Professional Documents
Culture Documents
GT2012
June 11-15, 2012, Copenhagen, Denmark
GT2012-68497
ABSTRACT The flow through a turbine cascade has been widely studied,
An experimental study of the hub leading edge contouring using documented and verified by many experiments (Langston [2],
fillets is performed in an annular sector cascade to observe the Sieverding [3], Langston [4], Wang et al. [5], Acharya and Mahmood
influence of secondary flows and aerodynamic losses. The investigated [6], Eckerle and Langston [7]). The secondary flow is generated due to
vane is a three dimensional gas turbine guide vane (geometrically the inlet boundary layer splitting in front of the leading edge coupled
similar) with a mid-span aspect ratio of 0.46. The measurements are with the passage pressure gradient between pressure side and suction
carried out on the leading edge fillet and baseline cases using side of a turbine vane or blade passage. When the inlet boundary layer
pneumatic probes. Significant precautions have been taken to increase approaches a leading edge, a total pressure gradient occurs along the
the accuracy of the measurements. The investigations are performed spanwise direction. As the flow stagnates at the leading edge, the total
for a wide range of operating exit Mach numbers from 0.5 to 0.9 at a pressure gradient then becomes a static pressure gradient along the
design inlet flow angle of 90°. Data presented include the loading, span. This spanwise static pressure gradient then causes the boundary
fields of total pressures, exit flow angles, radial flow angles, as well as layer fluid to flow towards the endwall. As the flow turns upstream, it
profile and secondary losses. The vane has a small profile loss of rolls into a vortex or system of vortices that wrap around the body,
approximately 2.5 % and secondary loss of about 1.1%. Contour plots forming the well-known horseshoe (HS) vortex, which has two distinct
of vorticity distributions and velocity vectors indicate there is a small legs, known as the suction side HS vortex and pressure side HS vortex
influence of the vortex-structure in endwall regions when the leading (Langston [2]). The pressure side HS vortex then mixes with the
edge fillet is used. Compared to the baseline case the loss for the endwall flow (due to the pressure gradient inside the passage) and
filleted case is lower up to 13 % of span and higher from 13% to 20 % grows in size and migrates towards the suction side, creating a passage
of the span for a reference condition with Mach no. of 0.9. For the vortex. This HS vortex produces large-scale, low-frequency bi-stable
filleted case, there is a small increase of turning up to 15 % of the span unsteadiness in the leading edge region (Devenport et al. [8]).
and then a small decrease up to 35 % of the span. Hence, there are no The secondary flow deteriorates the aerodynamic performance of
significant influences on the losses and turning for the filleted case. the turbine, which worsens as the blade loading is increased and the
Results lead to the conclusion that one cannot expect a noticeable blade aspect ratio is decreased (Mobarak et al. [9]). The secondary
effect of leading edge contouring on the aerodynamic efficiency for the flow depends on the prehistory (Herzig and Hansen [10]). The
investigated 1st stage vane of a modern gas turbine. secondary flow with trailing legs persists very far downstream, with
non-uniformities that strongly affect next blade rows (Devenport et al.
INTRODUCTION [8]).
Secondary flows are complex three-dimensional flows that arise There are few studies in the open literature reporting methods for
in a cascade of turbine vanes or blades. Such fluid flows exist in all reducing secondary losses. Leading edge (LE) contouring near the
axial flow turbines, machines that are used to produce most of the endwall is one of the methods that show potential to decrease
world’s electricity and provide thrust for most of its aircrafts. The secondary losses. Moreover, the similar application is well established
secondary flow is considered detrimental to the performance of gas in the aviation industry at the wing/body junction since many years.
