S. C.

Kacker A Mean Line Prediction Method

Staff Aerodynamicist.
Pratt & Whitney Aircraft of Canada Ltd., for Axial Flow Turbine Efficiency
U. Okapuu A mean line loss system is described, capable of predicting the design point ef-
ficiencies of current axial turbines of gas turbine engines. This loss system is a
Chief, Turbine Aerodynamics.
development of the Ainley I Mathieson technique of 1951. The prediction method is
Pratt & Whitney Aircraft of Canada Ltd., tested against the "Smith's chart" and against the known efficiencies of 33 turbines
Longueuil, Quebec, Canada of recent design. It is shown to be able to predict the efficiencies of a wide range of
axial turbines of conventional stage loadings to within ± 1'A percent.

Introduction
In choosing the gas path for a new turbine the designer has .10
to carry out an optimization study which involves the
8 0 = Ot2
calculation of velocity triangles. Blade lengths and radii are .08
thereby determined early in the design cycle, before blade
shapes are known. This is done by means of a "mean line"
velocity triangle calculation, which is based on the assumption .06
that the thermodynamic processes undergone by the working
fluid can be represented by velocity triangles at midspan. To
.04
produce an optimum gas path such a calculation must in-
corporate a system of aerodynamic losses expressed as a
function of blade row inlet and exit velocity triangles. Its .02 -
excellence is ultimately judged by its ability to predict the
aerodynamic efficiencies of known turbines of "competent"
.00
design.
0.0 0.2 0.4 0.6 08 10 1.2
Over the past 30 yr, a number of such turbine mean line loss
systems have been described in the open literature, Traupel S/C
[1], Craig and Cox [2], and Stewart [3]. Perhaps the best Fig. 1 Profile loss coefficient for /3-| = 0, JMAX' C = ° - 2 a " e r Ainley and
known and most completely documented of these is that due Mathieson [4]
to Ainley and Mathieson [4], published in 1951. Due to the
improvement in our analytical capability and the ac- .20 r" .70°= CX,
cumulation of test results on a variety of turbines, it is fitting •
s
that such a loss system be critically reviewed and updated at *
least once every decade. It is a tribute to the durability of the .16 -- ,65°
S
v\ ^ * *
Ainley/Mathieson system that it has become a foundation
worthy of subsequent refinement, the most notable of which .12
m.N ^ *>*'* ^> 60°
was published in 1970 by Dunham and Came [5]. Now, a ^ ^55°
decade later, may be an appropriate time to take a fresk look Yp ^ ^ ^ ^ 5 0 °
at the subject.
The present paper describes modifications to the
.08
^—r^-400
Ainley/Mathieson/Dunham/Came (AMDC) loss system.
These modifications are tested against experimental results .04 -
where possible. The complete loss system is finally tested
against design point efficiencies of 33 recent turbines. .00 i i i i i i

0.0 0.2 0.4 0.6 08 1.0 1.2

Overall Description of Loss System S/C
The loss of total pressure in a cascade of blades, expressed Fig. 2 Profile loss coefficient for /31 = a 2 . ' M A ' c = 0.2 after Ainley and
Mathieson [4]

in terms of cascade exit dynamic pressure, is assumed to be

Contributed by the Gas Turbine Division and presented at the International
Gas Turbine Conference and Products Show, Houston, Texas, March 9-12,
the sum of profile, secondary, trailing edge and tip leakage
1981, of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Manuscript losses, where the profile losses are corrected for Reynolds'
received at ASME Headquarters, December 9, 1980. Paper No. 81-GT-58. number effects

Journal of Engineering for Power JANUARY1982, Vol. 104/111

Copyright © 1982 by ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms
 INLET
ANGLE .28
GAS ANGLE o< 1
METAL ANGLE R* .24

> s .20
<
S © ©
.16

.12

t0 ~20 40 60 80 l^O 120 140 160

(3l + 32) DEGREES
GAS ANGLE <X2
Fig. 4 Thickness/chord ratio for typical turbine blade sections
EXIT METAL ANGLE B2
ANGLE
Fig. 3 Blade section terminology The Profile Loss Coefficient YP
The basis for the profile loss coefficient is a set of cascade
YT — F P / ( R e ) + Ys + YTET + YTC (1) test results incorporated in the AMDC loss system [4]. These
This formulation is seen to differ from that of the AMDC results are expressed in terms of pitch/chord ratio and cascade
system [4, 5], which is exit gas angle a2 for two special cases of /^ = 0 (Fig. 1) and
/3, = a2 (Fig. 2). For any other combination of angles, these
YT = [(YP + y s )REFAC + YTC] YTET (2) graphs are interpolated by means of the following equation:
where Y TET in Equ. (2) is a multiplier, not a loss coefficient.
In equation (1), blade Reynolds' number is taken to affect Y AY Ah\(h\\Y
only the profile loss coefficient, and the trailing edge loss <*2
coefficient is separated from the other loss terms. This would
appear a more logical arrangement, since it is difficult to - r w,=o,J](^-)«2 (3)
justify a connection between trailing edge losses and tip
clearance losses, for example. The foregoing equation is similar to the interpolation

