415

THERMODYNAMICS AND FLUID MECHANICS GROUP FLUID MACHINERY GROUP

SLIP FACTOR OF A CENTRIFUGAL COMPRESSOR AND ITS VARIATION WITH FLOW RATE
A. Whitfield,
BSc, MSc, PhD (Graduate)*

T o predict the complete performance map of turbocharger centrifugal compressors, it has been found essential to include a procedure for estimating the slip factor at off-design flow rates. The many correlations available for the slip factor only concern themselves with design-point operation and consequently only provide a single value for the slip factor. The approach presented here attempts, within the confines of a one-dimensional procedure, to recognize the changing flow pattern at the impeller discharge in order to predict the variation of slip factor with flow rate. The procedure is based upon the well-established jet-wake flow model of Dean and the overall slip factor calculated from the jet slip factor. Experimental and theoretical slip factors are presented for three impellers with 34, 30 and 12 radial blades, respectively. The complete performance map of the 12-bladed turbocharger compressor and the effect of using a constant and varying slip factor on the predicted performance are presented.

1 INTRODUCTION

THE SLIP FACTOR of the centrifugal compressor has received extremely detailed attention from numerous investigators over many years. Wiesner (I)treviewed the various methods that have been proposed for calculating slip factors and included the work of more than 20 investigators. This work was, however, restricted to the design point only, as most designers have confined themselves to design-point performance prediction. I n more recent years, detailed attention has been given to predicting the compressor performance for both design-point and offdesign operation (2) (3). This work has been aimed at matching turbomachinery components to internal combustion engines (4). This demands an accurate prediction of the compressor pressure ratio and efficiency for the complete operating range of the compressor. The Euler turbomachinery equation leads to the expression

p = 1 ; : (+ 7

Y K Y - 1)

17iuzc,,- U 1 C 4

for the compressor stagnation pressure ratio. Including the slip factor defined as

leads to

for the case where there is no inlet prewhirl.

This paper is intended for presentation at an Ordinary Meeting of the Thermodynamics and Fluid Mechanics Group and the Fluid Machinery Group in London on 30th October 1974. The M S . was received at the Institution on 13th October 1973 and accepted for publication on 18th June 1974. 33 * School of Engineering, University of Bath, Cloverton Down, Bath BA2 7 A Y . t References are given i n the Appendix.
@ IMechE 1974

The pressure ratio for any given speed is clearly a function of the efficiency and slip factor. The efficiency is a measure of the entropy gain attributable to the losses occurring within the stage. The treatment used to represent these losses was presented in (2)and (3). The variation of the slip factor with the compressor flow rate has been found to be an important parameter in the performance prediction procedure for turbocharger compressors. Stahler (5) presented the variation of slip factor with flow rate for a 19-bladed centrifugal compressor and concluded that the experimental results could be approximately predicted by assuming that the relative leaving angle Fz of the impeller was constant for a given impeller, i.e. was independent of the gas flow rate and the rotational speed. Sakai (6) also presented the change of slip factor with flow rate. By studying three different impellers with 17, 10 and 7 blades, Sakai found that the rate of change of the slip factor with the flow rate increased as the number of impeller blades was decreased. However, he did not include this variation of slip factor with flow rate into his efforts to provide a theoretical correlation for the slip factor. The published information on slip factors does not include any attempt to consider the real fluid conditions that exist at the exit from the impeller. Indeed, the work of Wislicenus (7), which provided for a velocity distribution in both the hub-shroud and the blade-to-blade plane, has largely been ignored as its inclusion generally leads to a poorer correlation between experimental and theoretical slip factors. Dean (8) (9) has presented details of the flow at the impeller exit. He has pointed out that the design compromise between the impeller and diffuser inevitably leads to an area ratio across the impeller which gives a diffusion rate that cannot be sustained. Consequently, most centrifugal compressors operate with the impeller flow separated.
Proc lnstn Mech Engrs Vol 188 32/74

416

A. WHITFIELD
-x--*--r

The model suggested by Dean then considers the flow pattern after the point of separation to be approximated by a high-velocity jet, which carries the bulk of the through-flow in a potential core, and a low-velocity wake. At the impeller tip, these jets and wakes rapidly mix giving rise to the now well-established mixing loss (10). Dean gives a relative velocity diffusion ratio Wl,/ W , of 1.8as a predicted ultimate that ought to be achieved without separation and considers that present state-of-the-art impellers achieve diffusion ratios of 1.3-1.4. For the impellers presented in this paper, relative velocity diffusion ratios in excess of 2.0 are attempted and consequently separation of the flow into a jet and wake is considered to occur. The flow through the jet is then considered to be similar to the potential flow theory of Stanitz (11) and consequently his derived slip factor can only be applied to the through flow jet. Considering the wide range of pumps and compressors presented by Wiesner (I), some probably not separated at all and the others all with differing degrees of fluid separation even at the design point, it can be considered somewhat remarkable that he was able to achieve any kind of correlation for the experimental data.
1.1 Notation A Flow area. Acoustic velocity. U C Absolute particle velocity. c m Torque coefficient. D Diffusion ratio. Diameter. d h Impeller blade height. K Non-uniform velocity distribution parameter. ri2 Mass flow rate. P Pressure ratio. R Radius and gas constant. T Temperature (absolute). U Impeller peripheral speed. W Relative velocity in impeller. z Number of impeller blades. P Relative flow angle from meridional direction. Ratio of specific heats. Y Isentropic stagnation efficiency. 17 0 Jet area angle at impeller tip (see Fig. 36). Slip factor. Y V Kinematic viscosity. Density. P
Subscripts J Jet. m Meridional direction. S Impeller inlet shroud. T Impeller tip. W Wake. 6 Tangential direction. Stagnation state (with appropriate subsidiary loca0 tion subscript). 0 Upstream of compressor. 1 Compressor impeller inlet. 2 Compressor impeller exit.
2 EXPERIMENTAL S L I P FACTOR

