A. J. Wennerstrom
Aero Propulsion and Power Laboratory,
Wright Research and Development Center
(AFSC),
Wright-Patterson Air Force Base, OH
45433-6563
Highly Loaded Axial Flow
Compressors: History and Current
Developments
This paper discusses approaches taken over many years to achieve very high loading
levels in axial-flow compressors. These efforts have been associated predominantly
with aircraft turbine engines. The objective has been to reduce the size and weight of
the powerplant, to increase its simplicity and ruggedness, and, whenever possible, to
reduce cost. In the introduction, some fundamentals are reviewed that indicate that
increased work per stage can only be obtained at a cost of increased Mach number,
increased diffusion, or both. The earliest examples cited are some ambitious
development programs of the 1950s and 1960s. Some innovative schemes to increase
diffusion limits are described that took place in the 1960s and 1970s. Major ad-
vancements in dealing with higher Mach number were made in the 1980s. Finally, a
few thoughts directed toward potential future developments are presented.
Introduction
One of the most obvious ways to reduce the weight of the
axial-flow compressor in an aircraft turbine engine is to reduce
the number of stages it requires. Inasmuch as the engine cycle
will dictate the overall pressure ratio required, a reduction in
the number of stages inevitably means that each stage must do
more work. Maximizing the work per stage while retaining ac-
ceptable overall performance has been a goal of the aircraft
engine designer ever since an axial-flow compressor was first
incorporated into an aircraft engine. In this section, I want to
illustrate what more work per stage means to the compressor.
Then in subsequent sections I will illustrate several research
and development efforts to which this has led and then will
conclude with a few comments concerning future prospects.
The most fundamental expression for work per stage is the
Euler Equation of Turbomachinery, which, neglecting radius
change, can be written for compressors
Ah, = mva-vn) (1)
This tells us that the work per stage is directly proportional to
the wheel speed and the change in absolute swirl velocity
across the rotor. The upper limit for wheel sped is most often
defined by some structural limitation. However, since as wheel
speed increases, relative flow velocities and hence Mach
numbers increase, in some instances such as high bypass tur-
bofans, wheel speed may be Mach number limited for reasons
of thermodynamic efficiency . The permissible change in swirl
velocity across a rotor is, on the other hand, almost always
limited by some aerodynamic constraint. This constraint may
be Mach number or some loading parameter related to
boundary layer separation.
Contributed by the International Gas Turbine Institute and presented as an
invited lecture at the Ninth International Symposium on Air Breathing Engines,
Athens, Greece, September 3-8, 1989. Manuscript received at ASME Head-
quarters February 5, 1990.
Losses, and hence efficiency, and also stall margin are also
related to work per stage. Although losses result from many
different phenomena, the majority of losses can be divided in-
to those related to Mach number (shock waves) and viscous
losses related to diffusion acting on boundary layers. Some of
the earliest indices of loading considered only inlet to exit con-
ditions across a blade row. These guidelines were arrived at by
examination of a very limited range of low-speed data. They
included the static pressure rise coefficient
Pi~P\
-<0.6
PW\
and the de Haller number
W1
W,
>0.72
(2)
(3)
The former defined the maximum practical static pressure rise
as a function of inlet dynamic head. The latter defined a
minimum exit-to-inlet relative velocity ratio to achieve ac-
ceptable performance. Neither of these indices were correlated
against losses. The most useful diffusion parameter to be in-
troduced was the diffusion factor presented by Lieblein et al.
(1953). This not only provided some guidance relative to prac-
tical upper limits for conventional blade rows, but it also
proved a useful correlation parameter for relative total
pressure viscous losses.
The diffusion factor, again neglecting radius change, can be
written
D=\ W2 Vn-Vn
2a W,
(4)
Through a series of simplifying assumptions in its derivation,
Journal of Turbomachinery OCTOBER 1990, Vol. 112/567
Copyright © 1990 by ASME
Downloaded 17 Feb 2012 to 159.226.48.114. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
the above factor is intended to be an approximate representa-
tion of the local suction surface diffusion factor defined by
^\nra] ~~ ~
W •W,
w
rr
avg
It is interesting to compare these three parameters for the ex-
treme case of an impulse blade row. For the sake of simplicity,
let us further consider the flow to be incompressible and to
have no inlet absolute velocity swirl, i.e., Vei = 0. Since im-
pulse blading has no static pressure change, the pressure rise
coefficient
P2-P1
-^-pw\
An impulse blade row in incompressible flow will have sym-
metric inlet and exit velocity diagrams in the relative frame.
ri and the de Haller number
W2
W,
= 1.0
Therefore, according to these criteria, the blade row has no
loading whatsoever. However, the diffusion factor can have a
substantial value and will be
D-
V- U
2a W, aW,
or, in terms of angles,
D =
sin /3,
Although the magnitude of this diffusion factor may be mean-
ingless because the existing correlations were derived for such
different blading, the diffusion factor does correctly indicate
that there can be a large amount of diffusion even in an im-
pulse blade row. The fluid will accelerate to a peak velocity
somewhere on the suction surface while turning and then must
inevitably diffuse to a much lower value to achieve
equilibrium exit conditions.
