You are on page 1of 12

Impact of Wall Temperature

on Heat Transfer Coefficient


and Aerodynamics for Three-
Roberto Maffulli1
Turbomachinery Department,
Von Karman Institute for Fluid Dynamics,
Dimensional Turbine
Chaussee de Waterloo 72,
Rhode St Genese 1640, Belgium
e-mail: maffulli@vki.ac.be
Blade Passage
The present work is aimed to examine how the heat transfer coefficient (HTC) and main
Li He three-dimensional (3D) passage aerodynamic features may be affected by a nonadiabatic
Osney Thermofluids Laboratory, wall temperature condition. A systematic computational study has been first carried out for a
Department of Engineering Science, 3D nozzle guide vane (NGV) passage. The impacts of wall temperature on the secondary
University of Oxford, flows, trailing edge shock waves, and the passage flow capacity are discussed, underlining
Osney Mead, the connection and interactions between the wall temperature and the external aerodynamics
Oxford OX2 0ES, UK of the 3D passage. The local discrepancies in HTC in these 3D flow regions can be as high as
e-mail: li.he@eng.ox.ac.uk 3040% when comparing low and high temperature ratio cases. The effort is then directed to
a new three-point nonlinear correction method. The benefit of the three-point method in
reducing errors in HTC is clearly demonstrated. A further study illustrates that the new
method also offers much enhanced robustness in the wall heat flux scaling, particularly rele-
vant when the wall thermal condition is also shown to influence the laminarturbulent transi-
tion exhibited by two well-established transition models adopted in the present work.
[DOI: 10.1115/1.4036012]

1 Introduction (e.g., for a generic case as shown in Fig. 1(b)) and the calculation
of the slope h does depend both on the magnitude of DT and on
The increase in performance of a turbofan engine strongly
which wall temperature Twc is taken. The validity of the two-
depends on the turbine entry temperature (TET). Common in
temperature method for HTC calculations shrinks to a neighbor-
modern designs are entry temperatures higher than the melting
hood of the analyzed wall temperature in the case of a nonlinear
temperature of the blade metal: the correct design of cooling sys-
dependence of q_ on Tw. The HTC represents then a local tangent
tems for hot components is thus of critical importance to prevent
of a general nonlinear qT _ w curve. And the two-point method
engine failures.
with a small enough temperature difference becomes effectively a
Convective heat transfer analysis is commonly based on the
finite difference approximation of the local differential.
heat transfer coefficient (HTC) definition given by Newtons law
The problem of the variability of HTC with wall temperature
of cooling of the below equation
has been discussed by several authors. The quest for heat transfer
q_ hTw  Taw (1) descriptors invariant with the wall temperature resulted in some
authors (see Ref. [1]) proposing the use of other parameters
where h is the HTC, Tw is the wall temperature, and Taw is the adi- instead of HTC like hadiabatic or the discrete Greens functions.
abatic wall temperature. The conventional wisdom is that the Such parameters are invariant with the wall temperature. How-
HTC is predominantly determined by the aerodynamics and hence ever, together with the complexity involved in calculating or
should not be dependent on the wall temperature. This reflects an measuring these heat transfer descriptors, the use of these
assumption that the thermal interaction between the solid and fluid approaches in turbomachinery external heat transfer has been so
domains is negligible, and as such, the aerodynamics fully deter- far quite limited because they are based on superposition principle
mines the heat transfer. and as such are formally valid only when the energy equation is
Figure 1(a) shows how in this condition, when h does not
depend on Tw, Eq. (1) gives a linear dependence of q_ on wall tem-
perature. In this case, the calculation of h follows the solution of a
linear system obtained calculating q_ for two completely arbitrary
wall temperatures with an interval DT. In the linear case of
Fig. 1(a), the calculated HTC will be the correct one regardless of
the used wall temperature or temperature difference.
This two-temperature method predicts the correct value of HTC
over a whole range of wall temperatures only if both h and Taw in
Eq. (1) can be treated as two aerodynamically determined
unknowns. If this is not the case, Eq. (1) is not linear anymore

1
Corresponding author.
Contributed by the Heat Transfer Division of ASME for publication in the
JOURNAL OF THERMAL SCIENCE AND ENGINEERING APPLICATIONS. Manuscript received Fig. 1 Schematics of the linear and nonlinear dependency of
March 21, 2016; final manuscript received February 5, 2017; published online April heat transfer with Tw: (a) linear variation of q_ with Tw and (b)
19, 2017. Assoc. Editor: Ting Wang. nonlinear variation of q_ with Tw

Journal of Thermal Science and Engineering Applications DECEMBER 2017, Vol. 9 / 041002-1
Copyright VC 2017 by ASME

