Blade Loading and Its

Application in the Mean-Line

Design of Low Pressure Turbines
In order to minimize the number of iterations to a turbine design, reasonable choices of
the key parameters must be made at the preliminary design stage. The choice of blade
loading is of particular concern in the low pressure (LP) turbine of civil aero engines,
where the use of high-lift blades is widespread. This paper considers how blade loading
should be measured, compares the performance of various loss correlations, and explores
the impact of blade lift on performance and lapse rates. To these ends, an analytical
John D. Coull1 design study is presented for a repeating-stage, axial-flow LP turbine. It is demonstrated
e-mail: jdc38@cam.ac.uk
that the long-established Zweifel lift coefficient (Zweifel, 1945, The Spacing of Turbo-
machine Blading, Especially with Large Angular Deflection Brown Boveri Rev., 32(1),
Howard P. Hodson pp. 436444) is flawed because it does not account for the blade camber. As a result the
e-mail: hph1000@cam.ac.uk Zweifel coefficient is only meaningful for a fixed set of flow angles and cannot be used as
an absolute measure of blade loading. A lift coefficient based on circulation is instead
Whittle Laboratory, proposed that accounts for the blade curvature and is independent of the flow angles.
University of Cambridge, Various existing profile and secondary loss correlations are examined for their suitability
1 J.J. Thomson Ave, to preliminary design. A largely qualitative comparison demonstrates that the loss corre-
Cambridge CB3 0DY, UK lations based on Ainley and Mathieson (Ainley and Mathieson, 1957, A Method of Per-
formance Estimation for Axial-Flow Turbines, ARC Reports and Memoranda No. 2974;
Dunham and Came, 1970, Improvements to the Ainley-Mathieson Method of Turbine
Performance Prediction, Trans. ASME: J. Eng. Gas Turbines Power, July, pp.
252256; Kacker and Okapuu, 1982, A Mean Line Performance Method for Axial Flow
Turbine Efficiency, J. Eng. Power, 104, pp. 111119). are not realistic, while the profile
loss model of Coull and Hodson (Coull and Hodson, 2011, Predicting the Profile Loss
of High-Lift Low Pressure Turbines, J. Turbomach., 134(2), pp. 021002) and the sec-
ondary loss model of (Traupel, W, 1977, Thermische Turbomaschinen, Springer-Verlag,
Berlin) are arguably the most reasonable. A quantitative comparison with multistage rig
data indicates that, together, these methods over-predict lapse rates by around 30%,
highlighting the need for improved loss models and a better understanding of the multi-
stage environment. By examining the influence of blade lift across the Smith efficiency
chart, the analysis demonstrates that designs with higher flow turning will tend to be less
sensitive to increases in blade loading. [DOI: 10.1115/1.4006588]

1 Introduction As the stage loading coefficient is increased (moving from the

bottom left to the top left of Fig. 1), the change in tangential ve-
1.1 The Smith Efficiency Chart. Perhaps the most famous locity across the blade row becomes larger. To achieve this addi-
correlation for turbine efficiency is that proposed by Smith [1]. tional turning one must either employ blades with higher
Using a large set of turbine test-rig data, he calculated the equiva- circulation (loading), or decrease the pitch (while maintaining cir-
lent efficiency for each turbine with zero tip gap. Plotting the culation). Both changes tend to increase loss, as does the high exit
results against stage loading coefficient and flow coefficient, he flow angle from these designs (see Eq. (14) below).
obtained the efficiency chart shown in Fig. 1 for designs with 50%
reaction. This plot shows that the efficiency tends to decrease as
1.2 Motivation and Scope of the Current Work. The Smith
either the stage loading or the flow coefficient are increased. To
chart provides a simple guide for the selection of flow angles, but
illustrate the changes across the Smith chart design space, four
it does not capture the influence of other key parameters. Modern
diagrams indicating the approximate flow angles and possible
LP turbines typically feature high-lift blade designs, which can
blade shapes have been added.
suffer a rapid drop in efficiency at the low Reynolds number con-
As the flow coefficient is increased (moving from the bottom
ditions experienced at cruise [8]. Ultimately, it is highly desirable
left to the bottom right of Fig. 1), the through-flow velocities
to accurately capture these effects at the earliest stages of design.
increase relative to the blade speed as the flow becomes more
The simplest model for turbomachine design is mean-line mod-
axial. For a given change in tangential velocity, this causes an
eling, where the flow is represented by the mean flow quantities at
increase in the dynamic pressures which tends to increase the
each inter-blade gap. Due to its simplicity, this analysis is well
losses. Furthermore, losses increase because there is a smaller
suited to understanding trends across the design space, but it must
overall acceleration through the blade row. Such acceleration is
rely on empirical correlations for profile, secondary and tip leak-
beneficial as it tends to minimize the growth of the boundary
age losses. Notable examples of loss correlations include those of
layers and secondary flows.
Craig and Cox [9], Traupel [7] and Ainley and Mathieson [3],
which was later updated by Dunham and Came [4] and Kacker
1
Corresponding author. and Okapuu [5]. While these models take account of blade load-
Contributed by the International Gas Turbine Institute (IGTI) of ASME for publi-
cation in the JOURNAL OF TURBOMACHINERY. Manuscript received August 24, 2011;
ing, none are influenced by the shape of the blade surface pressure
final manuscript received October 6, 2011; published online November 8, 2012. distributions, which can have a strong influence on LP turbine per-
Editor: David Wisler. formance (e.g., Refs. [10,11]). A profile loss model was recently

Journal of Turbomachinery Copyright V

C 2013 by ASME MARCH 2013, Vol. 135 / 021032-1

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 Table 1 Design and flow parameters for the datum repeating-
stage turbine; * indicates a parameter that remains constant
in the subsequent design space study

