Journal of Thermal Science Vol.24, No.

6 (2015) 526534

DOI: 10.1007/s11630-015-0818-7 Article ID: 1003-2169(2015)06-0526-09

Transition Modelling Implications in the CFD Analysis of a Turbine Nozzle

Vane Cascade Tested Over a Range of Mach and Reynolds Numbers

Michele Marconcini, Roberto Pacciani, Andrea Arnone

University of Florence, Department of Industrial Engineering Via Santa Marta, 3, 50139, Firenze, Italy,

© Science Press and Institute of Engineering Thermophysics, CAS and Springer-Verlag Berlin Heidelberg 2015

The aerodynamic performance of a gas turbine nozzle vane cascade was investigated over a range of Mach and
Reynolds numbers. The work is part of a vast research project aimed at the analysis of fluid dynamics and heat
transfer phenomena in cooled blades. In this paper computed results on the “solid vane” (without cooling devices)
are presented and discussed in comparison with experimental data. Detailed measurements were provided by the
University of Bergamo where the experimental campaign was carried out by means of a subsonic wind tunnel.
The impact of boundary layer transition is investigated by using a novel laminar kinetic energy transport model
and the widely used Langtry-Menter γ−Reθ,t model. The comparison between calculations and measurements is
presented in terms of blade loading distributions, total pressure loss coefficient contours downstream of the cas-
cade, and velocity/turbulence-intensity profiles within the boundary layer at selected blade surface locations at
mid-span. It will be shown how transitional calculations compare favorably with experiments.

Keywords: Gas turbine, CFD validation, transition model, secondary flows.

Introduction In cooled blades the showerhead and film cooling jets

in conjunction with the typical high Reynolds number
High pressure turbine nozzles are usually refrigerated operations and the high freestream turbulence coming
and thus characterized by leading edge showerhead, from the combustor are likely to promote early transition,
film cooling, and coolant ejections from the trailing edge. so that the boundary layer is generally assumed as fully
The interaction between cooling flows and the main flow turbulent. In order to assess the impact of the cooling
poses a serious challenge for the study of the flow phys- flows on the aerodynamics and heat transfer, also studies
ics either from the experimental point of view or the nu- on solid blades (without cooling holes and slots) in wind
merical one. tunnels are generally foreseen. In such circumstances,
In order to provide detailed investigations, experi- due to the strongly accelerating flow from the leading
mental approaches combined with CFD analyses are be- edge up to the passage throat, the blade boundary layer
coming more and more common. The works of Moore can remain laminar for a relevant fraction of the blade
and Gregory-Smith [16], Duden and Fottner [11], Martini chord. In these cases a correct interpretation of the expe-
et al. [15], Corriveau and Sjolander [9], Gomes and Nie- rimental data calls for the use of transition models in
huis [12], are only some examples of succeeding syner- CFD calculations. Moore and Gregory-Smith [16] ob-
gies between experiments and CFD aimed at the detailed tained good computed predictions for both loss and sec-
investigation of the flow physics in modern gas turbine ondary flowfield in a high turning cascade by using a
bladings. mixing layer model with specified intermittency to ac-

Received: August 2015 Michele Marconcini: PhD, Assistant Professor

[Link]
Michele Marconcini et al. Transition Modelling Analysis of a Turbine Nozzle Vane Cascade over a Range of M and Re numbers 527

Nomenclature
c blade chord ρ density
H blade height θ boundary layer momentum thickness
k turbulent kinetic energy ζ kinetic energy loss coefficient
lT turbulence length scale, lT = k 1/2 /ω Subscripts and Superscripts
M Mach number 1 cascade inflow
o throat 2 cascade outflow
p pressure ∞ freestream
Reis isentropic Reynolds number, U is c/ν is isentropic
Reθt local transition onset momentum thickness max maximum value
Reynolds number ms mid-span value
s blade pitch, curvilinear abscissa t stagnation quantity, transition onset
Tu turbulence intensity, Tu = (2k/3)1/2/U 2,is Acronymes
U velocity magnitude CFD Computational Fluid Dynamics
Zw Zweifel number DR Diffusion Rate
Greek FT Fully turbulent flow
α flow angle LKE Laminar Kinetic Energy
γ intermittency LM Langtry-Menter model
ν kinematic fluid viscosity TE Trailing Edge
ω specific turbulence-dissipation rate

