250 AIRCRAFT ENGINE COMPONENTS
5 Hole
0• 56 -
0 48 -
• .
O.40 -.~l
.
~Probe Measurement Limit~
~ l - - - - =
?-
0.32 Ll., x. o In
0.24 - I
0.16 - |
0.08 t ~ ~ ~
NsIS10 _ - - ~ ~ % ~ Curve Pi - P
Label qi
-0 08
a. " 0 .
SsISI
~
2
1 0.1
O.2
O.56 -
0.48
Y
I 3
4
5
O.3
0.4
O.5
X = 0.551 6 0.6
O.
Z O.24
7 O.7
8 0.8
NsISI ,EI6- 9 0.9
I0 L0
20 3.0
-0.08 , , l , { , L , ,
22 4. 0
b. 0.72 0.88 1.04 1.20 1.36 1.52
I. 957
i i i i i
64 0.80 0.96
Y
Fig. 4.10 Isobar plot of total pressure loss coefficients at three axial planes.
TURBINE AERODYNAMICS 251
temperature. This causes problems both in the axisymmetric meridional
plane streamline analysis, where the conventional assumption is that there is
no transport between the stream tubes, and in designing rotor blades where
it is desirable to have a particular spanwise temperature distribution so that
the highly stressed sections are not in the highest-temperature flow region.
The spanwise mixing is very intense close to the end wall due to the
vortices discussed above. Even well above the end wall, however, there can
also be relatively strong streamwise vorticity. The source of this is the
streamwise vorticity produced by putting relatively weak shear flows through
high-deflection cascades. This phenomenon is discussed in the following
section.
Classical Secondary Flow Analysis
The aspect of end wall aerodynamics to first receive analytical attention
was the secondary flow. Inviscid rotational flows produce a streamwise
component of vorticity when they are put through a deflection. The original
work by Squire and Winter 47 and the classic by Hawthorne 48 provide ample
clear evidence of the nature of this phenomenon. The theory of secondary
flow has been generalized by Hawthorne 49 to include the effects of strat-
ified, compressible fluids in a rotating frame of reference. This is an essential
feature when considering the secondary flow effects in turbine rotor blade
passages. The recent work by Came and Marsh5° presents a simple deriva-
tion, not only of the distributed secondary vorticity but also of the trailing
filament and shed vorticity. Taking the normal assumptions of secondary
flow, most notably that it does not alter the primary flow, Belik 51 performed
a very thorough parametric analysis of secondary flow losses. In this inviscid
analysis, the kinetic energy of the relative secondary motion was taken to be
lost. His analysis included variations in the inlet velocity profile shape,
velocity triangles, aspect ratio, and solidity. The results showed remarkably
good qualitative agreement with end wall loss data from a number of
sources. In an effort to provide a more generally applicable analysis, Stuart
and Hetherington52 devised a method that permitted a strong coupling
between the primary and secondary flows. The result was that severe
distortions of the Bernoulli surfaces could be accounted for.
Classical secondary flow analysis has the potential of being helpful in
understanding spanwise mixing in that viscosity may not necessarily have a
strong role away from the end walls. In light of the complex viscous nature
of the end wall flow described earlier, it would seem optimistic to expect this
mechanism to make anything more than a relatively small contribution to
the total end wall loss picture.
End Wall Boundary Layer Analysis
The application of boundary layer theory to the end wall problem in
turbomachinery has followed two distinctly different paths. The first con-
sists of tangentially averaging the annulus boundary layer equations and
then, with additional approximations, following the growth of the boundary
252 AIRCRAFT ENGINE COMPONENTS
layer axially through the machine. The work by Mellor and Wood 53 is an
example of this approach. This method has met with some success in
compressor applications where the solidity is high and the turning is low
relative to the turbines. Due to the very strong cross flows and the large
gapwise variation in the end wall boundary layer in a turbine cascade, it
would appear that the gap-averaged approach offers little hope.
The second approach consists of solving the three-dimensional boundary
layer equations over the cascade end wall. Mager 54 presented a study of the
momentum-integral form of the equations that was not limited to weak
cross flows. This work was applied to the case of a turbine cascade end wall
boundary layer by Dring. 55 The flow path between the cascade airfoils was
treated as a duct. In this way, the complexity of the leading edge flow, most
notably the horseshoe vortex, was avoided. The solution gave qualitatively
reasonable predictions of the end wall flow in that for the case presented the
incoming limiting streamlines (on the end wall surface) reached the suction
surface at an axial location approximately halfway through the cascade. The
predicted losses, however, were generally well below the measured values.
Booth 56 extended this approach to include the effects of compressibility. His
loss calculation included an additional effect due to the displaced inviscid
flow. For several first-vane test cases, his predictions were also significantly
below measured loss values. Carrick 44 applied the general integral method
of Smith 57 to the cascade end wall problem. Steps were taken to include a
degree of periodicity in the boundary layer ahead of the cascade leading
edge, but not to the extent necessary to account for the complex nature of
the flow observed by Langston et al. 43 Even more powerful three-dimen-
sional boundary layer analyses are available (e.g., Nash and Pate158), but
one must keep in mind that many aspects of the turbine end wall flow do
not satisfy the basic boundary layer approximations. The horseshoe vortex,
the back flow region behind it, and the alterations to the two-dimensional
pressure distribution are features beyond the scope of boundary layer
analysis.