Flow Angles

Reaction (K) 50%*

Flow Coefficient (/) 0.9
Stage Loading Coefficient (w) 2
Repeating Stages Assumed*
Non-dimensional flow parameters
Stator Reynolds number (ReS0 ) 250,000
Equivalent ReC 194,000
Rotor Reynolds number (ReS0 ) 177,000
Equivalent ReC 138,000
Stator reduced frequency (fr ) 1.31
Rotor reduced frequency (fr ) 0.67
Geometric Parameters
Fig. 1 Turbine stage efficiency normalized for zero tip gap, Hub-to-tip ratio (average for stage) 0.75
50% reaction designs, Smith [8] Rotor Aspect Ratio (h=Cx ) 6.50*
Stator Aspect Ratio (h=Cx ) 4.60*
Trailing edge thickness (tTE =C) 0.01*
developed for high-lift LP turbine blades by the current authors to Zero Tip gap 0*
capture such effects [6]. Mean radius constant through the stage*
This paper outlines a mean-line design study for a repeating- Vx constant through the stage*
stage LP turbine, making use of published loss correlations. The Blade Velocity Distributions
analytical methods are described in Sec. 3. Section 4 examines Circulation Coefficient (Equation (21)) 0.70
how blade loading should be quantified and demonstrates that the Peak Velocity Location (fraction of S0 ) 45%*
standard Zweifel lift coefficient [2] is unsuitable for this purpose. Leading Edge Integral 50%*
Section 5 evaluates various profile and secondary loss correla- Diffusion Factor (DF) 0.21
tions. This analysis first concentrates on the qualitative trends pre-
dicted for varying flow angles, loading and Reynolds number
before a quantitative comparison of lapse rates is performed. power output and shaft speed are fixed by the requirements of the
Finally, Sec. 6 examines the sensitivity of the Smith efficiency fan. The designer must choose the number of stages for the LP tur-
chart to increases in blade loading. bine, which sets the power output per stage. When comparing dif-
ferent turbine designs, one must therefore consider designs with
3 Analytical Methods for Design Study the same stage power output, mass flow rate and shaft RPM as the
datum case, for the same inlet conditions. As shall be demon-
3.1 Datum Turbine. In order to provide a reference for the strated, this requirement results in turbines of different mean ra-
subsequent analysis, a datum turbine design is considered that is dius and flow area as the flow angles are varied.
approximately representative of modern LP turbine designs. The
details of the design are presented in Table 1. The datum turbine 3.3 Flow Parameters and Gas Angles. Figure 2 shows a
is a repeating stage with 50% reaction. The flow coefficient (0.9) schematic of the two-dimensional flow through a turbine stage for
and stage loading (2) place the datum design approximately in the constant axial velocity. It also defines the sign conventions for the
middle of the Smith chart design space (Fig. 1). The Reynolds flow angles.
numbers are roughly representative of modern LP turbines at In this study the stage reaction is fixed at 0.5 (Table 1), so:
cruise conditions. The reduced frequency fr describes the fre-
quency of wakes arriving from the upstream blade row, which can a1 b2 ; a2 b3 (1)
be accounted for in the profile loss model of Ref. [5]. The influ-
ence of this parameter is largely second-order, and will not be dis- The stage is repeating, so that:
cussed further in this paper, though its effect has been included in
the analysis. The specified blade aspect ratios are based on the a3 a1 (2)
axial chord, since the designer will be interested in the overall
stage length. The chosen aspect ratios set the ratio of stator to The gas angles are therefore fixed by the choice of Flow Coeffi-
rotor blades at approximately 1:1.4. The stage has a constant cient / and Stage Loading Coefficient w:
mean radius and the area ratio through the stage expands to main-
tain constant axial velocity though the stage. The tip gap is
assumed to be zero throughout this analysis, which allows direct
comparison with the efficiencies on the Smith chart. For simplic-
ity, the analysis here has been performed at relatively low speed
and the Mach number is not included in Table 1 since it has only
a weak influence on performance. While analysis at higher speeds
shows that Mach number variations do have an influence on effi-
ciency, this does not affect the conclusions that can be drawn
from the analysis.

3.2 A Study of Comparable Designs. The current design

study has been conducted with the industrial design process in
mind. The multistage LP turbine has three key overall require-
ments which are set by the engine architecture: the core mass flow
rate is determined by the engine bypass ratio, while the required Fig. 2 Velocity triangles and angle conventions

021032-2 / Vol. 135, MARCH 2013 Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 w 2 0:5  / tan a1 (3)
w / tan a2  tan a1 (4)

3.4 Calculation Procedure. For a fixed stage power output

w_ x and mass flow rate m,
_ the blade speed U is set by the stage
loading coefficient:
s
w_ x
U (5)
m_ w

Noting that one is considering designs for a fixed rotational shaft

speed, the mean radius is thus given by:
s
U 1 w_ x
rmean (6)
xs xs m_ w

The flow coefficient sets the axial velocity: Fig. 4 Geometry estimation using a parabolic camberline
s
w_ x turning airfoils. It is first assumed that the camber-line of a blade
Vx / U / (7) can be approximated by a parabolic curve, which was shown to be
m_ w
reasonable by Horlock [12]:

Together with the flow angles, the inter-blade velocities can now ycamber A x2 B x (10)
be calculated:
The basic approach is illustrated in Fig. 4. The camber-line of the
Vi Vx = cos ai (8) blade is assumed to align with the flow angles at inlet and exit,
Wi Vx = cos bi (9) uniquely setting the constants A and B in Eq. (10). The length of
the camber-line b and the true chord C may then be calculated
where the index i represents each inter-blade gap (Fig. 2). Figure from the known axial chord Cx . The length of the blade suction
3(a) shows the variation of the exit velocity from the stator row surface S0 will largely depend on the camber-line of the blade,
(normalized by the datum value), which is largest for designs with and so the following simple relationship was proposed:
high flow coefficient and low stage loading.
Compressible flow relationships are then used to determine the S0 1:15 b (11)
Mach numbers, and hence the total and static values of pressure,
temperature and density at the inlet and exit of each blade row. To The factor of 1.15 is based on a survey of LP turbine blades
correctly calculate the total pressures throughout the stage, the from the literature. Figure 3(c) shows the ratio of this estimated
loss coefficients must be known: the calculation is therefore iter- surface length to the axial chord over the design space. In reality
ated to arrive at the correct result. the thickness distribution of the turbine blade will also make a
Together with the specified mass flow rate, the calculated den- contribution to the surface length, but Eq. (11) is sufficiently accu-
sity determines the flow area, which expands through the machine rate for the current purposes.
to maintain constant axial velocity. The flow area determines the For profile losses, the appropriate Reynolds number is based on
span: Fig. 3(b) shows the variation over the design space. The the suction surface length and exit velocity ([5]). Figure 3(d) dem-
span is largest for designs with high stage loading and low flow onstrates that this parameter is large for highly turning designs;
coefficients, which have low axial velocity. Together with the the variation in suction surface length (Fig. 3(c)) therefore domi-
specified aspect ratios (Table 1), the axial chords of the stators nates the variation in exit flow velocity (Fig. 3(b)) in setting the
and rotors are thus determined. Reynolds number.
It is necessary at this point to make some estimates of the blade
geometry, since some loss correlations rely on true chord, 3.5 Surface Velocity Distributions, Pitch and Loss
camber-line length and surface lengths. The relationships between Prediction. One of the profile loss models considered in this pa-
these geometric parameters will be very different for high and low per is based on the preliminary design correlation previously