Experimental shp factor Jet

slip

0 98

foctor

O 96)

---L --- Constant Stanitz slip foctor

i 0 91 4 b L

090i
0 88
a
14

~~

15

I6 17 16 19 Mass flow parameler. m G / P o ,

kq/sfl/bar

20

21

72 .

0 96

in 9 4 O
I

Constont Stanitz slip factor

10

II

I2 I3 14 Pass f l o w pararreter
~~

'

rn</e,
~~

IS

16

17

16

kq/svfC/bor

a Compressor A. b Compressor B.

Fig. 1. Experimental slip factors

were calculated from the experimentally-measured stagnation temperature rise across the impeller. The Euler turbomachinery equation gives

With no prewhirl, C,, = 0 and with the slip factor defined in the conventional manner for radial compressors,
Y = C,,/~2

l+(y--1)7p . . . (2) To1 a01 The measured experimental temperature rise, however, exceeds that given by the Euler equation by an amount attributable to disc friction. The contribution to the temperature rise due to disc friction was obtained using the correlation given by Daily and Nece (12)

-_

U2"

where Cmis a torque coefficient given by

1/10

c, = 3'7(i)
Re1', and

for Re
1/10

< 3 x lo5

c, =

,.lo,(

&)
Rel15

for Re

> 3 x lo5

where s is the gap between the rotating disc and the stationary casing and Re is the Reynolds number given by

The experimental slip factors presented in Figs 1 and 2

Proc lnstn Mech Engrs Vol 188 32/74

0 IMechE 1974

SLIP FACTOR OF A CENTRIFUGAL COMPRESSOR AND I T S VARIATION WITH FLOW RATE

09*r

417

090088-

090t

086L

E 084.
a \

"

Constant Stonitz slip foctor

I n

082r

080-

a
_ I

L Neqative
- -- .

0781
-

incidence 6 7

0 76

0 86
0 86
\
~

-x--x-

Experimntol d i p factoi slip foctor

-Jet

0781
I

Mass flow parameter, m a / P o , kq/sfl/bar

a 50 000 revlmin.

b 60 000 revlmin. c 70 000 revlmin.

Fig. 2. Experimental slip factors for compressor C

By subtracting the temperature rise due to disc friction from that obtained experimentally, the temperature rise attributable to the Euler work, To,-Tol, can be found. By substituting this into equation (2), the experimental slip factor can be found. The magnitude of the temperature rise attributable to disc friction cannot be precisely assessed and using alternative correlations will yield different slip factor magnitudes. For example, using the correlation given by Stahler (5) yields a slip factor of 0-8277 compared to the value of 0.8315 presented here. Assuming that there is no disc friction, a value of 0.8485 is obtained. The experimental magnitude of the slip factor clearly contains a degree of uncertainty.
3 THEORETICAL S L I P FACTOR

With an assumed theoretical slip factor, generally calculated from the Stanitz correlation p = 1-0-63n/Z the tangential component of velocity C,, can be found. The exit velocity triangle in Fig. 3a is then completely defined and all velocity magnitudes and directions can be readily found. The stagnation pressure ratio for the impeller is given by

3.1 Calculating the through-flow jet The calculation of the impeller inlet conditions was described in detail in (2) (3). T o compute the impeller exit conditions, the overall diffusion ratio D was specified (see section 5). The relative velocity can then be found from

The impeller stagnation efficiency 77 was calculated by considering the impeller losses as given in (2) and (3). Equation (3) then gives the impeller exit stagnation pressure and equation (2) the stagnation temperature. With the known absolute fluid velocity C,, the exit static conditions can be readily found. The jet flow area needed to pass the required mass flow rate was then found from the continuity equation

A.=-

CrnaPz

and the jet angle 0, Fig. 3b, calculated from

WI w2= 7
s

Q IMechE 1974

Proc lnstn Mech Engrs VoI 188 32/74

418
Without slip

A. WHITFIELD

Assume diffusion ratio

assuming flow completely fills the exit

Czlculate impeller exit velocity triangle

and efficiency

Calculate exit density and jet flow area through the jet

/ - -

Rotation

---Calculate slip factor based M jet size Calculate exit conditions assuming flow completely fills the passage

a Velocity triangle. b Jet wake.