count for transition. Pacciani et al. [19], using a laminar assess the prediction capabilities of the employed transi-
kinetic energy transport model [17, 18], showed how a tion models and provide a more realistic interpretation of
correct transition prediction was crucial for the reproduc- the experimental results relative to the use of a fully tur-
tion of the external heat transfer coefficient distribution bulent approach.
on a solid high-loading blade tested at high Mach number
and low Reynolds number conditions. Benyahia and Computational Framework
Houdeville [7] achieved reasonable external heat transfer
predictions on the well-known VKI LS89 vane cascade The TRAF code (Arnone et al. [2, 3]) was used in the
[4] for various Reynolds numbers and freestream turbu- present work. The unsteady, three-dimensional, Rey-
lence conditions using the Langtry-Menter correlation- nolds-averaged Navier-Stokes equations are written in
based transition model [13]. conservative form in a curvilinear, body-fitted coordinate
In 2011 the Italian Ministry of Instruction, University system and solved for density, absolute momentum
and Research (MIUR) funded a national research project components, and total energy. Convective fluxes are dis-
(PRIN 2011 INSIDE - Aerothermal Investigation of cretized by a 2nd order TVD-MUSCL strategy build on
cooled Stage turbIne: Design optimization and Experi- the Roe’s upwind scheme. A central difference scheme is
mental analysis) devoted to the study of the complex heat used for the viscous fluxes. The concept of artificial
transfer phenomena characterizing the high pressure compressibility of Chorin [8] is used to handle incom-
stages of gas turbines. The starting point was the study of pressible fluids by a time-marching approach.
a subsonic linear cascade with high-pressure nozzle Two different transition modelling frameworks have
guide vanes. It was tested at the University of Bergamo been considered in the present work, both are coupled to
in both the solid and cooled configurations [5,6]. In the the Wilcox k−ω turbulence model [21]. The widely-used
present paper the solid configuration was analyzed nu- γ−Reθt model proposed by Langtry and Menter (LM) [13]
merically. The TRAF code [2, 3] was used to carry out couples an intermittency and a transition-onset Reynolds
three-dimensional steady adiabatic transitional calcula- number transport equation to obtain a dynamic descrip-
tions over the whole range of tested Mach and Reynolds tion of the transition, which makes use of local variables
numbers. The LKE model of Pacciani et al. [17,18] and only. Such a model is conceived as a general framework
the Langtry- Menter γ−Reθt model [13] were used to si- to include correlation-based transition criteria in
mulate transition in conjunction with the Wilcox k−ω RANS/URANS solvers overcoming the difficulties asso-
turbulence model [21]. The purpose of the work was to ciated with the use of non-local variables.
528 J. Therm. Sci., Vol.24, No.6, 2015

