You are on page 1of 14

Proceedings of ASME Turbo Expo 2017: Turbomachinery Technical Conference and

Exposition
GT2017
June 26–30, 2017, Charlotte, NC, USA

GT2017-63268

EFFECT OF FREE-STREAM VELOCITY DEFINITION ON BOUNDARY LAYER


THICKNESS AND LOSSES IN CENTRIFUGAL COMPRESSORS

Jonna Tiainen∗, Ahti Jaatinen-Värri, Aki Grönman, Teemu Turunen-Saaresti, Jari Backman
Laboratory of Fluid Dynamics
LUT School of Energy Systems
Lappeenranta University of Technology
53851 Lappeenranta, Finland
Email: jonna.tiainen@lut.fi

ABSTRACT Cd dissipation coefficient [-]


The estimation of boundary layer losses requires the ac- cs blade surface length [m]
curate specification of the free-stream velocity, which is not p blade pitch [m]
straightforward in centrifugal compressor blade passages. This t trailing-edge thickness [m]
challenge stems from the jet-wake flow structure, where the free- t∗ dimensionless trailing-edge thickness [-]
stream velocity between the blades cannot be clearly specified. U2 tip speed [m/s]
In addition, the relative velocity decreases due to adverse pres- Uδ velocity at the boundary layer edge [m/s]
sure gradient. Therefore, the common assumption of a single U∞ free-stream velocity [m/s]
free-stream velocity over the blade surface might not be valid in w relative velocity [m/s]
centrifugal compressors. Greek alphabet
Generally in turbomachinery, the losses in the blade cas- β relative flow angle [◦ ]
cade boundary layers are estimated e.g. with different loss co- δ boundary layer thickness [m]
efficients, but they often rely on the assumption of a uniform δ∗ dimensionless displacement thickness [-]
flow field between the blades. To give guidelines for the esti- Ψ∗ dimensionless energy thickness [-]
mation of the mentioned losses in highly distorted centrifugal ρ density [kg/m3 ]
compressor flow fields, this paper discusses the difficulties in the ζ loss coefficient [-]
calculation of the boundary layer thickness in the compressor Abbreviations
blade passages, compares different free-stream velocity defini- BL boundary layer
tions, and demonstrates their effect on estimated boundary layer DES design
losses. Additionally, a hybrid method is proposed to overcome FB full blade
the challenges of defining a boundary layer in centrifugal com- LE leading edge
pressors. NC near choke
NS near stall
PE peak efficiency
NOMENCLATURE PS pressure side
Latin alphabet SB splitter blade
SS suction side

∗ Address all correspondence to this author.


1 Copyright
c 2017 by ASME
In incompressible flow cases such as low speed turbine cascades,
the inlet stagnation pressure can be combined with blade surface
pressure to calculate local free-stream velocities, as was done
by Harrison [6]. This method, however, is not well transferable
into centrifugal compressors due to compressibility effects and
a complex flow field even at the mid-span. In contrast, axial
compressors with sufficiently high aspect ratios can have a more
symmetrical total pressure distribution [7], which makes the def-
inition of the free-stream velocity more straightforward. Alterna-
FIGURE 1. SCHEMATIC PRESENTATION OF THE PITCHWISE tively, in the one-dimensional centrifugal compressor simulation
IDEALIZED SYMMETRICAL AND TYPICAL CENTRIFUGAL by Klausner and Gampe [8], the free-stream velocity required to
COMPRESSOR VELOCITY PROFILES. calculate the boundary layer blockage at the blade throat was de-
fined as an average of the blade inlet relative velocity and critical
speed at the throat.
TE trailing edge
Choi et al. [9] concluded that the free-stream velocity cannot
Subscripts
be defined in blade passages, and avoided the problem in an axial
1 impeller inlet
compressor by defining the boundary layer as a location from the
2 impeller outlet
blade surface where the velocity gradient between two adjacent
data points is less than one percent. Lynch and Thole [10] stud-
ied axial turbine endwalls with a cascade test stand. Their stud-
INTRODUCTION ies included a boundary layer analysis within the blade cascade.
The ever increasing demand to reduce green house gas emis- Lynch and Thole accounted for the nonlinear velocity profile in
sions and improve energy efficiency require more efficient en- the cascade by measuring the velocity profile in the spanwise
ergy conversion. Even though turbomachinery has existed for direction and then using the exact measured velocity at a cor-
almost a century, the design is more or less based on empirical responding spanwise location as a free-stream velocity for the
data which has been gained through measurements over decades. boundary layers on the endwalls. This works in blade cascades
To push the efficiency higher and higher, new insights about the with a uniform flow field in the spanwise direction [11], and to
different, highly complex flow phenomena in turbomachinery are some extent, this might work if the boundary layers on the hub
required. or shroud walls in centrifugal compressor blade passages were
The boundary layer is one of the major causes of losses in studied. However, this type of method does not work when one
turbomachinery. For example Boyce [1] attributed 5 percentage wants to study boundary layers on the blade surfaces due to a
points of the efficiency loss to skin friction. In low Reynolds jet-wake flow structure.
number compressors, the relative importance of boundary layer In the literature, conclusions about the boundary layer thick-
losses becomes even greater than it is in high Reynolds num- nesses in radial turbomachinery have also been drawn from CFD
ber machines [2, 3]. Typically, the losses in the blade cascade results even though the method for calculating the boundary layer
boundary layers are estimated with different loss coefficients thickness remains unclear to the reader [12–14]. This is also
(e.g. [4, 5]). However, Denton [4] suggests that an understanding one of the motivations for the current study since without be-
of the origins of loss is more valuable than a quantitative predic- ing able to reliably define the edge of a boundary layer on the
tion, whereas Prust’s [5] loss coefficients rely on an assumption blade surface, the post-processing of modeling results and com-
of a uniform flow field between the blades. The specification parison between different loss sources becomes user dependent.
of the free-stream velocity is not as straightforward in the cen- This dependency generates challenges in the reliable comparison
trifugal compressor blade passages as in cases of flat plates or of different studies.
stationary blade cascades. This complexity comes from the spec- Since literature provides no unified definition of the free-
ification of the free-stream velocity, which is difficult to specify stream velocity in the blade passage of a centrifugal compressor,
in the centrifugal compressor blade passages where the jet-wake this study attempts to shed light on this question. This article
flow structure results in a highly non-uniform velocity field, as compares different free-stream velocity definitions and demon-
illustrated in Fig. 1. strates how they affect the evaluation of boundary layer losses
Although boundary layers have been studied in the past var- by using two types of loss coefficients [4, 5] and studying the
ious times, there are still several discrepancies in the definition specific entropy in the boundary layers.
of boundary layers, which make it challenging to compare the To provide insights into different machines, the free-stream
studies and to conclude what would be the most suitable means velocity definitions are numerically investigated in the blade pas-
to define boundary layers, especially in centrifugal compressors. sages of two centrifugal compressors; one with splitter blades

