Comparison of Time-Resolwed

Turbine Rotor Blade Heat Transfer

Measurements and Numerical
R. S. Abhari 1
Calculations
G. R. Guenette
Time-resolved turbine rotor blade heat transfer data are compared with ab initio
numerical calculations. The data were taken on a transonic, 4-to-l pressure ratio,
A. H. Epstein uncooled, single-stage turbine in a short-duration turbine test facility. The data
consist of the time history of the heat transfer distribution about the rotor chord
at midspan. The numerical calculation is a time accurate, two-dimensional, thin
M. B. Giles shear layer, multiblade row code known as UNSFLO. UNSFLO uses Ni's Lax-
Wendroff algorithm, conservative boundary conditions, and a time tilting algorithm
Gas Turbine Laboratory, to facilitate the calculation of the flow in multiple blade rows of arbitrary pitch
Massachusetts Institute of Technology, ratio with relatively little computer time. The version used for this work had a simple
Cambridge, MA 02139 algebraic Baldwin-Lomax turbulence model. The code is shown to do a good job
of predicting the quantitative time history of the heat flux distribution. The wake/
boundary layer and transonic interaction regions for suction and pressure surfaces
are identified and the shortcomings of the current algebraic turbulence modeling in
the code are discussed. The influence of hardware manufacturing tolerance on rotor
heat transfer variation is discussed. A physical reasoning explaining the discrepancies
between the unsteady measurement and the calculations for both the suction and
pressure surfaces are given, which may be of use in improving future calculations
and design procedures.

Introduction
The accurate prediction of heat transfer in high-pressure transfer experience often does not follow that to be expected
turbine stages has long been recognized as a key to improved from steady, two-dimensional cascade flows (Sharma et al.,
gas turbine performance and engine life. To that end, a vast 1988, 1992). Rotor pressure side heat transfer may be the most
amount of engineering effort has been extended over the last different in this regard. Unsteadiness, three-dimensionality,
forty years to provide accurate experimental measurements of and nonuniform inflow all may be contributors to this dif-
the heat transfer distribution and to improve the accuracy by ference.
which that distribution can be predicted. This work has in- Unsteady effects in particular have been studied in some
cluded the measurement of two-dimensional airfoil heat trans- detail. Cascade experiments using rotating bars to simulate
fer in both steady-state (Turner et al., 1985) and transient NGV wakes and shock waves (Doorly and Oldfield, 1985;
cascades (Schultz et al., 1977) as well as stage measurements Ashworth et al., 1985; Johnson et al., 1989a, 1989b) have
in short duration test rigs (Dunn et al., 1984; Guenette et al., shown that wake impingement on the rotor can trigger laminar
1989). to turbulent transition, increasing the time average heat trans-
Progress in calculational techniques has been rapid over this fer to a level between that of steady-state laminar and fully
same period. Two-dimensional steady viscous calculations of turbulent. Experiments on low-speed turbine rotors tend to
the flow in turbines is now commonplace and three-dimen- confirm this observation (Blair et al., 1989; Dring et al., 1986;
sional steady techniques are also being applied. More recently, Sharmaetal.; 1988; Addison and Hodson, 1990a, 1990b). The
unsteady multiblade row codes have been developed (Rai, 1985), interaction of shock waves with the blade boundary layers can
including those capable of handling multiple blade rows with also increase rotor heat transfer. Rigby et al. (1989) constructed
arbitrary blade counts (Giles, 1988). a simple physical model explaining the unsteady heat transfer
It has long been observed that gas turbine engine rotor heat in terms of compressional heating in the boundary layer by
the moving shock waves.
'Current address: Textron Lycoming, Stratford, CT. The intent of this work is to examine the time-resolved aero-
Contributed by the International Gas Turbine Institute and presented at the
36th International Gas Turbine and Aeroengine Congress and Exposition, Or-
dynamics and heat transfer in a transonic turbine rotor with
lando, Florida, June 3-6, 1991. Manuscript received at ASME Headquarters a particular emphasis on elucidating the difference between
March 4, 1991. Paper No. 91-GT-268. Associate Technical Editor: L. A. Riekert. the unsteady flow and the time mean design intent. This is

818 / Vol. 114, OCTOBER 1992 Transactions of the ASME

Copyright © 1992 by ASME
Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 Table 1 Turbine design parameters

Turbine loading, AH/U2 -2.3

Total pressure ratio 4.2
Velocity ratio, Cx/U 0.63
Rotor aspect ratio 1.5
NGV exit Mach No. 1.18