Fig. 3 Calculated flow and geometry parameters across the design space: (a) stator exit velocity; (b) mean span; (c) ratio of
S0 =Cx ; (d) Reynolds number

Journal of Turbomachinery MARCH 2013, Vol. 135 / 021032-3

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 developed by the authors for high-lift blades featuring laminar used in the current study because there is insufficient data to pre-
separation bubbles [6]. This method relies on an empirical correla- dict this velocity depression for low or high turning designs.
tion for the boundary layer parameters at the suction surface trail- Instead, it has been assumed that the freestream velocity at the
ing edge, based on the freestream velocity distribution over the trailing edge is equal to the mean exit flow velocity from the blade
surface. Effectively, a Thwaites [16] calculation (or equivalent) is row. This assumption will have only a minor impact on the results
performed up to the point of laminar separation; a correlation then of this study.
relates the growth in momentum thickness through the separation In addition to the Coull and Hodson method, this paper consid-
bubble and downstream turbulent boundary layer. A separate cor- ers several other profile loss models ([35,9,13]), and secondary
relation predicts the shape factor and hence the trailing edge dis- loss models ([35,7,9,17]).
placement thickness.
Coull and Hodson [6] showed that, with certain assumptions,
the suction surface velocity distribution can be modeled by three 3.6 Relating Loss Coefficients to Stage Efficiency. It
key design parameters: the diffusion factor, the peak velocity should be noted that most of the loss correlations are based on
location on the surface and the leading edge integral, which total pressure loss coefficients obtained at low Mach numbers.
describes the loading up to the velocity peak and is defined in Such total pressure losses tend to have a relatively strong Mach
Eq. (12). number dependence, and this effect would therefore need to be
1  5   accounted for in the analysis. Fortunately this issue can be largely
Vs S avoided by using the energy loss coefficient, defined for the stator
Leading Edge Integral LEI d (12)
0 Vspeak Speak row as:
2
Comparison with the method of Thwaites [16] demonstrates that V2;is  V22
1 2
 Y for incompressible flow (15)
this parameter is proportional to the square of the momentum V2;is
thickness at the peak velocity location [6]. In this study the peak
velocity was fixed at 45% of the surface, which is relatively far where V2;is is the exit velocity obtained by isentropic expansion
forwards on the surface. Front-loading in this manner reduces the from the inlet total pressure to the exit static pressure. The energy
deceleration rate over the rear portion of the blade and minimizes loss coefficient has a much lower dependency on Mach number
the losses generated by the separation bubble. The leading edge than the total pressure loss coefficient, and may therefore be reli-
integral was fixed at 50%, which should be low enough to provide ably used for flows with a peak Mach numbers below around 0.85
reasonable incidence tolerance. The diffusion factor was varied to ([18,19]). This is an approximation, but it is sufficiently accurate
achieve the required blade loading. Sample velocity distributions for this study.
of the style assumed are shown in Fig. 5. The total-total isentropic efficiency for the repeating stage is
The blade pitch is related to the gas turning and the blade circu- defined as:
lation, e.g., for the stator:
    h01  h02 T01  T02
Vs S g (16)
d h01  h02;is T01  T02;is
s V2 S0
(13)
So tan a1  tan a2 cos a2 4 Lift Coefficients
The profile loss may be related to the predicted trailing edge 4.1 Zweifel Lift Coefficient. The Zweifel lift coefficient [2]
boundary layer parameters using an approximation to the analyti- is a widely-used measure of blade loading. It is defined as the ratio
cal solution of Denton [13]: of the tangential force on a blade to an ideal case, where the
flow on the pressure surface of the blade is stagnated (P P0 ),
 P  P   while the flow over the suction surface travels at the mean exit ve-
2 hTE dTE tTE 2
Yp  (14) locity from the blade row (P P2 ). The Zweifel coefficient is
s cos a2 s cos a2
therefore given by:
In Coull and Hodson [6], a correction was applied to Eqs. (13)
and (14) to account for the depression of the freestream velocity P dx
in the vicinity of the trailing edge. This correction has not been Zw (17)
Cx P01  P2

From a control volume analysis, the tangential force can be

related to the flow angles, giving:

m_ passage DVh
Zw (18)
hCx P01  P2

These equations can be simplified for the case of incompressible

flow. From Eq. (17), the Zweifel coefficient can be calculated by
integrating the square of the freestream velocity over the blade
surfaces:

0:5 q Vs2 dx
Zw Vs =V2 2 d x=Cx (19)
Cx 0:5 q V22

This definition is illustrated graphically in Fig. 6(a). For low

Mach numbers and constant axial velocity, Eq. (18) may be
Fig. 5 Sample surface velocity distributions rewritten as:

021032-4 / Vol. 135, MARCH 2013 Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 Fig. 6 Low speed definition of: (a) Zweifel lift coefficient Zw (Eq. (19)); (b) Circulation
coefficient Co (Eq. (21))

Fig. 7 Predicted efficiency for constant Zw 1:10 using the Fig. 8 Diffusion factor for constant Zw 1:10 using the models
models of Refs. [6] and [7] of Refs. [6] and [7]

to undergo turbulent separation. At the other end of the design

  space, the diffusion factor is only around 0.05, and these designs
m_ passage DVh s tan a2  tan a1 will therefore tend to exhibit attached laminar flow over the full
Zw (20)
Cx 0:5 q V22 Cx 0:5 sec2 a2 length of the suction surface. (The reason for the blanked region in
Fig. 7 is that the profile loss model is not valid for permanently
Mean-line efficiency predictions have been performed for a attached flow.) Therefore, the same Zweifel coefficient simultane-
constant value of Zweifel coefficient (Zw 1:1) and are presented ously describes very highly loaded blades suffering turbulent sepa-
in Fig. 7. The profile and secondary loss models used are those of ration, very lightly loaded blades with laminar attached flow, and
Coull and Hodson [6] and Craig and Cox [9], respectively. (The everything in-between. The Zweifel coefficient is therefore an inap-
justification for using these methods is presented in the following propriate parameter to compare designs with different flow angles.
section.) There are some similarities to Smiths efficiency chart
(Fig. 1), in particular the decrease of efficiency with increased
stage loading. However, the decrease of efficiency with increasing 4.2 Circulation Coefficient. A new lift coefficient based on
flow coefficient (towards the right of Fig. 1) has not been cap- the blade circulation is proposed. In a manner analogous to the
tured. There is also a region in the bottom right without data, the Zweifel coefficient, the Circulation Coefficient Co is defined as
reason for which will be made clear in the following plot. For con- the ratio of the blade circulation to an ideal circulation, with
stant a Zweifel coefficient, the trends in the Smith chart could not Vs V2 on the suction surface and stagnated flow (Vs 0) on the
be satisfactorily reproduced using any of the loss correlations con- pressure surface:
sidered in this paper.
Figure 8 shows the variation in diffusion factor (DF) across the Vs dS    
Smith-chart design space; this parameter varies to maintain a con- actual circulation Vs S
Co d (21)
stant Zweifel coefficient as the flow angles change. The plot ideal circulation V2 S0 V2 S0
shows that the diffusion factor varies significantly across the
design space. The maximum value of around 0.6 occurs for the This low-speed definition is illustrated graphically in Fig. 6(b). A
designs with the highest turning (low /, high w), which are likely suitable scaling for compressible flow must also be considered.

