Professional Documents
Culture Documents
net/publication/319169576
CITATIONS READS
15 165
2 authors:
All content following this page was uploaded by Marcel Staats on 17 October 2017.
GT2017-63246
ABSTRACT 1 INTRODUCTION
We present results of experiments on a periodically unsteady Introducing pressure gaining combustion concepts in gas
compressor stator flow of the type which would be expected turbines has the potential to increase the overall efficiency of the
in consequence of pulsed combustion. A Reynolds number of engine by up to 20% compared with state of the art single cy-
Re = 600000 was used for the investigations. The experiments cle gas turbines [1]. Using constant volume combustion (CVC)
were conducted on the two-dimensional low-speed compressor instead of constant pressure combustion increases the thermal
testing facility in Berlin. A choking device downstream the trail- efficiency of the gas turbine cycle. Air breathing engines usu-
ing edges induced a periodic non-steady outflow condition to ally operate with continuous combustion. A pulse detonation en-
each stator vane which simulated the impact of a pressure gain- gine (PDE) is one type of a propulsion concept that uses CVC
ing combuster downstream from the last stator. The Strouhal and has a highly non-steady characteristic [2]. Implementing the
number of the periodic disturbance was Sr = 0.03 w.r.t. the sta- PDE concept in a gas turbine can be managed by the use of can-
tor chord length. Due to the periodic non-steady outflow con- annular combustion chambers, where the tubes close sequentially
dition, the flow-field suffers from periodic flow separation phe- when combustion is carried out. The upstream located last com-
nomena, which were managed by means of active flow control. pressor stage then suffers from a non-steady outflow condition,
In our case, active control of the corner separation was applied leading to the risk of compressor stall. However, a stable com-
using fluidic actuators based on the principle of fluidic ampli- pressor operation is crucial in gas turbine application. In this con-
fication. The flow separation on the centre region of the stator tribution, a compressor stator flow of the type that is expected in
blade was suppressed by means of a fluidic blade actuator lead- a PDE was investigated. An active flow control (AFC) approach
ing to an overall time-averaged loss reduction of 11.5%, increas- was used that helped to stabilize the compressor stator flow-field.
ing the static pressure recovery by 6.8% while operating in the The two-dimensional low speed compressor test rig in Berlin
non-steady regime. Pressure measurements on the stator blade was used for the investigations. It consists of seven highly
and the wake as well as PIV data proved the beneficial effect of loaded stator blades and was operated at a Reynolds number of
the active flow control application to the flow field and the im- Re = 600000. A periodic non-steady outflow condition was im-
provement of the compressor characteristics. The actuation effi- posed by a choking-device mounted downstream from the trail-
ciency was evaluated by two figures of merit introduced in this ing edges, simulating the impact of a pulse detonation combuster
contribution. to the compressor stator.
