Proceedings of ASME Turbo Expo 2010: Power for Land, Sea and Air

GT2010
June 14-18, 2010, Glasgow, UK

GT2010-23604

COMPUTATIONAL STUDY OF GEOMETRIC PARAMETER INFLUENCE ON

AGGRESSIVE INTER-TURBINE DUCT PERFORMANCE

Paul T. Couey, Craig W. McKeever, Malak F. Malak

Honeywell Aerospace
Phoenix, Arizona, USA

Balamurugan S., Raju Veeraraghava H., Dhinagaran R.

Honeywell Technology Solutions
Bangalore, India

ABSTRACT To help assess the interaction of duct slope and pitch-wise

turning with area ratio and length, an analytical Design of
Modern direct-drive turbofan engines typically have the Experiments (DOE) was run using approximately sixty
fan turbine designed at significantly higher diameter than the different duct configurations. The DOE was carried out using
gas producer turbine. Furthermore, the gas turbine industry is 3D, steady CFD analysis. The results of the DOE are presented
being pushed to shorten engine length with the goal of reducing with insights provided into how the Cp* line may shift as a
weight. This results in a need to design very aggressive inter- function of duct slope. Of particular interest is that slope by
turbine-ducts (ITD’s) that have high endwall slopes. itself does not work particularly well as a risk indicator.
The gas turbine design cycle typically begins with However, a combination of new area ratio-length and slope-
conceptual design where many engine configuration iterations length parameters was found to segregate ducts between
are made. During conceptual design, there usually is little firm separated and non-separated cases.
geometric definition or time for detailed CFD studies on
aggressive ITD’s. This can cause a large amount of risk to the
engine development schedule and cost if the space allocated for NOMENCLATURE
the ITD during conceptual design is found to be insufficient
A [m2] area
later in the design cycle. Therefore, simple analytical tools for
AR [-] area ratio
accurately assessing the risk of an ITD in conceptual design are
important. Cp [-] static pressure recovery
The gas turbine industry is familiar with the Sovran and h [m] duct height
k [m2/s2] turbulent kinetic energy
Klomp annular diffuser performance chart [1] as a conceptual
L [m] duct meridional length
design tool for assessing ITD’s. However, its applicability to
P [Pa] static pressure
modern gas turbine ducts with high endwall slope is limited.
The location of the maximum pressure recovery for a given Pt [Pa] total pressure
length, the Cp* line, considers only two geometric parameters: R [mm] mean duct radius
t [s] time, spacing
area ratio and normalized length. The chart makes no
Tt [K] total pressure
distinction of risk of flow separation regarding the level of
Tu [-] turbulence intensity
slope or the pitch-wise turning in the duct. However, intuition
y+ [-] dimensionless, sublayer-scaled distance
would suggest that a high slope duct would have more risk of
separation than an equivalent area ratio duct with low slope. V [m/s] velocity
Similarly, a duct that turns the flow from axial to radial would ε [m2/s3] turbulence dissipation rate
Θ [deg] duct mean slope
be expected to be riskier than a pure axial duct.
Φ [deg] mean-line pitch angle

1 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 μT/μ [-] turbulent viscosity ratio upstream HPT stage and an LPT vane at the exit of the duct.
ζ [-] total pressure loss coefficient The testing showed the design to be free of shroud separation.
η [-] effectiveness Axelsson et al [5] investigated experimentally the time-
averaged mean flow field and turbulence development in an
Subscripts aggressive intermediate turbine duct (downstream a rotating
1 Duct Inlet turbine stage) using a 5-hole probe and 2-component hot-wire
2 Duct Exit anemometry. The testing revealed the development of the
r radial vortical structures throughout the ITD. As the flow moves
t tangential downstream in the ITD, the shear-layer that was formed at the
x axial inlet (due to the tip gap) breaks up and forms two co-rotating
vortices. Axelsson et al [6] outlined a procedure for obtaining
Abbreviations the dissipation of turbulence kinetic energy using the measured
ANOVA Analysis of variance one-dimensional energy spectrum, and laid out criteria for
CFD Computational Fluid Dynamics when the procedure can be legitimately used.
DOE Design of Experiments
HPT High Pressure Turbine Sovran and Klomp chart
ITD Inter-Turbine Duct including original data

LDV Laser-Doppler-Velocimeter
LPT Low Pressure Turbine
SST Shear Stress Transport turbulence model 0.8