Fig. 3. Impeller exit

The Stanitz correlation for the slip factor was then applied to the jet and the calculation repeated with the jet slip factor replacing the initially-assumed slip factor. If it were found that the area of the jet A , exceeded the total flow area available then the calculation was repeated assuming that the flow completely filled the passage, with a suitable allowance for boundary layer thickness, and the diffusion ratio D calculated. The calculation procedure is illustrated by the flow chart, Fig. 4. This procedure then provided a means of calculating the magnitude of the through-flow jet to which the Stanitz correlation for slip factor can be applied.
3.2 Calculating the jet slip factor Equation (4) enabled the size of the jet to be calculated This is necessary to determine the jet slip factor. The jet flow, which is considered to carry the bulk of the throughflow in a potential core, should be amenable to analysis by the potential flow theory of Stanitz (11). The Stanitz correlation for slip factor of a radial flow compressor

Compare new slip factor with previous value

diffusion ratio

To calculation of diffusion system

Fig. 4. Flow chart for calculation procedure

reduction in the slip factor. I n particular, when the velocity distribution in the hub-shroud plane at the impeller tip was assumed to be sinusoidal, Wislicenus showed that the slip factor was given by
p. =

I-- 8 (1-

n 2

Pth)

. .

(7)

applies when the flow is assumed to fully fill the impeller passage. When applied to the jet of Fig. 3b, which only partially fills the flow passage, the Stanitz correlation reduces to 0 p+,h = 1-0.63. . . . (6) 2 When applied to the jet width, the Stanitz procedure, however, yields a higher slip factor to that obtained experimentally. Wislicenus (7) has shown that including a non-uniform velocity distribution in the through-flow jet leads to a
Proc lnstn Mech Engrs Vol 188 32/74

wherc pth is the slip factor that would be obtained if the flow through the jet had a uniform velocity distribution, given by equation (6). It was considered unlikely that the sinusoidal velocity distribution would adequately describe the flow pattern in the jet and therefore equation (7) was generalized to
pj = 1--K(l-ptJ . . . (8) where K is a factor to allow for the non-uniform velocity distribution.

3.3 Jet-wake mixing theory for the slip factor The overall slip factor was obtained by combining the jet and the wake. Dean estimates that up to 20 per cent of the through-flow is carried by the wake. By considering the jet and wake, Fig. 3b, to mix instantaneously just outside
@ IMechE 1974

SLIP FACTOR OF A CENTRIFUGAL COMPRESSOR AND ITS VARIATION WITH FLOW RATE

419

the impeller, circumferential equilibrium can be applied across the mixing zone such that the change in the rate of the moment of momentum of the fluid is zero, leading to : kjR,(U , W, pi)+mwR,(Ut W wsin pw) sin = d?,(U,+ Wt sin pi) (9) From the velocity triangle at the impeller tip, Fig. 3a,
Ce = Ut+ W s i n p

3ve-LIl dif'usion ratio

Diffusion r a t i o alter acceleration due t o negative incidence

4r 1
23

o - - y

22
21

\\

Equation (9) can be reduced to

mat,coi+m2,c,, . . = ri?c,,
Now from the jet slip factor
c j = pjUt e

. (10)

and from the overall slip factor

ct = put o
Similarly a slip factor can be defined for the wake; however, as the magnitude of the relative velocity is low the wake slip factor was assumed to be unity. The overall slip factor is then given from equation (10) as
~ c so w _ l _ a t~ parameter, md<i,'<, _

~~

kq/sJ@'bur

Fig. 6. Impeller diffusion ratio

The overall slip factor, therefore3 deviates from the jet slip factor as the magnitude of the wake increases; this is illustrated in Fig. 5. The degree of separation and hence the wake magnitude was considered to be a function of the overall diffusion rate W,,/W, attempted in the impeller, Dean (8) considers a diffusion ratio of 1.8 as the ultimate that can be achieved without separation; however, it is likely that the fluid separated from the impellers considered here at diffusion ratios below this figure of 1-8. The computed diffusion ratios presented in Fig. 6 show that the degree of separation and hence the magnitude of the wake flowni, can be expected to be large at low flow rates and reduce as the flow rate is increased. From Fig. 5 it can be seen that the rate of change of the

overall slip factor p with the wake flow rate increases as the jet slip factor is reduced. Consequently, the rate of change of the overall slip factor with the flow rate will be greater when the jet slip factor is low (i.e. for impellers with a small number of blades) than when the jet slip factor is high (i.e. for impellers with a large number of blades). This jet-wake approach therefore qualitatively agrees with the experimental results presented by Sakai (6). This result is clearly dependent upon the assumed slip factor for the wake. If the wake slip factor were assumed to be equal to the jet slip factor, then the overall slip factor would also be equal to the jet slip factor. By combining equations (8) and (11), the overall slip factor can be obtained

This equation enables the overall slip factor to be calculated from the ideal slip factor of the through-flow jet, given by equation (6), provided that magnitudes can be specified for the parameter K and the wake flow rate proportion Istw/m.
4 PRESENTATION OF RESULTS

The results presented were all obtained using an assumed diffusion ratio D of 1.5 to calculate the jet size. The flow rate through the wake was assumed to be zero and equation (12) reduced to p = l--K1(1-,AtJ . . . (13) where the parameter Klallows for the velocity distribution and the neglected flow rate through the wake. Attempts were made to include the wake flow rate by assuming that m,/m varied linearly from zero at a diffu08 0 1 / . A p , 080 082 084 086 088 090 092 sion ratio of 1.5 to 0.2 at a specified upper diffusion ratio J e t slip factor limit, usually 2.5. This had the effect of increasing the magnitude of the ideal slip factor of the jet, as the jet size Fig. 5. Deviation of overall slip factor from was reduced, when it carried only 80 per cent of the total jet slip factor
I /
I

