Conjugate Heat Transfer Measurements and Predictions of A Blade Endwall With A Thermal Barrier Coating
Conjugate Heat Transfer Measurements and Predictions of A Blade Endwall With A Thermal Barrier Coating
Conjugate Heat Transfer Measurements and Predictions of A Blade Endwall With A Thermal Barrier Coating
Amy Mensch
Department of Mechanical
Measurements and Predictions
and Nuclear Engineering,
The Pennsylvania State University, of a Blade Endwall With
University Park, PA 16802
e-mail: [email protected] a Thermal Barrier Coating
Karen A. Thole Multiple thermal protection techniques, including thermal barrier coatings (TBCs), inter-
Department of Mechanical
nal cooling and external cooling, are employed for gas turbine components to reduce
and Nuclear Engineering,
The Pennsylvania State University,
metal temperatures and extend component life. Understanding the interaction of these
cooling methods, in particular, provides valuable information for the design stage. The
University Park, PA 16802
current study builds upon a conjugate heat transfer model of a blade endwall to examine
e-mail: [email protected]
the impact of a TBC on the cooling performance. The experimental data with and without
TBC are compared to results from conjugate computational fluid dynamics (CFD) simu-
Brent A. Craven lations. The cases considered include internal impingement jet cooling and film cooling
Department of Mechanical
at different blowing ratios with and without a TBC. Experimental and computational
and Nuclear Engineering,
results indicate the TBC has a profound effect, reducing scaled wall temperatures for all
The Pennsylvania State University,
cases. The TBC effect is shown to be more significant than the effect of increasing blow-
University Park, PA 16802
ing ratio. The computational results, which agree fairly well to the experimental results,
e-mail: [email protected]
are used to explain why the improvement with TBC increases with blowing ratio.
Additionally, the computational results reveal significant temperature gradients within
the endwall, and information on the flow behavior within the impingement channel.
[DOI: 10.1115/1.4028233]
T1 TTBC
s¼ (5)
T1 Tc;in
Since TTBC is on the same heat transfer circuit used to derive Eq.
(1), it follows that s depends on the same parameters as /TBC.
The nondimensional parameters matched to an engine for the
endwall and TBC layer for both the experiments and simulations
in this study are given in Table 1. The blowing ratio, Mavg, which
affects vg, defines the average ratio of coolant to mainstream
mass flux. The endwall h1 distribution is obtained from measure-
ments made by Lynch et al. [22]. An engine matched Bi is
R
achieved by making the endwall from CorianV, a DuPont material.
To estimate hi for each blowing ratio, a Nu correlation, derived by
Hollworth and Dagan [23] for an impingement array with stag- Fig. 2 Depiction of (a) the large-scale low-speed wind tunnel
and (b) the test section containing the Pack-B linear blade cas-
gered coolant extraction, is used. The thermal effect of a TBC is cade and conjugate endwall
reproduced with a thin layer of cork. The conductivity and thick-
ness of the cork have been chosen to replicate the RTBC/Rw within
the range of a typical engine. view of the test section including the shaded part of the endwall
constructed of Corian. Outside of the passage, the endwall was
constructed from medium density fiberboard (MDF). The center
Experimental Methods passages, 3 and 4, had film and impingement cooling and were
Overall effectiveness (/) and TBC effectiveness (s) were meas- used for this study. A summary of the blade geometry and main-
ured for the endwall of a linear cascade in the corner test section stream flow conditions is given in Table 2. At 0.5Cax upstream of
of the large-scale, low-speed, closed loop wind tunnel shown in the blade leading edge in the axial direction, the d99/S was 0.089
Fig. 2(a). This facility split the flow upstream of the cascade into and the freestream turbulence was 7%, as predicted by the CFD
mainstream and coolant streams. The mainstream section was solution from this study. The details of the simulations are
heated by a heater bank and passed through flow conditioning ele- described in the Computational Methodology section.