bx axial chord
C = true chord Yr = total loss coefficient for a blade row
cL == airfoil lift coefficient YP = profile loss coefficient
c axial velocity component Ys = secondary loss coefficient
CFMx = supersonic drag rise multiplying ^SHOCK = component of profile loss coefficient
factor to profile loss coefficient YP due to leading edge shock
f(AR) = aspect ratio function YjET = trailing edge loss multiplier, in
/(Re)= Reynolds number correction factor AMDC loss system
h = blade or vane height 1
TET = trailing edge loss coefficient
AH = enthalpy drop YTC = tip clearance loss coefficient
k = tip clearance a = absolute/relative gas angles for
k' = equivalent tip clearance for shrouded vane/blade
blade <*m = mean gas angle defined in equation
K\ ,K2 ,A"3 ,Kp ,KS = correction factors defined in text (13)
M = Mach number 0 = metal angle for vane or blade
0 = throat opening 7 = ratio of specific heats
P = total pressure A<p T E T = trailing edge K.E. loss coefficient
P = static pressure Vt-t = turbine efficiency (total-to-total)
q = dynamic head (P-p) Vo = 1),., at zero tip clearance
REFAC = Reynolds number correction factor * = stagger angle
for AMDC loss system 4>2= kinetic energy coefficient = (actual
Rec = Reynolds number based on true gas exit velocity/ideal gas exit
chord and exit gas conditions velocity)2
R= radius Subscripts
s= pitch H = hub
U= blade velocity at mid height T = tip
t
= trailing edge thickness (TET) 1,2 = inlet and exit conditions
^MAX = blade maximum thickness Sub = subsonic
=
y = loss coefficient AP/q where q is AMDC as per Ainley, Mathieson, Dunham
(generally) taken at blade exit and Came

112 / Vol. 104, JANUARY 1982 Transactions of the ASME

Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms

 22 ROTORS

2.0

1.8

1.6

1.4 NOZZLES

1.2

1.0
H^ 5 .6 .7 8 .9 1.0
R H /R T - HUB/TIP RADIUS RATIO
Fig 6 Inlet Mach number ratio for nonfree-vortex turbine blades
-40 "20 0 20 40 60
BLADE SECTION INLET ANGLE P i DEGREES
Fig. 5 Stagger angle for typical turbine blade sections