In order to see the combined effect of wheel speed and dif-
fusion on stage work capacity, we can combine the Euler
equation, equation (1), with the equation for diffusion factor,
equation (4). This leads to
W-, Mil = 2aUWAD-\+-~)
We see from this that work capacity is directly proportional to
wheel speed, is nearly proportional to inlet relative velocity,
and varies in some proportion with diffusion factor; if
W2/W1 = 1.0 it would be directly proportional. One also
observes that stage work capacity is directly proportional to
solidity. However, this is subject to several practical con-
straints in that skin friction losses will increase with wetted
surface, increased blade blockage will reduce choking mass
flow, and finally the weight will increase in more than direct
proportion to the blade count. Hence, it is readily apparent
why all efforts to increase stage work capacity have concen-
trated on increasing wheel speed, Mach number (relative
velocity), and diffusion.
Supersonic Compressors
Early History. As was pointed out in the previous section,
the most obvious way to increase work per stage was to in-
crease wheel speeds, whirl velocities, and hence Mach
numbers. This was pointed out many years ago and the idea of
operating a compressor at supersonic relative velocities with
normal shock waves in the blading is generally first credited to
Weise (1937) in Germany. Two excellent summaries of early
work with supersonic compressors are presented by Klapproth
(1961) and Erwin (1964).
Klapproth made the astute observation that the early design
approach for supersonic compressors was developed more
from supersonic diffuser design criteria than from conven-
tional compressor design criteria. As a consequence, early
work with supersonic compressors was not a logical evolution
of work with lower speed machines but rather attempted to
make a quantum leap forward with little regard for past tur-
bomachinery experience. Hindsight has shown that many of
the loading criteria developed for lower-speed machinery have
retained a remarkable validity for high-speed machinery, as il-
logical as it might seem. The more successful machines built
over the years have combined concepts of dealing with higher
Mach numbers with older, more conventional concepts of
blade aerodynamic loading limits.
Most work concerning supersonic compressors has already
been published in the open literature. What I would like to
present here, in approximately chronological order, are four
early, very serious development efforts that have not previous-
ly been published because they were originally classified. Each
of these incorporated a supersonic compressor, which went
through extensive engineering development. All four
represented extremely ambitious exploratory or advanced
development programs, had varying degrees of success
because of that, and hence never found their way into any pro-
duction engine.
The J-55 Turbojet. This is probably the first gas turbine to
run as an engine incorporating a supersonic compressor. Its
development was initiated in 1947 with the objective of power-
ing a target drone. Its compression system was unique in in-
corporating a single-stage supersonic axial compressor of the
shock-in-rotor type, designed to provide a 2.75 total pressure
ratio at a corrected tip speed of about 488 m/s (1600 ft/s). It
included inlet guide vanes, which provided some counterswirl
at the hub and incorporated a tandem-bladed stator. The
overall engine never achieved enough of its design goals to be
put into production. However, by the conclusion of the con-
tract, the compressor had achieved a total pressure ratio of ap-
proximately 2.9 at an isentropic efficiency of about 0.76 at
design speed. Lower in its speed range its efficiency peaked at
about 0.82. One of the major development prob-
lems was achieving adequate performance from the supersonic
compressor. In 1947, the contractor, Fredrick Flader, Inc.,
was optimistic enough to state, "we are convinced that this
development has progressed to a point where its incorporation
Nomenclature
D = diffusion factor
Aocai = local diffusion factor
h = enthalpy
p = static pressure
U = wheel speed
Ve = absolute swirl velocity
W = relative velocity
W
W
P =
peak suction surface
relative velocity
average suction surface
relative velocity
relative flow angle
measured from axial
static density
a = solidity (chord/spac
Subscripts
1 = blade-row inlet
2 = blade-row exit
t = stagnation quantity
568 / Vol. 112, OCTOBER 1990 Transactions of the ASME
Downloaded 17 Feb 2012 to 159.226.48.114. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Table 1 Counterrotating compressor design characteristics at
sea-level takeoff conditions
Low-pressure rotor
Design tip speed
Total pressure ratio (design)
Adiabatic efficiency
Inlet diameter ratio
Exit diameter ratio
High-pressure rotor
Design tip speed
Total pressure ratio (design)
Adiabatic efficiency
Inlet diameter ratio
Exit diameter ratio
Overall compressor
Overall pressure ratio
(excluding inlet losses)
Overall adiabatic efficiency
Compressor o.d.
