Professional Documents
Culture Documents
Introduction lower Reynolds numbers, the thicker boundary layers will result
in larger wakes. Concerning Mach number, Stanitz 共in the discus-
The rate at which fans, compressors, and pumps do flow work
sion attached to Wiesner 关1兴兲 reported that his results showed only
is less than that calculated with the assumption that the relative
flow at the exit of a rotor follows the blade trailing edges. The a small effect up to tip Mach numbers of 2.0.
angular momentum imparted to the flow is reduced by a factor Directly, or by implication, textbooks generally treat the rela-
known as the slip factor in radial flow machines. tive eddy as the major factor causing slip in radial flow turboma-
Despite the recent advances in computational fluid dynamics, chines, for example, Stodola 关2兴, Eckert and Schnell 关3兴, Ferguson
engineers and students still need a reliable method for first esti- 关4兴, Wislicenus 关5兴, Osborne 关6兴, Eck 关7兴, Dixon 关8兴, Watson and
mates of the slip factor in centrifugal impellers. Such a method Janota 关9兴, Cumpsty 关10兴, Logan 关11兴, Johnson 关12兴, Wilson and
should be direct 共no iteration or conditional procedures兲, have a Korakianitis 关13兴, Aungier 关14兴, and Saravanamuttoo et al. 关15兴.
sound fluid dynamic basis, be widely applicable in terms of basic At least, they generally do not attempt to model the other contrib-
impeller geometry such as blade number, blade angle, and impel- uting factors. Dean and Young 关16兴 and Japikse and Baines 关17兴
ler radius ratio, and be relatively accurate. do, however, consider the effect of the wake region in the blade
passage, but jet-wake models still require a slip factor correlation
in the jet flow region where viscous effects do not dominate.
Background Stodola 关2兴 presented a simplified and popular approximate
The main mechanism usually considered when predicting the derivation followed by many textbooks. He inserted a circular-
slip factor in radial flow machines is the so-called relative eddy. shaped control volume between the blades, near the outer radius
This is an inviscid flow effect. A fluid element entering a radial of the rotor. The circle touches the suction side trailing edge of
flow impeller does not rotate around its own axis with an angular one blade and is tangent to the pressure surface of its neighbor.
velocity equal to that of the rotor, but moves around the machine For a rotor with exit radius re and number of blades, Z, the blade
axis while maintaining a constant orientation relative to the ma- spacing is 2re / Z, the eddy diameter is 2e = 共2re / Z兲cos , with
chine casing. Relative to the rotor, however, the fluid element  the blade exit angle, measured from the radial direction. Stodola
rotates at an angular velocity equal but opposite to the angular assumed the slip velocity caused by the relative eddy to be equal
velocity of the rotor. The relative vorticity of the flow entering the in magnitude to the speed of rotation of the eddy at its rim: ⌬w
rotor will set up a recirculating flow pattern relative to the rotor. In = ⍀e = ⍀re共cos 兲 / Z = Ue共cos 兲 / Z. A recent example of such
centrifugal impellers it affects the primary flow by causing under- an approach is the paper of Paeng and Chung 关18兴. The present
turning across the rotor exit plane. study was started because the assumption that the eddy rim veloc-
Other mechanisms that cause slip are the relaxation of the cross ity ⌬w may be applied along the rotor perimeter 共the edge of
passage pressure gradient near the blade trailing edges and the another control volume兲 as the so-called slip velocity was difficult
unequal boundary layer displacement thicknesses on the pressure to justify, especially in a teaching situation.
and suction sides of the blades. The thicker boundary layers asso- Busemann 关19兴 proposed a remarkable slip factor prediction
ciated with lower Reynolds numbers enhances this effect. The method that was sensitive to the blade radius ratio. The blade
existence of a wake region in the passages of radial flow machines radius ratio is the radial distance of the blade leading edge from
may also play a part, as may hub and shroud drag and tip leakage the axis divided by that of the blade trailing edge. He analytically
flow. The wake develops in the suction side shroud corner of the solved the inviscid flow field through a series of two-dimensional
flow passage, as the pressure gradients associated with the flow impellers with logarithmic spiral blades. He generated maps of
turning from axial to radial, and the Coriolis force turns low mo- slip factor versus the impeller radius ratio, with the blade number
mentum fluid in the boundary layers more than the main flow. At as a parameter, for various blade sweep angles for logarithmic
spiral blades. Wislicenus 关5兴 and Wiesner 关1兴 reproduced these
Contributed by the Computational Fluid Dynamics Committee of ASME for pub-
maps 共for example, Fig. 3兲. The Busemann maps indicated that
lication in the JOURNAL OF TURBOMACHINERY. Manuscript received April 1, 2004; final slip factor depends on RR, but below a critical value of RR it is
manuscript received August 10, 2005. Review conducted by M. Casey. relatively constant. The popular method of Wiesner 关1兴 was de-
⌫cv = 冕 cv
V ds 共1兲
but depends on them collectively insofar as they affect so-
lidity.