Table 1 Cascade geometry and operating conditions. ble Zweifel number of 1.18. An experimental campaign
c 142.1 mm 1 0° was performed on the solid vane with the aim of collect-
ing a detailed set of measurements to be used as a refer-
s/c 1.04 M2,is 0.20.6
ence case for investigating the cooled configuration, a
H/c 0.69 Re2,is 6.5x1051.6x105 summary of the tested operating conditions is given in
Zw 1.18 Tu 1.6% Tab. 1 while a detailed description of the measurement
cos-1(o/s) 73.6 lT, /cx 1% techniques may be found in [5, 6]. The vane operating
condition at design is M2,is = 0.6 and Re2,is = 107.
Isentropic exit Mach numbers going from 0.2 to 0.6
were tested, and detailed experimental data are available
at low speed condition in terms of loading distribution,
boundary layer velocity and turbulence intensity profiles
along both the suction and the pressure side at mid-span.
Downstream of the trailing at x/cx = 1.53 the distributions
of the kinetic energy loss coefficient and the exit flow
angle are available. The measured inlet boundary layer
was imposed upstream of the cascade inlet plane at
x/cx=−1.6 together with a freestream turbulence level of
1.6%.
A region of constant profile losses near mid-span is a
Fig. 1 Three dimensional view of the H-O-H grid. necessary condition for two-dimensionality of the flow.
Sieverding [20] recommended a minimum aspect ratio
The laminar kinetic energy (LKE) transition model for reliable two-dimensional performance in a turbine
proposed by Pacciani et al. [17, 18] is used to sensitize cascade as a function of the velocity ratio across the cas-
the two-equation turbulence model to the transition onset. cade. A minimum aspect ratio is required in order to ob-
Transition is triggered via the energy transfer between the tain reasonably two-dimensional mid-span flow for typi-
laminar fluctuation regime and the turbulence field, and cal values of inlet endwall boundary layer thickness.
is accomplished by means of a suitable coupling between Based on Sieverding’s recommendations, the current
the LKE transport equation and the eddy-viscosity model. cascade would require an aspect ratio of about 1.6 at de-
A detailed description of the model can be found in [18]. sign Mach number, given that the velocity ratio equals
The computational domain was discretized using a 0.24. This is higher than the present cascade aspect ratio
H-O-H grid with 29×101×121, 401×81×121, and of 0.69, thus the secondary flows are expected to affect
81×101×121 mesh points respectively, for a total of the aerodynamic performance at mid-span.
about 5.3 million cells. The mesh was selected on the
basis of previous grid-dependence analyses. The y+ val- Results and discussion
ues of the mesh nodes closest to the wall were below 0.3
along the whole blade surface at the highest Reynolds Mid-span flow
number. A three-dimensional view of the mesh is shown A comparison between experimental data and calcula-
in Fig. 1. tions carried out using the LKE model is reported in Fig.
The MISES code [10] was used as comparison for 2 in terms of normalized isentropic Mach number distri-
predicting transitional flow at mid-span. MISES incor- butions on the blade surface. The agreement is good for
porates a modified version of the Abu-Ghannam and all the three tested Mach numbers. It must be noticed that
Shaw [1] by-pass transition model blended with the en the impact of transition on the blade loading distributions
criterion for natural transition. An H-type 311×20 grid is practically negligible. This can be clearly observed
was generated using the code standard parameters, and a from the comparison between Mach number distributions
constant streamtube thickness was adopted. obtained at M2,is = 0.2 from transitional and fully turbu-
lent calculations as shown in Fig. 3. In fact the two cal-
The Linear Cascade culated distributions are almost identical along the entire
blade surface. Only minor differences are visible between
The vane geometry was tested in a 6-blade linear cas- s/stot = 0.35 ÷ 0.45 where a slight local pressure recovery
cade configuration at the Turbomachinery Laboratory of occurs in the transitional distribution but not in the fully
University of Bergamo. The airfoil is representative of a turbulent one (see the enlargement in Fig. 3). This is as-
highly loaded first stage nozzle of a modern heavy-duty sociated with the change in the boundary layer blockage
gas turbine, with a deflection of 70 and an incompressi- due to the transition onset.
Michele Marconcini et al. Transition Modelling Analysis of a Turbine Nozzle Vane Cascade over a Range of M and Re numbers 529

Fig. 4 Measured [5] and computed loss coefficient distribu-

Fig. 2 Measured [5] and computed (LKE) blade load distribu- tions at mid-span (M2,is = 0.2, Re2,is = 6.5×105 ).
tions for M2,is = 0.2 ÷ 0.6.