turbines. This is due to the fact that the secondary flow causes total Many of the previous studies focused on the wing/body junction where
pressure drop and increase the entropy. In some cases it can account only the horseshoe vortex presents. Very few studies have reported the
for 30-50% of the total aerodynamic losses in a blade or vane row use of leading edge fillets in turbine cascades, where there are not only
(Sharma and Butler [1]). horseshoe vortices but also passage vortices. Sauer and Wolf [11] and
LE fillet
Figure 2: CAD model of LE fillet Experimental Facility and Cascade Arrangement: The
annular sector cascade (ASC) facility was designed to provide a test
The design of the leading edge fillet (Figure 2) is based on the facility for different kind of experimental investigations to improve the
work reported in [13-18]. The recommended design criteria from the efficiency of modern gas turbines. It is installed at the Division of Heat
literature [17-18] are taken to design the LE fillet. Their main design and Power Technology at KTH, Sweden. A schematic representation of
criteria have been used here: the maximum height of the fillet is one the wind tunnel arrangement is shown in figure 4.
(1) Inflow, (2) Settling chamber, (3) First radial contraction, (4) Turbulence grid (5) Second
radial contraction, (6) Test sector with NGVs and 7(Outflow)
Figure 7: Axial section view of test sector with LE fillet through
NGV0
Figure 5: Annular sector cascade arrangement
Pneumatic pressure taps are located at 55.7% cax,hub upstream
The ASC is equipped with a fully automatic traverse mechanism
from the NGV leading edge in the upstream casing. The downstream
which is able to scan the upstream and downstream flow field. The
traverse arc is located at 127.5% cax,hub downstream from the NGV
flow temperature is measured by a platinum resistance temperature
leading edge. The endwall pressure measurements are performed at the
sensor (Pt100) which has an accuracy of ± 0.15°C at 0°C. The
hub by 9 pressure taps at 136.5% cax,hub downstream from the NGV
temperature sensor is placed in the settling chamber and connected to a
leading edge denoted by ‘31’ in figure 7. The downstream
Keithley 2701 data acquisition and logging unit. An axial cross
measurements are performed by a 5-hole L-probe (figure 8) which is
sectional view of the test sector is depicted in figure 7. It shows all
calibrated for the Mach number range of 0.1 to 0.95. The same parallel
main locations. The inlet measuring plane is located at 55.7% cax,hub
bar turbulence grid (figure 9) with turbulence intensity (Tu) of 1.5%
upstream from the NGV leading edge denoted by ‘20’ in figure 7. The
(Putz [22]) is used for the filleted and the baseline case. The 5-hole
inlet total pressure is measured by a 3-hole cobra probe which has
probe is an L-probe and the tip is positioned at 108.6° from the axial
been calibrated before the run to get the inlet total pressure profile and
direction. Therefore, the actual probe tip traverse plane is at 107.1%
this inlet total pressure profile is used for the downstream loss
cax,hub and all the downstream measurements are performed at 107.1%
calculation. However, this inlet measuring plane is not used while
cax,hub. Each full area traverse of the 5-hole probe represents 1170
running the downstream area traverse in order to avoid the upstream
measurement points with 39 tangential and 30 radial points between
disturbance. Instead, the inlet total pressure is derived with the data
0% and 100% span for one pitch. All three NGVs are equipped with
0.9
0.8
Mach number
0.7
Figure 8: View of 5-hole Figure 9: A parallel bar turbulence
probe grid
0.6
Baseline(75% span)
0.8 Fillet(75% span) 0.75 0.8 0.85 0.9 0.95 1
0.6 0.4
0.35
TE of NGV 0
0.4
0.25
0.2
SS
PS
0.2
0 0.15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized surface length, x/cax 0.1
0.05
Figure 12: Profile Miso distribution at different spans for filleted and
baseline case at operating point Miso3= 0.9 0
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
Pitchwise location, s/s
Baseline,Miso=0.90
Fillet,Miso=0.90 Figure 14a: Normalized total pressure (p30/p20) distribution for
1
Baseline,Miso=0.70 baseline case at Mach number Miso3= 0.9
Isentropic Mach number, Miso
Fillet,Miso=0.70
0.8 Baseline,Miso=0.50 0.75 0.8 0.85 0.9 0.95 1
Fillet,Miso=0.50
0.6
0.4
0.35
0.4 TE of NGV 0
Spanwise location, z/h
0.3
0.2 0.25
SS
PS
0.2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.15
Normalized surface length, x/cax
0.1
0.05
Figure 13: Profile Miso distribution at 25% span for baseline and
filleted case at different operating points 0
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
Pitchwise location, s/s
From the flow field picture, it can be concluded that the NGV has a
smooth acceleration for all exit Mach numbers. The acceleration
increases with the increase of the exit Mach number. The maximum Figure 14b: Normalized total pressure (p30/p20) distribution for
velocity in the suction side is close to the trailing edge. This essentially filleted case at Mach number Miso3= 0.9
means that one can expect a small level of secondary loss.
