TURBOMACHINES
Over more than five decades, the VKI has
specialized in the area of compressors,
pumps and turbines, covering theoretical,
experimental and numerical teaching and
research activities. The VKI is recognized
as one of the “centres of excellence” in the
area of turbomachinery-related activities and
has attracted world-renowned specialists in
the field. The principal research activities
are related to design methods, low- and
high-speed compressor and turbine flows
and heat transfer. Significant contributions
were made in the areas of inverse design
and multi-disciplinary optimization meth-
ods, unsteady flows and blade row inter-
actions, aero-thermal investigations in HP
and LP turbomachinery components, micro
gas turbines, blade-cooling applications
and instrumentation development.
DETERMINATION OF THE EFFICIENCY achieved on the rotational speed results in a high ac-
curacy on the acceleration and hence the power.
OF A COOLED HP TURBINE IN
A COMPRESSION TUBE FACILITY Inlet total pressure and temperature as well as stage
exit total pressure were determined with accurate
probes (Figure 1). Each quantity is the result of aver-
aging several heads placed at different circumferen-
The experimental verification of the efficiency of any tial locations around the test section. The knowledge
turbine stage is a crucial step in the development of of the overall mass flow combined with that of the pitch-
any new turbine. The isentropic efficiency is defined wise and radial distribution of total temperature and
as the ratio between the flow angle allows the
power effectively extracted -a- -c- -d- -e- -f- calculation of mass-
from the fluid to the power averaged values using
obtained from an isentropic quantities determined
expansion, which requires at each test and di-
testing the turbine stage un- mensionless profiles
der adiabatic conditions. -b- resulting from several
Efficiency is usually meas- tests.
ured in continuously run- -g-
ning facilities that allows Results are presented
stabilizing the turbine in Figure 2 indicating
regime and performing de- an efficiency of
tailed area traverses with ac- 0.9088 at design
curate instrumentation. In Figure 1: -a,f,g- combined probes with a Kiel head and point. Figure 2 also
such type of facilities, two a thermocouple used as reference probes upstream and shows that the evolu-
independent techniques can
downstream of the stage and inside the cavities, -b,c- tion of the efficiency
be used to measure the as a function of the
isentropic efficiency: the
rake of Kiel heads, -d- rake of thermocouples, -e- five
rotational speed is
thermodynamic and the me- hole probe in line with the 1D pre-
chanical method. diction method. A lev-
el of repeatability between 0.27 and 0.51% is achieved,
Since the late 70’s, the use of short-duration facilities depending on the test conditions. The uncertainty
has enabled testing at the actual engine levels of analysis provides a more conservative value of 0.68%
Reynolds number, Mach number, gas-to-wall and gas- for the random error. The largest contributors are the
to-coolant temperature ratios. The determination of mass flow, the stage exit and inlet total pressure.
the efficiency in such facilities is challenging because Regarding the systematic error, an uncertainty level of
all the needed quantities should be measured during the 1.2% is achieved mostly due to the contributions of the
running time, i.e., usually in less than 1s. Temperature mass flow, the inlet total pressure and the inertia.
fluctuations are usually quite large, implying tempera-
ture inaccuracies and thus the mechanical method is
preferred, based on accurate torque measurements.
TIME-AVERAGED AND TIME-RESOLVED geometries are tested on a unique rotor disk divided
into three sectors (Figure 2). Measurements included
MEASUREMENTS ON A FILM-COOLED the rotor mid-span static pressure (Figure 1c) and the
ROTOR BLADE IN A TRANSONIC heat transfer at 15%, 50% and 85% span.
HIGH PRESSURE TURBINE The increase of the coolant rate causes an increase of
the stator downstream static pressure, i.e. an increase
in the stage degree of reaction. As a consequence, the
The high-pressure turbine is located down- stator exit velocity triangle changes and the
stream of the combustion chamber; rotor static pressure field is modified ac-
hence, the blades experience a harsh cordingly. In the presence of
environment. Turbine blades must coolant, the heat flux down-
be made of super-alloys and be stream of the coolant holes is
actively cooled. Rotor blades of slightly reduced. Figure 3
modern high-pressure tur- presents the effect of the ro-
bines are equipped with tor cooling on the time-re-
several span-wise rows of solved static pressure on
cooling holes. This research the rotor crown at 50%
focuses on the effect of rotor span. The large and
film cooling on the perform- abrupt pressure increase
ance of a highly loaded tran- at gauge 3, φstator ≈ 0.8 is
sonic turbine stage.The three due to the vane trailing
different film-cooling config- edge shock that sweeps
urations investigated are dis- the rotor blade suction
played in Figure 1. The present side from gauge 5 towards
investigation was carried out for the leading edge. At gauge 3
rotor coolant flow rates of 0%, the amplitude of the fluctua-
0.5% and 0.8% of the mainstream tion amounts to ~20% of P01. In
mass flow. spite of the small increase in cool-
Figure 2: View from upstream of ing rate from 0.5% to 0.8%, the shock
the rotor disk with the three passage tends to occur sooner and
different sectors of cooled blades sooner. This means that the rotor
coolant ejection affects the vane trail-
ing edge shock position and intensi-
ty. However this effect is also partly
attributed to the ejection of the rotor
hub disk leakage flow.
