2019 - J. Klinner - Transonig Blade Compressor CFD

Journal of Turbomachinery.
Received September 11, 2018;

Accepted manuscript posted March 26, 2019. doi:10.1115/1.4043329
Copyright © 2019 by ASME
THE PRESENT CHALLENGE OF TRANSONIC

COMPRESSOR BLADE DESIGN
d
Alexander Hergt∗
te
J. Klinner, J. Wellner, C. Willert, S. Grund, W. Steinert, M. Beversdorff
di
German Aerospace Center (DLR), Institute of Propulsion Technology
e
51147 Cologne, Germany
py
Co
Abstract
ot
The flow through a transonic compressor cascade shows a very complex structure
tN
due to the occuring shock waves. In addition, the interaction of these shock waves with
rip
the blade boundary layer inherently leads to a very unsteady flow behaviour.
The aim of the current investigation is to quantify this behaviour and its influence on
sc
the cascade performance as well as to describe the occuring transonic flow phenomena
nu
in detail. Therefore, an extensive experimental investigation of the flow in a transonic

Ma
compressor cascade has been conducted within the transonic cascade wind tunnel of
DLR at Cologne. In this process, the flow phenomena were thoroughly examined for an
ed
inflow Mach number of 1.21. The experiments investigate both, the laminar as well as
pt
the turbulent shock wave boundary layer interaction within the blade passage and the
ce
resulting unsteady behaviour. The experiments show a fluctuation range of the passage
Ac
shock wave of about 10 percent chord for both cases, which is directly linked with a
change of the inflow angle and of the operating point of the cascade.
∗
Corresponding author: Address: German Aerospace Center (DLR), Institute of Propulsion
Technology, 51147 Cologne, Germany; phone: 0049-2203-601-2217; fax: 0049-2203-64395; e-
mail:[email protected]
Hergt TURBO-18-1244 1
Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Journal of Turbomachinery. Received September 11, 2018;
Thereafter, RANS simulations have been performed aiming at the verification of the re-
producibility of the experimentally examined flow behavior. Here it is observed that the
dominant flow effects are not reproduced by a steady numerical simulation. Therefore,
a further unsteady simulation has been carried out in order to capture the unsteady
d
flow behaviour. The results from this simulation show that the fluctuation of the pas-
te
di
sage shock wave can be reproduced but not in the correct magnitude. This leads to a
e
remaining weak point within the design process of transonic compressor blades, because
py
the working range will be overpredicted.
Co
The resulting conclusion of the study is that the use of scale resolving methods such
as LES or the application of DNS is necessary to correctly predict unsteadiness of the
ot
tN
transonic cascade flow and its impact on the cascade performance.
rip
1 Introduction
sc
nu
The primary goal in the development of modern turbomachinery is the increase of efficiency
with a corresponding reduction of fuel consumption and has not changed significantly during
Ma
the past decades. In this context, the improvement of the aerodynamic efficiency of tran-
ed
sonic fan and compressor blading imposes high demands on the design process due to the
pt
very complex flow structures which are caused by shock waves and their interaction with the
ce
blade boundary layer.

Ac
One of the first investigations on shock waves on a single airfoil were performed by Ferri [1]
in the late 1930’s. In this study the shock formation and behavior is described depending
on the aerodynamic loading. Since that time, the research on transonic airfoils or compres-
sor bladings is focused on the description and understanding of the flow effects which are

related to the existence of shock waves [2, 3]. Within the studies of Liepmann [4] as well
as of Fage and Sargent [2] the shock wave boundary layer interaction is charcterized as an
essential transonic flow effect. This means that depending on the shock strength the foot
of the shock wave interacts with the surface boundary layer which leads to particular flow
d
te
situations. In this context within the investigation of Mundell and Mabey [5] as well as Lee
di
[6] three flow situations were defined. In the case of a weak shock the interaction with the
e
py
boundary layer leads to a thickening of the downstream boundary layer development. An
increase of the shock strength results in a separation under the shock foot and a downstream
Co
reattachment of the boundary layer. If the case of a very strong shock wave the boundary
ot
layer is separated up to the trailing edge. Of course, the described behaviour influences the
tN
performance of a compressor as investigated by Epstein [7].
rip
In addition, the unsteadiness of the shock wave location is also an inherent flow phenomena
sc
which is already shown in the study of Ferri [1]. In the publication of Hilton and Fowler
nu
[8] a visualization of this shock wave movement is shown. A concluding description and
Ma
quantification of the reason for the shock wave oscillation could not yet be given [9]. Dif-
ferent theories exist in the literture which have been investigated in the past. Beresh et al.
ed
[10, 11] investigated the influence of the the boundary layer upstream of the shock wave.
pt
They found that there is no fluctuation of the thickness of the incomming boundary layer
ce
but there is a correleation between the shock movement and the boundary layer shape. An
Ac
upstream shock motion correlates with a slimmer boundary layer and a downstream shock
motion correlates with a fuller boundary layer.
Furthermore, the studies of Ross [12], Thomas et al. [13] and Dussauge et al. [14] show
that in the case of the occurance of a separation bubble underneath the shock foot, the

unsteadiness of this separation influences the shock wave movement. In a further study of
Dussauge and Piponniau [15] it is stated that the shock wave oscillation also depends on the
turbulence intensity of the inflow as well as on an existing separation at the trailing edge or
on the turbulence structure of the wake. The influence of the downstream flow behavior on
d
te
the shock wave oscillation was also shown by Hartmann et al. [16].
di
In addition to the studies described above the investigation of Hergt et al. [17] show that
e
py
the shock oscillation range in a transonic fan cascade is in the region of up to 10 percent
chord length. A comparable magnitude of the shock oscillation in a singel rotor is shown by
Co
Strazisar [18]. In this study the shock oscillation range is in the region of up to 8 percent
ot
chord length. Based on the literature review as well as on the experimental experiences at
tN
the DLR Transonic Cascade Wind Tunnel it becomes clear that the shock oscillation has a
rip
significant impact on the working characteristics (performance, working range) of a transonic

sc
compressor blade cascade. In order to achieve a deeper understanding of the effects of shock
nu
wave boundary layer interaction and the their influence on the transonic compressor blade
Ma
design the lack of very highly accurate experimental data has to be solved [9].
Therefore, the aim of the investigation is to quantify the unsteady shock wave behaviour
ed
and its influence on the cascade performance. Furthermore, the occuring transonic flow
pt
phenomena is described in detail by means of an extensive experimental investigation of the

ce
transonic compressor cascade flow. The performed experiments at the DLR Transonic Cas-
Ac
cade Wind Tunnel include standard pressure measurement. In addition the velocities of the
shock boundary layer interaction (SBLI) on the blade’s suction side have been captured by
particle image velocimetry (PIV). The unsteady aerodynamic loading of the cascade results
in flexure of the blade surface (i.e. buffeting) which has been measured and compensated

by a correlation-based approach prior to PIV processing. Aside from motion of the passage
shock position, single PIV shots also indicate shape variations of the lambda shock system
and of the associated separation region. In order to quantify the mean flow velocities of the
SBLI region large PIV data sets are conditionally averaged upon the instantaneous passage
d
te
shock positions at a resolution of 1% of chord length. Furthermore, a frequency analysis of
di
passage shock motion was conducted with high spatial and temporal resolution using time
e
py
resolved shadowgraphs.
Finally, steady and unsteady numerical simulations are carried out in order to answer the
Co
question how well state-of-the-art numerical methods can predict the transonic flow be-
ot
haviour in the blade passage.
tN
rip
sc
nu
Ma
ed
pt
ce
Ac

