THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS

345 E. 47th St., New York, N.Y. 10017 97-GT-525

The Society shall not be responsible for statements or opinions advanced hi papers or cliScussion 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. Authorization to photocopy
material lot internal or personal use under circumstance not falling within the fair use provisions of the Copyright Act is granted by ASME to
libraries and other users registered with the Copyright Clearance Center (CCC)Transactional Reporting Service provided that the base fee of $030
per page is paid directly to the CCC. 27 Congress Street Salem MA 01970. Requests for special permission or bulk repmduction shoutl be addressed
to the ASME Tectmical Publishing Department

CopyrIgM 0 1997 by ASME All Rights Reserved Printed in U.S.A

ASPIRATED COMPRESSORS ' 1 111 11 11111R1 1111 111111

J. L. Kerrebrock, D. P. Reijnen, W. S. Ziminsky, L M. Smilg

Gas Turbine Laboratory
Massachusetts Institute of Technology
Cambridge, MA 02139 USA

ABSTRACT ior to be expected of a rotor with suction on all blades. They ex-
The performance of compressors can be improved in two hibited improved efficiency and increased mass flow. The rotor
ways by judicious removal from the flow path of the viscous flow as a whole with suction showed different stall behavior than its
in the boundary layers. First, removal of the boundary layer fluid counterpart without boundary layer control. Future plans include
just prior to or in a region of rapid pressure rise, either at shock the fabrication and experimental evaluation in the MIT Blowdown
incidence or more generally at the point of rapid pressure rise on Compressor, of one of the two stages discussed above. In this
the suction surface of the airfoil, can enable significant increases experiment suction will be applied to all the blades in both rotor
in the diffusion, hence in the work done by a stage for any given and stator, so that the increased work enabled by suction can be
blade speed. Second, removal of the high entropy fluid in the realized.
boundary layer minimizes the required compression work in sub- This research was supported by AFOSR, Dr. James
sequent stages of compression, thereby raising the compression McMichael, and by AlliedSignal Aircraft Engines, Dr. Arun Sehra.
efficiency. Analysis has shown that the latter effect can result in
approximately one half point increase in efficiency for each per- INTRODUCTION
cent of (high entropy) fluid removal. Design studies have been This paper reports progress in an investigation of the propo-
carried out for two different stages to assess the increase in pres- sition that the performance of compressors can be improved by
sure ratio that may be achieved. One stage that has been designed control of the viscous flows at points in the flow path where their
would produce a pressure ratio of 2 at a tip speed of 1000 ft/sec, behavior limits the diffusion attainable, the method of control
and may be very attractive for the fan stage of high-bypass turbo- being selective removal of the boundary layer flow. The investi-
fan engines. The other stage would produce a pressure ratio of 3 gation has to this point included experimental studies of the ef-
at a tip speed of 1500 ft/sec, and should be attractive as the first fects of suction at selected locations on an existing transonic com-
stage of a core compressor or the fan stage of a low-bypass ratio pressor rotor, studies of the thermodynamic effects of the fluid
engine. An experiment has been completed, to examine the ef- removal on compressor efficiency, and design studies of the gains
fect of boundary layer removal just prior to shock impingement in pressure ratio that can be achieved by designing stages to take
on the suction surface of blades in a transonic rotor. The suction advantage of the higher levels of diffusion enabled by boundary
was implemented on 5 of the 23 blades of the rotor, providing a layer control. It has not yet progressed to the construction and
direct comparison of the flow behavior with and without suction. test of such a stage, although this is intended as the next step. The
Analysis of the data has shown that the blades with suction have results are so far sufficiently encouraging that we wish to share
increased mass flow and that the flow more closely followed the them with the technical community.
suction surface near the trailing edge. The differences between Certainly it is not necessary to elaborate the desirability of
aspirated and normal blades were most pronounced when the ro- increasing the stage pressure ratio of compressors, since this has
tor was very close to stall. The third and fourth of the blades in been a prime objective of compressor designers since the gas tur-
the group with suction appear to be representative of the behav- bine was invented. To date most progress toward increased pres-