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


linear with the temperature. This may be appropriate when there Also even in a loosely coupled CFDfinite element analysis
are low wall-fluid temperature differences, and the dependency of (FEA) approach, the HTC is commonly used. As such, the way
air properties on the temperature can be neglected. This is not the HTC is defined and calculated and its dependency on Tw can have
case for a modern high pressure turbine (HPT) design. Further- a considerable impact on the convergence and stability of the iter-
more, a linear assumption of the energy equation is more chal- ation process in a loosely coupled method.
lenged for a high speed (transonic flow) when the energy equation The background as introduced above underlines that the
is strongly coupled with the continuity and momentum equations, dependence of HTC on the wall temperature deserves to be more
and the convective terms are likely to be highly nonlinear. systematically recognized and addressed. There seems to be a gap
Corrections on HTC as defined in Eq. (1) to account for its between attempting to correct this effect and the increasing use of
dependency on Tw may be made by using correlations, either CFD for 3D HPT aerothermal analysis for situations where the
based on a boundary layer model or entirely empirical. Different main performance differentiators come from the endwalls and sec-
authors (see Refs. [25] among the others) proposed to correct Nu ondary flow regions. The questions we face then are
with an exponential function of the wall to gas temperature ratio
(a) for 3D endwalls and secondary flow regions, is there a
(TR) in the form
meaningful dependence of relevant aerodynamic features
and HTC on the wall temperature?
Nu Nu0 TRn (2)
(b) if there is, given the limitation of a boundary layer
approach in these highly 3D flow regions, how can we cor-
where n is an empirical correction factor. Despite their wide
rect it effectively in conjunction with 3D CFD solutions?
usage, especially in a context of experimental research, these cor-
relations are inherently limited by the empiricism involved in the Furthermore, the way HTC is worked out needs to be suffi-
correct modeling of the value of the exponent n. Furthermore, ciently robust as required in a CFDFEA loosely coupled calcula-
these correlations typically are formed in a global correction tion. A relevant phenomenon in this regard is the
form, difficult to reflect local variations. On the other hand, a laminartransition in relation to the wall temperature condition.
boundary layer based correction would be limited by the basic The present work is motivated by the need to address these issues,
two-dimensional (2D) boundary layer assumption. extending the recent 2D analysis [6].
From the open literature, it seems that the problem of the
dependence of HTC with wall temperature should be systematically
addressed. It is of particular interest to ask how HTC should be 2 Solver and Validation
worked out and used consistently in the context of modern compu- Solution of steady Reynolds-averaged NavierStokes equations
tational fluid dynamics (CFD) applications for HPT aerothermal (RANS) has been obtained using FLUENT second-order, pressure-
design and analysis. This problem has been recently analyzed for a based steady solver.
2D NGV by Maffulli and He [6]. The authors observed the marked Calculations have been performed on the domain shown in
influence of wall temperature on HTC levels, near-wall aerodynam- Fig. 2. Mesh used is a 4  106 nodes hexahedral multiblock mesh,
ics, and trailing edge shock positioning. They also highlighted how created by ANSYS ICEM CFD. Due to the necessity of correctly
the use of a global TR correction (Eq. (2)) can lead to very errone- modeling heat transfer, y has been kept below five on all the
ous results. Consequently, the authors introduced a novel three- walls, particularly y 1 on blade surface. No wall functions
point nonlinear method for heat flux scaling, showing much have been used. A global view of the mesh and a close up of the
improved accuracy. The investigation of the effect of TR on heat trailing edge at the hub are shown in Fig. 3. Results have been
flux and external aerodynamics has been carried out also by Zhang checked to be not dependent on the grid chosen. The results of the
and He [7], who analyzed the effects of the thermal boundary con- mesh dependency study are discussed later in this section.
dition on rotor tip leakage flows. The authors showed a clear Computational results have been validated against the experi-
dependency of both HTC distributions and tip gap flow behavior on mental data obtained by Chana et al. [18]. Boundary conditions
the wall temperature. These studies highlighted how the interac- used, to match the experimental conditions, are summarized in
tions between the fluid domain and the solid wall condition affect Table 1. All the walls have been kept isothermal at ambient tem-
the local flow properties as well as the downstream aerothermal perature due to the test rig being a short duration facility. The
field. The importance of the solidfluid link has also been analyzed radial profile values for the outlet boundary condition have been
by Starke et al. [8], who highlighted how neglecting this coupling obtained by pitchwise-averaged mixing-plane calculations on the
can lead to heat load predictions with large errors. The topic of the same geometry by Rahim et al. [19]. The MT1 NGV, under the
wall-flow interaction for film cooling flows was studied in Refs. conditions analyzed, shows a laminar boundary layer region on
[9,10]. The authors highlighted how the change in the wall thermal
boundary condition and the subsequent change in thermal boundary
layer would affect heat transfer levels.
Harrison and Bogard [11] studied the validity of conventional
uncoupled analysis of film cooling flows. Bohn et al. [12] showed
by means of conjugate heat transfer (CHT) analysis, how the
blade conduction affects the secondary flows structure in film
cooling flows. Another important result of the coupling between
external aerodynamics and wall thermal boundary condition is
how the latter influences the boundary layer velocity profile and
shape factor as well as boundary layer transition point [13,14].
Due to the interactions between aerodynamics and wall heat
transfer, a correct modeling of the aerothermal behavior of the
flow should include simultaneous solution of the fluid and solid
domain. Despite the continuous increase of CHT applications to
turbomachinery [9,12,15,16] and the improvements in CHT solu-
tion methods (e.g., see Ref. [17]), the use of a fully coupled CHT
for design practice is, however, quite limited due to both extra
computational resources required and the complexity of involving
internal cooling parts. For these reasons, an uncoupled analysis
represents still a common tool in heat transfer analysis and design. Fig. 2 Computational domain