Table 2 MIT blowdown turbine scaling

Full Scale MIT Blowdown
Fluid Air Ar-Fr 12
Ratio specific heats 1.28 1.28
Mean metal temperature 1118°K(1550°F) 295°K (72°F)
Metal/gas temp, ratio 0.63 0.63
Inlet total temperature 1780°K (2750°F) 478°K (400°F)
True NGV chord 8.0 cm 5.9 cm
Reynolds number* 2.7 x 10 6 2.7 x 10 6
Inlet pressure, atm 19.6 4.3
Outlet pressure, atm 4.5 1.0
Outlet total temperature 1280°K(1844°F) 343°K (160°F)
Prandtl number 0.752 0.755
Eckert numbert 1.0 1.0
Rotor speed, rpm 12,734 6,190
Mass flow, kg/sec 49.0 16.6
Power, watts 24,880,000 1,078,000
Test time Continuous 0.3 sec
Fig. 1 Transonic turbine geometry investigated * Based on NGV chord and isentrapic exit conditions
t (y-l)M2T/AT
HEAT FLOW
TEMPERATURE
SENSORS

d INSULATOR
thick polyimide insulator (Fig. 2). The sensing area is rectan-
gular (1.0 x 1.3 mm), oriented such that the longer dimension
i_ is in the chordwise direction. The insulator is adhesively bonded
ADHESIVE
to, and completely covers, the blade profile (preventing thermal
BLADE PROFILE SURFACE discontinuities). At low frequencies, the temperature drop
across the insulator is a direct measure of the heat flux to the
(Not To Scale) wall (the device is essentially a thermal shunt). This direct
proportionality between heat flux and temperature difference
Fig. 2 Schematic of thin film multilayer heat flux gage; total
thickness is about 25 jim
extends from d-c to about 20 Hz, at which point the thermal
waves within the insulator begin to damp. Above 1 kHz, how-
ever, the 25-fim-thick insulator appears infinitely thick to the
done by a detailed comparison between time-resolved meas- top surface. Thus, above this frequency, a quasi-one-dimen-
urements and a viscous, multiblade row numerical calculation. sional assumption can be used to infer the heat flux from the
top surface temperature history (blade passing frequency is 3.6
Experimental Arrangement kHz). Using a numerical data reduction technique, the entire
frequency domain from d-c to 100 kHz is reconstructed. The
The test article (Fig. 1) is a 4:1 pressure ratio single-stage, gages were calibrated using a pulsed laser. The relative gage
0.5-m-dia, transonic turbine whose design parameters are given calibrations are accurate to better than 5 percent. Absolute
in Table 1. It was tested in the MIT Blowdown Turbine Facility. calibration accuracy is about 10 percent. Uncertainty is eval-
This is a short-duration (0.3 s test time) rig, which maintains uated for each transducer. Details of the gage theory, data
real engine similarity of Reynolds number, Mach number, reduction, and calibration may be found from Epstein et al.
Prandtl number, gas-to-wall temperature ratio, and flow ge- (1986).
ometry (Table 2). Corrected speed and weight flow are main-
tained constant to better than 0.5 percent over the test time. For the data presented herein, the signals from the heat flux
The turbulent intensity at the NGV inlet is less than 1 percent gages were digitized at a 200 kHz sampling rate. Unless oth-
for these tests. The stage results reported herein are for an erwise specified, the digital signal was then ensemble-averaged
uncooled geometry. More details may be found from Epstein for 360 vane passing periods (Guenette et al., 1989).
et al. (1984) and Guenette et al. (1989).
Temperature and pressure rakes measure the flowfield at The Numerical Procedure
the inlet to the nozzle guide vanes (NGVs) and the outlet of The numerical procedure used to model this flow was a two-
the rotor. Shaft rotational speed, rotor torque, and various dimensional, Reynolds-averaged, unsteady multiblade row Na-
wall static pressures are recorded as well. Of interest here is vier-Stokes code, UNSFLO. This is a coupled viscous/inviscid
the measurement of the time-resolved heat flux distribution code in which the thin shear layer Navier-Stokes equations are
about the rotor blade. This was done using thin film heat flux solved on a body-fitted boundary layer grid using an implicit
gages distributed about the blade profile. These transducers algorithm, while the Euler equations are solved on an outer
measure both the d-c and a-c components of heat flux. The in viscid grid using an explicit algorithm. The interface between
heat flux gages consist of two thin film (140-nm-thick) nickel the two regions is handled in a conservative manner. The code
temperature transducers mounted on either side of a 25-^im- utilizes an innovative space-time coordinate transformation,

Journal of Turbomachinery OCTOBER 1992, Vol. 114 / 819

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 3000 -. DESIGN INCIDENCE — Steady-State Laminar Calc.
— Steady-State Turbulent Calc.
— Time-Ave. Unsteady Turb. Calc
o Time-Ave. Rotor Meas.

2000
o

V)

1000-

-1 1
Suction Surface ° Pressure Surface
Fractional Wetted Surface
Fig. 5 Comparison of measured heat transfer with fully laminar and
fully turbulent steady-state calculations, and the time average of the
Fig. 3 Grid used in unsteady nozzle guide vane-rotor calculation unsteady multiblade row calculation for the design point

