Sunteți pe pagina 1din 31

Progress in Aerospace Sciences 47 (2011) 249–279

Contents lists available at ScienceDirect

Progress in Aerospace Sciences


journal homepage: www.elsevier.com/locate/paerosci

Research on the aerodynamics of intermediate turbine diffusers


Emil Göttlich n
Institute for Thermal Turbomachinery and Machine Dynamics, Graz University of Technology, Inffeldgasse 25 A, 8010 Graz, Austria

a r t i c l e i n f o abstract

Available online 14 April 2011 Intermediate turbine diffusers represent the flow path between the high pressure and the low pressure
turbine of a high-bypass ratio turbofan aero engine. Caused by the different rotational speeds of high
and low pressure spool, these components have to diffuse and guide the flow safely to a larger diameter
without disturbances or boundary layer separations. The large radial offset between in- and outlet of
intermediate turbine diffusers leads to a pronounced S-shape. The trend for further increased bypass
ratios will require more attention to this component since its shape influences the overall weight of
engine and nacelle considerably. The complicated aerodynamics of these annular ducts has to be
understood to realize short S-shaped diffuser designs. This article tries to review the flow evolution
through intermediate turbine diffusers and discusses the influence of the different effects in a
systematic way. Investigations by various researchers are presented and test turbine rigs for
experiments under engine realistic duct inlet conditions are described. Special focus is laid on different
measures for the designer to produce more aggressive diffuser layouts whilst keeping the losses low.
The application of flow control, shape optimization and endwall contouring are promising actions to
shorten the diffuser length and, furthermore, to gain an engine weight reduction. The paper ends with a
discussion of new design concepts for turbine ducts as well as for future engine architectures. It can be
concluded that intermediate turbine diffusers will become a key component for keeping the overall
engine weight and fuel burn low.
& 2011 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
2. Fundamental aerodynamics of intermediate turbine diffusers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
2.1. Influence of S-shaped meridional flow path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
2.2. Influence of wakes, swirl and diffusion rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
2.2.1. Influence of wakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
2.2.2. Influence of swirl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
2.2.3. Influence of diffusion rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
2.3. Influence of struts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
3. Aerodynamics of intermediate turbine diffusers arranged downstream of turbine stages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
3.1. Test programs with ducts downstream of turbine stages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
3.1.1. Oxford Rotor Facility at the Thermo-Fluids Laboratory of the University of Oxford
(Southwell Laboratory, Osney Laboratory) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
3.1.2. Turbine Test Facility (TTF) at the Gas Turbine Laboratory of the Ohio State University . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
3.1.3. Large-Scale Facility at the Division of Fluid Dynamic of Chalmers University of Technology. . . . . . . . . . . . . . . . . . . . . . . . . . 257
3.1.4. Transonic Test Turbine Facility (TTTF) at the Institute for Thermal Turbomachinery and Machine Dynamics of
Graz University of Technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
3.2. Downstream effect of a turbine stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
3.3. Influence of Mach number and swirl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
3.4. Influence of rotor tip gap size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
3.5. Separated duct flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
4. Potential measures for increasing the diffuser design space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
4.1. Shape optimization and endwall contouring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267

n
Tel.: þ43 316 873 7231; fax: þ 43 316 873 7234.
E-mail address: emil.goettlich@tugraz.at

0376-0421/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.paerosci.2011.01.002
250 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

4.2. Flow-controlled intermediate turbine diffusers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269


4.3. Integrated concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
5. Conclusions and future perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

1. Introduction which the load-carrying struts and support lines are protected
from the hot gas path by aerodynamically optimized large non-
In modern commercial aero engines small specific fuel consump- lifting fairings in the shape of airfoils [5]. The flow channel walls
tion (SFC) together with low life cycle cost has to be realized to allow and fairings are designed as segments made of heat resistant
economical air transportation. Furthermore, an environment-friendly material and usually protected by cooling air on the outside
aero engine should have a reduced CO2 emission by less fuel burn and against the high heat loads of the turbine flow.
generate low noise levels during take-off and landing. The use of This paper reviews the progress in the field of S-shaped inter-
high-bypass ratio turbofan engines is essential to cope with these mediate turbine diffusers by discussing the past and recent research
requirements. The trend to increase the bypass ratio, which means to work performed on this topic. Details of aerodynamics and losses
decrease the specific thrust of an engine is still ongoing. While the are described together with their difficulties to predict them by
CF6 and the CFM56 engine families use a bypass ratio between 5 and means of CFD, but adequate discussions on aircraft engine system-
6, more recently developed designs apply values, for example, of level parameters are also presented. These include weight, length,
about 9 for General Electric’s GE90 [1], 8.7 for the GP7000 developed mechanical constraints, airfoil counts, etc. Moreover, the future
by the Engine Alliance [2] or 10–11 for the turbofans powering topics in this field are highlighted, which take account of finding
Boeing’s Dreamliner, e.g. the Rolls Royce Trent 1000 [3]. The demand measures to design shorter components together with larger radial
of further increased bypass ratios of aero engines will lead to larger, offsets but without a shortfall in performance.
but slower rotating fans due to the tip speed limit caused by The need for more aggressive turbine transition ducts as an
structural and noise considerations. enabling factor for high-bypass turbofan engines has already been
Fig. 1 gives an estimation of the differences between two-spool defined as one supporting technology program within the Energy
turbofans with a bypass ratio of 6 and 10, under the assumption Efficient Engine Component Development and Integration Program
of a common engine core [4]. As one can see the shape of the (E3 program) contracted between NASA and Pratt and Whitney in
meridional flow path of both designs differs considerably: 1978 [6]. On the basis of this cooperation the performance objec-
The flow has to be guided from the low pressure (LP) system to tives of a transition duct with an area ratio of 1.5, a length to height
the high pressure (HP) system and vice versa by annular ducts ratio of 3.0 and a non-working strut fairing were defined.
with a respectable radial offset. For a higher bypass ratio these A lot of research work has been carried out on diffuser flows
radius changes are much more pronounced due to the LP blading but only fewer studies of the flow within annular S-shaped ducts
running on larger mean diameters, which are necessary in order are available in open literature. Many of them are dealing with the
to reduce the blade loading coefficient by an increased circumfer- flow in intermediate compressor ducts (see e.g. [7]). Therefore the
ential speed and thus to reduce losses. In this study [4] it is shown fundamental work of Sovran and Klomp in 1967 [8] providing a
that the higher bypass ratio leads to more LP stages but the disk performance chart for straight-walled annular diffusers is still
bores can be much larger due to lower stresses caused by the used as a reference to classify intermediate turbine diffusers
lower rotational speed. Therefore the weight penalty of the 9 stage regarding their criticality. Fig. 2 presents a diagram comparing
LP turbine is not as large as one might expect [4]. designs of intermediate turbine diffusers of selected aero engines
Comparing the shape of the turbine transition ducts between HP and rig test setups recently investigated at Chalmers University of
and LP turbine it becomes clear that the slope angle has to become Technology and Graz University of Technology to the optimum
steeper for the larger bypass ratio engine to avoid an excessive lines for maximum pressure recovery coefficient at prescribed
increase in overall engine weight. The ducts between LP and HP non-dimensional duct length (CP*) or duct area ratio (CP**) taken
system have to be designed as short as possible. The requirements of from [8]. A diffuser design described by the non-dimensional duct
short annular flow channels will lead to a pronounced S-shape. length (L/hin) and the area ratio (AR¼ Aout/Ain), which is between
Therefore these components are also known as swan-necked inter- these lines can be seen as optimal, while a layout below and above
mediate ducts. While the duct connecting the LP compressor stages is classified as conservative and aggressive, respectively. However,
with the HP compressor inlet is designed with no or only marginal this can only be a rough assessment as the effects of curvature,
diffusion, the channel between the HP and LP turbine may show a flow swirl and inlet wakes, which occur in a real S-shaped inter-
large increase in area and hence a considerable amount of diffusion. mediate duct are not considered in the investigations of Sovran
The strong curvature and the adverse pressure gradient will pro- and Klomp [8]. The area ratios and non-dimensional duct lengths
mote boundary layer separation. Nevertheless, the flow leaving the for this comparison were estimated from engine cut-off drawings
HP turbine has to be guided to the following LP turbine without any available in open literature or on the World Wide Web according
flow disturbances and provide the proper flow field for high effi- to the dashed lines in the sketch of Fig. 2. It can be seen that
ciency there. The same considerations are also true for the transition the engine manufactures placed their designs within or close to
ducts between the intermediate pressure (IP) and the LP turbine the optimal region but rather tight to the conservative border. The
of three-spool aero engines such as the Trent family of Rolls Royce, test rig setups C1 to C5 will be described later in the paper.
but there the situation can be seen as less critical due to the smaller Within the 6th framework program funded by the European
radial offset and the reduced wall curvature. Commission two projects dealing exclusively or partly with the
Depending on the overall aero engine design the intermediate topic of intermediate ducts were launched, namely AIDA—Ag-
turbine ducts can either be clean annular channels like for the gressive Intermediate Duct Aerodynamics for Competitive and
PW6000 aero engine (Pratt and Whitney) or be equipped with Environmentally Friendly Jet Engines (contract no. AST3-CT-2003-
load-carrying struts for bearing support and oil and cooling air 502836) [9] and AITEB-II—Aerothermal Investigations on Turbine
service lines leading radially inward through them such as in the Endwalls and Blades II (contract no. AST4-CT-2005-516113) [10].
turbine center frame of the GP7000 aero engine (Engine Alliance). These strategic research projects should help to eliminate the lack
The turbine center frame (TCF) is a star-like static structure, for of information on intermediate turbine diffuser aerodynamics
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 251

Nomenclature AIDA C1, AIDA C3, AIDA C4, AIDA C5 AIDA test configurations
AITEB-II Aerothermal Investigation on Turbine Endwalls (EU
Symbols
project)
CFD computational fluid dynamics
AR Aout/Ain exit to entry area ratio of duct (dimensionless) C, C1, C3, C5, D, E and F measurement planes
C chord length (m) CW clockwise
CP, CP,ideal pressure rise coefficient, idealized pressure rise DOE design of experiments
coefficient (dimensionless) H.L.V. hub leakage vortex
h blade height (mm) HP high pressure
hVG vortex generator height (mm) IC integrated concept
h enthalpy (J/kg) IP intermediate pressure
L axial duct length (mm) LDV Laser-Doppler-Velocimeter
L/hin non-dimensional duct length (dimensionless) L.P.V. lower passage vortex
LVG vortex generator length (mm) LP low pressure
p, pt pressure, total pressure (N/m2) MTF mid turbine frame
T blade pitch (m) NGV nozzle guide vane
u circumferential velocity (m/s) PS pressure side
Dh/u2 loading coefficient (dimensionless) RMS root mean square
T/C pitch to chord ratio (dimensionless) SFC specific fuel consumption
Ma Mach number (dimensionless) SS suction surface
n strut count (dimensionless) TCF turbine center frame
r1h hub radius at inlet (m) TE trailing edge
r2/r1 exit to entry mean radius ratio (dimensionless) TEC turbine end casing
Re Reynolds number (dimensionless) TMTF turning mid turbine frame
x/L non-dimensional axial distance (dimensionless) TTF Turbine Test Facility (Ohio State University)
a swirl angle (deg) TTTF Transonic Test Turbine Facility (Graz University of
aVG angle of attack for vortex generator (deg) Technology)
DsVG distance of vortex generator trailing edge to separa- U.P.V. upper passage vortex
tion line (mm)
ZD diffuser efficiency (dimensionless)
z loss coefficient (dimensionless)

Abbreviations

AIDA Aggressive Intermediate Duct Aerodynamics (EU project)

through cooperative numerical and experimental studies per- If the bypass ratio is increased above 10 in the future, it can
formed by all European aero engine manufacturers together with be expected that the intermediate compressor and turbine duct
universities and key research institutions. designs will have a much larger influence onto the engine weight of

Fig. 1. Common core turbofan engines with bypass ratio of 6 and 10 (flow paths Fig. 2. Sovran and Klomp performance chart with aero engines and test rig setups
and disc shapes from [4]). from e.g. [28,43,48].
252 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

a conventional turbofan. But the weight penalty of more LP stages


running at larger radii can be kept moderate by applying shorter
and thus more aggressive duct designs, which are far beyond the
classical limit of Sovran and Klomp, as shown in Fig. 2. The other
option is to apply the geared turbofan concept, where a gearbox
makes the fan rotational speed independent of the booster and the
LP turbine. The number of LP stages can be reduced whilst main-
taining the low levels of aerodynamic loading together with no need
for shifting the flow annulus outwards [4]. Due to the fact that the
part count of the geared turbofan can be reduced, although these
parts will be technologically more challenging, the conventional
turbofan with further reduced flow path length is still an option. To
achieve dependable ultrahigh-bypass ratio engine layouts it will be
necessary to apply new shape optimized intermediate turbine
diffusers together with endwall contouring, active or passive flow
control and integrated concepts. Examples for all of these measures
will be discussed later in this work.

2. Fundamental aerodynamics of intermediate turbine


diffusers

2.1. Influence of S-shaped meridional flow path

The pressure distribution in an S-shaped diffuser depends not


only on the area change such as in straight-walled channels, but also
upon the curvature of the endwalls and hence of the streamlines.
The effective local curvature is also linked to the boundary layer
growth at the surfaces. If the flow is turned from the axis-parallel
direction radially outward in the first bend of an S-shaped duct, a
pressure gradient from the casing (low pressure) to the hub (high
pressure) will be created according to the radial equilibrium within
the flow field. This cross-passage pressure gradient is superimposed
to the adverse pressure gradient, evoked by the diffusion due to an
increase in flow area. An optimized design of the endwall curvature
can help to control these pressure gradients, but both of them play a
significant role in an aggressive intermediate turbine diffuser. At the
end of the duct the flow has to be turned back towards the engine
axis in order to match the channel shape with the inflow of the
following LP turbine stage in a real engine or with the rig exit
section of an individual test setup. This will cause a cross-passage
pressure gradient but now with the higher pressure at the casing
and the lower pressure at the hub. The pressure distribution in an
S-shaped diffuser has a major effect on the boundary layer behavior
at the endwalls, which is most critical due to the strongest adverse
Fig. 3. Mach number distribution in a typical S-shaped diffuser without (a) and
pressure gradients right after strong wall curvatures. with (b) swirl.
Fig. 3a shows an example for the flow through an S-shaped duct
without swirl and other inlet distortions. The results were gained by
a two-dimensional CFD simulation with FLUENT 6.2.16 using the
k e turbulence model. The boundary conditions were defined by a Fig. 4a shows the pressure recovery along the duct surfaces by
mass flow rate of 13.2 kg/s and a constant inlet total pressure of means of the pressure rise coefficient, which is defined as
1.05 bar set approximately one diffuser length upstream in a
CP ¼ ðpn pin Þ=ðpt,in pin Þ ð1Þ
parallel-walled cylindrical channel. It should be mentioned that
the duct shape was adopted from a real test setup discussed later in The subscript ‘‘n’’ relates to the local position and ‘‘in’’ refers to
this paper where an additional LP vane row is arranged at x/L¼0.62 the mass-averaged values at the duct inlet. The zero position
(leading edge, midspan). In a real engine the LP vanes are positioned corresponds to the end of the parallel-walled cylindrical flow
within the duct area of still increasing radius. Therefore in this channel (geometrical diffuser inlet), while x/L¼ 1 marks the begin-
example the hub surface shows two convex curvatures before and ning of the also parallel-walled cylindrical exit duct (geometrical
after the vane position. The second one is to turn the flow towards diffuser exit). It can be seen that the curves for hub and casing start
the test rig outlet. For the CFD simulation the cylindrical annular to diverge prior to the diffuser inlet due to the upstream influence of
channel was extended by further 2.5 diffuser lengths. Three regions the first bend. The pressure field at the exit of a HP turbine stage
can be defined as susceptible to flow separations: The first one is the arranged upstream of an intermediate turbine duct will be modified
low Mach number region at the casing right after the first bend in a similar way, which has to be considered during its design.
(marked with ‘‘A’’) and the other ones are at the hub directly after The curve for the casing flow shows the peak suction point
the convex surface zones (marked with ‘‘B’’ and ‘‘C’’). (CP ¼  0.78) at x/L¼0.18 in the area with the highest curvature of
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 253

2.2. Influence of wakes, swirl and diffusion rate

The inflow of an intermediate turbine diffuser downstream of


a turbine stage differs significantly from the ideal boundary
conditions of the above mentioned example. The inflow and
hence every other downstream position is heavily influenced by
periodical unsteady effects caused by the passing rotor blades
modulated in their strength by stator–rotor interaction. These
effects although not independent of each other can be defined as

 rotor wakes causing a periodic change in the flow angle in the


absolute frame of reference due to their velocity deficit,
 secondary flow vortices having a significant influence onto the
boundary layers,
 rotor tip leakage with high energy flows at the casing surface,
 swirl angles different from the axial direction and possibly a
strong variation in blade spanwise direction and
 trailing edge shocks in cases of turbine stages with transonic
flow regime in the relative frame of reference.