Journal of Turbomachinery MARCH 2013, Vol. 135 / 021032-5

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 Several authors (e.g., Refs. [18] and [19]) have demonstrated that,
for peak Mach numbers less than around 0.85, a given high-speed
blade will have the same performance as a low-speed blade with
the same exit angle and a matching surface distribution of the
pressure coefficient:
P01  P
Cp (22)
P01  P2
Since the high-speed blade and its low-speed equivalent have the
same performance, they must have the same circulation coeffi-
cient. For the equivalent low-speed blade, one notes that:
 2
P01  P Vs
Cp (23)
P01  P2 V2

The equivalent low-speed blade therefore has a circulation

coefficient:
r    
P01  P S p S
Co d Cp d (24)
P01  P2 S0 S0

The distribution of Cp is identical for the high-speed blade and its

low-speed equivalent, so this definition is appropriate for high-
speed blades. Fig. 9 Predicted stage efficiency for constant circulation coef-
At low Mach numbers and constant Vx the circulation may be ficient Co 0:70 using the models of Refs. [6] and [7]
related to the flow angles and pitch (Eq. (13)) to give:
  cannot be directly evaluated without a large experimental dataset,
s tan a1  tan a2 any successful model should at the very least be able to predict
Co (25)
S0 sec a2 the correct trends across the design space. The following sub-
sections examine the influence of flow angles on profile and sec-
There are clear similarities between the circulation and Zweifel ondary losses respectively. This analysis is performed for the case
coefficients. Comparing Eqs. (20) and (25), the two coefficients shown in Fig. 9, with a constant circulation coefficient of
are related by the following parameter: Co 0:70. The sensitivity of the various models to blade loading
  and Reynolds number are then examined. Finally a quantitative
Co 0:5 Cx comparison between predicted and measured lapse rates of multi-
(26) stage turbines is performed to examine absolute accuracy.
ZW S0 cos a2

This parameter varies depending on the camber of the blade.

Therefore the fundamental difference between the circulation 5.1 The Influence of Flow Angles - Profile Loss
coefficient and Zweifel number is that the former takes into Models. While the overall trend of the Smith chart is well estab-
account the camber of the blade, while the latter does not. For a lished, the contribution of profile and secondary losses is not im-
turbine comprised of inclined flat plates (i.e., no camber) the mediately obvious. However, Denton [13] employed a simple
expression in Eq. (26) takes a value of 0.5; for extremely cam- analytical approach to examine the trends in profile loss across the
bered blades it may be as high as 0.8. Equation (26) therefore Smith chart design space. From entropy considerations, the profile
implies that the circulation coefficient will always be lower than loss coefficient may be calculated by integrating Vs3 over the blade
the Zweifel coefficient. This may at first seem odd since they are surfaces:
both defined relative to an ideal case; however Fig. 6 demonstrates X 1
graphically that the integration of Vs2 in the Zweifel coefficient S0
fp  2 Cd Vs =V2 3 dS=S0 (27)
will be larger than the integration of Vs in the circulation ssps
s cos a2 0
coefficient.
Mean-line predictions were performed for designs with constant The dissipation coefficient Cd will vary over the surface, depending
circulation coefficient. With a fixed pressure side velocity distri- on the boundary layer development and the Reynolds number.
bution, peak velocity location and leading edge loading, Co However an approximate prediction may be made by assuming that
uniquely sets the diffusion factor (Fig. 5). Thus a constant circula- it takes a constant value typical of fully turbulent flow, Cd 0:002.
tion coefficient implies constant surface velocity distributions (For the fixed velocity distributions in the current analysis, this
over the Smith chart design space. Figure 9 shows the predicted assumption effectively excludes Reynolds number influences.) The
efficiencies for Co 0:70 (implying DF 0:21), again using the result in Fig. 10(a) shows that the lost efficiency increases as the
profile and secondary loss models of Coull and Hodson [5] and flow turning increases, reaching a maximum in the top left of the
Craig and Cox [9], respectively. Remarkably good agreement Smith chart. This trend may be largely understood from Eq. (27).
with the trends of the original Smith chart in Fig. 1 is achieved. The loss coefficient rises strongly towards the top left of the Smith
This result supports the use of the circulation coefficient rather chart, where the exit flow angle a2 increases and the ratio of pitch
than Zweifel. In many ways this conclusion is unsurprising: blade to suction surface length s=S0 decreases (in order to perform more
design typically focuses on the shape of the surface velocity or flow turning with constant circulation).
Mach number distributions, rather than the Zweifel coefficient. The profile lost efficiency predicted using the method of Coull
and Hodson [5] is presented in Fig. 10(b). As for the Denton
method, the losses increase as the flow turning increases, reaching
5 Assessment of Loss Correlations a maximum in the top left of the Smith chart. The predicted loss is
This section examines several profile and secondary loss corre- higher than the Denton predictions, which is unsurprising given
lations from the literature. Although the accuracy of each model that the dissipation coefficient will tend to be higher for laminar,