共3兲 The factors that determine the influence coefficient F must
The average vorticity is: still be determined, and may include any or all of the
above.
= ⌫cv/Acv . 共2兲 共4兲 When blade solidity is zero 共no blades or infinitely short
The blade length is: blades兲, the equation correctly predicts that the slip velocity
is equal in magnitude to the rotor rim speed.
c = 共re − ri兲/cos  共3兲 共5兲 Impellers with splitter vanes can be handled by using a
The blade spacing is: control volume that includes two adjacent blade passages,
containing two rotor perimeter pitches and a suction and
se = 2re/Z 共4兲 pressure side of each of the main and splitter blades. This
The next step is to calculate the circulation by adding the contri- will be equivalent to using the mean blade length.
butions along each section of the control volume boundary. There 共6兲 The assumption of logarithmic spiral blades turns out to be
is no net contribution from the two boundary sections extending unnecessary at this point, but the blade shape may affect the
inward toward the rotor axis from the blade-leading edges since value of F.
the velocity components along these two boundaries are equal due
to periodicity, but the integration direction is opposite when inte- Definition of Slip Factor
grating around the boundary. The circulation is then: There are basically two definitions of slip factor. Both are equal
⌫cv = f e ⌬w 2re/Z + 共f p ⌬w + f s ⌬w兲共re − ri兲/cos  to one minus the normalized slip velocity. In the one definition the
冉 冊
amount of slip is normalized by dividing the slip velocity by the
f p + f s 共1 − ri/re兲Z rotor rim speed and in the other by the ideal 共slipless兲 circumfer-
= 共f e ⌬w 2re/Z兲 1 + 共5兲
f e 2 cos  ential fluid velocity component. The second one introduces the
complication that the circumferential fluid velocity component is
and
dependent on the flow through the impeller, except in the case of
1 ⌫cv ⌫cv radial blades 共e = 0兲, when the two definitions are equivalent. As
⍀= = = 共6兲 the second definition contradicts the assumption made in the deri-
2 2 Acv 2r2e /Z
vation that eddy-induced slip is independent of throughflow, we
⬖⍀ =
共f e ⌬w 2re/Z兲
2r2e /Z
冉 1+
共f p + f s兲 共1 − RR兲Z
fe 2 cos 
冊 共7兲
shall follow Wiesner 关1兴 and use the first definition:
s = 1 − 共⌬ws/U兲 = 1 − 1/关1 + F共c/se兲兴 共12兲
with RR= ri / re. Then: Since the magnitude of the other factors affecting slip, like the
冉 冊
trailing edge pressure gradient relaxation and boundary layer
⌬ws f e ⌬w 共f p + f s兲 共1 − RR兲Z blockage effect 共including the existence of wakes兲 are also prima-
= = 1+ 共8兲
Ue ⍀re fe 2 cos  rily dependent on solidity, solidity should correlate measured slip
factors well, at worst with a different coefficient F for each family
where the average slip velocity along the exit boundary is ⌬ws
of impellers.
= f e ⌬w, and ⍀re = Ue.
It is known that in practice slip factors are not independent of
The normal definition of blade row solidity is the blade chord
flow, but the relative eddy can, in terms of its definition, not be the
divided by the spacing, but to keep things simple, we shall replace
cause of these variations.