2 Copyright
c 2017 by ASME
and the other without. The compressor with splitter blades is ex- LOSS COEFFICIENTS
perimentally and numerically investigated at Lappeenranta Uni- To demonstrate the effect of four free-stream velocity defini-
versity of Technology [15] and the compressor without splitter tions on the boundary layer losses, two kinds of loss coefficients
blades is the Radiver in RWTH Aachen [16]. The studied com- are used. The boundary layer entropy loss coefficient of Den-
pressor geometries, operating points, and numerical methods are ton [4] is defined as
described in detail in the previous paper by the authors [3].
Z 1  3  
cs w x
ζBL = 2 ∑ Cd d . (1)
FREE-STREAM VELOCITY DEFINITIONS p cos β1 0 w1 cs
The boundary layer is usually considered as the flow area
close to solid walls where the velocity is less than 99.5% of the Equation (1) integrates the entropy increase at one spanwise
free-stream velocity. The definition is simple and quite easy to location over the blade surface in the meridional direction and
use with simple flows. However, in complex flows, the boundary sums the loss generation on both blade surfaces. The dissipation
layer investigations are much more challenging. In the follow- coefficient Cd is 0.002 as suggested by Denton [4], the relative
ing, four different methods for defining the free-stream velocity flow angle β1 is defined from the axial direction, and the relative
in the centrifugal compressor blade passages are presented and velocity w is defined at the edge of the boundary layer.
compared, taking into account that the velocity on the suction The boundary layer kinetic energy loss coefficient of Prust
and pressure sides should be different. [5] is defined as
1. The boundary layer is defined as 99.5% of the free-stream
velocity, and the free-stream velocity is defined as an area- Ψ∗TE
ζBL = ∗ − t∗ , (2)
averaged absolute velocity at the impeller inlet: Uδ = 1 − δTE
0.995U∞,inlet .
2. The boundary layer thickness is defined as the distance
between the blade surface and the location where the ve- where the dimensionless energy, displacement, and trailing-edge
locity is 99.5% of the velocity of the adjacent data point: thicknesses are
dU/dn = 0.005 ⇒ Un−1 = 0.995Un , where n is the normal to
the blade surface. Hence, the free-stream velocity U∞ = Un
"  #
w 2
Z 1    
∗ ρw y
equals the velocity of the data point next to the boundary Ψ = 1− d (3)
0 U∞ ρ∞U∞ p
layer edge and varies along the meridional location.
Z 1   
3. The free-stream velocity in different meridional locations ρw y
δ ∗ = 1 − t∗ − d (4)
from the blade leading edge to the trailing edge is estimated 0 ρ∞U∞ p
based on the second method. The boundary layer thickness t
is defined as 99.5% of the free-stream velocity and the free- t∗ = . (5)
p cos βTE
stream velocity is defined as an average of the free-stream
velocities along the meridional direction Uδ = 0.995U∞,ave ,
where U∞,ave = N1 ∑N1 Un and Un is evaluated based on the Equation (2) integrates the kinetic energy loss at one meridional
second method. location over the blade passage in the pitchwise direction. The
4. The boundary layer is defined as 99.5% of the free-stream density ratio ρ/ρ∞ ≈ 1 is assumed here, and according to the nu-
velocity, and the free-stream velocity is defined as the rel- merical results, the assumption seems to be valid. The relative
ative velocity in the middle of the blade passage Uδ = flow angle βTE is defined from the radial direction. In Prust’s
0.995U∞,middle which corresponds to the symmetrical flow loss coefficient, the flow in the blade passage is assumed to be
approximation. symmetrical. Therefore, at the trailing edge kinetic energy loss
occurs only in the boundary layers, and free-stream flow is as-
Relative velocity is used to estimate the boundary layers, and sumed isentropic.
the above-mentioned methods are referred to in the rest of the The effect of different free-stream velocity definitions on
paper as the first, second, third, and fourth method, respectively. boundary layer losses is also demonstrated by examining the
The first method is commonly used in flat plates and the second change of the specific entropy from the impeller inlet to the
method is similar to what Choi et al. used in an axial compressor boundary layers at the trailing edge. Both the loss coefficients
study [9]. Different free-stream velocity is used for full and split- and the specific entropy are investigated at the impeller trailing
ter blades and the values differ also between the blade pressure edge at the mid-span, where the influence of endwall boundary
and suction sides. layers and tip leakage flow on blade boundary layers is minimal.

3 Copyright
c 2017 by ASME
0.8 TABLE 1. TECHNICAL DATA FOR DESIGN POINT OF COM-
Compressor with splitter blades
Pressure coefficient ψ [-]
Compressor without splitter blades PRESSOR WITH SPLITTER BLADES AND PEAK EFFICIENCY
NS POINT OF COMPRESSOR WITHOUT SPLITTER BLADES.
0.6 DES
n/nDES = 1.0
n/nDES = 0.8
NC
With Without
NS splitter splitter
0.4 PE
blades blades
NC
Number of blades 7+7 15
0.2
0.00 0.02 0.04 0.06 0.08 0.10 Relative blade height (b2 /D2 ) 0.058 0.041
Flow coefficient φ [-]
Relative tip clearance (t/b2 ) 0.052 0.045
Flow coefficient (φ = UqD v
2) 0.065 0.051
FIGURE 2. DIMENSIONLESS COMPRESSOR MAP FOR BOTH 2 2

