You are on page 1of 8

THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS

345 E. 47th St., New York, N.Y. 10017

93-GT-101

The Society shall not be responsible for statements or opinions advanced in


papers or discussion at meetings of the Society or of its Divisions or Sections,
or printed in its publications. Discussion is printed only if the paper is published in an ASME Journal. Papers are available from ASME for 15 months
after the meeting.
Printed in U.S.A.

Copyright 1993 by ASME

AERODYNAMIC DESIGN AND ANALYSIS OF


VANELESS DIFFUSERS AND RETURN CHANNELS

Ronald H. Aungier
Product Development
Elliott Company
Jeannette, Pennsylvania

L - passage length

ABSTRACT

Procedures are presented for the aerodynamic design and


performance analysis of vaneless diffusers, crossover bends and
return channels. The design of the crossover bend and return
channel is formulated to permit a computerized interactive design
system, which has provided dramatic improvements in both
quality of designs and engineering productivity. Mean streamline
performance models are employed to fully support the interactive
design system. These performance models are qualified against
experimental data from several centrifugal compressor stage tests.
A recent development program which used these procedures is
reviewed to demonstrate their benefits.
NOMENCLATURE

A - area
A R - area ratio
a chordwise location of the point of maximum camber
b - passage hub-to-shroud width
C - velocity
c - vane chord length
cr - skin friction coefficient
C o, - meridional velocity component
C, - tangential velocity component
D - passage divergence parameter, equations (5) and (7)
d - minimum metal thickness of return channel, figure 8
d H - hydraulic diameter
E - diffusion efficiency, equations (9)
h - gas enthalpy
l c - curvature loss term, equation (13)
I D - diffusion loss term, equations (10) and (12)
i - vane incidence angle
K - blade loading parameters
K B - area blockage factor (fraction unblocked)
-

LC - total pressure loss coefficient (based on component


inlet velocity pressure)
m - meridional coordinate
M o - rotational Mach number (impeller tip speed divided by the
stage inlet total sound speed)
n, - dimensionless specific speed (Babe, 1981)
P - pressure
R - mean streamline radius of curvature
- radial coordinate
t - vane thickness
- gas specific volume
w - mass flow
x - distance along vane chord line
z - number of vanes
a - flow angle with respect to tangent
- vane inlet flow angle at minimum loss condition
- blade angle with respect to tangent
O" - vane deviation angle at minimum loss condition
0 - vane camber angle
B o - equivalent divergence angle, equations (5) and (43).
0 7 - vane trailing edge angle with axial direction
q5 - angle of C o, relative to the axis of rotation
a - vane equivalent solidity
Subscripts
av - average value or value at mid-chord
c - crossover bend parameter
h - parameter on the hub contour
m - a maximum condition
o - exit turn bend parameter
s - parameter on the shroud contour
t - total thermodynamic condition
1 - passage entrance parameter

Presented at the International Gas Turbine and Aeroengine Congress and Exposition
Cincinnati, Ohio May 24-27, 1993

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/22/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

are modeled by a classical diffuser analogy. The data of Reneau


et. al. (1967) shows the low loss regime can be identified by the
divergence parameter

4 - diffuser exit parameter


5 - crossover exit parameter
6 - return channel vane inlet parameter
7 - return channel vane exit parameter
8 - return channel passage discharge parameter

D = MA R

(5)

1) 1 L = 2tane,

where diffusion losses are low for values of D less than


D m = 0.4(b 1 / L)35 (6)

INTRODUCTION

The aerodynamic performance of centrifugal compressor stages


is strongly influenced by the internal flow and losses in the
stationary components. The return system stands out as
undoubtedly the least understood component of the centrifugal
compressor stage. The presence of the upstream 180 annular
crossover bend, and the interaction of the flow through the bend
with the downstream vane row, present a formidable fluid
dynamics problem. The literature offers little information
relative to specific design procedures or performance prediction
models for this component. Most published experimental and
theoretical investigations are directed toward developing a better
understanding of these complex flows (e.g., Davis, 1976, Fister,
et. al., 1982, Japikse and Osborne, 1982, Nykorowytsch, 1983,
Aungier, 1988a). Return system design has largely consisted of
applying basic engineering judgement to control diffusion rates,
passage curvatures and other design parameters, based primarily
upon empirical categorization of performance from past
development activity (e.g., Hohlweg, 1987).

