Beruflich Dokumente
Kultur Dokumente
G.B. Ashakiran1
M. S. Ramaiah School of Advanced Studies, Bangalore 560017, India
A. Gogoi2
ADA, PB No1718, Vimanapura Post, Bangalore 560 017, India
and
Q.H. Nagpurwala3
M. S. Ramaiah School of Advanced Studies, Bangalore 560017, India
Abstract
The next generation design of small gas turbines outshines the present capabilities in terms of improved
aerodynamics, superior technologies and high reliability to ensure air superiority and survivability. Auxiliary Power
Units (APUs) are such small gas turbines that provide many benefits to the modern military aircrafts. The
performance of each individual component of an APU is therefore significant. Efforts have been made by many
researchers through experimental and computational means for the improvement in performance in each of the
components of an APU, and especially that of a turbine stages. The design of a turbine stage that produces high
performance is therefore essential to meet the specific requirements of APU.
This Paper presents design and analysis of a high performance axial turbine stage for an Auxiliary Power Unit.
Preliminary free vortex based design, CFD analysis and mechanical analysis were carried out in an iterative manner
and optimisation processes were incorporated at each step of detail design process. A preliminary design code was
also developed in the course of the present work. The Computational Fluid Dynamics (CFD) software tool,
NUMECA, was validated by comparing the computational results with the experimental data of the Aachen turbine.
Appropriate computational grid for the full turbine stage with desirable features like orthogonality, clustering,
smoothness, etc was used and the CFD simulations were carried out using the Spalart Allmaras [1] turbulence
model and the central difference scheme of Jameson [2].
The results from the computational study were in excellent agreement with the preliminary design values and
the turbine stage produced impressive efficiency even at lower specific speeds. Effects of various parameters, like
tip clearance and blade stagger, which could not be addressed in the preliminary design, were studied through CFD
analysis and optimum values of these parameters were obtained. The off-design performance evaluation was also
carried out through CFD simulation and the results were compared with available loss modelling procedure[3,4].
The design was also checked for mechanical integrity in terms of stress levels in the blisk using ANSYS. The blisk
was found safe even with 15% overspeed to meet the specifications as per military standards.
Key Words: High Performance, Design Space, Artificial Neural Network, Multi-grid, Weight Optimization
1
Student, M.Sc. (Engg.), Rotating Machinery Design, MSRSAS
2
Scientist, Propulsion System, Aeronautical Development Agency
3
Professor, Centre for Rotating Machinery Design, MSRSAS
1
American Institute of Aeronautics and Astronautics
Nomenclature use of a number of guidelines and correlations as they
2 form the essential features of the overall design
A = Cross-sectional area, mm exercise of a turbomachine. They are now built
C = Secondary loss coefficient around more versatile FEA and CFD computational
D = Diameter, m tools that can reduce experimental lead time and
H = Enthalpy, kJ/kg operate at much lower costs.
N = Rotational speed, rpm
T = Temperature, K Problem Description: Small gas turbines have gone
U = Blade speed, m/s through large changes since 1950s. The evolution of
V = Absolute velocity, m/s small gas turbines has been primarily driven by the
Y = Blade loss coefficient influence of military requirements, as there was a
c = Blade chord, mm need for something better than piston engines.
h = Blade height, mm Due to smaller and lighter gas turbine engines, it
s = Blade space, mm has been possible to achieve higher operating
t = Blade thickness, mm pressures and temperatures in the core components.
α = Flow angle, radians The credit also goes to new concepts and processes
β = Blade angle, radians that have given unique solutions at lower life cycle
cost and for higher life of the components without
Abbreviations deteriorating their performance.
APU = Auxiliary Power Unit
The current design requirement of an axial turbine
AI = Artificial Intelligence
stage is based upon the potential future application in
ANN = Artificial Neural Network
a high performance small turbo-shaft engine, e.g. an
CFD = Computational Fluid Dynamics
APU. The primary objective here was to design an
FEA = Finite Element Analysis
un-cooled turbine stage having maximum
GGT = Gas Generator Turbine
aerodynamic performance while operating within the
GTS = Gas Turbine Starter
limits of aerodynamic and mechanical constraints.