equation given by the AMDC method except for the

terml|3i/a 2 1, which was introduced to allow for negative
inlet angle.
Figure 3 clarifies the terminology. Figure 4, a plot of
thickness/chord ratios of a number of recent blade designs,
proposes a general curve for this ratio, to be used when blade
sections are not yet designed. Figure 5, similarly of statistical
derivation, proposes values of stagger angle which can be used
to estimate true chords, c, from axial chords which are
generally established early in the design.
The cascade results of Figs. 1 and 2 are valid for vanes and
blades having a trailing edge thickness to pitch ratio (t/s) of
0.02. In the AMDC loss system YP is multiplied by Y TET to
obtain the combined profile and trailing edge loss at any other .2 A .6 .8
t/s ratio. In the present loss system, since Y' J^J- is a separate M-|,HUB
additive loss, YP from Figs. 1 and 2 is multiplied by 0.914, Fig. 7 Combined effect of leading edge shock on inner end wall flow
which is the value of YTET at t/s = 0 according to AMDC loss and the blade channel next to it
system, to obtain YP at zero trailing edge thickness.
The AMDC profile loss correlations, Figs. 1 and 2, were
perfectly valid at the time of publication of the original inner endwall of a blade can then be found from Fig. 7, which
AMDC loss system, and efficiencies of turbines of that era states
should be predictable by these data. However, advances in
aerodynamic analysis, made over the last three decades, = 0.75(M, , H U B - 0 . 4 ) ' (4)
suggest that a factor of 2/3 should be applied to these loss qx /HUB
coefficients for the efficiency prediction of present-day However, since this loss is only local, it would be unfair to
turbines. penalize the whole blade by this amount: a long blade would
be less affected, overall, by a loss here than would a short
The Subsonic Mach Number Correction. Cascade tests in blade. Therefore, the contribution to the mean line loss of
the decades following the publication of the AMDC loss blade inlet shock losses is taken to be
system have revealed that the profile loss coefficient YP is in
general not independent of Mach number, even in the sub- AP\
\ (5)
sonic flow regime. Compressibility can affect YP in two ways, qx /SHOCK \RTT/)
\R \V qxx /HUB
by causing shocks at blade leading edges and by affecting the
Finally, the subsonic shock loss coefficient, until now ex-
flow acceleration within blade channels.
pressed in terms of blade inlet mean line dynamic head, can be
Shock Loss. Shocks at blade leading edges can set in at expressed in terms of blade exit dynamic head by
relatively low average inlet Mach number levels, due to the
AP\ AP\
local flow acceleration adjacent to the highly curved leading • y*!
edges. Evidence of the possible detrimental effects of this can q 2 /SHOCK qx /SHOCK ( - )
be deduced from design rules of bygone days, which warn us \p2J
about designing for relative inlet Mach numbers greater than
approximately 0.6, and admonish us to design thinner leading
edges when inlet Mach numbers are high. (6)
Due to the radial variation in gas conditions, necessary for
the flow to be in equilibrium, incident Mach numbers are
always higher at the hub than at midspan. A sampling of
known turbines resulted in the curve shown in Fig. 6, which Channel Flow Acceleration. Flow in the passage formed by
estimates the incident Mach number at the hub of nonfree- two adjacent blades will undergo a larger velocity change
vortex turbine blades when midspan Mach number and when operated at a higher (subsonic) Mach number level. This
hub/tip radius ratio are known. The "shock loss" next to the is a consequence of the compressibility of the working

Journal of Engineering for Power JANUARY 1982, Vol. 104/113

Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms

 1.0
.10r 1

0.8 PRESENT PREDICTION /

AMDC LOSS SYSTEM J
0.6 .08 i
i

0.4
.06- o ( D A T A F R 0 M 4 DIFFERENT ;' A j ' o A
A (TRANSONIC CASCADE RIGS ,' J
0.2 Kt • 1-1.25|M2-0.2|FOfl Ma*0.2 ®) /A?
< .04
0.0 / LJ
/ r-f®
0.0 0.2 0.4 0.6 08 1.0
.02-
M, 9 0*°"
Fig. 8 Mach number correction factor K1 for the profile loss coef-
ficient, for accelerating cascades of M -, - M2 1
0. — i i i 1 1 1
Q2 0.4 0.6 0.8 1.0 1.2 1.4
M2(ISENTR0PIC)

Fig. 10 Comparison of transonic cascade test results with prediction,

blade profile 23

pears to be the major shortcoming of this correction in that it

ignores limit loading of airfoils.
Limit load is the condition at which a given cascade of
airfoils operating at a constant inlet pressure develops
maximum tangential force. When blade trailing edges are
more tangential, this condition occurs at a higher Mach
number, and therefore different cascades will have different
limit load exit Mach numbers. Limit load Mach number can
be estimated from blade geometry [6,7], but it can also be
Fig. 9 Mach number correction factor K2 for the profile loss coef- predicted from supersonic drag rise, if that is known [8].
ficient
Alternatively, the supersonic drag rise may be deduced from
known limit load Mach number calculated from blade
geometry. Work on this approach is currently in progress.
medium. In an accelerating flow passage, therefore, operation Early results look promising as shown in Fig. 10. This regime
closer to sonic exit velocities will tend to cause suppression of is as yet primarily of interest for efficiency prediction at off-
local separations and the thinning of boundary layers. This design operating conditions, however. Design point exit Mach
effect is most pronounced where inlet Mach numbers are only numbers of turbines in current gas turbine engines rarely
slightly lower than exit Mach numbers. assume magnitudes where supersonic drag rise plays a
Profile loss coefficients shown in Figs. 1 and 2 are derived significant part in the loss system.
from cascade tests carried out at low subsonic velocities and
are therefore pessimistic when applied to turbines operating at Reynolds' Number Correction. It is assumed that the
higher Mach number levels. Figures 8 and 9 correct for the profile loss coefficient YP is calculated at a reference
effects of exit Mach number and channel acceleration, Reynolds' number of 2 x 10 5 , based on true chord, and
respectively. The combined effect of these corrections is cascade exit gas conditions. At any other Reynolds' number,
KP = l-K2(l-Kl) (7) the correction / (Re) in equation (1) is
In the subsonic regime of operation of a cascade of blades,
therefore, the profile loss coefficient YP becomes: Rec
/((Re) ^ for Rec < 2 x l 0 5
\2xl05/
YP = 0 . 9 1 4 ( | YPAMDCKP + y S H oc K ) (8) = 1.0 for 2 x l 0 5 < R e c < 1 0 6 (10)
/ R e c \ -°- 2
As discussed in the forthcoming section, "Verification of for Rec>10 6
Loss System," these compressibility corrections were
necessary to make the AMDC loss system predict the correct
shapes of the efficiency islands on "Smith's chart." This formulation is similar but not identical to Denton's [9].
It is applied to the profile loss coefficient only as there is little
Supersonic Drag Rise. In the regime of supersonic exit evidence in the literature that other loss terms are affected.
velocities additional pressure losses occur as a result of shocks The Reynolds' number independence in the 2 x 105 to 106
originating in the trailing edge wake. Due to the lack of range is an approximation to the complex loss coefficient
cascade test results of adequate quality, a reliable loss model variation in the transitional regime. The true variation of loss
is not yet available. coefficient with Reynolds' number cannot be estimated
The AMDC loss system assumes a supersonic drag rise without a detailed knowledge of blade shapes.
according to
CFM = 1 + 6 0 ( M 2 - 1 ) 2 (9) The Secondary Loss Coefficient Ys
where CFM is applied as a multiplier to YP, when the exit The secondary loss coefficient calculation is the same as
Mach number exceeds unity. This factor is taken to be in- given by Dunham and Came [5], except for its dependence on
dependent of blade (metal) exit angle. This assumption ap- blade aspect ratio. In [5] it is assumed that Ys varies as the