@ IMechE 1974

Proc lnstn Mech Engrs Vol188 32/74

420

A. WHITFIELD

flow. Applying equation (1 1) then led to a further increase in the slip factor making it necessary to use large values of K (equation (8)), of the order of two or three, to obtain agreement between the theoretical and experimental slip factors. As introducing the wake flow rate only made minor changes in the rate of change of the slip factor with the flow rate, with the assumed maximum wake flow of 20 per cent of the total, and because of uncertainties associated with the magnitude of the wake flow rate, the complete procedure leading to equation (12) was abandoned in favour of equation (13). Figs l a and l b compare the calculated jet slip factors with those obtained experimentally for compressors A and B (34 and 30 radial blades at impeller exit, respectively). Both of these compressors had a vaned diffuser and the design-point operation was controlled by the incidence of the fluid on to the vaned diffuser. Consequently, both of the impellers operated with positive incidence on to the impeller and zero incidence never occurred. In contrast, compressor C had a vaneless diffuser, the impeller had 12 radial vanes and operated through both the positive and negative incidence regimes. Figs 2a, 2b and 2c present the theoretical and experimental results for three impeller speeds. The experimental results differ from those obtained with compressors A and B in that a minimum slip factor occurred, followed by an increasing slip factor with increasing flow rate. This minimum in the slip factor coincided with the predicted onset of negative incidence on to the impeller inducer, a regime not encountered with either compressor A or B. The calculation of the jet size and hence the jet slip factor used for compressors A and B had, therefore, to be modified and could not simply be extrapolated into the negative incidence regime.The calculation of the exit jet size in the positive incidence regime was based on an assumed overall diffusion ratio of 1.5, i.e. the exit relative velocity W , was given by

10

15 20 25 30 Number of impeller blades

35

40

Fig. 7. Variation of K1 w i t h number of blades

+ a

Experimental

_-_._ constunt Predicted

slip factor 0-044

-_,

Predicted variable slip factor

2 3

. L

r-deneqative
Impeller

M a s s i i o w parameter kq/sJK/bar

rnJz/e,

5 6 M o s s flow pornmeter, m G / F &

kq/s.&/bar

..
4

~~

Impeller neqotive incidence, , 7 8 9

Fig. 6 shows that as the flow rate increases, the overall diffusion ratio attempted decreases and, when this reaches 1.5, the through-flow jet completely fills the exit channel. However, before this state is reached the negative incidence regime of the impeller is encountered and this causes a rapid increase of the relative velocity in the inducer (2) (3), i.e. the relative velocity after incidence W ' l s exceeds W , and the actual rate of diffusion attempted in the impeller increases again as shown in Fig. 6. This leads to a reduction in the jet size and consequently to an increase in the jet slip factor. The predicted jet slip factors do, however, differ greatly from those obtained experimentally and parameter Klof equation (13) must be introduced. For impeller A, a value of Kl ranging between 2.5 and 2.8 was necessary to correlate the jet slip factor with the experimental slip factor; for compressor B a value of between 2-2 and 2.5 was needed, whilst for compressor C a value between 1.4 and 1.8 was needed. Figs 2a, 2b and 2c also include the calculated slip factor based on a Klvalue of 1.5 for all speeds. The variation with flow rate is accurately predicted, although the magnitude deviates from the experimental values because a constant value of K Lwith speed was used. However, the
Proc lnstn Mech Engrs Vol 188 32/74

Fig. 8. Performance map of compressor C a t 60 000 rev/min

values predicted are always better than those predicted with the simple constant value given by the Stanitz correlation. Fig. 7 presents a plot of the value of Kl against the number of impeller blades. This gives a guide to the value of Kl needed for impellers with different numbers of blades. This, however, can only be considered a tentative result, as impellers A and B are very similar and there is then a wide gap to compressor C; intermediate points are needed before definite conclusions can be drawn. Finally, Fig. 8 compares the predicted performance of the compressor obtained assuming a constant slip factor with that obtained with the variable slip factor predicted in Fig. 2b. The experimentally-measured compressor performance is also included for comparison.
5 CONCLUSIONS

The procedure presented to calculate the compressor slip factor is an integral part of a complete performance pre@ I M e c h E 1974

SLIP FACTOR OF A CENTRIFUGAL COMPRESSOR AND ITS VARIATION WITH FLOW RATE

42 1

diction procedure and cannot be separated from the impeller incidence model and the procedure used to account for the separation within the impeller. Until the complete three-dimensional flow through the impeller can be calculated to give the point of, and degree of, separation for both design-point and off-design operation, the diffusion ratio actually achieved by the impeller flow passage must be estimated. The magnitude of the diffusion ratio used (herein 1.5) governs the size of the through-flow jet and hence the calculated jet slip factor. Using a different diffusion ratio would clearly modify the results presented here. However, the jet size is not only dependent upon the estimated velocity, but also upon the calculated fluid density, which is a function of the estimated losses incurred within the impeller. This work is part of a continuing research programme to develop a complete performance prediction procedure for radial- and mixed-flow compressors. It is intended to study the effect of diffusion ratio, velocity distribution parameter and all the associated losses on the complete performance prediction rather than on the slip factor alone. The approach presented here, which accurately predicts the change of the slip factor with the flow rate, now provides a good foundation from which this study of performance prediction can be continued.
6 ACKNOWLEDGEMENTS

APPENDIX
REFERENCES

(I)

WIESNER, J. A review of slip factors for centrifugal imF. pellers, J . Engng Pwr, Trans. Am. Sac. mech. Engrs 1967 89,
558.