2.2.1. Influence of wakes


At the University of Durham a series of experimental investiga-
tions has been performed in order to decouple the complex flow
phenomena occurring downstream of a turbine stage in a systematic
manner. The measurements have been carried out in the University
of Durham Swan Neck Duct Facility, where a variable-speed fan
sucks air through a turbulence grid and an inlet contraction into a
parallel section, which houses a row of 34 fixed swirl vanes followed
by the annular diffusing duct to be tested and a plenum.
The first experiments deal with the pseudo-steady influence of
two-dimensional wakes extending from hub to the casing. For this
initial investigation of Dominy and Kirkham [11] flat plate vanes
were used to generate wakes comparable both in intensity and
extent to those of HP or IP blades but without swirl. The inter-
mediate turbine diffuser showed an area ratio of 1.5 and a hub-to-
casing radius ratio of 0.7. The Reynolds number based on the inlet
passage height was 3.9  105. Detailed information regarding test
setup, operating conditions and instrumentation are given in [11].
Fig. 4. Pressure rise coefficient along casing and hub surface without (a) and with The authors observed that the cross-passage pressure gradient
(b) swirl for the S-shaped diffuser of Fig. 3.
of the first radial bend drives the transportation of boundary layer
fluid from the hub surface to the casing within the low energy
wake flow. Fig. 5 shows these secondary flows evoked by wakes
the first bend, which is directly followed by a region of the strongest schematically. The radial velocity within the wake was found to
adverse pressure gradient due to the area increase together with the be the highest on the hub side giving rise to a pitch angle
influence of the first convex bend at the hub, which also leads to a variation within the passage of 131. Across most of the passage
pressure rise at the casing. Here it should be mentioned that the the wakes remain essentially two-dimensional and appear to
applied design compensates a too rapid increase in the flow area by influence the wall boundary layers only in the immediate vicinity
this convex shape of the hub surface marked with ‘‘C’’ in Fig. 3a.
Therefore no flow separation can be found at the critical zone at the
outer duct surface between minimum and maximum pressure,
although a rapid growth of the boundary layer and a low wall shear
stress will occur there. The maximum pressure at the casing
(CP ¼0.54) can be found in the second bend at x/L¼0.84 where the
flow is turned back to the axial direction.
On the hub surface a rapid increase in surface pressure (CP ¼0.52)
can be observed as the flow approaches the first bend x/L¼0.3
followed by two drops evoked by flow acceleration at the two
convex wall regions ‘‘B’’ and ‘‘C’’ (x/L¼0.49 and 0.96, respectively)
with a pressure recovery in-between. The first drop would not be so
pronounced in the real intermediate turbine diffuser due to the
blockage of the additional LP vane row at x/L¼0.62. At the geome-
trical diffuser exit x/L¼1 still a large cross-passage pressure gradient
is evident, which mixes out to CP ¼ 0.26 at x/L¼1.6.
It is obvious that a classification of the criticality of annular
S-shaped diffusers must be characterized by more parameters
than only the area ratio (AR) and the non-dimensional length. Fig. 5. Pressure-driven secondary flows evoked by S-shape and stationary wakes.
254 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

of the wake [11]. Further downstream the hub boundary layer prior to the inflow section at x/L¼0. The pressure at the casing peak
remains relatively uniform in circumferential direction while the suction point at x/L¼0.18 increases from CP ¼ 0.78 for zero swirl to
thickened casing boundary layer reveals strong evidence from the  0.27 due to the superimposed opposite pressure gradient. Addi-
transport of wake fluid into it. A re-energization of the boundary tionally, the effective wall curvature und thus the cross-passage
layer occurs there. Counter-rotating vortices start to form on pressure gradient is reduced in this region. The pressure rise over the
either side of the wake. As the fluid passes the second bend of the whole diffuser is significantly increased for the casing as well as for
duct the radial pressure gradient reverses, but the influence onto the hub. A mean exit CP of about 0.4 is reached compared to 0.26 for
the three-dimensional flow structures seems to be too small to the axial inflow. The maximum pressure at the hub in the first bend
change their character. At the duct outlet counter-rotating vor- is reduced from CP ¼0.52 to 0.16 and shifted from x/L¼0.3 to 0.36
tices on both sides of the wakes can be observed at the hub as due to the changed effective streamline curvature. The following two
well as at the casing, whereas the latter were found to be much flow accelerations at the corresponding convex wall regions ‘‘B’’ and
stronger. The secondary vortices induce local variations in yaw ‘‘C’’ are appearing earlier at x/L¼0.47 and 0.95, respectively, with a
and pitch angle of up to 741. The wake itself is observed to be less pronounced pressure drop and a stronger pressure recovery in-
very weak. The regions of highest loss were found at the casing between. In summary the cross-passage pressure gradient at the first
surface in the middle between two wakes where the boundary bend is halved for the swirling flow, which is in good agreement with
layer thickness shows its maximum. Another important conclu- the Mach number gradient shown in Fig. 3b.
sion of [11] is that a small flow deviation at the inlet may have a In general, it can be stated that a swirl creates, on the one
disproportionately large influence upon the flow at the entry of hand, less boundary layer loss due to the decreased diffusion and
the first turbine stage downstream of the duct due to the radially adverse pressure gradient along the streamlines caused by lower
driven wake flows promoted by the S-shape. Despite the growth effective wall curvature and, on the other hand, more friction loss
of the wake induced secondary flows within an S-shaped duct by the increased flow path length.
their influence upon the overall performance of the diffuser is Dominy and Kirkham [12] extended their work on the influence
remarkably slight [12]. No change in the surface static pressure of swirl together with non-rotating wakes as discussed above. The
distribution can be observed compared to a wake-free case. This same experimental facility was used with the same inlet conditions
means that only a redistribution of loss occurs instead of an except for a swirl angle of 151. The inlet flow shows thicker wakes
additional loss generation. and more pronounced secondary flows in comparison to the zero-
swirl case. Although the 34 vanes are simple and untwisted, the yaw
angle variation is small in the free stream outside the area where the
2.2.2. Influence of swirl interaction between the secondary vortices and the boundary layers
Additionally to the flow effects already discussed, a swirling flow occurs. Fig. 6 shows the secondary flows evoked by skewed wakes
at the diffuser inlet will further modify the pressure gradients and schematically. Downstream of the first bend of the S-shaped duct the
the boundary layers. The following part will discuss the influence of wake was found to be inclined to the radial direction. A cross-passage
swirl onto the intermediate turbine duct flow without as well as variation of the circumferentially averaged yaw angle of about 31 was
with wakes. observed. In contrast to the zero-swirl case the wake flow interacts
The effect of swirl in an annular diffuser is to superimpose a with the boundary layers over nearly the whole pitch. At the second
radial pressure gradient where the pressure at the casing is higher bend where the radial pressure gradient reverses the wake was
than at the hub. This gradient is the result of a radial equilibrium skewed further and the associated high loss fluid accumulated almost
between pressure and centrifugal force of a fluid particle. The entirely at the casing [12]. A loss core forms close to the casing wall
angular momentum has to be constant, if external forces can be and develops rapidly along the further duct. The boundary layer
neglected, so that the circumferential speed decreases with merges more intensively with the low energy wake fluid compared
increasing radius leading to the pressure rise. The radial pressure to the zero swirl case. Again a re-energization of the boundary layer
gradient caused by swirl is opposite to that evoked by the first occurs there. Also the vortex structure was found to change sig-
bend of the duct and additive to the effect of the second one. A nificantly: As a consequence of the wake skewing due to the swirling
further effect of swirl in an S-shaped diffuser is that the flow path flow hub and casing vortices of the same direction of rotation move
of fluid particle through the duct is increased, which leads to a relative to each other in the circumferential direction until they are
reduced effective curvature of the streamline in the radial bends adjacent. Consequently, the hub and casing vortices with the same
and thus to a reduced radial pressure gradient. Additionally, the direction of rotation merge but keep the casing vortices as the
friction losses increase slightly due to the longer flow path.
Bradshaw [13] reported that the turbulent energy production due
to the radial pressure gradient increases at the casing wall and
reduces at the hub wall. This difference implies a difference in growth
rate of the surface boundary layers and thus of their ability to sustain
streamwise pressure gradients without separation. For that reason
swirl stabilizes the casing and destabilizes the hub boundary layer.
Figs. 3b and 4b compare the flow through the example inter-
mediate turbine diffuser with an inlet swirl angle of 451 to the zero-
swirl case of Figs. 3a and 4a. The same mean inlet total pressure and
mass flow rate of 13.2 kg/s were applied for the flow simulation. The
Mach number distribution of Fig. 3b shows smaller cross-passage
gradients from hub to tip than for the zero-swirl case. The overall
level in Mach number is increased due to the additional circumfer-
ential velocity component.
In the axis-parallel inflow section the radial pressure gradient
caused by the circumferential velocity component is responsible
for the offset between the hub and casing pressure rise coefficient Fig. 6. Pressure-driven secondary flows evoked by S-shape, stationary wakes
lines. Again the upstream influence of the S-shape spreads the curves and swirl.
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 255

dominant structure [12]. Again no significant generation of additional wakes as well as inlet hub and casing boundary layers. Accord-
losses or loss redistribution was found for this relatively conservative ingly CP,ideal is 0.56, but in practice this theoretical value is
duct. This may change significantly for more aggressive designs. reduced by these blockage effects. The longer diffuser showed a
The flow effects discussed above were also summarized by Cp of 0.464 compared to a Cp of 0.420 for the shorter one. The
Dominy et al. [14] with a comparison of their measurements to diffuser efficiency ZD can be defined as
CFD and to results from an intermediate turbine diffuser arranged CP
downstream of a single-stage LP turbine stage. This test was carried ZD ¼  100% ð3Þ
CP,ideal
out in the Rolls Royce no. 4 Turbine Test Facility at a inlet Reynolds
number (based on the inlet passage height) of 3.2  105 together with Values for ZD of 83.5% and 75.6% can be calculated for the
a Mach number of 0.23 and a temperature of 370 K at the duct inlet. longer and the shorter diffuser, respectively. The CP distributions
The comparison of the turbine measurements with a CFD simulation for both setups are discussed in detail by Norris and Dominy [15].
based upon a circumferentially uniform inlet flow shows deviations It can be concluded that both the diffusion rate and the intensity
especially near the hub. The reason is that the flow downstream of a of the upstream wakes influence the structure and magnitude of
turbine stage is characterized by moving wakes and secondary flows the resulting secondary flow vortices significantly. The study also
from the rotor together with chopped wake segments and flow showed a decrease in pressure recovery as a result of shortening
effects from the stationary vanes. Only the latter can be captured by the diffuser by 30% in axial length.
time-averaged measurements in the intermediate turbine diffuser as
performed in this study [14]. The strong spanwise variation in swirl 2.3. Influence of struts
angle at the duct inlet due to the stage design and the secondary
flows cannot be considered. As an important conclusion of this work For multi-spool turbofan engines it may be necessary to equip
it can be stated that computational simulations generally give good the intermediate turbine diffuser with load-carrying struts to sup-
predictions of the flow development and of the loss generation port the inner shaft bearings and to provide a path for essential
within the diffuser when the truly three-dimensional nature of the engine services such as cooling air and engine lubrication oil. The
flow is represented. Calculations based upon a circumferentially influence of these struts onto the intermediate turbine diffuser flow
averaged inlet flow show clear limitations [14]. was the final topic within the series of experimental investigations
in the University of Durham Swan Neck Duct Facility (see Norris
et al. [16,17]). For this reason 26 evenly spaced struts have been
2.2.3. Influence of diffusion rate
added to the facility in addition to the 34 flat-plate vanes at the
The next topic dealt with in the series of investigations per-
diffuser inlet. The diffuser section used for this research corresponds
formed at University of Durham is the influence of the diffusion rate
to the shorter duct setup of [15]. Both experimental and numerical
on the intermediate turbine diffuser flow and its performance.
investigations have been applied to analyze the differences between
Norris and Dominy [15] used both experimental and numerical
the flow without and with struts. The strut surface was equipped
results to show the differences between two S-shaped ducts with
with a large number of static pressure taps in order to study the
the same area ratio and radial offset but with different axial length.
effect of one or two upstream wakes passing through the strut
Consequently, these intermediate ducts have different wall curva-
passage. Details on the instrumentation and the CFD predictions are
tures and diffusion rates to achieve the area increase over a much
given by Norris and Dominy [16].
shorter distance and thus a changed influence onto the secondary
The clean duct shows an unsteady separation bubble at the
flows generated by the upstream wakes.
casing right after the first bend where the diffusion rates are
For this investigation the University of Durham Swan Neck Duct
sufficiently high. In the measurements a thick and less clearly
Facility has been used with the flat plate vanes to generate wakes
structured low pressure region was found there, which is typical
without swirl and with a new inlet contraction in order to give a
for the energy dissipation of large eddies. The change of the local
more uniform radial total pressure gradient and a thinner and more
aerodynamic shape by this separation bubble might be the reason
stable casing boundary layer. Again the Reynolds number was
for the overpredicted peak suction value at the casing in the
adjusted to 3.9  105. The longer diffuser corresponds to the setup
region of the bubble by CFD [16]. The precise computational
used for the previous studies [11,12] while the shorter one has an
prediction of a separation bubble location and its extension is still
axial length reduced by 30%. Additionally, the intensity of the wakes
challenging. The existence of the bubble has been confirmed by
was varied by different lengths of the parallel inflow section.
surface flow visualization as well.
In the shorter duct the casing surface flow on the first bend
After installation of the struts in the same channel the surface
undergoes a larger acceleration and hence diffusion due to the
pressure distributions were found to change significantly. Their
smaller radius of curvature. This part of the flow was found to be
leading edges were positioned just ahead of the peak suction zone
responsible for a significant part of the overall duct total pressure
on the casing. Due to the fact that the strut has a symmetrical C4
loss caused by a thick low pressure region at the casing occupying
aerofoil shape, the strut diffusion (from 10% chord length onwards)
approximately 12% of the duct area compared to 8% in the longer
is located where the duct diffusion shows its maximum. The added
duct. This thicker boundary layer reduces the effective local duct
blockage by the leading edges is responsible for the fact that the
area and therefore less static pressure is recovered. For both
minimum CP value is decreased from 0.49 to  0.82 (see [16]). The
setups, the already described redistributed high-loss fluid in the
effect of the struts on duct blockage and hence flow acceleration is
mid-pitch casing region was observed but more pronounced for
also evident in the rest of the flow path. The overall diffuser pressure
the shorter diffuser [15].
recovery is reduced from 42% to 30% of the inlet dynamic head,
The theoretical maximum static pressure rise coefficient CP,ideal,
which means a diffuser efficiency drop from ZD ¼75.6% to 54%. This
which a diffuser can achieve, is only a function of its exit to inlet
smaller pressure recovery is responsible for a reduced work extrac-
ratio defined by
tion of a following LP turbine stage by decreasing the flow and stage
 2 loading coefficient [16].
1
CP,ideal ¼ 1 ð2Þ The area where the struts are positioned was found to be the
AR
part of the duct with the main losses. The overall duct loss with
AR is the area ratio of 1.5 for these setups. Of course, this installed struts is almost doubled. A massive flow separation
definition does not account for inlet blockage due to upstream occurring on the casing part of the strut was found to be the
256 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

contributions to the resulting flow field are much more complicated.


For example, the influence of rotor wakes is not an effect fixed in
space because of the blade movement. The influence of the
stationary vanes is disturbed by chopping their wakes by the rotor
into segments, which pass independently through the different
blade channels.
Also the following flow channel shape influences the spanwise
back-pressure profile of the turbine stage and unsteady interac-
tions with downstream vanes or struts have an upstream influ-
ence mostly onto the rotating blades. The decaying and mixing
out of all the effects from the upstream turbine stage occur in the
following intermediate turbine diffuser and change the flow
through it in contrast to a clean annular channel within a wind
tunnel test.