ments including a turbulence grid 11Cax upstream of the test Periodicity of the cascade was determined by taking measure-
section. A more detailed description of the wind tunnel and flow ments of the pressure distribution at the midspan of all seven
conditioning elements can be found in Ref. [22]. The mainstream blades. In Fig. 3, a typical set of data for the pressure coefficient,
temperature was measured 0.52Cax in the axial direction upstream Cp, is plotted versus normalized axial distance. The measured Cp
of the blades at multiple spanwise and pitchwise locations with a agreed well with the CFD results from this study, as well as with
thermocouple rake. The mainstream temperature varied by no an inviscid 2D CFD prediction from Ref. [22].
more than 60.6 C from the average T1. A Pitot probe, also The endwall configuration was designed with a generic internal
inserted 0.5Cax upstream, was used to measure the inlet main- and external cooling geometry to replicate the relevant nondimen-
stream velocity, U1. The standard deviation over the mean U1 sional parameters given in Table 1. Figure 4(a) shows the side
was less than 1%. view of the plenum, cooling features, endwall, and the layer of
The coolant passed through a desiccant drier and two heat TBC. A splash plate was used in the experiments to spread and
exchangers that chilled the coolant, before entering the plenum slow the plenum inlet flow. The coolant passed through an
located below the endwall. A laminar flow element, directly meas- impingement plate with 28 staggered holes. The nondimensional
ured the total coolant flowrate, which was adjusted to achieve the height between the impingement plate and the bottom of the end-
necessary blowing ratios. The blowing ratios reported in this wall, was H/D ¼ 2.9. Following impingement, the coolant flowed
study, Mavg, were an average of the local M at each film cooling into ten angled film cooling holes, which were staggered between
hole. The uncertainty in coolant flowrate was estimated for a 95% the impingement holes. The diameter, D, of the impingement and
confidence interval as 63%, using sequential perturbation [28].
To measure the internal coolant temperature, Tc,in, there were two
thermocouples located approximately 8.7D below the impinge- Table 2 Flow conditions and blade geometry
ment plate, which agreed within 6 3 C. The experimental setup
Scale factor 8.6 Inlet U1 10.5 m/s
achieved a typical mainstream to coolant temperature difference
Cax 0.218 m Inlet, flow angle 35 deg
of 40 C, which provided a coolant to mainstream density ratio of p/Cax 0.826 Exit flow angle 60 deg
about 1.15. S/Cax 2.50 Inlet Ma 0.029
The test section contained a seven blade linear cascade based Inlet Re 1.22 105 Exit Ma 0.047
on the low-pressure turbine Pack-B airfoil, which has been used in Exit Re 1.98 105
many other studies [2,8,22,29–37]. Figure 2(b) shows the top
The conduction heat loss from the downstream sides of the end- adiabatic boundary conditions in the simulations, the steady-state
wall into the fiberboard (k ¼ 0.3 W/m2) is estimated to be on the endwall temperature is higher than measured in the experiments.
same order as the convective heat transfer into the endwall from The data in Figs. 8(a)–8(f) give the local / along the length of
the mainstream. The lateral conduction in this part of the endwall the PS and SS streamlines shown in Fig. 4(b). The experimental /
is estimated to be an order of magnitude less because the tempera- without TBC and the predicted / (in red) have been extracted from
ture gradients in this region of the endwall are small. By using the measured and predicted contours in Figs. 6 and 7. The measured
Fig. 7 Overall effectiveness contours for Mavg 5 2.0 (a) measured without TBC, (b) predicted
without TBC, and (c) predicted under the TBC
Fig. 8 Comparison of overall effectiveness with and without TBC, showing measured and predicted values, along inviscid
streamlines, PS (a)–(c) and SS (d)–(f)
IR camera cannot measure /TBC on the endwall. Therefore, the added, there is less warming of the coolant because the TBC
measured /TBC is compared to the case without TBC in Figs. reduces the internal wall temperature.