Cp*
0.7
INTRODUCTION 0
10
AR-1
0.6

The demand for aggressive ITD’s in modern 0.5

turbomachinery engines has resulted in a flurry of new research 0.4
activity throughout the industry. The well known diffuser Transition duct
0.3 typical region
studies such as those by Sovran and Klomp [1] and ESDU [2, of interest
3] are becoming less useful for conceptual design analysis of 0.2
modern ITD’s. For example, Figure 1 shows the Sovran and -1
10
Klomp annular diffuser performance chart including the test 10
0
10
1

data that was used to develop the chart. Modern ITD’s are L/h1
pushing mean slopes with levels past 30 degrees and typically
have area ratios less than 1.3, or perhaps as high as 1.5, as Figure 1. Annular Diffuser Performance Chart (Sovran &
highlighted in the figure. If one reviews the original data-set Klomp) showing region of interest for ITD’s
used by Sovran and Klomp, it is seen that only a small portion
of the tested duct configurations lies in the typical area ratio Similarly, Marn et al [7] tested a high slope duct with area
region of interest. Furthermore, the majority of the ducts have ratio ~1.5 with a transonic HPT stage upstream at two different
slopes of 20 degrees or less. Out of over 120 configurations rotor tip clearance levels, 1.5% span and 2.4% span, using five
only four have slopes of 30 degrees, and only two of those hole probes with thermocouples, boundary layer rakes, and
points lie in the region of interest. Similarly, the ESDU report static pressure tappings. The test data was compared to analysis
has limited information regarding the effect of duct slope with using time averaged results. The study concluded that a larger
charts provided for inner wall slopes no greater than 5 degrees. tip clearance has a positive influence on the flow at the outer
Therefore it is impractical, and perhaps risky, to use the Sovran wall due to the increased swirl. Göttlich et al [8] tested the
and Klomp chart, or ESDU charts for conceptual design of same ITD that was used by Marn et al [7] using Laser-Doppler-
modern high slope ducts. Velocimetry (LDV) for measurements at duct inlet directly
Significant activity in the study of aggressive ITD’s has downstream of the HPT blades to obtain unsteady information
come about recently through the AIDA (Aggressive about the inflow and to quantify the differences between the
Intermediate Duct Aerodynamics) program. The AIDA two tip gaps. Additionally, the experiment included oil-film
program has resulted in recent studies by various institutions visualization to investigate the surface flow at the outer and
including Chalmers University of Technology, Graz University inner wall of the duct. The time-resolved results were compared
of Technology, ITP (Industria de Turbopropulsores S. A.), and with a numerical simulation. This experiment revealed a clear
MTU Aero Engines with various papers being written, as influence of a blade tip clearance variation onto the duct flow.
referenced in [4, 5, 6, 7, 8, 9, 10]. Marn et al [9] used another approach to shorten the duct and
Axelsson et al [4] tested an aggressive ITD with ~1.3 area thus the engine length by applying a so called integrated
ratio and slope of ~30 degrees. The testing included an concept where the struts, mounted in the transition duct replace