2
30°
P
3 5
seven 2D/3C-PIV planes
s
Blade actuator
(e) insert
Inflow
6
FIGURE 3. AFC setup and five-hole-probe measurement plane
FIGURE 2. POSITION OF THE CHOKING-DEVICE. (a) NO PAS- differential pressure sensors. Furthermore the pressure distribu-
SAGE IS BLOCKED φ = 360◦ . (b) PASSAGE TWO IS BLOCKED tion on the mid-section of the measurement blade was measured
φ = 90◦ . (c) PASSAGE THREE IS BLOCKED φ = 140◦ . (d) simultaneously. Thirty-two First Sensor: HCL 12X5 differ-
PASSAGE FIVE IS BLOCKED φ = 250◦ . (e) PASSAGE SIX IS ential pressure transducers were installed around the centre-line
BLOCKED φ = 280◦ on the measurement blade. Twenty-one of them were mounted
on the suction side of the profile and the rest were distributed
TABLE 2. PIV-PLANES along the pressure side. In the experiments one side-wall ac-
tuator was installed at s/smax = 14.5% on each end-wall of the
plane 1 2 3 4 5 6 7 six passages. The blade actuator outlet orifices were located at
s [mm] 160 200 240 280 320 360 415 s/smax = 52.29% on the blade suction surface in the passages
two, three, four and five. Both actuator-systems were based on
x/c 0.358 0.461 0.564 0.67 0.775 0.876 1 the principle of fluid amplification and follow the design rules
presented in [15]. The flow channels of the actuators are shown
in Fig. 4. The side-wall actuators, shown in Fig. 4 (a) gen-
erate pulsed air jets under a blowing-angle of ωSWA = 45◦ with
two outlet orifices on each device that were arranged in stream-
positions of the PIV planes are given. One Litron: nanoPIV wise direction. The two outlet orifices were BSWA = 20 mm apart
L 200-15 PIV laser and two PCO: pco.edge 5.5 cameras were and measured DSWA = 0.4 mm x LSWA = 20 mm. Further infor-
used for the stereoscopic setup. The standard PIV system was mation regarding the geometry and the working principle of the
triggered by a light barrier mounted to the choking device, allow- actuator is given in [14]. An ωBA = 30◦ blowing-angle was used
ing phase locked PIV snapshots. At every phase-angle, n = 50 in the design of the blade actuator. The device’s flow channels
pictures were evaluated and averaged. Figure 3 depicts the loca- are shown in Fig. 4 (b). It can be operated as a fluidic oscilla-
tion of the measurement planes that were positioned inside and tor by connecting port 1a with 2a and 1b with 2b. In our case,
downstream from passage four. A five-hole-probe was used to the ports 1a and 1b were plugged and a solenoid valve was con-
evaluate the static pressure recovery as well as the total pres- nected to the ports 2a and 2b in order to vary mass-flow-rate
sure loss reduction of one passage with respect to the input of and switching frequency independently. The six outlet orifices
the active flow control (AFC) devices. The wake measurement of the actuator measured DBA = 0.4 mm x LBA = 13 mm and had
plane was located at c/3 downstream the trailing edges. The five- a spacing of BBA = 13 mm. Extended explanations regarding the
hole probe was equipped with five First Sensor: HCLA 12X5 blade actuator are given in [16]. The outlet signals of the actu-
Output
Supply 2b 1b The local velocity (ulocal ) was calculated from the static pres-
1a sure coefficient measured at the x/c position of the actuators.
2a time Figure 5 (a) shows the total pressure at the outlet of the side-
wall actuator. The switching frequency of the actuator was set to
+
LSWA Control port factuator,SWA = 40 Hz (FSWA ≈ 0.5) and the mass-flow-rate was
BA-outlets: ṁSWA = 6.344 · 10−4 kg/s. Figure 5 (b) shows a correspond-
ing plot regarding the blade actuator. In this case the switching
ly
BBA Supp frequency was set to factuator,BA = 60 Hz, which corresponds to
+
BSWA DSWA LBA FBA ≈ 0.5. The mass-flow-rate was ṁBA = 1.106 · 10−3 kg/s. In
DBA order to evaluate the efficiency of the actuation it is necessary to
take the power of the jets into account. The jet-power is calcu-
lated using:
ators that were measured using a Pitot-probe are plotted in Fig. The mass-flow-rate of the actuators (ṁactuator ) was measured
5. In earlier work it was shown that the optimized switching fre- through Festo: SFAB-200 flow-meters. The variable v jet,max
refers to the maximum velocity at one actuator orifice and was
calculated using the equation of continuity. Figure 6 shows the
(a) (b)
1.06 1.04
side-wall actuator blade actuator
1.03 side-wall actuator blade actuator
1.04 20 150
pitot [pa]
pitot [pa]
1.02 15
100
1.02
1.01 10
50
ambient 5
ambient
1 1
0 0.05 0.1 0.15 0 0.05 0.1 0.15
t [s] t [s] 0 0
0 0.5 1 1.5 2
-3
FIGURE 5. TIME-RESOLVED SIGNAL AT ACTUATOR OUT- mactuator 10 [kg/s]
LET ORIFICES. (A) SIDE-WALL ACTUATOR SWITCHING WITH
FIGURE 6. CALCULATED JET ENERGY OF THE ACTUATORS
factuator,SWA = 40 Hz, MASS-FLOW-RATE ṁSWA = 6.344 · 10−4 kg/s.