Compressor weight flow (design)
Weight flow/frontal area
Impulse type
442 m/s (1450 ft/s)
3.69/1
0.847
0.65
0.772
Shock-in-rotor type
266 m/s (874 ft/s)
1.829/1
0.623
0.772
0.827
Statorless
6.75
0.775
406 mm (16 in.)
17.7 kg/s (39.0 lb/s)
135.7 kg/s/m2 (27.8 lb/s/ft2)
Fig. 1 Near-midspan cross section of rotor blading
in this design proposal is w a r r a n t e d . " History
showed this conclusion to be somewhat premature. These bits
of information were derived from Downs (1953) and Blanton
(1989).
Curtiss-Wright Counterrotating Compressor. This was
probably one of the most ambitious of all supersonic com-
pressor designs ever attempted. The effort took place in the
mid-1950s and the design was reported by Sabel and Sabatiuk
(1957a). It was designed for a flight Mach number of 3.0, to
accept an axially supersonic inlet Mach number at this condi-
tion utilizing an integral fixed-geometry free-stream inlet, and
to be able to take off from a sea-level static condition with a
choked inlet. One of its major goals was to eliminate the
necessity for a heavy, complex, variable-geometry inlet and its
controls and actuating mechanisims. Its objective was to
employ counterrotating rotors to avoid the necessity for any
inlet guide vanes or stators and to achieve an adequate
pressure ratio over its flight regime through this means.
Its projected design features are summarized in Table 1 for
the sea-level takeoff condition. Most notable is the fact that it
was proposed to do this with an impulse first rotor and a
shock-in-rotor second rotor. As with several of its
predecessors, supersonic diffuser design practice guided
design of the second rotor. Conventional compressor loading
criteria were completely ignored. Because of the sea-level
static takeoff requirement, the leading edges of the first rotor
were staggered at an angle of two or three degrees at this con-
dition. However, at the flight cruise condition, the first rotor
was anticipated to operate at negative incidence and with a
swallowed shock pattern.
The first rotor incorporated all of the technology con-
ceivable at the time. This included leading edge sweep de-
signed to keep the sonic line coincident with the surface of
revolution defined by the leading edge. An axisymmetric
stream filament design technique was used employing simple
radial equilibrium theory. As understood today, this would
exclude internal computing stations and streamline curvature •
effects. The choice of solidity and aspect ratio was based on
previous recent NACA experience. The mean solidity of the
first rotor was approximately 2.8. Maximum thickness varied
from 3.3 percent chord at the tip to 7.9 percent chord at the
hub. A method-of-characteristics computation was made to
insure a reasonable probability of passing design flow.
The second rotor was designed for an average relative inlet
Mach number of 2.34. The internal one-dimensional area con-
Fig. 2 Photograph of high-pressure rotor
traction of the blading was designed to permit starting a
relative approach flow of Mach 2.0. Twenty-five degrees of
leading-edge sweep was incorporated to enhance starting of
the supersonic flow. Today we would say that this rotor was
designed with significant precompression, and from the throat
to the trailing edge plane an equivalent cone angle of approx-
imately eight degrees was employed. The mean solidity was
approximately 2.05. The same aerodynamic design technique
was employed as for the first stage. An approximately
midradius cross section of both rotors is shown in Fig. 1. The
second rotor is shown assembled in Fig. 2.
Interestingly enough, diffusion factors were calculated. At
sea-level take-off conditions, they were 0.258 for the low-
pressure rotor and 1.00 for the high-pressure rotor. For the
first rotor of impulse design, the diffusion factor is not ter-
ribly meaningful. However, the diffusion factor of 1.00 for
the second rotor is typical of so many unsuccessful designs of
that era. When one looks back on the overall experience, it is
remarkable to what degree performance does correlate with
diffusion factor under circumstances far removed from those
under which it was derived.