workstation, equivalent to about 2 hours on a Cray XMP. The

stream-tube height distribution through the stages (an input
1.0 • to the code) was derived from a streamline curvature calcu-
lation (Norton, 1988).
Although the static pressure distribution about the rotor was
a>
eg not measured in this series of experiments, Ashworth et al.
m (1985; also Ashworth, 1987) did measure it for the same blade
+-* profile in a cascade. The measured values are shown in Fig.
J> 0.5 4, along with steady state calculations for the rotor and for
c
the cascade installation (which has a slightly different stream-
o Cascade Measurement tube height distribution). The agreement is quite good except
O) — Cascade Calc.
£ —- Rotor Steady State Calc. at the trailing edge, where natural instabilities such as vortex
GL 0.0 i
shedding are important but not resolved in this steady state
O -1.0 -0.5 0.0 0.5 1.0 solution. We infer from this comparison that the code in its
IS steady-state version does a good job of calculating the mean
Suction x/s Pressure
static pressure distribution about the rotor profile.
Fig. 4 Steady-state static pressure distribution calculations compared
to cascade measurement; the stream-tube height differs between the
rotor and cascade by about 3 percent
Comparison of Steady-State Heat Transfer Predictions
and Measurements
"time-tilting" (Giles, 1988), to permit arbitrary rotor-stator A comparison between the time average of the measured
pitch ratios while incurring little penalty in computation time. rotor heat transfer, the time average of the unsteady NGV-
The viscous grid is an O-type mesh about each blade for rotor calculation, and a steady-state solution for the rotor alone
which an ADI (alternating-direction-implicit) method with is shown in Fig. 5 for design point flow. The heat transfer
Roe's flux-difference splitting gives third-order upwinding for coefficient is presented as Nusselt number, defined in terms
the residual operator and first-order upwinding for the implicit of the rotor blade axial chord, inlet relative total minus local
operator. An algebraic turbulence model is used. Quasi-three- profile temperature, and thermal conductivity based on local
dimensional effects are included through the specification of temperature. The local profile temperatures used were those
a stream-tube thickness in the third dimension (axial velocity measured by the top sensors on the heat flux gages. Unless
density ratio, AVDR, or hade). The inviscid grid is an un- explicitly indicated, the relative uncertainties ( ± standard de-
structured finite element type of mesh composed of an arbitrary viation) in the measurements are less than the size of the sym-
mix of quadrilateral and triangular cells. Here, the Euler equa- bols. Surface length is defined with the axial leading edge
tions are solved using a generalization of Ni's Lax-Wendroff tangential chord intersection as the zero reference.
algorithm (Ni, 1981). Full second-order accuracy is achieved. There is relatively little difference between the steady-state
The Baldwin-Lomax turbulence model used in the viscous part turbulent and the time-averaged unsteady calculations. In gen-
of the solution is not capable of predicting boundary layer eral, the measured heat transfer is equal to or greater than the
transition. For the calculations presented herein, transition was fully turbulent calculation. Integrated about the profile, the
assumed at the leading edge unless otherwise specified. Many predicted heat transfer is 11 percent less than the measured
more details on this code and its checkout may be found in a value. (Note that the uncertainty in the absolute level of heat
companion paper by Giles and Haimes (1991). transfer in the measurement is 10 percent.) A similar trend can
For the calculations done here, the O grid had 18 points be seen in Fig. 6 for a different flow condition, - 10 deg rotor
across the boundary layer and a total of 16,000 were used incidence. Here, the predicted heat transfer is 12 percent lower
throughout the domain, Fig. 3. Only the midheight geometry than that measured. A comparison of the two flow conditions
of the stage was modeled and the code was relatively fast, (Fig. 7) shows that the differences in the. measurements are
requiring approximately 20 CPU hours on a Stellar GS1000 confined principally to the first part of the suction surface,

820 / Vol. 114, OCTOBER 1992 Transactions of the ASME

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 3000- -10° INCIDENCE — Steady-Slate Laminar Calc. flowfield. To do this, we shall start by looking at the unsteady
— Steady-State Turbulent Calc. flow.
— Time-Ave. Unsteady Turb. Calc
o Time-Ave. Rotor Meas.
Time-Resolved D a t a and Calculations
2000- A comparison of the time-resolved heat transfer data and
calculations is shown in Fig. 8 for the design point and for
2 — 10 deg incidence flow. Along both the suction and pressure

i 1000- Ai l^
\0^OQ;5#^
= O-= =f * *i V.s
surfaces, the flow is highly unsteady in both the calculations
and the measurements, which are in qualitative agreement.
The exception to this is at the leading edge where calculation
shows a much larger degree of unsteadiness than do the data.
° ° ^? / (Note that the calculational result was averaged over the same
<
6 \s \ surface extent as is covered by the heat transfer gage.) The
unsteadiness in the flow is a result primarily of two physical
0. \ _, \.'"' interactions between the blade rows: NGV wake impingement
• ^ ^ 1 i I I
1 1 on the rotor and potential coupling, including moving shock
" Suction Surface ° Pressure Surface
patterns.
Fractional Wetted Surface The complexity of the NGV-rotor shock pattern can be seen
Fig. 6 Comparison of measured heat transfer with fully laminar and in Fig. 9, which is an animation of the time unsteady calcu-
fully turbulent steady-state calculations, and the time average of the lational results (the times correspond to those in Fig. 8). At
unsteady multiblade row calculation for - 1 0 deg incidence
any moment, as many as six shock waves may be moving
through a single rotor passage—some traveling upstream, some
down, some strengthening, some attenuating. Simultaneously,
3000 -, there are interactions between the NGV wakes and the rotor.
— Design Incidence Calc. This is illustrated in Fig. 10. The wake convection is such as
o Design Incidence Meas.
— -10° Incidence Calc. to continuously keep an NGV wake along the rotor suction
A -10° Incidence Meas. surface. The pressure surface, however, sees intermittent wake
impingement. The overall complexity of the curved shock waves
2000. and convection of the wakes through the rotor are qualitatively
o similar to that observed in Schlieren photographs by Ashworth
Z (1987), using rotating bars to simulate NGV wakes and shock
waves upstream of a cascade of blades with the midspan rotor
(A
V) profile.
1000- Although flow visualization is not available for the NGV-
rotor flow, time-resolved heat transfer data are (Fig. 8). The
calculational results can be used to help understand the flow
structure evident in the measurements. For example, the three
peaks in the heat transfer data on the crown of the suction
surface (Fig. 8c) can be seen to be the result of a shock wave
sweeping forward along the suction surface, Fig. 11. By using
•1 Suction Surface 0 Pressure Surface 1 the calculational results in such a way, many of the features
Fractional Wetted Surface in the time-resolved heat transfer can be quantitatively ex-
Fig. 7 Comparison between measurements and time-averaged un- plained.
steady calculations at design and - 1 0 deg incidence