2 EXPERIMENTAL SETUP
The experimental part of the investigation was performed at the Transonic Cascade Wind
Tunnel [19, 20] at DLR in Cologne. This test facility allows investigations on compressor
d
blade cascades under real turbomachinery conditions regarding Mach number and Reynolds
te
number. An inlet Mach number of up to 1.4 can be achieved. Furthermore the Mach number
di
and the Reynolds number can be varied independently. Since the wind tunnel is operated
e
py
continuously in a closed loop it is possible to vary the Reynolds number from 1x105 to 3x106
Co
by setting the total pressure in the system.
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac

2.1 Cascade Configuration
The configuration studied is a transonic cascade consisting of 6 blades. The blades have a
chord length of 100 mm and a blade height of 168 mm. The Fig. 1 shows the test cascade
d
used. The design of the cascade was derived from an engine configuration and provided
te
for testing by Rolls Royce Germany as part of the EU research project TFAST (Transition
di
Location Effect on Shock Wave Boundary Layer Interaction).
e
py
Co
The Tab. 1 shows an overview of the design parameters of the cascade and the main test
conditions. All measurements reported are carried out at an inflow Mach number of 1.21.
ot
During the investigation, measurements of two flow cases were performed: The first is the
tN
laminar case which means that on the suction side of the blade an laminar boundary layer
rip
exist up to the chord length position where the shock wave interacts with the boundary layer
sc
nu
Ma
ed
pt
ce
Ac
Figure 1: TRANSONIC CASCADE WITH PLANAR ENDWALL

(laminar SBLI). In the second flow condition the transition takes place a certain distance
upstream of the shock wave and in which case the shock wave interacts with a turbulent
boundary layer (turbulent SBLI).
Due to the low turbulence level of the wind tunnel inflow the boundary layer typically is
d
te
laminar on the blade suction side. In order to achieve the test conditions where the transition
di
onset is shifted upstream of the shock boundary layer interaction region the turbulence level
e
py
of the inflow has to be increased. To this end a turbulence grid which has been placed
upstream of the test section (based on the work of Hoheisel [21]).
Co
Table 1: MAIN TEST CONDITIONS AND DESIGN PARAMETER OF THE CASCADE
ot
tN
rip
Inlet Mach number M1 = 1.21

sc
Reynolds number Re ≈ 1.4×106

nu
Inlet flow angle β1 = 145.6 deg

Ma
Flow turning at ADP ∆β = 14.9 deg
AVDR = 1.2
ed
Stagger angle βst = 134.4 deg

pt
Blade chord length c = 100 mm

ce
Pitch t = 60 mm
Ac
Blade span h = 168 mm
Pitch to chord ratio t/c = 0.6

3 EXPERIMENTAL METHODS
In addition to standard pressure measurements, the experiments also included L2F measure-
ments [22, 23, 24], PIV measurements and high-speed shadowgraphy [25, 26]. To determine
d
the adiabatic wall temperature, liquid crystals [27, 28] were used on the blade suction side,
te
allowing an assessment of the boundary layer condition and an occurring transition.
e di
py
The Fig. 2 gives an overview of the structure of the test section and the defined mea-
Co
surement levels. This figure also shows the position of the PIV lens used and the Region of
Interest (ROI). There the shock boundary layer interaction as well as the shock position were
ot
tN
investigated. In the inlet measurement level (MP 1) the static inlet pressure was measured
at the endwall. The wake measurement with a 3-hole probe was carried out in the exit
rip
measuring plane (MP 2). All pressure data given in this paper are time averaged. In the
sc
Fig. the adjustable upper transonic wall is to be emphasized further. This is used to achieve
nu
the supersonic inflow conditions. In addition, the reflection of the shock waves is reduced by
Ma
this wall.
ed
pt
ce
Ac
Figure 2: CROSS-SECTION DRAWING OF THE TEST SECTION

d
te
e di
py
Figure 3: DEFINITION OF THE MEASUREMENT PLANES AND CASCADE PARAM-
Co
ETERS (LEFT); DESIGN OF THE ENDWALL BOUNDARY LAYER SUCTION SLOTS
ot
(RIGHT)
tN
As shown in the Fig. 3, the distribution of the inflow angle in front of the cascade is
rip
measured at midspan in the L2F measuring plane. In this figure, the right side also shows
sc
the design of the two-sided suction slots in the passages of the cascade. These suction slots
nu
for the local suction of the boundary layer are necessary to be able to set the AVDR [29, 30]
Ma
for each measuring point. The setting of a given AVDR enables the comparability of the
results when different cascade configurations are used, since this is the only way to achieve a
ed
comparable cascade loading. Static pressure tabs in the blade midspan are used to determine
pt
the isentropic Mach number distribution.

ce
The uncertainty of the result data regarding the average inlet and outlet flow angle as
Ac
well as the losses are shown in the table 2.

Table 2: ESTIMATED UNCERTAINTIES OF THE TEST DATA AT MID-SPAN
Upstream flow angle β1 +/- 0.2 deg
Exit flow angle β2 +/- 0.2 deg
d
te
Mid-span loss coefficient ω +/- 0.002
e di
py
Co
Figure 4: SCHEMATIC DIAGRAM OF THE SHOCK SYSTEM AND LOCATIONS OF
ot
tN
PIV MEASUREMENT REGIONS; REGION A: DISTRIBUTIONS OF THE PASSAGE
rip
SHOCK POSITION; REGION B: MEASUREMENTS OF SBLI AND SEPARATION RE-

sc
GION
nu
3.1 PIV Instrumentation and Image Evaluation

Ma
A planar 2-Component PIV setup is applied using a classical normal viewing arrangement
ed
with the measurement plane located at midspan. Statistical evaluation of the passage shock
position based on a small field of a view of 22 × 2.5 mm2 (Region A in Fig. 4). PIV mea-
pt
ce
surements of the SBLI were conducted for a field of view of 24 × 9.5 mm2 that covers the
Ac
shock foot and boundary layer on blade’s suction side (Region B in Fig. 4). More details
regarding the PIV measurement setup are described in [26].
The unsteady aerodynamic loading of the thin blade results in its flexure (i.e. vibrations)

d
te
e di
py
Co
ot
tN
rip
sc
Figure 5: SINGLE PIV SHOTS OF THE SBLI FOR THE TURBULENT CASE (VECTORS
nu
AT u¿200 m/s ARE CLIPPED); NEARLY ATTACHED FLOW AND WEAK OBLIQUE
Ma
SHOCK
which is evidenced by vertical image displacements of several pixels. Using a correlation

ed
based approach this blade motion can be tracked and is used to register the data grid prior
pt
to PIV processing (see also [31] chapter 11.2.3). The PIV image data is processed using a
ce
coarse-to-fine multi-grid processing scheme with image deformation at each step to take into
Ac
account the strong shear in the flow. The sub-pixel correlation peak position measurement
was performed by a truncated sinc signal reconstruction algorithm. Due to this processing
and a sufficient particle image density an accuracy of correlation peak detection below 0.1

d
te
e di
py
Co
ot
tN
rip
sc
nu
Figure 6: SINGLE PIV SHOTS OF THE SBLI FOR THE TURBULENT CASE (VEC-
TORS AT u¿200 m/s ARE CLIPPED); LARGE FLOW SEPARATION WITH STRONG
Ma
OBLIQUE SHOCK AND TRANSITIONAL SEPARATION BUBBLE INCLUDING RE-