2 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 the following LPT vane row. Sanz et al [10] followed up the ITD GEOMETRY
testing that was done earlier, references [7 and 8], with a
numerical investigation of the two tip clearance levels. The To facilitate the creation of a range of geometries suitable
results of this work suggested that HPT rotor tip clearance for CFD analysis, four parameters were selected to define the
improves the duct pressure recovery compared to a zero tip ducts. These included the classic parameters of area ratio, AR,
clearance case. and the duct length normalized by the inlet annular height,
In addition to the recent work by the AIDA program, (L/h) from the Sovran & Klomp correlation. The duct mean
Dominy, et al [12] studied the effect of inlet swirl on ITD slope, Θ, and the exit mean-line pitch angle, Φ2, were added to
performance. Dominy et al [13] provides details of the flow complete the specification. Figure 2 shows the relationship
structure in an ITD tested both with and without an upstream between these parameters and the inlet and exit station hub and
turbine, including the development of secondary flow resulting shroud points.
from the cross passage pressure gradients acting upon the The basis for the duct geometries was the ITD geometry
wakes from the upstream turbine. Dominy, et al [14] reveals from a Honeywell LPT research rig. All the ducts created for
that both the intensity of the wakes entering the ITD and the the current study used the same inlet station geometry as the
diffusion rate within the duct, have a significant influence on rig: the inlet pitch angle was zero (axial) and the hub/tip radius
the development of secondary flow in the duct. ratio was 0.72. The hub and shroud inlet endwall slopes were
Honeywell’s approach to the research of aggressive ITD’s zero.
is to first perform analytical CFD studies to determine the key DUCT
geometric parameters driving the performance in the aggressive EXIT
ITD’s. Once these parameters are determined, the findings are 2
used to guide the selection of geometry for an experimental
program. As new knowledge of ITD behavior is acquired, the
learning is incorporated into tools and methods for conceptual h2
design as well as higher fidelity tools and methods. DUCT Ф2
The results of the analytical studies revealed that in INLET
addition to the classic duct parameters of area ratio and 1
L
normalized length, global geometric parameters that influence ΔR
the shroud curvature distribution are important. In addition, it Θ
was also seen that the inclusion of the LPT vane, the inclusion h1
of the upstream rotor, the upstream rotor tip clearance, and in Ф1
LX R2
the case of not using the upstream rotor, the inlet boundary
conditions can significantly alter the duct behavior. Some of
these same findings regarding, for example, whether to include
the rotor tip clearance have also been discussed in recent R1
literature [7, 10]. Furthermore, even for typical exhaust Figure 2. Duct Geometry Parameters
diffusers, rotor tip clearance was found to play a role in the duct
behavior [11], with the finding that the tip clearance (and the For a given set of parameters, the duct endwall geometry
associated leakage) can improve the diffuser pressure recovery. was constructed from a mean-line and a prescribed area
The study herein focuses on analysis of ducts with distribution. The duct mean-line was constructed with a 3rd
prescribed inlet boundary conditions representing the swirl order polynomial, constrained to match the inlet and exit mean
from an upstream rotor and with an LPT vane included at the radii and slopes, as defined by the inlet station, length, and
duct exit. Studies including the upstream rotor are ongoing. slope parameters. The hub and shroud endwalls were defined
These studies show the upstream rotor to have a significant by creating normal off-sets from the mean-line sized by the
impact on the duct behavior. This is perhaps a subject of a local mean radius and an area from the distribution. The area
future paper. distribution, as described by Equation 1, was parabolic
An ITD testing program is currently ongoing at the producing most of the area increase in the aft portion of the
University of Notre Dame under the direction of Professor duct. This avoided compounding the spanwise gradients
Thomas Corke. This testing is being done in two phases. The caused by the turn outboard in the forward portion of the ducts
first phase excludes the upstream rotor, while the second phase with additional diffusion associated with area change.
includes the rotor. The intent of the testing is to verify the
findings of the various analytical studies that have been
completed, and to provide information necessary to develop
new conceptual design tools and design guidelines. (1)
The duct construction technique provided no control over
the endwall curvature, nor did it necessarily produce optimal
area distributions. However, it allowed for rapid definition of

3 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 geometries of reasonable quality – an important attribute for a chart. Since the DOE considered slope and exit pitch angle
study with a large number of cases such as this. along with the classic parameters, many of the cases are
A representative LPT vane was placed downstream of the overplotted.
duct to provide the influence of the leading edge blockage on
the duct boundary layers. Previous Honeywell numerical Case AR L/h1 Θ Φ2