COMPRESSORS. Pressure coefficient (ψ = ∆hU22


s
) 0.520 0.450

ω qv
Specific speed (Ns = ∆h0.75 ) 0.830 0.830
s
U2
NUMERICAL METHODS Tip speed Mach number (MaU = a1 ) 0.920 1.170
Two centrifugal compressors presented in the previous paper
by the authors [3] are studied numerically. The main difference
between the compressors is that one has splitter blades and the RESULTS
other one only full blades. For the readers’ convenience, details To examine the overall influence of different free-stream ve-
of compressor geometries and important dimensionless numbers locities on the boundary layer edge location, the normalized rela-
at the design/peak efficiency point are shown in Table 1. The tive velocity distribution in the pitchwise direction at the trailing
compressor with splitter blades is modeled at the design operat- edge at the mid-span is shown in Figs. 3 and 4 for the com-
ing point (DES, n/nDES = 1.0, qm /qDES = 1.0) [17]. The com- pressors with and without splitter blades at their design/peak ef-
pressor without splitter blades is modeled at its peak efficiency ficiency points, respectively. Both figures show the wake (low
point (PE) at n/nDES = 0.8 [16]. velocity region) on the blade suction side. The results indicate
that the presence of the jet-wake flow structure in the blade pas-
Figure 2 shows the operating conditions under investigation.
sage causes the first method to predict equal and excessively low
Entire compressor maps are provided by Jaatinen-Värri et al. [17]
boundary layer edge velocities on both sides of the blade, which
for the compressor with splitter blades and by Ziegler et al. [16]
is physically incorrect. Qualitatively, the second method seems
for the compressor without splitter blades. The operating points
to give the best estimation for the boundary layer edge. Because
near stall (NS) and choke (NC) have been chosen by compar-
the third method is based on an average value of the free-stream
ing the measured operating maps and typical values used in the
velocity, it fails to capture the boundary layer edge at least on the
literature. In literature, the used near stall mass flow rates nor-
splitter blade suction side in Fig. 3. The fourth method gives ei-
malized by the design mass flow rate vary between 0.70 and
ther extremely thick boundary layers (Fig. 3) or equal velocities
0.91 [12, 18–24], and for the normalized near choke flow rates
on both sides of the blade (Fig. 4) due to the symmetrical flow
between 1.05 and 1.30 [12, 18–24]. In this study, the operating
approximation.
point near stall is 0.8qm,PE and near choke 1.2qm,PE for the com-
When the different methods are studied and compared, the
pressor without splitter blades. For the compressor with splitter
results are also compared to the boundary layer thickness distri-
blades, the operating point near stall is 0.6qm,DES and near choke
bution on a flat plate where one-seventh power law approxima-
1.3qm,DES . Additionally, the point 0.8qm,DES is studied in the
tion is used for the turbulent boundary layer profile. The turbu-
compressor with splitter blades.
lent boundary layer profile is used to give a rough estimation of
The commercial software ANSYS CFX 16.0 is used for nu- the boundary layer development on a flat plate. However, in the
merical calculations. The total pressure and total temperature numerical calculations the transition model is used.
are specified at the inlet boundary and the mass flow rate at the
outlet boundary. Turbulence is modeled using the SST k − ω
model in combination with the transition model of Langtry and Boundary layer thickness
Menter [25]. The use of the transition model in the centrifugal To compare the four free-stream velocity definitions, the
compressor is recommended by e.g. Aungier [26]. The numer- boundary layer thickness is estimated from numerical results at
ical model was validated against experimental data in the previ- different meridional locations from the full blade leading edge
ous paper by the authors [3]. Also a grid dependency study was (0%) to the trailing edge (100%) at the mid-span. The observa-
presented to assure grid independent results. tion locations are presented in Fig. 5.

4 Copyright
c 2017 by ASME
100% from the leading edge (trailing edge)
Normalized relative velocity [-] 1
Uδ = 0.995U∞,inlet
0.8 Un−1 = 0.995Un
Uδ = 0.995U∞,ave
0.6 Uδ = 0.995U∞,middle

0.4
SB FB
0.2
SS PS SS PS
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless pitchwise direction [-]

FIGURE 5. OBSERVATION LOCATIONS ALONG THE MERID-


FIGURE 3. NORMALIZED RELATIVE VELOCITY DISTRIBU- IONAL DIRECTION FROM THE FULL BLADE (FB) LEADING
TION IN THE BLADE PASSAGE OF THE COMPRESSOR WITH EDGE (LE) TO THE TRAILING EDGE (TE).
SPLITTER BLADES (DESIGN POINT, MID-SPAN).

100% from the leading edge (trailing edge) strongly from axial to radial (around 60% from the leading edge).
1.0 The decrease in the boundary layer thickness starts on the splitter
Normalized relative velocity [-]

Uδ = 0.995U∞,inlet
Un−1 = 0.995Un blade pressure side a bit later; 70% from the leading edge (Fig. 7)
0.8
Uδ = 0.995U∞,ave and in the compressor without splitter blades already 40% from
0.6 Uδ = 0.995U∞,middle the leading edge (Fig. 8). However, the decrease in boundary
layer thickness is estimated only by the first and third method,
0.4
which are based on a constant value of free-stream velocity over
FB
0.2 the blade length. The boundary layer thickness increases accord-
SS PS ing to the second method (as will be shown later in Fig. 21),
0.0 which accounts for varying free-stream velocity along the blade
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Dimensionless pitchwise direction [-] length.
The estimation of the decreasing boundary layer thickness
FIGURE 4. NORMALIZED RELATIVE VELOCITY DISTRIBU- is caused by the constant value of free-stream velocity. As the
TION IN THE BLADE PASSAGE OF THE COMPRESSOR WITH- relative velocity decreases in the axial part of the compressor,
OUT SPLITTER BLADES (PEAK EFFICIENCY POINT, MID- the constant value of free-stream velocity results in increasing
SPAN). boundary layer thickness until the flow direction changes from
axial to radial. In the radial part of the compressor, increase in
relative velocity and constant free-stream velocity result in de-
Figures 6, 7, and 8 show estimations of the boundary layer creasing boundary layer thickness. The decrease in boundary
thickness on the blade pressure sides in the compressors with and layer thickness on the splitter blade pressure side starts later than
without splitter blades at their design/peak efficiency points, re- on the full blade pressure side, because the relative velocity starts
spectively. The results of the first, second, and third free-stream increasing later. In the compressor without splitter blades, the
velocity definitions indicate that in the inducer, the boundary decrease in boundary layer thickness starts earlier, because the
layer thickness increases approximately as on a flat plate. From relative velocity starts increasing earlier (figure not shown). The
roughly 25-30% of the chord length, methods 1 and 3 predict a change in relative velocity on the full and splitter blade surfaces
strong increase in the boundary layer thickness. In the region will be shown later in Fig. 13, where free-stream velocity, i.e.
where the flow passage turns from axial to radial (30-80% of relative velocity on the boundary layer edge, varies along the
the chord length), methods 1 and 3 first predict increasing and meridional direction.
then decreasing boundary layer thickness. Near the trailing edge, On the blade suction side (Fig. 9), methods 1, 3, and 4 in-
the first three methods predict similar values for the boundary dicate a boundary layer separation at the meridional location of
layer thickness, but they are all far from the flat plate assump- 10% from the blade leading edge at the peak efficiency point.
tion. Since the fourth method is based on the symmetrical flow After the reattachment at the meridional location of around 20%
assumption, it predicts a boundary layer several orders of magni- from the leading edge, the boundary layer thickness remains al-
tude thicker. most constant, which is not in line with the flat plate assump-
On the full blade pressure side, the boundary layer thickness tion. Because the boundary layer thickness distribution on the
seems to start decreasing at the location where the flow is turning full blade suction side is similar in the compressor with splitter

5 Copyright
c 2017 by ASME
50.0

Boundary layer thickness [mm]


blades, the corresponding figure is not shown. Flat plate
The separation of the boundary layer can also be distin- Uδ = 0.995U∞,inlet
40.0
Un−1 = 0.995Un
guished in Fig. 10 presenting relative velocity vectors at the Uδ = 0.995U∞,ave
mid-span near the leading edge of the compressor without split- 30.0 Uδ = 0.995U∞,middle

ter blades at the peak efficiency point. The advantage of the sec-
20.0
ond method is visible in Fig. 3 on the splitter blade suction side
(SBSS), where the boundary layer edge lies between the blade 10.0
surface and the wake. However, unlike other methods, the sec-
ond method does not capture boundary layer separation on the 0.0
0 20 40 60 80 100
full blade suction side (Fig. 11, FBSS). The weakness in cap- Meridional location [%]
turing boundary layer separation comes from the definition of
the free-stream velocity. Since the capture of the boundary layer
FIGURE 6. BOUNDARY LAYER THICKNESS ON THE FULL
edge is based on the velocity gradient from the blade surface, the
BLADE PRESSURE SIDE (FBPS) OF THE COMPRESSOR WITH
limit for the boundary layer edge is achieved between the blade
SPLITTER BLADES (DESIGN POINT, MID-SPAN).
surface and the separation bubble (as between the blade surface
and the wake).
50.0