ed
VERSE FLOW
pt
pixel can be obtained for regions without strong laser background. Given a magnification of
ce
10.0 µm/pixel an absolute velocity uncertainty of 2 m/s can be achieved for measurements of
Ac
the SBLI in region B. The final vector spacing for region A (passage shock) is 0.18×0.18 mm2
at a interrogation window size of 0.70 × 0.70 mm2 . For region B (SBLI) the final vector
spacing is 0.12 × 0.06 mm2 at a interrogation window size of 0.48 × 0.24 mm2 . In order to

obtain statistical distributions of the passage shock position up to the 5 000 PIV samples
are evaluated for region A in Fig. 4. The unsteady shock position in each single shot is
obtained by matching the tracer velocity across the normal shock with an exponential fit.
The instantaneous shock positions is taken at the onset of the exponential decay of the tracer
d
te
velocity [32, 26].
di
Near the suction side surface both the shock foot and the separation region fluctuate by
e
py
±5 mm in chord-wise direction (±5% of chord or ±446 pixel). Furthermore, the SBLI is
highly unsteady which is evidenced in two single PIV shots at turbulent conditions shown in
Co
Figs.5 and 6. Neither the passage shock position nor the extension of the separation region
ot
are steady. Reverse flow partially becomes visible in single PIV shots as shown in Fig. 6.
tN
In order to specify whether the mean flow on the suction side is attached, separated or
rip
even exhibits reverse flow, conditional averaging was conducted based upon the position of
sc
the shock foot in each random PIV snapshot of the SBLI using a total of 10 000 PIV samples
nu
for region B (see Fig. 4). After PIV processing, including compensation of blade vibrations,
Ma
the shock foot position is estimated based on the velocity field in each sample. Conditional
averaging was performed over distinct PIV samples which are sorted after discrete shock po-
ed
sitions at a spatial resolution of 1 mm. To highlight the reverse flow region in the conditional
pt
averages, the dividing streamline at zero net mass flux is computed by direct integration of
ce
the velocity profile along the y-axis (c.f. [33]). In this way, the spatial extension of the shock
Ac
induced flow separation can be estimated.

3.2 Schlieren and High-Speed Shadowgraphy Setup and Process-
ing
A conventional Schlieren optic was used to assess the shock system in the middle measuring
d
passage between blade 3 and 4 and to check the flow periodicity in the pitch direction over
te
the entire cascade. A conventional Schlieren optic was used to assess the shock system in
e di
the middle measuring passage between blade 3 and 4 and to check the flow periodicity in
py
the pitch direction over the entire cascade. Since this optical setup does not allow spatial
Co
and temporal resolution of the shock oscillation in the measuring passage, a time-resolved
shadowgraphy system with pulsed LED back-illumination [34] was used. The temporal reso-
ot
tN
lution of this system is 20 kHz. The high speed camera (Photron SA-5) is equipped with an
Nikkor Micro f=200 mm lens at a magnification of 35 µmm/Pixel. The field of view covers
rip
a region of 24 × 4.42 mm2 at midspan which is centered with respect to region A in Fig. 4.
sc
More details on the high speed shadowgraphy system can be found in [26].
nu
The position of the passage shock is tracked in each image based upon the maximum absolute
Ma
intensity gradient along x. The power spectral distribution of shock motion is obtained by
ed
Fourier analysis of a shock motion over time using image sequences of up to 127 000 samples.
The spectral analysis is based on averaging the FFTs of a temporally sliding window of a
pt
ce
fixed width and overlap [35].

Ac

3.3 Liquid Crystal
Knowledge of the condition of the suction side boundary layer during the tests was necessary
to ensure that it was laminar or turbulent. For this purpose the liquid crystal measurement
d
technique [27, 28] was used. With this technique it is possible to visualize the separation and
te
transition behavior of boundary layers. The wall temperature difference is determined due
di
to the different heat transfer as described by Steinert and Starken [28]. In order to minimize
e
py
the measurement uncertainty, it is necessary to minimize the heat transfer from the flow
Co
to the blade. Therefore, a carbon fiber blade was used for the measurement. This results
in an almost adiabatic wall temperature distribution on the blade suction side. Since light
ot
reflections can also influence the visual measurement result, the blade was blackened. The
tN
liquid crystal mixture used was R20C10W from LCR Hallcrest. The liquid crystals have a
rip
temperature range of 10◦ C. The starting temperature is 20◦ C with a usable color spectrum
sc
from red (starting temperature) to yellow to green. Detailed information on the design of
nu
the required temperature spectrum for the measurements and evaluation can be found in
Ma
Schreiber et al. as well as Hergt et al. [36, 17].

ed
pt
ce
Ac

4 CASCADE STUDY RESULTS
The detailed measurements of the cascade at laminar and turbulent case were performed at
the ADP of the cascade. Figure 7 shows the measured inflow angle distribution for both
d
cases, measured at the L2F measurement plane (cf. Fig. 3). The figure shows that both the
te
e di
py
Co
ot
tN
rip
Figure 7: INFLOW ANGLE DISTRIBUTION OF THE LAMINAR AND TURBULENT

sc
CASE (L2F-MEASUREMENT PLANE AT MIDSPAN)

nu
Ma
ed
pt
ce
Ac
Figure 8: ISENTROPIC MACH NUMBER DISTRIBUTION AT MIDSPAN OF THE LAM-
INAR AND TURBULENT CASE

Table 3: MEASURED AVERAGED FLOW CONDITIONS OF THE LAMINAR AND
TURBULENT CASE
d
te
laminar turbulent
di
Inlet Mach number M1 = 1.21 = 1.21
e
≈ 1.38×106 ≈ 1.37×106
py
Reynolds number Re
Inlet flow angle β1 = 145.5 deg = 145.6 deg
Co
AVDR = 1.21 = 1.20
ot
Outlet flow angle β2 = 130.0 deg = 130.1 deg
tN
rip
shape of the distribution as well as the position of the bow shock is in a very good agreement.
sc
The averaged inflow angle of the cascade which was calculated by using conservation laws
nu
of fluid motions amounts to 145.5◦ for the laminar case and 145.6◦ for the turbulent case.
Ma
Table 3 shows the measured flow conditions of both cases and illustrate that both cases have
comparable test conditions.

ed
pt
In Fig. 8 the Profile Mach number distribution of the laminar and turbulent case is shown.
ce
This figure indicates that in both cases the average shock position on the blade suction side
Ac
seem to nearly coincide. In addition, the difference between a laminar and a turbulent shock
wave boundary layer interaction is also visible. In front of the deceleration region, which is
caused by the pressure rise over the shock, a Mach number peak is depicted for the turbulent

d
te
e di
py
Co
ot
tN
rip
sc
nu
Ma
ed
Figure 9: LIQUID CRYSTAL MEASUREMENT ON BLADE SUCTION SIDE OF THE

pt
LAMINAR (TOP) AND TURBULENT (BOTTOM) CASE

ce
case. This Mach number peak disappears for the laminar case and furthermore the shock
Ac
(deceleration region) is more smeared which means that the gradient of the Mach number
decrease is lower for the laminar case.
The Fig. 9 shows exemplary results of the liquid crystal visualizations. From this figure

d
te
e di
py
Co
Figure 10: TOTAL PRESSURE RATIO DISTRIBUTION IN THE WAKE BEHIND THE
CASCADE OF THE LAMINAR AND TURBULENT CASE (MIDSPAN)
ot
tN
the boundary layer behavior at both cases (top: laminar; bottom: turbulent) is observable.
The employed liquid crystals have a visible color range of 10◦ C starting from 20◦ C For adia-
rip
batic wall temperatures on the blade surface beyond this range the liquid crystal are black.
sc
Therefore, the black region close to the leading edge visible in the figure for both cases which
nu
results from the suction peak. Close to the endwalls the characteristic corner separation is
Ma
also visible as a black region. Furthermore, in the upper part of the figure the separation
bubble above the shock wave in the laminar case is depicted. This separation bubble disap-
ed
pears for the turbulent case.

pt
ce
As an intermediate result it can be stated, that both cases have comparable test conditions
and by means of the liquid crystal measurements it is confirmed that different boundary
Ac
layer conditions (laminar vs. turbulent) exist in front of the shock boundary layer interac-
tion region.