114/Vol. 104, JANUARY 1982 Transactions of the AS ME

Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms

 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
Fig. 14 Trailing edge loss [energy] coefficient correlated against the
ratio of trailing edge thickness to throat opening Cx/U
Fig. 15 Turbine stage efficiency at zero tip leakage, after Smith [12]
trailing edge losses are interpolated in a manner similar to
equation (3), i.e.,
r
A<t>TET = &<t>TET(
Wi=0) case of unshrouded blades was that tip leakage losses were
A TET A found to be overpredicted by that system on recent turbines.
<*2
\£) [ ^ ^ =«2> - *TET (ei=0) ] (17)
Verification of Loss System
The conversion from kinetic energy loss coefficient to the
pressure loss coefficient is given by The loss system outlined in the foregoing evolved to its final
7 formulation largely thanks to a 1965 paper by Smith [12],
[.-^«(: 1
-A07TE
7-1 1 which proposes a correlation of turbine efficiency on a
"loading diagram" having stage loading factor AH/If1 as one
(18) axis and stage flow factor Cx/U as the other (Fig. 15). Design
1
y- , „ \ ~^r
- 0+ ^ * 0 7-1 point efficiencies (corrected to zero tip clearance) of some 69
turbines, plotted on this diagram, strongly suggest concentric
The variation of Y'TET with Mach number is not well locii of constant efficiency. This correlation is so convincing
documented at present. For supersonic exit velocities the that it must be accepted as one of the tests which any mean
supersonic drag rise, as discussed under profile losses, is line loss system must satisfy.
assumed to include trailing edge losses. These 69 turbines were, clearly, all designed before 1965. In
addition they all share certain common design features
The Tip Clearance Loss Coefficient YTp described in [12], features which are a reflection of the
consistent design philosophy in the organization whence they
Overtip leakage losses on unshrouded rotor blades con- originated. Considering the probable sizes of these turbines,
stitute a major source of turbine inefficiency. A large body of the state of the (design) art at the time and the manner of
experimental data exists on this subject, relating changes in testing, one arrives at the conclusion that the efficiency
turbine efficiency to changes in rotor tip clearance. Most of contours shown in Fig. 15 are caused mainly by profile losses
the available results correlated within ± 15 percent with the and somewhat more weakly (due to the prevailing moderate to
following expression: large aspect ratios) by secondary losses. Therefore, if a
ATJ candidate loss system fails to duplicate the Smith's chart, the
fault is likely to lie with its profile and/or secondary loss
Vo components.
= 0.93 (19)
Ak Ri To test these loss system components, the equivalent of 103
h cosa-, RK
turbine stages were designed, using design rules and
mechanical limitations current in the 1950's. As a guide,
When the loss system is structured as equation (1) Arj has to multistage turbine gas paths and approximate cycle design
be converted to an equivalent loss coefficient YTC. This may points of three engines of this period were used as a
be done iteratively (by first calculating turbine efficiency with framework for these stages. The design point efficiency of
zero tip clearance, assessing the efficiency penalty due to each was then calculated by a candidate loss system and a plot
clearance and increasing the rotor loss coefficient until the similar to the Smith's chart was constructed.
process converges on efficiency, all the while recalculating the The Ainley/Mathieson loss system, while predicting quite
velocity triangles and other loss coefficients) or directly, by well the variation of efficiency with AH/U2, failed to
making certain simplifying assumptions. The exact process is adequately predict its variation with Cx/U at Cxi U values in
beyond the scope of this paper. excess of 0.6. In this respect our results were quite similar to
In the case of shrouded turbine blades the loss system in- those of Amann and Sheridan [13]. After examining various
corporates the relationship proposed in [5], i.e. alternatives we were forced to conclude that only the com-
pressibility and shock loss corrections to profile and secon-
y ™ = 0.37- —
cos z a 2 dary loss coefficiencts, described in the foregoing, can
(20)
/Ac/ \s/c) cos3 a,„ produce an adequate simulation of Smith's chart. It was
where found that the shock loss correction (FSHOCK) mainly affected
the high Cx/U regime of the chart, while the acceleration
k' = (21) correction (Kp) affected the general level of the efficiency
(number of seals) 0M contours. Figures 6, 7, 8, 9 and 13 were evolved by trial and
The reason for deviating from the \ M D C loss system in the error until a simulation of Smith's curves was obtained. The