Equation (6) is an empirical fit of the data in figure (8b) of


Reneau et. al. (1967). The analogy for the vaneless annular
passage used is

The present paper presents aerodynamic performance analysis


techniques for general vaneless annular passages and return
channels, including comparison with experimental data. Specific
aerodynamic design procedures for vaneless diffusers and return
systems are presented.

v dC
dip = -2(P, - P)(1 - E)

dm C dm

b dC
C dm

E=1;DO
E = 1 - 0.2(D / D,,) 2 ; 0 DD m
E =

scP
Cin
r

dc

m dm

h, = h +

1D;D

(10)

In addition to this streamwise diffusion loss term, an excessive


meridional gradient of the passage area can cause higher losses.
Again, a diffuser analogy is used to check for this situation at
each computing station. The maximum, stall-free, local area is
estimated by
(rb) m = (rb) 1 [1 + 0.16m / b 1 ]

which is equivalent to a diffuser divergence angle, 28 c , of 9 . If


the local area exceeds this value, a second estimate of the
diffusion term is generated

(1)

Ip

= 0.65v(P, - P)(1

r s.b

(12)
)

If this value exceeds the local value obtained by integrating


equation (3), it replaces that lower value. The passage curvature
loss term is given by

(13)
/c. = v(P, - P)C, 1 (13RC)

cICC, dip

dm

(9)

The diffusion term is given by

d(rC )
bC,,, dm" -

(8)

The flow angle term in equation (8) is an empirical factor derived


from comparisons of predicted and measured loss data from over
35 compressor stage tests. Based on this same comparison, an
empirical diffusion efficiency model was formulated as

Experience gained while predicting vaneless diffuser and


crossover bend losses with the author's three-dimensional analysis
(Aungier, 1988a) has provided insight into loss mechanisms to
permit formulating an accurate mean-streamline performance
analysis. The governing equations are

vdp
dm

(7)

D,,, = 0.4(b 1 1 L)35sina

VANELESS PASSAGE PERFORMANCE

21crbCif / v = w

Equation (13) was developed empirically from comparisons of


predictions with test data for 35 different vaneless diffuser/return
system combinations. This term has negligible effect on vaneless
diffuser performance, but is always significant (and sometimes
dominant) for crossover bends. Once the accuracy of the analysis
was established for vaneless diffusers, equation (13) was
developed to extend the analysis to crossover bends. Its validity
is further supported by the successful use of this same analysis in

C2

Except for the last two terms in equation (3), this set of equations
is conventional for mean-streamline analysis (e.g., Johnston and
Dean, 1966). The additional terms address loss contributions due
to flow diffusion and passage curvature. Flow diffusion losses

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/22/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

angle is equal to the flow angle in the vane throat. Since the
return channel analysis does not use area blockage directly, the
minimum loss incidence angle is corrected to account for the
entrance blockage to yield
The
area
blockage
factor
and
the
skin
friction
coefficient
are
(16)
tana*
=
KBtan[sin 1 (A, / A 6)]
computed using a simple boundary layer growth model, based on
where A, is the vane throat area. The incidence loss coefficient
a 1/7th power law for the boundary layer velocity profiles. The
is given by
analysis requires specification of an inlet boundary layer thickness
(typically supplied by an impeller performance analysis).
LCin, = 0.8[1 - C ii6 / (C6sina")] 2 (17)
Boundary layer growth is estimated from the change in angular
Skin friction loss is computed by
momentum predicted by integration of the governing equations.

The
boundary
layer
thickness
is
used
to
compute
the
blockage
(18)
LC
D = (4Lcf cl,)(Cm, C6) 2
factor. Skin friction is computed from an empirical correlation
where C., is the average of the discharge velocity and either the
of classical fully-developed pipe flow friction factors, including
inlet or throat velocity (whichever is larger), d H is the average of
laminar, transition and turbulent flow as well as surface
the throat and discharge hydraulic diameters and c f is computed
roughness effects (e.g., Schlichting, 1979). Pipe flow friction
from the correlation discussed previously. The average
factors are also valid for boundary layers if the pipe diameter is
blade-to-blade velocity difference is computed from the vane
replaced by twice the boundary layer thickness. Hence, the
circulation
thickness of the two boundary layers is used in place of the

passage width to compute the friction factors. This leads
to
(19)
2n(r6C.6 r7 C. 7) I (a)
AC
=
higher values for cases where the boundary layers do not fill the
and the blade loading loss coefficient is given by
passage, which is quite significant for high specific speed stages.