Presented at the International Gas Turbine & Aeroengine Congress & Exhibition
Orlando, Florida —June 2-June 5,2997
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 sure ratio has been made by increasing the blade speed, the tem- boundary layer, to reduce its thickness and modify its profile so
perature rise varying as blade speed squared for kinematically that it is more resistant to separation in subsequent diffusion.
similar flows. The aerodynamic loadings, quantified by a mea- The fluid extraction can be accomplished in a number of ways:
sure such as the Diffusion Factor, have not increased significantly. by scoops that intercept the flow at a fraction of the boundary
One central proposition under study here is that the loading can layer thickness, by slots or holes in a surface that otherwise re-
be increased significantly if the boundary layers are inhibited from tains its normal shape. All have their advantages and disadvan-
separation by removal at critical locations such as: just before the tages. As we shall explain later the scoop was judged to be most
most severe pressure gradient on the suction surface of the blades, advantageous for application at the point of shock impingement
at the point of shock impingement on the suction surface of tran- in a transonic rotor.
sonic blading, on the casing just ahead of the rotor or stator, or at
the tip on the suction surface near the trailing edge. A second THERMODYNAMIC IMPLICATIONS OF FLUID REMOVAL
proposition is that it is better from a thermodynamic viewpoint to One of the premises of the work reported here is that it can
remove the boundary layer flow from the flow path, rather than be beneficial to the overall performance of a compressor or to an
to mix it with the core flow, because this minimizes the entropy engine incorporating it, to remove the fluid influenced by vis-
of the flow from any stage of compression and therefore also the cous effects from the flow path, rather than mixing it with the
compression work of succeeding stages. inviscid flow and continuing to compress it. This is an uncon-
In spite of the attractive possibilities of boundary layer con- ventional approach; most other approaches to controlling separa-
trol in compressors, not much work has been done on control of tion envision using blowing or slotted airfoil sections that ener-
the type discussed here. Rather most of the past work has been gize the viscous-dominated flows to prevent separation. While
directed at slotted blades, trailing edge blowing or other ways of these viscous flow control schemes have much to recommend
reenergizing the boundary layer flow so as to prevent separation. them, they leave the high entropy fluid in the flowpath, raising
To the authors' knowledge the only systematic investigation of the average entropy (temperature) of the flow and increasing the
the use of boundary layer suction is that reported in Loughery et compression work in successive stages of compression. The
al. (1971). This work examined both blowing and suction as means present approach, of removing the viscous-dominated flow, low-
for increasing the turning capability of compressor stators. ers the entropy of the residual flow and so minimizes the work.
Broadly, the results were that blowing degraded the performance The following analysis quantifies these effects.
of the stator while suction enhanced it significantly. But it seems A complete investigation of the effects of fluid extraction
these interesting results were not developed further. would entail the consideration of ways of using the flow extracted
We begin our presentation with rather general discussions of from the compressor, and their impact on overall engine perfor-
the opportunities for use of flow control by fluid removal and the mance. We will not deal with this very complex systems problem
thermodynamic implications. The results of an experimental study here, rather we will assume that the extracted flow is expanded to
of suction at the shock impingement point in a transonic rotor compressor inlet pressure so as to recover as much energy as
will then be described, followed by the results of design studies possiblle from it. This gives an upper limit to the efficiency ben-
of two stages that take advantage of the increased diffusion en- efit of fluid extraction.
abled by boundary layer control. For this purpose the suction scheme is shown schematically
in Fig. 1. At any particular blade row a small flow 8m that has
OPPORTUNITIES FOR COMPRESSOR ENHANCEMENT suffered an entropy rise due to viscous effects, is extracted from
BY FLUID REMOVAL
The pressure rise of any compressor or pump is limited by
separation of the flow from the surfaces of the flow path at loca-
tions with adverse pressure gradients. Such locations include at
least: the suction surface of the blades where the pressure rises to
meet the downstream pressure, in transonic compressors, the lo-
cus of passage shock intersection with the suction surface of the
blades and the casing where its viscous layer is influenced by the
pressure field of the blading. Separation of the flow from the sur-
face at any of these points can lead to blockage of the flow in the
passage, reducing the diffusion from the design intent, and if se-
rious enough, to stall of the compressor. At such locations the
boundary layers will have thicknesses and velocity profiles de-
termined by their development to the point in question. The gen-
eral strategy advocated here, is to extract sufficient fluid from the Fig. 1: Schematic of removal of high entropy flows by suction.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 argument the important difference in the extracted and core flows
is in theft entropies.
The main flow at point 5 ends up at a lower entropy (and
temperature) than it would have (at point 3) in the absence of
suction, so the work required for a given pressure ratio is lower
and the resultant compressor efficiency is higher. Of course such
suction can be applied repeatedly in each stage. Here we analyze
the effect of suction at one location.
For a compressor without bleed, the work of compression
can be written

7- I 1
Wnb = me 1)( 7;3 - Ttl) = Int pTti [(ft1
Pil
) 111P -1

AS
where ri p is the polytropic efficiency and the subscript denotes
stagnation values in stationary coordinates. In the following ar-
gument we will assume that as a first approximation the poly-
tropic efficiency is uniform for all compression processes. With
bleed of amount 6m at a point 2, the work is

Wb = mcp (Ta — + (m — eim)cp (7;5 74 ) — eimcp (Tr o — m)

the last term expressing the recovery of energy from the extracted
Fig. 2: Schematic of the suction scheme on T-S coordinates. flow by its (ideal) expansion to compressor inlet pressure.
The entropy of the extracted flow relative to that of the aver-
age flow at point 2 can be written:
the flow path. The remainder of the flow m-8m continues through
the compressor. The thermodynamic effect of the flow extrac-
As, = c 1rea)
tion is shown on a temperature-entropy diagram of the compres- P Tr2
sion process on Fig. 2. Point 1 is the inlet and 3 the discharge of
the compressor. All of the thermodynamic quantities are stagna- The temperature of the core flow at 4 is related to the average
tion quantities in stationary coordinates unless otherwise speci- temperature at 2 and that of the extracted flow by:
fied. At some intermediate point 2 the suction is implemented.
The solid line 1-2-3, indicates the compression process in the mr12 ("I ("3)1;4 Smrt6
absence of suction, if the flow is treated as mixed. At point 2 With the usual ideal-gas relations between pressure and tem-
however the flow is divided into two portions, a small fraction perature, the work per unit mass of flow delivered to the exit of
that has undergone high viscous interaction hence has a high en- the compressor is,
tropy, indicated by condition 6, and the remainder with a low
entropy, at condition 4. This would be done for example by a {v-I
boundary layer scoop, as discussed previously. The high entropy
fluid would be extracted, and (conceptually) expanded back to
-( 7 )c
- r (Lir'P -1
the compressor entrance pressure at point 7 to recover as much
work as possible from it, while the remainder continues in the
Y-1 1
compression process to point 5. In application the bleed air might
be used for cooling of the engine hot section, for example, but
rather than complicate this analysis with such systems consider-
÷Cp Tr4 [(2 j-1 )1TIP -11--(-
et"
_ p 6 [1-(1) m
ations, the best possible situation from the viewpoint of compres-
sor efficiency is assumed. There is also a subtlety in the assump- Here the terms represent respectively the work done to bring the
tion that the fluid is divided at equal stagnation pressures. In fact entire flow to the point 2, the work done to finish the compres-
this division would occur at equal static pressures, but for this sion of the main flow to point 5, and the work recovered by ex-

