You are on page 1of 10

49th AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition AIAA 2011-498

4 - 7 January 2011, Orlando, Florida

Conjugate Heat Transfer Simulation of Cooled Turbine


Blades Using Unstructured-Mesh CFD Solver

Tetsuya Yoshiara1, Daisuke Sasaki2, Kazuhiro Nakahashi3


Department of Aerospace Engineering, Tohoku University, Sendai, Miyagi, 980-8579, JAPAN

Conjugate heat transfer simulation method was implemented in the unstructured-mesh


CFD solver to predict the thermal loads of cooled turbine vanes, Mark II and C3X. In this
simulation, the fluid domain and solid domain are solved simultaneously satisfying the
continuity of the heat flux and temperature on the boundary of the fluid and solid. In
addition, -Re transition model is applied to predict the boundary layer transition which
often occurs in the flow around turbine blades and affects the material thermal load
seriously. The predicted thermal load results from the two- and three-dimensional
computations are in close agreement with the measurement data capturing the transition on
the suction side of the vanes.

I. Introduction
In order to develop more efficient gas turbine engines, it is especially required to achieve higher turbine inlet
temperature. Highly increased turbine inlet temperature, however, causes the turbine materials to be damaged or
depleted because of the severe thermal loads. Until now, turbine inlet temperature has increased due to the
development of the high temperature materials and cooling systems. For the further requirements in the future, it is
important to predict the thermal loads of the turbine blades accurately to design the enhanced-performance cooling
systems and to estimate the damages, degradation, and lifetime of the materials. As the thermal loads prediction
method, the conjugate heat transfer (CHT)1 simulation, in which the fluid domain and solid domain are solved
simultaneously, is known to be effective. In the conventional calculation method, the flow domain is solved to
calculate the heat transfer coefficient distribution at first using the wall boundary temperature that is assumed before
calculation, and then the resultant heat transfer coefficient and fluid reference temperature are used as the boundary
conditions to calculate the solid heat conduction. The wall temperature is assumed by the empirical correlations,
thus there exists the uncertainty and this makes it quite difficult to predict the thermal loads particularly for the
complicated geometries. On the other hand, in CHT the wall temperature and heat transfer coefficient are not
prescribed as the boundary condition but are the solutions of the whole calculation. Thus the uncertainty can be
removed in CHT and can be applied to the arbitrary geometries and conditions.
The configurations of turbine blades are very complicated because of their cooling systems. For flow simulations,
the unstructured-mesh solvers have the advantage over structured-mesh solvers in terms of the easy mesh generation
for complex geometries. Also, the unstructured-mesh CFD solver, TAS-code (Tohoku university Aerodynamic
Simulation code2) developed at Tohoku University, has been used for the flow analysis and design of a commercial
airplane practically and is known to be a reliable flow simulation code.
It is necessary for the CFD code which is based on the CHT simulation to capture the flow characteristics around
turbine blades accurately. In particular, since the Reynolds number of the flow around turbine blades is about 105
106, there is the laminar boundary layer and the heat transfer coefficient increases sharply when the transition occurs.
Hence, the prediction of transition is important in terms of the thermal load prediction. In this study, -Re
transition model3, 4 proposed by Menter et al. is applied. Since this model can predict the transition using only the
local variables, it is easy to implement in the unstructured-mesh solver.
The objective of the present study is to develop and validate the conjugate heat transfer simulation code based on
the unstructured-mesh CFD solver applying -Re transition model aiming to predict the thermal loads of cooled

1
Graduate Student, Student member, yoshiara@ad.mech.tohoku.ac.jp.
2
Assistant Professor, Senior member AIAA.
3
Professor, Associate Fellow AIAA.
1
American Institute of Aeronautics and Astronautics

Copyright 2011 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
turbine blades. The code validation is conducted using two and three-dimensional turbine vanes, NASA Mark II and
C3X which have relatively simple geometries and internal convection cooling systems.

II. Numerical Method

A. Fluid calculation
The fluid domain is solved using three-dimensional unstructured-mesh CFD solver, TAS-code. The governing
equations are the compressible Reynolds-Averaged Navier-Stokes equations. The calculation domain is discretized
by means of the cell-vertex finite volume method. Inviscid flux evaluation method is HLLEW5 and LU-SGS6
implicit method is used for time integration. Spalart-Allmaras turbulence model7, SST turbulence model8, and-Re
3, 4
transition model coupled with SST turbulence model are used for the comparisons.

B. Solid calculation
The governing equation for the solid domain is the heat conduction equation. The solid domain is discretized by
the same method as the fluid domain using unstructured mesh. Thus, this conjugate code is homogeneous method.
The heat conduction equation is obtained from degenerating the Navier-Stokes equations setting velocity zero and
inviscid flux zero in the solid domain.