the chord by the blade length 共in a plane perpendicular to the rotor
axis兲 and use the spacing at the radius, re, of the blade trailing
The Dependence of F on Blade Angle
edges 共rotor rim兲. The solidity is then:
The next step is to determine the dependence of F on Z, , and
共re − ri兲/cos  共1 − RR兲Z RR. As a first approximation the dependence on  alone will be
c/se = = 共9兲
共2re兲/Z 2 cos  investigated, since Busemann drew separate figures for each blade
Define the solidity influence coefficient as: angle. Busemann 关19兴 共also Wiesner 关1兴 and Wislicenus 关5兴兲 pre-
sented graphs for each of the following blade angles B = 90 deg,
fp + fs 60 deg, 40 deg, 20 deg, 10 deg, and 5 deg 共measured from the
F= 共10兲
fe circumferential direction兲, corresponding to our angles  = 0 deg,
30 deg, 50 deg, 70 deg, 80 deg, and 85 deg 共measured from the
Since it depends on the relative magnitudes 兰V ds over the blade radial direction兲. Wiesner’s graphs show Busemann’s slip factor as
surfaces, compared to the value over the blade passage exit sur- a function of radius ratio for blade numbers 1, 2, 4, 8, and 16 with
face, F can be expected to be a function of the blade angle and of the graph for 30 blades estimated by Wiesner. Figure 3 shows the
the aspect ratio 共or solidity兲 of the blade passage. The normalized Busemann graph for  = 30 deg, with the line for 1 blade removed,
slip velocity is then simply: to make room for a legend, and Fig. 4 shows the corresponding
⌬ws 1 graph calculated from Eq. 共12兲 with F = 4.45. The graphs have
= 共11兲 roughly the same shape, but the SRE method predicts that the slip
Ue 1 + F共c/se兲
factor continues to increase with a decrease in RR, even at low
Typical values of solidity extracted from the data of Wiesner 关1兴 values of RR, say, below RR= 0.4. An inspection of the data re-
range from 0.5 to 2.5, with a few values as high as 3.5 and an ported by Wiesner, however, shows that, if we disregard the pump
average of 1.5. We shall see that the average value of F is about 4, data of Varley 共from Wiesner 关1兴兲, who presented a set of data for
implying that for c / se = 1.5 the normalized slip velocity 共⌬ws / U兲 RR= 0.338, then 90% of the rest falls within the range 0.4⬍ RR
is about 1 / 7 or 0.14. ⬍ 0.6.
冒冉 冊
only a 2.5% error in slip factor. Conversely, a small error in esti-
mating the Busemann values from Wiesner’s graphs would result Z共1 − RR兲
=1−1 1 + 5共cos 兲0.5
in a large variation in F. 2 cos 
The SRE slip factor equation then becomes:
=1−1 冒冉 1+5
共1 − RR兲Z
2共cos 兲0.5
冊 共14兲
Fig. 4 The SRE slip factor for  = 30 deg versus the radius ra- Fig. 6 Variation of the SRE slip factor with solidity, c / se, with
tio for various blade numbers blade angle, , as a parameter
Z 4 8 16 32 Z 4 8 16 32
Exact SRE 共50 deg兲 0.665 0.799 0.888 0.941 SRE,  = 0 deg, RR= 0.5 0.614 0.761 0.864 0.927
Approximate SRE 共50 deg兲 0.496 0.748 0.874 0.937 SRE,  = 0 deg, RR= 0.4 0.656 0.793 0.884 0.939
Stodola 共50 deg兲 0.495 0.748 0.874 0.937 Stanitz 0.706 0.828 0.906 0.951
Exact SRE 共65 deg兲 0.710 0.830 0.907 0.951
Stodola 共65 deg兲 0.668 0.834 0.917 0.959
s = 1 − 1 冒冉 1+5
共0.5兲Z
2共cos 兲0.5
冊⬇ 1 − 0.8冑共cos 兲/Z
The approximate SRE equation resembles the Wiesner equation in
its dependence on 冑cos , but not in its dependence on Z, but Eqs.
共16兲 共21兲 and 共22兲 predict exactly the same value of slip factor for Z
= 21.6, for any blade angle. The 0.7 exponent for Z in the Wiesner
The major difference between Eq. 共16兲 and Eq. 共15兲 is in the equation disagrees with the Stodola, Stanitz, and approximate
square root, but they are exactly equivalent for  = 50.2 deg for SRE equations. The fundamental Wiesner equation 共Eq. 共21兲兲, like
any number of blades. Dixon 关8兴 states that the Stodola equation the Stodola and Stanitz equations, predicts that as Z approaches
gives the best results in the range of 60 deg to 70 deg. Table 1 zero, the slip factor approaches minus infinity, but because of the
compares exact and approximate SRE predictions to Stodola. For lower valued exponent 共0.7兲 it does so more slowly. As the fun-
8 or more blades the agreement is within 0.01, but the Stodola damental equation also assumes that slip factor is independent of
equation is inaccurate for small Z, and incorrectly predicts that in RR, Wiesner proposed an empirical correction factor to force the
the limit of zero blades the slip factor equals minus infinity, slip factor to zero as RR approaches unity 共and blade length ap-
whereas the SRE method 共in its exact form兲 correctly predicts it to proaches zero兲. The correction factor is:
be equal to zero.