HCF = High Cycle Fatigue
NGV = Nozzle Guide Vane
MIL = Military Standards II. Preliminary and Detail Design
TIT = Turbine Inlet Temperature The axial flow turbine stage design began with an
l-D mean-line preliminary design calculation. A
I. Introduction spreadsheet was developed to execute the design
iterations, keeping the maximum performance goal of
G as turbines are typical power sources used in a
wide range of sizes for propulsion and power
the APU program. A number of design iterations
were carried out by varying the rotor mean speed and
generation systems. Recently, there has been
exit swirl angle. The findings of the preliminary
considerable interest shown in the development and
design study showed that the parameters which
application of small gas turbines to extort high power
improved the aerodynamic performance, tended to
density at lower costs. However, resulting from the
degrade the mechanical integrity of the turbine and
smaller size, their efficiency, pressure ratio and
vice versa. A ‘design space’ was therefore generated
specific work are mostly lower than those of the large
considering the acceptable aerodynamic and
scale systems. Therefore, more detailed studies are
mechanical limits, as shown in Figure 1.
required to effectively enhance their performance.
The designs in the lower left section of the
The design of a small gas turbine engine is a
acceptable region had the lowest mechanical risk in
fairly lengthy and complex process, involving many
terms of tip speed, whereas designs in the upper right
engineering disciplines, like material sciences, heat-
corner had the highest aerodynamic performance but
transfer, thermo-fluids, turbomachinery design,
with higher mechanical risks. Keeping in view the
testing for validation and finally the airworthiness for
maximum performance goal, a design point ‘P’ was
certification. Since all the above fields are employed
chosen with a rotor blade speed of 380 m/s and an
while designing an appropriate engine configuration,
exit swirl angle of 20°, which showed a reaction of
it requires several months, and sometimes years, of
45% at mean. It was observed that increasing the
detailed analyses and testing.
swirl angle beyond 20° caused the downstream free
The turbine design process can be broadly power turbine stage to suffer, where as reducing the
classified as preliminary design and detail design exit swirl angle, led to a lower reaction at mean.
process. Each step in these design processes makes Also, increasing the mean blade speed beyond 380
2
American Institute of Aeronautics and Astronautics
m/s was not found feasible due to the mechanical reaction, whereas the mechanical constraints were
limitations. In the present scenario, the performance evaluated in terms of blade tip speed, disc stresses
was measured in terms of efficiency and stage and vibration levels.
3
American Institute of Aeronautics and Astronautics
instance, was fixed at 4.75 mm, which was 25 % of Though a detailed understanding of the flow
the upstream axial chord. Upon fixing the axial conditions enables the designer to optimise the blade
length, the incline angle was adjusted to smoothen profiles for a given flow condition, testing of a
the flow path by fitting a spline curve as shown in the number of profiles and CFD iterations are required to
Figure 3, thereby improving the performance. It was arrive at the optimal profile. For instance, the profile
found that as the incline angle increased, the nozzle that showed an improved performance than the
blades became too short, making them difficult to former base profile is shown in Figure 4. Dotted lines
manufacture. indicate the preliminary or base profile.
4
American Institute of Aeronautics and Astronautics
The inputs for the ANN were a set of geometric shocks in the trailing edge region. For accuracy,
and aerodynamic parameters of the blade, like many nodes were taken in the streamwise direction to
lengths between control points, stagger angle, throat capture the flow physics at the trailing edge.
area and mass flow rate. The summation used
For meshing of rotor passages, the computational
sigmoid functions of values varying between 0 and 1.
domain was divided into sub-domains, like blade and
tip clearance regions, and each of these supported a
structured grid as shown in Figure 8. The approach
chosen to generate the grid for a rotor with tip
clearance was similar to the one generated for stators.
A radially stacked mesh, formed by 2-D multi-block
grid surfaces, extended from hub to shroud, thus
leaving a gap in the tip clearance region. The grid
lines were clustered towards hub and the casing to
capture the end wall boundary layers correctly.
Block Grid
Size (i*j*k) Description
Figure 7 Multi block grid generation showing No. Type
rotor and stator interface
1 33*41*217 Stator Blade O-H
First stage nozzle guide vanes of the gas generator 2 49*41*33 NGV Inlet H-H
turbine are subjected to both aerodynamic and
3 33*41*209 Rotor Blade O-H
thermal loads in spite of their smaller dimensions. A
multi-block grid was chosen to ensure adaptivity in 4 33*41*49 Rotor Outlet H-H
such fields. Multi-block grid provided the freedom of 5 9*9*161 Tip Clearance O-H
selecting the number of nodes on the profile as the 6 17*9*65 Tip Clearance H-H
flow was subsonic and there was no problem of
5
American Institute of Aeronautics and Astronautics
In order to describe the flow physics accurately and The results obtained from NUMECA software
to obtain good quality numerical results, special care were compared with those obtained from ‘TFLO’,
must be paid to the treatment of the boundary United Technologies’ solver ‘3D FLOW’,
conditions. The following were the boundary NASA/GE’s solver ‘APNASA-V5’ and the
conditions imposed on the axial turbine stage: experimental data. The circumferentially averaged
total pressure, total temperature and absolute flow
The stage inlet flow angle was set to zero. The total
angle at the rotor exit, 8 mm downstream of the
temperature and total pressure values were 1200 K
trailing edge of the NGV and rotor, were considered
and 583188 Pa respectively. A radial equilibrium
for comparison.
back pressure of 210000 Pa was imposed at the exit.