021032-6 / Vol. 135, MARCH 2013 Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 Fig. 10 Predicted lost efficiency due to profile loss for Co 0:70: (a) Denton [13]; (b) Coull and Hodson [6]; (c) Ainley and
Mathieson [3]; (d) Dunham and Came [4]; (e) Kacker and Okapuu [5]; (f) Craig and Cox [7]

separated and transitional flows [13]. Figure 10(b) also shows a high /) is far more rapid than either the Coull and Hodson or the
weak tendency for loss to increase at high flow coefficients, which simple Denton methods suggest, and therefore does not appear to
is evident as a slight curling of the contours in the bottom right of be realistic.
the chart. This effect is driven by variations in Reynolds number The profile lost efficiency predicted using the correlations of
(Fig. 3(d)) and exit dynamic pressure (from Fig. 3(a))2. Craig and Cox [9] is presented in Fig. 10(f). Although the loss
Figure 10(c) shows the profile lost efficiency predicted by the increases for high-turning designs, it also increases strongly as the
method of Ainley and Mathieson [2], who interpolated between flow coefficient rises, which does not match the simple Denton
loss charts for nozzle blades with axial inlet flow (i.e., a1 0 analysis in Fig. 10(a). (The kink in the data is due to extrapolation
for a stator), and impulse blades (a1 a2 ) for different flow errors from Craig and Coxs charts.)
angles and pitch-to-chord ratios. A correction is then made for the This comparison shows that the model of Coull and Hodson [5]
trailing edge thickness. The overall trend of Fig. 10(c) is some- gives the most realistic predictions of profile loss across the design
what similar to Fig. 10(a) and 10(b) except that high losses are space. This method also has higher fidelity than the others, since it
predicted for low turning designs (high / and low w), which is depends on the shape of the blade velocity distributions rather
caused by extrapolating the method to the high pitch-to-chord than just the pitch.
ratios of these designs.3 The predicted profile losses for designs
with high turning (low / and high w) are very high. In fact, even
if the secondary losses are zero, this model predicts lower effi- 5.2 The Influence of Flow Angles - Secondary Loss
ciency than was observed by Smith for such designs (Fig. 1). Models. Figure 11(a) shows the predicted lost efficiency due to
Figure 10(d) and 10(e) show two later modifications of the Ain- the secondary loss model of Craig and Cox [9]. This plot shows
ley and Mathieson method. Dunham and Came [3] accounted for that secondary losses are largely responsible for the observed drop
Reynolds number and Mach number variations, which makes only in efficiency at high flow coefficients. Together with the increase
small changes to the predictions (Fig. 10(d)). This model therefore in profile loss with the stage loading coefficient (Fig. 10(b)), the
suffers from the same problems as the original method. Kacker overall trend of the Smith chart is reproduced (Fig. 9). The varia-
and Okapuu [4] suggested multiplying the Dunham and Came pre- tion across the design space in Fig. 11(a) is due to a combination
dictions by 2/3 (Fig. 10(e)), which produces more reasonable lev- of competing effects. The Craig and Cox correlations are pre-
els of efficiency. However the increase in losses for high-turning sented graphically rather than as analytical expressions, but the
designs (high w and low /) and low-turning designs (low w and trends of Fig. 11(a) are driven by variation in the following pa-
rameters, in approximately decreasing order of influence:
2
(1) Velocity ratio through the blade row. Secondary losses tend
For a high speed datum turbine, this effect is slightly magnified due to the high
Mach numbers in this region of the Smith chart.
to be higher when there is a low overall acceleration
3
Note that these designs have low flow turning, so for a constant circulation they through the stage. Designs with high flow coefficient and
must have a high pitch-to-chord ratio. low stage loading therefore have high secondary losses.

Journal of Turbomachinery MARCH 2013, Vol. 135 / 021032-7

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 Fig. 11 Predicted lost efficiency due to secondary loss for Co 0:70, according to: (a) Craig and Cox [6]; (b) Ainley and
Mathieson [3]; (c) Dunham and Came [4]; (d) Kacker and Okapuu [5]; (e) Traupel [7]; (f) Benner et al. [17]

(2) Exit dynamic pressure. Pressure losses scale with the The details of the Ainley and Mathieson secondary loss model
dynamic pressure, which is high for designs with high flow are given in Ref. [14]. They chose the following form for the sec-
coefficient and low stage loading (see Fig. 3(b)). ondary loss coefficient, based on the work of Carter [15]:
(3) Aspect ratio. Secondary losses increase as the aspect ratio    
is decreased. A constant aspect ratio h=Cx has been speci- CL 2 cos2 a2
Ys k (28)
fied in this study, but the Craig and Cox correlation is sensi- s=C cos3 am
tive to the ratio of span to camberline h=b, which decreases
for highly cambered blades, driving a slight increase in sec- where k is a function of the effective inter-row flow areas and the
ondary losses towards the top left of the Smith chart. hub-to-tip ratio. The vector-mean flow direction am is defined as:
(4) Blade pitch. Increasing the blade pitch causes the secondary am tan1 tan a1 tan a2 (29)
flow structures to develop to a larger size (since they are
less restricted by the blade surfaces) and penetrate further The lift coefficient CL is analogous to that for an external wing
towards the midspan region. Craig and Cox use the pitch- section, and is defined using the vector-mean velocity:
to-camberline ratio, which tends to be larger for designs s
with low stage loading coefficient. CL 2 tan a1  tan a2 cos am (30)
(5) Lift coefficient. Highly loaded blades exhibit a stronger C
pressure difference between the suction and pressure surfa-
ces; this pressure difference drives the overturning of the The square of the lift coefficient in Eq. (28) drives the large
endwall flow. According to the Craig and Cox definition, increases in secondary loss at low flow coefficient and high stage
designs with a higher stage loading coefficient have higher loading (Fig. 11(b)). A similar effect was noted in point [5] above
lift coefficients, and so tend to have higher secondary for the Craig and Cox method, but here the influence of lift coeffi-
losses. cient dominates over the other effects.4
(6) Reynolds number. The Craig and Cox model assumes mod- One problem with the Ainley and Mathieson method is that the
est changes in secondary loss with Reynolds number, definition of lift coefficient CL (Eq. (30)) is not appropriate for
approximately following a turbulent trend ( Re0:2 ). moderate or highly cambered blades, as they themselves pointed
out in Ref. [14]. For example, the vector-mean velocity for a high-
The predicted secondary lost efficiency using the method of
turning impulse blade (a1 a2 ) is simply the axial velocity Vx ,
Ainley and Mathieson [3] is presented in Fig. 11(b). The observed
which is much smaller than the actual flow velocities in the cascade.
trend is very different to the Craig and Cox result, and very high
losses are predicted for designs with high stage loading coeffi-
cient. One is left wondering why the two methods give such dif- 4
It should be noted that Ainley and Mathieson account for the machine hub-to-tip
ferent trends, and which is more representative of reality. ratio rather than the aspect ratio.