3.1. Test programs with ducts downstream of turbine stages


Fig. 7. High-loss fluid distribution evoked by S-shape, stationary wakes and struts.
In order to investigate the effects of an upstream turbine stage
onto the aerodynamics of an intermediate turbine diffuser a much
explanation for that. The strut diffusion and the critical duct more sophisticated test arrangement is necessary. The test facil-
diffusion together were sufficient to cause the flow to separate. ities capable to host these experiments can be classified as low- or
The interaction of the stationary wake flow and the strut has been high speed test turbines operating in a short duration mode or a
analyzed by CFD. In the simulation, the inlet wake is circumferen- continuously running mode. The investigations of inter turbine
tially shifted so that no direct strut–wake interaction occurs. ducts downstream of a turbine stage, which are published in open
A schematic flow situation is given in Fig. 7. The re-energization of literature, have been performed at four different test sites. In the
the boundary layer by the pressure-driven secondary flows due to following paragraphs these test setups will be introduced and an
the inlet wake was observed on the side of the strut adjacent to the overview of all the performed research work related to them will
wake flow. There the separation was found to be suppressed and be given. After that a detailed discussion of important results will
hence any reversed flow. On the other side of the strut the separation be presented in a more systematic manner.
causes a low total pressure region together with negative velocities.
After passing the strut trailing edge the interaction of these two flow
regimes causes the low energy zone to lift off from the casing surface 3.1.1. Oxford Rotor Facility at the Thermo-Fluids Laboratory of the
forming a loss core, which is still visible at the diffuser exit section University of Oxford (Southwell Laboratory, Osney Laboratory)
[16] (see Fig. 7). The Oxford Rotor Facility [18] uses an isentropic light-piston
Additionally, time-dependent flow characteristics have been tunnel to provide fully scaled test conditions for approximately
investigated by Norris et al. [17]. RMS fluctuations of the static 100 ms. Over several years a test setup with a one and a half
pressure coefficient at each tapping were used as indication of the turbine stage was used for a series of research work where the
unsteadiness there. More details are given in the paper itself. The second vane was arranged in an S-shaped swan-necked turbine
separation bubble in the clean duct without struts was found to diffuser. This setup characterizes the engine representative design
be of a periodic low frequency unsteady nature. The installed struts of a Rolls Royce 2nd stage vane of very low aspect ratio. The large
evoke the above discussed stronger and more ordered casing surface size of this wide-chord vane is caused by the necessity to route
separation, which shows still a closed bubble character. After adding service lines through its core and to use it as a rear engine support
the influence of the inlet wakes a more stable flow on the strut strut [19]. The flow field inside the vane passage is very three-
surface close to the casing was found with a much reduced mag- dimensional and the presence of the upstream stage induces a
nitude and spatial extent of the flow unsteadiness correlating to the high level of unsteadiness. After the time-resolved investigation
pressure-driven secondary wake flows [17]. of the vane-rotor interaction of the preceding HP turbine [20] and
This investigation shows that the blockage effect has to be the wake, shock and potential field interaction of the 1.5 stage
considered during the design process of a well performing strutted setup [21,22] Miller et al. [19] focused their research onto the
diffuser. It is obvious that a reduction in flow area due to the struts rotor flow evolution within the swan-necked duct and its influ-
can be compensated by an area increase by changing the meridional ence onto the downstream IP vane [23]. Both experimental and
channel shape in this region. This means that the overall negative numerical investigations have been performed with and without
influence of load-carrying structures can be much reduced com- this second wide chord vane. The test setups of the facility are
pared to the findings in the works of Norris et al. [16,17]. shown in Fig. 8a and b, the stage parameters are given in Table 1.

3. Aerodynamics of intermediate turbine diffusers arranged 3.1.2. Turbine Test Facility (TTF) at the Gas Turbine Laboratory of
downstream of turbine stages the Ohio State University
This transient facility is a short-duration shock-tube and
All the effects occurring in the exit flow of a turbine stage blowdown facility and has been described thoroughly by Dunn
influence the inflow to an intermediate turbine diffuser and hence et al. [24]. During the approximately 100 ms run time of the
every other downstream position by periodical unsteady effects facility the upstream area-averaged total pressure and the wheel
caused by the passing rotor blades modulated in their strength by speed are approximately constant, varying by 8% and 0.6%,
stator–rotor interaction. As already mentioned these effects are respectively, over the 40 ms interval of data sampling for a given
wakes, secondary flows, swirl, tip leakage flow and possibly trailing tunnel run. This research program utilizes an uncooled one-and-
edge shocks. Some of them have already been discussed in detail in a-half transonic turbine stage of which all three airfoils are
a systematic way, mainly separated from each other. Downstream heavily instrumented. The test setup is representative of an early
of a turbine stage these effects occur simultaneously and their design iteration of the PW6000 engine [25]. The meridional flow
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 257

channel is depicted in Fig. 8c and the design parameters are given


in Table 1. In Clark et al. [25] the experimental results were used to
validate two unsteady CFD predictions with and without airfoil
scaling. When no airfoil scaling was used and thus the true airfoil
counts were modeled, the predicted levels of unsteadiness on the
blade were in very good agreement with the measurements in the
short-duration turbine rig. Significant discrepancies in the predicted
first vane/first blade potential interaction were found when airfoil
scaling was applied. Building on this investigation a further study of
Davis et al. [26] addresses the physics of the complex reflected-
shock interaction between the blade, the intermediate turbine duct
and the second vane. Haldeman et al. [27] investigated experimen-
tally the influence of vane-blade spacing and clocking on the
unsteady surface pressure loading in this setup.

3.1.3. Large-Scale Facility at the Division of Fluid Dynamic of


Chalmers University of Technology
Compared to both test turbines introduced above this facility
is continuously operating and allows the testing of large scale
S-ducts downstream of a low-speed turbine stage under engine
realistic Reynolds numbers. The design, construction and com-
missioning of this test facility were part of the EU project AIDA
and are described by Arroyo et al. [28]. Although modern turbines
operate at near sonic conditions, many of the phenomena are also
present at lower speeds. Therefore low-speed large-scale rigs are
of considerable value for studying the flow details, which would
otherwise not be obtainable. The experimental facility is a horizontal
closed-loop wind tunnel and consists of a centrifugal fan, a flow
conditioning section, the test section with the turbine stage, duct and
de-swirler, and a flow return channel with a heat exchanger. The
electrical power of the fan is 110 kW delivering a maximum mass
flow rate of 11.1 kg/s. The speed of the turbine rotor is adjusted by a
hydraulic brake to 1090 rpm.
Fig. 8d shows the meridional section of the test setup AIDA C1
of the EU project AIDA consisting of a turbine stage designed by
Rolls Royce Deutschland together with an intermediate turbine
diffuser geometry from Industria de Turbo Propulsores S.A.
(ITP). Table 1 gives some design parameters of this test case. In
the past years a detailed test series has been performed within
this test setup including a performance evaluation and endwall
flow structure investigation by Axelsson et al. [29], a detailed
investigation of the time-averaged flow field by Axelsson and
Johansson [30] and of the time-resolved flow field by Axels-
son [31] as well as studies concerning turbulence and the integral
length scale of the diffuser flow by Axelsson and George [32,33].
Measurements at off-design conditions have been presented by
Axelsson and Johansson [34]. A comparison between experimen-
tal and numerical results of the AIDA C1 intermediate turbine
duct was drawn by Johansson and Axelsson [35].
Within the EU project AITEB-II the design and optimization of an
Fig. 8. Test setups for intermediate turbine diffusers at different test turbine rigs. intermediate turbine duct with load-carrying struts has been

Table 1
Characteristics of test setups for intermediate turbine diffusers.

Oxford Ohio Chalmers Graz

Rotational speed rpm 8910 – 965/1088/1160 10100/11000


pt inlet/p outlet – – 5.19 – –
Inlet Ma – 0.45 – – 0.48/0.65
Rel. inlet Ma – 0.98 1.17 – –
Re – – – 170000 –
Vane/blade/vane count – 36/60/21 36/56/36 36/72/72 24/36/48
Mean diameter m – 0.5796 0.99 –
Tip clearance/blade h % 2.25 – 1.45 1.5/2.4
258 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

executed by Wallin et al. [36]. The final geometry was installed in leakage flow close to the casing wall and the rotor hub passage
the Chalmers Large-Scale Facility for a detailed measurement vortex. As an example of all the downstream effects evoked by an
program for comparison with CFD simulations. The geometry of upstream HP-turbine stage the results from an unsteady simula-
this setup is given in Fig. 8e. The outcomes of this investigation were tion of the aggressive intermediate turbine test setup of Fig. 8f
presented for the flow field at design inlet conditions by Wallin should be used (see also Fig. 2 for the classification of its agg-
et al. [37] and for the flow field at off-design inlet conditions by ressiveness). The computations were performed using the in-
Arroyo et al. [38]. The complete aerothermal investigation of this house Navier–Stokes code LINARS, developed at Graz University
setup was presented by Arroyo [39]. In his work a unique database of Technology. The CFD code uses pressure-sensitive wall-
has been created, which is already in use to improve design methods functions and the one equation turbulence model of Spalart and
in industry. The heat transfer measurements were conducted with Allmaras, which showed good results in previous numerical
the help of a technique based on infrared thermography. Two studies. The CFD and post processing tools used for this investiga-
different inlet boundary conditions out of the database from [39] tion are described in [51] and [52], respectively. Fig. 9 shows time
were used for the continuation of the numerical work within AITEB-II snapshots of entropy contours in different planes throughout this
by Wallin [40], where diverse turbulence models were applied to aggressive intermediate turbine duct. The position of these planes
analyze the flow field and heat transfer structures. is given in the upper left corner. High entropy appears in red
while low entropy is depicted in blue. The wake flows and the loss
3.1.4. Transonic Test Turbine Facility (TTTF) at the Institute for cores due to secondary flows can be seen very clearly as zones of
Thermal Turbomachinery and Machine Dynamics of Graz University high entropy. In order to identify the secondary flows a stream-
of Technology wise-vorticity plot together with relative streamlines are given in
This high-speed facility is a continuously operating cold-flow the upper right corner. The vortical structures are the lower
open-circuit plant, which allows the testing of rig inserts under passage vortex (A), a pair of trailing edge shed vortices (B) and
engine-representative Mach numbers. It is a free running turboset (C), the upper passage vortex (D) and the tip leakage vortex
consisting of the test turbine itself and a three-stage radial brake merged with a scraping vortex (E) and have been discussed in
compressor and is driven by pressurized air delivered by a detail in [46]. According to the findings of Miller et al. [20] the
separate 3 MW compressor station. The brake compressor deli- lower passage vortex and the tip leakage vortex are the most
vers additional air mixed to the flow from the compressor station pronounced vortical structures. Additionally the stage design of
and increases the overall mass flow through the test setup. The air this transonic high pressure turbine stage shows a strong varia-
temperature at turbine stage inlet can be adjusted by coolers tion of swirl angle with almost zero-swirl at the tip and approxi-
between 40 and 185 1C. The maximum shaft speed of the test rig mately 451 counter-swirl at the hub.
is limited to 11,550 rpm. Depending on the stage characteristic In plane C the rotor wakes are marked with black dashed lines,
a maximum coupling power of 2.8 MW at a total mass flow of the high loss regions corresponding to the vortical structures are
22 kg/s can be reached. Detailed information on the design and encircled. Between the wakes the entropy is low (rotor passage
construction of the facility can be found in Erhard and Gehrer [41], 2 and 3). In every third rotor passage the periodical appearance of
on the operation in Neumayer et al. [42]. high-entropy fluid from the chopped nozzle wake segments can
Within the EU project AIDA the Transonic Test Turbine Facility be observed (rotor passage 1). As already shown in the previous
(TTTF) has been adapted in order to investigate the flow physics in sections the wakes enable the transport of boundary layer fluid
intermediate turbine ducts. Details can be seen in Göttlich et al. [43]. from the inner duct wall outwards within the S-shaped duct. The
During this test program three different intermediate turbine duct swirl downstream of the rotor is responsible for the skewing of
setups have been tested at two different operating points with the wakes, which can be observed in plane C1. This strong swirl
different inlet Mach numbers together with a variation of the rotor variation enhances the tilting of the wakes on their way through
tip clearance size. The layout of these setups is shown in Fig. 8f and the duct.
g. The parameters are given in Table 1. A comparison of the results The radial pressure gradient due to the S-shape of the duct lifts
for the two Mach number cases in the so-called aggressive setup the wakes and secondary flow effects off from the inner wall,
AIDA C4 representing the flow downstream of a single-stage and a makes them drift outwards and merge with the outer duct’s
two-stage HP turbine, respectively, is reported on by Göttlich et al. boundary layer. In plane C1 all flow structures are already above
[44]. The study on the rotor tip gap size variation in the same test the dotted line at about 30% span. High-loss fluid from the vane
configuration is discussed by Marn et al. [45] using time-averaged wake is still visible downstream of every third rotor passage. The
results and by Göttlich et al. [46] by means of time-accurate results further stretching and tilting of the wakes can be observed in
at the duct inlet together with surface flow visualization. Addition- plane C3. Here the highest entropy values can be found due to the
ally, a numerical study on this topic was performed by Sanz et al. merging of the wake fluid of the blade and vane (marked with A).
[47] considering also a theoretical zero tip gap case to separate the The strong tilting and stretching of the blade wake flow was also
flow effects evoked by the tip leakage. A much shorter so-called observed in the unsteady measurement results of Axelsson [31].
super-aggressive intermediate duct providing a separated flow In his large-scale low-speed facility also the HP vane wakes are
(AIDA C5) was used to create a database for CFD code verification. detectable over a long distance in the diffuser and their evolution
This duct was also designed as a platform for the application of is discussed in his work.
passive flow control devices. The separated flow through this In plane D of the setup in Fig. 8f directly upstream of the LP
configuration is discussed for both inlet Mach number cases by vane a nearly continuous stripe of high entropy can be seen,
Göttlich et al. [48]. The tip gap size variation in the same test which is slightly intermitted by low-loss fluid at a distance of
configuration is discussed in detail by Marn et al. [49]. Finally the one HP vane pitch. Also Dominy and Kirkham [12] found this
flow differences between configuration AIDA C4 and AIDA C5 are accumulation of high loss fluid almost entirely at the casing right
shown by Marn et al. [50]. after the second bend of the duct. Downstream of the LP vane
(plane E) high loss fluid accumulates at the casing on the suction
3.2. Downstream effect of a turbine stage side of the vane wakes (B). The leakage flow through the gap
between LP vane and inner duct surface becomes visible (C). This
It has been shown by Miller et al. [20] that for an unshrouded gap was necessary to rotate the vane ring in the circumferential
HP-turbine rotor the inlet flow is dominated by the rotor tip direction during the measurements in this test setup.
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 259

Fig. 9. Entropy contours in different planes (C–E) and vortical structures (top right) for AIDA C4 configuration at Graz University of Technology. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)

Due to the vane–vane count of 24–48 some sort of clocking vane–vane position onto the time-accurate pressure distributions
effect can be seen in the form of a higher loss region in every of the transition duct and the LP vane was found. The results
second vane passage (D). That clocking can be interesting for indicate that there is no best clocking position that uniformly
turbine stages up- and downstream of an intermediate turbine reduces the unsteadiness on the transition duct or LP vane, and
diffuser has been demonstrated by Haldeman et al. [27] in the test certainly not simultaneously for both the duct and the vane.
setup of Fig. 8c. As a major outcome a large effect of the relative That means that if there is a critical location on the airfoil for
260 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

which the amplitude of the unsteady surface pressure is too high, coherent pressure wave fronts propagating downstream. The angle
clocking may be one solution for reducing the local level of of these pressure wave fronts (marked with red dash-dotted lines)
unsteadiness [27]. depends on the local speed of sound and the circumferential velocity
Downstream of a transonic turbine stage the duct inlet is also of the rotor. The third type of pressure gradients are fronts of
dominated by rotor trailing edge shocks. These shocks are a major upstream running pressure waves (marked with yellow dashed
contributor to the unsteady interaction between the turbine stage lines), which are mainly evoked by reflection of the rotor trailing
and the downstream components affecting the unsteady pressure edge shocks onto the LP vane surfaces. During the rotation the shock
field and the aerodynamic performance. Davis et al. [26] showed that hits the LP vane and sweeps forward along the pressure side.
pressure waves and shocks propagate from the blades downwards The duct inner surface can be divided into three zones: the
and are reflected back upstream to the blade itself. The authors also upstream one (marked with A) is strongly influenced by the down-
showed that the computed time-averaged pressure distributions stream running pressure wave fronts from the HP stator–rotor
generally agree very well with the experiment. Additionally, the interaction, the downstream one (marked with C) is heavily domi-
interactions between HP vanes and blades as well as blades and LP nated by the upstream reflections from the rotor TE shocks and
vanes are well predicted at midspan. But the simulation results the section between them (marked with B) where the interaction
disagreed with the measurements in the near-wall region since the between the up- and downstream running waves takes place. In this
inner duct wall is the most influenced part of the whole intermediate region there seems to be something like destructive interference
turbine diffuser by rotor trailing edge shocks. between them so that no strong pressure wave leaving this area is
This unsteady interaction can also be seen in Fig. 9 by means of clearly detectable.
the magnitude of the pressure gradient, which is projected onto Special focus should be laid onto the interaction of the down-
the duct endwalls and the blade surfaces. Fig. 10 shows a time- stream running wave fronts with the rotor trailing edge shocks.
snapshot of pressure gradients plotted onto the inner duct surface There are many points at the inner duct surface where these
to visualize this heavy interaction there. The view is radially two pressure gradients intersect (marked with green circles). Due
inward; the flow direction is from left to right. The trailing edges to the different propagation directions and velocities of both
of the rotor blades can be seen on the left, the LP vane leading involved pressure gradients these points are not stationary and
edges on the right. move slowly downstream in the relative frame of reference. The
Here it should be noticed that the contents of Figs. 9 and 10 are two pressure gradients superimpose in these regions and form a
also available as videos in the Electronic Annex 1 in the online stronger gradient with a different direction of propagation. There-
version of this article. The unsteady interaction mechanisms can fore the trailing edge shock seems to split from the further down-
be more easily captured in these moving pictures. stream part and changes its angle rapidly (marked with black
Supplementary material related to this article can be found dashed lines) at the encircled points of intersection. The picture in
online at doi:10.1016/j.paerosci.2011.01.002. the lower right corner of Fig. 10 shows the radial extension of this
There are different types of pressure gradients visible in this CFD interaction mechanism. The depicted surface almost perpendicular
result of Fig. 10. The dominating structures are the rotor trailing to the inner duct is measurement plane C1. It can be seen that the
edge shocks (marked with black dotted lines). The interaction of the trailing edge shock only splits over a relatively small part of the
trailing edge shocks emanating from the upstream nozzle guide channel height and sweeps back in its original direction as the point
vanes with the rotor blades also causes pressure waves entering the of intersection moves further downstream during the rotation in the
duct periodically, e.g. see Schennach et al. [53], and can be seen circumferential direction.
in Fig. 10. These pressure waves are generated through the reflection Apart from these unsteady interaction phenomena caused
of the HP vane trailing edge shocks on the moving rotor blades. mainly by the moving rotor blades, which can only be captured by
Whenever a shock hits the suction surface and sweeps forward to unsteady CFD or time-resolved measurements there are several
the blade leading edge an upstream running wave is generated. important flow features originating from the stationary HP vanes of
After that these structures are reflected back on the rear part of the the upstream turbine stage. On the one hand, the downstream
HP vane surfaces and travel downstream through the rotor passages influence of them is visible as the propagation of vane related flow
(see video in the Electronic Annex 1 in the online version of this features like wakes segments, secondary vortices and pressure
article). These waves are most pronounced below midspan. Due to fluctuations at fixed circumferential positions. On the other hand,
the phase lag between neighboring rotor blade passages they form they modulate the rotor relevant flow structures like secondary
vortices, tip leakages and trailing edge shocks in their strength by
stator–rotor interaction during the blade passing event. In an engine
relevant test setup with an upstream turbine stage the conse-
quences of both the propagation and the modulation mechanisms,
but not the interaction itself, can be captured by time-averaged
measurement techniques like pneumatic five-hole probes, static
pressure taps and oil-flow visualization. The latter is usually used to
study the flow on the diffuser walls and on the following LP vanes,
which are necessary to ensure a proper radial mass flow distribution
throughout the duct. As an example for this the findings of the
surface flow visualization of Axelsson et al. [30] can be seen where
two streaks of removed color pigments were found on the outer
duct surface within one HP vane pitch. These streaks correspond
well to the time-averaged static wall pressure distribution. In the
five-hole probe measurements two pairs of zones with counter-
rotating vorticity were expected to be responsible for this phenom-
enon. But only with the help of time-resolved hot-wire measure-
Fig. 10. Time snapshot of visualized pressure gradient on the inner duct surface from
ments in [31] the unsteady interaction of the tip leakage jet with the
unsteady CFD (AIDA C4 configuration). (For interpretation of the references to color in stationary outer wall could be found as the cause of these streaks in
this figure legend, the reader is referred to the web version of this article.) the time mean. These streaks played also an important role in the
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 261