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 pansion of the extracted flow. Note that we have assumed this 0.7 - Compression Ratio
latter process to be ideal. Bleed Pressure Ratio
0.6 -

Min imu m Mac hfor Gain

Eliminating T14 and Tt6 with the energy balance at point 2 4
and the definition of As v , and assuming (8m/m)«1, we find the 0.6
3
fractional difference of the works per unit of delivered flow with 0.4 -
and without bleed to be
0.3 -

7-1(1 02 -
Woth W6 Afv Pa Y dp ) -1
0.1 -
m (m eftn) _ ern) ( Prl
1-
W„b trn ) y-I I I , ,
rn (PajnIP -1 0 10 20 30 40 50
Compressor Pressure Ratio

Fig. 3: Minimum Mach No. for which bleed increases the

If the quantity in curly brackets is positive, there is a net reduc- efficiency, as a function of the compressor pressure ratio and
tion in work for a given pressure ratio, hence a corresponding the ratio, of compressor pressure ratio to pressure ratio at
increase in efficiency. It depends on the magnitude of the en- the bleed point (for ri p = 0.9).
tropy excess in the bled flow, on the compressor pressure ratio,
and on the pressure ratio at which the bleed occurs.
y-1[ I _ I )
To assess the likely magnitude of As v/cp, consider a bound-
ary-layer like flow over a compressor blade surface. Assume (Pa) 7 yip -
the static pressure in the core flow and at the surface are equal, y -1 Pti
M2 -

and that the surface is adiabatic. Then the temperature at the 2 mm y-1 y-I[
surface is close to the stagnation temperature in blade coordi- _(Pr2)
1& i m p °71
nates, and the entropy excess of the boundary layer fluid over
1911 PtI
the core fluid is
ós 1 ,11+-1 m2) Figure 3 shows that for most overall pressure ratios and bleed
k. 2 pressure ratios, the minimum Mach number is between 0.3 and
0.5. The inlet stages of most modem compressors satisfy this
where M is the Mach number relative to the blade. Therefore the criterion.
exponential coefficient containing the entropy rise is of the order Returning to the actual gain in efficiency to be expected from
of 1+ (y-1)M 212. Substituting this estimate in our general result, bleed, the quantity in the curly brackets is shown in Fig. 4 for a
we find relative Mach number of 1.5. We see that bleed at almost any

Wnb Wb 0.6 Compression Ratio

m (m-em) 4 Bleed Pressure Ratio
0.5 -
Wnb m )t. 2 2 ./
rn Fri 0.4
0
to
g 0.3 -
cc To
02 -
0
0.1 -

0
0 10 20 30 40 5 01
Compressor Pressure Ratio
From this form of the result it is clear that the principle gains
are to be realized for large relative Mach numbers, in the first Fig. 4: Fractional gain in efficiency per fractional bleed flow
stages of compressors. Indeed there is a Mach number below as a function of compressor pressure ratio and bleed
which no gain can be had by bleeding. This is given by pressure ratio, for relative Mach No. of 1.5.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 point in the compressor will produce four tenths of a percent in-
Casing
crease in efficiency per percent of mass bled. This is of course Boundary Test
only true if the bled fluid has the entropy level associated in this Layer Bleed Section
case with stagnation from a Mach number of 1.5. The benefit is Manifold for
Boundary Layer Bleed Dump Tank
quite insensitive to both overall pressure ratio and the pressure
ratio at which the bleed occurs, a somewhat surprising result.