(20)
LC B, = [AC / (6C6)] 2
This performance analysis is used for vaneless diffusers,
The maximum vane surface velocity is estimated assuming it
crossover bends and other vaneless annular passages. Its validity
occurs at mid-passage for a mid-loaded vane
has been established for a broad range of stage specific speeds

(21)
and operating conditions, including cases where D is far in excess
= 0.5(C6 + C7) + AC
of D,. Use of velocity pressure (P, - P) rather then velocity head
and C m > C6 is required to include the more common case where
as a loss coefficient base, insures applicability to a broad range
the return channel vane maximum surface velocity is the inlet
of Mach number levels. The analysis has been qualified against
value. When C m > 2C 7 , it is assumed that the boundary layer
test data for Mach number levels up to 0.85 with no observable
will separate at a velocity, C, = C m /2. Otherwise, the velocity
influence of Mach number level on prediction accuracy. But, it
when the boundary layer separates from the blade is set to C7.
should be noted that effects due to differences in inlet flow profile
The velocity after wake mixing is estimated from
distortion and turbulence intensity levels are not specifically

modeled by this analysis.
(22)
C
o = V(C7A 7 / A0)2 + C!7
overall stage performance prediction for other curved passages
such as the stage inlet passage and the exit turn from return
channels.

where A7 includes the vane metal blockage, while A, does not.


The wake mixing loss coefficient is given by
= [(Csp - Co) / C6] 2 (23)

RETURN CHANNEL PERFORMANCE

-The return channel vane passage performance shares many


common features with the author's vaned diffuser analysis
(Aungier, 1990), both of which are adaptations of a mean
streamline impeller flow analysis. A significant difference is the
treatment of incidence losses, which are strongly influenced by
flow distortion imposed by the upstream crossover bend. Two
estimates of aerodynamic area blockage factors at the vane
entrance are made, and the smaller value is used.
-

1
1 + g

KB =

(12R)

(r6b6)

The loss coefficient due to the exit turn into the eye of the next
stage is given by

(24)
LCo =
/ C6)2
which is derived from equation (13) for a 90 turn, assuming a
constant velocity. Our analysis also includes the choking loss of
Aungier, 1990. But, since the author has never seen choke in a
return channel, it will not be repeated here. The flow discharge
angle is computed from the transformed axial flow compressor
deviation angle model reported in Aungier (1990). The position
of the point of maximum camber, the vane solidity and camber
angle are estimated using the vane camberline vane angle at
mid-chord, O.,.

(14)

(15)

Equation (14) is a simple inviscid flow estimate using the average


radius of curvature, R, of the crossover bend. Equation (15)
estimates blockage due to stall, based upon equation (11). The
Minimum incidence loss is assumed to occur when the flow inlet

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/22/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

(13 , - Pe)
]
3 (P7 - Po)

Li

a -

parameter defined in equation (43) and evaluated at the crossover


discharge. The return systems considered represent both fairly
modern good design practice and obsolete designs containing very
undesirable features. The multiple integrations required, and
their use in the ratios of small differences to define the loss
coefficients, result in considerable data scatter. However, the
agreement between prediction and test data is considered
satisfactory to support both design and application activity.

(25)

z(r6 / r7 - 1)

(26)

2nsink,

(27)

= P7 - P6

RADIAL VANELESS DIFFUSER DESIGN

and the reference deviation angle is based on Howell (1947)


8*

6[0.92(a / c) 2 + 0.02(90 -

p 6)

The radial vaneless diffuser passage construction is constant


width, parallel wall style with a width adjustment in the first
15-20% of the passage to blend to the impeller tip width.
Specifications required are the inlet and discharge radii and
widths and the blend radius at the start of the constant width
section. A second-order polynomial variation of width with

(28)

FT - 0.026

Off-design incidence effects on flow deviation angle are included


by an empirical correlation of figure 177 of Johnsen and Bullock,
1965.
= exp[((1.5 - p, / 60) 2 - 3.3)0]

di

a77 = P7

(29

da

8 - (P4 - ar)

di

rn
Ho -

The analysis is a simple iteration procedure, computing the


losses and fluid turning while balancing mass at the discharge,
until convergence is achieved.

I-4
C.)

44
410
0

TABLE I: COMPRESSOR STAGE DESIGN DATA


STAGE

n.