Unfortunately, the loading near the hub endwall cannot be captured As the probe is traversed at 107.1% of cax, there is a minor phase
below 25% with the instrumented vane. shift in the positive circumferential direction between the wake
location and the vane trailing edge. Above 30% span, the wake is
Field of Total Pressure: Figures 14a and 14b display contour straight with clearly defined borders which are basically regions of the
plots of total pressure at the downstream plane for the baseline and the profile losses for both cases. However, some interesting secondary
filleted cases, respectively, at the reference operating point (Miso3 = flow phenomena, which can be attributed to the presence of LE fillet,
0.9). The horizontal axis corresponds to the normalized pitchwise are found between the hub endwall and 25% span. The dark blue total
location whereas the vertical axis corresponds to the normalized span. pressure loss core can be observed at about 13% span for the baseline
The total pressure is normalized by the upstream total pressure. The case (figure 14a) and at about 16% span for the filleted case (figure
24
clearance has to be kept between the probe head and the endwall. A
0.8
32
certain distance from the wall to the probe head is also necessary to
avoid the wall proximately effect (Treaster and Yocum [23]).
12
24
0.6
SS
PS
Exit Flow Angle Distributions: The exit flow angle
22
14
TE of NGV 0
0.4
distributions give information about the flow turning, which is a very
16
14
18
important design criterion of a nozzle guide vane. Lower values of exit
20
24
22
TE of NGV +1
V -1
flow angle (α) means higher turning. 0.2
14
TE of NG
34
12
0
-1 -0.5 0 0.5 1
0 5 10 15 20 25 30
Pitchwise location, s/s
1 Figure 15a: Exit flow angle distribution for filleted case at exit Mach
number Miso3= 0.9
0.8
Spanwise location, z/h
22
1
24
0.6 0.9
12
SS
PS
TE of NGV 0 0.8
20
14
18
16
16 20
22
0.4
14
0.7
14
TE of NGV +1
24 0
V -1
0.2 0.6
TE of NG
3
Baseline, Miso3=0.90
14
30 0.5
Fillet, Miso3=0.90
0
0.4
-1 -0.5 0 0.5 1
Pitchwise location, s/s 0.3
0.2
Figure 15a: Exit flow angle distribution for baseline case at exit
Mach number Miso3= 0.9 0.1
0
In general, a very similar picture is acquired for the baseline 15 15.5 16 16.5 17 17.5 18 18.5 19 19.5
(figure 15a) and the filleted (figure 15b) cases, although some local Exit flow angle [°]
variation is found. In the free stream, the exit flow angles remain
unaffected by the use of the LE fillet. The nature of the exit flow angle Figure 16a: Comparison of pitch-averaged exit flow angle
can be explained by the pressure gradient between the suction and distribution at exit Mach number Miso3 = 0.9
pressure side, sucking the flow towards the adjacent suction side. Very
high deviations from the mean exit flow angle can be observed in the
vortex regions where areas of both very high and very low exit flow
angles collide with each other. The phenomenon of underturning and
overturning in these secondary flow regions causes significant losses.