aerothermal investigation of [39,40] and were found to cause a swirling part load case. The longer effective flow path and a small
decreased heat transfer coefficient. separation on the suction side of the LP vane might be the reason
Time-averaged techniques are easy to apply, robust, not very time for that.
consuming and therefore cheaper in their application. In recent The situation changes significantly in the case of a strutted
research programs like AIDA and AITEB-II a huge amount of duct. The large airfoils within the annular channel can only be
measurements had to be carried out under variation of the operating designed optimally for one operating point of the upstream high
conditions influencing the Mach number and swirl, with different pressure turbine providing the corresponding Mach number and
rotor tip gap heights and for several duct designs being more or less swirl angle. Since the leading edges of these airfoils are closer to
aggressive. Therefore most of the measurements were performed the duct inlet and the chord length are very long, they are much
with time-averaged measurement techniques. The result is a data- more sensitive to flow distortions from the high pressure turbine
base for CFD code verification of the project partners. Most of them and off-design flows. The principles of the flow through a strutted
use steady numerical simulation tools during the engine design duct downstream of a turbine stage have been studied experi-
process with mixing planes between stator and rotor and further mentally as well as numerically by Wallin et al. [37] at design
downstream. Thus the HP vane effects are not visible in these results. conditions. The duct configuration corresponds to Fig. 8e, the
For instance, at the duct inlet only circumferentially averaged results operating speed of the turbine was adjusted to 1060 rpm corre-
from the time-averaged measurements can be compared to circum- sponding to a Reynolds number of 160,000 based on the inlet
ferentially averaged steady computational results from the rotating axial velocity and the inlet height. Measured two-dimensional
grid of the rotor. If there are deviations which occur mainly at the time-averaged contours of total pressure, total temperature, swirl
endwalls, it is hard to find out the discrepancies without knowing angle and pitch angle were taken as boundary conditions for the
the real flow physics with all unsteady effects. A numerical simula- numerical investigation of the duct so that the upstream turbine
tion of the test setup of Fig. 8d was performed by Johansson and stage could be omitted. A major finding of this investigation was
Axelsson [35] using a commercial flow solver without any adjust- the formation of a large vortex in the corner formed by the suction
ments to the turbulence models in order to show the current surface of the strut and the casing wall. This phenomenon is caused
capabilities of CFD for this kind of complex flows. As expected, the by the tip leakage flow of the upstream turbine stage, which
simulation managed to capture the total pressure and swirl angle impinges on the strut with high incidence and is forced away
profile fairly well between 20% and 80% span, but not close to the towards the hub. This dominating structure covers the entire span
walls. The static wall pressure distributions were also predicted well, at the duct outlet. The formation of these large vortices is shown in
but the total pressure losses were overestimated. principle in Fig. 11 with the help of streamlines released from 50%
Another difficulty is to find the correct boundary conditions for and 95% span at the diffuser inlet [37].
the numerical calculations. A procedure for obtaining the inlet Its position and size is captured satisfactorily by CFD although
dissipation using the one-dimensional energy spectrum gained by there is an over-prediction of the total pressure loss together with
means of hot-wire traverses is published by Axelsson and an under-prediction of the velocity. The circumferentially aver-
George [32]. The work presents the concept of integral length- aged radial loss coefficient profiles of CFD and experimental
scale and the differences between the physical length-scales and results agree reasonably well but the contour plots show larger
the pseudo-integral length-scale. differences. Therefore it is important to integrate the tip leakage
As a consequence of the complex intermediate turbine duct flow early in the design process and to consider not only radial
flows with many flow features not fully understood up to now, profiles when evaluating duct performance [37]. Another impor-
the future trend should go to unsteady multi-stage CFD simula- tant statement of the paper is that the implementation of inlet
tions and time-resolved measurement techniques despite their boundary conditions is difficult without a modeled turbine stage,
high demands in computational and testing time, respectively. either with a mixing plane analyses or much better with an
unsteady simulation if the computational resources are available.
In the aerothermal investigation of this test setup by Wallin [40]
3.3. Influence of Mach number and swirl also a strong circumferential heat transfer variation was found,
too. Therefore the author suggests including the upstream wakes
An intermediate turbine diffuser within a modern turbofan in the numerical analyses to account for their effects.
engine has to cope with different flow conditions corresponding The off-design situation of this test setup has been investi-
to the different phases of the flight like take-off, cruise and gated in the second part of this study, namely Arroyo et al. [38].
approach. The Mach number as well as the swirl angle will vary At lower and higher load of the high pressure turbine stage the
significantly between them. For that reason the duct should guide flow features change significantly together with the duct inlet
the flow without separation over the whole operating range in flow angle (nominal flow angle þ111 and  121 to nominal flow
order to ensure safe engine operation. angle, respectively). The flow situation for low-load (mid) and
In the EU projects AIDA and AITEB-II different studies with off- high-load conditions (bottom) can also be seen in Fig. 11 [38] and
design flow conditions have been conducted. In the Large-Scale compared to the design flow (top). For the lower load the duct
Facility at Chalmers University the AIDA C1 intermediate diffuser starts to separate (separation zone is marked in Fig. 11). For both
setup presented in Fig. 8d and classified in Fig. 2 was tested under off-design conditions the overall loss coefficient is larger (0.18
different representing operating conditions of the upstream turbine and 0.16 for lower and higher load, respectively) compared to the
stage (design point 1090 rpm and  211 swirl) providing a swirl design case (0.12). As the main source of loss the shear layer
angle variation of 7101. The speed was changed to 967 rpm for part produced by the tip leakage flow was found. While this effect is
load and 1160 rpm for peak load leading to a swirl angle of  311 and stronger for the higher-load case, the occurring separation seems
111, respectively [29]. The diffuser shows similar robust operation to be the main contributor to the losses of the lower-load case. For
for all swirl angles with the smallest peak suction occurring for the more details on the CFD, the experiments and the flow results
part load condition due to the largest effective wall curvature radius. see [37–39].
As expected, the strongest pressure gradient between inner and In the future there will be still two engine families with
outer duct wall in the first bend was found for peak load with the different HP turbine architectures in service, the first one with a
minimum swirl angle. The pressure loss from diffuser inlet to LP vane single-stage and the second one with a two-stage turbine driving
exit was smallest for the design point and largest for the highly the HP compressor shaft. To perform research work for both types
262 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

change for the lower Mach number case. This means that diffu-
sion took place earlier in the diffuser for this operating point. The
surface flow visualization showed stable flow conditions of this
aggressive diffuser geometry for both Mach numbers. It was not a
target to study off-design situations for both duct inflow environ-
ments in [44].

3.4. Influence of rotor tip gap size

One important flow feature provided by an upstream high


pressure turbine stage at the inlet to the intermediate turbine
diffuser is the tip leakage flow. In most of the cases this turbine
stage is unshrouded due to the necessity for film cooling of the
blades. Therefore the typical tip leakage vortices will be formed
by the roll-up of leakage fluid after crossing the blade tips from
the pressure to the suction side. These streamwise vortices at the
outer diffuser wall rotating in the circumferential direction
together with the rotor blades energize intermittently the casing
boundary layer by transporting higher momentum fluid from the
core flow to the near-wall region and low energy fluid away from
it whenever they pass by. As a consequence, the mixing process
is enhanced. Also a direct wall jet through the tip gap influences
the outer duct boundary layer similar to an injection of high
energy fluid. The main parameter influencing the strength of
these energizing effects is the rotor tip gap size. A larger gap will
increase the tip leakage flow and help to prevent the casing
boundary layer from separation due to the strong adverse
pressure gradient and thus increase the diffuser efficiency. Addi-
tionally, a larger gap reduces the power output of the high
pressure turbine. Moreover, the stronger the tip leakage vortex
is the higher are the mixing losses. On the contrary, a smaller gap
increases the turbine efficiency but reduces the positive effects
onto the intermediate turbine duct flow. There seems to be an
overall optimum for a given turbine/diffuser configuration. Also
the influence onto the following low pressure turbine has to be
taken into account.
The influence of the rotor tip gap variation onto the flow
through the aggressive AIDA C4 diffuser configuration (see Fig. 8f)
downstream of a transonic turbine stage has been experimentally
investigated by Marn et al. [45] and Göttlich et al. [46]. The duct
geometry was designed 20% shorter than the state-of-the-art for a
duct with this radial offset and is classified according to the
Sovran and Klomp performance chart in Fig. 2 where it is directly
on the classical limit for aggressive designs. The rotor tip gap
variation was realized by interchangeable liner rings around the
turbine rotor. In the first part of this work [45] time-averaged
flow measurements were performed by means of five-hole probe
area traverses, static pressure taps and boundary layer rakes. The
operating conditions of the high pressure turbine were set to
provide a diffuser inlet Mach number of 0.65 and an inlet swirl
angle of  151. A rotor tip gap size of 1.5% and 2.4% span were
applied. This influence of the tip gap variation varies within the
Fig. 11. Flow through strutted intermediate turbine diffuser for design (top), low- duct as shown by the data in the different measurement planes.
load (mid) and high-load conditions (bottom), streamlines released from 50% and
The measurement locations can be seen in the top left picture
95% span at inlet [37,38].
of Fig. 9. While immediately downstream of the rotor only the
upper part of the flow field is directly affected by the variation of
of HP turbines, measurements were performed within test setup the tip clearance size, in planes further downstream the whole
AIDA C4 (see Figs. 8f and 2) by Göttlich et al. [44] by means of flow field is concerned. For detailed results see [45], whereas here
pneumatic five-hole-probes. Measurements in the intermediate only the diffuser total pressure loss will be used to show the
turbine duct were performed at two different Mach numbers influence of the tip gap variation. The diffuser total pressure loss
(0.65 and 0.48) representing the flow downstream of a single- between inlet and outlet for the 1.5% span tip clearance is about
stage and a two-stage HP turbine, respectively. The flow behavior 2.8% and increases by about 7.5% up to 3.0% total pressure loss for
through this setup was found to be very similar for both operating the 2.4% span gap. Fig. 12 shows where the loss is generated
conditions. The main difference appeared at the end of the flow within the duct. It can be seen that the most part of the loss is
path. The flow parameters in the downstream part did hardly produced from plane C to C1 for both clearances in the same
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 263

magnitude. From plane C to C1 the flow is turned radially out- stationary wakes. Additionally, the outflow from the transonic
ward around the strong bend at the outer duct where the highest turbine stage was investigated and compared to numerical results
Mach number was observed. At the inner wall the flow channel of a steady Navier–Stokes solution. Thereby it was possible to
opens rapidly. In the following planes the loss production for the identify the formation of trailing edge shed vortices at midspan
larger gap is higher as is shown in Fig. 12. (see Fig. 9 top right). The strong tilted wakes and the modulation of
A detailed study of the duct inflow has been performed in the the flow effects due to stator–rotor interaction were quantified by
second part of the work [46]. A two-component LDV was used to the unsteady measurements and confirm the numerical results of
gain time-resolved results in order to better understand the Fig. 9. They can be found in the electronic annex of the online
structures and differences observed in the five-hole-probe mea- version of this paper. The surface flow visualization of Fig. 13 shows
surements of [45]. The tip leakage flow affecting only 20% span at no separation in this high diffusion intermediate turbine duct.
the duct inlet causes a different flow behavior over an increased A comparison with the steady CFD simulation confirmed that the
part of the channel height in the more downstream sections of the effects of the tip clearance variation could be captured precisely.
transition duct. As shown before the flow within such an S-shaped In the work of Sanz et al. [47] the unsteady flow through this
duct is very sensitive to relatively small flow variations at the duct whole configuration was numerically investigated for both gap
inlet which agrees very well with the findings of Dominy and heights as well as for a rotor with zero gap height. Again the in-
Kirkham [11] during their investigation within a setup with only house Navier–Stokes code LINARS was used [51]. The unsteady
results were compared at the stage exit and throughout the
intermediate turbine diffuser in order to study the flow physics.
The calculation for zero gap height allowed separating the
influence of the tip leakage flow.
Fig. 14 shows the radial distribution of the stage loading for
the three cases investigated: For non-zero gap height the loading
is the same in the lower part of the stage and differs slightly
above 75% span. There the loading is higher for the smaller gap
height of 1.5% span. For zero gap height the loading is slightly
smaller in the lower part of the stage due to a reduced stage
pressure ratio there, but remarkably higher in the upper part of
the stage because of the non-existing tip leakage flow.
In order to demonstrate the complex flow structure the simu-
lated time-averaged streamwise vorticity distribution in plane C is
shown in Fig. 15 for zero gap and 1.5% span gap height. Additionally
streamlines through the vortical structures are depicted. The upper
trailing edge shed vortex ‘‘C’’ is slightly stronger and moved
Fig. 12. Loss evolution in an intermediate turbine diffuser (AIDA C4 test setup) upwards in the zero-gap case because of the nonexistent tip leakage
depending on the rotor tip gap. flow. In the upper section of the channel the flow is dominated by

Fig. 13. Oil flow visualization in an intermediate turbine diffuser (AIDA C4 test setup) for two rotor tip gaps: outer wall (top) and inner wall (bottom).
264 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

the upper passage vortex ‘‘D’’ for zero gap and by the tip leakage case. For both gaps the principal trend is very similar. The minimum
vortex ‘‘E’’ for 1.5% span gap height. It can be seen that the tip static pressure at the outer wall is detected after the first bend
leakage flow displaces the opposite rotating upper passage vortex where the flow is accelerated. From there on the pressure con-
downwards and reduces its strength. In Fig. 9 also the entropy tinuously increases. At the inner wall the maximum static pressure
contours of plane C indicate that the tip leakage flow is the is found at a duct length x/L of about 0.46. A distinct minimum
dominant loss source [46]. occurs right after the change in curvature at the inner wall. There
For zero gap height a strong movement of the passage vortex for 1.5% span gap height the static pressure is slightly smaller. The
core ‘‘D’’ up- and downwards and a change of its shape could be comparison with measurement data shows a good agreement and
observed in the unsteady results during the blade passing cycle confirms the reliability of the numerical investigation, although
(see Fig. 9). Also the shocks emanating from the rotor extend from there are slight differences between the experiments and the
the hub to the tip in plane C with the highest strength in the tip computation for the outer duct wall for x/Lo0.25. These differences
region. For 1.5% span gap height the shock structure is very might be partly caused by a larger measurement uncertainty at the
similar up to 75% span, although it is differently inclined. But from beginning of the duct due to the high flow unsteadiness there.
there on the tip leakage flow dominates the Mach number distri- For the zero gap case the pressure recovery behavior is quite
bution and no shock occurs in the tip region (see [47]). different. At the outer wall it starts with a higher static pressure
It turned out that for this aggressive duct the tip leakage flow and a smaller velocity and accelerates around the first bend.
has a very positive effect on the pressure recovery by means of Because of the lower dynamic pressure and since the flow is not
energizing the outer duct boundary layer through the leakage able to follow the inner contour completely, the pressure mini-
wall jet combined with the wallward transport of high energy mum is less pronounced and occurs slightly earlier. From there on
fluid by the tip leakage vortex. This thinner boundary layer with the pressure increases and ends significantly lower than for the
less blockage effect is the reason for a larger effective flow area duct flow with gap. The higher velocities outside the stagnant
and consequently more diffusion. On the other hand, it has to be endwall flow and the blockage of local zones of separated flow
taken into account that the stronger tip leakage flow mixes out on (not shown here) cause the lower pressure recovery. At the inner
its way through the duct. For this reason the total pressure loss contour the trend is similar to the cases with gap up to nearly plane
increases for the larger tip gap as depicted in Fig. 12. Fig. 16 C3. From there on the pressure recovery is smaller due to the higher
shows the pressure distribution along the duct by means of the velocities at the hub and midspan section as described above.
pressure rise coefficient for both tip gaps and the zero tip gap Finally the efficiency of the whole system was evaluated. Looking
at the isentropic stage efficiency, the configuration with zero rotor
gap has – as expected – the highest efficiency of 93.2% due to the
additional work generated in the tip region (see also Fig. 14). With

Fig. 14. Stage loading at the inlet of an intermediate turbine diffuser (AIDA C4 test Fig. 16. Pressure rise coefficient of an intermediate turbine diffuser (AIDA C4 test
setup) depending on the rotor tip gap. setup) for different rotor tip gaps (0.8 mm¼ 1.5% span, 1.3 mm¼2.4% span).