This work forms part of a Science Research Councilsupported research project and I wish to acknowledge with thanks the award of an SRC research fellowship. I am also grateful for the experimental data supplied by industry, the turbocharger unit supplied by Holset Engineering Co. Ltd., and to Mr R. C . Atkey for his assistance with the experimental work at the University of Bath.

WALLACE, J. and WHITFIELD, A new approach to the F. A. problem of predicting the performance of centrifugal compressors, 2nd Int. J S M E symp. Fluid Machinery and Fluidics 1972 225 (Japan Society of Mechanical Engineers, Tokyo). (3) WHITFIELD, and WALLACE, J. Study of incidence loss A. F. models in radial and mixed flow turbomachinery Conference Publication 3 1973, Paper C55j73 (Institution of Mechanical Engineers, London). (4) WALLACE, J. and CAVE, R. Matching of high-output F. P. diesel engines with associated turbomachinery, Proc. Instn mech. Engrs 1973 187,535. (5) STAHLER, F. The slip factor of a radial bladed centriA. fugal compressor, J . Engng Pwr, Trans. Am. SOL.mech. Engrs 1965 87, 181. (6) SAKAI, WATANABE, FUJIE, and TAKAYANAGI, J., I., I. I. On the slip factor of centrifugal and mixed flow impellers, A.S.M.E. Paper No. 67-WA/GT-10. (7) WISLICENUS, F. Fluid mechanics of turbomachinery 1965 G. (Dover Publications, New York). (8) DEAN, C. On the unresolved fluid dynamics of the R. centrifugal compressors, A S M E Advanced Centrifugal Compressors 1971 (American Society of Mechanical Engineers, New York). (9) DEAN, C. The fluid dynamic design of advanced centriR. fugal compressors, Creare Inc. Technical Note 153, 1972. (10) JOHNSTON, P. and DEAN,R. C. Losses in vaneless difJ. fusers of centrifugal compressors and pumps,J. Engng Pwr, mech. Engrs 1966 88, 49. Trans. Am. SOL. (11) STANITZ, D. Some theoretical aerodynamic investigaJ. tions of impellers in radial and mixed flow centrifugal compressors, Trans. Am. Soc. mech. Engrs 1952 74,473. (12) DAILY, W. and NECE, E. Chamber dimension effects J. R. on induced flow and frictional resistance of enclosed rotating discsJ. bas. Engng, Trans. Am. Soc. mech. Engrs 1960
(2)

82, 217.

@ IMechE 1974

Proc lnstn Mech Engrs Vol 188 32/74

Discussion on A. Whitfield: SLIP FACTOR OF A CENTRIFUGAL COMPRESSOR AND ITS VARIATION WITH FLOW RATE

Discussion
F. S. Bhinder
Member

D. J. Myles Fellow

This is an interesting and important contribution on the phenomenon of slip in centrifugal compressors, The slip factor is a convenient empirical coefficient introduced in one-dimensional calculations to take into account the variations in absolute velocity vector due to the three-dimensional nature of the flow at the impeller exit. A reliable estimate of the variation of slip factor with mass flow rate is important in performance prediction calculations. But to retain the simplicity of the onedimensional procedure such an estimate must be made without resorting to a complete three-dimensional or quasi-three-dimensional analysis of flow in the impeller. The paper presents a very useful method by which the theoretical slip factor can be corrected for non-uniform velocity distribution and jet and wake mixing at varying flow rates. My queries are on equations (1 2) and (13), and Fig. 2. Substituting for K in equation (12) its value (l-pj)/ ( 1 prh)from equation (8) gives:

But uiz,

riI

rizj,

SO:

Since the assumption of uiz, = 0, i.e. rtij = rti, leads to p = pj, equation (13) seems unnecessary, unless the parameters K and K , in equations (12) and (13) respectively are supposed to be different. If K , and K are the same, there can be no advantage in using equation (13) instead of the above relationship, particularly if the values of K , must be chosen quite arbitrarily for agreement between predicted and measured results. The authors comments on this will be appreciated. In Fig. 2a, b and c the calculated and measured values of jet and overall slip factors are compared. ,It would be interesting to know how the j,et slip factor was measured. In our work at Hatfield we found that near to the impeller exit temperature and pressure measurements are very sensitive to location of the sensing instrument, besides being affected by high-frequency variations of the flow parameters in the jet and wake regions. Details of the authors measurement techniques and instrumentation would be useful. The Stanitz slip factor for impeller C from equation (5) would be 0.8351. The value shown in Fig. 2a, b and c appears to be 0.845. Although it is a minor point, the author may like to take note o this for future work. f
@ IMechE 1976