Discussion of the Difference Between Calculation and

Measurement
while the difference in predicted heat transfer is spread along
the blade. Both the predicted and measured heat transfer at Inviscid Versus Viscous Flow Effects. In a practical sense,
- 10 deg incidence are reduced by some 11 percent under that it is the differences between the calculation and measurements
for the design incidence. that are of the most interest since, if they were always in
The general picture these comparisons suggest differs sig- complete agreement, few experiments would need to be run.
nificantly from cascade and low-speed rotor data in that the One discrepancy is in the strength of the shock-induced heat
pressure side heat transfer does not lie between laminar and transfer spike observed on the crown of the suction surface
fully turbulent levels, but rather the data are above the pressure for the design point flow (Fig. 86). The difference in the area
side turbulent prediction. This difference bears examination. under the calculated and measured spikes accounts for the
The validity of the measurements was established earlier by difference in the time averages at that location observed in
Guenette et al. (1989) who showed that the midspan rotor Fig. 5. This difference could conceivably be either from an
measurements in the blowdown tunnel agreed closely with those inaccuracy in the inviscid flow (improper shock strength, for
made on the same profile in a cascade by Ashworth et al. example) or a problem with the viscous boundary layer cal-
(1985). This serves to both confirm the rotor measurements culation. To explore this, we made use of a one-dimensional
and suggest that flow in this midspan region is largely two model of the shock heating process developed by Rigby et al.
dimensional. Rather than simply dismiss the calculations then (1989), which explains the abrupt rise in surface heat flux as
as inaccurate (the calculational accuracy is really quite good being primarily due to compressional heating of the inner sub-
for an ab initio effort; it is not sufficient, however, for realizing layer. The modulation of heat transfer is given as
desired turbine life), we thought it instructive to examine in P
NU(Q~NUQ AI1 , \ . ^lpcp/kCg(t)
as much detail as possible the difference between the data and (1)
calculations in order to shed light on the modeling assumptions, Nu 0
test numerical approximations, and generally understand the

Journal of Turbomachinery OCTOBER 1992, Vol. 1 1 4 / 8 2 1

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 W
(b)

(c)

(d)

(h)

(e)

0 1 2 0 1 2 0 1 2
(g)
Time/Blade Passing Period Time/Blade Passing Period Time/Blade Passing Period

Calculation
Measurement
1600

(k) 800
(b)
0

1600

800

0.

Z 1600 (c)

wv
(d>
(h) -A^^WV^W'""

1600

(g) 800 j¥^j¥m^ (e)

0
0 1 2 0 1. 2 0 1 2
Time/Blade Passing Period Time/Blade Passing Period Time/Blade Passing Period

Fig. 8 Comparison of time-resolved measurements a n d calculations at the design point and at - 1 0 d e g incidence

where chord, and y the ratio of specific heats. The first term on the
(7-1) right-hand side of Eq. (1) represents the compression, while
g(s) = £lg(t)]=sW2£ (2) the second term is the isentropic heating in the sublayer. Thus,
the heat flux perturbation can be predicted from the static
pressure history. This was done using the unsteady wall static
The symbol 5 is the Laplacian operator and the subscript 0 pressure predicted by the computer code (Fig. 12) to evaluate
corresponds to the steady-state values. P is the static pressure, Eq. (1). Figure 13 shows that Eq. (1) predicts the same per-
Tg the gas temperature, T„ the wall temperature, p the density, turbation in heat transfer as does the code. This implies that
cp the specific heat, k the thermal conductivity, C the axial the problem with the code here lies in the prediction of the

822 / Vol. 114, OCTOBER 1992 Transactions of the ASME

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 /Ij2c NA
2e
l/ L~S\ /Zs^icA It 2e \\ (N4 /
ifls^^Z/—
\2d i?V\ 2d
A Lfi~i\c^ p^c^&r
\2d ^M
/MEM^£rN^
y ^ s ^ \
m /\ 2a/^TV
3
v' J \ J it— ^\ \2d Cv^sT/"""
/fr^
///(lb ~_^^\_.
(N3 J
(N3 /
\2o^^ N/