studies had shown the blockage impacting the tendency for B1 1.10 1.65 27.0 27.0
B2 1.10 2.33 27.0 27.0
separation in aggressive ducts. The baseline vane featured an B3 1.30 2.33 27.0 27.0
aspect ratio based on the axial chord of 1.7, a solidity of 1.3, B4 1.30 3.00 27.0 27.0
and an incompressible Zweifel coefficient of 0.79. The vane B5 1.50 3.00 27.0 27.0
B6 1.10 1.65 27.0 32.0
was positioned behind each duct with the leading edge slightly B7 1.30 2.33 27.0 32.0
downstream of the duct exit mean-line position. The B8 1.30 3.00 27.0 32.0
positioning required the vane to be shifted in the radial B9 1.50 3.00 27.0 32.0
B10 1.10 1.65 27.0 37.0
direction and pitched to align with the exit pitch angle of the B11 1.30 1.65 27.0 37.0
particular duct. The vane was scaled to match the duct exit B12 1.30 2.33 27.0 37.0
height and the flowpath blended to make a smooth transition B13 1.30 3.00 27.0 37.0
B14 1.50 2.33 27.0 37.0
from the duct through the vane and into the downstream B15 1.50 3.00 27.0 37.0
section. Finally, the vane count was adjusted to maintain B16 1.10 1.65 31.5 31.5
solidity near the baseline value. B17 1.10 2.33 31.5 31.5
B18 1.30 2.33 31.5 31.5
The current study began examining the influence of the B19 1.30 3.00 31.5 31.5
four duct parameters on duct performance with an analytical B20 1.10 1.65 31.5 36.5
B21 1.30 2.33 31.5 36.5
Design of Experiments (DOE) using a 2-level, 4-factorial face-
B22 1.30 3.00 31.5 36.5
centered composite design yielding 25 cases. The parameter B23 1.10 1.65 31.5 41.5
ranges were set to cover the typical ITD design space, with an B24 1.30 2.33 31.5 41.5
B25 1.30 3.00 31.5 41.5
emphasis on high duct slopes. The initial DOE matrix of duct
B26 1.50 3.00 31.5 41.5
parameters is presented in Table 1. B27 1.10 1.65 36.0 36.0
B28 1.10 2.33 36.0 36.0
Case AR L/h1 Θ Φ2 B29 1.10 3.00 36.0 36.0
A1 1.30 3.00 36.0 45.0 B30 1.30 3.00 36.0 36.0
A2 1.10 3.00 36.0 45.0 B31 1.10 1.65 36.0 41.0
A3 1.30 3.00 36.0 34.0 B32 1.10 2.33 36.0 41.0
A4 1.10 3.00 36.0 34.0 B33 1.30 2.33 36.0 41.0
A5 1.30 1.65 36.0 45.0 B34 1.30 3.00 36.0 41.0
A6 1.10 1.65 36.0 45.0 B35 1.10 1.65 36.0 46.0
A7 1.30 1.65 36.0 34.0 B36 1.10 2.33 36.0 46.0
A8 1.10 1.65 36.0 34.0 B37 1.30 2.33 36.0 46.0
A9 1.30 3.00 27.0 45.0 B38 1.30 3.00 36.0 46.0
A10 1.10 3.00 27.0 45.0
A11 1.30 3.00 27.0 34.0 Table 2. Extended DOE Geometry Parameters
A12 1.10 3.00 27.0 34.0
0.7
A13 1.30 1.65 27.0 45.0 Sovran and Klomp chart
A14 1.10 1.65 27.0 45.0 with DOE cases shown
0
A15 1.30 1.65 27.0 34.0 10 0.6
A16 1.10 1.65 27.0 34.0
A17 1.20 2.33 36.0 39.5
0.5
A18 1.20 3.00 31.5 39.5
A19 1.20 2.33 31.5 45.0
0.4
A20 1.30 2.33 31.5 39.5
A21 1.20 2.33 27.0 39.5
A22 1.20 1.65 31.5 39.5 Cp* 0.3
A23 1.20 2.33 31.5 34.0
AR-1

A24 1.10 2.33 31.5 39.5

0.2
A25 1.20 2.33 31.5 39.5
-1
10 Transition duct
Table 1. Initial DOE Geometry Parameters
typical region
of interest
The results of the initial DOE prompted a follow-on study
to extend the design space to higher area ratios. The second
analytical DOE was based on a 4-factorial, 3-level design. The 0 1
10 10
81 full-factorial cases were reduced to 38 as will be discussed L/h1
in the results section. The extended DOE duct parameters are
presented in Table 2. Figure 3. DOE Cases on Sovran & Klomp Chart
To help illustrate the range of ducts used in the DOE,
Figure 3 shows the cases plotted on the Sovran and Klomp