In figure 16a-16c, comparisons of the pitch-averaged exit flow angle
distribution along the span are depicted. As can be seen in the figures,
0.8
1
Spanwise location, z/h
0.7
0
2
-4
0.8
SS
PS
-6
-2
0.4
0.4
-4
-2
0.3
-6
TE of NGV +1
TE of NGV -1
0.2
0
0.2 -1-1 0
-4
-2
2
6
-2-4
-6
0.1
0
0
0
15 15.5 16 16.5 17 17.5 18 18.5 19 19.5 -1 -0.5 0 0.5 1
Exit flow angle [°] Pitchwise location, s/s
Figure 16b: Comparison of pitch-averaged exit flow angle Figure 17a: Radial flow angle distribution for baseline case at exit
distribution at exit Mach number Miso3 = 0.7 Mach number Miso3= 0.9
0.8
Spanwise location, z/h
0.7 1
0.6
0.8
Spanwise location, z/h
4 0
Baseline, Miso3=0.50
0.5
Fillet, Miso3=0.50 TE of NGV 0
0.6
0.4
SS
PS
0.3
-6
-2
0.4
-6
-4
-2
0.2
TE of NGV +1
TE of NGV -1
0
0.2 -8
0.1 6
0
0
0
-2
0 0
15 15.5 16 16.5 17 17.5 18 18.5 19 19.5
Exit flow angle [°] -1 -0.5 0 0.5 1
Pitchwise location, s/s
Figure 16c: Comparison of pitch-averaged exit flow angle
distribution at exit Mach number Miso3 = 0.5 Figure 17b: Radial flow angle distribution for filleted case at exit
Mach number Miso3= 0.9
Radial Flow Angle Distributions: The radial flow angle (also Vorticity and Velocity Vector Distribution: The three-
known as the pitch angle), together with the exit flow angle, form the dimensional velocity vectors are calculated in a Cartesian coordinate
structures of the flow: when the radial flow angle is negative, the flow system.
is a downward-directed flow. Similar to the exit flow angle
distribution, the radial flow angle distribution remains unaffected by
the use of a LE fillet. In the mainstream flow, especially close to the
suction side, the radial flow angle is around 0°. Close to the pressure The components uy and uz can be calculated from the velocity vector
side (Δs/s = 0.1 to 0.4), both figures (figure 17a and 17b) show a together with exit flow angle (α) and radial flow angle (β) according to
downwash, which is most likely caused by the tip endwall contouring. the following equations:
The regions of secondary flow are strongly influenced by the radial
flow. Changes in the endwall regions are discussed in the vorticity (1)
analysis. (2)
The axial component of the vorticity (ωx) can be calculated as:
(3)
SS
PS
and 0, a clearly visible counter-clockwise rotating passage vortex (A) 0.35
is present for both the filleted and the baseline cases. As discussed in 0.3
TE of NGV +1
0.1
TE of NGV -1
A
0.05 B
0
-5000 -4000 -3000 -2000 -1000 0 1000 2000 3000 4000 5000 -0.05
_
+ -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4
Pitchwise location, y/s
0.6 0.8 1
Δs/s
SS
PS
0.35
TE of NGV 0
0.3 E Figure 18b: Vorticity and flow vector distribution for filleted case at
Spanwise location, z/h
0.1
TE of NGV -1
0.9
However, the intensity of the vorticity contour is slightly smaller
in this region (A) for the filleted case compared to the baseline case. 0.8
Similar trends are observed in the total pressure contour (figure 14a
Spanwise location, z/h
0.7
and 14b) and also in the exit flow angle distribution where a small
increase in the turning is found for the filleted case. Directly right of 0.6
this region, a clockwise rotating vortex is found for both cases (B). Baseline, Miso3=0.90
0.5
This vortex originates from the suction side leg of the horseshoe Fillet, Miso3=0.90
vortex of the NGV0. An interesting phenomenon is found right above 0.4
the clockwise rotating vortex. Two small counter-rotating corner
0.3
vortices (C and D) are found for the baseline case, one with higher
intensity (counter-clockwise) (C) and another one with lower intensity 0.2
(clockwise) (D). The counter-clockwise rotating corner vortex (C)
0.1
might be a part of the passage vortex that is swept up due to the
suction side leg of the horseshoe vortex of the NGV0. The clockwise 0
1 2 3 4 5 6 7 8 9 10 11
rotating corner vortex (D) might be induced by the passage vortex as
kin.energy [%]
described by Wang et al. [6]. The intensity of the clockwise rotating
corner vortex (D) is slightly smaller for the filleted case compared to Figure 19a: Mass-averaged kinetic energy loss distribution at exit
the baseline case. However, the spanwise extension of the corner Mach number Miso3= 0.9
vortices is smaller for the baseline case compared to the filleted case.