041002-2 / Vol. 9, DECEMBER 2017 Transactions of the ASME

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


For the aerodynamic validation, levels of isentropic Mach num-
ber at 50% span have been compared with the experimental data.
Heat transfer validation has been carried out comparing the nondi-
mensional heat flux defined in the below equation with the
experiment

q_ _
qC
(3)
q_ ref T0in  Tw k

The results of the experimental validation are shown in Figs. 4


and 5. For the aerodynamic validation, CFD calculations and
experimental data are in good agreement. The two turbulence
models do not show marked differences. The matching between
heat transfer calculations and experimental data is comparable
with other CFD studies on the same blade [22,23] and is reason-
ably good in the context of the case considered. Is it also worth
pointing out that the experimental data are time-averaged from a
transient, full-stage experimental facility. Global heat transfer lev-
els showed to be independent on the turbulence models used,
though local differences especially in the leading edge area can be
observed. The two equations kx SST model will be used in most
of the following analyses. The results have been checked to be
grid independent. The node count has been varied in both span-
wise and pitchwise directions. During all the mesh tests, the near-
wall distance has not been varied as it was deemed necessary to
keep y < 5 to ensure a proper near-wall heat transfer resolution.

Fig. 3 Grid used for the performed calculations: (a) computa-


tional grid: laminar region highlighted and (b) grid detail: trail-
ing edge-hub

Table 1 Summary of the CFD boundary conditions used for


validation tests

P0in 460,000 Pa Fig. 4 Mis for 50% span section: comparison of CFD with
MisOut 1.05 hub0.91 tip experimental data
T0in 444 K
Tw 288 K
Turbulence intensity 6% at leading edge
Turbulence length scale 5  103 m

the suction side. Validation tests have been carried out using Spalart
and Allmaras [20] and kx SST [21] turbulence models. The
boundary layer laminarization and turbulent transition on the suc-
tion side have been artificially tripped for both Spalart Allmaras
and the standard kx SST models. This artificially tripped transi-
tion has been obtained by disabling turbulence production in one of
the boundary layer blocks (shown in Fig 3(a)). The present analyses
are first carried out with the fixed transition model as implemented.
A fixed tripped transition point treatment is commonly used in aero-
dynamic analyses. A similar treatment is also used in some recent
turbine aerothermal analyses (e.g., see Refs. [22,23]). The use of
more sophisticated transition models such as transitional SST [24]
and kklx [25] including the wall temperature influence on the Fig. 5 q_ =q_ ref for 50% span section: comparison of CFD with
transition will be discussed later. experimental data

Journal of Thermal Science and Engineering Applications DECEMBER 2017, Vol. 9 / 041002-3

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 6 Mesh dependency study. Surface heat flux (W/m2) for the two mesh densities.

Fig. 7 Mesh dependency study. Total pressure field (Pa) downstream of the trailing edge.

The results of the mesh dependency study for heat transfer cal- transfer predictions. A simple option (often adopted for its seem-
culations are shown in Figs. 6 and 7. Two mesh densities (4  106 ingly simple form) is
and 5  106 nodes) have been compared for both wall heat flux
and total pressure downstream of the vane. The results indicate q_
HTC (4)
a mesh independency for the analyzed case, and the mesh Tref  Tw
with 4  106 nodes has been used for the results shown in Secs. 4
and 5. With the definition given above, HTC can be worked out in one
CFD solution for a given Tw. The reference temperature Tref is
often taken as T0in.
3 HTC Calculation Procedure The main concern with this definition is potentially very incon-
First, it should be recognized that the definition of HTC may sistent values for HTC. For instance, we might need to work out
vary, which itself is a source of uncertainties in turbine heat HTC at a near-adiabatic condition. This need can be due to

