent No.1014593790000.
J. K. Kim, K. Y. Lee, C. Lee
, H. G. Kil, K. H. Chung and
S. M. Hwang, 2014, Development of a low-noise
regenerative blower for fuel cell application, The
KSFM Journal of Fluid Machinery, 17(2), 48-53.
LowNoiseDesignofRegenerativeBlowerbyCombiningtheFANDAS-RegenCode,OptimizationTechniqueand
Phase-ShiftCancellationConcept
475