This essentially implies that the LE fillet moves the position of the Figure 19a-19c illustrates the mass averaged kinetic energy loss
corner vortices. Above 13% span for the baseline case and 15% span distribution over the full span in the downstream plane. Both the
for the filleted case, a clearly distinct clockwise rotating vortex (E), is baseline and filleted cases give similar results at the mid-span. In
0.7
A closer investigation of the flow near the endwall reveals a
0.6
Baseline, Miso3=0.70
wider zone of total pressure loss for the suction side of the
0.5 baseline case (up to Δs/s = -0.14) compared to the filleted case
Fillet, Miso3=0.70
(up to Δs/s = -0.04). However, the opposite flow picture is found
0.4
on the pressure side where a slightly wider zone of total pressure
0.3 loss is found for the filleted case (up to Δs/s = 0.13) compared to
the baseline case (up to Δs/s = 0.12). The maximum total pressure
0.2
loss core can be observed at about 13% span for the baseline case
0.1 whereas at about 16% span for the filleted case.
0
From the exit flow angle distribution, it can be concluded that
1 2 3 4 5 6 7 8 9 10 11 there is no major change in the turning over the full span for the
kin.energy [%] filleted case compared to the baseline case, although some local
Figure 19b: Mass-averaged kinetic energy loss distribution exit variations do exist. For the filleted case there is a small increase
Mach number Miso3= 0.7 of turning up to 15 % of the span and then a small decrease up to
35 % of the span.
1 The radial flow angle distributions downstream of the vane
0.9
remain unaffected by the use of the LE fillet.
From the vorticity distribution the passage vortex, the suction side
0.8 leg of the HS vortex, two counter-rotating corner vortices and the
trailing edge shed vortex are found with only little difference in
Spanwise location, z/h
0.7
intensity for the filleted case compared to the baseline case in the
0.6 suction side. The passage vortex intensity is slightly smaller for
Baseline, Miso3=0.50
0.5
the filleted case compared to the baseline case. However, the
Fillet, Miso3=0.50 spanwise extension of vortices is slightly higher for the filleted
0.4 case compared to the baseline case. This essentially implies that
0.3
the LE fillet shifts the position of the vortices. However, the
changes in the overall vortex structure are very small.
0.2 A very similar picture of the kinetic energy loss coefficients over
0.1
the span can be observed at the three different operating points for
the baseline and the filleted cases. Losses from 30% to 70% span
0
1 2 3 4 5 6 7 8 9 10 11 indicate that it is the profile loss which accounts for
kin.energy [%] approximately 2.5% whereas the secondary loss is about 1.1% for
the reference operating point (Miso3 = 0.9). The secondary loss in
Figure 19c: Mass-averaged kinetic energy loss distribution at exit the hub region is higher for the baseline case up to 13% span
Mach number Miso3= 0.5 compared to the filleted case whereas it is lower from 13% to