4 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 NUMERICAL METHOD to curved. The exit rating plane was placed slightly upstream
of, and roughly parallel to, the vane leading edge. The distance
The aerodynamic performance of the duct and vane cases was approximately 2.5% of the duct inlet annulus height.
was evaluated with 3D, steady-state, viscous CFD analyses.
The simulations were performed with the ANSYS-CFX Version
11.0 tool. The SST k-ω turbulence model [15] was adopted for
these simulations. By including a blending function, the SST
model gave the advantages of both the k-ε and k-ω
formulations: the model switches between k-ω near the wall
and k-ε away from the wall. Near the wall, the SST model
enforces the Bradshaw equation [16] which predicts the shear
stress accurately. It was believed that this would lead to
improved capture of any flow separations within the diffusing
flowfield of the ducts. By contrast, k-ε models would be
expected to delay the on-set of separation or fail to capture
separation in adverse pressure gradient regions.
Simulation of the Honeywell research rig ITD using the
SST model agreed well with endwall static pressure
measurements and duct exit surveys of flow angle and total
pressure. Though the research duct did not exhibit separation,
the data-match provided some assurance that the model would
be suitable for the current study.
The domain was modeled with structured, hexahedral
grids, generated with Numeca’s AutoGrid 5. The
computational domain consisted of a periodic slice equal to a
single vane pitch. A cylindrical pre-duct section, with a length
1.4 times the inlet annular height was added upstream of the
ITD inlet to isolate the diffusion in the duct from the imposed
inlet boundary conditions. For similar reasons, the domain was Rating Planes
extended roughly a chord-length downstream of the vane
trailing edge. The default AutoGrid 5 O-H topology was used
for the stator blade-row with simple rectilinear blocks for the
annular sections up and downstream of the stator, including the
ITD. The final grids were on the order 500,000 cells with a y+
Figure 5. Typical Analytical Grid
below 13 for all cases. A representative grid is presented in
Figure 5.
For all cases, the total pressure, total temperature, and
velocity vector cosines were specified at the domain inlet, as 100 100
shown in Figure 6. These inlet boundary conditions were
Vx/|V|
extracted from survey data from the Honeywell research LPT
80 80 Vt/|V|
warm rig that used a preswirl vane to simulate the exit flow Vr/|V|
field of the upstream HPT. The domain exit average static
Tt/TtAVG
pressure was adjusted to achieve an inlet Mach number of 0.45,
Percent Span

Percent Span

60 60
the nominal value. At these conditions, the duct Reynolds Pt/PtAVG

number, based on the inlet hydraulic diameter was 4.6x105.

The inlet turbulence intensity, Tu, was specified at 1% with a 40 40

turbulent viscosity ratio, μT/μ, of 10 for the initial DOE cases –

the estimated turbulence at the turbine rig inlet. For the later
20 20
cases, the turbulence intensity was increased to 5% to be more
in line with the level expected at the exit of an HPT stage.
Inlet and exit rating planes were established for the 0 0
purpose of extracting duct performance parameters from the 0.80 0.90 1.00 1.10 -0.50 0.00 0.50 1.00
Normalized Total Parameter Normalized Velocity
solutions. The inlet rating plane was placed coincident with the
duct inlet. Because of how the geometries were constructed, Figure 6. Inlet Boundary Conditions
the inlet corresponds to the plane normal to the engine axis
where the hub and shroud endwalls transition from cylindrical

5 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 RESULTS curvature on the shroud wall, particularly near the duct inlet.
The exit pitch parameter was found to control the amount of
The results of the numerical studies are summarized in concave curvature introduced on the shroud wall. These
Table 3. For each duct solution, the static pressure recovery, features of the shroud wall curvature had a strong influence on
Cp, the total pressure loss coefficient, ζ, and the effectiveness, the boundary layer health.
η, were evaluated between the duct inlet and exit using the
mass-averaged pressure values at the rating planes. Case Cp ζ η Case Cp ζ η
A1 0.372 0.0147 0.674 B1 0.203 0.0104 0.534
The static pressure recovery coefficient, Cp, relates the A2 0.196 0.0167 0.522 B2 0.201 0.0118 0.522
static pressure rise in the duct to the dynamic head available at A3 0.308 0.0296 0.562 B3 0.305 0.0170 0.557
the duct inlet: A4 0.170 0.0227 0.466 B4 0.349 0.0164 0.631
A5 0.159 0.0432 0.284 B5 0.349 0.0246 0.532
A6 0.169 0.0337 0.458 B6 0.211 0.0089 0.573
A7 0.123 0.0328 0.223 B7 0.323 0.0146 0.595
A8 -0.048 0.0533 -0.127 B8 0.355 0.0151 0.650
(2) A9 0.373 0.0090 0.688 B9 0.363 0.0219 0.558
A10 0.178 0.0150 0.509 B10 0.212 0.0084 0.592
The total pressure loss coefficient, ζ, relates the drop in A11 0.370 0.0116 0.678 B11 0.274 0.0140 0.515
A12 0.186 0.0120 0.520 B12 0.334 0.0130 0.622
total pressure in the duct to the inlet dynamic head: A13 0.358 0.0098 0.645 B13 0.358 0.0143 0.664
A14 0.216 0.0098 0.594 B14 0.312 0.0196 0.486
A15 0.332 0.0137 0.592 B15 0.375 0.0196 0.580
A16 0.208 0.0103 0.561 B16 0.199 0.0143 0.953
(3) A17 0.242 0.0290 0.512 B17 0.209 0.0134 0.543
A18 0.303 0.0034 0.785 B18 0.286 0.0221 0.520
A19 0.308 0.0111 0.660 B19 0.348 0.0192 0.628
The effectiveness compares the actual static pressure A20 0.336 0.0225 0.612 B20 0.217 0.0112 0.588
recovery to that which would be achieved by a lossless duct in A21 0.312 0.0139 0.668 B21 0.310 0.0187 0.571
incompressible flow: A22 0.195 0.0251 0.409 B22 0.356 0.0176 0.650
A23 0.274 0.0168 0.583 B23 0.221 0.0099 0.615
A24 0.196 0.0156 0.542 B24 0.327 0.0160 0.610
A25 0.300 0.0185 0.641 B25 0.361 0.0160 0.666
B26 0.364 0.0240 0.561
B27 0.151 0.0256 0.399
B28 0.219 0.0164 0.569
(4) B29 0.214 0.0181 0.548
B30 0.350 0.0229 0.631
Additionally, the velocity vectors near the walls within the B31 0.210 0.0169 0.571
duct were examined for indications of separation. The cases B32 0.220 0.0145 0.588
exhibiting separation are highlighted in the table. B33 0.292 0.0250 0.538
B34 0.360 0.0203 0.656
Even though the domain inlet boundary conditions were B35 0.228 0.0131 0.635
held constant for all cases, the development of the boundary B36 0.216 0.0135 0.591
layer through the Pre-duct section was influenced by the duct B37 0.318 0.0205 0.592
B38 0.364 0.0186 0.670
flowfield. Though not examined specifically, the boundary
Highlight indicates duct exhibited separation.
layer characteristics at the duct inlet rating plane were not
expected to be the same across the cases. This difference
Table 3. Summary of Predicted ITD Performance
would have influenced the duct performance. Further
discussion of the results is presented in the following sections.