M.

b./r4

b./b4

b./R

A #1

0.776

0.70

0.081

1.448

0.922

5.20

A #2

0.776

0.70

0.072

1.767

1.995

15.41

0.854

0.88

0.058

1.444

1.453

8.09

0.272

0.50

0.011

3.064

0.717

8.74

0.699

0.70

0.049

2.338

1.747

17.65

1.110

0.70

0.097

1.294

0.913

3.78

0.-

20cs


O TEST DATA
0 - PREDICTION
0

10

15

20

25

30

35

DIFFUSER EXIT FLOW ANGLE - degrees


FIGURE 1: VANELESS DIFFUSER PERFORMANCE - COMPRESSOR

40
0

A41"

COMPARISON WITH EXPERIMENTAL DATA

Validation of the performance models was accomplished for


over 35 individual centrifugal compressor stage development
tests. These tests include traverse probes at the impeller tip,
vaneless diffuser exit and stage exit. These data are integrated
and mass averaged to permit calculation of the component loss
coefficients (where the crossover bend and the return channel
vane section are treated as a single component). Figures 1 to 7
are representative of the correlation between prediction and test
data for vaneless diffusers and return systems. As seen from
Table I, these cover a range of stage specific speeds and
rotational Mach numbers. The last column in Table I is a design

0 TEST DATA
0 - PREDICTION
40
35
30
25
20
15
DIFFUSER EXIT FLOW ANGLE - degrees
FIGURE 2: VANELESS DIFFUSER PERFORMANCE FOR COMPRESSOR "B"
0

10

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/22/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

radius is used at the entrance to match the two specified widths


and the zero gradient condition at the blend radius. Normal
practice is to use a radial hub contour with all width adjustment
imposed on the shroud contour. This is believed to be more
effective in improving the impeller tip flow profiles, which often
tend to deteriorate toward the shroud wall.

O TEST DATA
- PREDICTION

Design optimization involves selecting the discharge width and


the blend radius with the other specifications viewed as design
constraints. Loss curves similar to figures 1 and 2 are used to
evaluate the alternate choices of these parameters. An interactive
design program could be used for this purpose. But, direct use
of a mean streamline centrifugal compressor performance
program is more convenient, since impeller discharge flow
predictions are provided directly.

0-12.0

-4.0

-8.0

0.0

4.0

8.0

12.0

12.0

16.0

RETURN CHANNEL VANE INCIDENCE - degrees


FIGURE 5: RETURN SYSTEM PERFORMANCE FOR COMPRESSOR "B"

O TEST DATA
- PREDICTION

O
0

- PREDICTION
0 0 TEST DATA
0 50

60

FIGURE 3:

70

80

VANELESS

90

m
.::;

100 110 120 130 140

-6.0

0.0

8.0

4.0

RETURN CHANNEL VANE INCIDENCE - degrees


FIGURE 6: RETURN SYSTEM PERFORMANCE FOR COMPRESSOR "A42"

% DESIGN FLOW
DIFFUSER PERFORMANCE FOR COMPRESSOR "C"

O"

z o

zH 0

I-1

L . V)

ra,r
W

O r
W
0


0
0v

0
0

0 TEST DATA
- PREDICTION

-12.0

-6.0

-2.0

2.0

Cal
0

47O'

6.0

10.0

O TEST DATA
- PREDICTION
-8.0

-2.0

2.0

6.0

10.0

14.0

18.0

RETURN CHANNEL VANE INCIDENCE- degrees


FIGURE 7: RETURN SYSTEM PERFORMANCE FOR COMPRESSOR

RETURN CHANNEL VANE INCIDENCE - degrees


FIGURE 4: RETURN SYSTEM PERFORMANCE FOR COMPRESSOR "A#1"

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/22/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

"D"

RETURN SYSTEM DESIGN

A = - cotp 6 - B - 2C - 3D

Figure 8 illustrates the construction of the return system gas


path geometry. The 180 crossover bend is constructed with a
circular-arc hub contour and an elliptical-arc shroud contour.
The 90 exit turn uses a similar construction, but with the arc
types reversed for the hub and shroud contours. Straight line
contours are used for the return channel vane section, blended to
the crossover and exit turn contours by circular arcs with arc
radii equal to the local bend contour elliptical-arc axial semi-axis
or circular-arc radius. The vane camberline used is an adaptation
of the author's vaned diffuser camberline (Aungier, 1988b).
Leading and trailing edge blade loading parameters, 1( 6 and K7,
are specified to define the vane
camberline by
y
=
y r r6 (30)
cote =