In the following the differences in flow behaviour of both reference cases will be investi-
gated in more detail. Figure 10 shows the pitch dependent total pressure ratio distribution
of the laminar and turbulent case, which represents the loss behaviour. It is observed that
the distribution of the shock losses [37] is slightly influenced for the turbulent case. The
d
te
results in Fig. 10 also show that the shock loss level at the pressure side is lower than at
di
the suction side which represents a not fully periodic outflow behind the cascade. This is
e
py
an inherent effect at transonic and supersonic cascade wind tunnels and can only reduced
by test adjustment but not fully avoided. A detailed explanation of this effect is given by
Co
Schreiber et al. [19].
ot
In addition, the viscous losses represented by the wake are significantly reduced for the turbu-
tN
lent case. This means, that the shock interaction with the turbulent boundary layer produces
rip
lower losses than in the laminar case where a separation bubble occurs. The comparison of
sc
the Schlieren figures of Fig. 11 (laminar) and Fig. 12 (turbulent) shows slight differences of
nu
the shock structure for both cases, which corresponds with the slightly different shock loss
Ma
distributions. In Figs. 11 and 12 sketches of the shock structure which are derived from the
schlieren images are shown in order to achieve a better understanding of the shown shock
ed
system.
pt
In both cases there is a detached bow shock with a small subsonic region between the shock
ce
and the leading edge. Close to the leading edge a lip shock [38] on the suction and pressure
Ac
side is visible, which is typically caused by the accelerated flow around the leading edge and
it indicates a small separation there. The theory behind the lip shock is based on the flow
situation which occurs if transonic and supersonic flow meet a backward face step which
leads to local expansion. In this case the typical lip shock occurs which means a separation

d
te
e di
py
Co
ot
tN
Figure 11: SCHLIEREN PATTERN (TOP) AND SCHEMATIC DIAGRAM OF THE
rip
SHOCK STRUCTURE (BOTTOM) OF THE LAMINAR CASE

sc
shock. For a transonic compressor blade the occurring of the lip shock depends on the lead-
nu
ing edge shape. In the study of Hergt et al [17] it is shown that the lip shock not occurs
at the thin round leading edge of the baseline blade but in the case of a blunt leading edge
Ma
(extrem case) a very strong lip shock exist.

ed
pt
At the investigated operating point of the cascade, the bow shock and the passage shock
ce
meet each other at the shown bifurcation. Downstream of this bifurcation a slip line exists
Ac
[39], which represents a shear layer caused by the different downstream velocities above and
underneath the origin point of the bifurcation. In addition, the comparison of Figs. 11 and
12 shows two main differences which are responsible for the different loss distributions in
the wake. The lambda foot of the passage shock in the laminar case is an indication for the

d
te
e di
py
Co
ot
tN
Figure 12: SCHLIEREN PATTERN (TOP) AND SCHEMATIC DIAGRAM OF THE
rip
SHOCK STRUCTURE (BOTTOM) OF THE TURBULENT CASE

sc
existence of a separation bubble underneath it, which is responsible for the higher viscous
nu
losses. In the turbulent case, the inclination of the passage shock is slightly decreased and
Ma
at the moment of taking the picture also a bifurcation is visible above the suction side.
These changes in the passage shock structure leads to the slight differences of the shock loss
ed
distribution.
pt
In order to get more information on both the shock movement as well as the laminar sepa-
ce
ration bubble, PIV measurements were performed in two regions within the cascade passage
Ac
(cf. Fig. 4). Region A is located within the cascade passage which allows the statistical
analysis of the shock movement. Region B is located on the blade suction side near the
shock boundary layer interaction. During the PIV measurements a significant movement of
the passage shock was observable. Therefore, the analysis of the PIV measurements were

d
te
e di
py
Figure 13: STATISTICAL ANALYSIS OF THE SHOCK MOVEMENT AT PIV REGION
Co
A
ot
conditionally averaged as described earlier in the paper.
tN
rip
Figure 13 shows the statistical analysis of shock movement for both cases. A comparison
sc
of these results reveals a similar range of shock movement of about 10 mm in both cases
nu
that remains uninfluenced by the incoming boundary layer condition. Only the mean shock
Ma
position is changed from 52 mm at the laminar case to 55 mm for the turbulent case.
The shock movement and its interation with the suction side boundary layer is shown re-
ed
spectively in Figs. 14 and 15 for the laminar and turbulent case. These figures are based
pt
on the conditional averages of the PIV data and show the front, middle and rear position
ce
of the moving shock. In the laminar case the separation bubble is observable and visualized
Ac
by a dividing streamline. At the most upstream shock position the length of the separation
bubble on the blade surface amounts to 6 mm. As the shock position moves further down-
stream, the separation length increases rapidly and at rear position the separation length
amounts to 18 mm. The increase of the separation length leads also to an increase of the

boundary layer thickness behind the separation bubble. Nevertheless, the starting point of
the separation remains nearly unchanged.
Although evident from exemplary PIV shots in Fig. 15 a flow separation is not visible in
d
te
the mean flow for the turbulent case. From this it can be deduced that a fully laminar or
di
turbulent case does not really exist in this unsteady flow, but rather the laminar or turbulent
e
py
boundary layer should only be labeled as a rather predominant flow condition.
In addition to the shock motion, the unsteady shock behavior was measured by means of
Co
high speed shadowgraphy. The resulting spectra of the laminar and turbulent case are shown
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac
Figure 14: CONDITIONAL AVERAGED PIV RESULTS AT ROI B, FRONT SHOCK PO-
SITION (TOP), MID SHOCK POSITION (MIDDLE), REAR SHOCK POSITION (BOT-
TOM)

d
te
e di
py
Co
ot
tN
Figure 15: CONDITIONAL AVERAGED PIV RESULTS AT ROI B, FRONT SHOCK PO-
rip
SITION (TOP), MID SHOCK POSITION (MIDDLE), REAR SHOCK POSITION (BOT-
sc
TOM)
nu
in Fig. 16. The spectra are generally characterized by broadband noise. In addition to that,
Ma
the spectrum of the laminar case shows two dominant frequencies. The first frequency is at
ed
1.14 kHz and the second at 2.28 kHz which corresponds to the first harmonic frequency. The
pt
source of the frequencies is found by a periodic separation on the throttle behind the test
ce
section during this test and is not directly caused by the cascade flow. A further comparison
Ac
of the spectra shows that the shock movement square amplitude of the turbulent case is
larger than in the laminar case at frequencies below 500 Hz and above 2.5 kHz. The higher
squared amplitude at lower frequencies is equivalent with a much higher amplitude of the
shock movement. From this behaviour it can be assumed that the forces acting the on