C. Conjugate Heat Transfer method


On the boundary between the fluid domain and solid domain, it is necessary to satisfy the continuity of the heat
flux and temperature. Figure 1 shows the cells on the boundary of the fluid and solid. Fluid static temperature TF,
solid temperature TS, and wall boundary temperature TW (unknown) are defined as shown in Fig. 1. Vertical
distances LF and LS are measured from the wall boundary to the definition points of TF and TS respectively. The heat
flux qF between the fluid node and wall surface coincides with the heat flux qS between the solid node and wall
surface. This condition is represented as the following equation;
TF TW T TS
q k F k S W (1)
LF LS
where kF and kS are the thermal conductivities of the fluid and solid respectively. The unknown wall temperature is
obtained from solving Eq. (1) and this wall temperature is used for the boundary condition of the fluid domain and
solid domain, thus the continuity of the temperature on the wall is satisfied,
TW
k S / k F LF / LS TS TF (2)
1 k S / k F LF / LS
where kF is the thermal conductivity of the fluid and is represented as
the following relation;
(3)
kF Cp
Pr
where is viscosity coefficient and Cp is specific heat at constant
pressure. Pr is Prandtl number for air. Laminar viscosity coefficient l
is given by the following Sutherland formula:
3/ 2
l T T0 S
(4)
0 T0 T S Figure 1. The cells on the boundary of
where 0 = 1.7910-5 [kg/ms], T0 = 288.15 [K], S = 110.4 [K]. the fluid and solid.

D. -ReTransition model
-Re transition model proposed by Menter et al. uses the empirical correlations of the momentum-thickness
Reynolds number Re to decide the transition onset position. This model uses the relation (5) between the
momentum-thickness Reynolds number Re and the vorticity Reynolds number Re so that it is not necessary to
calculate the momentum thickness that is an integral parameter. This makes it easy to implement in unstructured-
mesh solvers and to handle a parallelization.
max Re (5)
Re
2.193
2
American Institute of Aeronautics and Astronautics
At first, the distribution of transition onset momentum-thickness Reynolds number Re t is obtained from the
empirical correlation based on the local turbulent intensity Tu. Ret is the function of Tu and pressure gradient
~
parameter. Then, the transport equation for local transition onset momentum-thickness Reynolds number Ret is
solved using resultant Ret to calculate the distribution of Ret in the boundary layer. Re is computed from the
relation (5) using Re that can be obtained from the local fluid variables and then the transition onset is evaluated by
~
comparing Re with Ret. The transition region is simulated by solving the transport equation of the intermittency
. In-Re transition model, the following two transport equations are solved in addition to the two equations of k
and of SST turbulence model.

~
Re t U ~
Re t
~
Re t
t t
j
P t (6)
t x j x j x j

U j

t P E
x
(7)
t x j x j f j
~
The effect of Tu is taken into account by the source term Pt in Eq. (6) to calculate the distribution of Ret in the
boundary layer. The term P is to accelerate the transition and D is to relaminarize the flow. Once the transition
onset position is determined, the intermittency rises and increases along the downstream.
-Re transition model is coupled with SST turbulence model through the intermittency as follows.

k U j k k ~ ~
k t Pk Dk (8)
t x j x j x j
Pk eff Pk ; Dk min max eff , 0.1,1.0 Dk
~ ~
(9)

The modified intermittency eff is to take into account the predictions of separated flow transition. The details of
the transition model including the correlations or the treatment of separated flow transition can be found in Ref. 3, 4.

III. Two-dimensional Calculations


The internally cooled turbine vanes, NASA Mark II and C3X9, are chosen because the detailed measurement
data are available and these models were often used to validate CHT calculations in the past1, 10, 11 . The
computational meshes are shown in Fig. 2. DotEditor12 was used to generate two-dimensional unstructured meshes.
On the boundary, the coordinates of nodes in the fluid domain correspond with those in the solid domain. The fluid
domain uses a hybrid mesh in order to capture the boundary layer and the solid domain uses only triangular elements.
The near-wall spacing of the fluid is below y+ of 1.0. The total pressure, total temperature and flow angle are given
at the inlet boundary and the static pressure at the exit boundary. The exit static pressure is calculated by the exit
Mach number using isentropic relation. For the circumferential direction, periodic boundary condition is imposed.
The flow conditions are shown in Table 1. Both the cases have ten cooling holes and the flows in all the cooling
holes are assumed to be fully developed turbulence. The Nusselt number for turbulent flow in a smooth pipe is
represented as the following empirical formula:
0.8
Nu D 0.022Cr Pr 0.5 Re D (10)
where Cr is the corrective coefficient and ReD is Reynolds number based on the diameter of each cooling hole. Pr is
Prandtl number of 0.9 for turbulent air flow. The heat transfer coefficient is obtained from Nusselt number Eq. (10)
and it is used as the boundary condition for the solid domain. The coolant condition for each hole including hole
diameter D, Cr, coolant temperature TC and ReD are shown in Table 2. The turbine material is Stainless steel ASTM
Type 310 and the thermal conductivity kS is specified as follows:
k S 0.0182 T 6.13 [W/m/K] (11)