Figure 1: Illustration
of vane trailing edge
shock system; three
rotor designs tested
for rotor/stator inter-
action
DETERMINATION OF THE MASS FLOW indicating that the predicted temperature history is re-
liable. The location of the piston is checked only in two
IN A COMPRESSION TUBE FACILITY points of the trajectory using a photodiode.
LOW-PRESSURE TURBINE AERODYNAMIC number (1.9 x 105 … 6.8 x 105) and Strouhal number
(0 … 0.9). The positive impact of the wakes can be ap-
PERFORMANCE preciated in Figure 2, especially at low Reynolds num-
ber where the losses are reduced by 46.7% when in-
creasing the Strouhal number.
The new generation of civil aircraft engines is charac-
terized by large fans and high by-pass
ratios. As the propulsion effi-
Figure 1: Wake generator
ciency is directly linked to
this component, the require-
ments imposed for its de-
sign are extremely se-
vere. Consequently, the Figure 2:
design of the low-pres- Profile loss
sure turbine, which is coefficient as a
driving the fan and the function of Sr
low-pressure compres-
sor, has to be performed
very carefully.
The boundary layer status and characteristics were de-
The current aerody-
termined, quantitatively, by means of a heat flux
namic performance
measurement technique. These measurements were
of LP turbines is already
carried out by means of platinum thin film gauges, de-
high but significant savings are still achievable
posited on a ceramic substrate. They operate in a tran-
through a reduction of the weight of this component
sient mode, based on the semi-infinite slab principle.
by e.g. using less airfoils per blade row. A reduced
number of blades implies a higher loading. A front
Because of their very small thickness, they provide
loaded blade design philosophy can address this re-
quantitative information in a broad frequency spec-
quirement even if strong adverse pressure gradients
trum and do not perturb the flow by, e.g. promoting
enhance the risk of separation along the rear suction
laminar to turbulent transition. They allow the charac-
side portion, considering the low Reynolds number
terization of the boundary layer status by differentiat-
conditions existing in this part of the machine. The ex-
ing the laminar from the turbulent state. In this way, it
tension of the profile losses, and therefore the per-
becomes possible to observe the temporal evolution
formance of the turbine, will be directly related to the
of the transition and to compute critical parameters
occurrence of separation and/or transition. It has been
such as the intermittency factor. A space-time diagram
shown that the unsteady/periodic effect of upstream
of the phase locked averaged suction side heat trans-
wakes and their interaction with the boundary layer
fer coefficient is presented in figure 3. The successive
may lead to a reduction of the profile losses.
wakes of the bars are identified in the leading edge re-
gion (small s/L values) along the time axis. This result
Research on the effect of periodically wake-induced
clearly shows the periodic suppression of the separa-
transition to allow high blade loading with acceptable
tion bubble (characterized by very low heat transfer
profile loss levels is conducted in the “Laboratory
coefficient values) in the wake path of the upstream
Jacques Chauvin” [AJ24, MP131, MP132]. This work
bars and its complex redistribution under these
is performed in the Light Piston Isentropic
unsteady conditions.
Compression Tube facility CT-2, allowing a correct sim-
ulation of Mach and Reynolds numbers as encoun-
tered in a real engine. A stationary linear cascade Figure 3:
model is subjected to wakes shed from upstream bars Phase locked
mounted on a fast rotating disk (Figure 1). This wake
heat transfer
generator provides the advantage of being able to
drive the bars at the velocity required to simulate the coefficient
correct inlet velocity triangle.
EXPERIMENTAL STUDY OF A VERY HIGH The characterization of the “compressible” profile was
performed in the Isentropic Compression Tube facili-
LIFT, LOW PRESSURE TURBINE BLADE ty, including the influence of periodic incoming wakes.
resulting from the cooling techniques is mandatory in events generated by the turbulent structures evolving
order to predict reliable metal temperatures and con- in the near wall regions. These events make the heat
sequently component lifetimes. The present contribu- transfer process more effective but also much more
tion therefore focuses on the analysis of the aero-ther- complicated to study and to understand. In order to
mal performance of a stationary rib-roughened get additional information on the flow behavior, some
cooling channel flow, as encountered inside high pres- numerical simulations using advanced turbulence
sure turbine vanes. modeling and Large Eddy Simulation are under way.