8(a)–8(f) for the discrete thermocouple measurements made The average increases in / due to changes in the blowing ratio
below the TBC along the inviscid streamlines. Along the inviscid and due to the addition of TBC are compared in Table 3. The
streamlines, higher cooling performance is observed with TBC change due to M, D/M , is an area-average across the impingement
compared to no TBC for all three blowing ratios. TBC increases area from Mensch and Thole [2]. The change due to TBC, D/TBC ,
the effectiveness on the PS line at all measurement locations by is the average difference between /TBC and / (without TBC) at
nearly a constant amount. The increase from / to /TBC varies on the measurement locations along the streamlines. The values in
the SS along the length of the streamline, especially between the table show that the improvement due to the addition of TBC is
0.2 < s/Cax < 0.7. The local peaks from film cooling seen on the about three times greater than the improvement from increasing
endwall without TBC have been smeared out in /TBC because the blowing ratio. Changing blowing ratio is less effective than TBC
TBC is insulating the endwall from the external flow, which at reducing endwall temperatures, which is consistent with the
includes film cooling. findings on the vane surface [10].
The predictions of /TBC in Figs. 6(c), 7(c), 8(b), 8(c), 8(e), 8(f), The improvement in / due to the TBC for the measurements
9(a), and 9(b) also show that the TBC provides a significant cool- and simulations has been plotted in Fig. 10. The values for the
ing effect on the endwall. The predicted contours of /TBC for experiment are those given in Table 3. The predicted data are an
Mavg ¼ 1.0, in Fig. 6(c) and for Mavg ¼ 2.0 in Fig. 7(c), are signifi- average across the entire conducting endwall surface. Figure 10
cantly higher at all locations on the endwall relative to / without demonstrates that D/TBC slightly increases with blowing ratio.
TBC. When the blowing ratio is increased from Mavg ¼ 1.0 The measured and predicted values for improvement with TBC
(Fig. 6(c)) to Mavg ¼ 2.0 (Fig. 7(c)), the area of high effectiveness agree very well, indicating that the discrete thermocouple loca-
under the TBC increases. The predicted /TBC along the inviscid tions used to measure the endwall temperatures under the TBC
streamlines in Figs. 8(b), 8(c), 8(e), and 8(f) (dashed red lines) provide a good indication of the effect throughout the passage.
compare fairly well with the discrete /TBC measurements. The
predictions show some influence of the film cooling on /TBC,
which is not observed in the measurements. Both /TBC and / are
underpredicted downstream in the passage for s/Cax > 0.7. These Table 3 Measured improvement in overall effectiveness due to
discrepancies are attributed to the conduction losses at the endwall increase in M or due to addition of TBC
sides previously discussed in reference to the / contours. When
DM D/M [2] Mavg D/TBC ¼ /TBC /
TBC is included in the slices of nondimensional temperature in
Fig. 9, the endwall temperatures are greatly reduced, even in the 0.6–1.0 0.05 0.6 0.13
areas not around the impingement and film cooling holes. In addi- 1.0 0.14
tion, the fluid in the channel and the film cooling holes remains 1.0–2.0 0.04 2.0 0.17
cooler with TBC than the cases without TBC. When TBC is
Fig. 11 Contours of TBC effectiveness at three blowing ratios, (a) measured Mavg 5 0.6, (b)
measured Mavg 5 1.0, (c) measured Mavg 5 2.0, (d) predicted Mavg 5 1.0, and (e) predicted
Mavg 5 2.0
The corresponding predictions for s show reasonable agreement effectiveness were reasonably close to the measurements, and the
to the experiments. Like the measurements, the simulations pre- correct trends were captured by the simulations. In conclusion, the
dict temperatures on the outside of the TBC to be hotter than the predictive model in conjunction with a suitable grid gave reasona-
bare endwall. However, the predicted s is slightly lower than ble predictions of conjugate heat transfer on an endwall.
measured, especially for Mavg ¼ 2.0. Also, the simulation for The simulations were used to reveal the temperature behavior
Mavg ¼ 1.0 in Fig. 11(d) shows more jet attachment than the within the impingement channel and the endwall. The temperature
experiments, which is consistent with the / comparison at the contours showed that the heat transfer was highly three-
same blowing ratio. The fully detached jets for Mavg ¼ 2.0 in Fig. dimensional, and steep temperature gradients existed in the wall.