Initial Design of Experiments

An analysis of variance (ANOVA) was performed on the

DOE results to evaluate the relative strengths of the geometric
factors on the duct performance parameters. Additionally, the
trends were examined for potential interactions between the
geometric parameters. As expected, the length and area ratio
parameters showed a strong effect on the duct performance. Of
particular interest was that the effects of slope and exit pitch
angle were of similar magnitude as the traditional parameters in
the region of interest for ITD’s: the difference in the mean
value of Cp for the cases at the maximum and minimum
parameter values was similar in magnitude for all four
parameters (Figure 7). Comparison of the duct geometries Figure 7. Effect of Geometry Parameters on Cp – Mean
showed that higher duct slopes tended to increase the convex Trends

6 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 An interaction was found between the length and area The duct slope had a strong interaction with each of the
ratio, in agreement with the trends observed in the Sovran & other parameters (Figure 8, upper row), the strongest of which
Klomp correlation. The rate at which Cp increased with area was with the duct length. For a given area ratio, the rate at
ratio was dependent on the length, as illustrated by the non- which the Cp drops with decreasing duct length was observed
parallel lines in the middle, right plot of Figure 8. To capture to be a function of slope. As the slope increased, the rate of
this interaction and reduce the number of parameters needed to change of Cp with length also increased. Since separation in
characterize duct performance, the Cp* line was examined. A the duct would produce a reduction in the Cp, the risk of
fit of the line in the log-log plane led to the realization that the separation for a given duct slope increased as the duct length
resulting slope was nearly unity (Equation 5). Simplifying the decreased. The physical interpretation is illustrated in Figure 9:
expression by forcing unity slope gave the expression shown in as the duct length is reduced at constant slope, the curvature on
Equation 6 and naturally led to (AR-1)/(L/h1) as a possible the shroud wall near the inlet to the duct increases. The high
correlating parameter. A value of 0.203 for this area ratio- curvature increases the local over-acceleration at the shroud
length parameter fit the Cp* line well, particularly at the inlet corner and the level of diffusion after the flow has
lengths typical for ITD’s. Higher values of the area ratio-length negotiated the corner. The thickened shroud boundary layer
parameter would tend to place a duct above the Cp* line, becomes more susceptible to separation. This finding validated
suggesting a stronger risk of separation. the initial hypothesis that duct slope would have an influence
on the location of the Cp* line. For correlation purposes, the
(5) parameter, Θ/(L/h), was adopted as a simple means of capturing
the effect of the duct slope and length on the shroud wall inlet
curvature.