Fig. 15. Flow structure at the inlet of an intermediate turbine diffuser (AIDA C4 test setup) without and with rotor tip gap.
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 265

increasing rotor tip gap height the efficiency decreases considerably This transition duct is further 20% shorter than the configuration
(90.5% and 89.3%, respectively). AIDA C4 and provides fully separated flow on the casing wall
Although the diffuser efficiency ZD is only 33.5% for the zero (see Fig. 8g). Again the duct geometry is classified by the Sovran
gap case compared to 67.4% and 67.0% for the 1.5% and 2.4% span and Klomp performance chart in Fig. 2, where it is highly above
gap, respectively, the efficiency of the whole configuration (HP the classical limit for aggressive designs. The intermediate turbine
stage and intermediate turbine diffuser) is remarkably higher for diffuser setup C5 is designed very steep and due to cost reasons it
the zero-gap case, and decreases with increasing gap height. The uses the same LP vane row like the former investigated aggressive
total pressure loss coefficient shows similar losses for the zero setup. Therefore the radial offset and area ratio is identical, but the
gap and the 1.5% span gap case. The losses generated by the second bend is much more pronounced in order to fit the meridional
stagnant flow at the outer wall for zero gap are balanced by losses duct shape to the existing downstream flow path components as it
generated by the tip leakage flow for 1.5% span gap height. The would be for a complete new test case. The same operating
increased gap height of 2.4% span leads to a significant loss conditions like in Göttlich et al. [44] (two different Mach number
increase in the duct. It is important to remark that the loss cases: 0.65 and 0.48) were used.
evaluation in [47] does not consider that for the zero-gap case the The measurements showed a very poor efficiency of this separ-
inlet profile to the succeeding LP stage has a significant deficiency ating intermediate turbine duct for both Mach numbers. The height
in the hub region, which will negatively influence the flow there. of the backflow region increases with increasing Mach number. This
Therefore for a study considering all effects of the tip clearance separation zone changes the effective duct shape massively so that
variation a downstream turbine has also to be included. the geometrical diffusion cannot be realized with such a short
Due to the strong influence of the tip clearance on the flow design. For more details on the core flow through this setup see [48].
field the entire system consisting of turbine stage, intermediate An interesting topic of this paper [48] was to get more insight
turbine duct and following parts has to be taken into account into the separated flow region. This was done by means of oil-film
during the design process of aero engines. Furthermore, it is also visualization as well as of a numerical simulation of the duct
necessary to keep the blade tip clearance constant during opera- without LP vanes and the measured inlet flow conditions as
tion particularly if the intermediate turbine diffuser is operating boundary conditions. The black surfaces of the duct and the LP
near separation. Although a larger tip clearance has a positive vanes were painted with a mixture of titanium oxide and mineral
influence on the flow at the outer wall due to the energization of oil. The CFD calculations were performed with FLUENT 6.2.16, using
the boundary layer the additional losses of an aggressive duct the k e turbulence model together with standard wall functions. At
must be much lower than the efficiency gain due to a higher the exit a constant back pressure for plane D was applied. Fig. 17
by-pass-ratio and due to the weight reduction of the engine. shows a comparison of the results of this oil-film visualization with
the CFD simulation for the higher Mach number case at the outer
duct wall. The flow direction is from the left to the right. The arrows
3.5. Separated duct flows indicate some shear stress lines. The flow is nearly axial at the duct
inlet with some variations over the HP vane pitch. Directly after the
Although no separated intermediate turbine diffuser will be first radial bend of the duct the flow starts to separate downstream
applied in a reliable aero engine or in other well-performing of a region of low wall shear stress. In principal a separation within
technical applications there has also been some research work the annular duct can be seen in the oil flow visualization as a
performed on this topic. For CFD applications dealing with separated deflection of the flow in the circumferential direction. In this test
flows or flows close to separation it is necessary to evaluate if the setup a massive recirculation region is formed. The separation is
solver is able to predict the separation start and its extension in a indicated by an upstream red line parallel to the separation lines
correct way. The super-aggressive intermediate turbine diffuser (convergent shear stress lines). In this separated flow vortical
setup AIDA C5 of the TTTF at Graz University of Technology has structures in the shear stress lines are clearly visible (foci, F). The
been designed as a test case for this purpose (see Göttlich et al. [48]). centers of these periodically circumferentially shifted vortices have a

Fig. 17. Oil flow visualization (left) and CFD simulation (right) of separated duct flow (AIDA C5 test setup). (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)
266 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

distance of exactly one HP vane pitch from each other. The sense of spots on the casing surface, thus possibly leading to severe engine
rotation and the formation of these foci have been verified by the damage and therefore their formation must be avoided [48].
CFD simulation and observed through an acrylic glass window The influence of the rotor tip clearance size onto the separated
during operation of the facility. A video of this experiment is flow through the super-aggressive intermediate turbine duct
available within the Electronic Annex 1 of the online version of this setup AIDA C5 has been investigated by Marn et al. [49]. The
article. Within these vortical structures the pressure is low so that outer duct flow was found to be separated for both gaps (1.5% and
the oil is drawn away from the duct surface like in a twister. 2.4% span at the duct inlet). However, it seems that the separation
Therefore the black surface of the duct is visible there. The separa- is less extended for the bigger gap at the casing. The positive
tion lines end up in the neighboring foci. Upstream of the LP vane effect of the tip clearance flow is responsible for that. At the hub
the flow reattaches again. The reattachment lines (divergent shear the duct becomes separated at different circumferential positions
stress lines) marked in green are the connection between singular for the larger gap. It seems that this circumferential separation
points (O). At these points all shear stress lines have their origin. position is also affected by the first stator. Further, the formation
Between them another type of singular points can be found on the of the vortical structures discussed in [48] at the outer casing
reattachment lines namely saddle points (green shaded regions, S1). downstream of the first bend, which are imposed by the HP-vane
There the flow from both adjacent origins collides: A part of the flow (and therefore, the distance between two vortex cores is exactly
then forms the backflow into the separation zone and the remaining one HP-vane pitch) could be observed for both gap sizes. Only the
part is traveling downstream and enters the LP vanes. The backflow circumferential position changed for the different tip clearances
splits up again in the saddle points S2 and then enters one of the due to a slightly changed flow angle at the outer duct wall. This
adjoining foci (F). influence has been confirmed by a CFD simulation. The simulation
All these structures can only be seen upstream of every second was run with and without the upstream stator whose influence is
LP vane in the flow visualization. The same flow features can be then considered by inlet boundary conditions based on the
seen as well in the results of the CFD simulation without modeled measured flow. The vortex formation is only seen in the case that
LP vanes and given constant back pressure (Fig. 17 on the right). the HP-vane influence is given by the inlet boundary condition.
That means that these effects are independent from the down- Fig. 19 shows a comparison of the pressure rise coefficient for both
stream components. Only one singular point O and S1 can be tip gap sizes. The minima are at the same axial positions. At the
found within the simulated sector. This indicates that the poten- casing the static pressure is slightly lower and at the rear part of the
tial field of the downstream LP vane row has a direct effect onto duct the pressure is higher for the larger gap. This is due to a
the reattachment behavior but not on the start of separation. reduced separation zone at the casing. Also a higher pressure at the
Therefore it was stated that the formation of a vortical structure duct outlet and a more pronounced peak suction was found for the
within the separation zone is imposed by the first stator, which is
also confirmed by the distance between two foci, which is the HP
vane pitch [48].
Fig. 18 shows a comparison of the oil flow visualization at the
casing between the two Mach number cases. The green arrows
indicate the inflow direction. The two foci in a distance of one HP
vane pitch marked with a yellow arrow cannot be seen for
Ma¼0.48. That means that the effects that evoke the formation
of these vortical structures are less pronounced in this case. For
both operating conditions the singular points (marked with O, S1
and S2) where the flow splits up and spreads out in different
directions can be seen upstream of every second LP vane. In
between the flow direction cannot be defined clearly for the lower
Mach number. For both cases the separation lines (red) are clearly
visible.
Another important point is that, if one runs a CFD calculation
with a mixing plane boundary condition downstream of the rotor
it cannot capture these flow phenomena and thus most likely
underpredicts the duct losses. Further, a more serious problem Fig. 19. Pressure rise coefficient for separated AIDA C5 diffuser setup with
can result in a real engine where these structures produce hot different rotor tip clearances (1.5% and 2.4% span).

Fig. 18. Separated surface flow for two Mach number cases at outer duct wall (AIDA C5 test setup). (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 267

possibilities to delay or suppress the negative influence of some


flow effects with the overall target to increase the design space of
intermediate turbine diffusers.

4.1. Shape optimization and endwall contouring

The early preliminary design phase of an intermediate turbine


duct is often done with the help of experimentally generated
standard diffuser performance maps like the well known Sovran
and Klomp diagram [8] which is used, beside the classification of
the aggressiveness, to correlate a pressure rise coefficient CP for a
given area ratio AR and a non-dimensional duct length L/hin. But it
is only valid for non-swirling flows. Also the theoretical max-
imum pressure recovery represented by

Fig. 20. Pressure rise coefficient for AIDA C4 (aggressive) and AIDA C5 (super-
CP,ideal ¼ 1AR2 ð4Þ
aggressive) test setups. which is only a function of AR does not consider swirl. On the
other hand, an intermediate turbine duct downstream of a
larger gap. At the hub the pressure was found to be equal over most turbine stage usually works with a significant inlet swirl and is
part of the duct. The diffuser efficiency ZD, which is calculated equipped with struts. Therefore the total pressure loss of such a
according to Eqs. (2) and (3), is much better for the larger rotor tip design has to be estimated with the help of a database where
clearance (ZD ¼12.7%) compared to the value for the smaller one parameters like swirl angle, duct blockage and strut losses have
(ZD ¼ 34.0%). The measurement results show that the diffuser been added and which is usually gained by a systematic compu-
efficiency ZD is much higher for C4 (ZD ¼51.4%) while for C5 the tational study. As an example the work of Gräsel et al. [54] should
duct even accelerates the flow due to the blockage of the large be named where a linear increase of the pressure loss z with the
separation zone (ZD ¼ 34.0%). The negative value indicates that the thickness to chord ratio T/C of a strut and also a linear contribu-
averaged static pressure at the exit is lower than at the inlet. tion of the strut count n was found. With such correlations and a
Finally, a comparison between the flow through an aggressive good database a preliminary design can be defined together with
and a super-aggressive intermediate turbine duct has been drawn an estimation of the expected pressure rise and loss (see Fig. 21).
by Marn et al. [50]. Similar to the study of the influence of the After the fixation of the preliminary design the final design
diffusion rate by Norris and Dominy [15], two different duct should be developed by shape optimization. The first step is the
configurations have been investigated using the same area ratio parameterization by means of a multilevel parameterization with
but different duct lengths, namely the aggressive AIDA C4 and the familiar engineering parameters for the preliminary design and
super-aggressive AIDA C5 intermediate turbine duct. As already purely mathematical parameters for the subsequent detailed
mentioned configuration C5 is 20% shorter than C4, which is design [54]. For the duct optimization the characteristic para-
already 20% shorter than a state-of-the-art duct, and thus pro- meters are the area ratio AR, the duct length L, the inlet hub
duces a large separation zone at the casing and therefore much radius r1h and the mean radius ratio r2/r1 between exit and inlet.
higher losses. For C5 the separation obstructs such a large amount The second parameter set defines the contour of the hub and
of the passage height that the flow is accelerated and therefore casing walls between the duct inlet and exit and is suitable for the
the averaged static pressure at the outlet is lower than at the
inlet. This means that the diffuser is working as a nozzle. Fig. 20
indicates that the pressure at the duct outlet at the casing as well
as at the hub is higher for C4 than for C5. Further, it can be seen
that the flow is strongly accelerated around the first bend at
the hub.
The aim of all these investigations [48–50], which is unique in
literature was to gain a more detailed insight into a duct flow
with heavy separation. If a duct is designed close to separation
and the flow starts to separate, for example, due to a changed
(smaller) HP-rotor tip clearance, the following turbine stage is
provided with totally different inflow conditions even when the
separation reattaches before entering the vane row. The calcula-
tions show that it is essential to model the influence of the HP
turbine exit flow field. Without this information the duct flow
cannot be predicted accurately and the losses will be predicted
too low.

4. Potential measures for increasing the diffuser design space

As shown in the previous parts the aerodynamics of inter-


mediate diffusers are complex and it seems that clear limits are
set for the designer to create an aggressive duct: The flow will
start to separate and/or the loss situation of the diffuser itself or
of one of its neighboring components will become more and more
critical. Therefore it should be further discussed, if there are Fig. 21. Optimization procedure.
268 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

design optimization. With the optimal shape of the diffuser walls boundary conditions at the inlet and the exit. For more details of the
the risk of boundary layer separation at the casing can be parameterization, the DOE with a face-centered composite design
suppressed. (FCCD), the flow solver (an in-house RANS solver with high-Re
The two most common parametric representations for aero- number or low-Re number k–e closure [55]) and the second-order
dynamic shape optimization together using open cubic B-splines response surface see the work of Wallin and Eriksson [56].
with five control points and the C1-continuity at the borders have Besides a 3D strutted compressor duct a 2D axisymmetric
been applied in the work of Gräsel et al. [54], namely camber line turbine duct has been studied as a test case for this optimization
and thickness as well as hub and casing parameterization. For both procedure. The geometry of the baseline intermediate turbine
methods a total of eight parameters p¼8 were retained supporting diffuser fits to the Chalmers Large-Scale Facility inlet and outlet
of the design of experiments (DOE) in order to determine the radii and was optimized for minimum pressure loss. Two design
sensitivity of each parameter on the pressure recovery and the spaces of different sizes were evaluated with the high- and the
pressure loss, to identify significant parameter interaction, to deter- low-Re number turbulence models, in order to evaluate the most
mine the dependence between pressure recovery and loss as well as practicable procedure. Here a poor loss prediction for some of the
to define the optimization strategy [54]. Two different sampling extreme designs of the larger design space was observed for the
techniques for the DOE have been investigated, the Central Compo- high-Re number model leading to a more unreliable regression
site Design (CCD) and the Latin Hypercube Sampling (LHS), where model. The low-Re number model response surface showed a
the first of which generated more extreme designs. After that the very good agreement with the loss coefficient obtained from a
automatic creation of duct geometry and computational mesh as detailed CFD analysis of the optimum geometries for both design
well as CFD calculation (FLUENT V6.2 with k–o SST model) and spaces, which are very similar for this axisymmetric non-separat-
analyses were managed by a batch procedure. A second-order ing duct. For the smaller design space the main result was a
response surface approximation (RSA) was conducted for all four decrease in total pressure loss by 19% for the high-Re and 16% for
DOEs. It was found that the RSA is independent of the sampling the low-Re number model analyses. An important optimization
technique, which clearly favors the LHS over CCD because of its mechanism found in [56] is to shift the mean line curvature towards
robustness and lower sample number. Also the tight correlation the exit of the duct. But the obtained loss values are too low because
between pressure recovery and pressure loss makes the design of a of the missing effects from an upstream turbine stage.
strutless intermediate turbine diffuser with a high pressure recovery If an initially vane-less intermediate turbine diffuser is
and a low pressure loss challenging. Additional details on the equipped with struts like in [16,17], the effective flow area will
methods and programs used are given in their paper [54]. change. Modifying the flow area distribution in that way to get
An important result for the intermediate turbine diffuser one similar to the initial vane-less design and accounting for the
design is that the observed CP and z variations in the found corre- strut blockage is sometimes referred to as area ruling [57]. The
lations [54] are only in the range of 735% and 750%, respec- area change could be obtained by increasing the hub-to-casing
tively, of the mean values of the investigated test case (AR ¼1.35 radial distance or by increasing the endwall distance only in the
and L/hin ¼4). This indicates the limitations of the preliminary vicinity of the strut, leading to an axisymmetric or a non-
design assessment using only the characteristic parameters as axisymmetric endwall re-design, respectively (see Fig. 22).
well as the potential gain by an optimization of the duct endwalls. In Wallin et al. [36] the design of the strutted intermediate
Another result of the DOE study is that the decoupling of the turbine diffuser of the test geometry for the EU project AITEB-II
casing and hub endwall contour allows a reduction of about 50% given in Fig. 8e is described. The duct is equipped with nine non-
of the design parameters with a close coupling between design lifting modified NACA four-digit series airfoils and has been
parameters and objective function. Also the ducts are very optimized for minimum total pressure loss by an axisymmetric
sensitive to changes of the casing endwalls. endwall modification of a baseline duct design. The applied
Additionally, a shape optimization of a preliminary design of optimization procedure was similar to those presented in [56].
the test setup of Fig. 8f has been performed by Gräsel et al. [54] A baseline duct design was constructed by using a fifth-order
and resulted in a significant increase in pressure recovery without polynomial as mean line and another fifth-order polynomial as
degradation of the pressure loss. However, the duct loss calcu- area distribution, which was slightly modified to account for
lated with the applied profile of total pressure, which is constant some strut blockage. The struts were designed based on a given
all around the circumference, was underpredicted due to the duct inlet swirl profile using low-cost streamline calculations. For
missing mixing out of the circumferential flow variations caused every new candidate in the optimization procedure a new strut
by stationary and rotating blades. Therefore the computational design fitting the endwalls had to be generated, which is only
domain should also contain the HP turbine to assess the effect of a coupled with the inlet swirl and the endwall shape of that duct.
change in radial pressure gradient due to the optimization.
Another approach was applied by Eriksson [55] for their tran-
sition duct shape optimization using CFD and response surface
methodology. Again a design space was defined by design para-
meters and a low number of candidate designs were chosen
according to the DOE theory. With a CFD simulation of each design
the objective function was evaluated and a response surface was
approximated. In order to keep the number of design parameters
at a minimum, but still obtain maximum flexibility of the duct
geometry an efficient way was introduced by applying perturbations
to a reference duct. Therefore an approximation to a perturbation,
which optimizes the duct geometry with respect to the objective
function was sought. There was no need to parameterize the original
geometry. A linear combination of orthogonal polynomials in the
interval 0rxrL was used to construct the approximate perturba- Fig. 22. Optimized axisymmetric (top left) or non-axisymmetric endwalls (top
tion in such a way that the radius, the flow angle and the curvature right) and non-axisymmetric endwalls designed with parts of harmonic functions
defined by the upstream and downstream components were used as (bottom).
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 269