Boundary layer thickness [mm]


In the middle of the blade in the meridional direction (Fig. Flat plate
12), the boundary layer thickness estimated by the third method 40.0
Uδ = 0.995U∞,inlet
Un−1 = 0.995Un
is close to that estimated by the second method, since the aver- Uδ = 0.995U∞,ave
age of the free-stream velocities along the meridional direction 30.0 Uδ = 0.995U∞,middle
in the third method is based on the free-stream velocities of the
20.0
second method (U∞,ave = N1 ∑N1 Un and Un is evaluated based on
the second method). 10.0
However, near the trailing edge especially on the splitter
blade suction side (Fig. 3), there is a big discrepancy between the 0.0
0 20 40 60 80 100
third and second methods. Due to the constant free-stream veloc- Meridional location [%]
ity in the third method, white circles are at the same locations on
y axis (normalized relative velocity) at every meridional location.
FIGURE 7. BOUNDARY LAYER THICKNESS ON THE SPLIT-
Because different free-stream velocity is used on full and splitter
TER BLADE PRESSURE SIDE (SBPS) OF THE COMPRESSOR
blades, white circles are at different locations on y axis on ev-
WITH SPLITTER BLADES (DESIGN POINT, MID-SPAN).
ery blade surface. As the free-stream velocity is fixed and does
not vary with the meridional direction in the third method, the
boundary layer edge is highly dependent on the velocity distri-
bution, which causes a big discrepancy at the trailing edge on the reminds boundary layer development on a flat plate more than
splitter blade suction side (3), since the free-stream velocity is e.g. boundary layer development in a pipe. The current analy-
not achieved close to the splitter blade suction side, but almost in sis, however, indicates that constantly increasing boundary layer
the middle of the blade passage. Due to the constant free-stream thickness along the flat plate has its weaknesses in the centrifu-
velocity over the blade length, the third method is not applicable gal compressors due to a varying pitchwise velocity distribution
for estimating the boundary layer thickness over the whole blade inside the blade passage.
length. To explain the discrepancies between the flat plate and cen-
trifugal compressor, Fig. 13 shows the free-stream velocity nor-
malized by the tip speed in the meridional direction on the full
Free-stream velocity (FB) and splitter (SB) blade pressure (PS) and suction (SS) sides
On the one hand, the results presented above indicate that in the compressor with splitter blades. The free-stream velocity
there is no significant variation in the boundary layer thickness in Fig. 13 is based on the second method, which uses the gradi-
when the boundary layer thickness is defined as a gradient be- ent between two adjacent points to estimate the boundary layer
tween two adjacent data points without specifying a value for the edge. As Fig. 13 indicates, the velocity on the blade suction
free-stream velocity (Figs. 6–9). On the other hand, one arbi- side (dashed lines) is higher than that on the blade pressure side
trary value for the free-stream velocity predicts both increasing (solid lines), and after around 40–60% of the meridional length,
and decreasing boundary layer thickness along the blade surfaces the relative velocity increases. The combination of the increased
in the meridional direction. velocity and the jet-wake flow structure might be the reason for
In the past, it was assumed that the turbomachinery blading the different behavior of the boundary layer thickness along the

6 Copyright
c 2017 by ASME
30.0
Boundary layer thickness [mm] Flat plate 10% from the leading edge
1

Normalized relative velocity [-]


Uδ = 0.995U∞,inlet
Un−1 = 0.995Un
Uδ = 0.995U∞,inlet
20.0 Uδ = 0.995U∞,ave 0.8 Un−1 = 0.995Un
Uδ = 0.995U∞,middle Uδ = 0.995U∞,ave
0.6 Uδ = 0.995U∞,middle

10.0
0.4
FB
0.2
0.0 SS PS
0 20 40 60 80 100
Meridional location [%] 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless pitchwise direction [-]
FIGURE 8. BOUNDARY LAYER THICKNESS ON THE FULL
BLADE PRESSURE SIDE (FBPS) OF THE COMPRESSOR WITH- FIGURE 11. NORMALIZED RELATIVE VELOCITY DISTRI-
OUT SPLITTER BLADES (PEAK EFFICIENCY POINT, MID- BUTION IN THE BLADE PASSAGE OF THE COMPRESSOR
SPAN). WITHOUT SPLITTER BLADES (PEAK EFFICIENCY POINT, MID-
SPAN).
6.0
Boundary layer thickness [mm]

Flat plate 50% from the leading edge


1.0

Normalized relative velocity [-]


Uδ = 0.995U∞,inlet
Un−1 = 0.995Un Uδ = 0.995U∞,inlet
4.0 Uδ = 0.995U∞,ave 0.8 Un−1 = 0.995Un
Uδ = 0.995U∞,middle Uδ = 0.995U∞,ave
0.6 Uδ = 0.995U∞,middle

2.0
0.4
FB
0.2
0.0 SS PS
0 20 40 60 80 100
Meridional location [%] 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Dimensionless pitchwise direction [-]

FIGURE 9. BOUNDARY LAYER THICKNESS ON THE FULL


BLADE SUCTION SIDE (FBSS) OF THE COMPRESSOR WITH- FIGURE 12. NORMALIZED RELATIVE VELOCITY DISTRI-
OUT SPLITTER BLADES (PEAK EFFICIENCY POINT, MID- BUTION IN THE BLADE PASSAGE OF THE COMPRESSOR
SPAN). WITHOUT SPLITTER BLADES (PEAK EFFICIENCY POINT, MID-
SPAN).

1.00
FBPS
FBSS
Normalized U∞ [−]

0.75 SBPS
SBSS
0.50

FIGURE 10. RELATIVE VELOCITY VECTORS AT THE MID- 0.25


SPAN NEAR THE LEADING EDGE OF THE COMPRESSOR WITH-
OUT SPLITTER BLADES (PEAK EFFICIENCY POINT). BOUND- 0.00
ARY LAYER SEPARATION IS VISIBLE ON THE BLADE SUCTION 0 20 40 60 80 100
Meridional location [%]
SIDE.

FIGURE 13. NORMALIZED FREE-STREAM VELOCITY DIS-


blade surface compared to that on the flat plate. The velocity TRIBUTION IN THE MERIDIONAL DIRECTION IN THE COM-
distribution in the compressor without splitter blades does not PRESSOR WITH SPLITTER BLADES (DESIGN POINT). THE
markedly differ from that in the compressor with splitter blades. FREE-STREAM VELOCITY IS BASED ON THE SECOND
Hence, it is not shown here. METHOD.