QN2J CN2y (\\1J /N2 /

Time = 0.82 Time = 1.11 Time = 1.37 Time = 1.6
Fig. 9 Animation of the shock structure predicted by the time-resolved code at the design point; the limes correspond to
those of Fig. 8

0.0 0.5 1.0 1.5 2.0

Blade Passing Periods

2400

0.0 0.5 1.0 1.5 2.0

Blade Passing Periods

Shock 2c

J)1600H
A A
z
A/A A/A
v/ V
0) 800-

0-
0.0 0.5 1.0 1.5 2.0
Blade Passing Periods

Fig. 11 Correlation of heat transfer measurement features with

calculated shock motion for suction surface location correspond-
ing to Fig. 8(c)

inviscid flow (the shock strength) rather than in the treatment

of the viscous boundary layer.
Fig. 10 Calculated wake trajectory at design point using entropy Disagreement between the code and measurements takes a
as a marker different form on the pressure side of the blade. Shock waves

Journal of Turbomachinery OCTOBER 1992, Vol. 114 / 823

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 do not reach the forward portion of the pressure surface and
3
this is reflected in the more sinusoidal nature of the unsteady
m
0) heat transfer. Wakes do periodically impinge on the pressure
surface but this process is not well modeled by the code for
c
g two reasons. The first is that since this is a hybrid Navier-
+-*
co Stokes/Euler method, the inviscid part of the flowfield does
c not have an accurate way of representing the convection of
the turbulence characteristics of the wakes. The second prob-
5)
+-» lem is that the algebraic turbulence model used in the viscous
j>
e part of the flow does not transfer turbulence from the free
stream to the boundary layer.
An example of how these modeling deficiencies may influ-
3
<n
in
ence the solution can be seen with reference to Fig. 14(a),
a
i.
which replots the data-calculation comparison of Fig. 8(/). The
a. instantaneous Mach number distribution through the bound-
o
+-<
ra ary layer at two instants of time is also shown, as is the in-
•»-*

W
stantaneous ratio of turbulent to laminar viscosity predicted
by the code at that location. At t = 0.75, the local Mach
number is high, the turbulent viscosity high, and the heat
Time / Blade Passing Period transfer large. At / = 1.1, the calculation shows the local Mach
Fig. 12 Calculated static pressure history at suction surface number is reduced, the turbulent viscosity has dropped by an
location corresponding to Fig. 8(b) order of magnitude, and the heat transfer reduced by a factor
of two. This behavior is consistent with the limitations of an
algebraic turbulence model. On the pressure surface of a tur-
bine rotor blade, the influence of the turbulence convected in
by the wake is not modeled here. In the code, the boundary
5000 layer turbulence is only a function of the instantaneous local
velocity, which fluctuates rapidly. In the physical flow, how-
ever, the turbulence has a physical time constant, which serves
to keep the turbulence level more uniform with time, smoothing
4000 out the short-term response to inviscid flow variations such as
those predicted by the code. This is less of a problem on the
blade suction surface, where more of the turbulence is gen-
3000 erated in the boundary layer.

Influence of Airfoil Geometric Variations. There is always

a question in the numerical modeling of fluid flow as to whether
"? 2000 -
the geometry modeled is an accurate representation of the
physical device. This can often be important in turbomachinery
since the design intent is not always realized at operating con-
1000 - ditions and because of manufacturing tolerances. To examine
the importance of NGV airfoil geometric variations in this
turbine, the raw data were ensemble averaged over 30 rotor
revolutions for each of the 36 NGV passages (phase lock av-
0 0.5 1 1.5 2 eraging), yielding the time average flow in each passage. These
data, shown in Fig. 15 for a transducer on the crown of the
Time / Blade Passing Period
suction surface corresponding to Fig. 8(b), illustrate the spatial
Fig. 13 Comparison of measured heat transfer, that calculated by the
variation in heat transfer that a rotor blade sees during one
code, and that inferred from the calculated static pressure history using revolution. The average passage heat flux (replotted from Fig.
the shock compression model 8b) is shown next to the variation about the annulus. From

- Calculation
Data
Time = 0.75
I

Time = 1.1

— i 1 1 1
0 1 2 o 5 10 15 20
Time / Blade Passing Period Turbulent / Laminar Viscosity

(a) Heat flux measurement (b) Mach No. distribution (c) Turbulent viscosity distribution
and prediction through the boundary layer through boundary layer
at t = 0.75 and t = 1.1 at t = 0.75 and t = 1.1

Fig. 14 Algebraic turbulence model only responds to instantaneous velocity fluctuations overestimating the drop in heat
transfer due to decreasing free-stream velocity on the rotor pressure surface

824 / Vol. 114, OCTOBER 1992 Transactions of the ASME

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 DESIGN INCIDENCE DESIGN INCIDENCE
3000-
6000-, NGV NGV
#1 .? Max. Passage Ave.
#36 Average M Time-Averaged Meas.
NGV • Min. Passage Ave.
— Time-Averaged
E 2000 Unsteady Calc.
E
3