FRR = 1 − 关共RR − RRlim兲/共1 − RRlim兲兴3 共23兲
Comparison to Equation of Stanitz where RRlim = exp关8.16共cos 兲 / Z兴 is the limiting radius ratio. It
A commonly used expression for the slip factor of radial bladed must be applied only when the limiting radius ratio is exceeded.
impellers 共 = 0 deg兲 is one proposed by Stanitz 关21兴, based on Note that Wiesner 关1兴 did not attempt to model the physics by
numerical solution of flow fields in impellers with radius ratio correcting the slip velocity, but applied the correction to the slip
0.445: factor directly.
Table 3 compares SRE predictions with Busemann and Wiesner
s = 1 − 0.63/Z 共17兲 for a typical RR value of 0.5, and a wide range of blade angles.
Note that this equation does not agree with the Stodola equation The Busemann data is given to two decimals only, having been
共15兲 for  = 0 deg, which gives: measured off figures in Wiesner 关1兴. The asterisk 共*兲 denotes val-
s = 1 − /Z 共18兲
Table 3 SRE slip factors compared to values predicted by
For  = 0 deg and RR= 0.5, the approximate SRE equation 共16兲 Busemann and Wiesner, for RR= 0.5
above reduces to:
Z 4 8 16 30
s ⬇ 1 − 0.8/Z 共19兲
This equation gives values that are about halfway between the Busemann,  = 0 deg 0.56 0.76 0.86 0.91
Stodola equation 共18兲 for  = 0 deg, and the Stanitz equation. For SRE,  = 0 deg 0.614 0.761 0.864 0.923
 = 0 deg and RR= 0.4, however, the approximate equation 共14兲 Wiesner,  = 0 deg 0.575* 0.763* 0.856 0.908
above reduces to: Busemann,  = 50 deg 0.63 0.79 0.88 0.92
SRE,  = 50 deg 0.665 0.799 0.888 0.937
2 Wiesner,  = 50 deg 0.676* 0.813 0.885 0.926
s ⬇ 1 − /Z = 1 − 0.67 . 共20兲
3 Busemann,  = 70 deg 0.71 0.83 0.91 0.94
SRE,  = 70 deg 0.731 0.845 0.916 0.953
The agreement between Eqs. 共17兲 and 共20兲 is striking. The full Wiesner,  = 70 deg 0.778 0.864 0.916 0.946
SRE is compared to the Stanitz equation in Table 2 and shows that Busemann,  = 80 deg 0.83 0.92 0.96 0.96
for RR= 0.4 agreement is within 0.025 for 8 or more blades. The SRE,  = 80 deg 0.793 0.884 0.939 0.966
Stanitz equation, unlike the Stodola equation, consistently predicts Wiesner,  = 80 deg 0.841 0.903 0.940 0.961
higher values of slip factor than the SRE method. It is gratifying
cases. Both figures support the notion that a prime factor that
ues corrected for exceeding Wiesner’s limiting radius ratio. The determines the slip factor is the modified blade solidity,
共c / se兲冑cos .
average value of the SRE value minus the Busemann value is
+0.0068 and the standard deviation of the difference is 0.054,
compared to +0.0113% and 0.057 for Wiesner. For the 16 points
considered, SRE performs slightly better in terms of the average Comparison to Equation of Eck
difference and the standard deviation of the difference when com- In the Discussion section attached to Wiesner’s paper 关1兴, Love
pared to Busemann. Although Wiesner typically predicts slip fac- proposed an equation for the slip factor derived by Eck 关7兴, who
tors to be about 0.005 higher than SRE, the standard deviation derived his equation by using pressure gradient arguments. The
based on the difference between them is only 0.002. equation is:
冒冉 冊
Figure 6 is a graphical representation of the SRE slip factor
prediction, Eq. 共14兲. It presents the slip factor as a function of 共/2兲cos 
s = 1 1+ 共24兲
solidity, with the blade angle as a parameter. All the lines in the Z共1 − RR兲
figure can, however, be collapsed onto a single line if the x-axis It may be rewritten as:
value is changed to 共c / se兲冑cos , the so-called modified solidity.