Papers on the estimation of fluid slip in centrifugal impellers have been numerous since the mid 1930s and, as the equations advanced for calculationg slip coefficient are almost as numerous, there is abundant scope for selection. This prompted Weisner (1) to compare them and, though slip can generally be accurately predicted in impellers with low backward-curved blades at the design point, its variation with flow rate and the prediction of slip in radial-bladed impellers even at the design point remains enigmatical. The author has approached these problems by recognizing that radial-bladed impellers can perform reasonably well even though the flow may be separated from the suction faces of the blades. Under these circumstances the work capacity of the impeller is more adversely affected than current theories suggest, and the variation of whirl-to-blade ratio C,/U with flow rate is far from constant. I share these views though experimentally they are hard to prove. The author has obviously experienced this difficulty, since he has chosen to measure the total temperature rise and estimate disc friction loss so as to calculate experimental C,/U values. What levels of accuracy does he put on these measurements and the calculated disc loss? Also, from what number and location of temperature readings was the mean assessed? These questions are particularly relevant, as extrapolations of the linear portion of the experimental slip factor curves for compressor B (Fig. lb) intercept the m = 0 line at above unity; a situation I have not previously met. There are also considerable variations in the curves of compressor C (Fig. 2) when reduced to a common speed (Fig. 9). The high-valued and variable coefficients needed to reduce theoretical C,JU values to coincide with experimental results are of concern. Rewriting equation (13) as

implies that, when the curve is linear (Fig. l a and b) K , / C m , is constant and the intercept on the p axis is unity. Curves of Fig. l a and those shown in (13) suggest this intercept is less than one and is in agreement with the well-known Busemann relative eddy theory which is not flow-dependent. If the relative eddy effect and that caused by boundary layer growth are regarded as separate but interdependent effects on the value of slip coefficient, the expression in (13) reduces to
Kp
D95

cm2 pCL,,--tan6

U2

Proc lnstn Mech Engrs Vol 188 32/74

D96

DISCUSSION ON A. WHITFIELD
1
I

0.90,

x 50 000 rev/min
0 60000 revlmin 0 70 000 rev/min

0.86?08 :.4
m

As insufficient data are available in the paper to evaluate this expression could the author please make this comparison with his experimental values to see if better correlation obtains?
REFERENCE

xo
0 0

0
0

(13) MYLES, J. 'An analysis of impeller and volute losses in centriD. fugal fans', Proc. Instn mech. Engrs. 1969-70 184 (Pt 1) (14). 253.

3 0.82U

L-

2
3 0.80Q

F. 0 . J. Otway Fellow In Fig. 1 of the paper I assume the different lines are for different speeds, and request confirmation of this with the range of speeds covered. In my experience much better correlation of slip factor with speed comes from plotting against flow coefficient 6,, defined as:
1.0

00 0.2 0.4 0.76 '< 7

0.6

0.8

12 .

14 .

16 .

1.8

6, =

cv2
XND;B,

Mass flow paraJpara. a t zero incidence

Fig. 9. Experimental slip factors for compressor C plotted on a common speed basis

for radial blades, where tan 6 is the slope of the curve, which may or may not vary in the useful range of the impeller according to whether the incidence changes is sign. pLth the relative eddy effect corresponding to an effective blade number Z given by 2' = 2n/$. Using the modified Stanitz equation (6) and making the approximation tan 6 = R {(27c/2)- B}/(R, R for separation starting at the leading edge, p reduces to
~

The form of this equation implies that, as separation increases, 8 decreases and the relative eddy effect becomes smaller owing to the reduced main flow volume. However, because of separation, the slope of the curve becomes more negative leading to a net reduction in pa result to be expected logically.
1

where C = constant (taken as 310.4), V, = volume flow at impeller outlet (ft3/min), N = rotational speed (rev/min), D, = impeller diameter (in) and B, = blade width at outlet (in). This coefficient is a constant times the ratio of radial velocity to blade tangential speed. Fig. 10 illustrates this correlation with varying speed; both Figs. 10 and 11 are for radially bladed impellers with vaneless diffusers, for comparison with compressor C of the paper. Fig. 10 is based on results from the N.G.T.E. no. 13 impeller tested with a vaneless diffuser (14). The slip factor has been calculated assuming a power input factor of 1.04. Disc friction has been ignored. , Note that it has a linear relationship with 6 , falling continuously over a range comparable with that shown for compressor C. This compressor impeller had 29 blades and a diameter of 20.7 in (52.6cm). While this compressor was associated with jet engine development so maximum pressure ratio was desirable. I believe many centrifugal compressor impellers have too many blades to achieve the highest efficiences. This particular impeller when operating with a vaned diffuser showed a maximum efficiency of 81 per cent while 83 per cent has been obtained with a 6.5 in (16.5 cm) diameter impeller having only 17 blades.

.o-

0 8L
"t

KEY f t lscc.

$
' c

0
0.6-

cp

Q .-

0.4-

1000 1127 1200 1298 1412 1493

I3

0
V
Q

0.21

0 0.10

0.11

0.12

0 13

0.14

0.15

0 16

0.17

0.18

0.19

0.20

6 2

Fig. 10. Slip factor vs flow coefficient; N.G.T.E. no. 13 impeller in vaneless diffuser
Proc lnstn Mech Engrs Vol 188 32/74