041002-4 / Vol. 9, DECEMBER 2017 Transactions of the ASME

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


experimental constraints or can also be caused by large local var- been varied from time to time to match the desired TR. Other
iations in fluid driving temperatures either at a high-speed flow boundary conditions used have been kept as described in Table 1.
with an appreciable recovery effect or due to a nonuniform inlet The midspan distribution of HTC for different wall temperatures
temperature traverse. Equation (4) may then return a zero (or even is shown in Fig. 9. Heat transfer coefficient at 50% span increases
negative) HTC, as the temperature difference in the denominator with decreasing wall temperature. The differences between the
does not reflect the true local driving temperature difference. near-adiabatic case (TR 0.99) and a lower temperature ratio one
Given that HTC is the reciprocal of the equivalent thermal resist- are also locally variable, highlighting the local effect of flow his-
ance, a zero or negative HTC would be completely meaningless. tory on HTC and the difficulties of using a global correction
This scenario would also violate the basic physical consideration approach as in many existing empirical correlations. The results
that heat transfer is driven by a temperature difference, thus a zero of Fig. 9 are in line with the observations from 2D tests by the
heat flux can only result from a zero temperature difference and same authors on the same NGV profile [6]. The analysis of the
HTC cannot be zero. Even from a practical perspective, a zero HTC contours for blade and endwalls shows a higher dependence
HTC would make the scaling in Eq. (3) nonworkable. Therefore, of HTC on wall temperature in regions interested by highly 3D
the HTC definition as given in Eq. (4), though applicable in a sim- flows.
ple low-speed flow with a uniform inlet total temperature, is not The following figures show the comparison of HTC contours
pursued here. for the nearly adiabatic and the cooled cases on blade wall. The
In the present work, the calculation of HTC has been carried HTC percent difference between the two cases is calculated as
out using the two-temperature method as a finite difference
approximation of the local tangent of a qT_ w curve. Referring to HTCTR0:7  HTCTR0:99
 100 (6)
the general case of HTC being dependent on wall temperature HTCTR0:99
(Fig. 1(b)), the local HTC is calculated using
Differently from the pressure side (Fig. 10), the suction side
Dq clearly shows the signature of secondary flows on heat transfer
h (5)
DT distributions for both temperature ratios. From Fig. 11, it can be
seen that suction sideendwall regions, dominated by secondary
where Dq is the difference in wall heat flux between two cases flows, do show the highest differences in HTC (up to 40%). The
with two wall temperatures with a difference of DT. similar HTC difference levels can be seen in the 50% span region,
As previously mentioned in the general case of HTC changing where the flow field is largely 2D. Here, the differences can be
with the wall temperature, the choice of DT does influence the explained by the influence of wall temperature on the positioning
predicted HTC values as the two-temperature method is rigor- of the suction-side shock wave. This aspect has been observed
ously valid only as a finite difference operator. For this reason, a already on the 2D MT1 case by Maffulli and He [6] and can be
study of the sensitivity of the HTC predictions to DT has been car- inferred by looking at the midspan HTC plots in Fig. 9. The influ-
ried out. ence of wall temperature on secondary flows and suction-side
The results of sensitivities are summarized in Fig. 8. For the shock will be further discussed in this section.
three temperature differences tested, the calculated HTCs are Similar observations can be made by comparing the endwalls
almost the same. The 5 K temperature interval has thus been used HTC contours of Figs. 12 and 13. The highest differences in HTC
throughout the following cases, being it small enough for negligi- levels are concentrated in the regions where secondary flows are
ble truncation errors of the finite difference approximation, but dominant. These results underline that a link between aerodynam-
still not too small to be affected by round-off errors. ics and heat transfer does exist, and HTC changes with wall tem-
perature cannot be ascribed only to the changes in gas properties.
For this reason, any global HTC correction, not taking into
4 Effect of TW on HTC and Flow account local flow history, cannot be considered effective in pre-
The effect of wall temperature on HTC distributions has been dicting correct HTC levels. Also, the strong dependence of HTC
analyzed for three fluid to wall temperature ratios (TR) ranging on TR in regions characterized by highly 3D flows makes inappli-
from 0.99 (quasi-adiabatic case) to 0.7 (cooled case). As the cable any boundary layer based correction.
research has been focused also on the effects of wall temperature The dependency of the shock position on the wall temperature
on secondary flows and shock position, the blade loading has been can be more closely observed by looking at the contours of
increased for the cases shown in what follows with respect to the Mn M$p=k$pk, where M is the Mach number vector, and $p
validation tests previously shown. Outlet isentropic Mach number is the pressure gradient. Such shock detection variable is equal to
ranged from 1.15 at the hub and 1 at the tip. Wall temperature has one across the shock. The flood lines for the two-temperature
ratios 0.99 and 0.7 are shown in Fig. 14.

Fig. 8 Midspan HTC distribution for the three analyzed temper-


ature differences Fig. 9 HTC dependency on TR, 50% span cut

Journal of Thermal Science and Engineering Applications DECEMBER 2017, Vol. 9 / 041002-5

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 10 HTC dependency on TR, on the vane pressure side

Fig. 11 HTC dependency on TR, on the vane suction side

A downstream shift of the suction-side shock for the cooled streamlines at the suction sideendwalls corner. Streamlines are
case is observed. The shift distance is of the same order of magni- consistently generated from the same points in the two cases.
tude as the trailing edge thickness. This is consistent with what Some differences are appreciable between the quasi-adiabatic and
has been previously observed on the 2D case for the same geome- cooled case with a more pronounced streamline convergence
try [6]. toward the midspan for the higher TR case.
A further examination of the effect of wall temperature on A close look at the changes in secondary flows has been possi-
secondary flows has been carried out. Figure 15 shows flow ble also by visualizing the direction of wall shear stresses in the

041002-6 / Vol. 9, DECEMBER 2017 Transactions of the ASME

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 12 HTC dependency on TR, on the shroud endwall

Fig. 13 HTC dependency on TR, on the hub endwall

form of limiting streamlines. The results are shown in Fig. 16. casing shroud for 0.99 temperature ratio is seemingly stronger
Differences in the wall shear stress direction due to different than that for 0.7. This might be attributed to the higher density
developments of secondary flow structures with wall temperature due to the cooling (hence the higher local fluid inertia, and resist-
are noticeable. In particular, Fig. 16 suggests that the radial ance to the secondary flow transport) in the near-wall region of
inward movement on the rear blade suction surface near the the cooled case (TR 0.7).