021032-8 / Vol. 135, MARCH 2013 Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 Fig. 12 Predicted lost efficiency with increasing lift for the datum flow angles
(/ 0:9, w 2).: (a) Profile Loss; (b) Secondary Loss

The reference velocity is therefore artificially low, and thus CL is ultra-high-lift blade designs for LP turbines. Higher lift designs
artificially high for such high-turning designs. Furthermore, Eq. (28) require fewer blades and therefore offer a weight and cost saving
was formulated for relatively low turning blades and it is therefore of for the engine, but tend to have lower efficiency. This section
questionable validity. In comparison, Craig and Cox used results examines the trade-off between lift and efficiency.
from a much wider range of blade designs to formulate their model. Calculations are shown in Fig. 12 for designs with the same
Figure 11(c) shows the predictions according to Dunham and flow angles as the datum turbine (/ 0:9, w 2) but varying
Came [3], who updated the Ainley and Mathieson secondary loss lift. Figure 12(a) shows the lost efficiency due to profile loss
model to account for aspect ratio. Predictions using the subse- alone. (Noting that these designs have velocity distributions of the
quent modification by Kacker and Okapuu [9] are presented in style shown in Fig. 5, the diffusion factor for each design is also
Fig. 11(d). Both methods produce predict lower losses than the indicated on the horizontal axes.) The method of Denton [13] pre-
original method, which is arguably more realistic, but still predict dicts the lowest loss, which is unsurprising since it assumes fully
the wrong efficiency trend across the design space. turbulent boundary layers. The Coull and Hodson [6] model indi-
Predictions using the correlation of Traupel [7] are presented in cates a modest increase in profile loss with increasing lift, consist-
Fig. 11(e). This method successfully captures the loss of effi- ent with such forward-loaded designs. However, there is a slight
ciency at high flow coefficient and low stage loading, and the pre- drop off in loss for designs with low lift (DF < 0:20), which is
dictions are very similar to those of Craig and Cox. (The very unlikely to be physical. Such designs are close to the limit of the
slight kinks at high stage loading coefficient are a result of extrap- high-lift experimental data used to formulate this method ([5]). It
olating from Traupels graphs.) should also be noted that the Coull and Hodson method does not
Figure 11(f) shows the predictions of the recent secondary loss account for the base pressure loss on the turbine trailing edge,
correlation of Benner et al. [17], assuming that the displacement which can be significantly higher for lower lift designs. Figure 12
thickness of the inlet endwall boundary layer is d 0:05h. This shows that the method of Craig and Cox [9] predicts a rapid rise
method correctly captures the increase in loss at high flow coeffi- in loss above Co  0:77, probably due to extrapolating beyond
cients. Although promising, it is not clear how to select an appro- their experimental design space. The Ainley and Mathieson [3]
priate value of d at the preliminary stages of design, which has a and Dunham and Came [4] methods both predict very high profile
large influence on the predictions. losses. For this particular set of flow angles, the Kacker and Oka-
Considering the trends across the design space, the secondary puu [5] correlation predicts similar values to the Coull and Hod-
loss models of Craig and Cox and Traupel are the most reasonable son method.
of those considered. This analysis says nothing of the absolute ac- The secondary lost efficiency is presented in Fig. 12(b) for the
curacy of these correlations, which may only be determined by models considered in the previous section. Of these, the method
direct comparison of predictions and test data. It should also be of Ainley and Mathieson [3] predicts very high losses in line with
noted that both methods are inherently approximate as they take Fig. 11. The models of Dunham and Came [3], Kacker and Oka-
no account of the shape of the surface velocity distributions, puu [5] and Benner et al. [17] (assuming an inlet endwall bound-
which are known to have a significant impact on the secondary ary layer with d 0:05h) predict that secondary loss stays
losses. For example, Gier et al. [11] compared two high-lift LP almost constant with increasing lift. These predictions are con-
turbine cascades with similar flow turning and loading. The veloc- trary to experimental experience (e.g., Ref. [11]). The Craig and
ity peak on the T162 design was relatively far forward on the sur- Cox [9] and Traupel [7] models predict a rise in secondary flow
face, and this exhibited a 30% higher secondary loss coefficient losses with lift, but the increase is significantly less than that
than the aft loaded T161. The mean-line secondary loss models observed by Gier et al. [11] and Praisner et al. [20] for high-lift
considered here lack the fidelity to capture this variation, and will LP turbine blades. Despite this inadequacy, these methods give
therefore be prone to (significant) errors. There is therefore con- the most realistic trends of the models considered.
siderable scope to develop improved methods in the future.

5.4 The Influence of Reynolds Number. As an aircraft

5.3 The Influence of Blade Loading. In recent years a great increases its altitude, the lower ambient pressure causes the blade
deal of research has examined the performance of high-lift and Reynolds numbers to drop which reduces the efficiency of the low

Journal of Turbomachinery MARCH 2013, Vol. 135 / 021032-9

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 pressure turbine. Since cruise performance is of crucial impor- effect which is captured in Fig. 13(a) by the Coull and Hodson [5]
tance, it is desirable to predict this Reynolds number lapse at the correlation. This method is therefore the most realistic.
preliminary design stage. The loss in efficiency due to secondary loss alone is presented
In the International Standard Atmosphere model, the variation in Fig. 13(b). Again, significant differences between the models
of temperature with altitude is approximated by assuming that it are evident. The models based on Ainley and Mathieson ([24])
decreases linearly with height: and the Benner et al. [17] correlation indicate no variation with
Reynolds number. Craig and Cox [9] predict a turbulent trend
T Tsea  0:0065  altitude in m (31) across the whole Reynolds number range, while Traupel [7]
assumes a laminar trend ( Re0:5 ) below ReC 2  105 . In real-
The pressure is then given by: ity, measurements of secondary losses generally show some
increase in secondary losses at low Reynolds number (e.g., Refs.
 5:256 [21] and [22]) but there is no universally-observed trend. It is
P T
(32) therefore unclear from this comparison whether the Reynolds
Psea Tsea
number dependency of the Craig and Cox [9] and Traupel [7]
methods are realistic.
where Psea and Tsea are the sea-level pressure and temperature, To illustrate the influence of blade loading on the Reynolds
and the exponent (5.256) is related to the earths gravitational field number lapse, Fig. 14 shows the Reynolds dependency for designs
and the properties of air. with the datum flow angles (/ 0:9, w 2) using the profile and
To examine the influence of Reynolds number, it is assumed secondary loss models of Coull and Hodson [6] and Traupel [7]
that the nondimensional operating point of the turbine is constant, respectively. A similar plot may be obtained using the Craig and
so that the same temperature and pressure ratios exist through the Cox secondary model. The predicted efficiency when considering
engine. The conditions at the inlet of the repeating turbine stage profile and secondary losses alone is presented together with the
(P01 , T01 ) at altitude are then governed by Eq. (32). Furthermore, overall efficiency for four levels of lift. In line with experimental
for a given stage geometry the following nondimensional groups experience [10], the profile efficiency drops strongly at low Reyn-
will remain constant: olds numbers particularly for the highest lift design. The second-
p ary efficiency shows a similar trend but with a weaker
xs rmean m_ cRT01 w_ x dependency on Reynolds number.
p ; 2
; 2 p
cRT01 rmean P01 rmean P01 cRT01
5.5 Quantitative Predictions of Reynolds Number Lapse
These groups control the variation of dimensional shaft speed, Rates. The analysis presented above focuses on the qualitative
mass flow rate and stage power as the stage inlet total pressure performance of the correlations, but it is also necessary to con-
P01 and temperature T01 vary. sider the absolute accuracy of the predictions. To this end, mean-
Calculations have been performed for the datum turbine operat- line analysis has been performed for the 3-stage high-lift (HL) and
ing at varying Reynolds number. Figure 13(a) shows the predicted ultra-high-lift (UHL) turbines examined by Haselbach et al. [8].
lost efficiency due to profile loss alone. The influence of Reynolds The two turbines had similar performance at sea level conditions,
number on each model may be summarized as follows: Denton but the UHL design had a far more rapid drop-off in performance
[13] and Ainley and Mathieson [3] have no Reynolds number de- at lower Reynolds numbers. Table 2 compares the measured lapse
pendency; Dunham and Came [4] and Kacker and Okapuu [5] rates to predictions using the profile method of Ref. [6] and the
assume a turbulent trend ( Re0:2 ) for Reynolds numbers below secondary loss models of Craig and Cox [9] and Traupel [7]. Both
200,000; Craig and Cox [9] predict a turbulent trend across the methods correctly predict a larger efficiency drop for the UHL
whole Reynolds number range. In fact the measurements of Coull design, but over-predict the lapse rates. The model of Traupel
et al. [10] demonstrated that high-lift LP turbines of this style tend gives more realistic values than that of Craig and Cox.
to exhibit a laminar trend ( Re0:5 ) at low Reynolds numbers The over-predicted lapse rates suggest that the Traupel and
and a turbulent trend ( Re0:2 ) at high Reynolds numbers, an Craig and Cox methods are oversensitive to Reynolds number