Three-Dimensional Flow Analysis
The recent availability of larger and faster computational facilities has
made possible the fully three-dimensional flow analysis. This area of compu-
tational fluid dynamics is described at length in Chap. 6. Attention here
shall be restricted to an example of such a calculation as it has been applied
to the turbine cascade end wall flow problem. The calculation by Water-
man 37 and by Waterman and Tall 59 is a fully viscous solution of the
three-dimensional, incompressible fluid flow equations. The calculation was
executed for several geometrically complex first-vane annular cascades. The
full equations were "split" into separate equations containing the viscous
and pressure effects. The viscous equation was made parabolic by the
assumption that the streamwise diffusion terms are negligible. This per-
mitted a standard marching solution with each step containing a solution of
the elliptic in-plane equation. The equation containing the pressure effects is
fully elliptic and is solved by relaxation. It includes a displacement effect
due to the viscous equation. By this means a coupling has been achieved
TURBINE AERODYNAMICS 253
between the viscous and inviscid phenomena. The advantages over the
inviscid secondary flow and viscous boundary layer analyses become clear.
Although there are m a n y features of the analysis that one might argue
about, the fact remains that this approach has the potential of accounting
for more of the observed features of turbine end wall flows than any other
approach currently available.
Unshrouded Rotor Tip Leakage
The clearance between the tip of a turbine blade and the outer air seal, or
rubstrip, has long been recognized as a source of large turbine inefficiencies.
Every effort is made to reduce this clearance mechanically. However, as a
result of mechanical and thermal transients, eccentricities, and other rea-
sons, the running clearance on most gas turbine rotor tips is 1-5% of the
blade span. In turbines where for cycle reasons the flow area is to be varied,
a similar situation occurs. A relatively simple means of varying the flow area
is to rotate the vanes about a radial axis. The resulting mechanical con-
figuration can have clearances between the airfoil and the flow path annulus
at b o t h the root and the tip. Thus, as m a n y as three clearances can be
present in a single turbine stage. The topic of clearance losses in shrouded
rotors is discussed in Sec. 4.7. This mechanism is included under parasitic
loss since it involves flow that leaves and re-enters the flow path.
The impact of these clearances on turbine performance has been under
continuous study for years. This source of end wall loss has a unique feature
that makes it relatively easy to study experimentally. That is, that the
clearance in a rig or cascade test can usually be easily varied and the impact
of clearance on performance can be isolated. The result is that there is a
large b o d y of cascade and rotating rig data available (e.g., see Refs. 60-62)
that permits one to construct a relatively accurate correlation of the perfor-
mance penalty as a function of relative clearance and other variables. A
simple order-of-magnitude study will indicate that, except for very small
clearances, the flow in a clearance gap is dominated by the pressure
difference across an airfoil and not by the viscous effects in the gap. For this
reason, the effects of relative wall motion, as in the case of a rotor tip, are
usually weak. Hence, a correlation based on rotor tip data would be
expected to be adequate for the prediction of variable-area vane clearance
losses where there is no relative wall motion.
It can be shown from the references cited above that there is a reduction
in turbine stage efficiency of approximately 1.5% for every percent of
relative clearance. Thus it is an unusual unshrouded turbine stage that is not
giving up at least 2% in performance to this mechanism. This powerful
impact has led turbine designers to strive for minimal clearances, as well as
to minimize the penalty occurring for a given clearance. This and other end
wall loss reduction schemes are discussed in the following section.
Loss Reduction Devices
As a result of the powerful impact end wall loss mechanisms have on
turbine performance, a number of attempts have been made to reduce these
254 AIRCRAFT ENGINE COMPONENTS
penalties. Some success has been seen in these attempts, but it is difficult to
extrapolate a successful device from one turbine configuration to another.
There is the distinct risk that a loss reduction scheme, if improperly
executed, m a y result in a significant loss increase. With this in mind,
consider a few of the successes that have been reported.
A method of contouring the annulus of a first vane to reduce end wall loss
has been presented by Deich et al. 63 Incorporation of this scheme into
turbine designs has produced conflicting results. Ewen et al. 64 applied the
scheme to a small axial turbine and saw a remarkable improvement in
performance. The results of Okapuu, 65 however, indicate no effect on
p e r f o r m a n c e at the design point, but some improvement at lower pressure
ratios.
Another trial-and-error approach to the reduction of end wall loss has
been followed by Prumper. 66 He employed boundary layer fences of varying
size, shape, and location in an attempt to reduce the end wall loss in an
annular cascade of first vanes. The result was that after testing over 400
configurations, an optimum was established that gave a significant improve-
m e n t in performance. The underlying physics and its universality, however,
remain to be determined.
The area of unshrouded rotor tips and rubstrips has also received
considerable attention. Grooves and recesses have been cut into the airfoil
tips in an attempt to make them behave as labyrinth seals. Rubstrips have
been recessed radially in an attempt to keep them in a low-flow, separated,
environment. Rubstrips have also been made out of honeycomb in an effort
to reduce the running clearance, as well as to reduce the toss for a given
clearance.
Turbine end wall loss reduction is at this point a black art and without
any real design discipline. It is at best a trial-and-error approach with no
guarantee that success in one application can be extrapolated to another
application. The underlying problem is the basic lack of understanding of
the composite of the numerous mechanisms that go into producing end wall
loss.
Conclusions
At this point, the conclusions with regard to the state-of-the-art of end
wall aerodynamics relative to that of profile aerodynamics must be obvious.
Although the mechanisms by which one can lose total pressure can be
counted on the fingers of one hand, these individual effects have not been
tied together into a meaningful, unified picture.
The loss correlations that have been derived are at best inadequate.
Scatter on the order of 100% is not uncommon. This is not surprising
considering the questions of net and gross loss and the large increase in end
wall loss downstream of a cascade. Accurate end wall loss correlations may
eventually become available, but there is little evidence of it today. One
notable exception is unshrouded rotor tip clearance loss, which appears to
be well within the realm of correlation.
Analytical attacks on the problem stem from a number of different
approaches. Those based on conventional secondary flow and three-dimen-