116/Vol. 104, JANUARY 1982 Transactions of the ASME

Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms

 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
Fig. 14 Trailing edge loss [energy] coefficient correlated against the
ratio of trailing edge thickness to throat opening Cx/U
Fig. 15 Turbine stage efficiency at zero tip leakage, after Smith [12]
trailing edge losses are interpolated in a manner similar to
equation (3), i.e.,
r
A<t>TET = &<t>TET(
Wi=0) case of unshrouded blades was that tip leakage losses were
A TET A found to be overpredicted by that system on recent turbines.
<*2
\£) [ ^ ^ =«2> - *TET (ei=0) ] (17)
Verification of Loss System
The conversion from kinetic energy loss coefficient to the
pressure loss coefficient is given by The loss system outlined in the foregoing evolved to its final
7 formulation largely thanks to a 1965 paper by Smith [12],
[.-^«(: 1
-A07TE
7-1 1 which proposes a correlation of turbine efficiency on a
"loading diagram" having stage loading factor AH/If1 as one
(18) axis and stage flow factor Cx/U as the other (Fig. 15). Design
1
y- , „ \ ~^r
- 0+ ^ * 0 7-1 point efficiencies (corrected to zero tip clearance) of some 69
turbines, plotted on this diagram, strongly suggest concentric
The variation of Y'TET with Mach number is not well locii of constant efficiency. This correlation is so convincing
documented at present. For supersonic exit velocities the that it must be accepted as one of the tests which any mean
supersonic drag rise, as discussed under profile losses, is line loss system must satisfy.
assumed to include trailing edge losses. These 69 turbines were, clearly, all designed before 1965. In
addition they all share certain common design features
The Tip Clearance Loss Coefficient YTp described in [12], features which are a reflection of the
consistent design philosophy in the organization whence they
Overtip leakage losses on unshrouded rotor blades con- originated. Considering the probable sizes of these turbines,
stitute a major source of turbine inefficiency. A large body of the state of the (design) art at the time and the manner of
experimental data exists on this subject, relating changes in testing, one arrives at the conclusion that the efficiency
turbine efficiency to changes in rotor tip clearance. Most of contours shown in Fig. 15 are caused mainly by profile losses
the available results correlated within ± 15 percent with the and somewhat more weakly (due to the prevailing moderate to
following expression: large aspect ratios) by secondary losses. Therefore, if a
ATJ candidate loss system fails to duplicate the Smith's chart, the
fault is likely to lie with its profile and/or secondary loss
Vo components.
= 0.93 (19)
Ak Ri To test these loss system components, the equivalent of 103
h cosa-, RK
turbine stages were designed, using design rules and
mechanical limitations current in the 1950's. As a guide,
When the loss system is structured as equation (1) Arj has to multistage turbine gas paths and approximate cycle design
be converted to an equivalent loss coefficient YTC. This may points of three engines of this period were used as a
be done iteratively (by first calculating turbine efficiency with framework for these stages. The design point efficiency of
zero tip clearance, assessing the efficiency penalty due to each was then calculated by a candidate loss system and a plot
clearance and increasing the rotor loss coefficient until the similar to the Smith's chart was constructed.
process converges on efficiency, all the while recalculating the The Ainley/Mathieson loss system, while predicting quite
velocity triangles and other loss coefficients) or directly, by well the variation of efficiency with AH/U2, failed to
making certain simplifying assumptions. The exact process is adequately predict its variation with Cx/U at Cxi U values in
beyond the scope of this paper. excess of 0.6. In this respect our results were quite similar to
In the case of shrouded turbine blades the loss system in- those of Amann and Sheridan [13]. After examining various
corporates the relationship proposed in [5], i.e. alternatives we were forced to conclude that only the com-
pressibility and shock loss corrections to profile and secon-
y ™ = 0.37- —
cos z a 2 dary loss coefficiencts, described in the foregoing, can
(20)
/Ac/ \s/c) cos3 a,„ produce an adequate simulation of Smith's chart. It was
where found that the shock loss correction (FSHOCK) mainly affected
the high Cx/U regime of the chart, while the acceleration
k' = (21) correction (Kp) affected the general level of the efficiency
(number of seals) 0M contours. Figures 6, 7, 8, 9 and 13 were evolved by trial and
The reason for deviating from the \ M D C loss system in the error until a simulation of Smith's curves was obtained. The