11(e) have less influence on s compared to the experiments, a The effects of individual impingement jets and film cooling holes
trend that was also observed for /. were observed inside the wall through the thermal gradients. The
The external TBC temperatures, s, along the inviscid SS and PS predicted temperature distributions inside the wall and impinge-
streamlines are given in Fig. 12(a) for Mavg ¼ 0.6, Fig. 12(b) for ment channel demonstrated the importance of matching the
Mavg ¼ 1.0, and Fig. 12(c) for Mavg ¼ 2.0. The data within the film relevant nondimensional parameters in a conjugate model.
cooling hole outlets have been removed. As blowing ratio
increases, the SS data (dashed lines) do not increase because the
film cooling jets become detached from the TBC surface. The Acknowledgment
decreased cooling by the film jets is balanced by the increased The authors would like to acknowledge support from the U.S.
cooling by internal impingement. In contrast, the PS data increase Department of Energy (DOE), National Energy Technology Labo-
with each increase in blowing ratio because the PS streamline ratory (NETL) through the University Turbine Systems Research
does not cross the path of the film cooling jets, and is influenced (UTSR) program. Any opinions, findings, conclusions, or recom-
by internal cooling. As discussed in reference to the contours, the mendations expressed herein are solely those of the authors and
simulations predict lower s than measured and less influence on s do not necessarily reflect the views of the DOE. The writers would
by the internal cooling than the experiments indicate. This like to thank Mark Zelesky of Pratt & Whitney, Dr. David Bogard
is more apparent on the PS after s/Cax > 0.25, and on the SS after of the University of Texas, and Robin Ames of DOE-NETL for
s/Cax > 0.7, downstream of the film cooling jets. Despite the their continued communication and support regarding this
underprediction, the trends of the data in the passage are well research.
captured by the simulations.
Nomenclature
Conclusions Bi ¼ Biot number (h1t/kw)
Overall and TBC effectiveness measurements were compared Cax ¼ axial chord length
to conjugate CFD predictions for the cases with and without TBC Cp ¼ pressure coefficient
at different blowing ratios. The improvement in overall effective- D¼ hole diameter
ness due to TBC was evaluated and found to be significant. h¼ convective heat transfer coefficient
Adding TBC produced a greater improvement in overall effective- H¼ impingement gap height
ness than the improvements achieved by increasing blowing ratio k¼ turbulent kinetic energy or thermal conductivity
alone. The TBC protected the endwall from the hot mainstream, L¼ length
reduced heat transfer, and allowed the internal cooling to be more M¼ blowing ratio (qcUc/q1U1)
effective. As blowing ratio increased, a greater improvement in Ma ¼ Mach number
overall effectiveness was observed because the TBC was more Nu ¼ Nusselt number (hD/kfluid)
effective at reducing heat transfer when the heat transfer was p¼ pitch length
higher without TBC. The reduction in heat transfer with TBC also P¼ pressure
caused the outer TBC temperature to be higher in comparison to PS ¼ pressure side streamline
the endwall temperature without TBC. q¼ heat flux
The predicted overall effectiveness with no TBC showed good R¼ thermal resistance (t/k)
agreement with the measurements. The main differences were Re ¼ Reynolds number (q1U1Cax/l1)
observed in the prediction of film cooling jet attachment for s¼ distance along a streamline starting at x ¼ 0
Mavg ¼ 1.0 and diffusion for Mavg ¼ 2.0, which were consistent S¼ blade span
with previous literature findings. The predictions of TBC SS ¼ suction side streamline