According to the experiments the kinetic energy loss

coefficient is used to evaluate the performance:
 1  1
 p2    p  
    2 
U 22  pt 2   p t1 
 1   1
(1)
U 22,is ,ms
 p2,ms  
1   
 pt1 
where U2,is,ms is the exit isentropic velocity evaluated at
mid-span, and p2,ms the static pressure at mid-span. Fig. 5 Predicted skin friction coefficient distributions at mid-
Figure 4 compares the measured and computed loss span with transitional and fully turbulent flow (M2,is =
coefficient pitch-wise distributions at mid-span for M2,is = 0.2, Re2,is = 6.5×105).
0.2. The solutions obtained with the transition models are
in good agreement with the experimental distribution suction side boundary layer is predicted to remain lami-
both for the peak value and the wake width. The fully nar up to the 65% of the axial chord by both the LM and
turbulent solution (FT) predicts a thicker wake with LKE models, while transition is only slightly anticipated
higher losses, this behavior is probably associated with in the MISES calculation. More pronounced discrepan-
the presence of a laminar boundary layer over a certain cies between predictions can be observed along the blade
portion of the blade surface. The LKE model exhibits pressure side. Here the skin friction coefficient shows a
slightly higher losses with respect to the LM one, sug- laminar behavior for both the LM model and MISES.
gesting a different prediction of boundary layer transition Transition onset is instead detected around the 60% of
path on the blade surface. the axial chord by the LKE model. However, as the fully
In order to investigate further on this issue, blade sur- turbulent level is never reached by the skin friction coef-
face skin friction coefficient distributions are reported in
ficient, transition is not expected to be completed at the
Fig. 5, for M2,is = 0.2. Results obtained with the MISES
blade trailing edge.
code are also reported for the sake of comparison. The
Along the blade suction side the predicted transition
location moves upstream as the exit isentropic Mach
number is increased. This can clearly be observed in
Fig.6 which compares the skin friction coefficient distri-
butions for the transitional calculations at M2,is = 0.2 and
M2,is = 0.6.
In this latter case all the computations give almost
identical transition onset locations at about the 55% of
the blade axial chord. Such a behavior is related to the
higher adverse pressure gradient experienced by the suc-
tion side boundary layer downstream the blade throat as
the exit isentropic Mach number is increased.
Fig. 3 Measured [5] and computed blade load distributions for
In fact, going from M2,is = 0.2 to M2,is = 0.6 the peak
M2,is = 0.2 with transitional (LKE) and fully turbulent load on the suction side moves downstream with an in-
(FT) flow. crease in the diffusion rate:
530 J. Therm. Sci., Vol.24, No.6, 2015

M max  M TE sTE The boundary layer thickness appears to be slightly

DR  (2)
M TE sTE  smax underestimated in ①, and slightly overestimated in ②.
which passes from 0.64 to 0.90. This circumstance is confirmed by the comparison
The impact of transition modeling is particularly evi- between calculated and measured turbulence intensity
dent in the boundary layer velocity profiles. These, for distributions within the boundary layer reported in
both the investigated transition models, are compared to Fig.8(a) for the same locations along the blade suction
experiments, in selected locations along the blade surface, side, while the ones relative to the pressure side are re-
in Fig. 7. Such locations correspond to the following axi- ported in Fig. 7(b). In the location ① both the boundary
al chord fractions: ① - x/cx = 0.67, ② - x/cx = 0.82, ③ layer thickness and the Tu peak near the wall appear to be
underestimated. In the location ② the Tu level is well
- x/cx = 0.89, ④ - x/cx = 0.95 on the blade suction side
reproduced by the transitional calculations, but the com-
and ① - x/cx = 0.59, ② - x/cx = 0.92 on the pressure
puted boundary layer thickness is actually slightly larger
side. Only results for M2,is = 0.2 are discussed for the
than the experimental one. Both the LM and LKE predic-
sake of conciseness. Computed and experimental results
tions are in good agreement with experiments in the loca-
at specified suction side locations are reported in Fig.
tions ③ and ④. This indicates that despite the differ-
7(a). Here the LM and LKE models yield almost coinci-
ences probably occurring in the calculated and actual
dent results. Discrepancies between computations and
transition paths, that affect the locations ① and ②, the
measurements are detected in the locations ① and ②.
post transition evolution of the boundary layer in the
These seem to be associated with different boundary layer
numerical predictions is similar to the one in the experi-
thicknesses between calculations and experimental data.
mental setup. The fully turbulent calculation is substan-
tially off both in terms of velocity and turbulence inten-
sity profiles within the suction side boundary layer. The
strong mismatch with measurements that is detected in
locations ① and ② is somewhat attenuated in loca-
tions ③ and ④, where the actual suction side boundary
layer has already undergone transition and is growing
with a turbulent trend. Barigozzi and Ravelli [6] drew
similar conclusions by using a fully turbulent calculation
with a realizable k- turbulence closure.

Fig. 6 Predicted skin friction coefficient distributions at mid-

span with M2,is = 0.20.6.