EFFECTS OF BOUNDARY LAYER REMOVAL AT

SHOCK IMPINGEMENT IN A TRANSONIC ROTOR
flu
Valve Driver
This experiment was conceived as part of the program of
research outlined in the Introduction. As such it had the specific
and limited purpose of determining the effects of removal of the
fluid in the boundary layer just prior to the shock impingement Fig. 5: Blowdown compressor with flow-initiation valve.
on the suction surface of the blades of a transonic rotor, this being
one situation where the impact of boundary layer control was Pesign of Experiment.
expected to be large. This experiment was performed by one of the authors
This particular experiment was not intended to show the full (Reijnen) as his doctoral program. It is described in complete detail
benefits that boundary layer control may have for compressors, in Reijnen (1997), so that the present description will be brief.
since these benefits can be fully realized only if the stage design The modifications to the blades were designed with the aid of
exploits the higher work enabled by boundary layer control. Given MISES, a streamline solver (Youngren, 1991) that treats the core
the limited objective of the experiment, it was decided that bound- flow in a quasi three dimensional way while accounting for two-
ary layer control would be applied only to a portion (5/23) of the dimensional boundary layer growth in the pressure field imposed
blades on an existing MIT B lowdown Compressor rotor. by the core flow. This consistent interactive treatment of the
The existing rotor was used because it was available and be- boundary layer and inviscid flow is known to give very good
cause there was a large inventory of data on its behavior, including results for isolated airfoils. In the present application the
three-dimensional visualizations of its shock structure (Epstein, axisymmetric stream surfaces were computed by a streamline
1977). Previous measurements of the performance of this rotor in- curvature approach (Hearsey, 1975). Later, this same calculation
dicated that it suffered from separated flow over the outer portion of procedure was adapted to design of the stages with fluid removal,
the span, presumably because of unfavorable shock-boundary layer but for the experiment design it was used to estimate the bound-
interaction. The flow visualizations of Epstein (1977) had provided ary layer thickness at the point of shock impingement. The bound-
detailed information on the shock structure and in particular on the ary layer scoop was set at a height such that it would intercept
location of the impingement on the suction surface. For these rea- about two thirds of the momentum thickness of the boundary layer.
sons, it seemed a good vehicle for testing the hypothesis that bound- A MISES result for a typical blade section (without boundary
ary layer removal could be beneficial in such situations. layer removal) is shown in Fig. 6. This resulted in scoop heights
The boundary layer control was applied to only a fraction of of about 1 millimeter (on a blade chord of about 8 cm). The scoops
the blades for three reasons. First, it was thought useful to have a

/
direct comparison of the behavior of blades with and without
boundary layer removal, other conditions such as the corrected
speed and pressure ratio being equal. Second, incorporation of
the system for handling the flow removed from the rotor was L
much easier for a small number of blades than for the full comple-
ment. And finally, the modification of the blades themselves by
addition of boundary layer scoops and flow passages was time-
consuming and expensive, so it was attractive to minimize their
a
number.
ltk‘tr
In connection with the experiment the MIT Blowdown Com-
pressor was modified by the addition of a fast-acting valve that titc
A 4Y.
replaces the explosively-cut diaphragm used formerly. Since this
valve greatly improves the performancae of the facility and has
not been described previously, a schematic of the modified facil- Contours of Mach increment = 0.025
ity is presented as Fig. 5. A detailed account of the valve is given
in Ziminslcy (1996). Fig. 6: Contours of Mach No. from MISES for r/rT = 0.8-0.85.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 extended from 76 to 95 percent of span, the inner end being set at Casing Boundary Layer Bleed
approximately the inner termination of the shock, the outer being Spider
set to ensure the stiuctural integrity of the blade. The suction
passages in the blades were machined from the pressure surface,
then closed out by a thin layer of fibreglass, returning the pres-
sure surface nearly to its original contour. The scoops were fabri- Er !I-
Era
cated of graphite-epoxy, with virtually all fibers oriented in the
streamwise direction to give the very thin scoops sufficient stiff-
ness. Because the suction manifold cut the aluminum structure
of the blade in two, it was terminated at 95 percent span. A photo- 1.11=11.111
graph of a completed blade is included as Fig. 7.
The removed fluid was carried out of the rotating system Rotating Seal
through passages on the rotor, a rotating seal at a small radius,
and a spider upstream of the rotor, as shown in Fig. 8. The flow
discharged to an auxiliary dump tank sized so that the suction Fig. 8: Upstream spider and suction flow removal system in
system was choked throughout the experiment. Measurements of blowdown compressor.
the suction mass flow were made by total to static preeeure mea-
surements at points in the rotor and spider and also by monitoring
A large number of experiments were carried out with this
the pressure variation in the auxiliary dump tank to which the
rotor configuration, at a selection of operating points. In addi-
bleed flow discharged.
tion, experiments were carried out with a (nominally) identical
The instrumentation consisted of high-frequency response
rotor without the fluid removal system, to form a baseline. From
silicon pressure transducers used in a number of ways. Some were
the analysis of these data, three sets of data emerged as most
installed flush in the casing at points upstream, downstream and
meaningful. These are for: the unmodified rotor, operating at its
over the tip of the rotor. A four-way probe contained four trans-
nominal design point (termed the baseline in the following dis-
ducers, one mounted so as to face the nominal flow direction and
cussion), the modified rotor operating at the same nominal de-
three others at 45 degree angles to this, two in the axial-circum-
sign point as the baseline (termed the design point) and the modi-
ferential plane and one in the axial-radial plane. Together these
fied rotor operating with the same throttle setting but a slightly
four transducers yield measurements of the stagnation and static
slower speed, hence at a point very close to the stall line for the
pressures, and the flow angles in the two planes. This probe was
unmodified rotor(terrned the high loading point).
the principal source of information about the rotor outflow. With
a diameter of about 2.5 mm, it has sufficient spatial resolution to
Results of Experiment
resolve most features of the blade wakes.
In anticipation of the discussion to follow, we believe the
experimental results show that the effects of the fluid removal
were to increase the capacity of the blades for turning the flow,
increase the mass flow and improve the efficiency. The behavior
of the third and fourth blades of the group of five with suction, is
thought to be representative of the behavior that this rotor would
have exhibited with suction applied to all blades. Because of com-
plexities due to the application of suction to only 5 of the 23 blades
however, these conclusions must be developed from a discussion
of the measured features of the flow.
The effects of fluid removal from the 5 blades were mani-
fested primarily in:
a) the shock structure of the rotor, as observed from the array of
casing pressure transducers upstream of the rotor and over
the rotor tip,
b) the static pressure, stagnation pressure and flow angles in
the rotor outflow at a radius near the center of the scoops and
c) the behavior of the axial Mach number of the inlet flow as
the operating point crosses the stall line at the end of the
Fig. 7: Photograph of blade with suction scoop. experiment.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 40
Aspirated Blades
AxialPosition (deg)