d
te
e di
py
Figure 16: SPECTRUM OF THE SHOCK WAVE OSCILLATION OF THE LAMINAR
Co
AND TURBULENT CASE
ot
the blade are also higher which lead to higher deflections of the blade. This interpretation
tN
is based on the results from the PIV measurement and the image evaluation as already
rip
described above. Because in the image evaluation process the occurring blade displacement
sc
is tracked. The results show that a larger shock movement width is linked with an higher
nu
blade deflection. In addition, the maximum blade deflection in the turbulent case is higher
Ma
than in the laminar case. Hence, it can be not generally stated that the turbulent shock
boundary layer interaction should be preferred in the operation of transonic blades. In fact, it
ed
has to be investigated how to reduce the shock movement in the laminar as well as turbulent
pt
case and how this effects the behaviour of the losses and shock boundary layer interaction.
ce
Ac

5 DISCUSSION
Concerning the experiments it can be stated that a wide basis for the description of the
unsteady flow behaviour including laminar and turbulent shock boundary layer interaction
d
in a transonic compressor cascade has been acquired through the experiments. Furthermore,
te
the cascade performance in terms of loss is also discussed in the study. Nevertheless, it
di
becomes clear that more detailed unsteady data about the shock boundary layer interaction
e
py
and the development of the separation bubble in the laminar case is needed. In this regard,
Co
high-speed PIV could be helpful to close this gap by providing temporally contiguous PIV
data.
ot
tN
The recapitulation of all the steady and unsteady results leads to the fundamental out-
come of the experimental investigation. The shown movement of the shock wave simultane-
rip
ously represents an oscillating inflow angle and this means a time variation of the cascade
sc
operating point around an averaged operating point. The importance of this fact becomes
nu
clear in the principle sketch shown in Fig. 17. From part a) to c) this figure shows in princi-
Ma
ple how the numerical stall margin of the cascade is reached by increasing the back pressure,
starting from the aerodynamic design point. Furthermore, from part d) to f) it is shown
ed
how the experimental stall margin of the cascade is reached by increasing the back pressure.
pt
Thereby the experimental stall margin means the averaged operating point, where flow sep-
ce
aration occurs. In part e) of Fig. 17 it is depicted that, due to the shock wave oscillation the
Ac
experimental stall margin is reached earlier than in the numerical simulation. There exist a
∆ in working range between the shock position where the numerical stall margin is reached
and the averaged shock position of experimental cascade stall margin. Thus, it is important

to accurately reproduce the shock oscillation during numerical simulations. Otherwise the
achievable working range will not be correctly predicted which is a critical fact. This means
d
te
e di
py
Co
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Figure 17: PRINCIPLE SKETCH OF THE NUMERICAL AND EXPERIMENTAL

Ac
SHOCK POSITION AT DIFFERENT OPERATING POINTS (a: NUM. AERODYNAMIC
DESIGN POINT, b: NUM. STALL ONSET, c: DEFINITION OF NUM. STALL MARGIN,
d: EXP. AERODYNAMIC DESIGN POINT, e: EXP. STALL ONSET, f: COMPARISON
OF THE NUM. AND EXP. STALL MARGIN)

that an accurate numerical prediction of the surge limit is the current challenge within the
design process of transonic compressor blades.
d
te
e di
py
Co
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac

6 NUMERICAL STUDY
Finally, a numerical study is carried out in order to answer the question how well state-of-
the-art numerical methods can predict the unsteady shock movement.
d
te
di
Numerical Approach and Setup
e
py
The solver applied for this study is TRACE, the DLR in-house solver for turbomachinery
Co
flows developed at DLR’s Institute of Propulsion Technology in Cologne. TRACE is a par-
allel flow solver for the compressible Navier-Stokes equation on structured and unstructured
ot
tN
meshes. For a detailed overview of the capabilities of the solver see [40]. The turbulence
is modeled using Menter’s two equation SST k-ω model as documented in [41]. This tur-
rip
bulence model is combined with Menter-Langtry’s γ-ReΘ transition model in its version of
sc
2009 [42]. Inviscid fluxes are evaluated using Roe’s flux-differencing-splitting method. The
nu
upwind states are computed using the second- or third-order MUSCL (Monotonic Upstream-
Ma
Centered Scheme for Conservation Laws) family of schemes in combination with a modified
van Albada limiter to avoid the Gibbs phenomenon near shocks. Viscous terms are dis-
ed
cretized using second-order accurate central differences. To efficiently perform unsteady

pt
ce
simulations in the time-domain the accurate and robust multistage implicit RungeKutta
Ac
(IRK) method of third-order accuracy is used. The cascade was simulated as a stream tube
with 3% chord thickness. The mesh of the cascade was generated with G3DHexa and fea-
tures a structured OCH multi block topology consisting of hexahedral cells. To assess the
influence of insufficient resolution of flow features with a significant influence two grids were

designed. Boundary layers on solid surfaces are resolved with a maximal y+ of 3 for both
configurations to meet the requirements of the transition model. But the coarse version con-
sists of approx. 350 000 cells which introduces excessive damping due to the poorly resolved
flow (as shown later). This represents a grid resolution that is typically used in blade design.
d
te
The fine version contains around 735 000 and is mainly refined the in region of the shock
di
and wake of the blade.
e
py
Co
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac

Numerical Results
The numerical study was starting with steady simulations on both grid configurations. While
the simulation on the coarse grid converged to a steady state solution quite quickly, this was
d
not the case for the finer grid. Here the flow keeps oscillating due to reduced numerical dissi-
te
pation. This could give a turbomachinery designer a good indication of the high unsteadiness
di
of the blade configuration. Starting with these solutions for the stationary problem the un-
e
py
steady simulations were initialized. Using a time step size of ∆t = 1.04 · 10−5 the simulations
Co
were continued until they converged to a periodic shock oscillation. An instantaneous Mach
contour in the blade passage is shown in Fig. 18. From this result the numerical Schlieren
ot
pattern is extracted. The comparison of the experimental and numerical Schlieren pattern
tN
is shown in Fig. 19. In this figure it can be observed that the shock wave structure in the
rip
cascade is well predicted by the numerical simulation.

sc
Further analysis of the numerical results are focused on the unsteady behaviour. During two
nu
Ma
ed
pt
ce
Ac
Figure 18: MACH NUMBER DISTRIBUTION OF ONE INSTANTANOUS TIME WITH
THE FINE MESH; BLACK LINE INDICATES THE POSITION OF RECORDED DATA

d
te
e di
py
Co
ot
tN
rip
sc
Figure 19: EXPERIMENTAL (TOP) AND NUMERICAL (BOTTOM) SCHLIEREN PAT-

nu
TERN OF THE LAMINAR CASE

Ma
additional periods, time resolved data along a line in the passage are recorded as indicated in
Fig. 18. As already observed for the steady simulation the excessive dissipation of the coarse
ed
grid leads to a quite steady state flow in contrast to the described experimental results. The
pt
shock position in Fig. 20 contains only insignificant unsteady shock movement. Here the
ce
shock position is quite unsteady as seen in Fig. 21. The location changes within 26 mm in
Ac
its extrema with a frequency of f ≈ 2965 Hz. The comparison with the experiments show
that shock movement range is over-predicted and also the frequency is not detected in ex-
periments. The reason for this shock behaviour in the numerical simulation could be the not

fully resolved turbulent spectrum.
d
te
e di
py
Co
ot
tN
Figure 20: TEMPORAL EVOLUTION OF THE MACH NUMBER; COARSE GRID CON-
rip
FIGURATION
sc
nu
Ma
ed
pt
ce
Ac
Figure 21: TEMPORAL EVOLUTION OF THE MACH NUMBER; FINE GRID CONFIG-
URATION