Fig. 7 Measured [6] and computed velocities along the blade Fig. 8 Measured [6] and computed turbulence levels along the
(a) suction side and (b) pressure side (M2,is = 0.2, Re2,is blade (a) suction side and (b) pressure side (M2,is = 0.2,
= 6.5 × 105). Re2,is = 6.5 × 105).
Michele Marconcini et al. Transition Modelling Analysis of a Turbine Nozzle Vane Cascade over a Range of M and Re numbers 531

On the blade pressure side the best agreement with the secondary flow towards mid-span. All the calculated
experiments in terms of boundary layer velocity profiles distributions are characterized by two distinct loss peaks.
is attained with the LM model (Fig. 7(b)). Fully turbulent The first one is located at about z/H = 0.15 where a peak
and LKE calculations yield almost identical results that in the experimental distribution is also found. The second
show a higher velocity gradient near the wall and a larger one is located approximately between z/H = 0.22 and z/H
boundary layer thickness with respect to experiments. = 0.26 for the various calculations and a correspondent
The turbulence intensity distributions of Fig. 8(b) con- peak is not so clearly visible in the experimental distribu-
firm that the LM model predicts a laminar boundary tion.
layer along the whole pressure side. The turbulence in- In order to further investigate on this issue, computed
tensity within the boundary layer is similar for the LKE loss coefficient contours in the measurement plane are
and fully turbulent calculations at both locations ① and compared to experimental ones in Fig. 10. The double
②. However experiments show a higher Tu level with peak structure of the predicted secondary flow is, more or
respect to all the calculations and with a peak value less, visible in all the calculations. The first peak in the
which is very similar in the two considered locations. tangentially averaged loss distributions is associated with
The measured Tu distribution in the location ② is well the strong passage vortex, while the second one is asso-
reproduced by fully turbulent and LKE calculations even ciated with the horseshoe vortex (e.g. [14]). While the
if the corresponding velocity profile is somewhat mis- structure of the secondary loss cores is quite similar be-
matched. This is suspected to be related to differences in tween computations and experiments, the measured high
the boundary layer response to the freestream turbulence loss regions on the suction side of the wake show only a
between the calculations and the experimental setup. distinct loss peak. As expected the best reproduction of
the secondary loss structure is obtained with the LM
Secondary flows model (Fig. 10(c)). Actually the position of the loss core
The vane aerodynamic analyses were completed with associated with the passage vortex is correctly captured
the secondary flow study. A detailed experimental survey both in the pitch-wise and span-wise direction by all the
was available at a section located at 53%cx downstream predictions, but the LKE (Fig. 10(d)) and fully turbulent
the blade trailing edge. For M2,is = 0.2, the span-wise (Fig. 10(b)) calculations seem to overestimate the effects
distortion in the blade-to-blade flow angle is reasonably associated with the horseshoe vortex. Close to the end-
captured by all the calculations, as can be observed in Fig. wall, the strong loss core associated with the corner vor-
9(a), where the deviation from the flow angle at mid-span tex is correctly captured by all the calculations.
is reported as a function of the blade span for half of the At M2,is = 0.6 the span-wise distortion of the exit flow
blade height. The fully turbulent calculation only slightly angle is again correctly reproduced by all the calculations
overestimates the overturning near the endwall. Span- (Fig. 11(a)). As far as span-wise loss distributions are
wise secondary loss coefficient distributions are reported concerned, the slight reduction in the secondary flow
in Fig. 9(b). The secondary loss coefficient is defined by penetration with respect to the case at M2,is = 0.2 is re-
subtracting the mid-span ζms loss coefficient from the produced by all the calculations (Fig.11(b)). Relative to
total: ζs = ζ - ζ ms. Both ζ and ζms are tangentially the case at M2,is = 0.2 now predicted span-wise loss dis-
mass-averaged values. tributions look very similar. This is suspected to be re-
Computed secondary loss coefficient distributions lated to the anticipated transition (see the skin friction
show a much pronounced dispersion with respect to the coefficient distributions of Fig. 6). The decreased fraction
flow angle ones. The best agreement with experiments is of laminar flow in the suction side boundary layer would
obtained with the LM model, while LKE and fully tur- have a reduced effect on the secondary flow development
bulent calculations tend to predict a higher penetration of along the blade passage. The double-peak structure is still