Journal of Thermal Science and Engineering Applications DECEMBER 2017, Vol. 9 / 041002-7

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


The changes of fluid density in the blade passage due to the
wall temperature have been examined also. A different wall tem-
perature alters the temperature distribution in the boundary layer
and, consequently, the density field. Basic thermodynamic consid-
erations bring to the conclusion that the maximum difference in
the density field is at the wall, where temperature difference is
maximum. When considering a comparison between the quasi-
adiabatic (TR 0.99) and the cooled (TR 0.7) cases, the differ-
ence in near-wall density is up to about 30% for the present case.
This value follows from the consideration that the pressure field is
largely invariant with the wall temperaturesomething that has
been observed also in the present caseand density variations fol-
low directly from the temperature field (the relative difference
between the Tw for the two cases considered is about 30%). These
differences reduce farther from the wall due to diffusion but do
remain far from negligible as shown in Fig. 18. The plots show
Fig. 14 Suction-side shock position for the quasi-adiabatic the density contours at a cut plane positioned after the trailing
(TR 5 0.99) and cooled (TR 5 0.7) case edge. Non-negligible differences (of the order of 1015%) can be
observed comparing the two density fields. A remarkable observa-
Also, a comparison of the contour plots of wall shear stress tion is the fact that such differences are not confined to the wall
between the two cases is reported in Fig. 17. It is possible to see regions but penetrate to a larger extent in the whole passage due
how the wall temperature impacts not only the position of the lim- to secondary flows. In general, it can be seen how the largest dif-
iting streamlines but also the levels of shear stress in the trailing ferences in density can be observed in the regions where most of
edge region. This again suggests the presence of a direct coupling the aerodynamic losses are generated (near-wall regions, trailing
between aerodynamics and heat transfer. The rear part near casing edge wake, and secondary flows). As such, this effect should not
shroud for the near-adiabatic case (TR 0.99) again shows a be overlooked in the aerodesign process.
stronger radial inward movement than that for TR 0.7, consist- Considering the differences in fluid density as well as second-
ent with the expected change in local fluid inertia. ary flows across the blade passage, the influence of wall

Fig. 15 Streamlines at suction sideendwalls corners for quasi-adiabatic (TR 5 0.99) and
cooled (TR 5 0.7) case

Fig. 16 Wall shear stress direction on suction side for quasi-adiabatic (TR 5 0.99) and
cooled (TR 5 0.7) case

041002-8 / Vol. 9, DECEMBER 2017 Transactions of the ASME

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 17 Wall shear stress magnitude on suction side for quasi-adiabatic (TR 5 0.99) and
cooled (TR 5 0.7) case

Fig. 18 Density contours at TE cut plane for quasi-adiabatic (TR 5 0.99) and cooled
(TR 5 0.7) case

temperature on the overall blade passage capacity has been also applied to correct heat transfer data both on a 2D NGV section [6]
assessed. All the CFD runs examined have mass flow converged and a 3D tip leakage flow field [7].
to less than 0.1%. The 0.7 TR case showed a 0.5% increase in The new nonlinear method is based on hypothesizing a linear
mass flow with respect to the quasi-adiabatic case. Considering relation between the HTC and the wall temperature. Newtons law
that the general trends of blading performance increasing nowa- of cooling can then be expressed by
days are aimed at fractions of a percentage point, the impact of
such a change in blade mass flow should thus not be overlooked, q_ h0 h1 Tw Tw  Taw (7)
also in noting that temperature ratios lower than 0.7 can be typical
of heavily cooled NGVs. where h0, h1, and Taw are the unknowns for each surface mesh
point. The three parameters are determined simply from the solu-
tion of a set of three linear equations, with the heat fluxes obtained
by CFD solutions at three different wall temperatures. The method
5 Nonlinear Method for TR Scaling does not require any numerical approximation techniques. It is
The results shown in Sec. 4 highlight a strong dependence of purely analytical, simple with very little extra processing cost.
HTC on the wall thermal condition. This is an inherently local The method is equivalent to considering a quadratic depend-
effect which is ignored when trying to correct HTC data using a ence of heat flux on wall temperature. The formulation starts from
global correction method. The previous research [6] showed the the hypothesis that the variability of Taw with wall temperature is
inadequacy of global corrections for the 2D midspan section of much smaller compared to the one shown by HTC. Its accuracy to
MT1 NGV. While global corrections are inapplicable even on 2D predict HTC at different wall temperatures has been also tested
cases, correction methods based on a boundary layer approach recently on a case where high variability of Taw is expected [7].
may be able to be applied for the case of a pure 2D flow field. The model showed comparable accuracy (and a more robust com-
However, when a 3D flow field is examined, the definition of a putational stability) with respect to a higher order nonlinear
boundary layer integration direction becomes cumbersome if not method including wall temperature dependency also for Taw.
impossible. The distinctive characteristic of such a method with respect to
The three-point nonlinear method recently introduced by Maf- the traditionally used two-temperature method is that while the
fulli and He [6] gives the possibility to apply a local correction of latter is valid only when used as a finite difference approximation
HTC with the wall temperature. The method has been successfully _ w curve at a given Tw (see Fig. 1(b)),
of the local slope of the qT