WITH RESPECT TO THE INPUT MASS-FLOW-RATE
(b) BLADE ACTUATOR SWITCHING WITH factuator,BA = 60 Hz,
MASS-FLOW-RATE ṁBA = 1.106 · 10−3 kg/s
y/H
In this equation, the RMS value of the jet velocity is used. 0
The pulsed jet signal is approximated with a square wave to 0.4 0.8 TE
x/c
allow the √ RMS value of the jet velocity to be calculated as (a) no passage blocked (b)
v jet,RMS = 0.5 · v jet,max [8]. The variable q1 represents the in-
flow dynamic pressure in the cascade experiments and A1 refers
0.5
to the inflow area of one passage. separated flow
y/H
separation
Measurement uncertainties backflow bubble
The manufacturer of the flow meters used in the experi- 0
0.4 0.8 TE
ments states an accuracy of ±0.8 % with respect to the full scale x/c
span, and the inflow dynamic pressure was measured with an (c) passage 2 is blocked (d)
error of ∆q1 = ±1.25 [Pa], as specified for the First Sensor:
HCLA 12X5 differential pressure sensor. Based on these val- 0.5
ues, the deviation from the quoted momentum coefficient was separated flow
always less than ∆cµ,SWA = ∆cµ,BA ≤ ±0.064 %. The static pres-
y/H
sure on the stator blade was measured with a maximum error
trailing edge
of ∆c p ≤ ±0.0052. All five-hole-probe data were evaluated in separation
terms of static pressure rise cP and total pressure loss ζ be- 0
0.4 0.8 TE
x/c
hind the measurement passage, which were calculated with com-
bined errors for the integrated values of ∆cP ≤ ±5.53 · 10−4 and (e) passage 3 is blocked (f)
∆ζ ≤ ±3.03 · 10−4 with respect to repeatability.
0.5
separated flow
y/H
3 REFERENCE FLOW
Figure 7 shows results obtained from 2D/3C-PIV measure- trailing edge
ments in the stator cascade. The periodic motion of the choking separation
0
device forces highly dynamic and recurring flow structure in the 0.4 0.8 TE
x/c
passages, which were resolved using a triggered standard PIV (g) passage 5 is blocked (h)
system. In the contour plots of Fig. 7 (a) the axial velocity ratio
regarding the phase-angle φ = 360◦ is shown, where no passage
was blocked. At this phase-angle the flow-field was least dis- 0.5
separated flow
turbed. The strong structures of the end-wall flow are indicated
y/H
z/P
z/P
hit the suction surface further upstream and took up more space 0.4 0.4
in the passage. Lower axial velocity occurs close to the blade
surface. The surface streamlines indicate a grown corner separa- 0.2 0.2
tion that is already reaching the midsection of the measurement SS 0 SS
0
blade. An iso-line for vax = 0 m/s shown in purple indicates the 0 0.2 0.4 0 0.2 0.4
flow separation line. Along the vortex separation line a separa- 2y/H 2y/H
tion bubble started to form from x/c = 0.5 on.
FIGURE 8. PRESSURE RECOVERY OF THE MEASUREMENT
Increasing flow separation was observed as the choking de- PASSAGE FOR PHASE-ANGLE φ = 250◦ (PASSAGE FIVE IS
vice blocked passage number three as shown in Fig. 7 (e) and (f). BLOCKED). (a) REFERENCE FLOW (b) WITH ACTIVE CON-
The surface streamlines strongly bent from the end wall towards TROL: MOMENTUM COEFFICIENT OF SIDE-WALL ACTUATION
the midsection of the measurement blade and the separation bub- Cµ,SWA = 1% AND MOMENTUM COEFFICIENT OF BLADE AC-
ble broke down and reached the trailing edge, resulting in a large TUATOR Cµ,BA = 0.5%
pressure induced trailing edge flow separation.