Fig. 9 Span-wise distributions of measured [5] and computed (a) flow angle (b) loss coefficient (M2,is = 0.2, Re2,is = 6.5 × 105).
532 J. Therm. Sci., Vol.24, No.6, 2015

Fig. 10 Kinetic energy loss coefficient contours at x/cx= 1.53, M2,is = 0.2, Re2,is = 6.5 × 105

Fig. 11 Spanwise distributions of of measured [5] and computed (a) flow angle (b) loss coefficient (M2,is = 0.6, Re2,is = 1.6 × 106).

present and the loss peak associated with the passage

vortex appears to be underestimated by the calculations.
Figure 12 reports computed axial distributions of total
and net secondary loss coefficients. Net secondary loss
coefficient is defined by subtracting the inlet ζ1 and
mid-span ζms loss coefficient from the total: ζs = ζ- ζms- ζ 1.
As already mentioned the higher loss level is entitled to
the fully turbulent calculation, while LM and LKE pre-
dictions show very similar total loss distributions. As far
as net secondary loss coefficient is concerned, all the
calculations yield practically the same distribution. This
suggest that the major contribution to the total loss dif- Fig. 12 Mass-averaged total (ζ) and net secondary loss (ζs)
ferences between the calculations comes from the profile coefficient (M2,is = 0.2, Re2,is = 6.5 × 105).
loss while the secondary flows are relatively unaffected
by the transition path for given exit isentropic Mach and performed on a gas turbine nozzle vane cascade using the
Reynolds numbers. TRAF code. A novel laminar kinetic energy transport
model and the widely used Langtry-Menter γ−Re,t model
Conclusions were employed in order to account for transition.
Computed results were assessed against detailed mea-
Three-dimensional steady transitional calculations were surements concerning blade loading distributions, boun-
Michele Marconcini et al. Transition Modelling Analysis of a Turbine Nozzle Vane Cascade over a Range of M and Re numbers 533