3
American Institute of Aeronautics and Astronautics
1
3 1
2 2

4 3

4
5
5
6
6

7
7
8
8

9
9

10
10

Figure 2. Two-dimensional computational meshes.

Table 1. Computational conditions


Mark II C3X
Test Code 5411 4311
Inlet Total Pressure [MPa] 0.337 0.245
Inlet Total Temperature [K] 788 802
Outlet Static Pressure [MPa] 0.70 0.143
Outlet Mach number 1.04 0.91
6
Reynolds number 3.3010 3.07106
Inlet Turbulence Intensity [%] 6.5 6.5
Inlet Flow Angle [deg] 0 0
Solidity 1.05 1.24
Number of Nodes (Fluid) 46000 51000
(Solid) 12000 14000

4
American Institute of Aeronautics and Astronautics
Table 2. Coolant conditions
Hole No. D Cr TC ReD
[cm] [K] [10-4]
Mark II C3X Mark II C3X
1 0.630 1.118 336.390 360.630 24.686 15.494
2 0.630 1.118 326.270 360.630 24.408 16.162
3 0.630 1.118 332.680 346.230 24.097 15.974
4 0.630 1.118 338.860 352.500 24.735 15.877
5 0.630 1.118 318.950 341.410 24.335 17.071
6 0.630 1.118 315.580 380.080 24.048 16.158
7 0.630 1.118 326.260 352.470 24.453 16.234
8 0.310 1.056 359.830 387.160 14.967 10.434
9 0.310 1.056 360.890 421.810 9.285 6.396
10 0.198 1.025 414.850 466.790 9.205 4.685

Figure 3 shows Mach number contours by -Re transition model for Mark II and C3X. The pressure
distributions on the vane surfaces from the three turbulence models are shown in Fig. 4. The horizontal axis is the
axial distance from the leading edge normalized by the axial chord of each vane. The two shock waves are observed
at the fore of the vane and near the trailing edge on the suction side in the case of Mark II. The one near the axial
distance of 0.4-0.5 is relatively strong and this indicates the great increase of the surface pressure. In contrast, the
surface pressure of C3X vane shows the mild distribution since C3X has the different flow path from that of Mark II
and the smaller exit Mach number. The results of the pressure distributions from the three turbulence models are
almost identical and are in good agreement with the experimental data. Thus, the fluid calculation part in the CHT
simulation is validated in the two-dimensional simulation. Also, these results are almost the same as those of the
flow only simulation assuming the adiabatic wall boundary condition as shown in Fig. 4. The difference of the wall
thermal condition little affects the main flow field and the surface pressure distribution.
Figure 5 shows the static temperature contours by-Re transition model for Mark II and C3X. Figure 6 is the
external heat transfer coefficient distributions and Fig. 7 is the surface temperature distributions compared with the
experimental measurement data. The large fluctuations of the heat transfer coefficient and surface temperature near
the trailing edges are due to the effect of the cooling holes in the vanes and simulated the test results well. These
results show the differences among the three turbulence models unlike the results of the pressure distributions.
Especially the results from SA and SST turbulence models assuming the fully turbulent flow show the different
characteristics of heat transfer coefficient and temperature on the suction side from the measured data since these
models cannot capture the transition from laminar to turbulent flow near the axial distance of 0.4. On the other hand,
the result from-Re transition model shows the close characteristic to the data capturing the increase of heat
transfer coefficient due to the boundary layer transition. In the case of Mark II, the boundary layer transition occurs
at the position of the fore shock wave and this causes the sharp increase of the heat transfer coefficient. However,
the amount of the increase is evaluated excessively compared with the measurements. In C3X geometry, where no
shock waves exist in the flow field, the calculation using -Re transition model also captured the qualitative trend
well and this denotes that the transition model is important for the prediction of the heat transfer and wall
temperature. With the slight different predicted values from the test data, the positions of the local minimum of the
heat transfer coefficient and wall temperature coincide with the test data and this means that the position of the
transition was captured accurately. The results from these turbulence models show the close values to the data
except for the transition region on the suction side, so the validity of the CHT simulation was confirmed.