The flow separation extended to its maximum when passage
five was blocked as shown in Fig. 7 (g) and (h). The data show static pressure coefficient cP was defined by
an extended area with backflow, indicated by the purple iso-line
in Fig. 7 (h) from x/c = 0.75 until the trailing edge for the blades p2 (y, z) − p1
cP = , (4)
centre section. From y/H = 0 to y/H = 0.2, a two dimentional q1
flow separation was observed. Between y/H = 0.2 and y/H =
0.5, a strong impact of the passage vortex became evident. The where p2 represents the measured static pressure at a certain y
contour plot in Fig. 7 (g) shows the three dimensional structures and z position in the five-hole-probe measurement plane, where
of the flow-field when passage five was blocked. N = 240 positions were evaluated. The variables p1 and q1 rep-
The trailing edge flow separation reached far into the pas- resent the inflow conditions. Considering Fig. 8 (a), no flow con-
sage flow-field displacing the indicated streamlines away from trol was applied. The pressure recovery was limited due to the
the blade’s suction surface. corner separation and the pressure induced flow separation on the
The stator vane recovers from the strong flow separation stator blade as found in Fig 7 (g) and (h). The lowest values of cP
phenomena when passage six was blocked. Figure 7 (k) shows were found in the direct wake of the centre blade flow separation.
the pressure induced flow separation from x/c = 0.8 until the Applying AFC through the side-wall- and the blade actuators an
trailing edge. The reduced flow separarion also becomes evident overall increase in pressure recovery was observed as shown in
in the passage flow-field shown in Fig. 7 (j). Fig. 8 (b). In this case, the overall momentum coefficient was set
to cµ = 1.5%, where the side-wall actuators operated at a mo-
mentum coefficient of cµ,SWA = 1% and the blade actuators with
cµ,BA = 0.5%. The dimensionless actuating frequency of both ac-
tuator systems was kept to the corresponding Strouhal numbers
4 ACTUATED FLOW + +
of FSWA = FBA = 0.5. Similar to the non actuated case, the data
The compressor stator flow-field showed a dynamic corner field shows the highest pressure recoveries in the centre of the
stall and periodic pressure induced trailing edge flow separation. passage for pitch wise values from z/P = 0.5 to z/P = 1. Below
Two actuator systems were installed to prevent both flow sepa- that pitch wise coordinate, the three dimensional flow structures
rations. The influence of the side wall actuation on the dynamic in the passage still led to lower local pressure recoveries. Adjust-
flow-field was discussed in [14] and [18]. ing the mass-flow-rate through the actuators results in a variation
Figure 8 depicts the pressure recovery of the stator vane at of the momentum coefficient cµ . The benefits of the actuating
a critical phase-angle of φ = 250◦ with and without active flow momentum were evaluated in terms of static pressure recovery
control. At this phase-angle, passage five was fully blocked. The as presented in Eq. 4 and total pressure loss coefficient measured
0.68
The corrected inflow total pressure pt,1,corrected was calculated as
follows. 0.18
0.66
(pt,1 · ṁ1 + pt, jet · ṁ jet ) SWA only
reference flow
cP,AFC /ζAFC
FM2: 1.1 1.2 best FM2 FM2 = (11)
2 best FM1 cP,noAFC /ζnoAFC
FM1 -0.5
transition
High pressure recoveries combined with low total pressure losses
cµ,SWA [%]
0.5 to the reference case without AFC. The value for ζAFC was cal-
1.
increasing pressure culated using Eq. 8. In the contour plot of Fig. 10 (a) the re-
0 sult of the second figure of merit is shown. Only values regard-
0 1 2 0 SWA BA 1
cµ,BA [%] LE x/c TE ing FM2 >= 1 are shown in the contour plot. The optimum
AFC setup is now found for cµ,SWA ≈ 0.6% and cµ,BA ≈ 0.3%.