116/Vol. 104, JANUARY 1982 Transactions of the ASME

Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms

 30

2.6

2.2

^1.8
^***^ _*--f525 KZh**^ .912 \

1.4

1.01 J T / ///
0.6
0.4 0.5 0.6 0.7 0.8 0.9 10 1.1 1.2 1.3
cx/u
Fig. 16 Turbine stage efficiency at zero tip leakage calculated by the
present method

EFFICIENCY EFFICIENCY
DIFFERENCE DIFFERENCE
.025 TO . 0 3 0 EXPERIMENTAL EFFICIENCY LESS .025 TO .030 H 1
. 0 2 0 TO . 0 2 5 THAN EFFICIENCY FROM CONTOURS CALCULATED EFFICIENCY LESS
.020 TO .025
.015 TO . 0 2 0 THAN EFFICIENCY FROM CONTOURS
n .015 TO .020
. 0 1 0 TO .015
. 0 0 5 TO .010
12. .010 TO .015
1 14 .005 TO .010
0 TO . 0 0 5 0 TO .005 131
0 TO .005 0 TO 73oT T _
.005 TO .010
.010 TO .015 r? .005
.010
.015
TO . 0 1 0
TO .015
TO . 0 2 0
&&p*mi>#to*m$f"

EXPERIMENTAL EFFICIENCY MORE CALCULATED EFFICIENCY MORE

THAN EFFICIENCY FROM CONTOURS THAN EFFICIENCY FROM CONTOURS
Fig. 17 Histogram for experimental data used to define contours in Fig. 18 Histogram for calculated data used to define contours in Fig.
Fig. 15, after Amann and Sheridan [13] 16

result is shown in Fig. 16. Calculations were carried out with speculations aside, the quantitative agreement may never-
the whole AMDC value of YP (rather than 2/3) in equation theless be considered satisfactory in the light of the inevitable
(8) to simulate the state of blade design art of the time. data scatter inherent in both figures. Figure 18 shows the
A qualitative comparison of Figs. 15 and 16 indicates that a corresponding differences between data points and contours
satisfactory duplication of the shapes of efficiency contours of Fig. 16, for completeness.
has been achieved. At low values of AH/U2 there is The second test of a loss system is the demonstration of its
unquestionably an initial increase in efficiency with AH/U2, ability to predict the design point efficiencies of existing
at constant Cx/U. The lack of this feature has been the major turbines of more recent vintage. Design point data were
shortcoming of the AMDC loss system as demonstrated in collected for 33 turbines ranging from the gas generator
Fig. 9 of [13]. turbine of a small automobile engine to the low pressure
Quantitatively, Fig. 16 differs from Fig. 15 in that a given turbine of a 45,000-lb thrust turbofan. In view of the ad-
efficiency curve of Fig. 16 attains a lower peak AH/U2 value vances made in blade profile design over the last 25 yr, the loss
but extends to a higher peak Cx/U value. At 90 percent ef- system was operated with only 2/3 of the AMDC profile loss
ficiency, for example, the maximum discrepancy amounts to coefficient. Since none of these turbines had bladerow exit
1 Vi efficiency points in each regime, while at 94 percent the Mach numbers in excess of 1.17 the supersonic drag rise
discrepancy is Vi efficiency point. This is partly explained by correction (by the present prediction, Fig. 10) had only a
an examination of the data from which efficiency contours small effect on the efficiencies of transonic stages. Figure 19
were derived. Figure 17 describes the differences between data shows the comparison. The closed symbols describe P&WA
points and contours of Fig. 15, as reported in [13], It is clear turbines. The open symbols describe turbines from other
that a considerable uncertainty exists in the positioning of sources. Table 1 lists the more important design parameters of
efficiency contours. It is also evident that a bias exists in the these turbines.
contours, which are predicting efficiencies approximately Vi The agreement is generally very good, the majority of
point higher than supported by data points. This would tend predictions falling within an error band of 1 Vi efficiency
to reduce the discrepancy between Figs. 15 and 16, at peak points. Of the few turbines which fall significantly outside this
AH/U2. As to the region of peak Cx/U, Fig. 15 has too few error band two can readily be shown to be "incompetent"
data points at Cx/U > 1.0 to substantiate the curves drawn. designs, and the experimental techniques used in the case of
Unpublished test data from other sources suggest that these two more are open to criticism.
curves may in fact extend to higher Cx/U, as in Fig. 16. These Figure 20 shows an identical comparison of the same 33