7 RECOMMENDATIONS AND OUTLOOK
Based on the shown results it can be stated that there is a need for high prediction quality
of the unsteadiness of the transonic compressor flow. Instead of using finer grids in the
d
numerical turbomachinery design as utilized within this study, the use of high-order spatial
te
schemes is also recommended and most likely more promising in terms of efficiency. Recently
di
new numerical methods were developed specifically designed to meet the requirements of a
e
py
shock/boundary layer interaction simulation, e.g. [43] and [44]. A review of methods for the
Co
numerical simulation of turbulent compressible flow in the presence of shock waves can be
found in [45]. Since the boundary layer contains a large spectrum of turbulent structures,
ot
tN
there is also the need to resolve these frequencies in the simulations to cover more unsteady
effects important for shock/boundary layer predictions. This can be accomplished by the
rip
use of scale resolving methods such as LES. The use of scale resolving methods with simpler
sc
flow configuration as performed by Bernardini et al. [46] or Lele and Larsson [47] or even
nu
DNS together with fundamental experimental data can lead to a better understand of the
Ma
shock/boundary layer interaction mechanisms.
Furthermore, the control of the unsteady behaviour of the shock boundary layer interaction
ed
is an important field of research because it is enable the possibility to combine loss reduction
pt
with a decrease of the flow unsteadiness. At first, the development of new design concepts
ce
of for transonic blades is be an promising approach [48]. In addition to that, the application
Ac
of passive and active flow control concepts could be investigated.

8 CONCLUSION
A very extensive experimental study on the flow phenomena in a transonic compressor cas-
cade was performed. The measurements were focused on the laminar and turbulent shock
d
boundary layer interaction and the unsteady flow behaviour which is mainly caused by this
te
interaction. The results of the study can be concluded as follows:
di
Concerning the cascade performance it was shown that the cascade losses of the turbulent
e
py
case decrease compared to the laminar case by reduction of the viscous losses on the blade
Co
suction side. This was directly traced back to the disappearance of the laminar sepparation
bubble.
ot
tN
Furthermore, the shock boundary layer interaction was detailed investigated by means
of PIV. Due to this interaction the cascade flow was very unsteady and therefor a signif-
rip
icant shock oscillation occurs in both cases. Hence, a new advanced PIV post processing
sc
method, the conditional averaging was used in order to average the PIV images depending
nu
on the shock position. Thus, the development of the laminar separation bubble underneath
Ma
the passage shock was identified. It was shown that the, chord length position of the point
of separation onset is nearly constant but if the shock moves downstream the size of the
ed
separation bubble increases rapidly.

pt
In addition to that, the power spectral density of the shock movement of the turbulent
ce
case shows that at lower frequencies the square movement amlpitude is higher which could
Ac
lead to an unwanted increase of blade deflection. Hence, it could not be stated in general that
the turbulent case is the favourable one, in spite of the loss reduction which were achieved.
The fundamental outcome from the experimental results of the investigation is that an

accurate numerical prediction of the surge limit is the present challange within the design
process of transonic compressor blades. Because, due to the shock wave oscillation the stall
onset at the surge limit shifted compared to a steady shock wave behaviour. This fact has
to be considered within the design process in order to achieve an accurate prediction of the
d
te
surge limit.
di
Finally, it was depicted by the numerical study that the introduced damping of the nu-
e
py
merical methods or unsufficient grid resolution supresses the unsteady effects with are crucial
to predict the efficiency losses. However, the shock wave oscillation phenomenon is clearly
Co
shown for this cascade by numerical approches even though the used numerical methods in
ot
this study are not fully capable to predict all shock wave details and turbulence motion.
tN
rip
sc
nu
Ma
ed
pt
ce
Ac

ACKNOWLEDGMENTS
This research was supported by the seventh EU framework and was performed within the
research project TFAST (Transition Location Effect on Shock Wave Boundary Layer Inter-
d
action, Grant Agreement number 265455).
te
The authors would like to thank the colleagues from Rolls Royce Deutschland and the Sze-
di
walski Institute of Fluid-Flow Machinery Polish Academy of Sciences (IMP PAN) for the
e
py
very good cooperation and discussions within the project.
Co
ot
tN
9 Nomenclature
rip
Latin
sc
c profile chord length

nu
h blade span, height
i incidence angle = β1 -β1,OP 0

Ma
M Mach number
ed
n,N number
pt
p pressure
ce
Pr Prandtl number
Ac
r radius
Re Reynolds number based on chord length
t pitch
T temperature

Tu turbulence intensity
v mean-flow velocity
x, y, z cartesian coordinates
Greek
d
te
β flow angle with respect to cascade front
di
cascade deflection angle = β1 -β2
e
py
κ isentropic coefficient
pt,1 −pt,2
ω total pressure loss coefficient =
Co
pt,1 −p1
Abbreviations
ot
ADP aerodynamic design point
tN
ρ2 ·v2 sinβ2
AVDR axial velocity density ratio = ρ1 ·v1 sinβ1
rip
BL boundary layer
sc
CFD computational fluid dynamics

nu
DNS direct numerical simulation

Ma
LE leading edge
LES large eddy simulation

ed
MA PIV measurement area

pt
MP L2F measurement plane laser 2 focus

ce
MP1 measurement plane 1 (inlet)

Ac
MP2 measurement plane 2 (exit)
PIV particle image velocimetry
PS pressure side
PSD Power Spectral Density

SBLI shock boundary layer interaction
SS suction side
RANS Reynolds-averaged Navier-Stokes
ROI region of interest
d
te
TE trailing edge
di
Subscripts
e
py
0 reference state
1 inlet plane
Co
2 exit plane
ot
ax axial
tN
is isentropic
rip
l laminar
sc
s stagger
nu
t total, stagnation value, turbulent

Ma
W wall
ed
References
pt
ce
[1] Ferri, A., 1938, “Untersuchungen und Versuche im Überschallwindkanal zu Guidonia,”

Ac
Jahrbuch 1938 der deutschen Luftfahrtforschung, pp. 112 – 138.
[2] Fage, A. and Sargent, R. F., 1947, “Shock Wave and Boundary Layer Phenomena Near
a Flat Surface,” Proceedings of the Royal Society of London, Series A: Mathematical
and Physical Sciences, 190, pp. 1–20.