0 IMechE 1976

SLIP FACTOR O F A CENTRIFIJGAI, COMPRESSOR A N D ITS VARIATION WITH FLOW RATE

1.0,

D97

0 1

04

jI
0

0 05

0 10

0 15

020

62

Fig. 11. Slip factor vs flow coefficient; gas circulator for research reactor

Fig. 1 1 shows the results for a relatively small highspeed and high-temperature gas circulator running on carbon dioxide. The impeller was 154 in (38.7 cm) diameter and had 23 vanes. Blade outlet width was 0.440 in ( 1 1.2 mm) and axial clearance 0.060 in (1.52 mm), which may be considered high, but this circulator operated at 475C. Power input factor was taken as 1.06 and again disc friction was neglected. The slip factor shows a levelling ofT above a flow coefficient 6, of 0.14. Possibly this too would have shown a rise at higher flows as observed with compressor C of the paper. The author states that the work reported is part of a continuing research programme. However, from the information presented the methods propounded do not seem as yet to give sufficient accuracy for predicting centrifugal compressor performance direct from the drawing board. This still leaves the designer dependent on empirical formulae based on earlier test results. Does he agree with this statement? Has he plans to refine the methods described t o give more accurate prediction'!'
KP;FlCKIiNCl;

are trapezoidal and that angular momentum is conserved. Theoretical distribution is uniform and that found experimentally is shown to be very nearly so. In Fig. 2c at a mass flow parameter of 7 the calculated slip factor is 0.82. On either side of 7 the slip factor is higher so, if a scroll having a non-uniform distribution were used, the average slip factor would be increased to about 0.85. Flow distributions were measured by pitot-yawmeter and hot-wire anemometer axial traverses made at a radius equal to the rotor tip radius but with the rotor absent. In practice the revolving rotor would partly smooth out flow non-uniformities by propagation of pressure waves in the rotor channels, thus creating a nonuniform static pressure distribution at the rotor tips.

-0-e-

- Theoretical
Experimental

(14)

SIMONS,

A . and S 1 A K M l : K . R. 'Tests on power jets no. 13: impcller

fitted in thc diffuserless blower casing', N.G.T.E. report R12.

R. D. Pearson Bath
The paper presents some useful and interesting results. I confine my remarks to the non-linear variation of slip factor with mass flow shown in Fig. 2 for compressor C having a vaneless diffuser, and speculate upon .the possible effect of design of outlet scroll in such cases. I have no results from compressor scrolls to present but, instead, summary curves showing peripheral flow distribution of inlet casings for radial flow turbines (see Figs. 12 and 13). The distributions for compressor scrolls designed by thr same rules ought to be very similar. In Fig. 12 the distribution measured for a standard 01 turbocharger casing is compared with a theoretical calculation based on the nominal geometrical form. This scroll was based on a design rule assuming linear variation of cross-sectional area with azimuth angle. The measured distribution has a similar shape to that predicted and both are extremely non-uniform, having a peakto-peak variation of almost 3:l. Fig. 13 shows a distribution for a casing based on improved design rules which assume all cross-sections
@ IMechE 1976 j )

0.6

\
\

' ,
O\
\*

0.4

d b-'

o'2

0 1 "

90

'

"

"

180

270

"

360

degrees

Fig. 12. Flow distribution in a standard volute

Proc lnstn Mech Engrs Vol 188 32/74

D98

DISCUSSION ON A. WHITFIELD

It is easily shown that a trapezoidal volute designed according to the rule described is specified for incompressible flow by the equation

A correction is needed for the thickness of the tongue end by adding to the outer wall a radius increment Ar where
Ar
=

t , ("l-RB)@) ____
RO-RB

. (22)

where

RB = R,-

WN/(2tanP)

. . (15)
=

and for compressible flow by

((%/aO)

This allows for both trapezoidal shape and conservation of angular momentum. The correction is blended to zero to meet the calculated scroll form over an azimuth angle of 30 degrees. Notation Gas flow angle at rotor tip radius relative to circumferential direction and based on W. B Half-angle of trapezoidal cross-section measured from median plane. 0 Azimuth angle at radius R measured from start of scroll. R Inside wall radius of scroll outer wall measured from rotor axis. R Value of R at 0 = 0. O Rl Value of R at 0 = 2n. R B Base radius, i.e. radius of intersection of sides of trapezoid. Tip radius of rotor. Tip width of rotor. Projected width of trapezoid at RT (less than Wto allow for circumferential corner radii). t2 Tongue end thickness. Density ( p o is upstream stagnation density). P Sonic velocity when u = 0. a 0 Absolute gas speed. U Y Ratio of specific heats.
a

(~/a~a cos )

cos aT - _ "B) tan a RT

Cos a can be assumed 0.99 for first approximation and a found for specified value of u/ao, whence
r/RT = (uT/aO) cos @T (u/ao)cos CI If a linear variation of tan CI with r is assumed for small finite 0 increments between specified values O1 and O,, then for intermediate values of 0 we have
fI-6,
=

fn(?--)

A+Br, A + Br,

liA

. . .

(18) (19)
. (20)

where
A =

t a n a = A+Br . . tan a, - ( r , / r , ) tan a,

1-(r1/r2) tan a,-tan a, r2-rl

and
1.6

B =

. . .

(21)

/
/

/I
1.4-

P-.
', 0
'\

,
P
\
\

Subscript Measured at rotor tip radius.