V)
1000-

1 1 1
— i
12 24 36 0 2 Suction Surface Pressure Surface
Time / Vane Passing Period Fractional Wetted Surface

6000 -, INCIDENCE 3000-, -10° INCIDENCE

Max. Passage Ave.
Time-Averaged Meas.
Min. Passage Ave.
Time-Averaged
Average -a 2000 Unsteady Calc.
o NGV |
~® 3000 z
(A
0) 0)
w
3 (0
100
I °

36 0 2
Suction Surface Pressure Surface
Time / Vane Passing Period
Fig. 15 Rotor blade heat flux distribution measured about the annulus
Fractional Wetted Surface
on the suction surface crown (location as in Fig. 8b) compared to that Fig. 16 Time-averaged heat flux about the rotor blade showing the
for the average NGV passage. There are 36 NGVs in this turbine maximum and minimum average (d-c) value corresponding to the pas-
sages illustrated in Fig. 15

this figure, we note two things. The first is that the passage- Inspection of the NGV ring revealed a ± 4 percent variation
to-passage variation in heat transfer due to shock impingement in nozzle throat area. The area ratio could not be readily
is quite noticeable, and the second is that there is a much wider checked to the necessary accuracy, but a ± 3 percent variation
variation at - 1 0 deg incidence ( ± 6 5 percent of mean) than would be within the manufacturing tolerances for this exper-
at design (±14 percent). The mean (d-c) heat flux averaged imental blade set.
over the passage varies by about ±13 percent at both inci- The influence of the throat area variation was examined by
dences. The magnitude of the passage variations about the running UNSFLO with a 4 percent change in throat area (nozzle
rotor is shown in Fig. 16, where the average as well as the pitch). At - 1 0 deg incidence, the predicted integrated heat
minimum and maximum heat flux averaged over each passage transfer about the rotor changed by 5 percent. The height of
is delineated. the shock-induced heat flux spike on the suction surface crown
We have considered two simple mechanisms that could ac- (Fig. 15) changed by 8 percent. Since this is much less than
count for the observed heat flux variations. The first is a the observed 13 percent change in integrated heat flux and 43
variation in NGV throat area, which would change both the percent change in peak-to-peak shock-induced heat transfer,
mass flux (and thus the rotor Reynolds number) and the NGV we conclude that NGV throat area variation alone is not re-
exit Mach number (and thus shock strength, since the choked sponsible for the observed differences.
rotor flow makes this a two-throat problem). The influence We have crudely estimated the magnitude of NGV area ratio
of Reynolds number would not be much different at the two effects as follows. From the one-dimensional continuity equa-
rotor incidence angles, however. A second source of heat flux tion (Shapiro, 1953), the change in Mach number (M) can be
variation could be differences in area ratio between the NGV related to the change in area (A) by
throat and trailing edge. This would change the strength of
the NGV trailing edge shock waves that impinge on the rotor dM1 2 + ( 7 - l ) M 2 dA
(3)
and enhance the heat transfer. Both mechanisms will alter the M M2-l A
rotor relative incidence angle. The pressure change, AJP, across a weak shock wave is