7 Copyright
c 2017 by ASME
Boundary layer losses 2.0
Denton (1993)

Normalized loss coefficients [−]


The results presented above indicate that the second method 1.5
1.0
for defining the free-stream velocity is qualitatively the most 0.5 BL
promising one since it does not include the wake in the bound- 0.0
ary layer. To bring the analysis one step further, the effect of 2.0
Prust (1973)
1.5 19.8 47.6
four free-stream velocity definitions on boundary layer losses is 1.0
investigated next in terms of loss coefficients and the specific en- 0.5
0.0
tropy. od od od od
Figures 14 and 15 present the entropy and kinetic energy loss m et h me
th
me
th
me
th
1st 2n
d
3r
d
4t
h
coefficients in the compressors with and without splitter blades at
their design/peak efficiency points, respectively. Both the splitter
and full blade are accounted for in the evaluation of the loss coef- FIGURE 14. NORMALIZED LOSS COEFFICIENTS IN THE
ficients of the compressor with splitter blades, whereas only one COMPRESSOR WITH SPLITTER BLADES (DESIGN POINT). THE
full blade is accounted for in the case of the compressor with- NORMALIZED LOSS COEFFICIENT BASED ON THE SECOND
out splitter blades. The loss coefficients are normalized by the FREE-STREAM VELOCITY DEFINITION EQUALS UNITY.
loss coefficient based on the second method. The entropy loss
coefficient of Denton is integrated over the blade surface from 1.5

Normalized loss coefficients [−]


Denton (1993)
the leading edge to the trailing edge at the mid-span, whereas 1.0
the kinetic energy loss coefficient of Prust is integrated over the 0.5 BL
blade passage from the blade surface to the boundary layer edge
0.0
at the mid-span at the trailing edge. Prust’s loss coefficient is not 1.5
integrated over the whole blade passage since it would otherwise Prust (1973)
1.0
include the kinetic energy loss due to the wake in addition to that
0.5
due to the boundary layer.
Figure 14 shows that the under-prediction of the boundary 0.0
od od od od
layer thickness by the first method results in the lowest value e th th th th
tm me me me
1s d d h
of the loss coefficient. The third method does not differ signif- 2n 3r 4t
icantly from the second method if the loss coefficient is calcu-
lated over the blade length (Denton) since the third method is FIGURE 15. NORMALIZED LOSS COEFFICIENTS IN THE
based on an average value of the free-stream velocity from the COMPRESSOR WITHOUT SPLITTER BLADES (PEAK EFFI-
second method. However, if the loss coefficient is calculated at CIENCY POINT). THE NORMALIZED LOSS COEFFICIENT
one meridional location (at the trailing edge, Prust), the over- BASED ON THE SECOND FREE-STREAM VELOCITY DEFINI-
prediction of the boundary layer thickness including the wake TION EQUALS UNITY.
on the splitter blade suction side (Fig. 3, SBSS) results in over-
predicted kinetic energy loss. The over-prediction of the bound- 1.75
Normalized specific entropy [−]

ary layer thickness by the fourth method results in the highest 1.50
value of the loss coefficient. With splitter blades
1.25
The results in Fig. 15 indicate that there is not as marked 1.00
a difference between different methods in the compressor with- 0.75
1.75
out splitter blades as there is with the splitters. The kinetic en- BL
1.50
ergy loss coefficients (Prust) are of the same order, which results Without splitter blades
1.25
from the velocity distribution at the trailing edge (Fig. 4). Since 1.00
the boundary layer thicknesses based on different free-stream ve- 0.75
locity definitions do not vary much at the trailing edge, the loss od od od od
m et h m et h m et h m et h
coefficients are also of the same order. 1s
t
2n
d
3r
d
4t
h
The length-averaged change in specific entropy from the im-
peller inlet to the boundary layers at the trailing edge at the mid-
FIGURE 16. NORMALIZED CHANGE IN THE SPECIFIC EN-
span are presented in Fig. 16 for the compressors with and with-
TROPY FROM THE IMPELLER INLET TO THE BOUNDARY LAY-
out splitter blades at their design/peak efficiency points. In the
ERS AT THE TRAILING EDGE (DESIGN POINT). THE SPECIFIC
compressor with splitter blades, both the splitter and full blade
ENTROPY IS NORMALIZED BY THE VALUE BASED ON THE
are accounted for, whereas in the compressor without splitter
SECOND METHOD.
blades, only one full blade is accounted for. The values of spe-

8 Copyright
c 2017 by ASME
100% from the leading edge (trailing edge) of the wake on the splitter blade suction side (SBSS), it also ac-
Uδ = 0.995U∞,inlet counts for the low entropy values on the full blade pressure side
Specific entropy [J/(kgK)]
60.0
Un−1 = 0.995Un
Uδ = 0.995U∞,ave
(FBPS). As the third method only accounts for the high entropy
40.0 Uδ = 0.995U∞,middle values on the splitter blade suction side due to over-predicted
boundary layer thickness, it results in an over-predicted specific
20.0 entropy in Fig. 16. The boundary layer thicknesses of the same
SS PS SS PS order of magnitude at the trailing edge of the compressor without
0.0 SB FB splitter blades results in almost equal values of specific entropy,
as can be seen also in the specific entropy distribution in Fig. 18.
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Dimensionless pitchwise direction [-]
Sensitivity analysis
FIGURE 17. SPECIFIC ENTROPY DISTRIBUTION IN THE The above-presented results show that the first, third, and
BLADE PASSAGE OF THE COMPRESSOR WITH SPLITTER fourth method are very sensitive to the value specified for the
BLADES (DESIGN POINT). THE BLADES LIE BETWEEN THE free-stream velocity. Since the second method uses the velocity
DISCONTINUITIES OF THE CURVE. gradient between two adjacent data points to capture the bound-
ary layer edge and to evaluate the free-stream velocity, it is sen-
sitive to the density of the observed data points on the line per-
100% from the leading edge (trailing edge)
pendicular to the blade surface. The observed data points are
Uδ = 0.995U∞,inlet
Specific entropy [J/(kgK)]