Journal of Engineering for Power JANUARY 1982, Vol. 104/117

Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms

 .95
PRESS. V cc
NO. 1 SOURCE STAGES STAGE MVU2 Cx/O 1
RATIO j NOZZLE
• PWA TURBINES ROTOR

O OTHER TURBINES 1 PSWA 1 1 1.88 .54 2.67 .366 1.75

>•
2 ; PSWA 1 1 1.51 .533 4.00 .612 1.98
%
W . 9 0 L- **>. 3 PSWA 1 1 1.45 .565 3.89 .571 1.72

U 4 PSWA 1 1 1.55 .570 2.95 .437 2.188

u. •6. 5 OTHER 1 1 1.25 .613 2.98 .520 .671
iii 27®.23
6 OTHER 1 1 1.55 .675 2.90 .55
50 .825
a e / 7 OTHER 1 1 1.59 .642 2.90 .725 1.26
IS Oio/ Pa / 8 OTHER 1 1 1.48 1.50
1.11 .38 3 1.50

«.85l- ° 5 1 f ^ V • ' On 9 OTHER 1 1 1.656 .595 2.77 1.0 1.0

g '1 '
10 OTHER 2 1
2
.989
.987
.309
.318
1.11
1.11
1.44
1.32
1 . 4 56
1.489
Q.
11 OTHER 1 1 1.672 .696 1.82 .71 1.01

12 PSWA 1 1 1.19 .397 2.00 .317 .739

13 PSWA 1 1 1.19 .531 2.00 .247 .777

.80 ^ 14 PSWA 1 i 1.19 . 4 97 2.00 .253 .670

- i t'N
15 PSWA 1 I 1.139 .322 3.40 .24 .486
: j _ _i_
.80 .85 .90 16 OTHER 2 I 1.65 .690 1.71 1.47 1.87
.<« 2 1.65 .813 1.87 1.54 2.16
EXPERIMENTAL EFFICIENCY 17 OTHER 1 1 2.00 .721 1.99 .468 .847
Fig. 19 Comparison of predicted efficency with experimental ef- 18 P&WA 1 1 1.44 .566 2.76 .637 1.9
ficiency of 33 turbines (new loss system) 19 PSWA 2 1 1.50 .924 1.45 1.27 2.09
2 1.77 .990 1.65 1.95 2.47