The optimization resulted in a duct design with a loss reduction the EU project DREAM—Validation of Radical Engine Architecture
of 25%. Details of the procedure performed in two steps are given Systems [59].
in [56]. Primarily the new designed duct intends to compensate the
strut blockage and opens up in its region thus reducing the
otherwise following strong adverse pressure gradient. An early rapid 4.2. Flow-controlled intermediate turbine diffusers
diffusion seems to be favorable in reducing the losses. Also a
significant contraction in the strut trailing edge region was found Flow control enables the designer to generate more aggressive
resulting in local flow acceleration in order to prevent boundary transition duct geometries with larger radial offsets as suggested,
layer separation in this most critical area of the diffuser (compare for example, by Lord et al. [60]. Flow separation can be sup-
also to the findings regarding the strutted duct in the work of pressed by active or passive flow control measures. The first type
Norris et al. [16,17]). The shape optimization for reducing the total can either be the energization of the boundary layer by injection
pressure loss might influence the inflow situation to a downstream of high energy fluid or the removal of low energy fluid from the
component like the LP turbine. This should be the topic of a critical wall regions (see Fig. 24).
separate study. Due to the high temperature level within an intermediate
The next logical step is to apply non-axisymmetric shape opti- turbine diffuser the extraction of hot gas from the flow may be
mization techniques for strutted intermediate turbine ducts. In challenging. In spite of that, a patent held by General Electric, Kirtley
turbine design endwall contouring is used as a tool for reduction and Graziosi [61], suggests the design of a self-aspirating high-area-
of secondary flow effects in turbine stages and hence loss. In ratio inter-turbine duct suitable for high-bypass ratio engines, where
Wallin and Eriksson [58] non-axisymmetric endwall contouring the low energy fluid is removed from the critical areas at the casing
has been applied at the diffuser hub endwall in order to investi- surface through boundary layer suction ports formed by a porous
gate the effects on secondary flows, distortion, local separations material and led via bypass channels to blowing ports on the suction
and associated losses. The same parameterization with perturba- side of the following LP turbine vanes. The natural static pressure
tions to a baseline design like in [56] was applied to maintain difference results in a self-aspirating assembly. This invention will
an axisymmetric inlet and outlet. The non-axisymmetry was combine two positive effects: The first one is to prevent the bound-
obtained by adding circumferential cosine and sine functions in ary layer from separation along the outer transition duct surface
the strut area of interest. Again a DOE approach of face-centered leading to the ability to shorten the length or to increase the dia-
composite design (FCCD) has been adopted. For more details and meter of a LP turbine. The second one is that blowing on the suction
the used CFD code see [58]. surface of the LP vanes will also prevent separation there. Higher
As baseline design a duct corresponding to the Chalmers aerodynamic loads for these vanes are feasible thus leading to a
Large-Scale Facility main parameters was used together with 16 reduced airfoil count and consequently to a reduced overall weight
lifting struts. The purpose of such a layout will be discussed in the of an aero engine [61].
chapter ‘‘Integrated concepts’’. In order to minimize the up- and Inter turbine diffusers with different types of blowing at the
downstream effects of the re-design optimization onto the tur- first part of the outer wall are the subject of two patents held by
bine stages, the same distances were used between the strut General Electric, namely Graziosi and Kirtley [62] as well as
leading and trailing edges to the duct inlet and exit, respectively. Widenhoefer et al. [63]. In both inventions secondary air is
Especially the downstream LP turbine must be prevented from injected to energize the boundary layer in order to prevent it
unwanted outflow changes. A more aggressive diffuser design from separation. The air will be taken from an upstream part of
was realized by shortening the strut chord length by 10% leading the gas turbine, possibly from the compressor. Due to a suitable
to an overall axial length reduction of 6% compared to the static pressure ratio between the suction port and the injection
baseline design. This results in a more loaded strut for the same slot the fluid flow will need no additional power supply by a
amount of turning. pump. The idea of [63] is to use several slot openings arranged
The design target was to reduce loss and to improve outlet flow along the outer contour rather than one large slot far upstream
quality; therefore a constraint on the outlet swirl angle was like in [62]. This multi-slot arrangement should require less
imposed, too. All details of the optimization process are presented blowing momentum to prevent separation and thus saving mass
in [58], here only the major results will be described. The optimiza- flow bypassing the combustor or the high pressure turbine.
tion resulted in a non-axisymmetric redesigned turbine duct for The injection of high-energy fluid by means of wall jets in
which the outer wall contour remained unchanged. The inner order to eliminate flow separation within aggressive intermediate
surface was modified by adding an optimized perturbation function turbine diffusers is demonstrated by Florea et al. [64]. It was
with six parameters. The hub modification procedure illustrated by shown that flow control alone might not be sufficient to eliminate
radius contours is shown in Fig. 23. Blue represents a radius flow separation and to obtain the ideal flow quality for the
decrease and red a radius increase. The turning strut profiles are following turbine stage. Therefore a novel methodology for
depicted in black. A comparison of optimized and baseline diffuser designing flow-control-enabled turbine transition ducts coupled
showed a reduction in loss coefficient from 0.237 to 0.213, which with shape optimization was presented. A numerical procedure
means a reduction by 10%. The outlet flow angle distribution of the with a design of experiments approach (DOE) was used in order to
optimized duct was less distorted but showed the same flow eliminate flow separation and to maximize system performance.
features [58]. Again the absolute loss values might be too small An LP turbine model based on performance maps for different
due to the missing unsteady effects. inlet radii and flow conditions was included in the optimization
Additionally, an optimization of the same configuration, but procedure to capture the system performance effects correspond-
only with an axisymmetric parameterization was performed. For ing to duct shape, losses and weight savings. Eleven parameters
this case it was found that the optimization had problems to for a given duct length, exit radius and span are used in order to
achieve low loss at maintaining the same flow turning. Finally it define the channel shape as well as the position of the two
can be concluded that non-axisymmetric contouring can reduce injection sites. A wall-shear-stress-driven criterion was formu-
the overall losses and suppress the strength of secondary flows lated to specify these locations. The transition duct results from
together with shortening the axial diffuser length. A validation of the numerical simulation of 430 candidate designs were coupled
this conclusion is planned by means of the experimental inves- to the LP turbine model through transfer functions for the system-
tigation of such a turbine diffuser in a two-shaft test rig within level analyses. Two global objective functions were considered,
270 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

Fig. 23. Baseline and shortened duct (top), hub modification procedure (mid), non-axisymmetric optimized hub (bottom), blue contour lines represent a radius decrease
and red a radius increase (by courtesy of Wallin). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

namely range and total fuel burn. More details of the optimization also shown that range and fuel burn benefits depend on the radial
procedure and the flow solver are given in [64]. offset of the intermediate turbine diffuser. A range benefit of
It was demonstrated that the first jet delays separation and the approximately 0.6% due to weight reductions through a smaller
second one prevents separation completely. The optimum wall LP turbine stage count or a fuel burn decrease of 0.4% for a fixed
curvature allows the first jet to penetrate the boundary layer and mission profile together with an optimized duct/LP turbine design
impinge on the core flow leading to a relatively uniform total resulting in an increased radial offset was confirmed by Florea
pressure profile at the duct exit. As a result of this study it was et al. [64]. There an injection mass flow of approximately 3% of
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 271

Fig. 25. Vane-type vortex generator geometry and main parameters.

perpendicular to the wall but inclined to the flow direction, according


to Fig. 25. A vortex is formed behind these small vanes as the flow
has to pass the tip region from the pressure to the suction side.
Dependent on the arrangement of the vortex generators co-rotating
or counter-rotating streamwise vortices can be produced. Of course
these devices will also generate additional drag and thus total pre-
ssure loss. For that reason the extension of low-profile vortex
generators into the flow is only 10–50% of the boundary layer
thickness to minimize their losses. These devices can still transfer
sufficient momentum wallward over a distance several times their
height for effective flow-separation control [65].
Fig. 24. Flow control mechanism: active (a and b) and passive (c) measures for Regarding numerical simulations it is difficult to model these
re-energization of boundary layer.
small structures with an appropriate mesh because it would need
a high number of nodes and therefore lead to time consuming
the core mass flow per slot was assumed. It is implied that for CFD runs. For optimization procedures, e.g. for finding the most
satisfying designs the flow control authority can be further suitable vortex generator configuration from many candidate
reduced and minimized together with unsteady flow actuation. designs, this would not be practicable.
Eventually this might lead to a closed-loop control, which is also A reliable vortex generator model could shorten the computa-
difficult due to the requirement of sensing and actuating in the tional time significantly by reducing grid size and complexity.
still hot environment downstream of the HP turbine. Therefore a new tuning-free body-force model has been devel-
In contrast to active flow control the application of passive flow oped and validated for a low-profile vortex generator installation
control is less complex because there is no need for handling addi- in an intermediate turbine diffuser by Wallin and Eriksson [66].
tional fluid streams possibly at unsteady flow rates. The installation They demonstrated that a realistic result could be obtained at a
of fixed components in order to reenergize the boundary layer in grid size reduced by approximately 80% by applying a body force
critical areas would be very beneficial (see Fig. 24). For that reason to the Navier–Stokes equations to replace the actual geometry of
one work package of the EU project AIDA [9] was installed to the vortex generator. This body force will reproduce the same net
evaluate the application of passive flow control devices in compres- effect onto the flow like a vortex generator. The new model was
sor as well as in turbine transition ducts. Numerical and experi- implemented into a CFD code for compressible flows [55] using
mental investigations have been conducted in order to assess their both a high-Reynolds number (HRN) k–e turbulence model with
performance. On the turbine side low-profile vortex generators have standard wall-functions and a low-Reynolds number (LRN) clo-
been designated as the most promising candidate for going into a rig sure with well resolved boundary layers. The application of the
test in an engine representative turbine transition duct environ- new model works as follows: Firstly, the procedure will specify
ment. As test vehicle for their application the super-aggressive the location of the vortex generator installation and identify the
intermediate turbine diffuser setup AIDA C5 of the TTTF at Graz adjacent grid cells. Then the flow condition at each location is
University of Technology (see Fig. 8g) has been chosen. As already obtained by linear interpolation in circumferential direction from
discussed this transition duct is fully separated on the casing wall the two cells closest to this position. After that the local force is
and therefore suitable for the application of these passive flow calculated and finally distributed back to these cells. The magni-
control devices in order to show possible improvements by pre- tude of the body force added to the adjacent cells is defined by
venting or reducing the separated flow region. their relative position to the vortex generator mean surface. For
A thorough review of low-profile vortex generator applications validation purposes CFD simulations with fully resolved grids and
and their ability to prevent flow separation is given in Lin [65]. The both turbulence models have been performed, too. The distance of
working principle is to transport high momentum fluid from the core the vortex generator trailing edge to the area in a clean duct
flow into the boundary layer by means of streamwise vortices. The where rapid boundary layer growth starts has been chosen to be
same mechanism was observed above in the previous discussion of  15  hVG, where hVG ¼4 mm represented the vortex generator
the positive effect of the tip leakage vortices but in a much larger height, which was 35% of the boundary layer thickness at the
scale. There are many types of possible geometries. A simple one is vortex generation position. The installation consisted of 120 pairs of
the vane-type vortex generator, which consists of thin flat plates counter-rotating vortex generators all around the circumference of
272 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

the casing wall. For more details see [66]. The results gained by the geometries with low-profile vanes similar to those defined in the
vortex generator model showed good agreement with the fully work of Canepa et al. [69] together with an angle of attack of 251
resolved analysis using both the HRN and LRN turbulence approaches were used. Within the cooperation of AIDA Wallin performed a
although there were some expected differences between the HRN variety of steady CFD simulations of the duct and the LP vane row
and LRN closures in loss predictions. This newly developed vortex but without the upstream turbine stage in order to the find the
generator model should also be applicable for unsteady simulations optimal vortex generator position and height. The body-force
of transient flows. model [66] and circumferentially averaged measurement results
The vortex generator model of [66] has been adopted for the from [48] were used as inlet boundary conditions. The objective
investigation of a large number of different configurations within was to minimize the overall total pressure loss through the setup.
a design of experiments (DOE) based response surface methodol- The optimum was found for a counter-rotating vortex generator
ogy CFD optimization procedure for a flow-controlled intermedi- installation with a height of 0.7 mm and a length of 5 mm leading
ate turbine diffuser by Wallin and Eriksson [67]. There the vortex to 240 vane pairs around the circumference at approximately one
generator devices were optimized with respect to their position, vortex generator length upstream of the separation line of the
height, length and angle of attack but it was not within the scope clean duct. Fig. 26 shows a comparison between the baseline duct
of the work to present the complete design process for vortex (no vortex generator) and the optimal vortex generator arrange-
generator controlled intermediate turbine diffusers. A very ment out of Wallin’s pre-test simulations. Besides the contours of
aggressive and separating duct design for the large scale facility stagnation pressure (low pressure in blue, high pressure in red) an
at Chalmers University with an area ratio AR of 1.62 and an L/hin iso-surface of negative axial velocity (in purple) is shown for each
of 2.56 was chosen, too, as a test case for this numerical study. configuration. A reduced separation zone is clearly visible for
In the Sovran and Klomp diagram of Fig. 2 this configuration is the setup with the passive flow control devices. Because of the
clearly located in the aggressive region. The boundary conditions reattachment of the flow of the baseline configuration upstream
for the simulation were circumferentially averaged measured of the LP vane the changes in the vane passages due to reduced
flow parameters taken from the investigations of Axelsson et al.
[29]. The same CFD code like in the authors’ previous works was
used [55]. The computational domain was a 51 sector of the
diffuser including one downstream LP vane and a prolonged
channel of 0.5  L with atmospheric pressure at the end. The
vortex generator configurations were modeled with the intro-
duced body-force model [66] and with a circumferential node
spacing of 0.0721.
For the optimization procedure four parameters with influence on
the vortex generator performance were allowed to vary within lower
and upper bounds: (1) a non-dimensional location of the trailing
edges relative to the baseline separation line (DsVG/hVG ¼0–11),
(2) height (hVG ¼1.3–2.9 mm, the minimum height corresponds with
the boundary layer thickness, due to manufacturing restrictions),
(3) non-dimensional length (LVG/hVG ¼2.5–5.5  hVG) and (4) angle of
attack (aVG ¼ 10–261). The target was to find a design with a total
pressure loss as low as possible. Details of the optimization are given
in the paper [67]. The procedure resulted in a optimum design para-
meter settings of: hVG ¼ 1.9 mm, DsVG/hVG ¼0, LVG/hVG ¼4, aVG ¼51.
It was shown that the loss obtained was very sensitive to the angle of
attack and that the trailing edge of the vortex generator should be
placed just upstream of the baseline detachment line, while the
height and the length seemed to be of lower importance [67].
Furthermore, the loss coefficient decreased from 0.232 for the base-
line duct to 0.198 for the optimum configuration which means a
decrease of 15%. The largest flow differences were found in the upper
70% of the channel height at outlet. Also a positive influence onto the
loss development in the LP vane passage could be observed. In
comparison with former experiments where the optimal angle of
attack is around 251, here the value for the optimum design was
found to be only 51 which is rather small.
The application and the performance of a low-profile vortex
generator installation within a super-aggressive (very high diffu-
sion) intermediate turbine diffuser have been investigated experi-
mentally by Santner et al. [68] within the AIDA project. As already
mentioned the test setup for these passive flow control devices
was the separating AIDA C5 duct downstream of a transonic test
turbine stage within the TTTF at Graz University of Technology
(see Fig. 8g). The basic flow evolution through this setup has
already been discussed and is also given in Göttlich et al. [48]. The
main parameters of the installation, which have a high influence
on the efficiency of these low-profile vortex generators, are the Fig. 26. Steady CFD simulation of AIDA C5 duct without (top) and with (bottom)
angle of attack and the position relative to the separation accord- vortex generators (by courtesy of Wallin). (For interpretation of the references to
ing to Wallin and Eriksson [67]. In this test vortex generator color in this figure legend, the reader is referred to the web version of this article.)
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 273