FIGURE 10. (a) EFFICIENCIES OF THE AFC APPROACH FM1
The increase in the total pressure loss coefficients after reach-
AND FM2. (b) STATIC PRESSURE DISTRIBUTION ON THE MEA-
ing cµ,SWA + cµ,BA ≈ 1% leads to decreasing values regarding
SUREMENT BLADE AT PHASE-ANGLE φ = 250◦ .
the second figure of merit FM2. Outside the optimum effective
AFC setups still occur regarding a large range of momentum co-
efficients.
pressure was used and the first figure of merit is calculated by: Figure 10 (b) shows the pressure distribution around the
midsection of the measurement blade for the reference case com-
∆PT E,AFC pared to two actuating cases for the phase-angle φ = 250◦ (pas-
FM1 = [16]. (9) sage five was blocked). In all cases the transition takes place
∆PT E,noAFC + (Pjet,BA + Pjet,SWA )
through a laminar separation bubble at the position x/c = 0.17.
For the reference case it was found that a large pressure induced
The variable ∆PT E describes the physical energy conversion trailing edge flow separation occurred at x/c = 0.6 in the mid-
through the stator based on the static pressure difference at the section of the measurement blade that limited the pressure re-
trailing edge of the stator blade to the inflow static pressure and covery from that point on.
is expressed by: When the actuation was operated corresponding to the op-
timum found in the second figure of merit, the pressure distri-
bution shown by the green line in Fig. 10 (b) resulted. Higher
v21
∆PT E = c p,T E · · ṁ. (10) pressure recovery on the trailing edge was measured and the re-
2 gion with flow separation was reduced but not fully suppressed.
A steady increase of static pressure until reaching the trailing
The static pressure coefficient c p,T E was measured at the posi- edge was measured when the optimum found by the first figure
tion x/c = 0.956 in the mid-section of the measurement blade. of merit was used for AFC.
In Eq. 9 the energy conversion of the stator passage with AFC The two different AFC setups led to static pressure recov-
∆PT E,AFC is related to the energy conversion of the stator vane eries that are depicted in Fig. 11 (a). In the polar plot, phase-
without active flow control ∆PT E,noAFC and the jet energies in- averaged static pressure recoveries are shown. Comparing the
serted by the side-wall and blade actuator Pjet,BA and Pjet,SWA . reference case to the actuated ones it becomes apparent that an
Figure 10 (a) shows the evaluation of Eq. 9 in the contour lines actuation with the optimum found in the first figure of merit and
with regard to the momentum coefficients of the side-wall and the second figure of merit led to increases in static pressure re-
the blade actuator. The contour line for FM1 = 1.0 depicts the covery at all phase-angles.
break even, where the amount of input actuation energy equals The green line depicts the case when the second figure of
the increase of the energy conversion that was measured at the merit was maximum. In this case the overall momentum coeffi-
trailing edge. Increasing values indicate more efficient AFC set- cient of the actuators was set to cµ,BA + cµ,SWA = 0.9%. Increas-
ups. The optimum became apparent for momentum coefficients ing the momentum to cµ,BA + cµ,SWA = 1.4% led to a further in-
of the side-wall actuator of cµ,SWA ≈ 1% and cµ,BA ≈ 0.4% re- crease regarding the static pressure recovery and depicts the case
garding the blade actuator. where the first figure of merit had its maximum.