Air was considered as the real gas and the values of Figure 9 shows the circumferentially averaged total
Cp and specific heat ratio were taken as 1160 J/kg/K pressure for the rotor. The predicted results are in
and 1.32 respectively at a temperature of 1200 K. good agreement with the experimental values and
with the results from other codes like TFLO,
The one equation Spalart-Allmaras turbulence
APNASA V5 and 3 D Flow.
model was used in all the computations and the
equations were solved by central difference scheme.
The hub wall of the rotor was given rotation along V. CFD Analysis
with the blade while the NGV hub and entire shroud CFD analyses were then carried out on the
were stationery. Implicit Residual Smoothening and designed turbine stage for the optimum point ‘P’ on
Multi-Grid Acceleration were used for convergence the design space. At the design point, a mass flow
acceleration of the solution. rate of 1.96 kg/s at an adiabatic efficiency of 90%
was obtained for an expansion ratio of 2.34. The
IV. Validation Study turbine stage generated a torque of 82.92 N-m and a
power output of 440 kW, which was adequate to
The validation of the numerical simulation
drive the 6:1 pressure ratio centrifugal compressor.
procedure was performed by analysing the Aachen
The results of the CFD analysis at the design point
1.5 stage turbine [7] and comparing the NUMECA
are discussed in the following sections:
results with the published experimental values and
those from other turbo-machinery codes. A) Static pressure: A smooth drop of static pressure
was observed at the mean section. At the hub and the
The inlet total pressure and outlet total pressure
tip sections, a modest flow deceleration was observed
were 166871 Pa and 112500 Pa respectively, with a
on the suction surface as shown in Figure 10 and
rotor speed of 3500 rpm. The grid was generated
11(a). Figure 10 shows the static pressure variation
taking 4 blocks with 890124 points and radial
for the baseline and improved profiles that involved
equilibrium boundary conditions were imposed at the
several CFD iterations (refer Figure 4 also).
outlet. The Jameson-Schmidt-Turkel [4] scheme and
the one equation Spalart Allmaras [5] turbulence
model were used for computations.
6
American Institute of Aeronautics and Astronautics
recovery for the nozzle was around 96 to 97 %. A varied between 35° and 45°. The second parameter,
further improvement in pressure recovery of up to i.e. the tip clearance was varied over three different
98-99% can be obtained in the nozzle through values, viz. 0.5%, 1% and 1.5% of the blade span and
extensive CFD experiments. its effect was evaluated with respect to the
performance characteristics of the turbine stage.
C) Mach number: Smooth flow acceleration was
observed towards exit of the rotor and stator The problem associated with the re-staggering
passages. Though the pockets of local acceleration process was the reduction in throat area that led to
were observed on the suction side of the blade, they choking of the flow passage. Studies were carried out
did not lead to any shock as shown in Figure 11 (c). to evaluate the turbine performance for higher stagger
Again, since the flow is subsonic at the nozzle hub, angles without altering the throat area. The CFD
the passage did not get choked. studies were executed with stagger angle of 35°, 40°
and 45° at a constant N /√T value of 1530.
D) Entropy: Production of entropy in the free stream
flow was found negligible but there was considerable Figure 12 shows the variation in efficiency with
entropy production at the hub and the tip sections as change in stagger angle at a constant blade speed. As
shown in Figure 11 (d). The entropy production at expected, the performance improved in terms of
the hub could be attributed to the random flow near efficiency as the stagger was increased from 35° to
casing and the entropy in the tip was mainly due to 40°. But some interesting results were obtained when
tip clearance vortex. the stagger was further increased to 45°. It was found
that the efficiency was tending to decrease with an
increase in stagger beyond 40° at meanline. This
could be attributed to the fact that even a small
change in throat area can lead to mass flow variations
across the turbine stage.
7
American Institute of Aeronautics and Astronautics
the immediate vicinity of the blade tip but spread realised when mass flow rate becomes constant for
radially inwards with increasing tip clearance, further change in back pressure as shown in Figure
covering nearly 25% of the blade span. This becomes 14 by a highlighted block.
evident from Figure 13, which shows the variation of
This study shows the importance of a careful
efficiency and hence performance distribution.
control of the airfoil loss characteristic as well as the
vane and blade area matching.