(6)

Figure 8. Effect of Geometric Parameters on Cp – Interactions between Parameters

7 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Influence of Duct Length @ Constant Slope

7.0
Extended Design of Experiments

The above results prompted a follow-on study to refine the

6.0 trends and understand the parameter combinations that would
produce separation. Accurate prediction of separation was seen
as key for improving the conceptual design tools for modern,
5.0
aggressive ITD’s.
Y/h1

The full-factorial matrix of 81 cases for the 4-factor, 3-

4.0 level DOE design was reduced to 38 partially because the
ranges overlapped the previous study, but also for effective use
Case B7 of computing resources. The selection was made by estimating
3.0 Case B8 the risk of separation based on the initial DOE results. Cases
Duct Inlet
with parameter combinations that were too conservative or too
2.0 aggressive were eliminated.
-2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 As described previously, the area ratio-length and slope-
X/h1 length parameters were found to be significant drivers in how
an aggressive IT-duct behaves. Therefore, the extended DOE
Figure 9. Effect of Duct Length at Constant Slope on and initial DOE cases were combined on a graph of these two
Duct Geometry parameters (Figure 11). The cases that exhibited separation are
circled. For the most part, the separated cases fell in the upper
right side of the plot, suggesting that higher values of the area
The final significant interaction revealed by the analysis ratio-length and slope-length parameters increased the risk of
was between the exit pitch angle and the duct slope: as the exit separation.
pitch angle became lower than the duct slope, the potential for Grouping the data by slope showed that higher values of
separation increased rapidly. Figure 10 illustrates this the area ratio-length parameter could be sustained at lower
interaction: a pitch angle less than the duct slope tends to slope before separation occurred. Therefore, it was reasoned
introduce an inflection in the duct curvature and the duct shape that the location of the Cp* line on the Sovran & Klomp chart
resembles an “S”. The concave curvature on the shroud wall would be a function of slope. A family of Cp* lines as a
increases the local deceleration of the flow, thickening the function of slope was deduced from the DOE data and plotted
boundary layer and increasing the risk of separation. on a modified Sovran & Klomp chart, as shown in Figure 12.

25 Θ = 27 deg.

Θ = 31.5 deg.
Influence of Duct Exit Pitch @ Constant Slope

20 Θ = 36 deg
7.0
Separated

S&K Cp* Line

Θ/(L/h1) [deg.]

6.0 15 (Approx)

5.0
10
Y/h1

4.0
5
Case B18
3.0 Case B24
Duct Inlet
0
2.0 0.00 0.05 0.10 0.15 0.20 0.25 0.30
-2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 (AR-1)/(L/h 1)
X/h1
Figure 11. Relationship of the Area Ratio-Length and
Figure 10. Effect of Duct Exit Pitch Angle at Constant Slope-Length Parameters on IT-Duct Separation
Slope on Duct Geometry

8 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 aggressive
0.7 IT-duct. In parallel, the authors are extending the
Sovran and Klomp chart indicating separation risk study by modeling an upstream, unshrouded rotor and will re-
associated with duct slope
0
10 0.6 examine the sensitivity of duct performance to the geometric
parameters.
0.5
ACKNOWLEDGMENTS
0.4
The authors would like to thank Greg Heitland for
Cp* 0.3
initiating the aggressive IT-duct research program at
AR-1

Honeywell. We would also like to acknowledge the efforts of

0.2 Vikram Mangat on the geometry generation and CFD analyses.
-1
10
In addition the authors would like to thank Honeywell senior
0
Transition duct management for their continuous support to this research
typical region program.
of interest
10
30
40
REFERENCES
20