/,a,
\

LD
1.2

a\

-'

% . +'

\
\

'0T. Sakai Tokyo

cm
VO

0.8 -

c,

Mean r a d i a l v e l o c i t y = Ideal r a d i a l velocity

by i n t e g r a t i o n o f a x i a l traverses

0.6-

0.4 -

0.2-

07

"

90

180

270

Fig. 13. Flow distribution in a trapezoidal volute

Proc lnstn Mech Engrs Vol 188 32/74

The paper gives a good explanation for the variation of slip factor with flow rate. Slip factor is normally defined under the assumption of inviscid flow. In the circumstances studied however, flow within the impeller can be taken as inviscid when flow rate through the wake is assumed to be zero-because in this case we can consider the blade to be extended to occupy the whole area of the wake. The assumption that the area of the potential core is a function of flow rate, and that the rate of change of the potential core with flow rate is greater for impellers with few blades, seems quite appropriate. In our own experimental study of radially bladed impellers, the lowest value of slip factor was found to occur in the neighbourhood of zero incidence. This result coincides with the author's theory and experiments. Determination of the parameter K , seems very difficult because not only the diffusion factor but also impeller geometry may influence it. The author's efforts to evaluate slip factor at offdesign flow rates is indeed much appreciated.
0 IMechE
1976

SLIP FACTOR OF A CENTRIFUGAL COMPRESSOR A N D ITS VARIATION WITH FLOW RATE

D99

M. J. C. Swainston Member
Absolute 0.9 v e l o c i t y 0.8 C/CmCix

16 R e l a t i v e 15 e x i t

This is a most interesting paper. Variation of slip factor at off-design conditions seems to have had too little attention in the past. This limiting of the field of study has doubtless contributed to the uncertainty still surrounding its evaluation at the design point. Some years ago I initiated work on this topic in connection with centrifugal pumps. The pump studied was a double-entry machine of standard commercial type rated at 400 gal/min (1800 litre/min) and 40 ft (12.2 m) head rise at a rotational speed of 1410 rev/min. A inch (9.5 mm) diameter cylindrical yawmeter was fitted parallel to the machine axis at impeller exit and this was traversed across the full width of the volute. Results at the design flow coefficient are given in Fig. 14, and show clear evidence of the boundary layer shed by the impeller hub section. The relative flow angle shows little variation across the section. Fig. 15 shows slip coefficients calculated from

Fig. 14. Flow at exit from impeller of a centrifugal pump

0 04

0 08
Flow coefficient

0 12

0 16

Fig. 15. Slip factor variation in a centrifugal pump

cp cot a2 1-cpcotp2 where 4 = C,,/U, and a, and p2 are respectiv~elythe absolute and nominal relative impeller exit angles measured from the circumferential direction. In this form the slip coefficient is defined as the ratio of the actual absolute whirl velocity to the whirl velocity based on the nominal impeller exit angle. For radial blades (p2 = 90) this becomes identical with the authors definition. These results confirm the trend he found for compressor C, to the extent that the curve has a minimum. However, his suggestion that the minimum slip coefficient relates to zero incidence for the inlet section is not supported by my data. Also shown in Fig. 15 is the slip coefficient obtained from relaxation solutions of the two-dimensional equations for flow in the meridianal and blade-to-blade planes, after the Stanitz methods (11). This does not agree closely with the measured value. Note that the slip coefficient at the design condition corresponds closely to the value 0.691 recommended by Stepanoff (15) for pumps. Fig. 16 shows the variation of relative exit angle over the flow range. It is far from constant and approaches the metal angle only at very high flow coefficients. This appears to show that Stahlers claim (5)-that the relative exit angle is virtually constant at all flow conditions in centrifugal compressors- cannot be applied to pumps.
=
REFERENCE

(15) STEPANOFF, J. Centrifugal and axial flow pumps, 1957 (John A. Wiley and Sons).

1. Watanabe Tokyo

Flow coefficterit
Fig. 16. Relative exit angle variation in a centrifflgal pump

This paper is interesting in that it adopts the concept of jet-wake region within the impeller channel in deriving the slip factor of centrifugal compressors. Recent experimental research on flow patterns within centrifugal impeller channels at Keio University has revealed that, for an impeller with straight radial blades, the wake domain or low-energy region exists at the shroud/ suction side of the impeller channel irrespective of
Proc lnstn Mech Engrs Vol 188 32/74

0 IMechE 1976

DlOO

DISCUSSION ON A. WHITFIELD

impeller configurations (16). Thus, the jet-wake concept is applicable on the stream surface near the shroud, but not on that near the hub, because of the absence of the wake region so far as the domain near the impeller exit is concerned. So I suggest to the author that better correlation will probably (if not certainly) be obtained by adopting the above assumption. Apart from this, I have derived (17) a forrriula for the slip factor of centrifugal impellers having straight radial blades as follows

where rz-ro tan c! l+b, p being the slip factor, a the apex angle between shroud contour and hub contour of the blade on the meridional plane, z the number of blades, b, the blade height at impeller exit (subscript 2 denoting impeller exit con-

dition), r2 the impeller exit radius, o the angular velocity of the impeller, V, the volume flow at impeller exit and ro the outer diameter of the inducer inlet. In deriving this formula, I also used a concept of the wake region (relative flow velocity being zero) and the jet region, though the jet-wake concept was not so established in those days. In addition, I used the method adopted by A. Stodola in deriving his formula for slip factor. If the volume flow at impeller exit V, is not known, use of V,, the volume flow at impeller inlet, leads to negligible error. Would the author therefore, if possible, compare the above formula with his experimental results and with his own formula?
REFERENCES

(16) MIZUKI, ARIGA, and WATANABE, S., I. I. Investigation concerning the blade loading of centrifugal impellers, A.S.M.E. Paper 74-GT-143. (17) WATANABE, the type of impeller of a centrifugal blower with 1. On special references to the problem concerning the number of impeller blades, Proc. Faculty Enging (Keiogijuku Univ.) July 1949 2 ( 5 ) , 49-71.

Proc lnstn M e c h Engrs Vol 188 32/74

@ IMechE 1976