8
American Institute of Aeronautics and Astronautics
the blisk and the working environment was set to The morphed model was then optimised to
950° C. Aditionally, the gas loads on the blade were minimize the disc weight and still keeping the stress
imposed. levels within the yield limit of the material at an
operating temperature of 1200 K. There were only
From manufacturing point of view also, the process
two main variables for disk weight optimisation , viz.
should not become too complex. Keeping in view
web thickness and hub radius at the shaft.
various constraints, it is just not possible to increase
or decrease the thickness of discs as desired. It is,
therefore, required to find a minimum thickness of
bladed disc that overcomes mechanical limitations as
well as results in saving the weight of the component.
The component weights in the airborne engines have
high significance because saving the weight on each
individual component can save overall weight of the
aircraft, and hence overall programme cost.
A parametric study was taken up to find the
minimum thickness of the web. It involved a number
of manual iterations. The parameters considered for
Figure 17 Shape morphing to generate blisk of
the design were ‘t1’, the thickness of the web; ‘t2’ the
uniform strength
thickness of the blade land; and ‘t3’ the thickness of
the shaft hub; as shown in Figure 16. A few initial The optimisation process involved a technique
iterations were tried with ‘t1’ = 7.5 mm, 8 mm and called Design of Experiments (DOE). As a starting
8.5 mm based on experience. These designs were not design point, the optimizer was deliberately given a
considered as they showed a high level of stress, design band of variation for each input parameters
which was not acceptable as a solution. that generated a diverse solution for the disc.
9
American Institute of Aeronautics and Astronautics
criteria in terms of stage reaction (45.5%) and [3] Ainley, D. G., and Mathieson, G. C. R., (1951)
efficiency (90.01%). The other features at the “A Method of Performance Estimation for
design point included an aerodynamic loading < Axial Flow Turbines”, ARC technical report,
4.0, exit swirl angle of 20°, mean blade speed < Ministry of Supply, Aeronautical Research
380 m/s and un-choked nozzles Council reports and memoranda, R & M No
2974.
A turbine design code is developed based on free
vortex theory and the output of this forms the [4] Ainley, D. G., and Mathieson, G. C. R., (1951)
basis of the blade generation. “Examination of the Flow and Pressure Losses
in Blade Rows of Axial Flow Turbine”, ARC
The results obtained from the CFD computations
technical report, Ministry of Supply,
are in good agreement with the design targets of
Aeronautical Research Council reports and
thermodynamic cycle and preliminary design. A
memoranda, R & M No 2891
mass flow rate of 1.964 kg/s is obtained as against
expected mass flow rate of 1.896 kg/s [5] Baskharone Erian, A., “Principles of Turbo
machinery in Air-Breathing Engines”,
Studies on tip clearance suggested a minimum tip
Cambridge University Press, ISBN-13 978-0-
clearance of 1%, which resulted in best
521-85810-6
performance of the turbine.
[6] Pierret, S., and Braembussche Van den, R.A.,
Studies with varying stagger angles (30° to 45°)
(1999) “Turbo machinery Blade Design using a
for the NGV suggested a stagger angle of 40° that
Navier-Stokes Solver and Artificial Neural
gave the best performance without choking.
Network”, RTO-MP-8, AC/323(AVT)TP/9,
Mechanical analysis of the blisk was carried out, ISBN 92-837-0005-8.
both at the design speed and the burst speed of
[7] Yao, J., Jameon, A., Alonso, J.J., and Liu, F.,
115% and the stress levels were found well within
(2000) “Development and Validation of a
the safe limit.
Massively Parallel Flow Solver for Turbo
Design optimization on the blisk was carried out machinery Flows”, 38th Aerospace Science
for its weight and a disc of 1.4 kg was selected as Meeting and Exhibit.
the optimal design, keeping the stress levels
within allowable limits.
Shape morphing of the blisk was performed and
the blisk of uniform strength was generated.
The results have demonstrated the utility of an
integrated approach involving preliminary design,
CFD analysis and mechanical analysis to arrive at
a satisfactory design of the turbine stage.
Acknowledgments
The authors would like to thank all the members of
the Propulsion Systems, ADA and the Centre for
Rotating Machinery Design, MSRSAS, Bangalore for
their helpful suggestions.
References
[1] Spalart, P. R., and Allmaras, S.R., “A One
Equation Turbulence Model for Aerodynamic
Flows”, AIAA Paper 92-0439.
[2] Schmidt Jameson, W., and Turkel, E., (1981)
“Numerical Solution of Euler Solution by finite
volume methods with Runge-kutta time
stepping schemes”, AIAA paper 81-1259
10
American Institute of Aeronautics and Astronautics