dary layers, and secondary flows taken over a range of [4] T. Arts, M. Lambert de Rouvroit, and A. W. Rutherford.
Mach and Reynolds numbers. The analysis of the mid- Aero–Thermal Investigation of a Highly Loaded Tran-
span section of the blade with the MISES code was also sonic Linear Turbine Guide Vane Cascade. Technical note
carried out for comparison. The inclusion of transition 174, von Kármán Institute, Brussels, Belgium, 1990.
modeling was recognized to be crucial for the prediction [5] G. Barigozzi and S. Ravelli. Effects of Trailing Edge Film
of the aerodynamic losses that were found to be substan- Cooling on the Aerodynamic Performance of a Nozzle
tially overestimated by the fully turbulent approach. To Vane Cascade. 9th European Turbomach. Conf., Istanbul,
this purpose the performance of the different transition Turkey, pages 437–447, 2011.
models were found to be very similar. Transition model- [6] G. Barigozzi and S. Ravelli. The Effect of Turbulence
ing was also recognized as determinant for the correct Models on CFD Predictions of the Flowfield in a Turbine
predictions of boundary layer velocity/turbulence inten- Nozzle Vane Cascade. paper n.104, 2011. ISAIF10, July
sity profiles. The best agreement with experiments was 4–7, Brussels, Belgium.
achieved with the LKE model which correctly predicted [7] A. Benyahia and R. Houdeville. Transition Prediction in
the occurrence of the boundary layer transition on the
Transonic Turbine Configurations Using a Correlation-
blade pressure side. Reasonable agreement with mea-
Based Transport Equation Model. 45th Applied Aerody-
surements was also obtained in terms of span-wise angle
namics Symposium Conf. Proc., Marseille, France, 22–24
distortions and secondary losses. Transition modelling
March., 2010.
allowed to reproduce the correct span-wise penetration of
[8] A. J. Chorin. A Numerical Method for Solving Incom-
the secondary flows. Also the predicted secondary vortex
pressible Viscous Flow Problems. J. Comput. Phys., 2:
structures were in better agreement with the experiments
12–26, 1967.
with respect to the fully turbulent approach, with the
[9] D. Corriveau and S. A. Sjolander. Influence of Loading
Langtry-Menter model giving a slightly more reliable
prediction. Net secondary losses were found to be largely Distribution on the Off-Design Performance of High-
insensitive to boundary layer transitions, while the extra Pressure Turbine Blades. J. Turbomach., 129(3), 2007.
losses resulting from the use of a fully turbulent approach [10] M. Drela and H. Youngren. A User’s Guide to MISES
mainly comes from the overestimation of the blade 2.63. MIT Aerospace Computational Design Laboratory,
boundary layers and wake contribution. 2008.
[11] A. Duden and L. Fottner. Influence of Taper, Reynolds
Acknowledgements Number and Mach Number on the Secondary Flow Field
of a Highly Loaded Turbine Cascade. Proc. I. Mech. E.
The work presented in this paper was performed Part A: J. Power and Energy, 211(4): 309–320, 1997.
within the framework of the project INSIDE (Aerother- [12] R. Gomes and R. Niehuis. Aerothermodynamics of a
mal Investigation of cooled Stage turbIne: Design opti- High-Pressure Turbine Blade With Very High Loading
mization and Experimental analysis) PRIN 2011 and Vortex Generators. ASME paper GT2010-23543, 2010.
n.2010K3B4RL funded by the Italian Ministry of In- [13] R. B. Langtry and F. R. Menter. Correlation-Based Tran-
struction, University and Research (MIUR). sition Modeling for Unstructured Parallelized Computa-
The authors would like to thank Prof. Giovanna Bari- tional Fluid Dynamics Codes. AIAA J., 47(12): 2894–
gozzi and Dr. Silvia Ravelli of University of Bergamo for 2906, December 2009.
the useful discussion on the experimental data. [14] P. Marchal and C. H. Sieverding. Secondary Flows on
Turbomachines - 11 Secondary Flows Within Turboma-
References chinery Bladings. AGARD-CP-214, 1977.
[15] P. Martini, A. Schulz, H.-J. Bauer, and C. Whitney. De-
[1] B. J. Abu Ghannam and R. Shaw. Natural Transition of tached Eddy Simulation of Film Cooling Performance on
Boundary Layers – The Effect of Turbulence, Pressure the Trailing Edge Cutback of Gas Turbine Airfoil. J.
Gradient and Flow History. Journal of Mechanical Engi- Turbomach., 128(2): 292–299, 2006.
neering Science, 22(5): 213–228, 1980. [16] H. Moore and D. G. Gregory-Smith. Transition Effects on
[2] A. Arnone. Viscous Analysis of Three–Dimensional Ro- Secondary Flows in a Turbine Cascade. ASME paper 96-
tor Flow Using a Multigrid Method. J. Turbomach., GT-100, 1996.
116(3): 435–445, July 1994. [17] R. Pacciani, M. Marconcini, A. Arnone, and F. Bertini.
[3] A. Arnone, M. S. Liou, and L. A. Povinelli. Navier– An Assessment of the Laminar Kinetic Energy Concept
Stokes Solution of Transonic Cascade Flow Using Non– for the Prediction of High-Lift, Low-Reynolds Number
Periodic C–Type Grids. J. Propul. Power, 8(2): 410–417, Cascade Flows. Proc. I. Mech. E. Part A: J. Power and
March–April 1992. Energy, 225(7): 995–1003, 2011.
534 J. Therm. Sci., Vol.24, No.6, 2015

[18] R. Pacciani, M. Marconcini, A. Fadai-Ghotbi, S. Lardeau, Turbomach., 134(6): 061037, 2012.

and M. A. Leschziner. Calculation of High-Lift Cascades [20] C. H. Sieverding. Advanced Methods for Cascade Testing
in Low Pressure Turbine Conditions Using a Three-Equa- - 2.3 Subsonic-Choked (Turbine) Cascades. AGARD-
tion Model. J. Turbomach., 133(3): 031016, July 2011. AG–328, 1993.
[19] R. Pacciani, F. Rubechini, A. Arnone, and E. Lutum. [21] D. C. Wilcox. Turbulence Modeling for CFD. DCW Ind.
Calculation of Steady and Periodic Unsteady Blade Sur- Inc., La Cañada, CA, USA, 2nd edition, 1998. ISBN 1-
face Heat Transfer in Separated Transitional Flow. J. 928729-10-X.