0
11 10 9 8 P7 6 5 4 3 2 S1 S23 S22 S21 S20 19 18 B17 16 15 14 13
20

-20

-40

60
0 50 100 150 200 250 300 350
Phase (deg)
Fig. 9: Upstream shock structure as indicated by casing pressures at locations upstream of rotor.

The significance of each of these requires some explanation (as an ensemble average) where the increase at the five sucked
before the results are discussed. First, the shock structure was blades (200 to 280 degrees) is clearly evident. From these fea-
available only at the tip in these experiments. On the other hand, tures we conclude that the effect of suction on the upstream flow
the rotor outflow measurements were most meaningful at a ra- field to first order is to turn the flow into the blade row, expand-
dius typical of the scoops, and not available at all very near the ing it slightly. This pattern of upstream flow occurred at all oper-
tip because of the limited radial spatial resolution of the probe. ating points for the rotor with suction.
Therefore some caution must be exercised in connecting the up-
stream and downstream features. We believe the discussion to Pownstream Flowtield: The downstream flowfield in
follow takes this into account. With regard to the behavior of the general is very unsteady, with fluctuations of the same order as
rotor at stall it must be understood that the rotor operates at a the variations attributable to individual blades. So to bring out
fixed operating point (in terms of corrected speed and weight flow) the features traceable to the suction, the data was ensemble aver-
during that part of the blowdown when the throttle orifice dis- aged. To do this the probe was positioned at a fixed radius through-
charging into the dump tank is choked. At the end of the experi- out the experiment (rather than traversed radially), and the data
ment as the dump tank pressure rises the orifice unchokes and the from several revolutions of the rotor were averaged. Most of the
corrected weight flow decreases so that the rotor's operating point data shown here is of this type, however the most distinct fea-
moves up a constant corrected speed line toward the stall line. tures of the flow were also identifiable in the unaveraged data.
Once the rotor stalls the flow enters a regime approximating
shutoff, which is of no real interest for our purposes.

Upstream Shock Structure: The principal features of

the upstream shock structure are shown in Fig. 9. Here circum-
ferential distanace is expressed in degrees, so to show the shock
angles in the axial-circumferential surface the axial coordinate is
expressed in equivalent degrees. The crosses indicate the posi-
tions (times) of maximum rate of pressure rise in the shocks. The
first of the blades with suction (indicated by S I) shows a very
strong shock, followed by a very strong expansion, the net of the
two being an expansion to a pressure some 4 percent below the
normal upstream pressure. The shocks from the next two blades
(S23, S22) are nearly normal (i.e. similar to those from the
unsucked blades). The final two blades (S21, S20) show a read-
justment to the normal state. The static pressure is about 4 per-
cent lower than normal across the full patch of blades with suc- 50 100 150 200 250 300 350
tion. From this static pressure and the stagnation pressure, which Rot. Position (deg)
is equated tothe supply tank pressure, the axial Mach number at
Fig. 10: Axial Mach number at face of rotor, showing increase
the rotor face can be computed. This is shown as M4 in Fig. 10 at aspirated blades (200° to 275°).

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Rel. Flow Angle (deg)
Downstream of the rotor at the mid-span of the scoops, the 55 Aspirated Blades
static pressure was higher than normal over two or three blades
preceding (in measurement time) the sucked blades. It then de- .14/17\11,1(
creased across the sucked patch to a value below the normal and 45.
finally relaxed to the normal value after the sucked blades.
The four-way probe gave a wealth of additional information s S20
about the downstream flow, including the local Mach number, 35 ,