Journal of Turbomachinery OCTOBER 1992, Vol. 114 / 825

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 proach in estimating the streamtube height distribution
^ = *> (M2 - i ) (4) throughout the turbine stage.
P, 7+1
Streamtube tube height variations may explain other dif-
Combining Eqs. (3) and (4) gives ferences between the code and measurements. At the leading
edge, for example, the calculation predicts a considerable mod-
d ulation of the heat transfer as a shock wave (shock 2a of Fig.
V 1 / _ M 2 [ 2 + (Y-- 1)M2] dA 9) sweeps around from the suction surface. However, the meas-
(5) urements show little evidence of this modulation, implying
j
/AP\ ~ ( M 2 - I) 2 A
either that the inviscid shock structure is calculated inaccurately
1 U; here (i.e., unlike the calculation, the shock wave never reaches
the leading edge in the turbine measurements) or that there is
From Eq. (1), the change in heat transfer will be proportional
to the change in shock pressure rise (assuming the shape of a problem with the boundary layer calculation near the stag-
the pressure time history remains the same and just the mag- nation point. We cannot otherwise explain the discrepancy at
nitude changes), thus this time.
One of the more puzzling results to us is the relatively small
ANu difference between the steady fully turbulent rotor calculation
<Nu and the time average of the unsteady calculation. It very much
M2[2 + ( 7 - l ) M 2 ] dA
(6) appears that, integrated over time, the unsteadiness plays a
ANu ' (M2-!)2 A relatively small role in this highly loaded transonic turbine,
Nu even though both the calculations and data show the unstead-
iness to be very large.
Note that Eq. (6) implies that, for a given geometric variation This is in many ways a progress report on an ongoing effort
(dA/A), the variation in peak heat transfer will be greater as as the code continues to evolve. For example, true grid in-
the Mach number approaches one. At the design incidence, dependence has not been rigorously demonstrated in the sense
the design value of the NGV exit Mach number is 1.22, while that the grid is continually refined in all directions until there
at - 1 0 deg incidence it is 1.13. Equation (6) quantitatively are not changes in the calculated flowfield, although a finer
explains the larger fluctuations observed at - 10 deg incidence. boundary layer grid has been run with little change in the
We can then use Eq. (6) to estimate the amount of blade-to- calculated heat transfer. Also, work is ongoing to move from
blade geometric variation required to account for the observed an algebraic turbulence model to a more sophisticated one.
heat flux variations. For both incidences, about a 2 percent Thus, many of the questions raised in this work will continue
variation in NGV area ratio will account for the ±10 to ±43 to be addressed.
percent variation in shock-induced heat transfer measured.
Thus, it appears to us that the measured passage-to-passage
Summary
variation in heat flux is fully explained as a combination of a
variation in NGV throat areas and in NGV throat-to-exit area Time-resolved heat transfer measurements and calculations
ratios. However, the variation does not explain any of the have been compared at two operating points. In both cases,
differences between calculation and measurement. the code predicts about 90 percent of the measured integrated
In production turbines, the total nozzle throat area is closely heat load to the blades. For one flow condition, the discrepancy
controlled since it establishes the engine working line. The is mainly on the first part of the suction surface and is thought
individual blade passages, however, may vary by the several to be due to an underestimation of the local shock strength.
percent discussed above. Thus, we might expect production At the reduced loading conditions, the shock waves in the
transonic turbine rotors to have concomitant annular varia- numerical solution are weaker and the predicted heat transfer
tions in heat load (on the order of ±10 percent). Since the more closely matches the measurements. At this - 1 0 deg
phenomena we have discussed are approximately linear, the incidence flow condition, however, the heat transfer on the
total heat load in a typical built-up turbine with geometric pressure surface is underpredicted. Examination of the time-
variation similar to the one studied here will be about the same resolved data showed ±13 percent variations in average heat
as a geometrically uniform one. The built-up machine sees a flux about the annulus. This was explained as due to 2 percent
once-per-revolution heat load fluctuation, which the uniform variations in NGV geometry. The importance of stream-tube
turbine does not, but we know of no reason why this should height on the detailed heat transfer was emphasized.
be detrimental since the thermal penetration depth at shaft Although this paper has focused on the differences between
frequencies is small. In other words, while academically in- the calculational results and the measurements in order to
teresting, we do not see any practical import to the observed assess the impact of various modeling assumptions and nu-
passage-to-passage difference in heat flux. merical approximations, it should be noted that, both overall
and in detail, the agreement between the two is quite good.
This good agreement using a quite unsophisticated turbulence
Unresolved Influences and Further Work. It is clear from model implies that accurate prediction of heat transfer in highly
the preceding section that small geometry changes can have a loaded turbines may be more sensitive than generally realized
large influence in the details of the heat transfer in a transonic to details of the inviscid flow prediction than to the particulars
turbine. This reinforces the importance of proper specification of the turbulence model employed.
of the axial distribution of streamtube height (AVDR or hade) Clearly, a relatively fast and accurate two-dimensional un-
in a two-dimensional calculation like this. It would be sur- • steady multiblade row code is a useful tool both for interpreting
prising if the streamline curvature calculation used to prepare experimental measurements and for elucidating the complex
the inputs for the numerical calculations presented here was flow in unsteady turbomachinery.
always accurate to better than the few percent level that caused
the large variations in the measured flow from the NGV pas-
sages. Inaccurate specification of the streamtube height is, we Acknowledgments
believe, the principal reason for the observed differences in The authors would like to thank Dr. R. J. G. Norton and
shock-induced heat transfer between the calculations and the Mr. R. Haimes for their useful comments and Mr. T. Shang
measurements. A steady, three-dimensional inviscid code may for his help in running the code..This work was supported by
provide an improvement over the streamline curvature ap- Rolls Royce Inc., Mr. A. Veninger, Technical Monitor.