separation are small. The overall loss from the inlet to the exit of altered but producible geometry is also able to reduce or even
the CFD domain did not change. This means that the additional suppress flow separation. The two-dimensional (2D) rectangular
drag generated by the vortex generators is compensated by the S-shaped wind tunnel model possesses the same curvature of the
flow improvements due to reduced separation. After these CFD outer duct contour and diffusion rate and a similar Mach number
optimization process the next task was to realize and investigate level as in the annular duct of the turbine test rig but with steady
this test configuration in the turbine environment. inflow conditions. For that reason the duct contour at the casing
For reasons of costs and time a simple method for the has been retained but a change of the inner duct shape was
installation of the vortex generators on the existing duct geome- necessary to achieve an area distribution and diffusion rate (due
try was needed. Therefore, they were manufactured by stamping to the geometry), respectively, such as in the annular duct. The
their shape into a band of copper foil with a thickness of 0.1 mm. rectangular inlet section of the wind tunnel was reduced to a
The foils were glued on the surface, which allowed an easy height of 46.13 mm, which corresponds to the minimum section
application and removal. On the upper right of Fig. 27 a picture in the test turbine at the duct inlet. The width of the flow channel
of the manufactured vortex generators can be seen. A drawback of (100 mm) was given by the test rig. A comparison of both
this production process is that the small vanes have a more geometries is given in Fig. 27 in the middle. To achieve similar
trapezoid cross section instead of parallel walls and the edges are flow conditions like in the annular duct the Mach number of 0.6 at
rounded and not sharp as the suggested design (see Fig. 27 top). the inlet was kept the same. In order to reproduce also the radial
Therefore the decision was taken to pre-test these foils in a pressure distribution and the blockage effect generated by the
rectangular 2D duct with steady inflow and without secondary downstream LP-vanes at the duct exit, three simple straight
effects emanating from a preceding rotor to find out if this slightly profiles were placed at the equivalent axial position. Details of

Fig. 27. Vortex generator geometry (top), installation in 2D and annular duct (mid), oil flow visualization at outer wall contour in 2D duct (bottom left) and annular duct
(bottom right).
274 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

the vortex generators, the test rigs and the instrumentation can row and are consequently less pronounced due to mixing and
be found in [68]. decay.
The results of the wind tunnel test showed a distinct positive It can be concluded that the different types of flow control will
influence of the flow control devices under the steady flow become an enabling factor for the design of aggressive inter-
conditions there. As shown in Fig. 27 at the bottom left the sepa- mediate turbine diffusers for future aero engines. Flow control
ration onset was shifted downstream by approximately 10 mm has to be integrated into the optimization procedures during the
(marked with D) due to the vortex generators. Additionally, a design process. For reliable results it will be necessary to use
much smaller separation zone than in the configuration without time-accurate calculations to capture the real interaction
vortex generators was found. A video of the oil flow visualization mechanisms between unsteady main flow and flow control and
can be found in the Electronic Annex 1 of the online version of thus their real influence.
this article. But in the annular diffuser with the highly unsteady
inflow provided by the upstream turbine stage the surface flow 4.3. Integrated concepts
visualization showed a similar flow behavior such as the baseline
duct without vortex generators (Fig. 27 bottom right). The surface
Integrated concepts are novel ideas to integrate the function of
flow without vortex generators has already been discussed in
several components. In strutted intermediate turbine diffusers with
detail in [48] and is depicted in Fig. 17. Again, the video of this
load-carrying aerodynamically unloaded (non-turning) vanes
flow visualization is given in the Electronic Annex 1 of the online
usually a conventional LP vane row will follow to provide a suitable
version of this article. The only difference to the clean duct is a
inflow for the LP rotor. Therefore it would make sense to avoid the
slightly better behavior of the flow along the inner duct contour,
vanes by designing these struts as loaded airfoils in order to provide
which is very close to separation. The flow is less deflected into
flow conditions similar to those downstream of the LP vane row. An
the circumferential direction for the case without vortex gen-
intermediate diffuser setup without a conventional LP vane row will
erators. The measurements of Santner et al. [68] demonstrated
save axial duct length and thus weight. This concept can be seen in
that the loss through the intermediate turbine diffuser with the
Fig. 28a and b. There the difference in axial length is marked with D.
vortex generators installed is higher than without it. This is not
For instance, a configuration tested within the EU project AIDA
only a result of the additional losses due to the blockage effect of
should replace the first LP vane row of a front bearing engine
the vortex generators but also because of an even larger separa-
tion zone after installing these devices. The already poor diffuser
efficiency ZD ¼12.7% of the baseline configuration was further
reduced to 4.3%.
The main contributor to the inefficiency of the vortex gen-
erators has to be an unsteady effect such as the wakes or the
shocks emanating from the HP-rotor or secondary flows. This is
supported by the numerical investigations at Chalmers University
for finding the applied geometry using a mixing plane approach
or circumferentially averaged measurement results at the duct
inlet thus neglecting unsteady effects. This means that in order to
be able to predict the behavior and influence of vortex generators
accurately it is necessary to perform time-dependent numerical
simulations considering all these unsteady interaction effects.
This highlights the increased importance of unsteady CFD in the
design process of future aero engines.
It is concluded that the rotating wakes of the rotor blades are
responsible for the bad performance of these passive flow con-
trols in a real engine environment. The fluctuating slip velocity
component due to the velocity deficit in the wake relative frame
of reference prevents the steady formation of the streamwise
vortices at the flow control devices. In contrast to that the rotor
trailing edge shocks seem to be a second-order effect due to their
weakening at the outer duct surface [47]. The increase in loss can
be explained by the fact that the vortex generators form an
obstruction to the tip leakage vortices from the moving rotor blades.
Thus their positive effect onto the diffuser flow as discussed
in [45–47] is reduced and as a result the diffuser efficiency becomes
even worse. It can be concluded that the low-profile vortex gene-
rators investigated in [68] at the applied axial position within the
annular duct are not effective in stabilizing the flow through a three-
dimensional intermediate turbine duct with a highly unsteady three-
dimensional flow field although they showed promising results in the
2D S-shaped duct [68].
In compressor ducts where separation takes place at the inner
wall the application of vortex generators will be more promising
due to the weakening of the wakes at lower radii and the
occurrence of the tip leakage flow located on the opposite casing
surface. The shocks on a transonic compressor blading are posi- Fig. 28. Comparison between strutted diffuser (a), integrated concept or turning
tioned far away from the trailing edges; also all flow effects from mid turbine frame (b) and turning strutted diffuser concept with embedded LP
a compressor rotor have to convect through the stationary vane vanes (c).
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 275

architecture where the vane needs a big sectional area to hold distributions downstream of the test arrangement. The larger
bearing services. The rotor is located directly downstream of the spacing of the turning struts compared to the much thinner LP
struts, so that they have to provide the downstream blade row vanes evokes more pronounced secondary flows (vortices marked
with undisturbed inflow of suitable flow angle and Mach number. with dashed arrows in Fig. 29). In the plot of the base design and
Therefore these lifting struts have a distinct three-dimensional the integrated concept a thickening of the wake close to the
design in the aft blade region while in the front region they have casing was observed. The upper passage vortex (U.P.V.) and the
to be cylindrical to be able to lead the supply lines through. To lower passage vortex (L.P.V.) occupy nearly the whole flow
apply this concept a compromise must be found between the channel downstream of the turning struts. In the base design
number of struts (weight), vibration, noise and occurring flow the passage vortices are much smaller but there are more of them
disturbances due to secondary flows and losses. within the measurement domain. A larger yaw angle variation
The AIDA C3 setup for the Transonic Test Turbine Facility over the passage height must also be taken into account. In the
(see Fig. 8h) was designed by Industria de Turbo Propulsores S.A. plot of the integrated concept beside untypical hub leakage
(ITP) as integrated concept using the same inner diffuser geome- vortices (H.L.V., due to a gap for rotating the blade row during
try from AIDA C4 together with 18 turning struts with a thickness the measurements) another vortex can be observed at the hub
to chord ratio of 18%. This configuration has been investigated by which rotates in the clockwise (CW) direction. An oil flow
Marn et al. [70] in order to compare the outflow from this wide visualization showed that this clockwise rotating vortex is the
chord vane arrangement to the AIDA C4 test setup acting as a casing horse shoe vortex traveling below the U.P.V. and pushed
baseline for this study. It was shown that a similar flow field can radially inwards by the pressure gradient. For more details on the
be realized with this integrated concept as for the base design. flow development through the setup see [70]. A significant weight
Fig. 29 shows a comparison by means of the total pressure reduction of the blading between 20% and 39% depending on the

0.35

350 SS 350 SS 0.34


PS PS
340 U.P.V. 340
U.P.V. 0.33
330 330
0.32
320 320
0.31
310 310
0.3
300 300
0.29
290 290
0.28
280 280
H.L.V.
0.27
270 270
H.L.V.
260 260 0.26

0.25
-100 -80 -60 -40 -20 0 -110 -100 -90 -80 -70 -60 -50 -40 -30 -20 -10 0
Base Design Integrated Concept IC

1.0
0.9
0.8
0.7
Passage Height

0.6
0.5
0.4
Base Design
0.3
IC
0.2
0.1
0.0
0.28 0.29 0.3 0.31 0.32 0.33 0.34 0.35
pt/pt,inlet

Fig. 29. Comparison of non-dimensional exit total pressure distributions of diffuser with LP vanes (base design) to diffuser with turning struts (integrated concept) and
resulting circumferentially mass-averaged total pressure distribution (bottom).
276 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

airfoil wall thickness was predicted by the design of a configura- turbofan engines, their meaning for the whole aero engine
tion with a strut that also fulfils the functions of a LP vane without system and some future trends. The discussion begins with a
severe loss increase [70]. review of the fundamental aerodynamic studies of such diffusers
The flow effects observed by Marn et al. [70] agree very well without upstream turbine stages. Special focus is laid onto the
with the time-averaged results in the investigated low-aspect effect of the S-shape, of swirl, of non-rotating wakes and of the
ratio vane by Miller et al. [23]. In their analysis they also studied diffusion rate leading to radial pressure gradients, an outward
the influence of the rotor tip leakage, which was found to be the transportation of boundary layer fluid together with the forma-
strongest interaction mechanism in shroudless turbines with tion of vortices and more or less strong adverse pressure
large tip gaps. In order to separate the effects a CFD simulation gradients. The aerodynamic situation of ducts equipped with
with zero rotor tip gap has been performed comparable to that load-carrying struts is also shown and the additional difficulties
carried out by Sanz et al. [47]. Miller et al. [23] showed that a drop for the designer are highlighted. The blockage and the additional
in efficiency occurred between the leading edge and 20% axial diffusion within the channel formed by two struts can enforce
chord length of the wide chord vane when the tip gap was strong separations starting in the corners between strut and
removed. A local high entropy region was found there in the casing surface. Therefore a modification of the meridional channel
corner between casing and suction surface, which coincides with shape is necessary to compensate these negative effects during
a local flow separation. This is due to the fact that the vane is the design process.
designed for a flow with tip leakage. The tip leakage flow The next topic of this article focuses on the discussion of the
suppresses the fluid movement of casing fluid onto the vane effects resulting from an upstream turbine stage. The rotating and
suction surface and the formation of the upper passage vortex. non-rotating wakes, secondary flows together with the rotor
Due to the significantly altered fluid motion in a low aspect ratio trailing edge shocks and tip leakage vortices interact with each
vane by the presence of the upstream stage it is important to other and with the boundary layers on their way through the
integrate all their downstream effects already during the design downstream diffuser. The need for modeling all these effects in a
phase. Miller et al. [23] also stated in their work that the CFD simulation is discussed in order to predict the endwall flows
secondary flows within the vane could have been reduced by and losses correctly. Steady calculations using mixing planes
locating the second bend of the duct well downstream of its exit. underpredict the mixing interactions and thus the resulting
It has to be mentioned that the test setup of Miller et al. [23] losses. The positive influence of the rotor tip leakage flow is
represents the second stage IP vane of very low aspect ratio for a shown. The transportation of high momentum fluid towards the
three-shaft aero engine. For that reason the leading edge of this outer duct surface by the streamwise vortices energizes the
aerofoil is positioned closer to the HP rotor exit compared to an boundary layer there enabling it to better withstand the adverse
integrated concept causing much stronger interactions between pressure gradient. Also the effect of swirl stabilizes the flow at the
the components. casing surface but the overall loss of a strutted duct depends
The influence of a rotor tip clearance size variation on such an strongly on flow angle variations corresponding to off-design
integrated concept discussed above has been determined by Marn operating conditions at the diffuser inlet.
et al. [71]. But only marginal changes were found between the Then the paper focuses onto measures for the enlargement of
total and static pressure, Mach number, yaw and pitch angle the design space of intermediate turbine diffusers towards more
distributions for the gap sizes of 1.5% span and 2.4% span down- aggressive layouts with shorter axial length or larger radial offset.
stream of the strut. Only the stagnation point on the strut leading Compared to a straight-walled annular diffuser S-shaped ducts
edge was shifted a little towards the pressure side for the larger tip have many more design parameters like the effective flow area
gap case due to the altered flow angle in the casing region. Here at distribution and the curvature of the flow path formed by the
the outer duct between the struts a strong deflection of the surface shape of their endwalls. The parameterization and the computa-
flow for the smaller tip gap indicated the flow to be close to tional optimization of the duct shape is the first measure
separation [71]. It seems that the positive effects of the tip leakage discussed. The result of such a numerical procedure can be an
flow are not as pronounced in a strutted duct as in a clean duct. intermediate turbine diffuser with optimized axisymmetric or
In contrast, the development of the tip leakage flow in a non-axisymmetric endwalls with either less loss production or
strutted diffuser with non turning vanes has already been with a shorter axial length. This endwall optimization also known
discussed above using the papers of Wallin et al. [37] and Arroyo as area ruling can also be performed whenever a baseline diffuser
et al. [38] and can be seen in Fig. 11. should be equipped with struts. The second measure discussed is
The optimization of such an integrated concept by means of the application of active or passive flow controls in order to
non-axisymmetric endwall shape optimization has been shown enable the duct boundary layers to withstand stronger curvatures
by Wallin and Eriksson [58] as already discussed in Section 4.1. in outward direction and thus larger adverse pressure gradients.
Fig. 23 shows the baseline and the optimized meridional geome- In the future new technologies like single dielectric barrier
try as well as the turning strut profiles. discharge plasma actuators could become important for the
To decrease the negative effect of the larger secondary flow application of separation control within intermediate turbine
structures of an integrated concept but at the same short axial diffusers (see [72,73]) besides the flow control measures dis-
extent one future solution could be to embed several thin LP cussed above. Numerical optimization procedures can also be
vanes between the fewer load-carrying turning struts. Fig. 28c applied for finding the correct position or flow control parameters
gives an idea of such a concept with embedded LP vanes. Of such as e.g. blowing rate or vortex generator height. In general, a
course the weight reduction of this layout might not be as combined optimization of the diffuser shape together with flow
beneficial although the number of thick load-carrying struts could control measures would be very beneficial. The behavior of the
possibly be further reduced. neighboring components can also be included in the optimization
procedure with the help of empirical models but the number of
design parameters influenced by the optimizer would increase
5. Conclusions and future perspectives drastically, which would exceed the possibilities of present
computational resources. The paper ends with the introduction
The purpose of this article is to review the aerodynamics of of integrated concepts, a measure to shorten an aero engine with
S-shaped intermediate turbine diffusers for high-bypass ratio a strutted intermediate turbine diffuser by integrating the
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 277

Fig. 30. Meridional section of TTTF in counter-rotating two-shaft arrangement for testing of turning mid turbine frames (TMTF).