The experimental results of sea-level static testing were
reported by Sabel and Sabatiuk (1957b). Choking in the first
rotor limited weight flow to approximately 79 percent of the
design value. Measured performance recorded with both
spools at design speed was an overall total pressure ratio of
3.24 and an isentropic efficiency based on temperature rise of
73.3 percent. This efficiency seems surprisingly high when it is
noted that the total pressure ratio is only 48 percent of the
design value. Observation of casing static pressures presented
in Fig. 3 shows that most of the static pressure diffusion ac-
tually occurred downstream of the second rotor rather than
within it. Inasmuch as the exit plane instrumentation was
within this region, it is likely than when exit mixing losses are
taken into account, the true efficiency would be much lower.
The casing static pressures tell another interesting story. The
static pressure ratio across the first impulse rotor was intended
to be about unity and this was the result achieved. However,
the static pressure ratio intended to exist across the second
rotor was about 11.7 according to the design. In fact, the ex-
perimental results show a negligible static pressure rise also
Journal of Turbomachinery OCTOBER 1990, Vol. 112/569
Downloaded 17 Feb 2012 to 159.226.48.114. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 3 Casing static pressure distribution Fig. 5 Close-up of assembled high-pressure rotor viewed along span
Fig. 6 TS-10 supersonic rotor
Fig. 4 DLD gas generator supersonic compressor blade
across the second rotor. From inlet to exit, an overall static
pressure ratio of only about 2.0 was achieved and all of this
occurred over a distance of several annulus widths
downstream of the second rotor.
Several additional tests were made with this rig. However, a
major facility-related mechanical failure destroyed the com-
pressor and this program was terminated without further
significant accomplishment.
Direct Lift Demonstrator. This was a program ac-
complished by the General Electric Company in the
mid-1960s. Its objective was to develop the major components
of a turbofan engine to a level of technology consistent with
direct lift (VTOL) requirements and then to assemble them as
an engine and test it in a typical environment. It was a two-
spool turbofan. The low-pressure spool had a single transonic
fan stage. The high-pressure spool had a single supersonic
compressor stage. My remarks will be limited to aerodynamic
aspects of the supersonic core stage. The data used were
presented by Conliffe (1971).
The supersonic stage was a shock-in-rotor/shock-in-stator
configuration. Its design characteristics included a design total
pressure ratio of 3.15 at a corrected tip speed of 420 m/s (1378
ft/s) and with a projected isentropic efficiency of 75 percent.
The inlet hub/tip radius ratio was approximately 0.82.
Relative Mach numbers ranged from about 1.2 to 1.4 across
the leading edge of the rotor and from about 1.4 to 1.5 across
the leading edge of the stator. Design static pressure ratios at
the outer casing for both rotor and stator were approximately
1.9. Rotor solidity was approximately 2.2; stator solidity was
about 4.8. Mean aspect ratios were approximately 0.88 and
0.27 respectively. An example of the rotor blading is shown in
Fig. 4. A close-up of the assembled rotor is shown in Fig. 5.
The performance achieved for the best configuration subse-
quently used in the gas generator was a total pressure ratio of
about 3.5 at 71.4 percent isentropic efficiency. A large number
of configurations were examined during the development pro-
gram. The majority were aimed at improving the stator per-
formance. Although the indicated performance of the rotor
had been quite good throughout the program, the stator total
pressure recovery and discharge flow distribution remained
quite poor. In this instance, the designers observed generally
accepted limits for loading parameters such as diffusion fac-
tor. However, once again, the inability of a stator to accept
Mach numbers and static pressure ratios as high as those of
rotors had been demonstrated and nothing the designers at-
tempted altered this situation substantially.
Turboaccelerator (TS-10). This was a progr
本文档为【Highly Loaded Axial-Flow Compressors - History And Current Development】,请使用软件OFFICE或WPS软件打开。作品中的文字与图均可以修改和编辑,
图片更改请在作品中右键图片并更换,文字修改请直接点击文字进行修改,也可以新增和删除文档中的内容。
该文档来自用户分享,如有侵权行为请发邮件ishare@vip.sina.com联系网站客服,我们会及时删除。
[版权声明] 本站所有资料为用户分享产生,若发现您的权利被侵害,请联系客服邮件isharekefu@iask.cn,我们尽快处理。
本作品所展示的图片、画像、字体、音乐的版权可能需版权方额外授权,请谨慎使用。
网站提供的党政主题相关内容(国旗、国徽、党徽..)目的在于配合国家政策宣传,仅限个人学习分享使用,禁止用于任何广告和商用目的。