Journal of Thermal Science and Engineering Applications DECEMBER 2017, Vol. 9 / 041002-9

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 19 Wall heat transfer (W/m2) for TR 5 0.6, direct CFD (center), conventional two-point method (left), and new three-point
method (right)

Fig. 20 Prediction error comparison between the two-point and three-point

the three-point method is valid for a range of wall temperatures as 25% in some regions. Using the proposed three-point method
being essentially a three-point parabolic curve fitting method. allows to reduce the prediction errors significantly. These compar-
This increased validity range makes the method particularly isons demonstrate consistently that the new three-point method
appealing when using HTC as a buffer parameter in a loosely offers, at the price of a 50% increase in computational time com-
coupled CFDFEA approach, allowing improved stability and pared to the two-point method, the possibility of much more accu-
convergence. In Sec. 6, we will first examine the validity and rately predicting the heat flux over a whole range of wall
effectiveness of the three-point method for the fully 3D case. We temperatures.
will then consider how the method may help in more challenging
cases considering the influence of wall temperature on the
laminarturbulent flow transition. 6 TW Effect on Laminar Turbulent Transition
The three-point method is applied to predict data at TR 0.6. The fixed transition treatment as described above, though sim-
The constants of Eq. (7) are calculated using CFD solutions at TR ple, does not allow to capture the effect of wall temperature on the
0.99, 0.8, and 0.7. In the same fashion, the two-point linear transition location. The results obtained in this section use transi-
method is applied to extrapolate the heat flux distribution on the tion models in which the transition location is determined as part
blade and endwalls at TR 0.6, using h and Taw calculated at the of the solution, allowing to further explore the effects of TR on
quasi-adiabatic case (TR 0.99). Figure 19 shows the comparison external aerodynamics.
between the direct CFD results and the two used prediction meth- The analysis is carried out on the 2D midspan section of the
ods (the two-point linear and the three-point nonlinear). The MT1 NGV analyzed in the previous paragraphs. Over 100,000
improved predictive performance of the three-point method with nodes mesh have been used for the study. Wall y has been kept
respect to the conventional two-points one is evident, especially lower than one. Validation and mesh independency tests have
in the regions characterized by highly 3D flows (corresponding, as been carried out on the used grid. The results of such studies are
seen previously, to high HTC variability). To better compare the not shown here for brevity. Calculations have been performed
predictive performance of the two models, the percentage errors using the two well-established transition models available in
of the predictions are calculated using the below equation FLUENT: transitional SST [24] and kklx [25].
Figures 21 and 22 show wall nondimensional heat flux obtained
using the kklx and the transitional kx SST, respectively.
q_ predicted  q_ CFD Despite some offset in the predicted transition point and length,

error  100 (8)
q_ CFD both transition models are consistent in predicting an early onset
of transition due to the cooling. This effect of cooling in promot-
Comparison of the errors is shown in Fig. 20. Heat transfer errors ing boundary layer transition as predicted presently is qualita-
obtained using the conventional two-point method can be as high tively in line with that observed experimentally by Back et al.

041002-10 / Vol. 9, DECEMBER 2017 Transactions of the ASME

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 21 q_ =q_ ref (Eq. (3)) for TR 5 0.8 and 0.6 using the kklx Fig. 23 HTC plots for TR 5 0.6 for kklx transition model,
transition model obtained using the traditional two-point method of Eq. (5)

Fig. 24 Comparison of the new three-point method and the


conventional two-point method for a transitional case using the
kklx transition model
Fig. 22 q_ =q_ ref (Eq. (3)) for TR 5 0.8 and 0.6 using the transi-
tional SST model convergence issues when a loosely coupled CFDFEA iterative
method is applied.
The results and discussions above lead to the following
[26] and is also consistent with the destabilizing effect of cooling
observations:
on bypass transition described by Reshotko and Tumin [27].
A quantitative evaluation of the accuracy of the used transition (a) The introduction of a nonfixed transition would exacerbate
models is beyond the scope of this paper. This observed behavior the limitations of a linear scaling for heat transfer.
in the present calculation is interesting as a cooled boundary layer (b) The conventional two-point approach can result in very
is often perceived as being more stable (thus should accordingly oscillatory values of HTC due to transition point depend-
have a delayed transition instead). ence on Tw. This would affect not only the consistency and
Given the movement of the transition point and the drastic accuracy of HTC itself but also the stability and conver-
change of the heat transfer from the laminar to turbulent regimes, gence of CFDFEA iterations when HTC is used as the
a key question of interest is, how does this movement affect the buffer parameter.
HTC and its working procedure? The issue is twofold now, accu-
Given the previously highlighted limitations of the conven-
racy and robustness. The latter is particularly relevant in the con-
tional two-point method in the presence of transition, the three-
text of stability of CFDFEA loosely coupled methods using HTC
point method presented in Sec. 5 has been applied. The two-point
as a buffer parameter. Given the consistency in the response to
method has been used to linearly extrapolate heat transfer data
wall temperature shown by the two transition models used, the
obtained on TR 0.8. The three-point method instead has been
following analysis will be carried out by only using the kklx
used, based on CFD data at TR 0.99, 0.8, and 0.6. Predictions
transition model.
by both models are then made at TR 0.7. The results compared
In Fig. 23, the HTC distributions for TR 0.6 are shown for
to that of the direct CFD at TR 0.7 are shown in Fig. 24. It is clear
the kklx transition model. HTC is here calculated using the con-
that the conventional two-point method results in a very erroneous
ventional two-point method with DT 10 K (Eq. (5)). The move-
prediction of the transition point (predicted to be in the same place as
ment of the transition point with the wall temperature leads to a
for TR 0.8) and a large overshooting of the heat flux around the
seemingly abrupt jump in local heat flux seen by a fixed surface
transition. However, the result of the new three-point method is able
point. As a result, the curve for the surface point has almost a
to follow much more closely the direct CFD solution.
local discontinuity (i.e., locally nondifferentiable). This causes
HTC to overshoot around the transition point, as indicated in
Fig. 23. This characteristic of HTC around the transition point has
been observed for different temperature ratios and temperature 7 Conclusions
differences tested. This oscillatory behavior in HTC calculated There appears to be lack of consensus regarding if and how a
using the two-temperature method can possibly lead to wall temperature condition would affect blade passage external