By

2Cy 2

These equations apply to radii greater than the vane discharge


shroud radius, r7s . Below r 78 , a constant blade angle camberline
is used. The vane thickness distribution used is given as a
function of the distance, x, along the chord line between I -, and
r7s . Below r 7s , the vane thickness is held constant. The
thickness distribution is given by
t = to +

x / xm ; for x >

y = (1 - x) 1 (1 -

(38)

(39)

(40)

; forx> x1 ,

(41)

- 2)

(34)

3(Y - 1) 3

(cotp 6 - coti3 7 )(K7 - K6) 9


+
4(Y - 1) 2 4

+ I)

[0.95(1 + /x1

+ 0.05]

(42)

The blade maximum thickness, t,,, and its location, x,/c, are
specified, as are t 6 and t 7 . Figure 9 shows a typical return
channel vane designed with these equations.

(33)

Y = r7, / r6

C-

3Dy 3

0 = Aln(y) + B(y - 1) + C(y 2 - 1) + D(y 3

(cot0 6 - cot0 7 )(k6 +

- to)y e

to = rb + (t7 - t6)x /

D-

(37)

With reference to figure 8, the design constraints imposed on


the return system design are r b 4 , and the diffuser exit flow
angle (from the vaned or vaneless diffuser design). To complete
definition of the crossover contours, the designer must specify the
following

(35)

1) either b 6 or the crossover exit flow angle


2) either r 4 or (A,, - A 4 )/(A 5 - A 4 )
3) either R, or the average b/R over the bend

B = K6 (cot13 6 - cot(3 7) / (Y - 1) - 4C - 9D (36)

Axial Distance
FIGURE 8: RETURN SYSTEM GEOMETRY

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/22/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

where A., is the passage area mid-way through the bend and b/R
is the ratio of passage width to mean streamline radius of
curvature. The alternate choices provide direct control over vane
incidence, the passage area distribution and the distribution of
b/R. Their availability greatly accelerates the design process.

were used to design the stationary components for this stage.

Return system losses achieved are about 15% lower than our
standard designs. This is equivalent to a gain in overall stage
efficiency of over 1 percentage point. Within the usual test data
scatter, both the vaneless diffuser and the return system
performance are in good agreement with prediction.

With reference to figure 8, the designer specifies the return


channel gas path contours by supplying r 6 , r8h, rss , bs /R0 , d, and
the area ratio across the 90 exit turn. The parameter, d, is
useful since it controls the minimum metal thickness in a cast
return channel. This parameter sets the minimum axial length
achievable with the designer's other specifications. Design of the
vane requires supplying the number of vanes, the vane trailing
edge angle (0, of figure 8), t6, t7, t,,,, K6, K7, X,/e, 136 and 13 7 .

DESIGN RECOMMENDATIONS

Vaneless diffuser design optimization normally leads to diffuser


discharge flow angles in the range of 30 to 35 , where
component loss coefficients approach a minimum and a good
stable operating range can be expected (e.g., see figures 1 and 2).
But, each choice of blend radius and discharge width produces a
different loss curve. Only an investigation of alternatives can
identify the choices best suited to the design objectives. This is
very significant for low specific speed stages where loss curves
for different passage widths can be dramatically different.

Note that the entire gas path displayed on figure 8 is defined by


only 12 parameters, which provide direct control over the key
area and curvature distributions. The vane requires 10 more
parameters which directly control the vane passage area
distribution and blade loading style. The vane and all contours
are defined analytically. This reduces the problem to a form well
suited to a computerized interactive design system. Our design
system allows the user to interactively modify the design
parameters while supplying the following support functions on
demand

Crossover bend and exit turn design should control the ratio of
passage width to mean streamline radius of curvature, b/R.
Local values of b/R should be less than 1. An average over the
bend less than 0.8 is preferred. Based on experimental loss data
for numerous return systems, the passage equivalent divergence
angles, 20,, should not exceed 9 , where
20 c = 2tan l [k(A / A 4 1) / (m m4) /2] (43)
-

1) Screen displays of gas path geometry (figures 8 & 9).


2) Screen displays of graphical and tabular summaries of
geometry and key area and curvature distributions.
3) Aerodynamic performance analysis.
4) Linearized blade-to-blade flow analysis Aungier (1988b).