[3] Green, J. E., 1970, “Interactions Between Shock Waves and Turbulent Boundary Lay-
ers,” Progress in Aerospace Sciences, Pergamon, Oxford,, 11, pp. 235–340.
[4] Liepmann, H. W., 1946, “The Interaction Between Boundary Layer and Shock Waves
d
in Transonic Flow,” Journal of Aerospace Sciences, 13(12), pp. 623–638.
te
di
[5] Mundell, A. R. G. and Madey, D. G., 1986, “Pressure fluctuations caused by transonic
e
shock/boundary-layer interaction,” Aeronautical Journal, pp. 274–81.
py
[6] Lee, B. H. K., 1989, “Investigation of Flow Separation on a Supercritical Airfoil,” Jour-
Co
nal of Aircraft, AIAA, 26(11), pp. 1032–1037.
ot
[7] Epstein, A. H., Kerrebock, J. L., and Thomkins, W. T., 1979, “Shock Structure in
tN
Transsonic Compressor Rotors,” AIA Journal, 17, pp. 375–379.
rip
[8] Hilton, W. F. and Fowler, R. G., 1947, “Photographs of Shock Wave Movement,” Re-
sc
ports and Memoranda 2692, Aeronautical Research Council , Great Britain.

nu
Ma
[9] Dolling, D. S., 2001, “Fifty Years of Shock-Wave/Boundary-Layer Interaction Research:
What Next?” AIAA Journal, 39(8), pp. 1517–1531.

ed
[10] Beresh, S. J., Clemens, N. T., Dolling, D. S., and Comninos, M., 1997, “Investigation
pt
of the causes of large-scale unsteadiness of shock-induced separated flow using planar

ce
laser imaging,” No. AIAA Paper 97-0064 in 35th AIAA Aerospace Sceinces Meeting
Ac
and Exhibit, AIAA, Reno, NV, USA.
[11] Beresh, S. J., Clemens, N. T., and Dolling, D. S., 1999, “The Relationship between
Upstream Turbulent Boundary Layer Velocity Fluctuations and Separation Shock Un-

steadiness,” No. AIAA Paper 99-0295 in 37th AIAA Aerospace Sceinces Meeting and
Exhibit, AIAA, Reno, NV, USA.
[12] Roos, F. W., 1975, “Surface Pressure and Wake Flow Fluctuations in a Supercritical
d
Airfoil Flowfield,” No. AIAA Paper 75-66 in 13th Aerospace Sciences Meeting, AIAA,
te
Pasadena, CA, USA.
e di
[13] Thomas, F. O., Putnam, C. M., and Chu, H. C., 1995, “On the Mechanism of Unsteady
py
ShockWave/Turbulent Boundary layer Interactions,” Experiments in Fluids, 18(1/2),
Co
pp. 69–81.
ot
[14] Dussauge, J.-P., Dupont, P., and Debieve, J.-F., 2006, “Unsteadiness in shock wave
tN
boundary layer ininteractionsw separation,” Journal of Aerospace Science and Technol-
rip
ogy, 10, pp. 85–91.

sc
[15] Dussauge, J.-P. and Piponniau, S., 2008, “Shock/boundary-layer interactions: Possible
nu
sources of unsteadiness,” Journal of Fluids and Structures, 24, pp. 1166–1175.

Ma
[16] Hartmann, A., Klaas, M., and Schröder, W., 2012, “Time-resolved stereo PIV measure-
ed
ments of shock-boundary layer interaction on a supercritical airfoil,” Experiments in

pt
Fluids, 52(3), pp. 591–604.

ce
[17] Hergt, A., J.Klinner, Steinert, W., Grund, S., Beversdorff, M., Giebmanns, A., and
Ac
Schnell, R., 2015, “The Effect of an Eroded Leading Edge on the Aerodynamic Perfor-
mance of a Transonic Fan Blade Cascade,” ASME Journal of Turbomachinery, 137, pp.
021006–1 – 11.

[18] Strazisar, A. J., 1984, “Investigation of Flow Phenomena in a Transonic Fan Rotor
Using Laser Anemometry,” No. 84-GT-199 in ASME Turbo Expo, ASME, Amsterdam,
Netherlands.
d
[19] Schreiber, H. A., Starken, H., and Steinert, W., 1993, “Transonic and Supersonic Cas-
te
cades,” AGARDOgraph - Advanced Methods for Cascade Testing, AGARD AG 328,
di
pp. 35–59.
e
py
[20] Steinert, W., Fuchs, R., and Starken, H., 1992, “Inlet Flow Angle Determination of
Co
Transonic Compressor Cascade,” ASME Journal of Turbomachinery, 114(3), pp. 487–
ot
493. tN
[21] Kiock, R., Laskowski, G., and Hoheisel, H., 1982, “Die Erzeugung höherer Turbulenz-
rip
grade in der Meßstrecke des Hochgeschwindigkeits-Gitterwindkanals, Braunschweig, zur

sc
Simulation turbomaschinenähnlicher Bedingungen,” DLR Forschungsbericht, (FB82-

nu
25).
Ma
[22] Schimming, P., 1976, “Experimental Investigation of Supersonic Inflow of Compres-
sor Cascade by the Laser-2-Focus Method,” Symposium of Measuring Techniques in

ed
Transonic and Supersonic Cascade Flow, Ecole Polytechnique Federale de Lausanne.

pt
ce
[23] Schodl, R., 1980, “A Laser-Two-Focus (L2F) Velocimeter for Automatic Flow Vevtor
Ac
Measurements in the Rotating Components of Turbomachines,” ASME Journal of Flu-
ids Engineering, 102(4), pp. 412–419.
[24] Schodl, R., 1989, “Laser Two Focus Techniques,” VKI-Lecture Series 1989-05 Measure-
ment Techniques in Aerodynamics.

[25] Klinner, J., Hergt, A., Beversdorff, M., and Willert, C., 2012, “Visualization and PIV
Measurements of the Transonic Flow around the Leading Edge of an eroded Fan Air-
foil,” 16th Int Symp on Applications of Laser Techniques to Fluid Mechanics, Lisbon,
Portugal.
d
te
[26] Klinner, J., Hergt, A., and Willert, C., 2014, “Experimental Investigation of the tran-
di
sonic flow around the leading edge of an eroded fan airfoil,” Experiments in Fluids,
e
py
55(9), 1800.
Co
[27] Mee, D. J., Walton, T. W., Harrison, S. B., and Jones, T., 1991, “A Comparison of Liq-
ot
uid Crystal Techniques for Tranition Detection,” No. AIAA-91-0062 in 29th Aerospace
tN
Science Meeting, AIAA, Reno, Nevada, USA.
rip
[28] Steinert, W. and Starken, H., 1996, “Off-Design Transition and Separation Behavior of
sc
a CDA Cascade,” ASME Journal of Turbomachinery, 118(2), pp. 204–210.

nu
[29] Schreiber, H. A., 1976, “Comparison between Flows in Cascades and Rotors in the
Ma
Transonic Range,” Lecture Series in Transonic Blade-to-Blade Flows in Axial Turbo-
machinery, von Karman Institute for Fluid Dynamics.

ed
pt
[30] Schreiber, H. A. and Starken, H., 1981, “On the Definition of the Axial Velocity Density
ce
Ratio in Theoretical and Experimental Cascade Investigation,” Symposium on Measur-

Ac
ing Techniques in Transonic and Supersonic Flow in Cascades and Turbomachines,
Lyon, France.