The analysis of the numerical results and a deeper in-
The analysis is carried out on the aerodynamic (ve- terpretation of the available data will lead to the final
locity field) and convective heat transfer data recent- definition of the flow mechanisms responsible for the
ly acquired in a large scale, stationary cooling chan- heat transfer field development.
nel test section. The selected channel and rib
geometries constitute a quite detailed test case under Figure 1: Comparison between heat transfer and
study at the VKI for the last few years [AJ67,MP6, velocity fluctuations normal to wall between two
MP30,MP35]. Several, sometimes redundant, meas- ribs
urement techniques were used for this purpose in or-
der to provide high quality data. Both the aerodynamic
and heat transfer data are acquired in between two
consecutive ribs. The use of a Digital Particle Image
Velocimetry (DPIV) technique allows the acquisition of
highly detailed flow field measurements over differ-
ent and mutually perpendicular planes in the full
interrib space. A quasi 3D view of the flow field is
therefore available and its interpretation provides an
exhaustive description of the mean flow topology.
The heat transfer data were gathered with the help of
a Liquid Crystal Thermometry technique. It was ap-
plied on all the channel walls, rib surfaces included,
and provided in this way the complete heat transfer
Figure 2: Comparison between velocity
coefficient distribution along the investigated channel fluctuations on top of the rib
region.
advanced materials and cooling techniques must be In both channels, a global three dimensional view of
adopted for a safe operation of the HP gas turbine the flow was attempted by measuring the whole flow
blades. Apart from the progress made in the metallur- field in between two consecutive ribs over different
gical domain, a continuous cooling of the blades of the and mutually perpendicular planes. Where previous
turbine first stage allows operating at temperatures information was available, comparisons were made
which are far above the material’s melting point, with- with, e.g., results from alternative measurement tech-
out affecting the component integrity and geometry. A niques, showing most of the time a remarkable agree-
complete and correct understanding of the convection ment and providing complementary information.
mechanisms associated to the applied cooling tech- Moreover, the data analysis provides a lot of informa-
niques is therefore of major importance. The present tion regarding particular flow structures in terms of
experimental study therefore deals with the detailed their location and size, as shown in Figure 1 for the
aerodynamic investigation of the turbulent flow inside square section channel. For some of these structures, a
rib-roughened turbine blade cooling channels of vari- model describing their evolution was proposed [MP6].
ous geometries. Inside these passages, the forced con-
vection cooling process is significantly enhanced by Figure 2: Comparison
the presence of ribs (turbulence promoters) installed between wall heat trans-
on one or more walls of the channels.The effects of the fer and streamlines
rib size and orientation on the flow behaviour, and
therefore on the heat transfer and pressure distribu- The combined interpretation of
tion along the channel walls, are closely related to the the PIV results and of the avail-
safe operation of HP gas turbine blades. able heat transfer measure-
ments on the channel walls,
The present research effort is devoted to the experi- highlighted the existing links
mental analysis of the flow inside two cooling chan- between the aerodynamic and
nels of different geometries on the basis of detailed thermal behaviour of the flow
aerodynamic measurements performed by Digital [MP30, MP35]. A typical exam-
Particle Image Velocimetry (DPIV). The principal aims ple, obtained in the second
are to provide new information about the behaviour cooling channel, is reported in Figure 2. The mean
of such a complicated flow, useful for its understand- flow path measured by DPIV as close as possible to
ing, to complement the wall heat transfer coefficient the ribbed wall and the Nu number distribution
distributions already available and to create a wide and (coloured contour plot) measured by Liquid Crystal
reliable data base for numerical code validation [AJ67]. Thermometry on the ribbed wall are fully coherent with
respect to separation and reattachment area location.