Fig. 13 Predicted lost efficiency with varying Reynolds number for the datum turbine:
(a) profile loss; (b) secondary loss

021032-10 / Vol. 135, MARCH 2013 Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 Fig. 14 Reynolds number lapse for varying lift, showing effi-
ciency when considering profile loss alone [6], secondary loss Fig. 15 Comparison of experimental and predicted efficiency
alone [8] and the overall efficiency; datum flow angles (/ 0:9, lapse rates using the profile loss model of Coull and Hodson [6]
w 2) Traupel [7] secondary loss model

variations. For example, in the recent rig measurements of Vaz-

quez et al. [22] the secondary losses approximately followed a
 Re0:06 trend, which is significantly less sensitive than either of
the Traupel or Craig and Cox models (Fig. 13(b)). Table 2 also
shows the calculated loss of efficiency due to profile losses alone.
These values are the lapse rates that would be predicted by a sec-
ondary model with no Reynolds number dependency (e.g., Refs.
[24], and [17]). These profile-only predictions are closer to the
experimental results but still over-predict the lapse rates. This is
likely to be a consequence of the greater flow unsteadiness in the
multistage environment compared to the experiments of Ref. [5].
Higher disturbance levels will tend to reduce separation bubble-
induced losses at low Reynolds numbers, and increase losses at
high Reynolds number by increasing the turbulent wetted area
[10]. Both of these effects will tend to reduce the lapse rate.
Figure 15 shows a more extensive comparison between pre-
dicted and measured lapse rates for several multistage rig tests.
The significant scatter on the plot is in part due to the spread of
Fig. 16 Efficiency contours for three lift coefficients
the experimental data. Predicted lapse rates have been calculated (Co 0:70, 0.75 0.80), using the profile loss model of Coull and
using the Traupel secondary loss model [7] and by considering Hodson [6] and secondary model of Traupel [7]
profile losses alone [6]. In line with the results in Table 2, these
methods tend to over-predict the lapse rates, on average, by
around 30% and 15% respectively. This result reinforces the need
to improve the understanding of secondary loss variation with shows efficiency contours for three circulation coefficients
Reynolds number, and the impact of the highly complex multi- Co 0:70; 0:75; 0:80 (implying DF 0:21, 0.28, 0.36) using
stage environment on performance. the profile loss model of Coull and Hodson [6] and secondary
model of Traupel [7]. (A similar chart may be obtained using the
Craig and Cox secondary loss model.) As the lift is increased the
6 The Influence of Blade Loading on the Smith Chart contours retreat towards the bottom left of the chart, indicating
Efficiency Contours that the efficiency decreases for a given set of flow angles. The
Finally, it is possible to estimate the influence of the lift coeffi- retreat is not uniform, being far more gradual for the highly-
cient on the efficiency contours of the Smith chart. Figure 16 turning blades towards the top left of the plot. These designs have
low pitch-to-chord ratios and high exit angles, causing high profile
losses. Equation (14) shows that profile loss tends to reduce with
increasing pitch (due to a reduction in wetted area), which partly
Table 2 Predicted and measured Reynolds number lapse for mitigates the increase in loss with lift.
the turbines examined by Haselbach et al. [9] While the low sensitivity to lift is most apparent at the extreme
top left of Fig. 16, the same trend is evident in the center of the
Efficiency change, sea level to cruise design space. Two design points have been marked on Fig. 16
roughly indicate the practical limits of typical LP turbines flow
Secondary Loss Model HL UHL
angles: point A(/ 0:8, w 2:2) and point B(/ 1:0,
Craig & Cox 2.35% 2.52% w 1:8). Increasing the lift coefficient Co from 0.70 to 0.80
Traupel 1.39% 1.53% causes an efficiency drop of 0.56% at point A, but a significantly
Profile Losses Only 1.29% 1.45% larger drop of 0.86% at point B. Therefore the lower-turning
Experiments [9] 0.9% 1.3% design B is around 50% more sensitive to increases in lift than
design A.