20 PSWA 2 1 .9 .622 1.44 1.102 2.23

.95 2 .83 .502 1.53 1.81 3.02

21 PSWA 2 2 1.28 .758 1.70 1.102 2.23

% PWA TURBINES 1 1.02 .698 1.73 1.81 3.02

22 PSWA 1 1 1.94 .529 2.24 1.23 2.67

O OTHER TURBINES
23 PSWA 1 1 2.388 .712 2.68 1.23 2.67

24 PSWA 1 1 1.45 .565 3.89 .571 1.08

25 PSWA 1 1 1.45 .561 3.89 .984 1.72

26 PSWA 1 1 1.516 .583 2.81 .421 1.95

27 PSWA 1 1 1.463 .492 3.23 .527 2.155

28 PSWA 1 1 1.38 .485 1.99 .710 .921

29 PSWA 1 1 1.58 .545 1.94 1.328 1.111

30 PSWA 4 1 2.44 1.50 1.33 2.61 3.94

31 " 2 2.32 1.28 1.35 4.07 4.98

32 " 3 2.38 1.25 1.45 5.01 5.76

33 " 4 2.71 1.50 1.68 5.68 5.80

novel techniques to minimize overtip leakage losses and fully

three-dimensional flow analysis methods are already showing
promise of a bright future.
A mean line efficiency prediction method is the sum of a
large number of loss components. While some of them may
prove to be quantitatively imperfect, the manner in which
they are combined may cause errors to cancel. The final proof
.75- of a loss system must be its ability to correctly predict the
efficiencies of well documented turbines.
Oe
_1_ »13 -1_ _1_ Conclusion
.80 .85 .90 .95 Regardless of the degree of sophistication of analytical
EXPERIMENTAL EFFICIENCY tools available for the detail design of turbine blading, a mean
Fig. 20 Comparison of predicted efficiency with experimental ef- line design technique will always be needed for the op-
ficency of 33 turbines (AMDC loss system) timization of the turbine gas path and the prediction of at-
tainable turbine efficiency of a new engine.
The mean line loss system described, a development of the
Ainley/Mathieson/Dunham/Came system, appears to be
turbines carried out with the AMDC loss system. The capable of predicting design point efficiencies of current
agreement is not satisfactory. turbines of conventional stage loadings and "competent"
The design point efficiencies under discussion here are design to within ± 1 Vi efficiency points. A major departure
those obtainable on rotating aerodynamic rigs. In the engine from AMDC is the restructuring of the loss system and the
environment these turbines will generally show a lower ef- introduction of compressibility effects and shock losses into
ficiency for a variety of reasons such as effects of secondary the calculation of profile and secondary loss coefficients.
air injection, leakages, nonuniformities in the turbine inlet
flow, etc. These features are not accounted for the in the loss Acknowledgments
system described in the foregoing. Also not accounted for are
the inevitable future advances in the art of turbine design: The work reported herein has been partially funded by the

118/Vol. 104, JANUARY 1982 Transactions of the ASME

Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms

 Research and Development Centre of Transport Canada, 6 Hauser, C , and Plohr, H.W., "Two-Dimensional Cascade Investigation
of the Maximum Exit Tangential Velocity Component and Other Flow Con-
Contract No. 107380 under the program management of W. ditions at the Exit of Several Turbine Blade Designs at Supercritical Pressure
H. D. Hanchet. Technical contributions by V. Ozarapoglu, K. Ratios," NACARME51F12, Aug. 1951.
K. Lam and M. Mueller are hereby also acknowledged. 7 Fruchtman, I.. "The Limit Load of Transonic Turbine Blading," ASME
Paper No. 74-GT-80, Apr. 1974.
8 Okapuu, U., "Limit Load and Supersonic Drag Rise of Transonic
References Turbine Cascades," Pratt & Whitney Aircraft of Canada, Internal Memo, July
1974.
1 Traupel, W., "Thermische Turbomaschinen," Springer-Verlag, Berlin, 9 Denton, J.D., "A Survey and Comparison of Methods for Predicting the
1966. Profile Loss of Turbine Blades," Institute of Mechanical Engineers Conference
2 Craig, H.R.M., and Cox, H.J.A., "PerformanceEstimateof Axial Flow Publication 3, Paper C76/73, 1973, pp. 204-212.
Turbines," Proceedings of the Institution of Mechanical Engineers, Vol. 185, 10 Rogo, C , "Experimental Aspect Ratio and Tip Clearance Investigation
No. 32, 1971, pp. 407-424. on Small Turbines," SAE Paper 680448, Mid Year Meeting Detroit, Mich.,
3 Stewart, W.L., "A Study of Axial Flow Turbine Efficiency Charac- May 20-24, 1968.
teristics in Terms of Velocity Diagram Parameters," ASME Paper No. 61-WA- 11 Okapuu, U., "Some Results from Tests on a High Work Axial Gas
37, 1961. Generator Turbine," ASME Paper No. 74-GT-81, Apr. 1974.
4 Ainley, D.G., and Mathieson, G.C.R., "A Method of Performance 12 Smith, S.F., " A Simple Correlation of Turbine Efficiency," Journal of
Estimation for Axial Flow Turbines," British ARC, R&M 2974, 1951. Royal Aeronautical Society, Vol. 69, July 1965, pp. 467-470.
5 Dunham, J., and Came, P.M., "Improvements to the Ainley/Mathieson 13 Amann, C.A., and Sheridan, D.C., "A Comparison of Some Analytical
Method of Turbine Performance Prediction," ASME JOURNAL OF and Experimental Correlations of Axial Flow Turbine Efficiency," ASME
ENGINEERING FOR POWER, July 1970, pp. 252-256. Paper No. 67-WA/GT-6, Nov. 1967.

Journal of Engineering for Power JANUARY 1982, Vol. 104 /119

Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 05/10/2014 Terms of Use: http://asme.org/terms