Sovran, G., and Klomp, E. D., 1965, “Experimentally

0 1
10 10 [1]
L/h1
Determined Optimum Geometries for Rectilinear
Figure 12. Modified Sovran & Klomp Chart with Cp* Diffusers with Rectangular, Conical or Annular Cross-
Lines as a Function of Duct Slope Section,” General Motors Corporation Research
Publication, GMR-511.
SUMMARY AND CONCLUSIONS [2] ESDU, 1975, “Performance of Circular Annular Diffusers
in Incompressible Flow,” Item Number 75026.
The numerical study of IT-duct geometries has shown that [3] ESDU, 1975, “Introduction to Design and Performance
the influence of slope and exit pitch angle can be as significant Data for Diffusers,” Item Number 75027.
as the area ratio and length parameters traditionally used to
correlate duct performance. The level of duct slope was shown [4] Axelsson, L.-U., Arroyo Osso, C., Cadrecha, D.,
to affect the location of the Cp* line on the Sovran & Klomp Johansson, T. G., 2007, “Design, Performance Evaluation
annular diffuser performance chart. The duct exit pitch angle and Endwall Flow Structure Investigation of an S-Shaped
was found to contribute to the amount of concave curvature on Intermediate Turbine Duct,” ASME Turbo Expo GT2007-
duct shroud endwall. Significant interactions between the four 27650.
duct geometry parameters were also found. [5] Axelsson, L.-U., Johansson, T. G., 2008, “Experimental
Two new parameters have been proposed to characterize Investigation of the Time-Averaged Flow in an
duct performance, particularly the potential for separation in the Intermediate Turbine Duct,” ASME Turbo Expo GT2008-
duct. The area ratio-length parameter, (AR-1)/(L/h1), was 50829.
derived from the Sovran & Klomp Cp* line. The slope-length
parameter, Θ/(L/h1), captures the tendency of increased [6] Axelsson, L.-U., George, W. K., 2008, “Spectral Analysis
curvature at the duct inlet shroud wall which drives over- of the Flow in an Intermediate Turbine Duct,” ASME
acceleration and increased diffusion along the shroud surface. Turbo Expo GT2008-51340.
Taken together, these parameters were shown to segregate ducts
[7] Marn, A., Göttlich, E., Pecnik, R., Malzacher, F. J.,
between separated and non-separated cases.
Schennach, O., Pirker, H. P., 2007, “The Influence of
Sanz et al have shown the significance of the flow field Blade Tip Gap Variation on the Flow Through an
produced by an upstream rotor tip clearance gap on the duct
Aggressive S-Shaped Intermediate Turbine Duct
performance [10]. However, the current study considered only
Downstream of a Transonic Turbine Stage - Part 1: Time-
the influence of the downstream vane on the duct. Had a rotor
Averaged Results,” ASME Turbo Expo GT2007-27405.
been included, it is expected that the study would have shown
the onset of separation occurring at more aggressive parameter [8] Göttlich, E., Marn, A., Pecnik, R., Malzacher, F. J.,
levels. The current results are likely to be conservative in Schennach, O., Pirker, H. P., 2007, “The Influence of
predicting separation. Blade Tip Gap Variation on the Flow Through an
The current study has guided the selection of geometry for Aggressive S-Shaped Intermediate Turbine Duct
the University of Notre Dame experimental program which is Downstream fo a Transonic Turbine Stage - Part 2: Time
testing aggressive ducts in a flow rig. The experimental data Resolved Results and Surface Flow,” ASME Turbo Expo
will be used to validate the CFD analyses and for down- GT2007-28069.
selecting geometry for the subsequent test of an HPT stage and

9 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 [9] Marn, A., Göttlich, E., Cadrecha, D., Pirker, H. P., 2008,
“Shorten the Intermediate Turbine Duct Length by
Applying an Integrated Concept,” ASME Turbo Expo
GT2008-50269.
[10] Sanz, W., Kelterer, M., Pecnik, R., Marn, A., Gottlich, E.,
2009, “Numerical Investigation of the Effect of Tip
Leakage Flow on an Aggressive S-Shaped Intermediate
Turbine Duct,” ASME Turbo Expo GT2009-59535.
[11] Willinger, R., Haselbacher, H., 1998, “The Role of Rotor
Tip Clearance on the Aerodynamic Interaction of a Last
Gas Turbine Stage and an Exhaust Diffuser,” ASME 98-
GT-94
[12] Dominy, R. G., Kirkham, D. A., 1995, “The Influence of
Swirl on the Performance of Inter-Turbine Diffusers,”
VDI BERICHTE NR. 1186.
[13] Dominy, R. G., Kirkham, D. A., Smith, A. D., 1996,
“Flow Development Through Inter-Turbine Diffusers,”
ASME 96-GT-139.
[14] Dominy, R. G., Norris, G., 1997, “Diffusion Rate
Influences on Inter-Turbine Diffusers,” Proc Instn Mech
Engrs Vol 211 Part A.
[15] Menter, F. R., 1994, “Two-equation eddy-viscosity
turbulence models for engineering applications,” AIAA-
Journal., 21(8), pp. 1598-1605.
[16] Bradshaw, P., Ferriss, D.H, and Altwell, N.D. 1967,
“Calculation of Boundary Layer Development Using the
Turbulent Energy Equation,” J. Fluid Mech., vol. 28, pp.
593-616.

10 Copyright © 2010 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use