Figure 1: 3D flow
topology upstream
of the rib COMPARISON OF TURBINE TIP LEAKAGE
FLOW FOR FLAT TIP AND SQUEALER TIP
GEOMETRIES AT HIGH-SPEED CONDITIONS
Figure 2: Multidisciplinary optimization flow chart Micro gas turbines have experienced a growing interest
during the last decade. Their large energy density
A new topic of re- (Whr/kg) makes them an attractive replacement for
search is multi-point much heavier batteries for the propulsion of small air-
optimization, where planes (UAV) or portable power units [MP207,AJ64].
the optimum geome- Designing micro gas turbines by a simple scaling of
try is a function of the large high performance gasturbines will not provide
performance at design good results because the scaling conditions are
and off-design condi- not satisfied:
tions. This has been - The small dimensions result in a large heat
achieved without an transfer between the hot turbine and cold com-
increase of the design pressor and the assumption of adiabatic flow is
time by a paralleliza- no longer valid [MP109,MP168]. A conjugate
Figure 3: 2D rotor stress dis- tion of the calcula- heat transfer model (CHT) is needed to evaluate
tribution at reference blade tions, using as many the heat flux and its impact on the compressor
processors as the op- and turbine performance.
height and thickness
erating points [MP217, - Micro gas turbines operate at very low Reynolds
MP178]. numbers (10000) and machining techniques may
The system is presently being extended to Multi not allow very smooth surfaces. The increase of
Disciplinary Optimization. It is an extension of the ex- rotational speed results in larger Coriolis forces.
isting method (Figure 1) combining the mechanical and The large heat fluxes may induce buoyancy flows.
aerodynamic optimization (Figure 2) [IB18, MP250]. The Hence the flow structure may be very different
GA searching for the optimum geometry gets its input from the traditional one and it is doubtful that the
from a Finite Element Stress Analysis (FEA) as well as turbulence models are still valid at the micro gas
from the Navier-Stokes flow analysis. turbine operating conditions.
Both topics are the subject of intense research in - The possibility to mix simpler axisymmetric and
recent years at VKI. more expensive full 3D calculations, as is demon-
strated in the next example.
Time resolved PIV in a rotating channel between 0.1 and 0.55 and Buoyancy numbers up to
0.73. A protective cage, connected to an aspiration
An accurate prediction system, collects the seeding particles at the channel
of the micro gas turbine exit.
performance requires a
correct estimation of Relative velocity fields in the rotating channel are
the impact of Coriolis measured by means of Particle Image Velocimetry us-
forces and heat transfer ing a high speed camera and a continuous laser, both
on low Reynolds num- rotating with the channel. The CMOS camera can
ber flows over rough record up to 7.8kHz with a 640 x 480 pixel resolution
surfaces. The RC-1 facil- and has a 2Gb internal memory to store the pictures.
ity has been specially The air-cooled laser diode provides up to 25W laser
built to study these ef- light at 806nm wave length. It is coupled by an optical
fects and to evaluate fibre cable to the line generator. This compact system
their impact on turbu- allows a direct and hence more accurate measure-
lence models. It allows ment of the relative velocity as well as time-accurate
Figure 4: Close up high resolution, time re- measurements of the flow variations [MP236,MP229].
of the rotating channel solved Particle Image
Velocimetry in rotating Measurements are taken in a plane parallel to the
diverging channels. The aim is to build a database for channel bottom wall, halfway between inlet and out-
validation/modification of existing turbulence models let. It corresponds to the blade plane in the micro gas-
of the CFD codes. turbine blade. Mean and instantaneous flow fields are
obtained in the stationary channel and at different ro-
The facility consists of a 0.7m long divergent Plexiglas tational speeds.
channel, mounted on a rotating disk. It has a 6° total The increasing influence of rotation on the boundary
aperture angle, a 0.079m inlet hydraulic diameter and layer mean flow field can be noticed in Figure 5.
represents a large scale model of a micro gas-turbine Boundary layer thickness increases on the suction side
impeller passage. The volume flow is measured with and decreases on the pressure side. The flow is more
a Venturi connected to the rotating channel and ad- turbulent and less likely to separate on the pressure
justed by means of an upstream fan / throttle valve side than on the suction side.
combination. It respects the main scaling parameters Figure 6 shows how at Ro = 0.33 the vorticity concen-
of the radial impeller passage: i.e. Reynolds number trates closer to the pressure side wall but spreads over
between 3 x 103 and 3 x 104, rotational number (Ro) a large part of the domain near the suction side.
a) b) c)
© 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Genèse, Belgium 53
Flow In Turbomachines
DESIGN OF A HIGHLY LOADED AXIAL The blade profiles have been designed as Controlled
Diffusion Blades by the use of an inverse method de-
COMPRESSOR STAGE veloped at VKI starting from initial subsonic NACA 65
FOR THE INVESTIGATION profiles.
OF THE INFLUENCE OF CASING A final design and off-design 3D Navier-Stokes calcu-
TREATMENTS ON THE STABILITY MARGIN lation of the geometry permitted the validation of the
design against specifications and computed the oper-
ating map of the stage (Figure 2).