60.0
Un−1 = 0.995Un
specified by the user when post-processing the data, and they
Uδ = 0.995U∞,ave are not based on the computational mesh. Here, the number of
40.0 Uδ = 0.995U∞,middle evenly distributed data points in the pitchwise direction from the
SS PS full blade to the full blade is 10 000, corresponding to approx-
20.0
FB
imately 140 data points per millimeter in the compressor with
splitter blades and approximately 180 data points per millimeter
0.0 in the compressor without splitter blades at the trailing edge. The
closer the blade leading edge is, the larger the density of the data
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Dimensionless pitchwise direction [-] points is.
The sensitivity analysis results of the optimum number of
data points in Figs. 19 and 20 indicate that the boundary layer
FIGURE 18. SPECIFIC ENTROPY DISTRIBUTION IN THE
thickness increases with a decreasing density of data points. On
BLADE PASSAGE OF THE COMPRESSOR WITHOUT SPLITTER
the blade pressure side, the largest differences in boundary layer
BLADES (PEAK EFFICIENCY POINT). THE BLADE LIES BE-
thickness occur when the number of data points is less than 45
TWEEN THE DISCONTINUITIES OF THE CURVE.
per millimeter. On the blade suction side, the boundary layer
separation near the leading edge is not captured with 90 or more
cific entropy are normalized by the value calculated based on the data points per millimeter.
second method. The results indicate that the normalized specific
entropy based on the third method in the compressor with the Hybrid method
splitter blades differs mostly from the other values. To overcome the challenges of defining a boundary layer in
The reason for the over-predicted specific entropy by the centrifugal compressors, a hybrid method is proposed. It com-
third method is visualized in Figs. 17 and 18, where the specific bines the advantages of the second and third methods, being able
entropy distributions in the blade passages are presented for the to both predict separation near the blade leading edge and ex-
compressors with and without splitter blades at their design/peak clude the wake from the boundary layer thickness. The hybrid
efficiency points, respectively. The discontinuities in the curves method is less sensitive to the density of the data points than the
show the locations of the blades. Figures 17 and 18 show the second method since it captures the boundary layer separation
change in the specific entropy from the impeller inlet to the trail- near the blade leading edge also with a higher density of the data
ing edge at the mid-span. The entropy is increased mostly in the points. The hybrid method is defined as
boundary layers and in the wake.
The length-averaged values of the specific entropy are cal- 
culated from the blade surface to the boundary layer edge. Fig-  dU
= 0.005 ⇒ Un−1 = 0.995Un for attached flow (6a)
ure 17 shows that even though the fourth method over-predicts dn
the boundary layer thickness including the high entropy values 
Uδ = 0.995U∞ , ave for separated flow,(6b)

9 Copyright
c 2017 by ASME
3.0 3.0

Boundary layer thickness [mm]


Boundary layer thickness [mm] 9 points per mm 2nd method
18 points per mm 3rd method
45 points per mm Hybrid method
2.0 60 points per mm 2.0
90 points per mm
180 points per mm

1.0 1.0

0.0 0.0
0 20 40 60 80 100 0 20 40 60 80 100
Meridional location [%] Meridional location [%]

FIGURE 19. EFFECT OF THE DATA POINT DENSITY OF THE FIGURE 21. BOUNDARY LAYER THICKNESS ON THE FULL
OBSERVED DATA POINTS ON THE SENSITIVITY OF THE SEC- BLADE PRESSURE SIDE (FBPS) OF THE COMPRESSOR WITH-
OND METHOD ON THE FULL BLADE PRESSURE SIDE OF THE OUT SPLITTER BLADES CALCULATED USING A HYBRID
COMPRESSOR WITHOUT SPLITTER BLADES. METHOD. WHEN FLOW DOES NOT SEPARATE, HYBRID
METHOD EQUALS TO SECOND METHOD.
3.0
Boundary layer thickness [mm]

9 points per mm
18 points per mm
45 points per mm 3.0

Boundary layer thickness [mm]


2.0 60 points per mm 2nd method
90 points per mm 3rd method
180 points per mm Hybrid method
2.0
1.0

1.0
0.0
0 20 40 60 80 100
Meridional location [%]
0.0
0 20 40 60 80 100
Meridional location [%]
FIGURE 20. EFFECT OF THE DATA POINT DENSITY OF THE
OBSERVED DATA POINTS ON THE SENSITIVITY OF THE SEC-
OND METHOD ON THE FULL BLADE SUCTION SIDE OF THE FIGURE 22. BOUNDARY LAYER THICKNESS ON THE FULL
COMPRESSOR WITHOUT SPLITTER BLADES. BLADE SUCTION SIDE (FBSS) OF THE COMPRESSOR WITH-
OUT SPLITTER BLADES CALCULATED USING A HYBRID
METHOD.
where U∞,ave = N1 ∑N1 Un and Un is evaluated based on an attached
flow assumption.
The calculation procedure is iterative and is performed at an
arbitrary number of meridional locations specified by the user:
blades are shown in Figs. 21 and 22, respectively. The hybrid
1. Flow is assumed attached: the free-stream velocity and method calculates the free-stream velocity using Eq. (6a) when
boundary layer thickness are calculated using Eq. (6a). the flow is attached and Eq. (6b) when the flow is separated (Fig.
2. The average value for the free-stream velocity U∞,ave is cal- 22, SBSS 10% of the chord). In Figs. 21 and 22, the results of
culated. the second and third method are shown for comparison. In Fig.
3. The velocity profile is plotted and flow separation on the 21, hybrid method equals to second method because the flow is
blade suction side near the leading edge is qualitatively ana- not separated.
lyzed from the plot by the user.
For the hybrid method, the normalized entropy loss coeffi-
4. The locations of flow separation are specified by the user.
cients (Denton) are 1.04 and 1.07 for the compressors with and
5. The free-stream velocity and boundary layer thickness are
without splitter blades, respectively. The normalized kinetic en-
calculated for attached and separated flow using Eqs. (6a)
ergy loss coefficient (Prust) and normalized specific entropy are 1
and (6b).
for both compressors since they are calculated at the blade trail-
The boundary layer thickness distributions on the blade ing edge where the flow is attached, and therefore, the second
pressure and suction sides of the compressor without splitter method is used for the free-stream velocity.

10 Copyright
c 2017 by ASME
Effect of operating condition on the hybrid method

The results presented above are obtained at the design / peak


efficiency points of the compressors. If other operating condi-
tions are under investigation, the advantage of the hybrid method
is its applicability in the entire compressor map for the sake of the
velocity gradient, which is not dependent on the velocity profile
like the other methods. To support the claim of the applicability
of the hybrid method, the numerical calculations near stall and
near choke are performed.

Figures 23 and 25 show the boundary layer thickness distri-


bution at different operating conditions for the compressors with
and without splitter blades, respectively. Figure 23 indicates that
near choke (NC) the flow separation at the location 10% from
the full blade leading edge is less severe than at the design point.
With decreasing mass flow rate (0.6 and 0.8qm,DES ), the flow
FIGURE 23. BOUNDARY LAYER THICKNESS ON THE FULL
separates more than at the design point as the relative velocity
BLADE SUCTION (UPPER) AND PRESSURE (LOWER) SIDES OF
contours and vectors indicate in Fig. 24. However, the hybrid
THE COMPRESSOR WITH SPLITTER BLADES CALCULATED
method fails to capture the severe flow separation at the lowest
USING A HYBRID METHOD AT DIFFERENT OPERATING CON-
studied mass flow rate (0.6qm,DES ) as can be seen in Fig. 23. In
DITIONS. VELOCITY PROFILES 10% FROM THE FULL BLADE
the velocity profile (Fig. 23, NS), the boundary layer edge on
LEADING EDGE ARE SHOWN ON THE LEFT.
the full blade suction side is predicted between the separated low
velocity region and the blade surface.
TABLE 2. CHANGE IN LOSS COEFFICIENTS AND SPECIFIC
The reason for this under-prediction of the boundary layer ENTROPY BETWEEN DESIGN POINT AND OFF-DESIGN OPER-
thickness is that near stall, the relative velocity is low in the en- ATING CONDITIONS IN THE COMPRESSOR WITH SPLITTER
tire blade passage, not just in the separated flow region or in the BLADES.
wake. Therefore, the average free-stream velocity becomes low
near stall, resulting in under-predicted boundary layer thickness 0.6qm,DES 0.8qm,DES 1.3qm,DES
at the flow separation location (10% from the full blade lead- Denton (1993) −34% −33% −23%
ing edge) of the compressor with splitter blades. However, in
the compressor without the splitter blades (Fig. 25), more se- Prust (1973) −4% +4% +8%
vere boundary layer separation near the full blade leading edge Specific entropy +110% +70% +258%
is captured with a decreasing mass flow rate.