5
American Institute of Aeronautics and Astronautics
(a) Mark II (b) C3X
Figure 3. Mach number contours.

1 1
0.9 0.9
0.8 0.8
0.7 0.7
p/pt

p/p t

0.6 0.6
Adiabatic SA Adiabatic SA
0.5 0.5
CHTMark II SA CHTC3X SA
0.4 CHTMark II SST 0.4 CHTC3X SST
CHTMark II SST+-Re CHTC3X SST+-Re
0.3 EXPMark II 0.3 EXPC3X

0.2 0.2
-1 Pressure Side 00 Suction Side 1 -1 Pressure Side 00 Suction Side 1
(a) Mark II (b) C3X
Figure 4. Surface pressure distributions.

6
American Institute of Aeronautics and Astronautics
(a) Mark II (b) C3X
Figure 5. Static temperature contours.

1 1
CHTMark II SA CHTC3XSA
0.9 CHTMark II SST 0.9 CHTC3XSST
CHTMark II SST+-Re CHTC3XSST+-Re
0.8 EXPMark II 0.8
EXPC3X
h/ (1135 W/m2/K)

0.7 0.7
h/ (1135 W/m 2 /K)

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
-1 Pressure Side 00 Suction Side 1 -1 Pressure Side 00 Suction Side 1
(a) Mark II (b) C3X
Figure 6. External heat transfer coefficient distributions.

0.85 0.85
CHTMark II SA CHTC3X SA
CHTMark II SST CHTC3X SST
0.8 0.8 CHTC3X SST+-Re
CHTMark II SST+-Re
EXPMark II EXPC3X
0.75 0.75
T/ (811 K)
T/ (811 K)

0.7 0.7

0.65 0.65

0.6 0.6

0.55 0.55
-1 Pressure Side 0
0 Suction Side 1 -1 Pressure Side 0
0 Suction Side 1
(a) Mark II (b) C3X
Figure 7. Surface temperature distributions.

7
American Institute of Aeronautics and Astronautics
IV. Three-dimensional Calculations
The C3X turbine vane which models the experimental circumstance was used for the three-dimensional
calculation. The three-dimensional computational mesh was obtained by extruding the two-dimensional mesh
toward the spanwise direction. MEGG3D13 was used for the three-dimensional unstructured mesh generation. The
computational mesh is shown in Fig. 8 and consists of hexahedral elements near wall and prismatic elements for the
rest of the domain. Although the mesh near the side wall in the spanwise direction in the main flow domain is very
fine in order to capture the boundary layer, such high resolution in this direction is not required in the solid domain
and the mesh is coarsened as shown in Fig. 8. In this near-wall area where the nodes of the main flow domain and
those of the solid domain do not coincide, the wall temperature is interpolated from the solid domain to the main
flow domain. The number of nodes is 2,200,000 for the main flow domain, 300,000 for the solid domain.
The coolant conditions are the same as that of the two-dimensional calculation, in which the Nusselt number
based on the empirical correlation is imposed on the internal wall surfaces. The calculation conditions of the main
flow domain are the same as that of the two-dimensional calculations and the side wall is assumed to be an adiabatic
non-slip wall.

Figure 8. Three-dimensional computational mesh.

Figures 9 and 10 show Mach number contours and static temperature contours results respectively from -Re
transition model at midspan. The pressure distributions at midspan are shown in Fig. 11 and are in close agreement
with the measurement data. The difference from the two-dimensional result is seen near the trailing edge on suction
side predicted by SA turbulence model.
Figure 12 shows the external heat transfer distribution at midspan and Fig. 13 is the surface temperature
distribution at midspan. Although the wall temperature result from SA turbulence model is lower than the data near
the stagnation point and on the suction side, these results show the similar characteristics to those of the two-
dimensional calculations about each turbulence model. The boundary layer transition on the suction side is captured
also in the three-dimensional simulation using the -Re transition model. The overall view of the surface
temperature contours is shown in Fig. 14 and the distribution is symmetric because the identical coolant condition
along the spanwise direction is set using the empirical correlation of Nusselt number. The slightly higher
temperature shown near the side wall is due to the secondary flow generated by the side wall and vane surface.

8
American Institute of Aeronautics and Astronautics
Figure 9. Mach number contours at midspan. Figure 10. Static temperature contours at midspan.