Journal of Thermal Science and Engineering Applications DECEMBER 2017, Vol. 9 / 041002-11

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


aerodynamics. Even when such effect is recognized to exist, how Tu turbulence intensity
it should be accounted for in predicting convective heat transfer, TR Tw =T0in wall to inlet temperature ratio
particularly in the context of using modern CFD tools, remains U velocity (m/s)
unclear. Built on a recent study on this issue for a 2D configura- l dynamic viscosity (Pas)
tion, the present computational study focuses on the effect for a q density (kg/m3)
fully 3D HP NGV blade passage.
The present results for the midspan are in line with the results
of the previous 2D work. However, stronger effects of tempera- References
ture ratio on HTC have been observed in the regions dominated [1] Moffat, R. J., 1998, Whats New in Convective Heat Transfer? Int. J. Heat
by endwall and secondary flows, where the differences in HTC Fluid Flow, 19(2), pp. 90101.
are as high as 40% between the cooled and quasi-adiabatic cases. [2] Kays, W., and Crawford, M., 1981, Convective Heat and Mass Transfer, 2nd
ed., McGraw-Hill, New York.
The observed changes in heat transfer characteristics in highly [3] Fitt, A., Forth, C., Robertson, B., and Jones, T., 1986, Temperature Ratio
three-dimensional and complex flow regimes would strongly chal- Effects in Compressible Turbulent Boundary Layers, Int. J. Heat Mass Trans-
lenge those correction methods based on a boundary layer fer, 29(1), pp. 159164.
approach for which a clearly defined 2D upstream path would [4] Eckert, E., 1955, Engineering Relations for Friction and Heat Transfer to
Surfaces in High Velocity Flow, J. Aeronaut. Sci., 22(8), pp. 585587.
have to be a prerequisite. [5] Petukhov, B. S., 1970, Heat Transfer and Friction in Turbulent Pipe Flow
The blade passage secondary flows are shown to respond to With Variable Physical Properties, Advances in Heat Transfer, Vol. 6, J. P.
wall temperature condition, which in turn affects the HTCTw Hartnett and T. F. Irvine, eds., Academic Press, New York, pp. 503564.
dependency. The fluid density increase in near-wall regions for a [6] Maffulli, R., and He, L., 2014, Wall Temperature Effects on Heat
Transfer Coefficient for High-Pressure Turbines, J. Propul. Power, 30(4),
cooled case leads to considerable rise in the local flow inertia. pp. 10801090.
Consequently, a clearly reduced radial inward secondary flow on [7] Zhang, Q., and He, L., 2014, Impact of Wall Temperature on Turbine Blade
the rear suction near casing region can be identified. Also, in line Tip Aerothermal Performance, ASME J. Eng. Gas Turbines Power, 136(5),
with previous observations on the 2D case, the cooled wall is seen p. 052602.
[8] Starke, C., Janke, E., Hofer, T., and Lengani, D., 2008, Comparison of a Con-
to move the suction surface shock wave downstream by a distance ventional Thermal Analysis of a Turbine Cascade to a Full Conjugate Heat
comparable with trailing edge thickness. The passage flow Transfer Computation, ASME Paper No. GT2008-51151.
capacity is also seen to vary with wall temperature by a non- [9] Dees, J. E., Bogard, D. G., Ledezma, G. A., and Laskowski, G. M., 2011, The
negligible amount, compared with typical performance changes of Effects of Conjugate Heat Transfer on the Thermal Field Above a Film Cooled
Wall, ASME Paper No. GT2011-46617.
practical interest for an HP turbine. [10] Dees, J. E., Bogard, D. G., Ledezma, G. A., Laskowski, G. M., and Tolpadi, A.
The use of the new three-point nonlinear correction allows to K., 2012, Momentum and Thermal Boundary Layer Development on an Inter-
correct HTC data on a local basis, simply by solving three linear nally Cooled Turbine Vane, ASME J. Turbomach., 134(6), p. 061004.
equations. The three-point method demonstrates the capability in [11] Harrison, K. L., and Bogard, D. G., 2008, Use of the Adiabatic Wall Tempera-
ture in Film Cooling to Predict Wall Heat Flux and Temperature, ASME Paper
predicting HTC distributions for the NGV passage subject to 3D No. GT2008-51424.
complex passage flows, with a substantial reduction in errors, [12] Bohn, D., Ren, J., and Kusterer, K., 2003, Conjugate Heat Transfer Analysis