To conclude, the hybrid method is applicable in the range


0.8 − 1.3 of the design mass flow rate in the compressor with CONCLUSIONS
splitter blades and in the range 0.8 − 1.2 of the peak efficiency The aim of this paper was to show how the generally used
mass flow rate in the compressor without splitter blade. The hy- definitions for the free-stream velocity affect the estimation of
brid method has a limitation near stall, but more results are re- the boundary layer thickness in the blade passage of a centrifu-
quired to make a general conclusion. gal compressor with a complex flow field. The effect of four
free-stream velocity definitions is demonstrated inside two cen-
The changes in entropy loss coefficient (Denton), kinetic en- trifugal compressors: one with splitter blades and the other with-
ergy loss coefficient (Prust) and specific entropy between the de- out. Also the effect of the free-stream velocity definitions on
sign / peak efficiency point and off-design operating points are the boundary layer losses is demonstrated using two loss coeffi-
presented in Tables 2 and 3 for the compressors with and with- cients: an entropy loss coefficient accounting for the losses from
out splitter blades, respectively. The results indicate that the loss the blade leading edge to the trailing edge in the meridional di-
coefficients based on the velocity ratio (Denton and Prust) are rection and a kinetic energy loss coefficient accounting for the
not comparable in a centrifugal compressor at different operating losses from the blade surface to the boundary layer edge in the
conditions, resulting from a change in a mass flow rate. However, pitchwise direction. In addition to the two loss coefficients, the
the loss estimation based on specific entropy distribution predicts specific entropy values in the boundary layer are compared be-
increased losses at off-design conditions. tween four free-stream velocity definitions. The pros and cons of

11 Copyright
c 2017 by ASME
FIGURE 25. BOUNDARY LAYER THICKNESS ON THE FULL
BLADE SUCTION (UPPER) AND PRESSURE (LOWER) SIDES
OF THE COMPRESSOR WITHOUT SPLITTER BLADES CALCU-
LATED USING A HYBRID METHOD AT DIFFERENT OPERAT-
ING CONDITIONS. VELOCITY PROFILES 10% FROM THE FULL
BLADE LEADING EDGE ARE SHOWN ON THE LEFT.

FIGURE 24. RELATIVE VELOCITY CONTOURS IN THE COM-


PRESSOR WITH SPLITTER BLADES. VELOCITY VECTORS
FROM THE FULL BLADE LEADING EDGE TO 20% OF THE
CHORD ARE SHOWN ON THE LEFT.

TABLE 3. CHANGE IN LOSS COEFFICIENTS AND SPECIFIC


ENTROPY BETWEEN PEAK EFFICIENCY POINT AND OFF-
DESIGN OPERATING CONDITIONS IN THE COMPRESSOR
WITHOUT SPLITTER BLADES.

0.8qm,PE 1.2qm,PE
Denton (1993) −49% −19%
Prust (1973) −16% −13%
Specific entropy +70% +186%

the free-stream velocity definitions are summarized in Table 4. FIGURE 26. RELATIVE VELOCITY CONTOURS IN THE COM-
The four free-stream velocity definitions were compared to PRESSOR WITHOUT SPLITTER BLADES. VELOCITY VECTORS
the fully turbulent flat plate boundary layer profile. In this study, FROM THE FULL BLADE LEADING EDGE TO 20% OF THE
the fully turbulent approximation seemed to be adequate since CHORD ARE SHOWN ON THE LEFT.
the numerical results accounting for the transition (SST k − ω +

12 Copyright
c 2017 by ASME
TABLE 4. SUMMARY OF THE INVESTIGATED FREE-STREAM [2] Mateescu, D., and Abdo, M., 2010. “Analysis of Flows Past
VELOCITY DEFINITIONS. Airfoils at Very Low Reynolds Numbers”. Proceedings of
the Institution of Mechanical Engineers, Part G: Journal of
Method Pros, cons, and effect on loss coefficients Aerospace Engineering, 224(7), pp. 757–775.
1st + [3] Tiainen, J., Jaatinen-Värri, A., Grönman, A., and Backman,
J., 2016. “Numerical Study of the Reynolds Number Effect
− under-prediction, equal velocity on PS and SS
on the Centrifugal Compressor Performance and Losses”.
→ under-predicted loss coefficients In Proceedings of the ASME Turbo Expo 2016: Turbine
2nd + BL edge captured between the blade surface Technical Conference and Exposition. Paper No. GT2016-
and the wake 56036, p. 10. June 13–17, 2016, Seoul, South Korea.
[4] Denton, J. D., 1993. “Loss Mechanisms in Turboma-
− BL separation not captured
chines”. Journal of Turbomachinery, 115, pp. 621–656.
→ most promising method [5] Prust Jr., H. W., 1973. “Boundary-Layer Losses”. In Tur-
3rd + BL separation captured bine Design and Application, A. J. Glassman, ed., Vol. 2.
ch. 7, pp. 93–122. Washington, D.C.: National Aeronau-
− wake might be included in the BL thickness tics and Space Administration. NASA SP–290.
→ loss coefficients might be over-predicted [6] Harrison, S., 1990. “Secondary Loss Generation in a Lin-
ear Cascade of High-Turning Turbine Blades”. Journal of
4th +
Turbomachinery, 112, pp. 618–624.
− over-prediction, wake might be included in the [7] Weber, A., Schreiber, H.-A., Fuchs, R., and Steinert, W.,
BL thickness 2002. “3-D Transonic Flow in a Compressor Cascade With
→ loss coefficients might be highly over-predicted Shock-Induced Corner Stall”. Journal of Turbomachinery,
124, pp. 358–366.
[8] Klausner, E., and Gampe, U., 2014. “Evaluation and En-
Langtry and Menter) did not markedly differ from the numerical hancement of a One-dimensional Performance Analysis
results with a fully turbulent flow assumption (SST k − ω). Method for Centrifugal Compressors”. In Proceedings of
As a result of this study, a hybrid method for defining the ASME Turbo Expo 2014: Turbine Technical Conference
free-stream velocity and calculating the boundary layer thickness and Exposition. Paper No. GT2014-25141, p. 11. June 16–
is proposed. It combines the advantages of the second and third 20, 2014, Düsseldorf, Germany.
methods: the hybrid method is able to capture both the bound- [9] Choi, M., Baek, J., Chung, H., Oh, S., and Ko, H., 2008.
ary layer separation and the boundary layer edge lying between “Effects of the Low Reynolds Number on the Loss Charac-
the blade surface and the wake. As only two compressors in de- teristics in an Axial Compressor”. Proceedings of the Insti-
sign and off-design operating conditions have been under inves- tution of Mechanical Engineers, Part A: Journal of Power
tigation, future work among different applications and operating and Energy, 222, pp. 209–218.
conditions would make the results more general. [10] Lynch, S., and Thole, K., 2016. “Comparison of the
Three-Dimensional Boundary Layer on Flat Versus Con-
toured Turbine Endwalls”. Journal of Turbomachinery,
138, pp. 041008–1–10.
ACKNOWLEDGMENTS [11] Hergt, A., Meyer, R., and Engel, K., 2006. “Experimental
The authors would like to acknowledge the financial contri- Investigation of Flow Control in Compressor Cascades”. In
bution of the Academy of Finland. This research is part of the Proceedings of ASME Turbo Expo 2006: Power for Land,
project funded by the Academy of Finland under grant number Sea and Air. Paper No. GT2006-90415, p. 10. May 8–11,
274897. 2006, Barcelona, Spain.
[12] Bousquet, Y., Carbonneau, X., Dufour, G., Binder, N., and
Trebinjac, I., 2014. “Analysis of the Unsteady Flow Field
REFERENCES in a Centrifugal Compressor from Peak Efficiency to Near
[1] Boyce, M. P., 1993. “Principles of Operation and Perfor- Stall with Full-Annulus Simulations”. International Jour-
mance Estimation of Centrifugal Compressors”. In Pro- nal of Rotating Machinery, 2014, pp. Article ID 729629, 11
ceedings of The Twenty-Second Turbomachinery Sympo- pages.
sium, Turbomachinery Laboratory, Texas A&M University, [13] Zheng, X., Lin, Y., Gan, B., Zhuge, W., and Zhang, Y.,
College Station, Texas, pp. 161–178. September 14–16, 2013. “Effects of Reynolds Number on the Performance of
1993, Dallas, TX, United States. a High Pressure-Ratio Turbocharger Compressor”. Science