Journal of Turbomachinery MARCH 2013, Vol. 135 / 021032-11

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms

 7 Conclusions d displacement thickness
The Zweifel coefficient cannot meaningfully compare the lift of / flow coefficient Vx =U
two blades with different flow angles. The root of the problem is g isentropic stage efficiency
the failure of Zweifel to account for blade curvature. An alterna-  kinematic viscosity
tive lift coefficient based on circulation has been defined in an q density
analogous fashion to the Zweifel coefficient. This Circulation h momentum thickness
Coefficient is directly related to the freestream velocity distribu- xs shaft rotational speed
tions over the blade surfaces, providing a meaningful measure of w stage loading coefficient DVh rotor=U
blade loading that is independent of the flow angles. f total energy loss coefficient
The loss correlations derived from Ainley and Mathieson [35] Subscripts
do not capture the correct trends for profile and secondary losses 0 stagnation quantity
over the Smith chart design space and should therefore not be 1 stator inlet plane
used for preliminary design. In particular the secondary loss mod- 2 stator exit/rotor inlet plane
els predict the opposite trend to that expected. 3 rotor exit plane
The profile loss model of Coull and Hodson [6] and the second- is isentropic
ary models of Craig and Cox [9] and Traupel [7] produce realistic p profile loss
efficiency trends with respect to flow angles, lift coefficient and peak peak velocity on the suction surface
Reynolds number lapse rates, but tend to over-predict lapse rates. s secondary loss
The Traupel secondary model gives more realistic quantitative sea sea level
predictions of lapse rates than the Craig and Cox method, but the TE (suction surface) trailing edge
Reynolds number dependency of both models needs to be revis-
ited. The influence of the true multistage flow environment on per- References
formance also needs to be considered. [1] Smith, SF., 1965, A Simple Correlation of Turbine Efficiency, J. R., Aero-
For typical LP turbine flow angles, designs with higher flow naut. Soc., 69, pp. 367370.
[2] Zweifel, O., 1945, The Spacing of Turbomachine Blading, Especially with
turning will tend to have lower sensitivity to increases in blade Large Angular Deflection Brown Boveri Rev., 32(1), pp. 436444.
loading. [3] Ainley, D. G., and Mathieson, G. C. R., 1957, A Method of Performance Esti-
mation for Axial-Flow Turbines, ARC Reports and Memoranda Paper No.
2974.
Acknowledgment [4] Dunham, J., and Came, P. M., 1970, Improvements to the Ainley-Mathieson
Method of Turbine Performance Prediction, Trans. ASME:, J. Eng. Gas Tur-
The authors would like to thank Dr. John Adamczyk and Pro- bines Power, July, pp. 252256.
fessor John Denton for their comments on the analysis. Frank [5] Kacker, S. C., and Okapuu, U., 1982, A Mean Line Performance Method for
Haselbach and Julien Lefaivre of Rolls-Royce plc kindly provided Axial Flow Turbine Efficiency, J. Eng. Power, 104, pp. 111119.
the details for the turbine rig tests. Funding from EPSRC and [6] Coull, J. D., and Hodson, H. P., 2011, Predicting the Profile Loss of High-Lift
Low Pressure Turbines, J. Turbomach., 134(2), pp. 021002-1 021002-14.
Rolls-Royce is gratefully acknowledged. [7] Traupel, W, 1977, Thermische Turbomaschinen, Springer-Verlag, Berlin.
[8] Haselbach, F., Schiffer, H. P., Horsman, M., Dressen, S., Harvey, N., and Read,
S., 2002, The Application of Ultra High Lift Blading in the BR715 LP
Turbine, ASME J. Turbomach. 124(1), pp. 4551.
Nomenclature [9] Craig, H. R. M., and Cox, H. J. A., 1971, Performance Estimation of Axial
Flow Turbines, Proc. Inst. Mech. Eng., 71, pp. 19701971.
Symbols
[10] Coull, J. D., Thomas, R. L., and Hodson, H. P., 2010, Velocity Distributions
b camber-line length for Low Pressure Turbines, J. Turbomach., 132(4), pp. 041006-1041006-13.
C, Cx true chord; axial chord [11] Gier, J., Franke, M., Hubner, N., and Schroder, T, 2008, Designing LP Tur-
Cd dissipation coefficient bines for Optimized Airfoil Lift, ASME Paper No. GT2008-51101.
[12] Horlock, J. H., 1966, Axial Flow Turbines: Fluid Mechanics and
Co circulation coefficient Thermodynamics, SBN 0-88275-097-6.
DF diffusion factor Upeak  UTE =UTE [13] Denton, JD, 1993, Loss Mechanisms in Turbomachines ASME J. Turbom-
fr S0 -based reduced frequency fwake S0 =UTE ach., 115(4), p. 621656.
LEI leading edge integral (Eq. (12)) [14] Ainley, D. G., and Mathieson, G. C. R., 1951, An Examination of the Flow
and Pressure Losses in Blade Rows of Axial-Flow Turbines, ARC Reports and
ReC chord-based exit Reynolds number UTE C=
Memoranda Paper No. 2891.
ReS0 S0 -based exit Reynolds number UTE S0 = [15] Carter, ADS, 1948, Three-Dimensional Flow Theories for Axial Compressors
m_ mass flow rate and Turbines, Proc. Inst. Mech. Eng., 159(1), pp. 255268.
P pressure [16] Thwaites, B., 1949, Approximate Calculation of the Laminar Boundary
Layer, Aeronaut. Q., 1, pp. 245280.
rmean mean radius [17] Benner, M. W., Sjolander, S. A., and Moustapha, S. H., 2006, An Empirical
s pitch Prediction Method For Secondary Losses In Turbines Part II: A New Second-
S surface distance from the leading edge ary Loss Correlation, J. Turbomach., 128(2), pp. 281291.
S0 suction surface length [18] Vera, M., Hodson, H. P., and Vazquez, R., 2003, The Effect of Mach Number
tTE trailing edge thickness on LP Turbine Wake-Blade Interaction 9th ISUAAAT, Sept. 48, Lyon,
France
T temperature [19] Marconcini, M., Rubechini, F., Pacciani, R., Arnone, A., and Bertini, F., 2010,
U blade velocity at mid-span Redesign of High-Lift LP-Turbine Airfoils For Low Speed Testing ASME
Vi , Wi gas velocity in the absolute and relative frame Turbo Expo, Glasgow, UK, 1418 June, ASME Paper No. GT2010-23284.
[20] Praisner, T. J., Grover, E. A., Knezevici, D. C., Popovic, I., Sjolander, S. A.,
Vs freestream velocity over blade surfaces Clarke, J. P., and Sondergaard, R., 2008, Towards the Expansion of Low-Pres-
Vx , Vh axial and circumferential velocity components sure-Turbine Airfoil Design Space, ASME Paper No. GT2008-50898.
w_ x stage power [21] Hodson, H. P., and Dominy, R. G., 1987, The Off-Design Performance of a
Y total pressure loss coefficient Low-Pressure Turbine Cascade ASME J. Turbomach. 109(2), pp. 201209.
Zw Zweifel lift coefficient [22] Vazquez, R., Torre, D., Partida, F., Armananzas, L., and Antoranz, A., 2011,
Influence of Surface Roughness on the Profile and End-Wall Losses in Low
ai , bi flow angles in the absolute and relative frame Pressure Turbines, ASME Paper No. GT2011-46371.

021032-12 / Vol. 135, MARCH 2013 Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/28/2014 Terms of Use: http://asme.org/terms