0 50 1
100 150 200 250 300 350
stagnation pressure and flow angles in the radial-axial and cir-
cumferential-axial planes. By a coordinate transformation these Rotational Position (deg)
quantities were also accessible in coordinates relative to the ro- Fig. 12: Outflow angle from rotor at highly loaded condition.
tor. By a technique originated by Thompkins (1976), the adia-
batic efficiency can also be computed locally. The local stagna-
tion temperature can be computed from the Euler equation be-
(highly loaded) in contrast, the normal blades have lost the clearly
cause there is no tangential velocity upstream of this rotor. The
defined outflow structure while the sucked blades retain it. Thus,
efficiency is then computed from the measured stagnation pres-
one effect of boundary layer removal is to cause the blades to
sure and the computed stagnation temperature. The results of this
behave as though they are still operating at the design condition,
computation were compared to direct measurements of the effi-
rather than at the highly loaded condition.
ciency via an aspirating probe by Epstein and Ng (1985) and found
A second point to be noted from these two data sets is that at
to be satisfactory.
the design point the two or three blades ahead of the sucked blades
The clearest evidence of the effects of boundary layer re-
show an outflow structure similar to that of the normal blades at
moval on the rotor outflow comes from comparison of the stag-
the more highly loaded operating point. This is traceable to the
nation pressure (in stationary coordinates) traces for the rotor with
excess of downstream static pressure at these blades, noted above.
suction, at two different operating points, the nominal design point
There is further confirmation that the boundary layer removal
and the heavily loaded point as described above. This compari-
increases the turning capacity of the blades, in the relative out-
son is shown in Fig. 11. Referring first to the top trace (design)
flow angle, shown in Fig. 12. Here it is clear that the flow off the
the flow off the suction surface of each blade is clearly defined as
suction surface of the sucked blades is turned closer to the axis
a region of large stagnation pressure, except for the three or four
(i.e. is turned more ) that from the normal blades. This is espe-
blades just ahead of the sucked blades. (The blade passing period
cially prominent for the first, fourth and fifth sucked blades.
is 360/23 = 15.7 degrees). The sucked blades, which appear in
These data plus complementary data on the stagnation pres-
the range of about 200 to 280 degrees, have outflow structures
sure in relative coordinates and the efficiency, lead to the an overall
not very different from the normal blades. In the lower trace
view of the flow as summarized in Fig. 13. In this figure ex-
cesses or deficiencies of static pressure are indicated by the circles
with plus or minus signs. The top diagram shows the information
gained from the casing static pressure transducers. This includes
the very strong shock and expansion from the first sucked blade,
the static pressure deficit upstream and downstream of the sucked
4 blades, and the upstream pressure excess ahead of the blades fol-
1.3 lowing the sucked blades.
The lower diagram shows the static pressure pattern down-
1.7. stream of the blades. It indicates an excess of relative stagnation
.0 pressure behind the several blades preceding the sucked ones,
together with a higher efficiency for these blades. It also indi-
cis 1.5
cates that the last three sucked blades show high relative stagna-
4 tion pressures and correspondingly high efficiency. This last point
a. 1.3 is considered to be of special importance, since the behavior of
0 these last three blades may be regarded as indicative of the be-
50 100 150 200 250 300 350
havior of a rotor (of this design)with suction applied to all blades.
Rotational Position (deg)
The proposition that the flow can adjust to the effects of suc-
Fig. 11: Comparison of outlet stagnation pressure of rotor tion in only two blade spacings, was tested by performing a MISES
with suction at design (upper) and heavily loaded (lower) calculation for six blades, three with boundary layer removal and
operating points. three without. The technique for modeling boundary layer removal

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Casing Data
7 6 4 3 2 SI S23 S22 S21 S20 19 18 17
< < <

\\\ \\\
Low Vel. Detect
High Ptr & Eft High Pt; & El
Ii tot

/
7 /
6 /
96 ,
94 / , "e
3#j -•••1 I "rdi
e I ,7
2 S1 523 19#41) S21
522 9
2i 23
S20 19 18 17
86% Span

Fig. 13: Summary of effects of suction on flow into and from rotor.

will be described in connection with the discussion of design stud- constant Mach number until it stalls at about 265 milliseconds.
ies. Only 6 rather than 23 blades were included because of com- Apparently the presence of the five blades with boundary
putational limitations. The upstream static pressure given by this layer removal modifies the stall behavior of the entire rotor. The
calculation is shown in Fig. 14. The important points are that a rotor with suction does not produce a higher stalling pressure ra-
static pressure deficit of about the right magnitude is exhibited tio than the one without it. This seems reasonable because the
for the sucked blades and that the changes occur over just one pressure ratio is limited by the blades without suction.
blade passage.
DESIGN STUDIES OF STAGES WITH
Stall Behavior With and Without Boundary Laver BOUNDARY LAYER REMOVAL
Removal: The stalling behavior of the rotor with boundary layer The intent of this portion of the program was to help quan-
removal from five blades, is compared to that of the baseline ro- tify the gains that may be realizable by designing compressor
tor in Fig. 15. The quantity compared is the Mach number at the stages to take advantage of the increased diffusion permitted by
inlet to the rotor. These are actual time traces, not ensemble aver-
ages, extending through the normal test time and into the time
when the throttle orifice has unchoked so that the rotor operating 10
point is moving up a corrected speed line toward the stall line. 0.75
Beginning at about 230 milliseconds the Mach number for the
baseline rotor decreases gradually until it stalls at about 250 mil- 2 0.5
liseconds. In contast, the rotor with suction maintains a nearly 0.25