In order to calculate the efficiency of the AFC by taking In Fig. 11 (b) the phase-averaged total pressure losses be-
static pressure recovery and loss reduction into account, a sec- come evident. The equation used to calculate the second figure of
sur f a c rfac
red
180° 0° 180° 0° - s u
0.5 0.6
uct
0.05 0.1 iso- iso
cP (a) (b)
ion iso-Q-surface
passage vortex
BA-outlets
270° 270°
attatched flow
FIGURE 11. IMPACT OF THE AFC TO THE FLOW-FIELD RE-
BA-outlets 0.5
GARDING THE OPTIMUMS OF THE FIRST- AND THE SECOND
FIGURE OF MERIT. (a) PHASE-AVERAGED STATIC PRESSURE (c) (d)
RECOVERY OF THE STATOR VANE EVALUATED IN THE FIVE-
HOLE-PROBE PLANE. (b) PHASE-AVERAGED TOTAL PRES- FIGURE 12. PASSAGE FLOW-FIELD MEASURED WITH 2D/3C-
SURE LOSS IN THE FIVE-HOLE-PROBE PLANE. PIV AT THE PHASE-ANGLE φ = 250◦ . (a) REFERENCE FLOW.
(B) ONLY SIDE-WALL ACTUATION Cµ,SWA = 1.6%. (c) SIDE-
WALL ACTUATION AND BLADE ACTUATION WITH Cµ,SWA =
merit accounts the total pressure losses in the stator vane. When 1.6% AND Cµ,BA = 1.6% (d) SURFACE STREAMLINES FOR THE
the optimum momentum coefficients based on the second figure ACTUATED CASE WITH Cµ,SWA = 1.6% AND Cµ,BA = 1.6%
of merit were applied as seen in the green line plot it was found
that the total pressure losses were reduced at every phase-angle.
The higher momentum used in the optimum based on the The pressure induced flow separation in the centre blade is not
first figure of merit led to loss reductions for a wide range of reduced at all by the use of the side-wall actuation. When the
phase-angles but between φ = 36◦ to φ = 180◦ increasing total blade actuator was added with the same momentum coefficient
pressure losses occurred, due to the higher actuation power input. as the side-wall actuator (cµ,SWA = 1.6% and cµ,BA = 1.6%) the
flow-field forms as shown in Fig. 12 (c). Here the passage vortex
Figure 12 shows the flow structures in the stator vane structure was still comparable to the case where only the side-
that were conducted in the measurement passage using the wall actuator was operative. Furthermore no region with sepa-
2D/3C-PIV setup for the phase-angle of φ = 250◦ (passage five rated flow was found on the stator blade. The surface stream-
was blocked). The Q-criterion was evaluated in every two- lines on the suction surface of the measurement blade are shown
dimensional PIV-plane to depict the passage vortex at every in- in Fig. 12 (d). Bent streamlines were found at the blade-wall
vestigated phase-angle. The result is shown in the green iso- junction that were induced by the passage vortex. Further to-
surface for Q = 3500 1/s with linear interpolation between the wards the centre section of the blade, the streamlines show fully
PIV-planes. The planes indicate the axial velocity ratio and a attached flow.
second iso-surface in grey for vax = 0 indicates the separation
line. In Fig. 12 (a) the reference case is shown. The intense trail-
ing edge flow separation shows in the vax = 0 iso-surface. The 5 CONCLUSIONS
early breakdown of the corner vortex and the mixing of it with Periodically occurring flow separation in a compressor sta-
the pressure induced flow separation leads to the Q iso-surface tor vane induced by a non-steady outflow condition were inves-
not showing a connected passage vortex structure. When the tigated. The wake of a two-dimensional low speed stator cas-
side-wall actuation was turned on with a momentum coefficient cade was periodically choked to simulate the impact of a con-
of cµ,SWA = 1.6% the flow-field formed as shown in Fig. 12 (b) stant volume combuster to the stator. 2D/3C-PIV measurements
regarding the same phase-angle. A reduction of the corner sepa- revealed the three dimensional flow structures and the struc-
ration was achieved and the clearly formed passage vortex is in- tures of the flow separation phenomena. Phase-averaged data
dicated by the iso-Q-surface of Q = 3500 1/s. The flow losses re- showed a separation bubble along the passage vortex separation
lated to the corner separation were shifted toward the end-walls. line that breaks down, reaches the trailing edge and forms the