Return channel vane design should typically feature front loaded


vanes to minimize discharge flow deviation. Since inlet Mach

numbers are normally moderate, this loading style is a reasonable


choice. The vane loading parameters, K6 and K 7 , define the local
radial gradient of the cotangent of the blade angle as a fraction of
the overall (average) gradient. This provides direct control of the
blade loading style (Aungier, 19886). Values of 10 6 = 1.6 and
K, = 0.4 have been found to be reasonable choices. The vane
maximum thickness, and its location along the chord line should
be used to control the vane passage area distribution to minimize
local mean velocity gradients.

0 TEST DATA
PREDICTION

E-
Z
'41ID
,

44,

CONCLUSIONS

Mean streamline aerodynamic performance models have been


presented for vaneless annular passages and return channels.
Predicted loss coefficients for vaneless diffusers and return
systems show excellent agreement with experimental data. These
performance models are well suited to interactive aerodynamic
design activity and can be incorporated into any mean line
centrifugal compressor performance analysis.

sr

U6.0

6:74, 14,

opppuse

M':
M
0

Cl

"

A systematic interactive design system for return system design


has been presented. It has dramatically reduced the engineering
time required for this design activity. Test results from a recent
return system design accomplished with this design system show

0
60
80
120
100
140
t DESIGN FLOW
FIGURE 10: COMPONENT PERFORMANCE FOR COMPRESSOR "E"
0

Ideal diffusers show increased losses for 20 c greater than about


11 , so this lower value for a bend is not unexpected.

Results to date from this design system have been impressive.


Return channel design activity has been reduced from several
days to a few hours. Figure 10 shows results from a recent high
specific speed stage development program. The present methods

20

40

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/22/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

that the improved return system performance predicted by this


aerodynamic design and performance analysis system was actually
achieved.

REFERENCES
Aungier, R.H., 1988a, "A Performance Analysis For The
Vaneless Components Of Centrifugal Compressors", Flows In
Non-Rotating Turbomachinery Components, ASME FED-Vol 69,
pp 35-43.
Aungier, R.H., 1988b, "A Systematic Procedure For The
Aerodynamic Design of Vaned Diffusers", Flows In Non-Rotating
Turbomachinery Components, ASME FED-Vol 69, pp 27-34.
Aungier, R.H., 1990, "Aerodynamic Performance Analysis Of
Vaned Diffusers", Fluid Machinery Components, ASME
FED-Vol 101, pp 27-44.
Balje, 0.E., 1981, Turbomachines, Wiley, New York,
pp 34-37.
Davis, W.R., 1976, "Three-Dimensional Boundary Layer
Computation on the Stationary End-Walls of Centrifugal
Turbomachinery" Trans. ASME, J. of Fluids Eng., pp. 431-442.
Fister, W., Zahn, G. and Tasche, 1982, "Theoretical and
Experimental Investigations About Vaneless Return Channels of
Multi-Stage Radial Flow Turbomachines", ASME Paper No.
82-GT-209.
Hohlweg, W.C., 1987, "Correlation and Application of
Centrifugal Compressor Return System Losses" , Fluid Machinery
for the Petrochemical and Related Industries, Proceedings of the
IMechE, pp 97-103.
Howell, A.R., 1947, "Development of the British Gas Turbine
Unit", Lecture: Fluid Dynamics of Axial Compressors, ASME
Reprint.
Japikse, D. and Osborne, C., 1982, Vaneless Drser, Return
Bend and Return Channel Investigation, Creare Inc., TN 346
(Proprietary).
Johnsen, I.A. and Bullock, R.O., editors, 1965, Aerodynamic
Design of Axial Flow Compressors, NASA SP-36.
Johnston, J.P. and Dean, R.C., 1966, "Losses in Vaneless
Diffusers of Centrifugal Compressors and Pumps", Trans. ASME,
J. of Eng. for Power, pp. 49-62.
Nykorowytsch, P. , editor, 1983, Return Passages of Multi-Stage
Turbomachinery, ASME FED-Vol 3.
Reneau, L., Johnston, J. and Kline, S., 1967, "Performance
and Design of Straight Two-Dimensional Diffusers", Trans.
ASME, J. of Basic Eng. , pp. 141-150.
Schlichting, H., 1979, Boundary Layer Theory, Seventh
Edition, McGraw-Hill, New York, Chapter 20.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/22/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like