13 Copyright
c 2017 by ASME
China: Technological Sciences, 56(6), pp. 1361–1369. Huenteler, J., and Li, Z., 2012. “Stability Improvement of
[14] Hildebrandt, A., Franz, H., and Jakiel, C., 2011. “Numeri- High-Pressure-Ratio Turbocharger Centrifugal Compressor
cal Analysis of Effects of Centrifugal Compressor Impeller by Asymmetric Flow Control-Part I: Non-Axisymmetrical
Design on Overall- and Flow Field Performance”. In Pro- Flow in Centrifugal Compressor”. Journal of Turboma-
ceedings of ASME Turbo Expo 2011: Turbine Technical chinery, 135(2).
Conference and Exposition. Paper No. GT2011-45014, [23] Xu, W., Wang, T., Gu, C., and Ding, L., 2011. “A Study on
p. 11. June 6–10, 2011, Vancouver, British Columbia, the Influence of Hole’s Diameter With Holed Casing Treat-
Canada. ment”. In Proceedings of ASME Turbo Expo 2011, Vol. 4,
[15] Jaatinen-Värri, A., Turunen-Saaresti, T., Röyttä, P., pp. 499–508. Paper No. GT2011-46167, 10 pages. June
Grönman, A., and Backman, J., 2013. “Experimental Study 6–10, 2011, Vancouver, British Columbia, Canada.
of Centrifugal Compressor Tip Clearance and Vaneless Dif- [24] Röyttä, P., Grönman, A., Jaatinen, A., Turunen-Saaresti, T.,
fuser Flow Fields”. Proceedings of the Institution of Me- and Backman, J., 2009. “Effects of Different Blade Angle
chanical Engineers, Part A. Journal of Power and Energy, Distributions on Centrifugal Compressor Performance”. In-
227, pp. 885–895. ternational Journal of Rotating Machinery, 2009, pp. Arti-
[16] Ziegler, K. U., Gallus, H. E., and Niehuis, R., 2003. “A cle ID 537802, 9 pages.
Study on Impeller-Diffuser Interaction: Part I - Influence [25] Langtry, R., and Menter, F., 2005. “Transition Modeling for
on the Performance”. Journal of Turbomachinery, 125(1), General CFD Applications in Aeronautics”. In 43rd AIAA
pp. 173–182. Aerospace Sciences Meeting and Exhibit. Paper No. AIAA
[17] Jaatinen-Värri, A., Turunen-Saaresti, T., Grönman, A., 2005-522. January, 10–13, 2005, Reno, Nevada.
Röyttä, P., and Backman, J., 2013. “The Tip Clearance [26] Aungier, R. H., 2000. Centrifugal Compressors: A Strat-
Effects on the Centrifugal Compressor Vaneless Diffuser egy for Aerodynamic Design and Analyses. The American
Flow Fields at Off-Design Conditions”. In Proceedings of Society of Mechanical Engineers. p. 315, ISBN 0-7918-
the 10th European Conference on Turbomachinery: Fluid 0093-8.
Dynamics and Thermodynamics, J. Backman, G. Bois,
and O. Leonard, eds., pp. 972–982. April 15-19, 2013,
Lappeenranta, Finland.
[18] Shahin, I., Alqaradawi, M., Gadala, M., and Badr, O., 2017.
“On the Aero Acoustic and Internal Flows Structure in a
Centrifugal Compressor with Hub Side Cavity Operating at
Off Design Condition”. Aerospace Science and Technology,
60, pp. 68–83.
[19] Weber, A., Morsbach, C., Kgeler, E., Rube, C., and Wedek-
ing, M., 2016. “Flow Analysis of a High Flowrate Cen-
trifugal Compressor Stage and Comparison With Test Rig
Data”. In Proceedings of ASME Turbo Expo 2016: Turbo-
machinery Technical Conference and Exposition, Vol. 2D-
2016. Paper No. GT2016-56551, 12 pages. June 13–17,
2016, Seoul, South Korea.
[20] Bousquet, Y., Carbonneau, X., Trébinjac, I., Dufour, G.,
and Roumeas, M., 2013. “Description of the Unsteady
Flow Pattern from Peak Efficiency to Near Surge in a Sub-
sonic Centrifugal Compressor Stage”. In Proceedings of
the 10th European Conference on Turbomachinery: Fluid
Dynamics and Thermodynamics, J. Backman, G. Bois,
and O. Leonard, eds., pp. 917–927. April 15-19, 2013,
Lappeenranta, Finland.
[21] Ding, L., Wang, T., Yang, B., Xu, W., and Gu, C., 2013.
“Experimental Investigation of the Casing Treatment Ef-
fects on Steady and Transient Characteristics in an Indus-
trial Centrifugal Compressor”. Experimental Thermal and
Fluid Science, 45, pp. 136–145.
[22] Yang, M., Zheng, X., Zhang, Y., Bamba, T., Tamaki, H.,

14 Copyright
c 2017 by ASME

You might also like