0
Static Pressure at 10% Chord Upstream 170 190 210 230 250 270
1.0
1.0
0.9 No Suction
Ps 0.75
0.8 vr
2 0.5 0-wwesass
osimis." 000s001%01
07 0.25
-14 -10 -6 -2 2 6
Rot. Direction in Blade Passages 0
160 180 200 220 240 260
Fig. 14: Upstream static pressure variation from MISES Time (msec)
calculation for six blades, three of which have boundary
layer removal. Fig. 15: Effect of suction on stall behavior.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 boundary layer control. Thus far attention has been paid only to solutions and in fact no guarantees that satisfactory designs can
the limitations that arise from separation on the suction surfaces be achieved even with boundary layer control. Our objective has
of the blading, either from shock -boundary layer interaction or been to identify designs that meet the requirements, thus provid -
from rapid 'diffusion ahead of the trailing edge. It is recognized ing an assessment of the feasibility of increased loading. Starting
that the casing layer and perhaps other locations will require treat- with the velocity triangles dictated by the pressure ratios, the ap -
ment as well, but this discussion will be limited to the blades. proach that has been taken is to begin with rather conventional
Further, as will become apparent from the following discussion, multiple-circular arc blade profiles that nominally should deliver
even the treatment of the flow over the blade surfaces, as de- the required turning, compute the flow field using the combined
scribed here, leaves a great deal to be desired. So the present de- streamline-curvature-MISES approach, add suction at the criti-
sign study must be viewed as only a first step toward evaluating cal points and iteratively modify the blade shape and location
the potential performance gains possible with boundary layer and amount of suction until a solution was obtained for each
control. saeamsurface. In general, incipient separation or choking caused
For such a design study an efficient tool is needed, that mod - the calculation procedure to fail to converge, so pragmatically,
els adequately the critical flow phenomena. The approach that success was equated in part to convergence. Some features of the
has been taken is to use the Multiple Interacting Streamtube Solver two designs follow.
(MISES) (Youngren, 1991), coupled to an axisymmetric stream-
line curvature method (Hearsey, 1975). The result is a quasi -three MT= 1, Pressure ratio = 2
dimensional calculation in which the effect of radial shift of the This design was carried out by L. M. Smilg and is discussed
streamsurfaces, as estimated by the streamline curvature approach, in detail in Smilg (1994). The rotor had an inlet hub/tip radius
modifies the diffusion of the core flow in the MISES calculation. ratio of 0.55, constant casing diameter and constant axial veloc -
MISES in turn provides a consistent treatment of the interaction ity. To illustrate the solutions, Fig. 16 presents for mid-span the
between the core flow and the boundary layers, which are treated variations of Mach number on the blade surfaces. the Mach num -
as streamtubes by an integral approach. This core flow -boundary ber variation through the passage, and the variations of displace -
layer interaction is especially critical in transonic blading because ment and momentum thickness of the boundary layer on the suc -
of the great sensitivity of the flow to small changes in effective
flow area. The MISES approach has been well validated for iso-
lated airfoils and two-dimensional cascades. 1.5
Modelling the effects of suction is the subject of ongoing
research. Within the limits of the MISES approach as used here it
Mach Num ber

is not possible to include a detailed geometric representation of 1.0

the scoops and their resulting flow field. For the present explor-
atory designs the approach has been to represent the effect of
suction on the boundary layer by a reduction in the momentum 0.5
thickness at the nominal location of the scoop, and the effect of
fluid extraction on the core flow by an adjustment of the core
streamtube adjacent to the blade. The details of these calculations 0
are available in Ziminslcy (1996) and Smilg (1994). Stages have 0
Fraction of Chord
been designed to meet two different sets of requirements: .06
a) A fan stage with pressure ratio = 2 and tip tangential Mach
number =1, which could offer lower noise and better match - xi .05
Cc
Ca
ing of jet velocities for high bypass turbofans. .04
b) A compressor first stage or fan stage with pressure ratio = 3 w
E
and tip tangential Mach number = 1.5, which could substi - E .03
tute for two or more stages in military turbofans, or for an E .02
equal number of stages in the core of high bypass engines. ao
E
Either of these is an aggressive requirement, well beyond the 5 o .01
2
diffusion limits of blading without boundary layer control. More 0
quantitatively, the D Factors for the successful designs, are in the
Fraction of Chord
range of 0.6 to as much as 0.8, the definition of success being that
the computed flow is well behaved, with reasonably low com- Fig. 16: Variations of surface Mach numbers, passage Mach
puted losses and with no large regions of separation. number, and momentum and displacement thicknesses for
Since these are design studies there are no uniquely "best" rotor at midspan.

10

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 tion surface. On the boundary layer plot, the effect of suction can to streamtube area changes. We are not sure that the design to be
be seen at x/c = 0.4. At this point the mass removal was 3.2 per- described is fully viable, but it does represent a first attempt to
cent of the passage flow and the scoop height would be 0.029 of design a high-tip speed stage with very large turning. And the
chord. With the suction, the momentum thickness and shape fac- calculations suggest that attached flow may be attainable in such
tor remained acceptable up to the trailing edge. Without it separa- a design.
tion would have been predicted shortly beyond mid-chord. The procedure used for this design, and some typical fea-
A summary of results of this type is given in Table 1. It can tures of the result, will be described briefly. The desired pressure
be seen that rather large amounts of suction were required to sup- ratio and an assumed axial velocity variation through the stage,
port the diffusion required in the stator, considerably less in the together with the assumption of radially constant temperature rise,
rotor. Since it is probably easier to deal with the bled flow from essentially dictate the velocity triangles. A constant tip radius was
the stator than from the rotor, this may still be an interesting de- selected, together with reasonable hub slopes in the rotor and sta-
sign from the engine systems standpoint. tor. These choices determine the meridional flow to first approxi-
mation.
Table 1: Amounts and Locations of Suction To initiate the blade design, multiple-circular arc blades were
for Low Tip Speed Stage assumed as a first approximation and a MISES calculation per-
formed in the inviscid mode. The blade leading and trailing edge
Suction Suction Scoop
Amount Position Mass angles, the locations of maximum thickness and the solidity (num-
Streamline Height Percentage
(% 0) (54. x/c) (x/c) ber of blades) were manipulated to arrive at an inviscid solution
with low shock losses and with the desired exit flow angles. In
Rotor hub --- --- --- ---
the process it was found that rather high solidities were required
Rotor 1/4 span 75 55 .0222 3.4
to avoid excessive turning in the inlet (supersonic) portion of the
Rotor 1/2 span 75 40 .0279 3.2
passage while meeting the turning requirement. Excessive turn-
Rotor 3/4 span 75 40 0521 4.9
Rotor tip 50 40 .0386 3.4 ing in the supersonic portion would lead to an increase in the
Stator hub 50 70 .0379 8.7 Mach number entering the passage and hence very large shock
Stator 1/4 span 50 75 .0379 7.2 losses. Once this inviscid design seemed reasonable, MISES cal-
Stator 1/2 span 50 80 .0501 8.3 culations were done with the boundary layer included, with suc-
Stator 3/4 span 75 80 .0507 8.2 tion at points judged to be appropriate from the inviscid calcula-
Stator tip 85 80 .0549 8.5 tions. The amount of fluid removal was determined as approxi-
mately the minimum that would prevent separation at each ra-
dius. Solutions that were satisfactory in this sense were obtained
The calculation also provides an estimate of losses, from at all points except the hub of the rotor, where 14 percent extrac-
which the efficiency can be computed. This is summarized for tion was required to prevent separation.
several streamsufaces in Table 2, where lc and nc are the stagna- The resulting blade shapes and their Mach number distribu-
tion temperature and pressure ratios. tions are shown for the mid span of the rotor and stator respec-
tively in Figs. 17 and 18. The locations of suction scoops and the
Table 2: Streamline Efficiency and Pressure Ratio mass removal fractions are summarized for the hub, mid-span
for Low Tip Speed Stage and tip of both the rotor and stator in Table 3.