8 2 6 / V o l . 114, OCTOBER 1992 Transactions of the ASME

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 References Giles, M. B., 1988, "Calculation of Unsteady Wake Rotor Interaction,"
AIAA J. of Prop, and Power, Vol. 4, No. 4, pp. 356-362.
Giles, M. B., and Haimes, R., 1991, "Validation of a Numerical Method for
Addison, J. S., and Hodson, H. P., 1990a, "Unsteady Transition in an Axial Unsteady Flow Calculations," ASME Paper No. 91;GT-XX.
Flow Turbine: Part 1—Measurements on the Turbine Rotor," ASME JOURNAL Guenette, G. R., Epstein, A. H., Giles, M. B., Haimes, R., and Norton, R.
OF TURBOMACHINERY, Vol. 112, pp. 206-214. J. G., 1989, "Fully Scaled Transonic Turbine Rotor Heat Transfer Measure-
Addison, J. S., and Hodson, H. P., 1990b, "Unsteady Transition in an Axial ments," ASME JOURNAL OF TURBOMACHINERY, Vol. I l l , pp. 1-7.
Flow Turbine: Part 2—Cascade Measurements and Modeling," ASME JOURNAL Johnson, A. B., Rigby, M. J., Oldfield, M. L. G., Ainsworth, R. W., and
OF TURBOMACHINERY, Vol. 112, pp. 215-221. Oliver, M. J., 1989a, "Surface Heat Transfer Fluctuations on a Turbine Rotor
Ashworth, D. A., LaGraff, J. E., Schultz, D. L., and Grindrod, K. J., 1985, Blade Due to Upstream Shock Wave Passing," ASME JOURNAL OF TURBOMA-
"Unsteady Aerodynamics and Heat Transfer Processes in a Transonic Turbine CHINERY, Vol. I l l , pp. 105-115.
Stage," ASME Journal of Engineering for Power, Vol. 107, No. 4, pp. XX- Johnson, A. B.,"Rigby, M. J., and Oldfield, M. L. G., 1989b, "Unsteady
00. Aerodynamic Phenomena in a Simulated Wake and Shock Wave Passing Ex-
Ashworth, D. A., 1987, "Unsteady Aerodynamics and Heat Transfer Proc- periment," AGARD-CP-468, p. 9.
esses in a Transonic Turbine Stage," D. Phil. Thesis, Oxford University, United Ni, R.-H., 1981, " A Multiple Grid Scheme for Solving the Euler Equations,"
Kingdom. AIAA Journal, Vol. 20, No. 11, pp. 1565-1571.
Blair, M. F„ Dring, R. P., and Joslyn, H. D., 1989, "The Effects of Tur- Norton, R. J. G., 1988, Private Communication.
bulence and Stator/Rotor Interactions on Turbine Heat Transfer: Part II— Rai, M. M., 1985, "Navier-Stokes Simulations of Rotor-Stator Interaction
Effects of Reynolds Number and Incidence," ASME JOURNAL OF TURBOMA- Using Patched and Overlaid Grids," AIAA Paper No. 85-1519.
CHINERY, Vol. I l l , pp. 97-103. Rigby, M. J., Johnson, A. B., Oldfield, M. L. G., and Jones, T. V., 1989,
Doorly, D. J„ and Oldfield, M. J., 1985, "Simulation of the Effects of Shock "Temperature Scaling of Turbine Blade Heat Transfer With and Without Shock
Wave Passing on a Turbine Rotor Blade,'1' ASME Journal of Engineering for Wave Passing," 9th International Symposium on Air Breathing Engines, Athens,
Gas Turbines and Power, Vol. 107, No. 4,, pp. 998-1006. Greece.
Dring, R. P., Blair, M. F., Joslyn, H. D., Power, G. D., and Verdon, J. M., Schultz, D. L., Jones, T. V., Oldfield, M. L. G., and Daniels, L. C , 1977,
1986, "The Effects of Inlet Turbulence and Rotor Stator Interactions on the "A New Transient Cascade Facility for the Measurement of Heat Transfer
Aerodynamics and Heat Transfer of a Large-Scale Rotating Turbine Model, Rates," AGARD CP-227.
Vol. I—Final Report," NASA CR 4079, UTRC-R86-956480-1. Shapiro, A., 1953, Dynamics and Thermodynamics of Compressible Fluid
Dunn, M. G., Rae, W. J., and Holt, J. L., 1984, "Measurement and Analysis Flow, Vol. 1, Ronald Press.
of Heat Flux Data in a Turbine Stage," ASME Journal of Engineering for Gas Sharma, O. P., et al., 1988, "Rotor-Stator Interaction in Multi-stage Axial
Turbines and Power, Vol. 106, pp. 229-240. Flow Turbines," AIAA Paper No. 88-3013.
Epstein, A. H., Guenette, G. R., and Norton, R. J. G., 1984, "The MIT Sharma, O. P., Pickett, G. F., and Ni, R. H., 1992, "Assessment of Unsteady
Blowdown Turbine Facility," ASME Paper No. 84-GT-116. Flows in Turbines," ASME JOURNAL OF TURBOMACHINERY, Vol. 114, pp. 79-
Epstein, A. H., Guenette, G. R., Norton, R. J. G., and Yuzhang, C., 1986, 90.
"High Frequency Response Heat Flux Gauge," Review of Scientific Instruments, Turner, E. R., Wilson, M. D., Hylton, L. D., and Kaufman, R. M., 1985,
Vol. 57, No. 4, pp. 639-649. "Turbine Vane External Heat Transfer," NASA CR-174827.

For Your ASME Bookshelf

Recent Innovations and Experience with Plant Monitoring
and Utility Operations
Editors: I. Fruchtman, S.R. Reid, P.A. Wagner

Includes: on-line monitor of life consumption for boiler drum; data interpolation for vibration
diagnostics using two-variable correlations; reexamination criteria for high energy piping in fossil
plants; design and operation of a portable turbine lube oil filtration skid; more.

1991 ISBN 0-7918-0823-8 PWR-Vol. 15 88 pp.

Order No. H00655 $30 LisV$15 ASME Members

To order write ASME Order Department, 22 Law Drive, Box 2300, Fairfield, NJ 07007-2300
or call 800-THE-ASME (843-2763) or fax 201-882-1717.

Journal of Turbomachinery OCTOBER 1992, Vol. 114 / 827

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use