function of the following LP vane into the load carrying airfoils in test setups together with up- and downstream components like in
the duct. the work of Antonov et al. [76]. The latest progress in this respect
All the topics discussed can help to improve the overall can be seen in Hubinka et al. [77]. An engine relevant test setup for
performance of an aero engine. The intermediate ducts and turning mid turbine frames with an up- and downstream turbine
especially the intermediate turbine diffuser will become a key stage is now available. Fig. 30 shows the adapted TTTF at Graz
component of future aero engine architectures. As mentioned University of Technology with a counter-rotating mechanically
above the further increase in bypass ratio will lead to more slowly independent two-shaft test turbine arrangement, which went suc-
rotating LP turbines requiring intermediate turbine ducts with cessfully into operation at the beginning of 2010.
larger radial offsets. On the other hand, a short intermediate duct It can be concluded that intermediate turbine diffusers are
will save weight of the whole engine and the nacelle and thus engine components, which are not as simple as they look at a first
help to save fuel. Other engine layouts suggest shifting the rear glance. For the designer it is important to understand their relevance
bearing of the LP shaft upstream of the LP turbine leading to an concerning high performance and to see them as an enabling factor
overhung arrangement; see, for example, the US patent of Seda for future aero engine concepts. It will be interesting to see which
[74]. Instead of a turbine exit casing (TEC) a mid turbine frame new engine concepts the future will bring, but intermediate turbine
(MTF) in the region of the intermediate turbine diffuser will function diffusers may play a significant role in any case.
as engine mount structure to the aircraft. The load of the two rear
bearings of both spools will be lead through thick struts to the outer
ring of this structure of the mid turbine frame. A weight reduction References
can be achieved by the smaller exit guide vanes, a shorter shaft and
a thinner LP turbine casing. Further reduction in axial length and [1] GE – Aviation History. The history of aircraft engines. /http://www.geae.
com/aboutgeae/history.html#section6S.
thus in engine weight can be realized by additionally applying the [2] Engine Alliance GP7200 Specifications. /http://www.enginealliance.com/
idea of an integrated concept. This will lead to so-called turning mid gpspec.htmlS.
turbine frames (TMTF). These structures are not only interesting for [3] Rolls Royce Trent 1000 fact sheet VCOM13797 Issue 5 March 2009. /http://
www.rolls-royce.com/Images/brochure_Trent1000_tcm92-11344.pdfS.
high-bypass ratio turbofan engines, but also be a key component for [4] Kurzke J. Fundamental differences between conventional and geared turbo-
the open rotor technology especially if a counter-rotating LP turbine fans. ASME paper GT2009-59745, 2009.
concept will be used to directly drive a counter-rotating propel- [5] Weber S, Hackenberg H-P. GP7000: MTU Aero Engine’s contribution in a
successful partnership. AIAA paper ISABE-2007-1283, 2007.
ler [59]; also for the geared turbofan concept turning mid turbine [6] Gardner WB. Energy efficient engine (E3) technology status. AIAA paper
frames could be of interest. In addition the aerodynamics of AIAA-82-1052, 1982.
S-shaped turbine ducts might be of great relevance for the design [7] Bailey DW, Britchford KM, Carrote JF, Stevens SJ. Performance assessment of
an annular S-shaped duct. ASME J Turbomach 1997;119:149–56.
of IP compressor exit S-duct in intercooled aero gas turbine engines, [8] Sovran G, Klomp ED. Experimentally determined optimum geometries for
but there additionally a transition from an annular to a branched duct rectilinear diffusers with rectangular conical or annular cross section. In:
connected to separate intercooler modules has to be realized [75]. Sovran G, editor. Fluid mechanics of internal flow. Elsevier; 1967. p. 270–319.
[9] EU Transport Research: Aggressive Intermediate Duct Aerodynamics for
On the experimental side the testing of intermediate turbine
Competitive and Environmentally Friendly Jet Engines. /http://ec.europa.
diffusers downstream of HP turbine stages is state-of-the-art as eu/research/transport/projects/article_3683_en.htmlS.
discussed in the paper. This is necessary to capture all effects [10] EU Transport Research: Aerothermal Investigations on Turbine End-walls and
occurring in a real aero engine. The test data is usually used to Blades II. /http://ec.europa.eu/research/transport/projects/article_3671_en.
htmlS.
validate new designs and, furthermore, to calibrate the numerical [11] Dominy RG, Kirkham DA. The influence of blade wakes on the performance of
tools for the engine design process. The next step is to test diffuser inter-turbine diffusers. ASME J Turbomach 1996;118:347–52.
278 E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279

[12] Dominy RG, Kirkham DA. The influence of swirl on the performance of inter- [43] Göttlich E, Malzacher FJ, Heitmeir FJ, Marn A. Adaptation of a transonic test
turbine diffusers. In: VDI Berichte 1186, 1995, p. 107–22. turbine facility for experimental investigation of aggressive intermediate
[13] Bradshaw P. Effects of streamline curvature on turbulent flow. AGARD turbine duct flows. AIAA paper ISABE-2005-1132, 2005.
AG-169, 1973. [44] Göttlich E, Marn A, Malzacher FJ, Schennach O, Heitmeir F. Experimental
[14] Dominy RG, Kirkham DA, Smith AD. Flow development through inter-turbine investigation of the flow through an aggressive intermediate turbine duct
diffusers. ASME J Turbomach 1998;120:298–304. downstream of a transonic turbine stage. In: Papailiou K, Martelli F, Manna
[15] Norris G, Dominy RG. Diffusion rate influences in inter-turbine diffusers. M, editors. Proceedings 7th European conference on turbomachinery fluid
IMechE J Power Energy 1997;211(Part A):235–42. dynamics and thermodynamics, 2007, p. 383–94.
[16] Norris G, Dominy RG, Smith AD. Strut influences within a diffusing annular [45] Marn A, Göttlich E, Pecnik R, Malzacher FJ, Schennach O, Pirker HP. The
S-shaped duct. ASME paper 98-GT-425, 1998. influence of blade tip gap variation on the flow through an aggressive
[17] Norris G, Dominy RG, Smith AD. Flow instability within a diffusing, annular S-shaped intermediate turbine duct downstream a transonic turbine stage—
S-shaped duct. ASME paper 99-GT-70, 1999. Part I: time-averaged results. ASME paper GT2007-27405, 2007.
[18] Ainsworth RW, Schultz DL, Davies MRD, Forth CJP, Hilditch MA, Oldfield MLG. [46] Göttlich E, Marn A, Pecnik R, Malzacher FJ, Schennach O, Pirker HP. The
A transient flow facility for the study of the thermofluid-dynamics of a full influence of blade tip gap variation on the flow through an aggressive
stage turbine under engine representative conditions. ASME paper 88-GT- S-shaped intermediate turbine duct downstream a transonic turbine
144, 1988. stage—Part II: time-resolved results and surface flow. ASME paper GT2007-
[19] Miller RJ, Moss RW, Ainsworth RW, Harvey NW. The development of turbine 28069, 2007.
exit flow in a swan-necked inter-stage diffuser. ASME paper GT2003-38174, [47] Sanz W, Kelterer ME, Pecnik R, Marn A, Göttlich E. Numerical investigation of
2003. the effect of tip leakage flow on an aggressive S-shaped intermediate turbine
[20] Miller RJ, Moss RW, Ainsworth RW, Horwood CK. Time-resolved vane-rotor duct. ASME paper GT2009-59535, 2009.
interaction in a high-pressure turbine stage. ASME J Turbomach 2003;125: [48] Göttlich E, Marn A, Malzacher FJ, Heitmeir F. On flow separation in a super-
1–13. aggressive intermediate turbine duct. In: Heitmeir F, Martelli F, Manna M,
[21] Miller RJ, Moss RW, Ainsworth RW, Harvey NW. Wake, shock, and potential editors. Proceedings 8th European conference on turbomachinery fluid
field interactions in a 1.5 stage turbine-Part I: vane–rotor and rotor–vane dynamics and thermodynamics, 2009, p. 1389–404.
interaction. ASME J Turbomach 2003;125:33–9. [49] Marn A, Göttlich E, Malzacher FJ, Pirker HP. The effect of rotor tip clearance
[22] Miller RJ, Moss RW, Ainsworth RW, Harvey NW. Wake, shock, and potential size onto the separated flow through a super-aggressive S-shaped inter-
field interactions in a 1.5 stage turbine-Part II: vane–vane interaction and mediate turbine duct downstream of a transonic test turbine stage. ASME
discussion of results. ASME J Turbomach 2003;125:40–7. paper GT2009-59934, 2009.
[23] Miller RJ, Moss RW, Ainsworth RW, Harvey NW. The effect of an upstream [50] Marn A, Göttlich E, Malzacher FJ, Heitmeir F, Santner C. Comparison between
turbine on a low-aspect ratio vane. ASME paper GT2004-54017, 2004. the flow through an aggressive and a super-aggressive intermediate turbine
[24] Dunn MG, Moller JC, Steele RC. Operating point verification for a large shock duct. AIAA paper ISABE-2009-1259, 2009.
tunnel test facility. Wright Research and Development Center technical [51] Pecnik R, Pieringer P, Sanz W. Numerical investigation of the secondary flow
report WRDC-TR-2027, 1989. of a transonic turbine stage using various turbulence closures. ASME paper
[25] Clark JP, Stetson GM, Magge SS, Ni RH, Haldeman CW, Dunn MG. The effect of GT2005-68754, 2005.
airfoil scaling on the predicted unsteady loading on the blade of a 1 and 1/2 [52] Post Processing—Data Visualization with VIEWER. /http://ttm.tugraz.at/
stage transonic turbine and a comparison with experimental results. ASME ?seite=facilities/numerical/DataVisS.
paper 2000-GT-0446, 2000. [53] Schennach O, Pecnik R, Paradiso B, Göttlich E, Marn A, Woisetschläger J. The
[26] Davis RL, Yao J, Clark JP, Stetson G, Alonso JJ, Jameson A, Haldeman CW, effect of vane clocking on the unsteady flow field in a one-and-a-half stage
Dunn MG. Unsteady Interaction between a transonic turbine stage and transonic turbine. ASME J Turbomach 2008;130:031022 (8 pp).
downstream components. ASME paper GT-2002-30364, 2002. [54] Gräsel J, Pierré M, Demolis J. Parametric interturbine duct design and
[27] Haldeman CW, Krumanaker ML, Dunn MG. Influence of clocking and vane/ optimisation. In: Proceedings of 25th international congress of the aero-
blade spacing on the unsteady surface pressure loading for a modern stage nautical sciences, ICAS, 2006.
and one-half transonic turbine. ASME J Turbomach 2003;125:743–53. [55] Eriksson L-E. Development and validation of highly modular flow solver
[28] Arroyo Osso C, Axelsson L-U, Håll U, Johansson TG, Larsson J, Haselbach F. versions in G2DFLOW and G3DFLOW. Internal report 9970-1162. Sweden:
Large-scale low speed facility for investigating intermediate turbine duct Volvo Aero Corporation; 1995.
flows. AIAA paper AIAA-2006-1312, 2006. [56] Wallin F, Eriksson L-E. Response surface-based transition duct shape opti-
[29] Axelsson L-U, Arroyo Osso C, Cadrecha D, Johansson TG. Design, performance mization. ASME paper GT2006-90978, 2006.
evaluation and endwall flow structure investigation of an S-shaped inter- [57] Wallin F. Flow control and shape optimization of intermediate turbine
mediate turbine duct. ASME paper GT2007-27650, 2007. ducts for turbofan engines. Doctoral thesis, Department of Applied
[30] Axelsson L-U, Johansson TG. Experimental investigation of the time-averaged Mechanics of Chalmers University of Technology, 2008. ISBN:978-91-
flow in an intermediate turbine duct. ASME paper GT2008-50829, 2008. 7385-205-0.
[31] Axelsson L-U. Experimental investigation of the flow field in an aggressive [58] Wallin F, Eriksson L-E. Non-axisymmetric endwall shape optimization of an
intermediate turbine duct. Doctoral thesis, Department of Applied Mechanics intermediate turbine duct. AIAA paper ISABE-2007-1300, 2007.
of Chalmers University of Technology, 2009. ISBN:978-91-7385-264-7. [59] The alternative approach to a cleaner future. Dream—Validation of Radical
[32] Axelsson L-U, George WK. Spectral analysis of the flow in an intermediate Engine Architecture Systems. /http://www.dream-project.euS.
turbine duct. ASME paper GT2008-51340, 2008. [60] Lord WK, MacMartin DG, Tillman G. Flow control opportunities in gas turbine
[33] Axelsson L-U, George WK, Johansson TG. Fourier decomposed turbulence engines. AIAA paper AIAA-2000-2234, 2000.
measurements downstream of a high-pressure turbine stage. ASME paper [61] Kirtley KR, Graziosi P. Self-aspirating high-area-ratio inter-turbine duct
FEDSM2008-55146, 2008. assembly for use in a gas turbine engine. United States patent 6851264,
[34] Axelsson L-U, Johansson TG. Evaluation of the flow in an intermediate 2005. /http://www.freepatentsonline.com/6851264.htmlS.
turbine duct at off-design conditions. In: Proceedings of 26th international [62] Graziosi P, Kirtley KR. High-area-ratio inter-turbine duct with inlet blowing.
congress of the aeronautical sciences, ICAS, 2008. United States patent 7137245 B2, 2006. /http://www.freepatentsonline.
[35] Johansson PBV, Axelsson L-U. Numerical and experimental analysis of the com/7137245.htmlS.
flow in an aggressive intermediate turbine duct. ASME paper GT2009-59299, [63] Widenhoefer JF, Graziosi P, Kirtley KR. Multi-slot inter-turbine duct assembly
2009. for use in a turbine engine. United States patent 7549282 B2, 2009. /http://
[36] Wallin F, Eriksson L-E, Nilsson M. Intermediate turbine duct design and www.freepatentsonline.com/7549282.htmlS.
optimization. In: Proceedings of 25th international congress of the aero- [64] Florea R, Bertuccioli L, Tillman TG. Flow-control-enabled aggressive turbine
nautical sciences, ICAS. 2006. transition ducts and engine systems analysis. AIAA J Propul Power 2007;23:
[37] Wallin F, Arroyo Osso C, Johansson TG. Experimental and numerical inves- 797–803.
tigation of an aggressive intermediate turbine duct: Part 1—flowfield at [65] Lin JC. Review of research on low-profile vortex generators to control
design inlet conditions. AIAA paper AIAA-2008-7055, 2008. boundary-layer separation. Prog. Aerosp Sci. 2002;38:389–420.
[38] Arroyo Osso C, Wallin F, Johansson TG. Experimental and numerical inves- [66] Wallin F, Eriksson L-E. A tuning-free body-force vortex generator model.
tigation of an aggressive intermediate turbine duct: Part 2—flowfield at off- AIAA paper AIAA-2006-0873, 2006.
design conditions. AIAA paper AIAA-2008-7055, 2008. [67] Wallin F, Eriksson L-E. Design of an aggressive flow-controlled turbine duct.
[39] Arroyo Osso C. Aerothermal investigation of an intermediate turbine duct. ASME paper GT2008-51202, 2008.
Doctoral thesis, Department of Applied Mechanics of Chalmers University of [68] Santner C, Göttlich E, Marn A, Hubinka J, Paradiso B. The application of low-
Technology, 2009. ISBN:978-91-7385-351-4. profile vortex generators in an intermediate turbine diffuser. ASME paper
[40] Wallin F. Numerical aerothermal analysis of an aggressive intermediate GT2010-22892, 2010.
turbine duct. In: Heitmeir F, Martelli F, Manna M, editors. Proceedings 8th [69] Canepa E, Lengani D, Satta F, Spano E, Ubaldi M, Zunino P. Boundary layer
European conference on turbomachinery fluid dynamics and thermody- separation on a flat plate with adverse pressure gradients using vortex
namics, 2009, p. 603–13. generators. ASME paper GT-2006-90809, 2006.
[41] Erhard J, Gehrer A. Design and construction of a transonic test turbine facility. [70] Marn A, Göttlich E, Cadrecha D, Pirker HP. Shorten the intermediate turbine
ASME paper 2000-GT-480, 2000. duct length by applying an integrated concept. ASME J Turbomach
[42] Neumayer F, Kulhanek G, Pirker HP, Jericha H, Seyr A, Sanz W. Operational 2009;131:041014 (10 pp).
behaviour of a complex transonic test turbine facility. ASME paper 2001-GT- [71] Marn A, Göttlich E, Leitgeb T, Heitmeir F. The influence of rotor tip clearance
489, 2001. size on the flow through an S-shaped intermediate turbine duct shortened by
E. Göttlich / Progress in Aerospace Sciences 47 (2011) 249–279 279

applying an integrated concept. In: Heitmeir F, Martelli F, Manna M, editors. [74] Seda JF. Aircraft engine with inter-turbine engine frame. United States patent
Proceedings 8th European conference on turbomachinery fluid dynamics and 6708482 B2, 2004. /http://www.freepatentsonline.com/6708482.htmlS.
thermodynamics, 2009, p. 1365–78. [75] Walker AD, Carrotte JF, Rolt AM. Duct aerodynamics for intercooled aero gas
[72] Ma R, Niewiarowski J, Corke T, Flint T, Neiswander B. Inter-turbine duct flow turbines: constraints, concepts and design methodology. ASME paper
separation control with SDBD plasma actuators: experiment. American GT2009-59612, 2010.
Physical Society, 61st annual meeting of the APS Division of Fluid Dynamics, [76] Antonov SS, Sadovnichii VN, Chernyakov VA. The operating characteristics of
2008. an intermediate diffuser between turbine stages under transient conditions.
[73] Neiswander B, Corke T, Flint T, Niewiarowski J, Ma R. Inter-turbine duct flow Therm Eng 1996;43:471–5.
separation control with SDBD plasma actuators: Simulation. American [77] Hubinka J, Santner C, Paradiso B, Malzacher F, Göttlich E, Heitmeir F. Design
Physical Society, 61st annual meeting of the APS Division of Fluid Dynamics, and construction of a two shaft test turbine for investigation of mid turbine
2008. frame flows. AIAA paper ISABE-2009-1293, 2009.

S-ar putea să vă placă și