Streamline lc Tic Itc TIpoiy

Hub 1.23 0.953 1.995 0.953

1/4 span 1.23 0.969 2.017 0.969 Table 3: Suction Parameters for Rotor and Stator
1/2 span 1.23 0.948 1.995 0.953
3/4 span 1.23 0.934 1.976 0.940 Section Location Slot Mass
Tip 1.23 0.919 1.957 0.926 Length Percentage
Removed
Average 1.23 0.945 1.991 0.950
r05 0.28 0.05 14.4
r09 (midspan) 0.248 0.05 2.25
r15 0.55 0.05 2.57
MT = 1.5, Pressure ratio = 3 s05 0.6 0.03 2.0
This design proved to be very demanding of the streamline- s09 (midspan) 0.7 0.05 4.41
curvature-MISES approach, because of the sensitivity of the ro- s15 0.75 0.05 3.0
tor behavior to the shock structure which is in turn very sensitive

Ii

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 2.0

1.5
Mach Number

1.0

0.5

Fraction of Chord Fraction of Chord

0.0080

0.0064
8
0.0048 00. 2

0.0032
1

0.0016
0
0 1
Fraction of Chord
Fraction of Chord

Fig. 17: Rotor blade surface Mach number and boundary layer Fig. 18: Stator blade surface Mach number and boundary layer
thickness variations for high tip speed stage with suction. thickness variations for high tip speed stage with suction.

CONCLUDING REMARKS
The potential improvements in pressure ratio per stage and Stage Experimental Evaluation of Boundary Layer Blowing and
in efficiency that may be attained with boundary layer removal Bleed Techniques for High Lift Stator Blades", NASA CR-54573.
are sufficient to warrant serious development of these possibili- Ng. W. F., and A. H. Epstein, 1985, "Unsteady Losses in Tran-
ties. The work reported here has quantified the thermodynamic sonic Compressors", ). Eng. for Gas Turbines and Power, Vol. 107,
effects, demonstrated the potential benefits through two prelimi- No. 2.
nary stage designs that take advantage of the increased diffusion Reijnen, D. P., 1997, "Experimental Study of Boundary Layer
enabled by boundary layer control, and most substantively, ex- Suction in a Transonic Compressor", PhD Thesis, Department of
perimentally demonstrated beneficial effects of boundary layer Aeronautics and Astronautics, MIT.
removal at the point of shock impingement in a transonic rotor. Smilg, L. M., 1994, "Design of a High Pressure Ratio Fan
The next logical step in development of these concepts is the de- Stage to Take Advantage of Boundary Layer Suction", SM The-
sign, construction and test of a full stage that exploits boundary sis, Department of Aeronautics and Astronautics, MIT.
layer control. Thompkins, W. T., 1976, "An Experimental and Computa-
tional Studey of the Flow in a Transonic Compressor Rotor - , PhD
REFERENCES Thesis, Department of Aeronautics and Astronautics, MIT.
Epstein, A. H., 1977, "Quantitative Density Visualization in a Youngren, H., 1991, "Analysis and Design of Transonic Cas-
Transonic Compressor Rotor", J. Eng Power, Vol. 99, pp 460-475. cades With Splitter Vanes", SM Thesis, Department of Aeronau-
Hearsey, R. M., 1975, "A Revised Computer Program for tics and Astronautics, MIT.
Axial Compressor Design", Aerospace Research Laboratories Ziminsky, W. S., 1996, "Design of a High Pressure Ratio Tran-
Report ARL-TR-75-0001, Wright Patterson AFB, Dayton Ohio. sonic Compressor Stage With Active Boundary Layer Control", SM
Loughery, R. J., R. A. Horn Jr., R. C. Tramm, 1971, "Single- Thesis, Department of Aeronautics and Astronautics. MIT.

12

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use