1 1
CHTC3XSA
0.9 0.9 CHTC3XSST
CHTC3XSST+-Re
0.8
0.8 EXPC3X
h/ (1135 W/m2/K)

0.7
0.7 0.6
p/pt

0.6 0.5
0.5 CHTC3X SA 0.4
CHTC3X SST 0.3
0.4 CHTC3X SST+-Re 0.2
0.3 EXPC3X
0.1
0.2 0
-1 Pressure Side 00 Suction Side 1 -1 Pressure Side 0
0 Suction Side 1

Figure 11. Surface pressure distributions at Figure 12. External heat transfer coefficient
midspan. distributions at midspan.

0.85
CHTC3X SA
CHTC3X SST
0.8 CHTC3X SST+-Re
EXPC3X
0.75
T/ (811 K)

0.7

0.65

0.6

0.55
-1 Pressure Side 0
0 Suction Side 1

Figure 13. Surface temperature distributions at midspan. Figure 14. Surface temperature contours.

9
American Institute of Aeronautics and Astronautics
V. Conclusion
The conjugate heat transfer method was implemented in the unstructured-mesh CFD solver TAS-code and the
prediction of thermal load of the internally cooled turbine vanes was conducted using three turbulence models. One
of them is the -Re transition model coupled with SST turbulence model. In two-dimensional calculations, it was
shown that the difference of the turbulence models had little effect on the pressure distributions. However, the
results from the transition model captured the thermal properties of the test data well for the two different turbine
geometries and flow conditions. In three-dimensional calculation, the results show the similar characteristics to the
two-dimensional calculation capturing the transition onset position accurately on the suction side. Thus, the method
in the present paper revealed the importance of the boundary layer transition prediction for accurate prediction of the
thermal properties of turbine blades using the conjugate heat transfer method.

References
1
Bohn, D, Krger, U, and Kusterer, K, Conjugate Heat Transfer: An Advanced Computational Method for the cooling
design of modern gas turbine blades and vanes, Heat Transfer in Gas Turbines, eds. B. Sundn and M. Faghri, pp. 57-108, WIT
Press, Southampton, UK, 2001.
2
Nakahashi, K., et al. Some Challenges of Realistic Flow Simulations by Unstructured Grid CFD, International Journal for
Numerical Method in Fluids, Vol. 43, 2003, pp. 769-783.
3
Menter, F. R., Langtry, R. B., Likki, S. R., Suzen, Y. B., Huang, P. G., and Vlker, S., A Correlation Based Transition
Model Using Local Variables Part 1: Model Formulation, Journal of Turbomachinery, Vol. 128, No. 3, 2006, pp. 413422.
4
Langtry, R. B., Menter, F. R., Correlation-Based Transition Modeling for Unstructured Parallelized Computational Fluid
Dynamics Codes, AIAA Journal, Vo. 47, No. 12, 2009, pp. 2894-2906.
5
Obayashi, S. and Guruswamy, G. P., Convergence Acceleration of an Aeroelastic Navier-Stokes Solver, AIAA Paper 94-
2268, 1994.
6
Sharov, D., and Nakahashi, K., Reordering of Hybrid Unstructured Grids for Lower-Upper Symmetric Gauss-Seidel
Computations, AIAA Journal, Vol.36, No.3, 1998, pp. 484-486.
7
Spalart, P. R. and Allmaras, S. R., A One-Equation Turbulence Model for Aerodynamic Flows, AIAA Paper 92-0439,
January 1992.
8
Menter, F. R., Two-equation eddy-viscosity turbulence models for engineering applications, AIAA Journal, Vo. 32, No. 8,
1994, pp. 1598-1605.
9
Hylton, L. D., Mihelc, M. S., Turner, E. R., Nealy, D. A., and York, R. E., Analytical and Experimental Evaluation of the
Heat Transfer Distribution Over the Surface of Turbine Vanes, NASA Paper No. CR-168015, 1983.
10
Bamba, T., Yamane, T., Fukuyama, Y., Turbulence Model Dependencies on Conjugate Simulation of Flow and Heat
Conduction, ASME Turbo Expo 2007, GT2007-27824.
11
Luo, J., and Razinsky, E. H., 2007, Conjugate Heat Transfer Analysis of a Cooled Turbine Vane With the V2F Turbulence
Model, ASME Journal of Turbomachinery, Vol. 129, Paper No. GT2006-91109, pp. 773781.
12
Ishida, T., Study of Grid Generation around Complex Geometry, Masters Thesis, Tohoku University, 2008. (in
Japanese)
13
Ito, Y. and Nakahashi, K., Improvements in the Reliability and Quality of Unstructured Hybrid Mesh Generation,
International Journal for Numerical Methods in Fluids, Vol. 45, Issue 1, May 2004, pp. 79-108.

10
American Institute of Aeronautics and Astronautics

You might also like