compared with the conventional two-point method. Finally, the for Film Cooling Configurations With Different Hole Geometries, ASME
influence of wall temperature on laminarturbulent transition and Paper No. GT2003-38369.
[13] Liepmann, H. W., and Fila, G. H., 1947, Investigation of Effects of Surface
its implication for the conventional HTC calculation and scaling Temperature and Single Roughness Elements on Boundary-Layer Transition,
have been examined. The study further underlines the link Technical Report, NACA Report No. 890.
between aerodynamics and heat transfer while highlighting the [14] Rued, K., and Wittig, S., 1986, Laminar and Transitional Boundary Layer
shortcomings of the conventional approach for heat transfer scal- Structures in Accelerating Flow With Heat Transfer, ASME J. Turbomach.,
108(1), pp. 116123.
ing. The advantages of the new three-point nonlinear method are [15] Verstraete, T., Alsalihi, Z., and Van den Braembussche, R., 2007, Numerical
further illustrated in the case of nonfixed transition modeling. The Study of the Heat Transfer in Micro Gas Turbines, ASME J. Turbomach.,
benefits of the new method can be seen in this case not only in 129(4), pp. 835841.
terms of an increased accuracy in heat transfer coefficient predic- [16] Heidmann, J. D., Kassab, A. J., Divo, E. A., Rodriguez, F., and Steinthorsson,
E., 2003, Conjugate Heat Transfer Effects on a Realistic Film-Cooled Turbine
tion but also in enhanced stability, closely relevant to CFDFEA Vane, ASME Paper No. GT2003-38553.
iterative coupling. [17] He, L., and Oldfield, M., 2011, Unsteady Conjugate Heat Transfer Modeling,
ASME J. Turbomach., 133(3), p. 031022.
[18] Chana, K., Patel, T., and Mole, A., 2001, A Summary of Measurements With a
Acknowledgment Non-Uniform Inlet Temperature Profile From the MT1 Single Stage HP
Turbine, TATEF Project No. BRPR-CT97-0519.
This work has been partly sponsored by UK Engineering and [19] Rahim, A., Khanal, B., He, L., and Romero, E., 2014, Effect of Nozzle Guide
Physical Sciences Research Council (EPSRC) (Grant No. EP/ Vane Lean Under Influence of Inlet Temperature Traverse, ASME J. Turbom-
G035245), whose support is gratefully acknowledged. ach., 136(7), p. 071002.
[20] Spalart, P., and Allmaras, S., 1992, A One Equation Turbulence Model for
Aerodynamic Flows, AIAA Paper No. 1992-0439.
Nomenclature [21] Menter, F., 1994, Two-Equation Eddy-Viscosity Turbulence Models for Engi-
neering Applications, AIAA J., 32(8), pp. 15981605.
C blade chord (m) [22] Khanal, B., He, L., Northall, J., and Adami, P., 2013, Analysis of Radial
k thermal conductivity (W/m K) Migration of Hot-Streak in Swirling Flow Through High-Pressure Turbine
l turbulence length scale (m) Stage, ASME J. Turbomach., 135(4), p. 041005.
[23] Lad, B., He, L., and Romero, E., 2012, Validation of the Immersed Mesh
M Mach number Block (IMB) Approach Against the Cooled MT1 NGV Application for Mesh
Nu Nusselt number Dependency Studies, ASME Paper No. GT2012-68779.
Nu0 Nusselt number at TR 1 [24] Menter, F. R., Langtry, R., Likki, S., Suzen, Y., Huang, P., and V olker, S.,
q_ wall heat flux (W/m2) 2006, A Correlation-Based Transition Model Using Local VariablesPart I:
Model Formulation, ASME J. Turbomach., 128(3), pp. 413422.
q_ ref reference wall heat flux (W/m2) [25] Walters, D. K., and Cokljat, D., 2008, A Three-Equation Eddy-Viscosity
Re Reynolds number Model for Reynolds-Averaged NavierStokes Simulations of Transitional
T temperature (K) Flow, ASME J. Fluids Eng., 130(12), p. 121401.
Tw wall temperature (K) [26] Back, L., Cuffel, R., and Massier, P., 1969, Laminar, Transition, and Turbulent
Boundary-Layer Heat-Transfer Measurements With Wall Cooling in Turbulent
Taw adiabatic wall temperature (K) Airflow Through a Tube, ASME J. Heat Transfer, 91(4), pp. 477487.
Tref reference temperature (K) [27] Reshotko, E., and Tumin, A., 2004, Role of Transient Growth in Roughness-
T0in inlet total temperature (K) Induced Transition, AIAA J., 42(4), pp. 766770.

041002-12 / Vol. 9, DECEMBER 2017 Transactions of the ASME

Downloaded From: http://thermalscienceapplication.asmedigitalcollection.asme.org/ on 10/25/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like