[31] Raffel, M., Willert, C., Scarano, F., Kähler, C., Wereley, S., and Kompenhans, J., 2017,
Particle Image Velocimetry, Experimental Fluid Mechanics, Springer International Pub-
lishing AG.
d
[32] Ragni, D., Schrijer, F., van Oudheusden, B., and Scarano, F., 2011, “Particle tracer
te
response across shocks measured by PIV,” Exp Fluids, 50(1), pp. 53–64.
e di
[33] Fitzgerald, E. J. and Mueller, T. J., 1990, “Measurements in a separation bub-
py
ble on an airfoil using laser velocimetry,” AIAA Journal, 28(4), pp. 584–592, URL
Co
http://dx.doi.org/10.2514/3.10433.
ot
[34] Willert, C., Mitchell, D., and Soria, J., 2012, “An assessment of high-power light-
tN
emitting diodes for high frame rate schlieren imaging,” Exp Fluids, 53, pp. 413–421.
rip
[35] Welch, P., 1967, “The use of fast Fourier transform for the estimation of power spectra:
sc
A method based on time averaging over short, modified periodograms,” IEEE Transac-
nu
tions on Audio and Electroacoustics, 15(2), pp. 70–73.

Ma
[36] Schreiber, H. A., Steinert, W., and Kuesters, B., 2002, “Effects of Reynolds Numbers
ed
and Free-Stream Turbulence on Boundary Layer Transition in a Compressor Cascade,”

pt
ASME Journal of Turbomachinery, 124(1), pp. 1–9.

ce
[37] Schreiber, H. A., 1986, “Experimental Investigation on Shock Losses of Transonic and
Ac
Supersonic Compressor Cascades,” No. AGARD-CP-401 in Conference Proceedings for
Transonic and Supersonic Phenomena in Turbomachines, AGARD, Munich, Germany,
pp. 11–1 –11–15.

[38] Merzkirch, W., 1971, “Der Ablösestoss bei der Expansion einer Überschall Gren-
zschicht,” Zeitschrift für Flugwissenschaften, 19. Jahrgang, Heft 1, pp. 1–12.
[39] Babinsky, H. and Harvey, J. K., eds., 2011, Shock Wave-Boundary-Layer Interactions,
d
Cambridge University Press.
te
di
[40] Becker, K., Heitkamp, K., and Kügeler, E., 2010, “Recent Progress in a Hybrid-Grid
e
CFD Solver for Turbomachinery Flows,” No. CFD-2010-01609 in Fifth European Con-
py
ference on Computational Fluid Dynamics, ECOMAS CFD 2010, ECCOMAS, Lisbon,
Co
Portugal.
ot
[41] Menter, F., Kuntz, M., and Langtry, R., 2003, “Ten years of industrial experience with
tN
the SST model.” Turbulence, Heat and Mass Transfer 4, K. Hanjalić, Y. Nagano, and
rip
M. Tummers, eds.
sc
[42] Langtry, R. and Menter, F., 2009, “Correlation-based transition modeling for unstruc-
nu
tured parallelized computational fluid dynamics codes,” AIAA J., 47(12), pp. 2894–
Ma
2906.
ed
[43] Lu, Y. and Dawes, W. N., 2015, “High order large eddy simulations for a transonic
pt
turbine blade using hybrid unstructured meshes,” Proceedings of ASME Turbo Expo
ce
2015: Turbine Technical Conference and Exposition, 56659, pp. V02CT44A005–, URL
Ac
http://dx.doi.org/10.1115/GT2015-42283.
[44] Gou, J., Su, X., and Yuan, X., 2016, “Adaptive mesh refinement for DDES simula-
tion on transonic compressor cascade with unstructured mesh,” Proceedings of ASME

Turbo Expo 2016: Turbomachinery Technical Conference and Exposition, 49712, pp.
V02CT39A034–, URL http://dx.doi.org/10.1115/GT2016-56925.
[45] Pirozzoli, S., 2011, “Numerical methods for high-speed flows,”
d
Annual Review of Fluid Mechanics, 43(1), pp. 163–194, URL
te
https://doi.org/10.1146/annurev-fluid-122109-160718.
e di
[46] Bernardini, M., Pirozzoli, S., and Grasso, F., 2010, “Analysis of unsteadiness in tran-
py
sonic shock/boundary interactions,” Proceedings Fifth European Conference on Com-
Co
putational Fluid Dynamics ECCOMAS CFD 2010, J. C. F. Pereira, A. Sequeira, and
ot
J. M. C. Pereira, eds., Lisbon, Portugal. tN
[47] Lele, S. K. and Larsson, J., 2009, “Shock-turbulence interaction: What we know and
rip
what we can learn from peta-scale simulations,” Journal of Physics: Conference Series,
sc
180(1), p. 012032, URL http://stacks.iop.org/1742-6596/180/i=1/a=012032.

nu
[48] John, A., Shahpar, S., and Qin, N., 2016, “Allevation of shock-wave effects on a highly
Ma
loaded axial compressor though novel blade shaping,” No. GT2016-57550 in ASME
Turbo Expo, ASME, Seoul, South Korea.

ed
pt
ce
Ac

2019 - J. Klinner - Transonig Blade Compressor CFD

Uploaded by

Copyright:

Available Formats

2019 - J. Klinner - Transonig Blade Compressor CFD

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2019 - J. Klinner - Transonig Blade Compressor CFD

Uploaded by

Copyright:

Available Formats

Journal of Turbomachinery.

Received September 11, 2018;

THE PRESENT CHALLENGE OF TRANSONIC

in detail. Therefore, an extensive experimental investigation of the flow in a transonic

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

as LES or the application of DNS is necessary to correctly predict unsteadiness of the

blade boundary layer.

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

motion correlates with a fuller boundary layer.

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

significant impact on the working characteristics (performance, working range) of a transonic

phenomena is described in detail by means of an extensive experimental investigation of the

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

2.1 Cascade Configuration

Figure 1: TRANSONIC CASCADE WITH PLANAR ENDWALL

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

boundary layer (turbulent SBLI).

upstream of the test section (based on the work of Hoheisel [21]).

Inlet Mach number M1 = 1.21

Reynolds number Re ≈ 1.4×106

Inlet flow angle β1 = 145.6 deg

Flow turning at ADP ∆β = 14.9 deg

Stagger angle βst = 134.4 deg

Blade chord length c = 100 mm

Blade span h = 168 mm

Pitch to chord ratio t/c = 0.6

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Figure 2: CROSS-SECTION DRAWING OF THE TEST SECTION

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

the isentropic Mach number distribution.

well as the losses are shown in the table 2.

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Table 2: ESTIMATED UNCERTAINTIES OF THE TEST DATA AT MID-SPAN

Upstream flow angle β1 +/- 0.2 deg

Exit flow angle β2 +/- 0.2 deg

SHOCK POSITION; REGION B: MEASUREMENTS OF SBLI AND SEPARATION RE-

3.1 PIV Instrumentation and Image Evaluation

regarding the PIV measurement setup are described in [26].

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

which is evidenced by vertical image displacements of several pixels. Using a correlation

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

OBLIQUE SHOCK AND TRANSITIONAL SEPARATION BUBBLE INCLUDING RE-

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

induced flow separation can be estimated.

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

3.2 Schlieren and High-Speed Shadowgraphy Setup and Process-

fixed width and overlap [35].

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

3.3 Liquid Crystal

Schreiber et al. as well as Hergt et al. [36, 17].

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

4 CASCADE STUDY RESULTS

Figure 7: INFLOW ANGLE DISTRIBUTION OF THE LAMINAR AND TURBULENT

CASE (L2F-MEASUREMENT PLANE AT MIDSPAN)

Figure 8: ISENTROPIC MACH NUMBER DISTRIBUTION AT MIDSPAN OF THE LAM-

INAR AND TURBULENT CASE

Downloaded From: https://turbomachinery.asmedigitalcollection.asme.org on 04/01/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use