Figures 7 and 8 present comparisons between Eq. 共14兲 and the
Busemann and Wiesner slip factors, respectively, for all the appli-
cable test cases in Wiesner 关1兴. The agreement is good in both
s = 1 冒冉 1+
2 cos 
4 Z共1 − RR兲
冊
=1−1 冒冉 1+4
Z共1 − RR兲
2 cos 
冊
共25兲
Equations 共14兲 and 共25兲 are very similar. The only difference is
that in the Eck equation the coefficient Z has a constant value of 4
instead of 5共cos 兲0.5. At  = 50.2 deg the two equations are iden-
tical. The comparisons with Stodola, Stanitz, and Wiesner have,
however, confirmed that the 5共cos 兲0.5 coefficient as a good ap-
proximation. As the coefficient Z varies between 1.00 at 0 deg and
0.30 at 85 deg, it is unlikely that a constant value would be ad-
equate over a wide range of blade angles.
Neither Eck nor Love pointed out that the fundamental param-
eter influencing slip factor is solidity.
blade angle and radius ratio. It is apparent from Fig. 9 that Eq. = 0.5兲 to those of Busemann 共measured from the graphs in
共26兲 above 共represented by the lines兲 predicts Csanady’s data Wiesner, for example, Fig. 3兲 for RR= 0. The agreement is gener-
points surprisingly well. Since these curves were derived from ally within three percentage points, but it deteriorates as the blade
Busemann’s data they demonstrate once more how well SRE ap- number decreases. The Wiesner correlation 共Fig. 12兲 is very good
proximates the Busemann calculations. Figure 10 shows a similar for 16 blades 共and for 30, where Busemann data did not exist, but
graph using Wiesner’s correlation, but the agreement is not nearly were estimated by Wiesner兲, but it is relatively inaccurate for
as good, except at blade angles exceeding 65 deg, and it cannot lower blade numbers or large blade angles.
model the effect of RR explicitly. Figures 13 and 14 present comparisons between Eq. 共14兲 and
the Busemann and Wiesner slip factors respectively, as before, but
Slip Factor For Low Radius Ratio with RR values of less than 0.5 changed to 0.5. The agreement is
The study so far has shown that the SRE method predicts slip improved and can be described as excellent in both cases. It is
factor well for RR⬇0.5, but initial comparisons 共Figs. 3 and 4兲 remarkable that the simple SRE correlation manages to represent
showed that it overpredicted at low values of RR. The simplest Busemann’s predictions accurately for all Wiesner’s data points.
way of removing this discrepancy is to assume that if RR⬍ 0.5, SRE agrees better with Busemann than with Wiesner’s prediction
then RR= 0.5. This is equivalent to assuming that blade-leading method.
edges extending inward beyond RR= 0.5, do not contribute to the Although assuming a fixed value of RR= 0.5 as the cutoff point
control of slip. Figure 11 compares SRE slip factors 共for RR may be regarded as simplistic, it works very well. The author has
derived a more complicated method, but does not consider it
Fig. 11 A comparison of SRE „with RR= 0.5… and Busemann Fig. 13 SRE „with RRÐ 0.5… and Busemann predictions of the
slip factors for RR= 0 slip factor for Wiesner test cases, versus „c / se… 冑 cos 
Fig. 15 SRE prediction and Wiesner test cases for slip factors Fig. 17 The slip factor comparison between the Wiesner pre-
versus „c / se… 冑 cos  diction and Wiesner test cases
Error Estimate For Experimental Data Reported by suming realistic radius ratios and excluding rotors with a few
Wiesner blades, where many trusted methods start to approach an unreal-
istic limiting value.
Experimental slip factors in Wiesner’s paper are calculated as: The SRE equation is, next to Eck’s, the only one to intrinsically
= / + 2 tan 2, where is the pressure or head coefficient, predict the zero slip factor for zero blades. It also correctly pre-
is the efficiency, 2 is the impeller discharge flow coefficient, and dicts the zero slip factor for blades of zero length 共RR= 1兲, inde-
2 is the impeller discharge angle. 共The subscripts 2 are omitted pendent of the number of blades, without the need to calculate a
from hereon.兲 If one can estimate the typical standard deviation Si limiting radius ratio, and it can handle splitter blades in a logical
in determining each of the four independent variables, the vari- way. It turned out that the prime variable that determines slip is
ance S2 and standard deviation of the value of the dependent the blade solidity, c / se:
variable, may be derived, by doing a sensitivity analysis. Ac-
cording to Granger 关22兴 the relationship between the standard de- 共1 − RR兲Z
c/se =
viations, S is for the present case: 2 cos 
S2 = 冉 冊 冉 冊 冉 冊 冉 冊
S
2
+
S
2
+
S
2
+

S
2 The proposed general equation for relative-eddy-induced slip fac-
tor is: