Siemens Sodenberg

Heat Transfer Correlations for Gas Turbine Cooling
LITH-IKP-EX--05/2313--SE

Jenny Sundberg
Acknowledgements
I would like to thank the members of the group working with aerodynamics and
cooling of blades and vanes, GRCTA at Siemens in Finspng for always taking time
to answer my questions. Also, special thanks to my supervisors at Siemens Mats
Annerfeldt and Yuri Mamon and to my supervisors at LiTH Dan Loyd and Johan
Svensson for your support. In the Siemens office in Lincoln, I would like to thank
John Maltson, Anthony Davis and Andrew Down for helping me with the Multipass
program code. I would also like to thank Mats Kinell for helping me with the Q3D
program code and Xiufang Gao and Daniel Lrstad for your help with correlations
Abstract
A first part of a Heat Transfer Handbook about correlations for internal cooling of
gas turbine vanes and blades has been created. The work is based on the cooling of
vanes and blades 1 and 2 on different Siemens Gas Turbines. The cooling methods
increase the heat transfer in the cooling channels by increasing the heat transfer
coefficient and/or increasing the heat transfer surface area. The penalty paid for the
increased heat transfer is higher pressure losses.

Three cooling methods, called rib turbulated cooling, matrix cooling and impingement
cooling were investigated. Rib turbulated cooling and impingement cooling are
typically used in the leading edge or mid region of the airfoil and matrix cooling is
mostly applied in the trailing edge region.

Literature studies for each cooling method, covering both open literature and internal
reports, were carried out in order to find correlations developed from tests. The
correlations were compared and analyzed with focus on suitability for use in turbine
conditions. The analysis resulted in recommendations about what correlations to use
for each cooling method.

For rib turbulated cooling in square or rectangular ducts, four correlations developed
by Han and his co-workers [3.5], [3.8], [3.9] and [3.6] are recommended, each valid
for different channel and rib geometries. For U-shaped channels, correlations of
Nagoga [3.4] are recommended.

Matrix cooling is relatively unknown in west, but has been used for many years in the
former Soviet Union. Therefore available information in open literature is limited.
Only one source of correlations was found. The correlations were developed by
Nagoga [4.2] and are valid for closed matrixes. Siemens Gas Turbines are cooled with
open matrixes, why further work with developing correlations is needed.

For impingement cooling on a flat target plate, a correlation of Florschuetz et al. [5.7]
is recommended for inline impingement arrays. For staggered arrays, both the
correlations of Florschuetz et al. [5.7] and Hglund [5.8] are suitable. The correlations
for impingement on curved target plate gave very different results. The correlation of
Nagoga is recommended, but it is also advised to consult the other correlations when
calculating heat transfer for a specific case.

Another part of the work has been to investigate the codes of two heat transfer
programs named Q3D and Multipass, used in the Siemens offices in Finspng and
Lincoln, respectively. Certain changes in the code are recommended.
Nomenclature
The original parameters of the references have been used for most cases, to facilitate
comparison to references and to adjust to commonly used parameters for each cooling
method. Therefore, one parameter sometimes has multiple meanings. For those cases,
the cooling method it is valid for is written after the explanation of the parameter.

General
a = channel cross-section area [m
2
]
a
A
= channel flow cross-section area at matrix side bound [m
2
]
A = area [m
2
]
A
f
=
open area ratio, i.e. ratio of jet hole area to opposing
heat transfer surface area
[-]
A
i
=
index of alloy heat resistance, see equation (4.11),
6
10 6 . 2 =
i
A for ally of interest here
[Pa]
A
r
=
coefficient in Q3D for Nu calculation on ribbed side
wall
[-]
A
row
= area of all holes in a spanwise impingement row [m
2
]
A
s
=
coefficient in Q3D for Nu calculation on smooth side
wall
[-]
b =
width of a 2D slot with equivalent area of a single row
of impingement
[m]
C = friction enhancement factor [-]
C
d
= discharge coefficient [-]
C
d,s
= discharge coefficient for a thin orifice [-]
C
p
= specific heat at constant pressure [J/(KgK)]
C
v
= specific heat at constant volyme [J/(KgK)]
cbs = vector in function Sfun in the Q3D program code [-]
cfs = vector in function Sfun in the Q3D program code [-]
d = hydraulic diameter of channel (matrix cooling) [m]
d = hole diameter (impingement cooling) [m]
D
= effect of centrifugal force on flow in matrix side bound [-]
D
h
= hydraulic diameter in ribbed channel [m]
D
p
= diameter of curved target plate [m]
e = rib height [-]
e = rib height [m]
+
e
=
roughness Reynolds number [-]
2 r
f

= average friction factor for flow in duct with ribs on 2
opposite sides
[-]
F
= total area increase for SR- and SSR- schemes [-]
f = friction factor [-]
C
F =
mass exchange effect of opposite sub channels through
the inter-ribs windows,
C
F =z
[-]
CD
F =
effect of flow contraction-diffusion in matrix side
bound
[-]
F
= heat transfer area of ribs themselves [-]
F
s
=
heat transfer surface area of a corresponding smooth
duct
[-]
F
1
= total leading edge air outlet area [m
2
]
F
4
= total trailing edge air outlet area [m
2
]
G
c
=
channel crossflow mass velocity based on channel
cross section area
[kg/(m
2
s)]
G
f
= summary of cooling air mass flow [kg/(m
2
s)]
G
g
= mass flow after the compressor [kg/(m
2
s)]
G
j
= jet mass velocity based on jet hole area [kg/(m
2
s)]
j
G
= average jet mass velocity [kg/(m
2
s)]
) (
+
e G =
heat transfer roughness function, non-dimensional
temperature at rib tip
[-]
h = rib height [m]
H = channel height [m]
I
c
= cross flow interference paramterer [-]
k = width of rib top [m]
K = heat transfer enhancement factor [-]
K
A
=
heat transfer enhancement due to increased heat
transfer surface area
[-]
K
=
heat transfer enhancement factor, including effect of
increased heat transfer surface area
[-]
l = length of matrix channel (matrix cooling) [m]
l = length of control section (rib turbulated cooling) [m]
l =
lenght of target surface arc on curved target plate
(impingement cooling)
[m]
L = length of matrix (matrix cooling) [m]
L = channel length (rib turbulated cooling) [m]
L = length of impingement hole (impingement cooling) [m]
L
c
= characteristic length according to Table 2.1 [m]
L
es
= lenght of equivalent slot [m]
LE = Leading Edge
m = distance between ribs, m = t-k (matrix cooling) [m]
m = distance from stagnation line (impingement cooling) [m]
M
*
=
cross flow-to-jet flow mass velocity ratio,
j c
m m & &
[-]
n =
index of alloy heat resistance in equation (4.11),
n = 6060 for alloy of interest here
[-]
n
c
= number of streamwise rows in an impingement array [-]
N
p
= number of holes in a spanwise impinement row [-]
n
r
= Reynolds number exponent in Q3D Nu calculation for
ribbed side wall
[-]
n
s
= Reynolds number exponent in Q3D Nu calculation for
smooth side wall
[-]
Nu = Nusselt number [-]
s Nu
=
average Nu over the full length of a smooth duct [-]
Nu
= averaged Nu over an impingement array [-]
Nu
1
= Nu for the first row, without cross flow [-]
p = pressure [N/m
2
]
p
0
= plenum pressure (impingement cooling) [N/m
2
]
p
2
= channel pressure (impingement cooling) [N/m
2
]
P = rib pitch [m]
p = pressure loss [N/m
2
]

Pr = Prandtl number [-]
PS = Pressure Side
Q = heat transfer [W]
R = radius of U-shaped channel (rib turbulated cooling) [m]
R = radius of curved target plate (impingement cooling) [m]
r = specific gas constant, r = 283.9 [kJ/(kgK)]
R.A.M = Reduced Area Method
Re = Reynolds number [-]
) (
+
e R =
heat transfer roughness function, non-dimensional
velocity at rib tip
[-]
s = width of rib base [m]
S
= perimeter of blade or vane cross-section (rib
turbulated cooling)
[m]
S =
area of blade cross section with intensifiers of the
cooling method (S<S
0
) (matrix cooling)
[m
2
]
S =
with of equivalent slot according to Nagogas
definition (impingement cooling)

S
*
= area of through flow section in matrix [m
2
]
S
0
=
area of shell cross section of initial smooth channel
without intensifiers
[m
2
]
SS = Suction Side
SSR = Segmented Rib cooling scheme
SR = Surrounding Rib cooling scheme
S
T
= rib top area [m
2
]
St = Stanton number [-]
t = rib pitch [m]
T = temperature [K]
T
diff
=
temperature difference between the core flow and in
the flow near the wall
[K]
f
T
=
cooling air temperature in the sub channels
[K]
G
T =
stagnation temperature
[K]
w
T =
metal temperature
[K]
T = temperature difference [K]
w
T
=
temperature change in the blade wall in control
section
[K]
0 , w
T
=
initial temperature in the blade wall in a smooth
channel
[K]
TE = Trailing Edge
u = mean axial velocity of fluid [m/s]
) (
+ +
e U
e
= ) (
+
e R [-]
v = velocity [m/s]
v = mean velocity [m/s]
w = cooling air velocity in channel (matrix cooling) [m/s]
w = rib width (rib turbulated cooling) [m]
x
w
,

=
maximum value of w
[m/s]
W = channel width (rib turbulated cooling) [m]
W/H = aspect ratio [-]
W = matrix width (matrix cooling) [m]
W
r
= ribbed wall width at a cross-section [m]
W
s
= smooth wall width at a cross-section [m]
x = local distance measured from channel inlet [m]
x =
for U-shaped channel: distance from duct inlet to the
control section (rib turbulated cooling)
[m]
x = distance from stagnation line (impingement cooling) [m]
x = x/d [-]
x
n
= streamwise hole spacing [m]
xbs = vector in function Sfun [-]
xfs = vector in function Sfun [-]
y
n
= spanwise hole spacing [m]
z = number of channels in matrix (matrix cooling) [-]
z =
jet plate-to-impingment plate spacing (impingement
cooling)
[m]
z
a
= streamwise location from impingement line [m]
Z = 2W/(W+H) [-]

Greek Symbols

= heat transfer coefficient [W/(m
2
K)]

= rib angle (rib turbulated cooling) [rad]

=
angle of longitudinal rib [rad]
= distance between matrix top and bottom [m]

= surface roughness
= matrix end clearance, see Figure 4.2 [m]

= rib flank angle [-]
= rotation parameter [-]

= angle of flow rotation [rad]
T
=
ability to carry load of section S
T
[-]

= rib efficiency [-]
=
relative depth of matrix channel, channel width-to-
height ratio
[-]
=
specific heat ratio,
v p
C C = [-]
= thermal conductivity of fluid [W/(mK)]
=
dynamic viscosity of fluid [kg/(ms)]

= kinematic viscosity of fluid [m
2
/s]

=
portion of the duct perimeter covered by the rib
system,
s
F F =
[-]
=
cooling effectiveness, relative cooling depth
( ) ( )

=
f G w G
T T T T
[-]

= density [kg/m
3
]
= effect of tension stresses of cooling method [-]
= blade life with cooling method used [cycles]
= effect on life on turbine blade or vane [-]
m

=
hydraulic resistance in the spatial turns at matrix
side bound
[-]
= streamfunction [-]

= hydrodynamic energy effectiveness [-]

= correction factor, comparisons of cooling methods [-]

Subscripts
air = cooling air
AV = average
b = bulk
b = rib base (matrix cooling)
rib = total rib heat transfer surface
c = corrected
c = cross flow (impingement cooling)
conv = convective
d = jet hole diameter
e = reduced area method
f = film (fluid)
j = jet flow
l = channel length
n = normal distance
new = value calculated with one of the correlations from the literature study
original = value used in Q3D today
r = ribbed side in channel with 1, 2, 3 or 4 ribbed walls
rc = circular channel
r4 = channel with four ribbed sides
s = smooth channel
s4 = channel with four smooth sides
ST = line along concave matrix side wall
w = wall
x = local value
x = local distance in channel from matrix side wall (matrix cooling)
x- = axial component
0 = smooth duct
0 = jet discharge condition (impingement cooling)

= value near the wall

= tangential component
Table of Contents
1 INTRODUCTION..................................................................................................................................1
1.1 BACKGROUND................................................................................................................................1
1.2 PROBLEM DESCRIPTION.................................................................................................................2
1.3 OBJECTIVE .....................................................................................................................................2
1.4 LIMITATIONS..................................................................................................................................2
1.5 METHOD.........................................................................................................................................2
1.5.1 Method Criticism ...................................................................................................................3
1.6 DISPOSITION AND READING INSTRUCTIONS ..................................................................................3
2 THEORY.................................................................................................................................................5
2.1 FLOW AND HEAT TRANSFER BASICS .............................................................................................5
2.1.1 Heat Transfer .........................................................................................................................5
2.1.2 Friction...................................................................................................................................5
2.1.3 Flow........................................................................................................................................6
2.2 GAS TURBINE VANES AND BLADES...............................................................................................8
2.3 REFERENCES THEORY..................................................................................................................10
3 RIB TURBULATED COOLING.......................................................................................................11
3.1 THEORY........................................................................................................................................11
3.2 CORRELATIONS FROM ARTICLES .................................................................................................12
3.2.1 Results ..................................................................................................................................12
3.2.2 Analysis ................................................................................................................................12
3.2.3 Conclusions..........................................................................................................................12
3.3 CORRELATIONS IN Q3D...............................................................................................................12
3.3.1 Results ..................................................................................................................................12
3.3.2 Analysis ................................................................................................................................12
3.3.3 Conclusions..........................................................................................................................12
3.4 CORRELATIONS IN MULTIPASS ....................................................................................................12
3.4.1 Results ..................................................................................................................................12
3.4.2 Analysis ................................................................................................................................12
3.4.3 Conclusions..........................................................................................................................12
3.5 REFERENCES RIB TURBULATED COOLING...................................................................................12
4 MATRIX COOLING...........................................................................................................................12
4.1 THEORY........................................................................................................................................12
4.1.1 What is Matrix Cooling? .....................................................................................................12
4.1.2 Fin Effect of Ribs .................................................................................................................12
4.1.3 Summary of Literature Study...............................................................................................12
4.2 CORRELATIONS FROM LITERATURE.............................................................................................12
4.2.1 Results ..................................................................................................................................12
4.2.2 Analysis ................................................................................................................................12
4.2.3 Conclusions..........................................................................................................................12
4.3 RIB EFFECTIVENESS IN Q3D........................................................................................................12
4.3.1 Results ..................................................................................................................................12
4.3.2 Analysis ................................................................................................................................12
4.3.3 Conclusions..........................................................................................................................12
4.4 REFERENCES MATRIX COOLING..................................................................................................12
5 IMPINGEMENT COOLING.............................................................................................................12
5.1 THEORY........................................................................................................................................12
5.1.1 What is Impingement Cooling? ...........................................................................................12
5.1.2 Summary of Literature Survey.............................................................................................12
5.2 FLAT TARGET PLATE ...................................................................................................................12
5.2.1 Results ..................................................................................................................................12
5.2.2 Analysis ................................................................................................................................12
5.2.3 Conclusions..........................................................................................................................12

5.3 CURVED TARGET PLATE..............................................................................................................12
5.3.1 Results ..................................................................................................................................12
5.3.2 Analysis ................................................................................................................................12
5.3.3 Conclusions..........................................................................................................................12
5.4 REFERENCES IMPINGEMENT COOLING.........................................................................................12
6 CONCLUSIONS...................................................................................................................................12
6.1 RIB TURBULATED COOLING ........................................................................................................12
6.2 MATRIX COOLING........................................................................................................................12
6.3 IMPINGEMENT COOLING ..............................................................................................................12
7 FUTURE WORK.................................................................................................................................12
APPENDIX 1 - RIB TURBULATED COOLING...............................................................................12
APPENDIX A CORRELATIONS..............................................................................................................12
APPENDIX B CORRELATIONS FOR U-SHAPED CHANNELS ..................................................................12
APPENDIX C RESULTS FOR CASE A....................................................................................................12
APPENDIX D EFFECT OF RIB ANGLE...................................................................................................12
APPENDIX E EFFECT OF PITCH-TO-RIB HEIGHT .................................................................................12
APPENDIX F EFFECT OF ASPECT RATIO..............................................................................................12
APPENDIX G EFFECT OF RIB HEIGHT..................................................................................................12
APPENDIX H Q3D CORRELATIONS.....................................................................................................12
APPENDIX I MULTIPASS RESULTS ......................................................................................................12
APPENDIX J INTERESTING ARTICLES..................................................................................................12
APPENDIX 2 - MATRIX COOLING...................................................................................................12
APPENDIX A ZINC TEST......................................................................................................................12
APPENDIX B CORRELATIONS AND RANGES OF [4.2] ..........................................................................12
APPENDIX C CORRELATIONS AND RANGES OF [4.5] ..........................................................................12
APPENDIX D RESULTS FOR A TYPICAL TURBINE BLADE ...................................................................12
APPENDIX E INTERESTING ARTICLES .................................................................................................12
APPENDIX 3 IMPINGEMENT COOLING.....................................................................................12
APPENDIX A CORRELATIONS AND RANGES........................................................................................12
APPENDIX B FLAT TARGET PLATE, CASE B.......................................................................................12
APPENDIX C FLAT TARGET PLATE, COMPONENTS.............................................................................12
APPENDIX D FLAT TARGET PLATE, RANDOM TESTS .........................................................................12
APPENDIX E FLAT TARGET PLATE, EFFECT OF PARAMETERS............................................................12
APPENDIX F ANALYSIS OF CORRELATIONS........................................................................................12
APPENDIX G CURVED TARGET PLATE, COMPONENTS .......................................................................12
APPENDIX H CURVED TARGET PLATE, RANDOM TESTS....................................................................12
APPENDIX I CURVED TARGET PLATE, EFFECT OF PARAMETERS .......................................................12
APPENDIX J INTERESTING ARTICLES..................................................................................................12

Table of Figures
Figure 2.1 Boundary Layer [2.3]. ______________________________________________________6
Figure 2.2 Velocity Profiles in Pipe Flow [2.1]. ___________________________________________7
Figure 2.3 Schematics of a Gas Turbine [[2.6]. ___________________________________________8
Figure 2.4 Schematic Drawing of SGT -800 [2.7]. _________________________________________8
Figure 2.5 Turbine Blade (Right), and Turbine Vane (Left ) [2.8]._____________________________9
Figure 2.6 Typical Cooling Concept [2.9] _______________________________________________9
Figure 3.1 A Ribbed Channel. ________________________________________________________11
Figure 3.2 Flow around Ribs [3.1]. ___________________________________________________11
Figure 3.3 Comparison of K and C. Ribbed Duct is Number 3 [3.3].__________________________12
Figure 3.4 Geometry of SR-Scheme, Front View (Left) and Top View (Right).___________________12
Figure 3.5 Geometry of SSR-Scheme, Front View (Left) and Top View (Right).__________________12
Figure 3.6 Secondary Flow induced by Angled Ribs[3.1]. __________________________________12
Figure 3.7 Flow Patterns for Different Rib Spacings [3.13]. ________________________________12
Figure 3.8 Triangular Ducts Investigated by Metzger & Vedula [3.1]. ________________________12
Figure 3.9 Triangular Ducts Investigated by Zhang et al [3.1]. ______________________________12
Figure 3.10 Ranges for Correlations and Channels in SGT -700 and SGT -800. _________________12
Figure 3.11 Nu on Ribbed Side Wall for Case A. _________________________________________12
Figure 3.12 Friction Factor on Ribbed Side Wall for Case A. _______________________________12
Figure 3.13 Nu on Smooth Side Wall in Ribbed Duct for Case A. ____________________________12
Figure 3.14 K on Smooth Side Wall in Ribbed Duct for Case A. _____________________________12
Figure 3.15 Results for SR- and SSR-Scheme Calculations. _________________________________12
Figure 3.16 Cross-Section 4, Vane 2, SGT -700. _________________________________________12
Figure 3.17 Temperature in Vane 2, SGT -700. __________________________________________12
Figure 3.18 Temperature Difference in Vane 2, SGT -700. _________________________________12
Figure 3.19 Q3D Interface for Ribbed Channels. _________________________________________12
Figure 3.21 Original Interface (Left) and Corrected Interface (Right). ________________________12
Figure 4.1 Example of a Matrix Geometry [4.1]__________________________________________12
Figure 4.2 Open Matrix. ____________________________________________________________12
Figure 4.3 Geometry of Longitudinal Rib. ______________________________________________12
Figure 4.4 Geometry in Matrix Side Bounds [4.2].________________________________________12
Figure 4.5 Line ST on Matrix Shell inner Concave Surface._________________________________12
Figure 4.6 Comparsion of Cooling Depht [4.2]. __________________________________________12
Figure 4.7 Three Matrixes Investigated by Jurchenko and Malkov [4.3]. ______________________12
Figure 4.8 Nusselt Number Enhancement for the Tests with End Clearances [4.3]. ______________12
Figure 4.9 Height and Chord of an Airfoil.______________________________________________12
Figure 4.10 Alfa as a Function of Re
d
in Initial Section, Re
d
= 15 000. ________________________12
Figure 4.11 Alfa as a Function of x/d in Inital Section, Re
d
= 15 000. _________________________12
Figure 4.12 Alfa as a Function of x/d for Intial and Basic Sections, Re= 11000._________________12
Figure 4.13 Rib effectiveness interface in Q3D. __________________________________________12
Figure 4.14 Q3D Rib Geometry. ______________________________________________________12
Figure 5.1 Impingement Cooling Setup. ________________________________________________12
Figure 5.2 Curved Target Plate[5.1]. __________________________________________________12
Figure 5.3 Impingement Cooling in Turbine Vane [5.2]. ___________________________________12
Figure 5.4 Impinging Jet (Left, [5.3]) and Free Jet (Rigth) [5.4]. ____________________________12
Figure 5.5 Re and G
c
/G
j
as a Function of Row Number for x
n
/d=y
n
/d =3, z/d=2.75 and Re
d
=29900 and
57800 [5.14]. _____________________________________________________________________12
Figure 5.6 Alfa for Case C. __________________________________________________________12
Figure 5.7 Q3D Comparison for Case C. _______________________________________________12

Introduction

1
1 Introduction
In this opening chapter, the background, problem description and objective of the
work is presented, together with an explanation of the methods used. The limitations
of the work and the disposition of the report are also offered.
1.1 Background
The efficiency and power output of a gas turbine increases with higher turbine inlet
gas temperature. Modern gas turbine vanes and blades are exposed to gas with
temperatures which far exceeds the melting point of the component material. Thus,
the blades and vanes have to be cooled in order to lower the temperature. When
cooling the component it is important to know the correct boundary conditions, to
avoid creating too large temperature gradients. Large temperature gradients cause
thermal stresses and significantly decrease the component life.

Both internal and external cooling is used in turbine blades and vanes. The cooling air
is extracted from the compressor. The cooling affects the gas turbine in two ways.
First, less mass flow is available for combustion in the combustion chamber. Second,
the trailing edge thickness has to be increased, which creates a larger wake behind the
trailing edge which affects the aerodynamics negatively.

The extraction of air decreases the efficiency of the turbine, since less air is available
for power generation. Maximum cooling with minimum cooling air is therefore
desired.

This work concerns internal cooling of turbine vanes and blades 1 and 2 on Siemens
Gas Turbines 700 and 800, shortly called SGT -700 and SGT -800. These components
are situated next to the combustor chamber, and are therefore exposed to the highest
temperatures. The report is primarily addressed to gas turbine cooling engineers.

Blades and vanes are cooled by internal channels, through which the cooling air flows
in different schemes and configurations. The cooling air decreases the channel wall
temperature by convective cooling. A number of cooling methods are applied to
different part of the vane or blade. To make the cooling systems more efficient and
spend a minimum of air, the cooling systems nowadays usually include features that
increase the heat transfer coefficient and/or increasing the heat transfer surface area.
The heat transfer coefficient is increased by enhancement of the flow turbulence and
by breaking the flow boundary layer. The penalty paid for the increased heat transfer
is higher pressure loss.

Many methods exist in theory, but only a handful are widely used in practice. Three
cooling methods are investigated here. They are called rib turbulated cooling, matrix
cooling and impingement cooling. The flow of these cooling methods is complicated.
Heat transfer coefficients and pressure losses can be calculated by Computational
Fluid Dynamics programs, or CFD, but modeling is very time consuming. Since it is
often necessary to investigate many different cases, it is convenient to use semi
empirical correlations based on experiments.

Introduction
2
To evaluate how effective these cooling methods are, the heat transfer and friction of
the cooling methods are compared to a corresponding smooth channel. The
augmentation of heat transfer and friction are expressed by the heat transfer
enhancement factor and the friction enhancement factor. These factors are used in
heat transfer programs for flow and temperature calculations of gas turbine
components.
1.2 Problem Description
The enhancement factors are calculated from correlations based on tests or
approximated from experiments. A large number of correlations exist in open
literature, and it is not obvious which ones are most suitable for different flow
schemes and geometries. There is a need to gather information about correlations and
to evaluate them, in order to find out which correlations to use for different cases.
1.3 Objective
The objective of this work is to create a first part of a Heat Transfer Handbook for
internal cooling of gas turbine vanes and blades. This part of the Heat Transfer
Handbook will contain recommendations for how to calculate the heat transfer- and
friction enhancement factors for rib turbulated cooling, matrix cooling and
impingement cooling. A sub objective is to investigate the code of two heat transfer
programs used at Siemens, in order to examine how well the code agrees with open
literature correlations.
1.4 Limitations
The effect of rotation is not considered. Only articles with correlations based on tests
are considered, so for example articles about CFD analysis are not in the scope of this
work.

Concerning ribbed channels, correlations for ribs with 30 to 90 angles to the air
flow velocity vector are investigated. Only continuous ribs with inline pattern are
considered. Investigated ducts have rectangular, U-shaped or square cross-section.

The limitations of matrix cooling have been that there are limited sources of
information and that the work has a time limit.

For impingement cooling, the average heat transfer of impingement of multiple jets on
a concave or flat surface is investigated. The target plate is solid without roughness
elements, film cooling holes or other irregularities. The jets have circular cross-
section and impinge perpendicular to the jet target. Both single rows of jets and
staggered or inline array patterns are investigated.
1.5 Method
One cooling method was investigated at the time, first rib turbulated cooling, second
matrix cooling and third impingement cooling. Literature studies were performed for
each area, which covered both open literature and in-house reports. The information
from the literature studies were summarized for each cooling method. Much
information was also gathered by personal communication with cooling engineers in
the Turbine Aerodynamic and Cooling department at Siemens, GRCTA. Articles and
Introduction

3
reports containing correlations were studied with focus on what type of experiment
had been performed, correlation ranges and the applicability to turbine conditions. For
each cooling method, the correlations were compared and analyzed. The agreement
between the correlations was also investigated, since if independent investigators have
come to the same conclusion the correlations are more likely to be accurate.
Conclusions about the correlations were drawn based on information in the sources of
the correlations and from analyzing the results of the comparisons.

Parallel to the investigation of correlations in literature, the codes of two in-house
programs used at Siemens were investigated for each cooling method. In Finspng, a
heat transfer program called Q3D is used for heat transfer calculations. In the Siemens
office in Lincoln, a heat transfer program called Multipass is used. Q3D was studied
for all three cooling methods. Multipass cannot be used for heat transfer for matrix
and impingement cooling, and thus was only studied for rib turbulated cooling. The
code was checked against the correlations it was based upon.
1.5.1 Method Criticism
The recommendations of correlations are partly based on the results of a limited
number of flow cases and geometries that are relevant for turbine conditions. It is
possible that the correlations behave somewhat different for other test cases. It has not
been doable to test all cases possible for all correlations.

Parts of the Q3D code consist of numerical approximations of correlations, why it is
hard to analyze. This can cause misinterpretations.

The Multipass program itself was not available for use, but the subroutines were
investigated separately from the program. That can cause mistakes in the code
interpretations. However, the results were checked and accepted by the Lincoln office.
1.6 Disposition and Reading Instructions
All information about each cooling method is gathered in one place, to facilitate use
of the report for cooling design. Chapter 1 gives an introduction to the work and
presents background, problem description, objective, limitations and the disposition of
the report. Chapter 2 shortly presents basic flow and heat transfer theory and also the
geometry and cooling fundamentals of turbine blades and vanes. The next three
chapters summarize the work on rib turbulated cooling, matrix cooling and
impingement cooling, in that order. Each chapter contains basic theory about the
cooling method, a summary of the literature study, results, analysis and discussion and
finally conclusions. Both the work about correlations and the program codes are
presented in each cooling method chapter. References for the cooling methods are
also included in each separate chapter. Next, Chapter 6 summarizes the conclusions
drawn for all cooling methods. Chapter 7 presents future work.

Theory

5
2 Theory
This chapter presents basic flow and heat transfer relationships. A brief introduction
to gas turbines and blade cooling is also offered.
2.1 Flow and Heat Transfer Basics
Definitions of flow, friction and heat transfer parameters and terms used are shortly
presented.
2.1.1 Heat Transfer
Much of the heat transfer in blades and vanes internal cooling systems takes place by
convective cooling, where heat is transferred from the hot wall to the cooling air. The
heat transfer coefficient - is defined by equation (2.1) [2.1].

( )
air w conv
T T A Q =
(2.1)
To generalize heat transfer correlations, it is common to use non-dimensional
parameters. The heat transfer coefficient is often made non-dimensional by the
Nusselt number, defined in equation (2.2).

c
L
Nu

= (2.2)
The characteristic length L
c
varies for different geometries and some examples are
given in Table 2.1 [2.2].

Table 2.1 Examples of Characteristic Lengths.
Flow Case L
c
=
flat plate, local value x
flat plate with length L, the hole plate L
cylindrical pipe d
non-cylindrical pipe D
h
Another parameter that describes convective heat transfer is the Stanton number, see
equation (2.3).

Pr Re
=
Nu
St (2.3)
2.1.2 Friction
The fanning friction factor is defined from the pressure loss according to equation
(2.4).
2 4
2
v
D
L f
p
h
=

(2.4)
The Darcy friction factor is defined according to equation (2.5).

2
2
v
D
L
f p
h
D

=

(2.5)

Theory
6
The relationship between the fanning and the Darcy friction factor is displayed in
equation (2.6).

4
D
f
f = (2.6)
2.1.3 Flow
Flow can be turbulent, laminar or in the transitional region between laminar and
turbulent. Laminar flow occurs for non-disturbed flow with relatively low velocities
and is characterised by even velocities and orderly motions. The opposite, turbulent
flow, occurs at higher velocities and is characterised by velocity fluctuations and
disordered motions. The Reynolds number describes the flow regime, see equation
(2.7). For example, flow in a pipe is laminar for Re < 2300 and turbulent for
Re > 4000 approximately.

The Reynolds number is defined as the ratio of the inertia forces to the viscous forces
in the fluid. The inertia forces depend on the flow kinetic energy and are a function of
the fluid density and the square of flow velocity. The viscous forces depend on the
fluid viscosity and the flow velocity. The characteristic length is the same as for the
Nusselt number, see Table 2.1.

c
L v
= Re
(2.7)
Flow in a pipe or over a flat plate can be divided into two regions. First the boundary
layer region, where the friction from the surface below affects the velocity profile.
Second, the inviscid flow region, where the friction effect is negligible, see Figure
2.1.

Figure 2.1 Boundary Layer [2.3].
In a pipe, where flow enters with uniform velocity, the boundary layer will grow
thicker and thicker further from the pipe entrance and thus, the inviscid region will
decrease. After a certain length from the inlet, the hydrodynamic entry length, the
velocity profile is uniform and the flow is said to be hydrodynamically developed, see
Figure 2.2 .
Theory

7

Figure 2.2 Velocity Profiles in Pipe Flow [2.1].
If also the temperature profile is constant, the flow is called fully developed. If neither
cooling nor heating occurs, hydrodynamically developed flow is equivalent of fully
developed flow. [2.1]

The Prandtl number describes the thickness of the boundary layer, see equation (2.8).

( )
p
C
=

Pr
(2.8)
The definition of mass flow is important for correlations. Mass flow is defined in
equation (2.9).

v A m = &
(2.9)
A discharge coefficient is often used for mass flow definitions in correlations. It is
defined in equation (2.10) [2.4].

ideal
real
d
m
m
C
&
&
=
(2.10)
Mass velocity G is defined in equation (2.11) [2.5].

A
m
G
&
= (2.11)

Velocity profile
Hydrodynamically
developed region
Hydrodynamic entry
region
x
Velocity boundary layer
Theory
8
2.2 Gas Turbine Vanes and Blades
Gas turbines consist of three main parts; the compressor, the combustor and the
turbine see numbers (1), (2) respectively (3) in Figure 2.3.

Figure 2.3 Schematics of a Gas Turbine [2.6].
The compressor increases the pressure of the inlet air before the air enters the
combustion chamber. In the combustor, air is mixed with fuel and the gas mixture
combusts at high temperatures. The temperature increase makes the gas expand. In the
turbine, the gas first reaches vane 1, in which the expanding gas is directed towards
blade 1. In blade 1, the gas stream is deflected, which causes a torque on the shaft.
The torque brings the shaft to rotate, which is the useful movement of the engine. The
rotational movement can then be used for various purposes, such as working oil and
gas pumps or, as in Figure 2.3 operating a generator (4) that produces electricity
which is transformed to high tension by a transformer (5). A picture of SGT -800 is
seen in Figure 2.4.

Figure 2.4 Schematic Drawing of SGT -800 [2.7].
Figure 2.6 shows the nomenclature for a blade or vane. The front part is called the
leading edge and the back part the trailing edge. Low pressure and high pressure sides
are called suction side and pressure side, respectively.
Theory

9

Figure 2.5 Turbine Blade (Right), and Turbine Vane (Left ) [2.8].
Figure 2.6 describes a typical cooling concept of a turbine blade.

Figure 2.6 Typical Cooling Concept [2.9]
Impingement cooling is often applied at the leading edge. In the mid chord region, rib
turbulated cooling is common and in the trailing edge, pin fin cooling or matrix
cooling is used. Platforms on vanes are often cooled by impingement. Exampes of a
blade and a vane are displayed in Figure 2.5.

Platforms
Theory
10
2.3 References Theory

[2.1] engel, Y. A. and Turner, R. H. (2001). Fundamentals of Thermal-Fluid
Sciences, McGraw-Hill Companies, New York (ISBN 0-07-118152-0)
[2.2] Storck, K., Karlsson, M., Andersson, I., Loyd, D. Formelsamling i termo- och
fluiddynamik,(2001), Institutionen fr Konstruktion och Produktion,
Linkpings Tekniska Hgskola
[2.3] Appelqvist, B. and Loyd, D. Grundlggande teknisk strmningslra, (1979),
Institutionen fr Konstruktion och Produktion, Linkpings Tekniska Hgskola
[2.4 ] Hglund, H. Experimental Investigation of Impingement Cooling Under a
Staggered Array of Circular Jets, Thesis Work at the Department of Energy
Tehnology Royal Institute of Technologi, KTH (1999)
[2.5] Holman, J. P. (2002). Heat Transfer Ninth Edition, McGraw-Hill Companies,
New York (ISBN 0-07-112230-3)
[2.6] Vattenfall (2005).
http://www.vattenfall.se/om_vattenfall/energikunskap/gas/gasturbin.asp
(2005-12-14)
[2.7] Siemens Industial Turbomachinery, Finspng
[2.8] Vontobel (2005). http://people.web.psi.ch/vontobel/images/turb8_tn.jpg
(2005- 12-13)
[2.9] Han, J.C., Dutta, S., & Ekkad S.V. (2000). Gas Turbine Heat Transfer and
Cooling Technology, Taylor & Francis, New York (ISBN 1-56032-841-X)

Rib Turbulated Cooling

11
3 Rib Turbulated Cooling
The concepts of rib turbulated cooling are described, together with a summary of the
literature study. After that, the results, analysis and conclusions of investigated
correlations and program codes are presented.
3.1 Theory
In turbine vanes and blades, ribs are mostly used in the internal cooling channels in
the middle of the component. The ribs are situated on opposite walls, almost always
towards the pressure side and suction side, see Figure 3.1 [3.1]. Sometimes only one
side has ribs, because the internal cooling has to match the external load, which can be
different on pressure and suction side.

Figure 3.1 A Ribbed Channel.
The ribs cause separation from the flow at the rib tops, and reattachment to the flow
between the ribs. This disturbs the boundary layer, which leads to increased heat
transfer, see Figure 3.2. Separation and reattachment increases the turbulence of the
flow, which mixes the fluid elements near the wall with the cooler ones in the middle
of the flow.[3.1] A new thin boundary layer is started at the reattachment point after
every rib.

Figure 3.2 Flow around Ribs [3.1].
The increase of heat transfer depends mainly on the aspect ratio of the duct, the flow
Reynolds number and the rib configuration [3.2]. Figure 3.2 shows the setup of a
ribbed duct, and the important rib parameters such as rib height - e, pitch, and rib
angle. Ribs that are orthogonal to the flow direction are called transverse ribs.

Due to the complex flow that the ribs create, the flow has to be described by empirical
correlations from experiments, instead of by analytical solutions [3.1].

To compare the results from different experiments and with turbine blade channels,
certain dimensionless ratios of the geometry is of extra interest. Important
H
W
Ribs
P

12
dimensionless parameters for ribbed ducts are the channel aspect ratio, the pitch-to-rib
height ratio and the rib height-to-hydraulic diameter ratio. The definition of rib
angle is also important, see Figure 3.2.
The disadvantage of ribbed duct is that the pressure drop is increased by the ribs.
However, since the ribs are relatively small the pressure drop is often acceptable. The
largest pressure drop usually occurs in the channel bends. A comparison of the heat
transfer to the pressure drop was made for different cooling methods [3.3]. Curve
number 3 in Figure 3.3 represents a duct with ribs on two opposing walls. It is clear
that the friction grows faster than the heat transfer.

Figure 3.3 Comparison of K and C. Ribbed Duct is Number 3 [3.3].
The Nusselt number in a smooth duct is highest near the inlet, due to the turbulent
developing flow. Further from the inlet the Nusselt number decreases. However, the
smooth wall Nusselt number in a ribbed duct is about 20% to 60% higher than that in
a smooth duct. In a ribbed duct, the Nusselt number on both the ribbed and smooth
walls are also fluctuating, due to the separation from and reattachment between the
ribs. The ribbed walls have higher fluctuations and higher Nusselt number than the
smooth walls. [3.1]

The leading edge channels are often U-shaped. Rib turbulated cooling of U-shaped
ducts were investigated by Nagoga, [3.3]. He tested two different cooling methods,
called the SR-scheme and the SSR-scheme. Both the SR- and SSR-scheme consist of
a periodic array of ribs that are spaced at a pitch P either perpendicularly or at angle
to the flow. The ribs are attached to the concave side of the duct. SR stands for
Surrounding Ribs, and the geometry can be seen in Figure 3.4. [3.4]

Figure 3.4 Geometry of SR-Scheme, Front View (Left) and Top View (Right).
The SSR-scheme consists of segmented ribs, which are placed in a semi-circular
longitudinal duct. Figure 3.5 describes the geometry of the SSR-scheme. [3.4]
C
K

13

Figure 3.5 Geometry of SSR-Scheme, Front View (Left) and Top View (Right).

14
Summary of Literature Study
The heat transfer enhancement decreases with increased Re [3.5] [3.6]. The reason is
that flow reattaches faster for higher Re [3.5]. For Re = 60 000, the flow reattached
approximately 2 rib heights downstream, compared to 6 rib heights for Re = 10 000
[3.5].

[3.7] found that for a rib angle of 45, Nu is increased but f remains the same as for
=90 . [3.8] and [3.9] reported that angled ribs create a

Figure 3.6 Secondary Flow induced by Angled Ribs[3.1].
secondary flow according to Figure 3.6. For = 60 to 30, the secondary flow
makes Nu decrease along the rib axes, from left to right in Figure 3.6 Secondary Flow
induced by Angled Ribs[3.1]. [3.9].

The ribs cause Nu to have a periodic distribution, due to the continuous separation and
reattachment of the flow. For a square duct with transverse ribs, Nu decreases from
the inlet along the channel in the streamwise direction, until x/D
h
> 3 where the
periodic Nu reaches a constant value. For a square duct with angled ribs, Nu also
decreases after the inlet, but for x/D
h
> 3 Nu increases again, due to the secondary
flow induced by the rib angle. For ducts with larger W/H than 1, this effect is
gradually decreased. [3.9]

Han et al. [3.8] investigated the same phenomenon as described above but in channels
with aspect ratios less than 1. They found that for = 90 or 30, local Nu is
periodically distributed between the ribs after x/D
h
> 3, i.e. Nu neither increase nor
decrease. For = 60 or 45 [3.8] found that the periodic Nu increases along the
channel due to the secondary flow described above.

Han and Park [3.9] also found that for a duct with square cross-section, highest Nu
and f were obtained for =60 . For a duct with W/H = 4, the highest Nu and f were
obtained for = 90. The best heat transfer performance, i.e. high Nu and low f, were
obtained for = 30 and 45 [3.10]. Ribs with 60 and 45 angle to the flow had 25
to 30 % better heat transfer performance than transverse ribs [3.8].
Concerning aspect ratio, [3.5] investigated ducts with aspect ratios from to 4 and
found that ducts with small aspect ratios had better heat transfer performance than
large aspect ratio ducts. [3.8] investigated ducts with aspect ratios less than 1, and also
found that channels with smaller aspect ratios perform better.

The effect of rib height was investigated by [3.7]. It was concluded that increased
e/D
h
led to increased friction factor. An experimental study took place at Siemens in
Finspng in 2002, which aimed to investigate the effect of rib height [3.11]. The
outcome was that the influence of rib height on the Nu heat transfer decreases with an

15
increased Re. An increase of rib height from 5 mm to 7 mm had little effect on the
heat transfer coefficient, but increased the friction factor by 2 [3.11].

The effect of rib pitch was discussed in many articles, and many investigators came to
the same conclusions. For P/e no less than 10, Nu increases with decreased P/e
[3.7],[3.5],[3.12]. Maximum heat transfer is obtained for P/e =10 [3.7], [3.10]. For
smaller rib spacing, the reattachment between ribs cannot occur, see Figure 3.7 [3.13].

Figure 3.7 Flow Patterns for Different Rib Spacings[3.13].
[3.6] investigated the effect of number of ribbed walls. They found that Nu and f
increased with the number of ribbed walls. For example, K increased from 2.04 for 1
ribbed wall to 2.63 for 4 ribbed walls when Re = 30 000. They also found that a duct
with 4 ribbed wall and W/H = , obtained 29% higher Nu than a duct with 1 ribbed
wall and Re = 30 000. A duct with 4 ribbed walls and W/H = 2 obtained 4% higher Nu
than a duct with 2 ribbed walls when Re= 30 000. [3.6]

In 1987, Metzger and Vedula studied heat transfer performance in ribbed channels
with a triangular cross-section. Different configurations of angled ribs were used,
which created secondary flow according to Figure 3.8. [3.1]

Figure 3.8 Triangular Ducts Investigated by Metzger & Vedula [3.1].
Metzger and Vedula found that the downstream angled ribs transported the main flow
towards the ribbed sides, and the upstream angled ribs carried the main flow towards
the smooth side wall. This resulted in higher heat transfer coefficient on the ribbed
16
sides for the downstream angled ribs, and higher heat transfer coefficient on the
smooth side wall for the upstream angled ribs. [3.1]

Experiments for the spiral ribs showed that for 60 angle, all three walls reached
nearly the same Nusselt number for fully developed flow. They also found that the
ribbed side wall for spiral ribbed had lower Nusselt number than the ribbed side wall
of the downstream and upstream angled ribs. [3.1]

Zhang et al. investigated channels with triangular cross-section in 1994. Experiments
were made on six geometries, see Figure 3.9.
Figure 3.9 Triangular Ducts Investigated by Zhang et al [3.1].
They found that Nu for the three-wall partial ribs were approximately 10 % higher
than for three-wall full ribs. For transverse ribs K=2 - 2.3 and C = 3.6-6.6 and for 45
angled ribs K = 1.6-1.8 and C = 1.6-2.5. In the 55 corner, K = 1.4-1.7 for transverse
ribs and K = 1.2-1.5 for 45 angled ribs. In the 35 corner, heat transfer enhancement
was low for all configurations, both transverse and for 45 angled ribs. [3.1]


17
3.2 Correlations from Articles
The results, analysis and conclusions drawn of the investigation of correlations from
articles are presented below.
3.2.1 Results concerning Correlations from Articles
Investigated Correlations
Following correlations from published articles were examined.

Webb, Eckert &Goldstein, 1970 [3.13]
Han, 1988 [3.5]
Chandra, Niland & Han, 1997 [3.6]
Han, Glicksman & Rohsenow, 1978 [3.7]
Han & Park, 1988 [3.9]
Han, Ou, Park & Lei, 1989 [3.8]
Han, 1984 [3.12]
Han, Park and Lei, 1985[3.10]
Nagoga, 2001 [3.4]

The correlations are summarized in Appendix A. The correlation ranges are showed in
Figure 3.10, together with ranges of interest for turbine cooling. The investigated
reports from Siemens did not contain correlations for ribbed ducts. More articles
about rib turbulated cooling are listed in Appendix J. The correlations are valid for
different ranges of Reynolds number - Re, rib angle, aspect ratio, pitch-to-rib height
ratio and rib height-to-hydraulic diameter ratio. Ranges for the correlations derived
from articles are shown in Figure 3.10, together with ranges that are of interest for
typical gas turbine blades and vanes in the first and second stages. The ranges of
Nagogas correlations for U-shaped ducts are displayed in Appendix B.
18
Reynolds Number
0 20 40 60 80 100 120 140 160
Re*10
-3

Aspect Ratio
0 5 10 15 20
W/H

Pitch-to-Rib Height Ratio
5 10 15 20 25 30 35 40
P/e

Webb et al. 1970
Han et al. 1978
Han 1984
Han 1985
Han & Park 1987
Han 1988
Han et al. 1989
Chandra et al 1997
Turbine components

Rib Height-to-Hydraulic Diameter Ratio
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
e/D
h

Rib Angle
20 30 40 50 60 70 80 90
Rib Angle

Figure 3.10 Ranges for Correlations and Channels in SGT -700 and SGT -800.

19
Comparison of Correlations
The correlations were compared and analyzed by using Matlab. The effect of rib
angle, aspect ratio, pitch-to-rib height ratio, rib height-to-hydraulic diameter ratio and
Reynolds number was investigated. To be able to compare as many correlations as
possible in their valid ranges, a test case with rib- and duct geometry according to
Table 3.1 was chosen. This case is referred to as Case A. All correlations except for
Webb et. als 1970, Chandra et. als 1997 and Han et. als 1978 are valid for Case A.
The latter are therefore analyzed separately.

Table 3.1 Case A, Test Case for Ribbed Ducts.
P [mm] 3.15
e [mm] 0.315
D
h
[mm] 5
W [mm] 5
H [mm] 5
Number of ribbed walls 2
Re [-] 10 000- 80 000
[] 90
W/H [-] 1
P/e [-] 10
e/D
h
[-] 0.063

Figure 3.11 and Figure 3.12 display the results for the Nusselt number and friction
factor for Case A. Nu for a smooth channel is calculated with the Dittus-Boelter
equation, see equation (3.1). The Q3D correlations will be handled in chapter 3.2.3
but they are also presented here to avoid displaying almost the same figures twice.
1 2 3 4 5 6 7 8
x 10
4
0
50
100
150
200
250
300
350
400
Nusselt Number for a Ribbed Duct
Reynolds Number
N
u
s
s
e
l
t

N
u
m
b
e
r
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984,OBS! Nu
average
Han et al. 1985
Q3D
Smooth Channel

Figure 3.11 Nu on Ribbed Side Wall for Case A.
20

4 . 0 8 . 0
Pr Re 023 . 0 = Nu (3.1)

As can be seen in Figure 3.11, Han 1984 results in a lower Nusselt number than the
other correlations. The reason for this is that this correlation concerns the average
Nusselt number for the ribbed and smooth walls in a duct with two ribbed walls. The
other correlations are for the Nusselt number on the ribbed side wall, which is of more
interest for turbine blade design.
1 2 3 4 5 6 7 8
x 10
4
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
Friction factor for a Ribbed Duct
Reynolds Number
F
r
i
c
t
i
o
n

F
a
c
t
o
r
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D
Smooth duct

Figure 3.12 Friction Factor on Ribbed Side Wall for Case A.
The results concerning the friction factor are displayed in Figure 3.12. All correlations
except Q3D result in approximately the same value for the friction factor. All
correlations of Han et al. calculate an average friction factor, which is based on a
weighted average between a totally smooth duct and a totally ribbed duct, see
Appendix A.

Figure 3.13 and Figure 3.14 presents Nu and K for a smooth wall in a duct with two
ribbed walls.

21
1 2 3 4 5 6 7 8
x 10
4
0
50
100
150
200
250
300
Nusselt Number on Smooth Wall in a Ribbed Duct
Reynolds Number
N
u
s
s
e
l
t

N
u
m
b
e
r
Han 1988
Han & Park 1988
Han et al. 1985
Q3D
Smooth Channel

Figure 3.13 Nu on Smooth Side Wall in Ribbed Duct for Case A.
1 2 3 4 5 6 7 8
x 10
4
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
Nusselt Number Enhancement K on Smooth Wall in a Ribbed Duct
Reynolds Number
H
e
a
t

T
r
a
n
s
f
e
r

E
n
h
a
n
c
e
m
e
n
t

F
a
c
t
o
r
Han 1988
Han & Park 1988
Han et al. 1985
Q3D

Figure 3.14 K on Smooth Side Wall in Ribbed Duct for Case A.

22
Figure 3.14 shows that there is quite a large difference on the enhancement factor
between the correlations. Results for K, C and eff of a ribbed side wall are presented
in Appendix E.

The effect of , P/e, e/D
h
and W/H on Nusselt number enhancement, K, friction
factor enhancement, C and efficiency, eff for Case A has been studied. The results are
presented in Appendices C to G.

U-shaped Channels
The friction factors and enhancement factors in U-shaped ducts cooled by SR- and
SSR-schemes are displayed in Figure 3.15 for the geometry according in Table 3.2.

Table 3.2 Geometry for SR- and SSR- Calculations.
e D
h
R P T
f
T
w
0.3 2.5 1.08 3.3 45 800 400
0 5 10 15
x 10
4
0
5000
10000
15000
Nusselt Number
Re
N
u
0 5 10 15
x 10
4
0
0.05
0.1
0.15
0.2
Friction Factor
Re
f
0 5 10 15
x 10
4
1
1.5
2
2.5
Nusselt Number Enhancement Factor K
Re
K
0 5 10 15
x 10
4
2
2.5
3
3.5
Friction Enhancement Factor C
Re
C
SR
SSR
SR
SSR
SR
SSR
SR
SSR

Figure 3.15 Results for SR- and SSR-Scheme Calculations.

Temperature Analysis
A temperature analysis was made in order to investigate the effect on surface
temperature for a change of K. Vane 2 on SGT -700 was chosen since the ducts
geometries are easily approximated to fit the correlations. A cross-section in the
middle of the vane, named section 4 in Q3D, was used, see Figure 3.16.

23

Figure 3.16 Cross-Section 4, Vane 2, SGT -700.
The K values in channels 1, 2 and 3 were altered, see Figure 3.16. The temperatures
are calculated for the 2D-section, and are shown as the temperatures of the external
surface. Temperature is calculated at points along the perimeter. The distance along
the external perimeter, S, starts with 0 at the suction side trailing edge and is defined
as S
max
at the pressure side trailing edge. The points along the perimeter are made
dimensionless by dividing S by S
max
.

For channel 1, only the Nagoga correlation was applicable, for channel 2, only Han &
Park 1988 was valid, and for channel 3 only Han et al. 1989 was valid. Table 2
display the geometries and both original and new K values. The dimensions of the
channel are taken from the CAD program Catia. Channels 2 and 3 were approximated
to have a rectangular cross-section. The Reynolds number and enhancement factors
were taken from a Q3D model of the vane. The formula used for the calculations of
the original K values is not known. Index s indicates value for smooth wall and index
r value for ribbed wall.
Table 3.3 Parameters for Channels in Vane 2, SGT -700.
Channel 1 Channel 2 Channel 3
Correlation Nagoga [3.4] Han & Park [3.9] Han et al. [3.8]
W - 6.33 3.89
H - 5.39 4.63
e 0.3 0.3 0.3
P 3.3 3.3 3.3
D
h
2.52 5.82 4.23
45 45 45
R 1.08 - -
W/H - 1.17 0.84
P/e 11 11 11
e/Dh 0.12 0.052 0.071
Re 68 591 58 025 48 881
K
r,orig
2.78 2.68 2.750
K
r,new
2.272 2.157 2.634
K
s,orig
1.670 1.610 1.650
K
s,new
1.670 1.463 1.430
K
r,new
/K
r,orig
0.817 0.805 0.958
K
s,new /
K
s,orig
1 0.909 0.867
24
The Nagoga correlations does not consider the smooth side wall, therefore the original
value was kept for channel 1 for the smooth side wall.

Figure 3.17 and Figure 3.18 display the results of the temperature calculations.
SGT 700 Vane 2 Section 4
800
810
820
830
840
850
860
870
880
890
900
0 0,2 0,4 0,6 0,8 1 1,2
Position S/S_max
T
e
m
p
e
r
a
t
u
r
e

[
K
]
Original SGT700
New SGT700

Figure 3.17 Temperature in Vane 2, SGT -700.

Temperature Difference, New-Original
-15
-10
-5
0
5
10
15
20
25
0 0,2 0,4 0,6 0,8 1
Position, S/Smax
T
e
m
p
.

D
i
f
f
e
r
e
n
c
e

[
K
]
Temp. differences

Figure 3.18 Temperature Difference in Vane 2, SGT -700.


25
3.2.2 Analysis and Discussion of Correlations from Articles
Correlations from eight articles have been investigated, in order to find which ones
are most reliable to use in turbine blade and vane cooling channels with ribs. Figure
3.10 gives an overview of the different ranges of the correlations together with the
ranges of interest for turbine design. Unfortunately, no correlation covers the total
area of interest. Therefore different correlations could be of use for different cooling
channels.

The friction factors in the correlations are the average friction factors, weighed
between f of a fully smooth duct and that of a fully ribbed duct. The friction factor on
a ribbed side wall is therefore higher than the values displayed in Figure 3.12.

Webb et al.s correlation from 1970 covers a lot of interesting ranges. However, the
experiments it is based on were performed on fully ribbed circular pipes, instead of
rectangular or square ducts with only two ribbed sides. This makes the correlation
unsuitable for turbine blade design.

The correlations of Han 1988 cover a wide range of parameters. J.C. Han has made a
large amount of research in this area, and when it comes to correlations of this kind, it
is difficult to find articles that he has not been involved in. His results from 1987,
1988 and 1989 are similar but valid for different duct- and rib geometries. The main
difference is that Han 1988 only is valid for transverse ribs and the other two for both
transverse and angled ribs, see Figure 4. The correlations from 1987 are valid for
aspect ratios between 1 and 4, and those from 1989 are valid for smaller aspect ratios,
between and 1. Experiments from these articles are comparable to each other and
all aim to gather information to use for turbine blade cooling.

Figure 3.11 shows that the Han 1988 correlations give a slightly lower value of Nu
compared with Han & Park 1988 and Han et al. 1989. Since it is safer to choose a too
low Nu than a too high, it would be wiser to use the correlation from 1988 for
transverse ribs.

Figure 11 in Appendix D shows that for W/H = 1, where both the correlations from
1987 and 1989 are valid, the later gives a higher value of the Nusselt number.
Therefore it would be safer to use the 1987 correlation for an aspect ratio of one.

His earlier work, in 1984 and 1985, also aims for turbine blade design. However, the
correlations from 1984 only concerns average values for the Stanton number and
friction factor, instead of the actual St and f of a ribbed or smooth wall which are used
in heat transfer programs. It is therefore not suitable for turbine blade design.

The correlation from 1985 concerns angled ribs, but is valid for a smaller range for
e/D
h
and W/H, than the correlations from 1987 and 1989. The values of St and f of
Han 1985 differs from those of Han et al. 1989 and 1987, Figure 3.11 and Figure 3.12.
The difference between the correlations is that in Han 1985, calculations are made
with the average roughness Reynolds number,
+
e and the average roughness
function, ( )
+
e R and in the other Han correlations calculations are based on the
roughness Reynolds number
+
e and roughness function R(e
+
). Definitions of these
26
parameters is found in Appendix A. Since the most recent and the majority of articles
use the later way of calculations, that method it likely to be the most reliable.

The article of Han, Glicksman and Rohsenow from 1978 covers a relatively small
range of Reynolds number and is valid for very large aspect ratios, see Figure 3.10.
Therefore the correlations are often not applicable for turbine vane and blade design.
The experiments were performed using parallel plate geometry instead of a
rectangular or square duct, which also contributes to the unsuitability of using these
correlations for the turbine components of interest. However, it can be of interest for
other areas, for example combustor cooling.

Chandra, Niland and Han performed experiments to investigate the effect of different
number of ribbed walls in 1997. The correlations are only valid for a small range of
W/H and e/D
h
and may therefore not be suitable to use for duct with two ribbed walls,
where there exist more applicable correlations. However, the correlations can be
useful for ducts with one, three or four ribbed walls.

Effect of Rib Angle
The correlations from articles all display similar trends concerning rib angle,
according to Figure 1 to Figure 3 in Appendix D. The highest Nusselt number appears
for = 60 for all correlations except for Han et al. 1989, which calculates the
highest Nusselt number for = 75 . However, the experiments from 1989 were only
made for = 30, 45, 60 and 90, therefore it is difficult to say how correct the
value is for = 75 .
The highest Nusselt number is unfortunately followed by the highest friction factor, as
seen in Figure 5, Appendix D. Therefore the efficiency, Figure 3 in Appendix D, has a
minimum for around 60 and 75. Even thought Figure 3 in Appendix D show a
minimum of eff for around 60 and 75, rib angles of 60 are often used in turbine
vanes and blades. This is because the pressure loss due to the ribs in cooling ducts in
are relatively small compared to other losses, for example that caused by the channel
bends. Therefore rib angle is often chosen with respect to the highest Nusselt number
instead of efficiency.

Effect of Pitch-to-Rib Height Ratio
Appendix E show the results concerning P/e. These correlations are not valid for P/e
< 10, but for the valid range it is clear that the highest heat transfer coefficient occur
for P/e = 10. The correlation of Han et al. 1978 is valid for a larger span of P/e.
Figure 4, Appendix E indicates that P/e = 10 is the optimal pitch-to-rib height ratio.
However, this correlation is not valid for Case A, so the result should be handled with
caution. The Han correlations agree well with the theory in chapter 3.2.

Effect of Aspect Ratio
The effect of aspect ratio is displayed in Appendix F. The Han correlations from
1987, 1988 and 1989 give similar results, whereas that from 1985 deviate from the
trend. The graphs confirm the theory that smaller aspect ratios results in higher
efficiency than larger aspect ratios. An increased aspect ratio leads to increased K and
C, according to Figures 1 and 2, Appendix F. The increase of C is larger than the
increase of K. Therefore, the highest efficiency is received for small aspect ratios.


27
Effect of e/Dh
Figure 14 to Figure 16 in Appendix G display the effect upon C, K and eff for various
e/Dh. Han 1988 and 1985 returns a lower value than Han et al. 1987 and 1989. Han
1984 returns lower values than the other correlations, which is the same pattern as for
previous results.
The effect on K is relatively low compared with the other dimensionless parameters
investigated. The friction factor is more sensitive to rib height according to Figure 2,
Appendix G. It is logical that increased rib height leads to increased pressure drop.
This agrees well with the theory about the effect of e/Dh. The fact that the friction
factor increases while the Nusselt number is rather unaffected by increased rib height,
makes the highest efficiency for small rib height-to-hydraulic diameter ratios, as seen
in Figure 16, Appendix G.

Temperature Analysis
As seen in Figure 3.17 and Figure 3.18, the changes of K values affect the surface
temperature on the vane. Since the values of K
new
where all lower than the original
values, it should result in higher surface temperatures. Since K
original
is higher than
K
new
, the model was not based on calculations from Q3D, since the Q3D enhancement
factors result in lower values than all Han correlations except that from 1984, see
Figure 3.11. The model of Vane 2, SGT -700 was built in Moscow several year ago,
and it is not known what correlation the enhancement factor is bases on.

The surface temperature became higher with K
new
implemented, except at the trailing
edge. At the trailing edge, the temperature in the original model is approximately
12 C higher than in the modified model. The explanation is that since the K values
are lowered in channels 1 to 3, less heat is transferred to the cooling air which results
in colder cooling air in the trailing edge. Colder trailing edge cooling air results in
colder surface temperature in this region, which is displayed in Figure 3.17 andFigure
3.18.

The largest temperature difference appears next to the leading edge at the pressure
side. This is natural since the K values are lowered most in channel 1 and 2. The new
revised models surface temperature is 20 C higher than the original model.

Since the enhancement factor K has such impact on the surface temperature, and the
difference of the original and new value of K for this example is as much as 18.3 %, it
is recommended to look into the routine of choosing Nusselt number enhancement
factors.

28
3.2.3 Conclusions concerning Correlations from Articles
In conclusion, following recommendations are made for what correlations to use.

Han 1988 [3.5] for transverse ribs in ducts with two opposing ribbed walls
Han et al. 1989 [3.8] for angled ribs in ducts with two opposing ribbed walls
with 1 4 1 < H W .
Han & Park 1988 [3.9] for angled ribs in ducts with two opposing ribbed walls
with 4 1 H W .
Chandra et al. 1997 [3.6] for duct with one, three or four ribbed walls.
Nagoga 2001 [3.4] for U-shaped ducts with SR- or SSR-schemes.

29
3.3 Correlations in Q3D
The results from an investigation of the Q3D program code are presented and
analyzed below.
3.3.1 Results concerning Q3D
The correlation in Q3D is an in-house correlation based on Han test data. Q3D
calculates the Nusselt number and the friction factor. The input variables are duct
height, duct width, rib height and the hydraulic diameter. The Nusselt number is
calculated according to equation (3.2). The Q3D correlations are used to choose A and
n.

n
A Nu Re = (3.2)

The Q3D interface for ribbed channels is showed in Figure 3.19.

Figure 3.19 Q3D Interface for Ribbed Channels.
The path to the Channel box is

H_Hydro Hydraulic Net HN Branches-FE Grid Channels

As seen in Figure 3.19, both the Nusselt number for a ribbed duct, Nu
r
and for a
smooth duct, Nu
s
are calculated. The exact values and expressions used in Q3D were
extracted from the code and are displayed in Appendix H.

Reduced Area Method
The area and hydraulic diameter used in Q3D to calculate the friction factor and
pressure drop are calculated by the so called Reduced Area Method - R.A.M. This
means that the area and hydraulic diameter are calculated by the reduced height H
e
,
instead of the total height H according to equation (3.3) and (3.5) and Figure 3.20.

30
e H H
e
= 2
(3.3)

e
e
e
H W
H W
D
+

=
2

(3.4)

e
H W A =
(3.5)

H W
H W
D
h
+

=
2
(3.6)

H W A = (3.7)

The alternative to the Reduced Area Method is the Total Area Method, when H and
D
h
are used. The Total Area Method is used in most of the open literature on the
subject. To compare Q3D with the other correlations, the code was recalculated to the
Total Area Method.

Whether Nu is calculated with the Reduced or Total Area Method depends on in what
form the hydraulic diameter input value is entered.

Q3D Compared to the Han Correlations
Nu and f calculations of Q3D compared to the different Han correlations are displayed
in Figure 3.11 toFigure 3.14 and in Appendices A to E. The Q3D Nusselt number is
lower than those of the Han correlations. The Q3D friction factor is considerably
higher than those of the other correlations, see Figure 3.12.

The high friction factor is probably due to a bug in the code. The code contains
calculations to adjust f according to the number of ribbed walls, but the calculations
are immediately overwritten. The overwritten equations use three parameters Lch,
LRibs and m_lbr. Lch represents the channel perimeter and LRibs the part of the
perimeter with ribs. The meaning of m_lbr is unknown. These parameters should be
input parameters, but they cant be entered into the program. A bug was found in the
interface. Three boxes were excluded from the interface concerning number of ribbed
walls. The original and the corrected interface are showed in Figure 3.21.
H
e
Figure 3.20 Cross-Section of Channel with 2 Ribbed Walls.

31

Figure 3.21 Original Interface (Left) and Corrected Interface (Right).
This is not the only reason for f being so high, because f is already too high before
this stage of calculation. When reasonalble values are entered in the three boxes for
Lch, LRib and m_lbr, they are immediately overwritten by some default value and are
not used in the calculations.
3.3.2 Analysis and Discussion of the Q3D Correlations
Figure 3.11 and Figure 3.12 show that the Nusselt number is lower and f is much
higher in Q3D than the Han correlations. The code is very different from the
equations used in the Han correlations, and is probably curve-fitted models of the
original equations and data. Whether the differences between the original correlations
and the code are deliberate corrections or mistakes is difficult to tell. One reason of
the high friction factor could be that the friction factor from Q3D is the one for a
ribbed side wall instead of an average between a fully ribbed and fully smooth duct.
The low value for the Nusselt number could be chosen due to safety reasons, since it
is better to underestimate the value of Nu than vice versa.
To localize the fault concerning both f and Nu is difficult, since the code is based on
numerical approximations which are hard to connect to a physical duct. However, the
three hidden input parameters is part of the solution. They need to be investigated
further.

The Q3D code also differs from the original correlations since P/e in Q3D affects
neither the value of f or Nu for Q3D, see Figure 39 Figure 40 in Appendix E. The
reason is that the code is based on a constant value of P/e = 10. Figure 14 in Appendix
Lch
LRib m_lbr
32
G shows that for Q3D, Nu is unaffected of e/D
h
, which is not the case for the Han
correlations.
It is also notable that Q3Ds correlation has a different Reynolds dependency than the
other correlations. Again, the reason for this is difficult to detect without more
knowledge about the basis of the correlations.

Q3D is based on the Reduced Area Method. For the Nusselt number, the only
difference from the Total Area Method is that the reduced hydraulic diameter, D
e
is
expected as input instead of D
h
, see equations (3) and (5). For the friction factor, the
difference from the Total Area Method is that H
e
is used instead of H, see equation
(2). Therefore it would be easy to change the code to use the Total Area Method,
which is used in most open literature.
3.3.3 Conclusions Concerning Q3D
Following conclusions were drawn from the investigation of the Q3D code.

Nu in Q3D is lower than Nu from the Han correlations it is based on.
The friction factor in Q3D is considerably higher than the Han correlations,
probably due to a bug that needs further investigation.
Three input parameters concerning number of ribbed walls are missing in the
interface. The parameters are part of the solution to the high friction factor, but
needs further investigations.
It would be easy to change the code from the Reduced Area Method to the
Total Area Method, which is recommended.

33
3.4 Correlations in Multipass
Correlations used in the Multipass program have been investigated. The results,
analysis and conclusions of the investigation are presented below.
3.4.1 Results concerning Multipass
Four subroutines calculate the friction factor and Stanton number, each based on
different correlations. The subroutines have been investigated and compared to the
original correlations. The results are presented graphically in Appendix I. The
comparisons concern the Nusselt number and friction factor for a ribbed wall in a duct
with two ribbed walls. The Nusselt number is obtained from equation (2.3).

Subroutine Ribbed 1
The equations in Ribbed 1 are based on the work of Han, Glicksman and Rohsenow
[3.7]. In Figure and Figure 18 in Appendix I, the Nusselt number and friction factor
from Han et al.s correlation is displayed together with that from Ribbed 1.

The geometry used for comparison is quite different from the other test geometries,
since this correlation has anogher validity range, see Figure 3.10.

Subroutine Ribbed 2
Calculations in Ribbed 2 are based on Han and Park [3.9]. A bug in the code was
discovered. Figure 19 and Figure 20 compare the original correlation with the Nu and
f used in ribbed 2, both with the bug corrected and not corrected.

Subroutine Ribbed 3
The code is based on correlations from work by Han, Ou, Park and Lei [3.8]. A bug
was detected for the calculation of the friction factor. Figure 21 and Figure 22 display
the Nusselt number and friction factor concerning Ribbed 3, both with and without the
bug corrected.

Subroutine Ribbed 4
This subroutine is based on a mixture of Ribbed 2 and Ribbed 3, i.e. it is a mixture of
correlations from Han & Park 1988 and Han et al. 1989. A bug in the Stanton number
due to a mix up of friction factors was detected and corrected. The results are
displayed in Figure 23 and Figure 24.
3.4.2 Analysis and Discussion of the Multipass Correlations
Subroutine Ribbed 1
The Stanton number agrees with the original correlation of [3.7].

The friction factor in Ribbed 1 is first calculated according to the work of [3.7], but
then manipulated numerically. The numerical manipulations results in a friction factor
that differs slightly from the original correlation, see Figure 18. In [3.7], the fanning
friction factor, f is derived. In Ribbed 1 and Ribbed 3, a subroutine called FF is used
for calculations of the friction factor, which calculates the Darcy friction factor, f
D
.
For better comparisons, the friction factors from Multipass are recalculated to the
fanning friction factors here.
34

Subroutine Ribbed 2
The equations in the code are almost identical with the Han and Parks correlations.
There is one difference, where Ribbed2 uses the friction factor of a ribbed wall in a
duct with two ribbed walls, f
r2
instead of duct with four ribbed walls, f
r4
in for
calculations of e
+
. This is probably a bug in the code and is easily corrected.

Subroutine Ribbed 3
The Stanton number is calculated almost exactly as the original correlation, whereas f
is numerically manipulated, which results in a much higher value than that of Han et
al., see Figure 22. Except for the numerical solution of f, there are also a few
differences between the codes calculations and that of Han et al.s correlation. It is
probably a bug and is easily corrected. Figure 21 and Figure 22 display the Nusselt
number and friction factor concerning Ribbed 3, both with and without the bug
corrected.

Subroutine Ribbed 4
One difference between the code and the correlation exists, similar with the one in
Ribbed 2. In one equation, Ribbed 4 uses the friction factor for a ribbed wall in duct
with 2 ribbed walls, f
r2
instead of the friction factor for a fully ribbed duct, f
r4
. This
effects the Stanton number but not the friction factor. Figure 23 and Figure 24 show
the results.
3.4.3 Conclusions concerning Multipass
In Ribbed 1, the Stanton number is the same as in [3.7]. The friction factor is
numerically manipulated and is not equal to the original correlation.

In Ribbed 2, the Stanton number differs from [3.9], due to a mix up of friction factors,
which are used in the Stanton number calculation. The friction factor is equal to [3.9].

In Ribbed 3, the Stanton number differs from the [3.8], due to a small difference from
the base correlation. The friction factor is numerically manipulated and differs from
the [3.8].

In Ribbed 4, the Stanton number also differs from the Han correlation, due to a mix
up of friction factors that are used in the St calculation. The friction factor is equal to
the original correlation.

35
3.5 References Rib Turbulated Cooling
[3.2] von Karman Institute for Fluid Dynamics, Heat Transfer and Cooling in Gas
Turbine, Lecture Series 1995-05, May 8-12 1995
[3.3] Nagoga, G. Intensification of the Heat Transfer in the Cooling Ducts of the
Gas Turbine Blade, Finspng HTC Database, Part I, Russia (2000)
[3.4] Nagoga, G. Intensification of Heat Transfer in the Cooling Ducts of Gas
Turbine Blade, Finspng HTC Database, Part IV, Moscow, 2001
[3.5] Han, J. C. Heat Transfer and Friction Characteristics in Rectangular Channels
With Rib Turbulators, Journal of Heat Transfer, 110, (1988), 321-328
[3.6] Chandra, P. R., Niland, M. E. and Han, J. C. Turbulent Flow Heat Transfer
and Friction in a Rectangular Channel With Varying Numbers of Ribbed
Walls, Journal of Turbomachinery, 119, (1997), 374-380
[3.7] Han, J. C., Glicksman, L. R. and Rohsenhow, W. M. An Investigation of Heat
Transfer and Friction for Rib-Roughened Surfaces, International Journal of
Heat Mass Transfer, 21, (1978), 1143-1156
[3.8] Han, J. C., Ou, S., Park, J. S. and Lei, C. K. Augmented Heat Tranfer in
Rectangular Channels of Narrow Aspect Ratios with Rib Turbulators,
International Journal of Heat Mass Transfer, 32, (1989), 1619-1630
[3.9] Han, J. C. and Park, J. S. Developing Heat Transfer in Rectangular Channels
with Rib Turbulators, International Journal of Heat Mass Transfer, 31,
(1988), 183-195
[3.10] Han, J. C., Park, J. C. and Lei, C. K. Heat Transfer Enhancement in Channels
With Turbulence Promoters, Journal of Engineering for Gas Turbines and
Power, 107, (1985), 628-635
[3.11] Larsson, T.(2002) Investigation of the Effect of Different Rib Configurations
on Heat Transfer in the GT10C Blade 1 Leading Edge, Siemens Finspng
Technical Report
[3.12] Han, J. C. Heat Transfer and Friction in Channels With Two Opposite Rib-
Roughened Walls, Journal of Heat Transfer, 106, (1984), 774-781
[3.13] Webb, R. L. Heat Transfer and Friction in Tubes with Repeated-Rib
Roughness, International Journal of Heat Mass Transfer, 14, (1971), 601-617

Matrix Cooling

37
4 Matrix Cooling
The basic concept of matrix cooling is presented, togheterher with a summary of the
literature study. Next, the results, analysis and conclusions concerning matrix cooling
are accounted for.
4.1 Theory
The flow and geometry of matrix cooling is described. After that, a summary of the
literarure study is presented.
4.1.1 What is Matrix Cooling?
A matrix consists of two layers of opposite angled longitudinal ribs. The ribs create a
system of channels, in which the cooling air flows and continuously changes direction
as it changes channel on its way to the matrix exit, see Figure 4.1.

Figure 4.1 Example of a Matrix Geometry [4.1]
The heat transfer coefficient is increased due to inlet effects, where a new thin
boundary layer is developed at the entrance of each channel. When the flow passes
from one channel to another, a swirl is created and the turbulence of the flow is
increased. Heat transfer is also increased due to increased heat transfer surface area
from the longitudinal ribs.

Another positive effect is that the longitudinal ribs enhance the strength of the
component. The angle of the channels has a large influence on the matrix heat
transfer enhancement.[4.2]

The matrix in Figure 4.1 is a so called closed matrix, where the channels reach the
side wall. In a closed matrix, the flow that reaches the end of a channel flows through
the bend and into the channel on the opposite side without mixing with the flow from
the other channels. An alternate design is called open matrix, where there is a
clearance between the channels and the side wall, see 1 and 2 in Figure 4.2 .[4.3]
In an open matrix, the flow that reaches the end of a channel mixes with the flow in
the clearances 1 or 2. The air that flows in to new channels starting at the
clearances, is taken from the flow in the clearance channel.

l

Matrix Cooling
38
Matrix cooling is often used in the trailing edge of turbine vanes and blades. In
Siemens Gas Turbines produced in Finspng, only open matrixes are used. Usually
the flow in open matrixes is axial and that in closed matrixes is radial.
W
1 2
W
1 2

Figure 4.2 Open Matrix.
This cooling method is relatively unknown in west, but has been used for many years
in the former Soviet Union. Therefore available information in open literature is
limited.
4.1.2 Fin Effect of Ribs
Fins, or in this case ribs, increase heat transfer due to increased heat transfer surface
area. However, the ribs are not of uniform temperature, but are cooler at the tip than
close to the hot wall. The effective temperature difference is therefore lower
compared to an ideal rib where the whole rib is at wall temperature. This effect is
described by the rib efficiency - , which is defined as

e Temperatur Wall at was Rib Entire if Rib from Transfer Heat Ideal
Rib the of Transfer Heat Actual
=

The equations for rib efficiency depend on whether the rib top is isolated or
convective, see equations (4.1) to (4.4) and Figure 4.3. [4.4]

Isolated Rip Top:
( )
h M
h M

=
tanh
(4.1)
where
s
M
2
(4.2)
Convective Rip Top:
( )
c
c
h M
h M

=
tanh

(4.3)
where
2
k
h h
c
+ = (4.4)
Matrix Cooling

39

Figure 4.3 Geometry of Longitudinal Rib.
The fin increases the heat transfer due to increased heat transfer surface area. The fin
effectiveness - is defined according to equation (4.5).

= =
b
rib
A
A
Rib No was there if Transfer Heat
Rib with Transfer Heat

(4.5)

The area A
b
is the rib base area, based on length s. A
rib
is the heat transfer surface of
the rib.

4.1.3 Summary of Literature Study
A summary of the investigated literature is presented below.
Bunker, 2004 [4.1]
Tests were made to study the pressure losses and local and average heat transfer
enhancements in matrix cooling channels. Two methods were used. First, acrylic
models were tested and the effect of increased heat transfer surface area was not taken
into account, since the rib material was insulating. Heat transfer on the matrix shell,
corresponding to suction and pressure sides, was investigated. Liquid crystal
technique was used for temperature measurements. Second, metal models were tested
and the effect of increased heat transfer area was included. Temperature was
measured with an infrared camera. The heat transfer in the channels was compared to
that in a smooth duct, calculated with Dittus-Boelter, equation (3.1). Following
conclusions were drawn from the results of the tests.

Average K on duct shell was equal to 2.5. After turns, local K could reach a
value of 3.
Narrow sub channels provide higher overall heat transfer enhancement, with
K 3, than wider sub channels which have less turn effects and K 3 .
The effect of increased heat transfer surface area is of great importance.
It is appropriate to treat the ribs as simple fins, and each rib surface has
approximately the same heat transfer coefficient as the shell of the matrix
channel.
k
h

s
Matrix Cooling
40
Nagoga, 2000 [4.2]
This is the only work found that contains heat transfer and friction correlations for
matrixes. The correlations and their ranges are summarized in Appendix B. To
describe heat transfer and friction enhancement factors, Nagoga compared the
matrixes investigated to a matrix with straight channels, i.e. for which = 0. He did
not think it was reliable to compare the matrixes to the flow in a long smooth duct.
Nagoga found that a matrix increases the heat transfer compared to a geometry with
straight channels, i.e. with = 0.

Heat Transfer on Base Shell
The base shell corresponds to the suction and pressure sides on a turbine blade. The
matrix can be divided into two parts, the initial and basic section. The initial section
consists of the channels that begin at the inlet of the matrix and end at the side bound.
The basic section consists of the channels beginning at a side bound and ending either
at a side bound or at the outlet of the matrix. Channels in the initial section have the
same heat transfer and flow behavior as a smooth duct.

Heat transfer in the basic section was higher than that in a straight duct. The local
enhancement factor varied from 1.28 to 2. For increased Reynolds number and x/d,
the enhancement factor decreased.

Both the local Nusselt number and the Nusselt number averaged on the length of the
channel depend on , and increase with increased up to = 45 . The average
Nusselt number decreases with increased channel length.

The local Nusselt number reaches a maximum right after the turn and then reduces
along the channel. Local Nu has a value similar to that in a smooth duct when
. 30 d x

Nagoga concluded that the rib pitch, relative channel depth, form of channel cross-
section and the type of side bound, concave or flat for example, did not affect local
and average heat transfer in the scope of interest for turbine blades and vanes.

Experiments showed that the effect of the relative depth of the channels must not be
connected with the effect of cross flow interaction with the co-planar sub channels on
opposing sides.

Heat Transfer on Side Bounds
In order to understand the heat transfer on the internal surface on the semi-cylindrical
side bounds, which corresponds to the leading edge on a turbine blade, a number of
new parameters were defined, see Table 4.1.
Table 4.1. Definition of Ddimensionless Paramteres.
D
Describes the effect of the centrifugal force on flow, defined
as ( ) d R D ) ( cot 1
2
+ = .
CD
F
Describes contraction-diffusion of the flow at the turn A in Figure 4.4.
a a F
A CD
= , see Figure 4.4 for a
A
and a.
C
F
Describes the mass exchange of opposite sub channels through the inter-
ribs windows. z t S F
C
= = cos 2 = number of matrix channels
Matrix Cooling

41

Figure 4.4 Geometry in Matrix Side Bounds [4.2].
The investigation concluded that the heat transfer on the side bounds were higher than
for a duct with straight sub channels. The heat transfer in this area depends on the
parameters D ,
CD
F and
C
F described above, which in turn depend on angle .

The local and average Nusselt numbers along a line ST on the inner concave surface
of the side bound shell according to Figure 4.5 were measured and correlations
developed, see equation (109) in Appendix B.

Figure 4.5 Line ST on Matrix Shell inner Concave Surface.
No explicit correlation of the local Nusselt number on the side bound is reported. The
local Nusselt number is also determined by parameters D ,
CD
F and
C
F , and the rate
of dependence is the same as for the average Nusselt number.

Hydraulic Losses
The pressure losses and hydraulic friction in the matrix were studied. The hydraulic
friction in the initial section did not differ from that in a smooth, straight duct with
= 0. In the basic section, the hydraulic friction exceeded that in a duct with
straight sub channels. The enhancement of hydraulic friction was always higher than
the heat transfer enhancements. Equations (94) and (104) in Appendix B display
friction correlations in the initial and basic section, respectively.

An energy efficency was defined according to equation (113), Appendix B. It was
concluded that the energy efficiency of heat transfer intensification for a matrix
depends only on the value 2 . Maximum energy efficiency was achieved for 2 =
90.

Matrix Cooling
42
Study of the Mechanism of Heat Transfer and Friction
Flow visualization showed that in the initial section, the flow was axial, i.e. the full
velocity vector w, coincided with the axis of the sub channels. The heat transfer was
identical to that in a straight smooth duct, and no increase of heat transfer or friction
occurs.

In the basic section, the flow was twisted due to the turn and overflows from one
channel to another at points A in Figure 4.4, why heat transfer and friction exceeded
that of a smooth straight duct.

Three parameters describe the twisted flow, the angle of flow rotation - , the
rotation parameter - and the maximum value of the axial velocity component -
x
w
,
. The angle is defined in equation (4.6) and
x
w
,
is defined in equation (4.7).
The relationship between and is displayed in equation (4.8) according to [4.2].

w
x
w
w
|
|
.
|
\
|
=
arctan
(4.6)
AV
x
x
w
w
w

=
,
,

(4.7)
76 . 0
18 . 1 tan

= (4.8)

The rotation parameter describes the ratio of the angular momentum flow to the
axial momentum flow. Experimental dependencies were developed for and
x
w
,
,
see equations (117) to (122) in Appendix B. Index indicates values near the wall.

The angle of flow rotation - , and the rotation parameter - both have a maximum
close to the side bound and reduces downstream the channel. The increase of heat
transfer and hydraulic friction depend on the flow rotation at the inlet of the channels.

The flow rotation is caused by the contraction and diffusion of flow in the spatial turn
at point A in Figure 4.4. The rotation intensity, described by , and
x
w
,
, depend
on and the distance x/d from the turn to the control section.

The local Nusselt number in the basic section channels turned out to be the same as
for a round straight channel with inlet rotation. Therefore, the same correlations for
the local Nusselt number can be used for these two cases. Rotation at the inlet of a
channel causes the cooling air velocity to rise in the areas close to the walls and the
mass centrifugal forces and secondary vortexes to form. Equations (123) and (124) in
Appendix B display the correlation for the local Nusselt number in the basic section.

The hydraulic friction in the basic section differed from that in a smooth straight
channel with inlet rotation. Therefore, the same friction correlations could not be used
for the two cases. The difference could be explained by the open U-type form of the
cross-section of the matrixs sub channels determining the interaction of the flow in
the opposite coplanar channel. Equations (125) and (126), Appendix B, display the
correlation for the friction factor in the basic section sub channels. However, these
correlations turned out only to be valid for Re > 10
6
. For Re < 10
5
, the hydraulic
Matrix Cooling

43
friction factor in sub channels in the basic section exceeded the friction factor
calculated by equation (125) by 30 % to 60 %.

Effectiveness of the Matrix
Three indexes are important for evaluating the effectiveness of a cooling method
concerning turbine blades and vanes:
the hydrodynamic energy effectiveness -
the relative depth of cooling or cooling effectiveness -
the effect on life of turbine blade or vane -

Hydrodynamic energy effectiveness
The matrix energy effectiveness - is defined in equation (113), Appendix B. It is
only depending on angle .

A number of comparisons of the hydrodynamic energy effectiveness of matrix to
other cooling methods were made. Main emphasis was put on rib turbulated cooling
and pin fin cooling. Pin fin cooling consists of cylindrical pins in a flat channel.

Enhancement factors K and C for a matrix are based on comparisons between a matrix
with angled sub channels and an analogous geometry but with straight sub channels,
i.e = 0 . The other cooling methods calculate K and C through comparisons
between the cooling method and a single smooth duct. Since not the same bases are
used, a correction factor
was calculated and included in the comparisons. Also,

the friction enhancement factor has to be corrected for an accurate comparison to
other methods.

The results showed that the matrix has lower heat transfer enhancement than the
methods mentioned above.

Cooling effectiveness
The cooling effectiveness is defined in equation (4.9).

=
f G
w G
T T
T T

(4.9)

An increased cooling depth without increase of the relative cooling air mass flow is
desired. Comparisons of cooling effectiveness depth with following other convective
cooling methods were made.

1. a single ribbed duct
2. several connected ribbed ducts with film cooling holes
3. closed matrix
4. impingement

The results showed that the matrix method had the largest relative cooling depth. This
is shown in Figure 4.6, where the cooling depth as a function of relative mass flow for
different cooling methods is shown. The numbers in Figure 4.6 correspond to the list
above.
Matrix Cooling
44

Figure 4.6 Comparsion of Cooling Depht [4.2].
The relative mass flow is defined in (4.10).
g
f
G
G
g =
(4.10)

Effect on Life of Turbine Blade or Vane
Since the blade is exposed to tension stresses - due to centrifugal forces, the
hydrodynamic energy effectiveness and relative cooling depth is not sufficient to
express the effectiveness of a convective cooling method. The effect of life of the
blade or vane - must also be investigated. Equation (4.11) defines .
0 , 0 ,
0
1
0
w w w
T
n
T T
i
A
(
(
|
|
.
|
\
|
= =

(4.11)
The effect of tension stresses for the cooling method is calculated according to
equation (4.12).
|
|
.
|
\
|
+
= =
0 0
0
1
S
S
S
S
T
T

(4.12)
The matrix method was compared to following convective cooling methods
concerning blade life.
pin fin turbulators
rectangular ducts with two opposite ribbed walls
dimple methods (spherical hollows)

Following results were obtained. Blade or vane life increased
Matrix Cooling

45
3,7 to 4 times for pin turbulator method
10-12.5 times for rib turbulated cooling
20 times and more for the whirlwind method
42 times for the matrix method

For matrixes, maximum was received for = 30. However, the report does not
cover the effect of material and temperature on the increased blade life due to matrix
cooling.

Experiments were made to compare the matrix method to the impingement cooling
method, with cross cylindrical pin turbulator in the leading edge. It was found that the
matrix method increased the operation duration of the blade in 8 to 9 times until
reaching the same radial strain for the screen insertion method.

Conclusions of Nagoga [4.2]
1. Heat transfer and flow studies of the matrix base shell and side bounds were studied
and correlations for heat transfer and friction were developed.

2. The intensification of heat transfer and friction in matrix channels are caused by the
flow rotation induced by the turn of the flow at the side bound of the matrix. At the
side bound, the cooling air overflows from one channel to the opposite in a spatial
turn, see Figure 4.4. The intensity of friction, heat transfer and rotation is maximal
right after the spatial turn, and then decreases with increased distance from the side
bound. All three factors depend on angle , with maximum for = 45 . The
friction, heat transfer and rotation in a sub channel are not affected by the coplanar
crossing flow of opposite sub channel.

3. For a scope valid for turbine blade, with = 28.6 to 120.3 and Re
d
= 5000 to
20 000, as an example, the heat transfer enhancement K = 2.0-3.1 and friction factor
enhancement C = 1.8-2.1 for a closed matrix, which is better than most known
cooling methods.

4. Following results were obtained for the heat transfer on the concave surface of the
matrix side bounds, corresponding to the leading edge on a turbine blade.

Heat transfer is increased for
o reduced relative width
C
F of the matrix
o increased curvature D
o increased contraction
CD
F of the turn near the side bound.
Heat transfer changes along perimeter of the concave surface and is maximal
right after the middle point of the turn, in the flow direction.
Maximum heat transfer is always received for = 45 .
Heat transfer is higher than that in a smooth straight channel and higher than
in serpentine-shaped channels.
Heat transfer is less than that of impingement on a concave surface

Matrix Cooling
46
5. The hydraulic resistance in the spatial turns at the side bound -
m
is independent
of Re
d
, increases at the reduction of number of sub channels and is not higher than
1.48.

6. It is established that the area of the rib tops can carry load and take static and
cyclical tension loads in full measure, if the angle between the force vector and ribs
does not exceed 30 . A matrix can handle a larger number of thermal cycles until
appearance of cracks in the shells than for example a blade with pin fins in the trailing
edge. This is difficult to prove theoretically, but experience at Siemens in Finspng
has shown that cracks have not been a problem for matrix cooled areas.

7. Comparisons of cooling effectiveness were made between matrices and a number
of other cooling methods used in turbine blades and vanes. The results showed that
matrices had better performance concerning cooling effectiveness than any other
method tested.

8. The effect on the blade and vane life was investigated. Numerical and practical
tests showed that the matrix increased the blade life in high pressure turbines in 40
times, which was 3 to 4 times more than the ability of for example the methods pin
fins and ribbed ducts.

Matrix Cooling

47
Jurchenko & Malkov, 1995 [4.3]
Three matrixes were investigated to gather information about heat transfer and
hydraulic resistance. The geometry of the models is displayed in Figure 4.7.

Figure 4.7 Three Matrixes Investigated by Jurchenko and Malkov [4.3].
Matrix geometries are described in Table 4.2 to
Matrix Cooling
48
Table 4.4. There was no direct channel connection between the inlet and outlet of the
matrixes.

Table 4.2 Rib heights of the Three Models.
model M1 model M2 model M3
h
1
[mm] 1.0 0.7 0.4
h
2
[mm] 1.0 1.3 1.6

Table 4.3 Geometry for test matrix.
W
[mm]
L [mm] H/L 2 [] t [mm] [mm]
60 35 0.58 120 2.0 2.0
The experiments with the three models in Table 4.2 were carried out in two stages,
first with clearances 1 and 2 between the matrix and the side wall, and second,
tests with the same models but without clearances between the matrix and the side
wall, see Figure 4.7.
The values of the through flow section areas and the end clearances are displayed in
Matrix Cooling

49
Table 4.4. In the cases without end clearances, i.e. 1 = 2 =0, the through flow
section area S
*
is equal to 32 mm
2
.

Matrix Cooling
50
Table 4.4 End Clearances and Through Flow Section Areas.
model M1 model M2 model M3
1 [mm] 2.8 1.4 3.0
2 [mm] 1.6 1.4 2.0
S
*
[mm
2
] 40.8 37.6 42.0
The hydraulic characteristic for models with end clearances was investigated. As
expected, the maximum mass flow was obtained in model M1. The mass flow of the
other models were 20 % to 30 % lower than for M1, due to the smaller through flow
area.

Next, the hydraulic characteristic for models without end clearances was studied. Also
here, model M1 had the highest mass flow. The mass flow for M1 was approximately
2 times lower than for the model with end clearance. Corresponding values for M2
and M3 were approximately 2.6 to 2.8.

Zinc tests were performed to investigate the heat transfer. For information about zinc
tests, see Appendix A. The highest zinc crust thickness, i.e. the highest heat flow, was
obtained in the middle of the profile on model M1. That was expected since this
model had the maximum mass flow.

The test showed that the heat transfer was lower at the inlet of the matrix, due to that
the flow in the initial channels has not yet been exposed to turning movements at the
matrix side bounds. The heat flux also reduces from the inlet to the outlet of the
matrix because the cooling air is heated.

The enhancement of heat transfer, here labeled

K was expressed through the
enhancement of the Nusselt number and the effect of increased heat transfer area K
A
,
see equations (4.13) to (4.15).

A
K Nu
Nu
K
0

(4.13)

where

( )
55 . 0 4 . 0 8 . 0
0
Pr Re 023 . 0

=
f w
T T Nu (4.14)

f
d
Nu

=
(4.15)

Figure 4.8 shows the results of the heat transfer enhancement factor for the three
models and a fourth model with similar angle as the other models but with aspect
ratio W/H = 3.33. The rib tip contact plane displacement turned out not to affect the
heat flow significantly.

Matrix Cooling

51

Figure 4.8 Nusselt Number Enhancement for the Tests with End Clearances [4.3].
For Reynolds number larger than 6000, the difference of
K between the models does

not exceed 7-10 %. For Re < 6000, model M2 had the highest heat transfer
enhancement.

For the tests with no end clearance, the highest values of heat transfer were observed
for model M1.

The results can probably not be used for matrixes with 2 < 120 and with direct
channels between inlet and outlet of the matrix.

It was found that for the same available pressure drop, heat transfer was higher for the
matrixes with end clearances, i.e. open matrixes, than without. However, heat transfer
intensification was found to be connected to an increase of cooling air mass flow.

For the matrixes without end clearances, heat transfer increased for constant mass
flow, but with higher hydraulic resistance compared to matrixes with end clearances.

This concludes that for a case where maximum heat transfer is wanted for specified
pressure loss, matrixes without channels that reach from matrix inlet to matrix outlet
without reaching an end wall, must have end clearances. A positive side effect of
matrixes with end clearances is increased blade strength. [4.3]
Matrix Cooling
52
Filipov and Bregman, 2005 [4.6]
Recommendations about matrix cooling are presented. Matrix should only be used for
airfoils with height-to-chord ratio 2, see Figure 4.9.

Figure 4.9 Height and Chord of an Airfoil.
For larger height-to-chord ratios, the inlet-outlet-area ratio of the matrix will be too
small. This ratio should not be less than 2. Advantages of matrix cooling are

increased blade strength
effective blocking of internal cavities
high air velocity, even for cases with low cavity blocking

Draw backs with matrix cooling are

lower average heat transfer enhancement factors, K1.5-1.7, than pin fins
(K2) and ribbed ducts (K2-2.5)
fin effectiveness low near trailing edge outlet due to small slot thickness
height
chord
Matrix Cooling

53
4.2 Correlations from Literature
Correlations for matrices were investigated and compared to correlations for smooth
channels. The results, analysis and conclusions of the investigation are presented in
this chapter.
4.2.1 Results concerning Correlations from Literature
Nagoga [4.2] is the only literature that presents matrix correlations, except for the
shorter summary of [4.2] presented by Shukin and Nagoga [4.5]. However, several
correlations for smooth ducts with inlet effect are available and can be used for
comparisons. More articles about matrix cooling are listed in Appendix E.

According to [4.2], the matrix can be divided into two sections. First, the initial
section, which are the channels beginning at the matrix inlet and ending at the side
wall. The flow in these channels is the same as in a smooth channel [4.2]. Second, the
basic section, which are the channels beginning at the side wall and ending at a side
wall or at the matrix outlet.

The heat transfer enhancement factors on a blade in a Siemens Gas Turbine that is
cooled by two matrixes are today based on the results of crystal and zinc tests. The
crystal test resulted in knowledge of gas temperature and the temperature profile on
the blade. To measure the gas temperature, thermal crystals are glued at the end of
small sticks situated on the blade. For measurements of the blade temperature profile,
small holes are EDMed and the thermo crystals are glued in the holes. The thermal
crystals measurements result in knowledge of the highest temperature that the crystal
has been exposed to during the tests.

It turned out that the blade temperature profile calculated in Q3D differed from the
measured temperatures. Among other things, the difference depended on an error in
the K value in Q3D. After the crystal test, a zink test was performed, which resulted in
that more correct enhancement factors were obtained. For information about zink
tests, see Appendix A. The enhancement factors in the channels of the large matrix
were calculated to 1.3.

The correlations from [4.2] were applied to two test channels in a typical cooling
system. Since these correlations were developed for a closed matrix, the results of the
test example should be handled with care.

In the correlations, the relevant input parameters are the channel length, the local
coordinate, the Reynolds number, the thermal conductivity of air and the channel
hydraulic diameter. The matrix in the example contains channels that reaches from the
matrix inlet direct to the matrix outlet. They are here called the direct section. Nagoga
[4.2] did not develop correlations for the direct section, but the flow is assumed to be
similar to that in a smooth duct. Input for each test channel is displayed in Table 4.5

Matrix Cooling
54
Table 4.5. Input parameter for Test Channels.
Initial Section Basic Section
l [mm] 18.7 31.2
x [mm] 1 1
Re
d
[-] 11000 6029
[W/m/K)] 0.0623 0.0658
d [mm] 1.4 1.2
T
w
[K] 1008 1093
T
f
[K]

783 847
The correlations of [4.2] and the results of the calculations in the example are
presented below.

Initial Section
Equations (90) to (131) in Appendix B describe correlations from [4.2] for the initial
section. The correlations from [4.2] were compared to other correlations for smooth
ducts, see below. The correlation named Q3D, equation (4.25), is used in Q3D as a
default value for Nu in a smooth duct.

Kays and London [4.7]:
( )
575 . 0 3 . 0 8 . 0
Pr Re 02 . 0
w f
T T Nu = (4.16)

Larsson [4.7]:
80 . 0
Re 0232 . 0 = Nu (4.17)

Cohen et al. [4.8]:
( ) ( )
1 . 0 55 . 0 4 . 0 8 . 0
Pr Re 034 . 0

= d x T T Nu
w f
(4.18)

ISBN5-21702338-4 [4.7]:
( )
(
(
|
|
.
|
\
|
|
.
|
\
|
+ =

7 . 0
7 . 0
7 . 0 4 . 0 8 . 0
1 Pr Re 024 . 0
w
f
w f
T
T
d
x
T T Nu
(4.19)

Dittus-Boelter [4.4]:
4 . 0 8 . 0
Pr Re 023 . 0 = Nu (4.20)

Sieder and Tate [4.4]:
( )
14 . 0 3 1 8 . 0
Pr Re 027 . 0
f w d
Nu = (4.21)

Gnieliski [4.4]:
( )
4 . 0 8 . 0
Pr 100 Re 0214 . 0 = Nu (4.22)

Nusselt [4.4]:
( )
055 . 0 3 1 8 . 0
Pr Re 036 . 0 l d Nu
d
= (4.23)

Petukhov [4.9]:
( )
( ) ( ) 1 Pr 8 7 . 12 07 . 1
8 Pr Re
3 2 5 . 0
+

=
f
f
Nu
(4.24)
Matrix Cooling

55
where
( ) ( )
2
64 . 1 Re ln 790 . 0

= f

QD3:
8 . 0
Re 018 . 0 = Nu (4.25)

Figure 4.10 and Figure 4.11 display the results of the comparison. For a better
comparison of different Nu definitions, the heat transfer coefficient is displayed.
Equation (91) is labeled Nagoga Nu
l
and equation (92) Nagoga Nu
d
since they are
both correlations for average Nu but based on different Nu definitions.
0.5 1 1.5 2 2.5 3 3.5 4
x 10
4
0
1000
2000
3000
4000
5000
6000
7000
8000
Re
d
A
l
f
a
Alfa in Initial Section
Nagoga, local
Nagoga, Nu
l
Nagoga, Nu
d
Larsson 2002
Kays & London 1983
Cohen et. al 1996
ISBN5-21702338-4
Dittus-Boelter
Gnielinski
Nusselt
Sieder and Tate
Petukhov
Q3D

Figure 4.10 Alfa as a Function of Re
d
in Initial Section, Re
d
= 15 000.

Figure 4.10 display the heat transfer coefficients from each correlation in the
following order with highest first; ISBN5-21702338-4, Nusselt, Cohen et al., Sieder
and Tate, Nagoga local i.e. equation (90), then Larsson and Dittus-Boelter close
togheter, Gnielinski, Q3D, Nagoga Nu
d
, Petukhov, Kays and London and last Nagoga
Nu
l
. Figure 4.10 is plotted for x =1mm, where the inlet effects are significant.
Equation (90) is calculated for K = 1, which is the case for the initial section [4.2].

Matrix Cooling
56
0 5 10 15 20 25
1000
1500
2000
2500
3000
3500
4000
x/d
A
l
f
a
Alfa in Initial Section
Nagoga, local
Nagoga, Nu
l
Nagoga, Nu
d
Larsson 2002
Kays & London 1983
Cohen et. al 1996
ISBN5-21702338-4
Dittus-Boelter
Gnielinski
Nusselt
Sieder and Tate
Petukhov
Q3D

Figure 4.11 Alfa as a Function of x/d in Inital Section, Re
d
= 15 000.

Basic Section
Equations (97) to (108) in Appendix B display the correlations developed by Nagoga
[4.2] for the basic section in a matrix. Shukin and Nagoga [4.5] summarized the work
[4.2]. The correlations presented in [4.5] are rewritten but are basically the same as in
[4.2], see Appendix C. The results of heat transfer and friction calculations are
presented in Appendix D.

The heat transfer coefficients for the initial and basic sections are compared in Figure
4.12. Correlations for a smooth duct are also displayed for comparison. For clarity,
not all correlations for a smooth duct earlier discussed are included.
Matrix Cooling

57
0 5 10 15 20 25
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
x/d
A
l
f
a
Alfa in Basic and Intial Sections, for the Intial Test Geometry
Nagoga,inital local
Nagoga,initial av. Nu
l
Nagoga, initial av. Nu
d
Cohen et. al 1996
ISBN5-21702338-4
Dittus-Boelter
Q3D
Nagoga average, basic
Nagoga local, basic

Figure 4.12 Alfa as a Function of x/d for Intial and Basic Sections, Re= 11000.
Equation (91) is labeled Nagoga Nu
l
and equation (92) Nagoga Nu
d
since they are
both correlations for average Nu but based on different Nu definitions.
Matrix Cooling
58
4.2.2 Analysis of Correlations from Articles
Initial Section
Figure 4.10 and Figure 4.11 show that the correlations give very different results. In
Figure 4.10 it is seen that equation (4.19) results in a much higher Nu than the other
correlations. The explanation is that it considers inlet effect and since Figure 4.10 is
taken at x = 1 mm the inlet effects significant. Figure 4.11 display the dependency of
x/d, and it shows that equation (4.19) is only much higher than the other correlations
near the channel inlet. Below equation (4.19) in Figure 4.10 comes the correlations of
Nusselt and Cohen et al., which also display high heat transfer coefficients. The
reason is that also these correlations consider the inlet effects. The Nusselt number
correlation is only valid at the entrance region. It calculates an average Nu for not
fully developed flow, why it results in higher values than the correlations for fully
developed flow. The Cohen et al. correlation considers inlet effects and is dependent
of x/d. It is seen in Figure 4.11 that it displays lower Nu for larger x/d, i.e. where there
are no inlet effects. Below Nusselt and Cohen et al. are the correlations of Sieder and
Tate, Larsson and Nagogas correlation for local Nu, equation (90) gathered in Figure
4.11. It is seen in Figure 4.10 that equation (90) is not similar to the Larsson and
Sieder and Tate for larger x/d, but is significantly lower. Sieder and Tate and Larson
display relatively high heat transfer coefficients. The reason that Larsson is high is
because the correlation was developed for not fully developed flow. The Sieder and
Tate correlation consider the effect of the temperature difference between fluid near
the hot wall and the cooler fluid in the core flow. Other correlations that also consider
this effect are Kays and London, Cohen et al. and equation (4.19), which also display
high values with exception of Kays and London. The other correlations are not valid
for high temperature differences between fluid and wall.

Dittus-Boelter, Petukov and Gnielinski do not condsider inlet effects or temperature
difference effects, which explain the relatively low values. Nagoga Nu
d
, Petukov and
Nagoga Nu
l
also display relatively low values. Nagoga Nu
d
and Nagoga Nu
l
consider
temperature difference effects but not inlet effects.

The Dittus-Boelter equation is probably developed for small temperature differences,
whereas the Q3D and Nagoga correlations are based on zinc tests and thus probably
valid for higher temperature differences.

Basic Section
Figure 4.12 display the heat transfer coefficient of the basic section compared to the
initial section and other smooth duct correlations as a function of x/d for Re = 15 000.
It is clear that for all correlations that consider inlet effects, in Nagogas case the
correlations for local values, the trends are similar. At the channel entrance, new thin
boundary layers are created which causes turbulent flow and high heat transfer
coefficients. The heat transfer then decreases along the channel length, as x/d
increases. When comparing the local Nagoga correlations for the initial and basic
sections, the basic section has approximately 2.4 times higher heat transfer coefficient
at the channel inlet. Closer to the channel end, for x/d > 15, the basic section has
approximately 1.5 times higher heat transfer coefficient than the smooth section. In a
comparison of the average heat transfer coefficients of Nagoga, it is seen that the
Matrix Cooling

59
basic has approximately 2.5 times higher heat transfer coefficient than the smooth
section.

More plots of correlations of the basic section are displayed in Appendix D.
Correlation of [4.2] and [4.5] differ only due to different round ups. Nagogass
correlations result in slightly lower values, why it is better to use, see Figure 26 to
Figure 29. The local heat transfer enhancement factor for the test channel in Table 4.5
is approximately 2.52 and the average is approximately 1.87 according to Figure 28
for Nagogas correlation. The average value of 1.87 is higher than K = 1.3, which is
used today. However, the value 1.3 was chosen conservatively to avoid an
overestimation of K. Open matrices have lower heat transfer than closed matrixes
[4.3]. According to Figure 31 that display Nagogas correlation, K goes from 2.6 at
the inlet to 1.6 at the end of the channel. The inlet effects are thus significant. Since
Nagogas correlations were developed for closed matrixes, they are expected to result
in higher values than what are used for open matrices.

The friction is increased by a factor of 3.4 according to [4.2] or 3.5 according to [4.5],
see Figure 27. According to Figure 29, the friction factor in the basic section test
channel is approximately 0.135 and in a corresponding smooth channel f would be
approximately 0.04.

Matrix Cooling
60
4.2.3 Conclusions concerning Correlations from Articles
Matrix cooling is a cooling method that is relatively unknown in west. The only
correlations found in the literature study are those form [4.2], which are valid for
closed matrixes. Flow in initial matrix channels can be approximated to flow in
smooth ducts. Several correlations for flow in smooth ducts are available. Since the
channels are relatively short, correlations that consider inlet effects are most suitable.
In Siemens Gas Turbines, open matrix cooling is used. The enhancement factors used
today are estimated based on tests, since there are no correlations available for open
matrix cooling. Further work is needed to develop reliable correlations for open
matrices.

Matrix Cooling

61
4.3 Rib Effectiveness in Q3D
Since a matrix consists of longitudinal ribs, it was of interest to investigate box Rib
Effectiveness in Q3D. It is reached by following pathway.

H_HydroToolbar, UtilitiesRib Effectiveness

The interface with input and output parameters are displayed in Figure 4.13.

Figure 4.13 Rib effectiveness interface in Q3D.
The Q3D code was studied in order to evaluate the rib effectiveness calculations.
Comparisons were made with correlations from open literature.
4.3.1 Results concerning Rib Effectiveness in Q3D
From the input parameters, a rib geometry according to Figure 4.3 is defined. The
channel area and the channel perimeter are calculated by a numerical method, which
is difficult to relate to a physical geometry. As a comment in the code, an easier
calculation of channel area and perimeter is noted. This may be interpreted as a hint of
what area and perimeter the other code refers to. The shape factor -K
f
is defined in the
code is displayed in equation (4.26).
t
L
K
f
1
= (4.26)

K
f
is a measurement of the heat transfer surface enhancement, where L
1
is the length
of the dotted line in Figure 4.14.

Matrix Cooling
62

Figure 4.14 Q3D Rib Geometry.
The rib effectiveness -
D Q3
is defined according to equations (4.27) and (4.28) in
Q3D.

( )
h mu
h mu
D Q

=
tanh
3
(4.27)
s
mu
2
(4.28)
4.3.2 Analysis of Rib Effectiveness in Q3D
Equation (4.27) and (4.28) correspond to the definition of rib efficiency for ribs with
isolated tops, see equations (2.6) and (2.9). In Figure 4.13, the term Rib
Effectiveness has been mixed up with Rib Efficiency. Q3D assumes that the rib
tops are isolated. In a matrix, approximately half the rib top area is in contact with
opposing rib top and can be approximated to be isolated. The other half is exposed to
convective cooling. Therfore, it would be better to implement at formula where X
percentage of the efficiency is calculated for convective rib tops cooling and the rest,
1-X, for isolated rib tops according to equation (4.29).

( ) ( )
h M
h M X
h M
h M X
c
c

=
tanh
100
1 tanh
100

(4.29)
4.3.3 Conclusions concerning Q3D Rib Effectiveness
The implementation in Q3D calculated the rib efficiency for isolated rib tops. Since
approximately half of the rib tops in a matrix is exposed to convective cooling, it
would be better to implement a formula according to equation (4.29), which considers
both the isolated and convective rib tops. The phrase Rib Effectiveness in the Q3D
interface should be replaced by Rib Efficiency.

t
h
s
L
1

Matrix Cooling

63
4.4 References Matrix Cooling
[4.1] Bunker, R. S. Latticework (Vortex) Cooling Effectiveness Part 1: Stationary
Channel Experiments, Proceedings of ASME Turbo Expo 2004, Power for
Land, Sea and Air, June 14-17, 2004, Vienna, Austria
[4.2] Nagoga, G. Intensification of the Heat Transfer in the Cooling Ducts of the
Gas Turbine Blade, Finspng HTC Database, Folder No 1, Alstom,Russia,
(2000)
[4.3] Jurchenko, V. and Malkov, V. Technical Report on Results of the Work:
Experimental Investigation of Hydraulic Resistance and Heat Exchange in
Vortex Matrices, Formed by Parallel Ribs and Channels with Assymetric
Location of Ribs Tips Contact:,(1995)
[4.4] Holman, J. P. (2002). Heat Transfer Ninth Edition, McGraw-Hill Companies,
New York (ISBN 0-07-112230-3)
[4.5] Shukin, S. and Nagoga, G. Comparative Efficiency Analysis of the Known
Schemes of Heat Transfer Intensification in the Cooling Channels of Gas
Turbine Blades, The Program of Further Development and Proposal on
Rational Cooling Scheme of Modern Gas Turbine Blades, Alsom, (2001)
[4.6] Filipov, V. and Bregman, V. SGT -800 Gas Turbine Criteria of the Matrix
Cooling Application, Ref.: 1102004/TR045 Rev. 1, Moscow 2005/09/30
[4.7] Larsson, T. Investigation of the Effect of Different Rib Configurations on Heat
Tranfer in the GT10C Blade 1 Leading Edge, RTT10_10/02 (2002)
[4.8] Cohen, H., Rogers, G. F. C. and Saravanamuttoo (1987). Gas Turbine Theory,
3
rd
Edition, Longman Scientific & Technical, Harlow (ISBN 0-582-30539-X)
[4.9] engel, Y. A. and Turner, R. H. (2001). Fundamentals of Thermal-Fluid
Sciences, McGraw-Hill Companies, New York (ISBN 0-07-118152-0)
Impingement Cooling

65
5 Impingement Cooling
The concepts of impingement cooling and a summary of the literature study are
presented below. The chapter also contains the results, analysis and conclusions of the
investigation of correlations and the Q3D code.
5.1 Theory
Impingement cooling is described below, and a summary of the literature study is
presented.
5.1.1 What is Impingement Cooling?
Figure 5.1 gives an example of impingement cooling. High pressure air flows through
holes in a perforated plate. When the jets formed by the holes hit the surface, the
surface is cooled.

Figure 5.1 Impingement Cooling Setup.

The cooled surface is called target plate and the perforated plate is called jet plate.
Different impingement cooling systems exist, for example with variations in the
number of jets, target plate configurations, jet hole configuration and jet angle of
attack. Figure 5.1 display an inline array of jets impinging on a flat surface. Figure 5.2
gives example of impingement on a curved target plate, which for example is the case
for cooling of the leading edge on a vane or blade.

Figure 5.2 Curved Target Plate[5.1].
Inline pattern

Staggered pattern
Cross Flow
Impingement Cooling
66
Where the jets hit the target plate, the flow is highly turbulent and the boundary layer
very thin, which leads to a very high heat transfer coefficient. Impingement is mainly
used where there is need for very high heat transfer coefficients. Best effect is
achieved for geometries with plenty of mass flow available but stricter limitations of
allowable pressure drop. The impingement holes often situated on inserts which are
not load-bearing, see Figure 5.3. It is also used on thicker parts of components such as
the leading edge and mid chord regions. Platforms are also often cooled by
impingement.

Figure 5.3 Impingement Cooling in Turbine Vane [5.2].
The impingement geometry is described by the hole diameter, the streamwise and
spanwise jet-to-jet distances and the distance between jet orifice and target plate,
according to Figure 5.1.

For a curved target plate, such as the leading edge, the target plate curvature also has
to be considered. For heat transfer calculations, these parameters are often made
dimensionless by division of the hole diameter in order to generalize the correlations.

The aerodynamics of a single impinging jet is described in Figure 5.4. The flow is
divided into three regions, the free jet region, the stagnation region and the wall jet
region.

Figure 5.4 Impinging Jet (Left, [5.3]) and Free Jet (Rigth) [5.4].
Before striking the target plate, the impingement jet acts as a free jet. The flow of a
free jet is divided into the potential core zone, the developing zone and the fully
developed zone, see Figure 5.4. When the free jet leaves the impingement hole, the
Impingement Cooling

67
outer part of the jet is mixed with the surrounding air and causes it to mix with the jet.
This increases the jet diameter and the turbulence of the flow. The part of the jet that
is not affected by the surrounding air is called the potential core. The velocity in the
potential core is constant and equal to the exit velocity of the jet. The influence of the
surrounding air increases as the distance from the impingement hole increases, and
finally the potential core is non-existent.[5.5]

In the developing zone, the axial velocity profile changes toward the fully developed
profile which is achieved in the fully developed zone.

The free jet region is followed by the stagnation region, which is located where the jet
impinges on the target plate. It is surrounded by developing boundary layers. After
impingement, the spent jet causes a highly turbulent flow which increases the heat
transfer. In theory, no heat transfer can occur in the stagnation point as the velocity is
zero, but in reality the stagnation point is very unstable and moves all the time, so the
heat transfer coefficient in the stagnation region is very high due to the thin laminar
boundary layer [5.6]. The heat transfer coefficient decreases with increased distance
from the stagnation point. However, when the wall jet changes from laminar to
turbulent flow, a second heat transfer maxima can occur for certain Reynolds number
and z/d distances.

In the wall jet region, the flow is parallel to the impingement plate, and its velocity
profile is described in Figure 5.4. The wall jet increases in thickness due to a build up
of the boundary layer.[5.2]

The impingement cooling system has two chambers, the jet creating or plenum
chamber and the target chamber. The jet creating chamber has higher pressure than
the target chamber.

In turbine cooling, multiple jets are often used, either as a row of jets or as arrays of
jets. This complicates the flow structure. Nearby jets affect each other, and the
upstream jets creates a cross flow, which flow perpendicular to the jet flow along the
target plate, see Figure 5.1. [5.6]
Impingement Cooling
68

5.1.2 Summary of Literature Survey
Flat Target Plate
Reynolds number -Re:
The discharge coefficienct C
d
is vertically independent of Re
d
[5.7]. A weak
dependence of Re exist depending of whether the hole has sharp edges or not [5.8].
Reynolds number based on the jet hole diameter gives a sufficiently good heat transfer
correlation [5.9], and is what is used by most investigators. All articles investigated
showed that the heat transfer increases with Re. According to Hglund [5.8], this
trend is stronger for higher Re. Most investigators found that the heat transfer
dependence of Re can be approximated by a power law [5.10], [5.11] and[5.7] to
mention a few, but Kercher and Tabakoff, [5.12] disagree. They claim that heat
transfer for a square inline array cannot be correlated by power functions equations of
dimensionless parameters, and that the Reynolds number exponent is strongly
dependent of x
n
/d. The heat transfer behavior changes distinctly at Reynolds number
near 3000 [5.12].

Crossflow ratio-G
c
/G
j
:
The cross flow decreases the impact from jets on the target plate surface and should
be minimized to achieve maximum heat transfer [5.8], [5.10], [5.12], [5.13]. The
Nusselt number of the last impingement row in an array is not significantly affected
by the cross flow according to [5.13] and [5.7]. [5.13] investigated arrays with 6 to 12
rows and [5.7] studied arrays with 10 rows. For an inline, dense array of jets, Baileys
and Bunker [5.14] found that G
c
/G
j
increased linerarly up to approximately row 6, see
Figure 5.5.

Figure 5.5 Re and G
c
/G
j
as a Function of Row Number for x
n
/d=y
n
/d =3, z/d=2.75 and Re
d
=29900
and 57800 [5.14].

Between row 6 and 8, G
c
/G
j
peaked due to a flow adjustment from impingement
dominating flow to channel like flow, and thereby much increased convective cooling.
A probable reason for the flow adjustment is the inline hole pattern. At the cross flow
peak, Re decreases. After row 8, G
c
/G
j
increases again and the Reynolds number
increases and peaks at the last row. This resulted in the highest Nu at the exit row and
a Nu minimum in the middle rows. The tendency was most obvious for higher
Reynolds number. A similar trend is reported by [5.8], for a staggered array. He found
that for low G
c
/G
j
, heat transfer monotonically decreases for increased crossflow.
Still, if the crossflow velocity is high, heat transfer will increase due to convection
and Nu will rise at the end of the target plate, creating a heat transfer minima in the
midsection of the array.
Impingement Cooling

69
Cross flow changes the impingement location of the jets downstream, because it
deflects the jets, and thus increase Nu downstream of the impingement location,
[5.8],[5.10] and decreases the heat transfer upstream [5.10].

Florshuetz et al. [5.7] found that for small x
n
/d and y
n
/d and large z/d, Nu/Nu
1

decreases for increased G
c
/G
j
. However, for larger hole spacing and smaller z/d,
Nu/Nu
1
increases slowly due to convective cooling effects of the crossflow, which
does not interfere much with the jets for this kind of geometry. [5.7] also concluded
that jets at the final rows are not much affected by the crossflow. Three complexities
of the crossflow phenomena are identified in [5.7]. First that G
c
/G
j
used in
calculations is average over the channel, instead of being lowest at the first row and
then increasing downstream. Second, the correlations are valid for different x
n
/d and
y
n
/d, with aspect ratios from 0.625 to 3.75. Geometries with small aspect ratios are not
affected by z/d but geometries with large aspect ratios are much affected by z/d. Third,
some jets impinge the target surface before they are developed, and some after. [5.13]
investigated the effect of minimum, intermediate and maximum crossflow. They
found that for a given massflow, the more densely placed holes are more sensitive to
the effect of decreased heat transfer with increased crossflow. The tendency is more
obvious for larger z/d, since it is easier for the cross flow to deflect a long and narrow
jet.

Jet-to-Impingement Target Spacing - z/d
[5.7] found that heat transfer was not much dependent on z/d, except for the highest
hole density tested, x
n
/d = 5 and y
n
/d = 4, where the average heat transfer decreased
with increased z/d for a given cross flow ratio. The trend of z/d was stronger for
higher x
n
/d and y
n
/d. The z/d range was 1 to 3 for the tests. The same trend was
reported by [5.9], which tested z/d from 2 to 8. [5.15] found that the average Nu
increases very slightly with increased z/d up to 4, where it reaches a maximum. For
z/d > 4,the average Nu decreases with increased z/d. [5.11] found by analytical
calculations that the optimal jet-to-target plate spacing is 5.43. [5.12] found that
without crossflow, heat transfer increases with increased z/d but with crossflow, heat
transfer decreases with increased z/d. [5.8] reported that for small z/d, the average Nu
is not much affected. The local Nu decreases at smaller z/d, but a second peak arises
which makes the average Nu approximately the same. For large z/d, heat transfer
decreases as z/d increases according to [5.8].

Jet-to-Jet Spacing x
n
/d and y
n
/d:
[5.9] reported that very large jet-to-jet spacings, i.e. small open area -A
f
, results in a
small average heat transfer. Up to A
f
= 0.012 heat transfer increases rapidly with A
f
.
For A
f
> 0.012, heat transfer still increases, but not as fast. Chance concludes that A
f
should be at least 0.012 to 0.015. The investigations of [5.13], [5.15], [5.12], [5.16]
and [5.8] also concluded that heat transfer increases with smaller jet-to-jet spacings.
[5.11] reports a calculated optimum open area of 0.385. [5.8] reported that closely
spaced jets results in that more of the cross flow channel is blocked by jets, which
intensifies the interaction between jets and cross flow.

Hole Pattern:
[5.7] found that for large hole spacing, there is not much difference in heat transfer
whether the array has inline or staggered hole pattern. For a decreased hole spacing
and increased z/d, the staggered array showed decreased heat transfer coefficient
Impingement Cooling
70
compared to the inline array. The reason was differences in the spanwise distribution
of cross flow between the inline and staggered array. For the inline array, the cross
flow becomes channeled between the rows of jets, which reduces the negative cross
flow effect.

[5.9] investigated the effect of hole pattern, and found that the difference in heat
transfer between staggered and inline arrays was almost insignificant.

Curved Target Plate
On the stagnation line, Nu decreases with increased z/d, with stronger trend for
smaller y
n
/d. For the average Nu, the same trend was observed, but with less z/d
dependency. Nu
x
/Nu
x=0
depends mostly on x
n
/d and showed only a weak dependency
of z/d, especially for z/d >>1. x indicates the distance from the stagnation line. [5.1]

For z/d = 2-8 correlations for average Nu should not display a dependency of z/d. A
second heat transfer maxima was observed at 45 angular distance from the
stagnation line. This maximum is strongly dependent on Re and occurs for relatively
small z/d. At the stagnation point, the heat transfer and flow for a curved target plate
is similar to that for a flat target plate. [5.17]

Tests showed that z/d = 1 resulted in a heat transfer no less than 95 % of the
maximum heat transfer obtained for other z/d tested. It was also found that a line of
circular holes resulted in better heat transfer performance than slot jets. Smaller value
of l/b led to increased average heat transfer. [5.18]

Tests were done to investigate the effect of an elongated leading edge. For a
semicircular leading edge, maximum heat transfer was obtained for z/d = 1 for
circular jets. For a given mass flow, the average heat transfer increased monotonically
as l/b increased. It was also found that slot jets and circular jets behaved similarly.
[5.19]

An investigation showed that the average Nu is proportional to Re
0.75
and inversely
proportional to z/d with exponent 0.25. The curvature of the target plate effects the
heat transfer for the hole curvature surface, i.e. 2 .[5.20]
Impingement Cooling

71
5.2 Flat Target Plate
The results, analysis and conclusions of the investiation of impingement on a flat
target plate are presented in this chapter.
5.2.1 Results concerning Flat Target Plate
Following correlations were investigated.

Florschuetz, Truman and Metzger 1981 [5.7]
Metzger and Korstad 1972 [5.10]
Bailey and Bunker 2002 [5.14]
Martin 1977 [5.11]
Florschuetz and Isoda 1982 [5.21]
Obot and Trabold 1987 [5.13]
Behbahani and Goldstein 1983 [5.15]
Hglund 1999 [5.8]
Woldersdorf [5.23]
Chance 1974 [5.9]
Goldstein and Seol 1990 [5.16]

The correlations and their ranges are summarized in Appendix A. The parameter Nu
represents the average Nusselt number over an impingement row. More articles about
impingement cooling are listed in Appendix J. The article of Woldersdorf has not
been found, because it is an intern ABB report. All articles contained correlations for
the Nusselt number, except for that of Florschuetz and Isoda that contained
correlations for the friction factor and discharge coefficient. Hglund also developed
correlations for the friction factor and discharge coefficient.

A test case called Case B, that covers most ranges of the correlations, was used to
compare the correlations for a flat target plate, see Table 5.1. Inline and staggered
array correlations are compared separately. Nusselt number, discharge coefficient and
friction factor were calculated for the fifth row in an array. The cross flow ratio could
be estimated to 0.2 for this row [5.8].

Table 5.1 Case B, Geometry used for Flat Target Plate Comparisons.
z/d [-] 2
y
n
/d [-] 4.5
x
n
/d [-] 5
N
p
5
n
c
10
T
w
[K]

1073
T
j
[K]

700
T
c
[K]

T
j
C
d
[-]

0.79
G
c
/G
j
[-]

0.2
L [mm] 0.4
Results for Case B are presented for inline arrays and staggered arrays in Appendix B.
Impingement Cooling
72
Calculations were also performed on the airfoil of a typical vane 1 cooling scheme, a
single row upstream and two rows with inline pattern downstream, both on the
pressure side. Results from the calculations are presented in Appendix C. Three tests
with random geometry were also performed in order to investigate the correlations
behavior further, see Appendix D.

In Q3D, the correlation of Florschuetz et al [5.7] is used for heat transfer calculations
on flat target plates. The Nusselt number correlation is identical with that of [5.7] and
the mass flow calculation is also based on the same article. The massflow in Q3D is
calculated according to equation (5.1).

row
j
j
j
A GSR
G
G
m = &
(5.1)
The ratio
j j
G G is calculated according to equation (5.2) from Florschuetz et al.

) sinh(
) cosh(
c
n p
j
j
N
x x N
G
G

(5.2)
where
( )
( ) ( ) d x d y
C
i x x
n n
d
n

=
=
4 2
) 5 . 0 (
and i = 1, 2, 3, , N
p (5.3)
The factor GSR is calculated according to equation (5.4), where G
j
in the numerator is
calculated according to equation (5.5) and
j j
G G in the denominator according to
equation (5.2).

j j
j d
G G
G C
GSR

=
(5.4)

T R
p p
G
j
=
2
2
2
0

(5.5)
This means that
j d
G C GSR = , so that equation (5.1) actually is represented by
equation (5.6).

d row j
C A G m = &
(5.6)

For detailed information about the flow model, see [5.7]. The flow model described
above is valid for incompressible flow. The streamwise pressure gradient is assumed
to be due to the acceleration caused by the impining jets, and the wall shear is
neglected. The pressure gradient can then be described according to equation (5.7)
according to [5.7].

c c
dG G
dP

=
2

(5.7)

Impingement Cooling

73
The impingement cooling flow in turbine components is compressible. Therefore, the
incompressible model described above was compared to calculations for compressible
flow, according to equations (5.8) and (5.9). The same expression for the pressure loss
was used, i.e. equation (5.10).

(
(
|
|
.
|
\
|
|
|
.
|
\
|
=
+

1
0
2
2
0
2
1
2
p
p
p
p
(5.8)

=
d row
C A
T r
p
m
0
&
(5.9)

The massflow for a geometry and flow case according to Table 5.2 was investigated.

Table 5.2 Test Case for Q3D Mass Flow Comparsion.
y
n
/d z/d T N
p
n
c
p
0
p
2
C
d
4.5 2 750 5 10 15.1*10
5
22.1*10
5
0.79

It turned out that the two calculatation came to somewhat different results, see Table
5.3.

Table 5.3 Results of Mass Flow Comparison.
Row m& , Q3D (Compressible) [g/s] m& , Incompressible Flow [g/s]
1 9.6 7.44
2 9.7 7.51
3 9.8 7.65
4 9.9 7.87
5 10.0 8.15
6 10.1 8.52
7 10.2 8.97
8 10.0 9.5
9 9.7 10.13
10 8.9 10.85
Figure Figure 35 in Appendix B describe the C
d
obtained by using Hglunds and
Florschuetz and Isodas correlations for Case B. They give quite different results. An
explanation could be that C
d
depends on a variety of parameters, such as how sharp
the hole edge is, cross flow, pressure difference and manufacturing method.

Effect of Different Parameters
Appendix E contains the results of variations of parameters x
n
/d, y
n
/d, z/d, G
c
/G
j
and
T
c
. The base geometry is called Case D, see Table 5.4.

Impingement Cooling
74
Table 5.4 Case D, Base Geometry for Effect of Parameters.
z/d 2
x
n
/d 5
y
n
/d 4
N
p
15
T
w
1073
T
j
700
T
c
700
G
c
/G
j
0.7
Hole Pattern Inline
5.2.2 Analysis of Flat Target Plate
The agreement between correlations, considered parameters and the validity of the
ranges are analyzed. Appendix F contains a summary of the considered parameters
and the validity of ranges.
Inline Arrays
Agreement between Correlations
Figure 32 in Appendix B presents three correlations by Florschuetz, Truman and
Metzger from 1981. A number of random tests were performed which are presented in
Appendix D. The lowest curve is the base correlation which is used throughout this
report. The correlation of the middle curved is a simplified version of the base
correlation. The dotted curve is for the first line in an array with no crossflow present,
subsequently it results in a higher Nu. Since the simplified correlation results in a
higher Nu, it is safer to use the base correlation.

Results from comparison of correlations are presented in Appendix B, C and D. The
relative positions of the inline correlations are not the same for different cases, which
complicates the choice of correlation. Still, some trends exist. Obot and Trabold and
Goldstein and Seol often show the highest Nu. The latter is developed without
concern of cross flow. The deviating behavior of Obot and Trabold could be
explained by that it is slightly out of range for x
n
/d and y
n
/d. Metzger and Korstad
always result in a relatively low Nu. The behavior of Wolfersdorf correlation is very
inconsistent and, since the article is not available, difficult to analyze. Florschuetz et
al., Martin, Chance and Bailey and Bunker often result in relatively similar curves, but
the relative positions vary with the geometry. The conclusion is that the inline
correlations of Florshuetz et al., Martin, Chance and Bailey and Bunker are those
most similar to each other. For design of inline arrays, these correlations should be
investigated and analyzed for the flow case and geometry of interest.

Parameters Considered
The considered parameters for each correlation are presented in Appendix F.
Parameter T
diff
corresponds to the effect of the temperature difference between the
cross flow and the impinging jets.

All correlations consider the distance between target and jet plate -z/d and none
considers the parameter T
diff
. The correlations of Martin and of Chance were
developed for square arrays and therefore only consider one of y
n
/d and x
n
/d. Neither
Wolfersdorf nor Bailey and Bunker consider the effect of y
n
/d. Obot and Trabold and
Florschuets et al. are the ones that considers most parameters.
Impingement Cooling

75

Validity Range
The parameter ranges of the investigated correlations together with ranges relevant for
gas turbines are displayed in Appendix A. The ranges for gas turbines are taken from
impingement systems on vane 1 on SGT 600 and SGT -700 and blade 1 and vane 1
and 2 on SGT -800. Appendix F contains a summary of whether the correlation ranges
matches those relevant for turbine blades and vanes.

Partly indicates that the correlation ranges match for approximately 50 % or more.
Hardly indicates that the correlation ranges match for less than 50 %. The ranges of
Chance and Obot and Trabold are not very suitable for turbine conditions. The Bailey
and Bunker and the Florschuetz correlation matches the ranges of interest well,
whereas the Martin and Wolfersdorf ranges matches the turbine ranges mediocrly.

Staggered Arrays
Figure Figure 34 in Appendix B and Figure 38 in Appendix C present Nu for
staggered arrays. The random tests in Appenix B also display results for a staggered
array. The relative positions of the correlations vary from case to case. Trends
observed are that Metzger and Korstad gives the lowest value for all cases tested.
Goldstein and Seol often results in a relatively high value, but not always. This
correlation was developed for Nu average over the hole array area, and thus deviates
from the other correlations which are average over an impingement row. Martins
correlation agrees well with that of Behbahani and Goldstein. The results for
Florschuetz, Chance, Wolfersdorf, Behbahani and Goldstein and Martin vary from
case to case. Therefore it is difficult to make a recommendation of the use of a
specific correlation based on agreement between them.

Hglunds correlation is the only one that considers T
diff
. Florschuets correlation is
the one that considers the most parameters. Martins correlation is for square arrays,
as mentioned earlier. Bebhahani and Goldstein and Wolfersdorf consider the same
amount of parameters.

Validity Range
Florschuetzs and Hglunds correlations match the turbine ranges the best. Martin,
Wolfersdorf and Bebhahani and Goldsteins ranges dont coincide with the turbines
ranges very well, see Appendix F.

Effect of hole spacing - x
n
/d and y
n
/d
Chapter 5.1.2 states that larger hole spacing decreases the average heat transfer.
Figures Figure 49 and Figure 50 in Appendix E supports that theory. Figure Figure 49
describes the effect of x
n
/d. All correlations consider the effect of x
n
/d and display the
same basic tendency, with decreased Nu for increased x
n
/d. Obot and Trabold deviates
from the other with a significantly stronger dependence of x
n
/d. It should be noted that
this correlation is slightly out of its validity range for open area ratio. The second
strongest dependency is displayed by Baileys and Bunker. The flattest curve is that of
Hglund. Behbahani and Goldstein coincides almost exactly with Martin.
Woldersdorf, Chance, Hglund and Florshuetz have almost the same inclinations, but
Woldersdorf display significantly lower Nu values. Largest agreement between
Impingement Cooling
76
correlations is displayed by Hglund, Chance and Florschuetz. Therefore, the
influence of x
n
/d is probably best described by these correlations.
For effect of y
n
/d, Obot and Trabold deviates from the general trend again, with much
stronger dependence. Goldstein and Seol, Florschuets et al., Metzger and Korstad and
Chance display a weak dependence of y
n
/d. Behbahani and Goldstein, Martin and
Wolfersdorf is independent of y
n
/d. In Hglunds correlation, y
n
/d is connected to x
n
/d
so that
n n
x y = 3 2 for all tested plates. This leads to the conclusion that the
spanwise jet-to-jet spacing is not of significant importance for the average Nusselt
number.

Effect of jet target-to-impingement target spacing - z/d
Cross flow mass velocity G
c
was implemented as inversely proportional to z/d to
consider the effect of increased cross-section of the crossflow channel. Kercher and
Tabakoff [5.12] reported that for no crossflow, heat transfer increased with increased
z/d but with crossflow heat transfer decreased with increased z/d. A similar trend is
displayed in Figure 47 by Florshuetz et al.. Additional to the Nusselt number
calculated from equation (45) , Florschuetz et al. developed equation (47) for a row
without crossflow. The correlation for no crossflow increases with increased z/d and
the equation that considers crossflow decreases with increased crossflow. Tests were
also done with lower G
c
/G
j
ratios and for those cases, the base equation presented
similar behavior as that with no crossflow. The effect of increased Nu for increased
z/d could be based on less jet-to-jet interference with larger z/d. A decreased Nu for
larger z/d is probably due to decreased impact of the impinging jets.

Hglund, Goldstein and Seol and Florshuetz et al. display a similar dependency of z/d.
Florshuetz et al reported the highest dependency for x
n
/d = 5 and y
n
/d = 4, which is the
case tested here. For denser spaced holes, the dependence should be less [5.7].
Behbhani and Goldstein differs form the other correlations in that Nu increases
slightly with increased z/d up to z/d = 4. Baileys and Bunker, Martin, Obot and
Trabold and Goldstein and Seol display a relatively high Nu dependency of z/d.
Chance display a medium dependency of z/d. The z/d dependency without cross flow
was also investigated, see Figure 48. The same pattern as above was displayed.

Investigators have come to quite different conclusions concering the effect of jet
plate-to-target plate spacing. The study of correlations led to the conclusion that for
small z/d, the effect of jet plate-to-target plate spacing is low. For larger z/d, the heat
transfer decrease with increased z/d, as seen in Figure 47 in Appendix E.

Effect of aspect ratio - x
n
/y
n

Figure Figure 46 in Appendix E display the effect of x
n
/y
n
. The only correlation that
tested different aspect ratios is that of Florschuetz et al, which show that increased
aspect ratio leads to decreased heat transfer. The tendency is stronger for smaller
aspect ratios. Chance did not report the range of aspect ratios used, why that
correlation is not compared.

Effect of Cross Flow Ratio G
c
/G
j

Martin, Behbahani and Goldstein and Goldstein and Seol dont considerate the effect
of cross flow. Obot and Trabold investigated three different cross flow sizes, which is
shown in the correlation by a parameter n obtained from Figure 31 in Appendix A.
The effect is not seen by varying G
c
/G
j
. Metzger and Korstad consider cross flow, but
Impingement Cooling

77
shows almost no effect on Nu. Hglund, Florschuetz et al., Chance and Baileys and
Bunker show a relatively similar dependence of G
c
/G
j
, with strongest dependency for
the two latter. Wolfersdorf deviates from the former mentioned group for G
c
/G
j
> 0.4
and is probably not valid for large cross flow since that results in negative Nu. For
effect of G
c
/G
j
, Hglund and Florschuetz et al. are primarily recommended to use.

Impingement Cooling
78
5.2.3 Conclusions concerning Flat Target Plate
The investigation of impingement on flat and curvet target plates are summarized in
Appendix F. After considering the agreement between correlations, considered
parameters, validity of range and effect of different parameters, following correlations
are recommended for impingement on flat target plates.

Staggered Array: Hglund [5.8] or Florschuetz et al. [5.7]
Inline Array: Florshuetz et al.
Q3D : Q3D calculates Nu and m& according to [5.7]

It is not obvious which one of Hglund or Florschuetz correlation that is most reliable.
The advantage of Hglund equation is that the experimental study was performed at
Siemens in Finspng and aimed for components of Siemens gas turbines. Since the
tests were performed in Finspng, it is possible to look directly at the test data to
locate the nearest measured value, instead of only using a curve fit of the measured
values.
The advantage of Florschuetzs correlation is that Florschuetz and his co-workers are
experienced researchers that have performed a large number of investigations in this
area. Hglunds report is a diploma work, and he has not presented any other material
on the subject. For heat transfer calculations on turbine component, it is advised to
compare the correlations and decide from case to case which one is more reliable,
depending on range and geometry.
Impingement Cooling

79
5.3 Curved Target Plate
The results, analysis and conclusions of the investigation of impingement on curved
target plates are reported below.
5.3.1 Results concerning Curved Target Plate
Following correlations were investigated.

Chupp, Helmst, McFadden and Brown 1969 [5.1]
Hrycak 1980 [5.17]
Metzger, Yamashita and Jenkins 1969 [5.18]
Nagoga 2000 [5.20]
Kopelev 1988 [5.22]
Metzger, Baltzer and Jenkins 1972 [5.19]

The correlations and ranges are displayed in Appendix A. More articles about
impingement cooling are listed in Appendix J. The correlations are developed for a
single row of impingement. The literature about Kopelevs correlation is only
available in Russian, why the correlation has not been genuinely analyzed. The range
of Kopelevs correlation has not been found. No correlations for impingement arrays
for curved target surfaces were found.

A geometry called Case C was used for testing correlations for impingement on
curved target plates, see Table 5.5.

Table 5.5 Case C, Geomtery for Comparison of Curved Target Plate Correlations.
D
p
/d [-] z/d [-] y
n
/d [-] T
w
[K]

T
j
[K] TE outlet area [mm
2
] N
p
[-]

10 2 4 1000 600 91.62 15
The correlations were tested with Reynolds number based on the width b of a so
called equivalent slot. The equivalent slot is an imagined slot with the same area as
the total area of all holes. The length L
es
and width b of the equivalent slot are
defined in equation (5.10) and (5.11).

d N L
p es
= (5.10)

es
p
L
N
d
b
=
4
2

(5.11)

Impingement Cooling
80
Figure 5.6 display the result for a comparison of the heat transfer coefficient for Case
C.
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
0.5
1
1.5
2
2.5
x 10
4
Re
b
A
l
f
a
Alfa as a Function of Re
b
Chupp et al. 1969
Hrycak
Q3D
Metzger et al. 1969
Nagoga
Kopelev 1988
Metzger et al. 1972

Figure 5.6 Alfa for Case C.
The leading edge of vane 2 on SGT -800 is cooled by three impingement rows. Heat
transfer calculations were performed for the middle row and the results are presented
in Appendix G. The heat transfer coefficient used in the in-house cooling design
program 3Dhtcp is approximately 2200 W/m
2
/K, which can be compared to that in
Figure 53 of 5280 W/m
2
/K. The results are displayed in Appendix G.

In Q3D, Nu for impingement on curved target plates are calculated by use of Chupp et
al.. An investigation of the Q3D code showed that for a curved target plate, the code
agreed completely with the base correlation, as seen in Figure 5.7.
Impingement Cooling

81
2000 4000 6000 8000 10000 12000 14000 16000
4
6
8
10
12
14
16
18
Re
d
N
u
Q3D Comparison for Curved Target Plate
Chupp et al. 1969
Q3D

Figure 5.7 Q3D Comparison for Case C.
Since the correlations varied much, three random tests were also performed to
investigate the tendencies, see Appendix H. The effect of different parameters was
investigated and the results are displayed in Appendix I.
Impingement Cooling
82
5.3.2 Analysis of Curved Target Plate
Since the Nusselt number is based on different characteristic lengths, the correlations
are best compared by the heat transfer coefficient . Results for the tested geometries
in Appendix G and H show that the relative positions of the correlations remains
approximately fixed. The values of and the dependency of Re differs significantly
between the correlations. The correlations of the two articles of Metzger et al [5.18]
and [5.19] give very similar results and the highest values. The correlations of
Hrycak and Kopelev also result in similar values of the heat transfer coefficient,
which are lower than that of Metzger et als. Lowest is obtained by the correlations
of Chupp et al. and Nagoga, which results in very similar values.

Since many correlations result in different , it is hard to choose a correlation.
However, both Nagoga and Chupp et al. came to similar results. So did the two
investigations by Metzger et al., but those correlations results in much higher
than Chupp et al.s and Nagogas. It is better to estimate a too low value than a too
high. Therefore, Nagogas or Chupp et als correlations are of interest. However,
Chupp et al.s correlation is not valid for the range that is important for turbine
cooling. Therefore, the Nagoga correlation is probably better to use.

All correlations consider D
p
/d and y
n
/d. Only Nagoga consider T
diff
. Only Chupp et al.
and Nagoga consider the effect of z/d. Nagogas correlations is the one that consider
the most parameters, which promotes the use of Nagoga over the other in this aspect.

Validity Range
The range of Nagogas correlation matches that of turbine components the best.

Effect of Different Parameters
Appendix I display the effect of different parameters. All tested correlations show
approximately the same dependence of the hole spacing. Metzger, Baltzer and
Jenkins correlation deviates a little from the the other however. The strange behavior
of the correlation is probably due to the form it is presented in, see Appendix A.

Only Nagoga and Chupp et al. display a dependency of z/d within valid z/d range. Nu
decreases slightly with increased z/d for both Nagoga and Chupp et al, with a little
stronger dependency for Chupp et al.. Valid range for Kopelevs correlation is not
known, but it displays the same trend as Nagogas.

Hrycak, Chupp et al. and Nagoga display approximately the same dependency of
target plate curvature within valid ranges. Nu decreases with increased D
p
/d for all
three correlations, except for small curvatures where Nu increases with increased D
p

according to the correlation of Chupp et al..

Impingement Cooling

83
5.3.3 Conclusions concerning Curved Target Plate
The investigated correlations give very different results. However, taking into account
ranges, parameters considered and the agreement between correlations, the Nagoga
correlation [5.20] is recommended to use. If possible, it is also recommended to
consult the other correlations when calculating Nusselt number for a specific case.
The Q3D code agrees with the correlation it is based on, i.e. that of Chupp et al. [5.1].

Impingement Cooling
84
5.4 References Impingement Cooling
Literature
[5.1] Chupp, R. E., Helmst, H. E., McFadden, P. W. and Brown, T. R. Evaluation of
Internal Heat-Transfer Coefficients for Impingement-Cooled Turbine Airfoils,
Journal of Aircraft, 6, (1969), 203-208
[5.2] Nashar, Impingment Heat Transfer and Flow Characteristic in Gas Turbine
Blade Cooling, ABB Technical Summary (1992)
[5.3] Viskanta, R. Heat Transfer to Impinging Isothermal Gas and Flame Jets,
Experimental Thermal and Fluid Science, 6, (1993), 111-134
[5.5] Strand, T. Strmning och vrmeverfring i luftstrlar anstrmmande
vinkelrtt mot en vggyta, Chalmers Tekniska Hgskola, Institutionen fr
Tillmpad Termodynamik och Strmningslra, (1972)
[5.7] Florschuetz, L.W., Truman C. R. and Metzger D.E. Streamwise Flow and Heat
Transfer Distributions for Jet Array Impingement with Crossflow, Journal of
Heat Transfer, 103, (1981), 337-342
[5.8] Hglund, H. Experimental Investigation of Impingement Cooling Under a
Staggered Array of Circular Jets, Thesis Work at the Department of Energy
Tehnology Royal Institute of Technologi, KTH (1999)
[5.9] Chance, J. L. Experimental Investigation of air Impingement Heat Transfer
Under an Array of Round Jets, Tappi, 57, (1974), 108-112
[5.10] Metzger, D.E. and Korstad, R. J. Effects of Crossflow on Impingement Heat
Transfer, Journal of Engineering for Power, January (1972), 35-42
[5.11] Martin, H. Heat and Mass Transfer between Impinging Gas Jets and Solid
Surfaces, Advances in Heat Transfer, Academic Press, 13, 1-60 (1977)
[5.12] Kercher, D. M. and Tabakoff, W. Heat Transfer by a Square Array of Round
Air Jets Impinging Perpendicular to a Flat Surface Including the Effect of
Spent Air, ASME Journal of Engineering for Power, 73-82, (1970)
[5.13] Obot, N.T. and Trabold, T.A. Impingement Heat Transfer within Arrays of
Circular Jets: Part 1-Effects of Minimum, Intermediate, and Complete
Crossflow for Small and Large Spacings, Transactions of the ASME, 109,
(1987), 872-879
[5.14] Bailey, J. C. and Bunker, R. S. Local Heat Transfer and Flow Distributions for
Impinging Jet Arrays of Dense and Sparse Extent, Proc. of ASME TURBO
EXPO, Amsterdam, June 3-6 2002, Research and Development Center
General Electric Company, Niskayuna, NY, (2002)
[5.15] Behbahani, A.I, and Goldstein, R.J. Local Heat Transfer to Staggered Arrays
of Impinging Circular Air Jets, Journal of Engineering for Power, 105,
(1983), 354-360
[5.16] Goldstein, R.J. and Seol, W. S. Heat Transfer to a Row of Impinging Circular
Air Jets Including the Effect of Entrainment, Int. J. Heat Mass Transfer, 34,
(1991), 2133-2147
[5.17] Hrycak, P. Heat Transfer from a Row of Impinging Jets to Concave
Cylindrical Surfaces, International Journal of Heat Mass Transfer, 24, 407-
419, (1981)
Impingement Cooling

85
[5.18] Metzger, D.E., Yamashita, T. and Jenkins, C. W. Impingement Cooling of
Concave Surfaces With Lines of Circular Air Jets, Journal of Engineering for
Power, 91, 149-158, (1969)
[5.19] Metzger, D. E., Baltzer, R. T. and Jenkins, C. W. Impingement Cooling
Performance in Gas Turbine Airfoils Including Effects of Leading Edge
Sharpness, Journal of Engineering for Power, July, 219-225, (1972)
[5.20] Nagoga, G. Intensification of Heat Tranfer in the Cooling Channels of Gas
Turbine Blades, Part III, Moscow, (2001)
[5.21] Florschuetz, L. W. and Isoda, Y. Flow Distributions and Discharge Coefficient
Effects for Jet Array Impingement with Initial Crossflow, ASME Journal of
Engine Power, 105, 296-303, (1983)
[5.22] Kopelev (1988), ISBN-5-217-00118-6

Personal Communication
[5.23] Xiufang Gao, department GRCCC at Siemens, Finspng, October 2005
Conclusions

87
6 Conclusions
Following conclusions were drawn for the three investigated cooling methods.
6.1 Rib Turbulated Cooling
Recommended Correlations:

Han 1988 [3.5] for transverse ribs in ducts with two opposing ribbed walls
Han et al. 1989 [3.8] for angled ribs in ducts with two opposing ribbed walls
with 1 4 1 < H W .
Han & Park 1988 [3.9] for angled ribs in ducts with two opposing ribbed
walls with 4 1 H W .
Chandra et al. 1997 [3.6] for duct with one, three or four ribbed walls.
Nagoga 2000 [3.4] for U-shaped ducts with SR- or SSR-schemes.

Conclusions concerning Q3D:
Nu in Q3D is lower than Nu from the Han correlations it is based on.
Friction in Q3D is considerably higher than the Han correlations due to a bug
that needs further investigation.
Three input parameters concerning number of ribbed walls are missing in the
interface. The parameters are part of the solution to the high friction factor, but
needs further investigations.
It would be easy to change the code from the Reduced Area Method to the
Total Area Method, which is recommended.

Conclusions concerning Multipass:
Four sub routines, Ribbed 1 to Ribbed 4 are used.
Ribbed 1 is based on [3.7]. St the same as in [3.7] but f is numerically
manipulated and differs from [3.7].
Ribbed 2 is based on [3.5]. St is the same as in [3.5] but f is numerically
manipulated and differs from [3.5].
Ribbed 3 is based on [3.8]. St differs from [3.8] due to a small bug and f is
numerically manipulated and also differs form [3.8].
Ribbed 4 is based on a mixture of [3.5] and [3.8]. St differs from the original
correlations due to a small bug in the code but f agrees with the original
correlations.

Conclusions
88
6.2 Matrix Cooling
Conclusions about Correlations:
Literature about matrix cooling is limited.
Only [4.2] contained matrix correlations. The correlations are valid for closed
matrixes.
Flow in initial or direct matrix channels can be approximated to flow in
smooth ducts with inlet effects.
Further work is needed to develop correlations for matrix cooling.

Conclusions about Q3D:
Q3D calculates rib efficiency for ribs with isolated tops.
The code should be changed according to equation (4.29), which considers
both the isolated and convective rib tops.
The phrase Rib Effectiveness in the Q3D interface should be replaced by
Rib Efficiency.

6.3 Impingement Cooling
Flat Target Surface:
Use Hglund [5.8] or Florschuetz et al. [5.7] for correlations for staggered
impingement arrays. Compare the correlations and decide from case to case
which one is more reliable, depending on range and geometry.
Use Florschuetz et als [5.7] correlation for inline impingement arrays.
Q3D calculates Nu and m& according to [5.7]

Curved Target Surface:
The investigated correlations give very different results.
Use the Nagoga correlation, but it is also recommended to consult the other
correlations when calculating Nusselt number for a specific case.
Q3D is based on the correlations of Chupp et al. [5.1]. The code agrees with
the base correlation.
Future Work

89
7 Future Work
The heat transfer handbook will also contain correlations and information about other
cooling methods used for internal cooling of blades and vanes. Examples of such
cooling methods are listed below.

Pin Fin Cooling
Dimple Cooling
Cyclone Cooling

Pin fin cooling consists of single or arrays of pins that increases heat transfer due to
increased turbulence and increased heat transfer surface area. Dimple cooling consists
of small circular cavities in the surface, dimples, which disturb the boundary layer and
thus increases heat transfer. Cyclone cooling cools the leading edge under through
cross supply of cooling air. For more information about these and other cooling
methods, see [4.2].

Another part of the further work is to investigate the Q3D code more concerning the
reported problems that are not fully solved. The next step of this work is also to
implement the recommended correlations into the heat transfer program Q3D.

Appendix 1 - Rib Turbulated Cooling

91

92
Appendix A Correlations
Valid ranges for the correlations are displayed in Figure 3.10.
Webb, Eckert and Goldstein 1970 [3.13]
The correlations below are valid for ribbed tubes with circular cross-section. The
estimated errors of the correlations were approximately 12 % for the friction factor
and 15 % for the Stanton number.
(
(
(
(
(
|
|
.
|
\
|
+ |
.
|
\
|
=
+
2
) ( 75 . 3
2
ln 5 . 2
2
e R
e
D
f
rc

(1)

53 . 0
95 . 0 ) ( |
.
|
\
|
=
+
e
P
e R (2)

| | ) ( Pr ) (
2
1
2
57 . 0 + +
+
=
e R e G
f
f
St
rc
rc
rc

(3)

28 . 0
) ( 50 . 4 ) (
+ +
= e e G (4)

5 . 0
2
Re |
.
|
\
|
=
+ rc
f
D
e
e (5)

Han 1988 [3.5]
The correlations below are valid for rectangular ducts with ribs on two opposing
walls. The deviations from measured data were 6 % for R(e
+
), 8 % for G(e
+
)
and 10 % for ) (
+
e G .
2
4
5 . 2
2
ln 5 . 2 ) (
2
|
|
.
|
\
|
|
|
.
|
\
|

=
+
Z
D
e
e R
f
h
r

(6)

H W
W
f
W
H
f f
s r
+
|
.
|
\
|
+ =
4 4

(7)

2 . 0
4
Re 046 . 0

=
s
f (8)

35 . 0
10 2 . 3 ) ( |
.
|
\
|
=
+
e
P
e R (9)


93
H W
W
Z
+
=
2
(10)

( )
(
(
+ |
.
|
\
|
=
+ +
1
2
) ( ) ( 2
2 1
4
4
2
r
r
r
f
e R e G
f
St
(11)

28 . 0
) ( 7 . 3 ) (
+ +
= e e G (12)

( )
4 4 s r
r
s
r r
f f
W
W
f f + =
(13)

5 . 0
4
2
Re |
.
|
\
|
=
+ r
h
f
D
e
e (14)

Chandra, Niland and Han 1997 [3.6]
The correlations below are valid for rectangular ducts with 1, 2, 3 or 4 ribbed walls.
The deviations from measured data were 4 % for R(e
+
) and 4 % for G(e
+
).
(
+
+
=
4 4
1
1
s
r
s
r
r s
r
f
W
W
f
W W
f
(15)

2
4
5 . 2
2
ln 5 . 2 ) (
2
|
|
.
|
\
|
|
|
.
|
\
|
=
+
Z
D
e
e R
f
h
r

(16)

41 . 3 ) ( =
+
e R (17)

H W
W
Z
+
=
2
(18)

2 . 0
4
Re 046 . 0

=
s
f (19)

5 . 2
2
ln 5 . 2 ) (
2
4
|
|
.
|
\
|
=
+
Z
D
e
e G
f
St
h
r
r

(20)

156 . 0
435 . 0
) ( 427 . 1 ) (

+ +
|
.
|
\
|
=
W
W
e e G
r
(21)

94
5 . 0
4
2
Re |
.
|
\
|
=
+ r
h
f
D
e
e (22)

Han, Glicksman & Rohsenow 1978 [3.7]
2
75 . 2
2
ln 5 . 2 ) (
2
|
|
.
|
\
|
|
|
.
|
\
|
=
+
h
D
e
e R
f
(23)

n
e P
e R |
.
|
\
|
|
.
|
\
|
=
+
10
45
9 . 4 ) (
57 . 0
(24)
where n = -0.13 if P/e < 10
n = ( )
71 . 0
90 53 . 0 if P/e 10

| | 2 2 ) ( ) ( +
=
+ +
f e R e H
f
St
r

(25)

( )
( )
j
e
e H
45
35 10
) (
28 . 0
+
+

= (26)

where j = 0.5 for 45 <
j = -0.45 for 45

5 . 0
2
Re |
.
|
\
|
=
+
f
D
e
e
h
(27)

Han & Park 1988 [3.9]
+
), 8 % for G(e
+
) and
10 % for ) (
+
e G .

2
4
5 . 2
2
ln 5 . 2 ) (
2
|
|
.
|
\
|
|
|
.
|
\
|
=
+
Z
D
e
e R
f
h
r

(28)

( )
(
(
|
.
|
\
|
+ |
.
|
\
|
=
+
2
35 . 0
90
86 . 17
90
07 . 27 31 . 12 10 ) (

m
H
W
e P e R
(29)

where m = 0 if = 90
m= 0.35 if < 90
if W/H > 2 let W/H = 2


95
H W
W
f
W
H
f f
s r r
+
+ =
4 4 2

(30)

2 . 0
4
Re 046 . 0

=
s
f (31)

| |

)
+
=
+ +
1
2
) ( ) ( 2
4
4
2
r
r
r
f
e R e G
f
St
(32)

( ) ( ) ( ) ( )
n m
e P e H W e G 10 90 24 . 2 ) (
35 . 0
1 . 0
=
+ +

(33)

where m = 0.35 & n = 0.1 for square ducts
m = 0 & n = 0 for rectangular ducts

Smooth Side Wall:

) ( 2 . 1 ) (
+ +
= e G e G (34)

| |

)
+
=
+ +
1
2
) ( ) ( 2
4
4
r
r
f
e R e G
f
St
(35)

( )
2 r s
St St
H
W
St St + = (36)

5 . 0
2
Re |
.
|
\
|
=
+
f
D
e
e
h
(37)

Han, Ou, Park & Lei 1989 [3.8]
+
) and 8 % for G(e
+
).
2
4
5 . 2
2
ln 5 . 2 ) (
2
|
|
.
|
\
|
|
|
.
|
\
|
=
+
Z
D
e
e R
f
h
r

(38)

m
H
W
e R |
.
|
\
|
(
(
|
.
|
\
|
+ =
+
2
90
86 . 17
90
07 . 27 31 . 12 ) (

(39)

where m = -0.5 if 60 90
m = ( )
2
60 5 . 0 if 30< < 60
m = 0 if 30

96
H W
W
f
W
H
f f
s r r
+
+ =
4 4 2

(40)

2 . 0
4
Re 046 . 0

=
s
f (41)

| |

)
+
=
+ +
1
2
) ( ) ( 2
4
4
2
r
r
r
f
e R e G
f
St
(42)

( )
n
e C e G
+ +
= ) (
(43)
where
for H W 1 n = 0.35
C = 2.24 if =90
C =1.80 if 60 < 90

for < H W n = ( )
44 . 0
35 . 0 H W
C = ( )
76 . 0
24 . 2

H W if =90
C = ( )
76 . 0
80 . 1

H W if 30 < 90

5 . 0
4
2
Re |
.
|
\
|
=
+ r
h
f
D
e
e (44)

Han 1984 [3.12]
The estimated errors of the correlations were approximately 10 % for the friction
factor and 10 % for the Stanton number.

2
4
2
ln 5 . 2 5 . 2
2
ln 5 . 2 ) (
2
|
|
.
|
\
|
|
.
|
\
|
+

|
|
.
|
\
|
=
+
W H
H
D
e
e R
f
h
r

(45)

( )
53 . 0
95 . 0 ) ( e P e R =
+
(46)

H W
f H f W
f
r s
r
+
+
=
4 4
2
(47)

W H
St H St W
St
r s
r
+
+
=
4 4
2
(48)

( )
2 . 0 6 . 0 2 . 0
4
2 Pr Re
023 . 0
h av
s
D R
St

=
(49)


97
( ) ) ( ) ( 2 1
2
4
2
4
+ +
+
=
e R e G f
f
St
r
r
r

(50)

( )
57 . 0
28 . 0
Pr 5 . 4 ) ( =
+ +
e e G
(51)

5 . 0
4
2
Re |
.
|
\
|
=
+ r
h
f
D
e
e (52)

Han, Park & Lei 1985 [3.10]
+
) and 10 % for H(e
+
).

2
2
5 . 2
2
ln 5 . 2 ) (
2
|
|
.
|
\
|
|
|
.
|
\
|
=
+
h
r
D
e
e R
f
(53)

( ) ( )
n
e e P e R ) 003 . 0 10
90
1 . 21
90
6 . 31 6 . 15 ) (
35 . 0
2
+ +

(
(
|
.
|
\
|
+ =

(54)

where n = 0 if 45
n = 0.17 if < 45

5 . 0
2
2
Re |
.
|
\
|
=
+ r
h
f
D
e
e (55)

( ) ( )
28 . 0
3 . 0
90 74 . 3 ) (
+ +
= e e H
(56)
| |

)
+
=
+ +
1
2
) ( ) ( 2
2
2
2
r
r
r
f
e R e H
f
St
(57)
98
Appendix B Correlations for U-shaped Channels
The correlations are valid for 180000 Re 28000
d
13 5 e P , 90 0 and
1 . 0 =
h
D e .

SR-scheme
s Nu r
Nu C Nu = for 20
h
D x
(58)

x s Nu x r
Nu C Nu
, ,
= for 20
h
D x
(59)

( )
2 . 0
1

|
.
|
\
|
+ =
e
P
K C
Nu

where
K is found in Table 1 (60)

55 . 0 4 . 0 8 . 0
,
Pr Re 035 . 0

=
w x x s
T Nu for 20
h
D x
(61)

55 . 0 4 . 0 8 . 0
Pr Re 023 . 0

=
w s
T Nu for 20
h
D x
(62)

( )
R e
R e
F
2
1 1
1

+ = 1 94 . 0 < (63)

Table 1 Values of K
.
0 30 45 60 90
K (SR) 1.00 2.20 3.40 2.90 1.25
K (SSR) - 4.4 3.9 2.6 1.3
K (SSR)
- 6.00 9.30 5.90 3.35

s f r
f C f =
(64)

25 . 0
Re 3164 . 0

=
l s
K f or
25 . 0
Re 838 . 0

=
l s
f (65)

( )
25 . 0
65 . 2

=
h l
D l K (66)

( )
2 . 0
3 . 3 1

|
.
|
\
|
+ =
e
P
C
f
(67)

SSR-scheme
Following correlations are valid for 13 67 . 1 e P , 90 30 , 5 . 1 5 . 0 R e
and 000 140 Re 000 25 .

s Nu r
Nu C Nu = for 20
h
D x , for
s
Nu , see equation (61)
(68)

s
Nu
r Nu C Nu = for 20
h
D x (69)

99

x s Nu x r
Nu C Nu
, ,
= for 20
h
D x ,for
x s
Nu
,
, see equation (62)
(70)

( )
25 . 0 33 . 0
1 |
.
|
\
|
|
.
|
\
|
+ =

R
e
e
P
K C
Nu

where
K is found in Table 1 (71)

55 . 0 4 . 0 8 . 0
2 . 0
Pr Re 044 . 0

|
|
.
|
\
|
=
w
h
s T
D
x
Nu (72)

( )
+ =
s r
s r
f
f f

(73)

( )
+ =
1
1

C
C
f
(74)

( )
8
1 arccos 2
1
+

=
R e
for R e in current study

(75)

8
2 1
1
+
+
=
R e
for R e in current study

(76)

s s
f =
(77)

s f r
f C f =
(78)

s r
C

=
(79)

25 . 0
Re 3164 . 0

=
l s
K for 20
h
D x

(80)

25 . 0
Re 84 . 0

=
x s
for 20
h
D x (81)

25 . 0
65 . 2

|
|
.
|
\
|
=
h
l
D
x
K (82)

( )
25 . 0 33 . 0
1 |
.
|
\
|
|
.
|
\
|
+ =

R
e
e
P
K C

where
K is found in Table 1

( 83)
100
Appendix C Results for Case A
1 2 3 4 5 6 7 8
x 10
4
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
Nusselt Number Enhancement K for a Ribbed Duct
Reynolds number
H
e
a
t

T
r
a
n
s
f
e
r

E
n
h
a
n
c
e
m
e
n
t

F
a
c
t
o
r
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 1 K for Ribbed Side Wall for Case A.

1 2 3 4 5 6 7 8
x 10
4
4
6
8
10
12
14
16
18
Friction Factor Enhancement C for a Ribbed Duct
Reynolds number
F
r
i
c
t
i
o
n

E
n
h
a
n
c
e
m
e
n
t

F
a
c
t
o
r
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984,OBS! Nu
average
, Nu
average
Han et al. 1985
Q3D

Figure 2 C for Ribbed Side Wall for Case A.

101
1 2 3 4 5 6 7 8
x 10
4
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0.55
0.6
Efficiency, eff=K/C for a Ribbed Duct
Reynolds Number
e
f
f
=
K
/
C
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984,OBS! Nu
average
Han et al. 1985
Q3D

Figure 3 eff for Ribbed Side Wall for Case A.

102
Appendix D Effect of Rib Angle
Effect of rib angle,

30 45 60 75 90
0
0.5
1
1.5
2
2.5
3
3.5
Effect of alfa on K for a Ribbed Duct. Re = 30 000
alfa
K
Han & Park 1988
Han et. al 1989
Han et al. 1985
Q3D
Han 1988

Figure 4 K as a Function of alfa, Case A.
30 45 60 75 90
2
4
6
8
10
12
14
16
18
20
22
Effect of alfa on C for a Ribbed Duct. Re = 30000
alfa
C
Han & Park 1988
Han et. al 1989
Han et al. 1985
Q3D

Figure 5 C as a Function of alfa, Case A.

103

30 45 60 75 90
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Effect of alfa on Efficiency, eff=K/C for a Ribbed Duct. Re = 30 000
alfa
C
Han & Park 1988
Han et. al 1989
Han et al. 1985
Q3D

Figure 6 C as a Function of alfa, Case A.

104
Appendix E Effect of Pitch-to-Rib Height
Effect of P/e
10 15 20
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
Effect of P/e on K for a Ribbed Duct. Re = 30 000
P/e
K
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 7 Effect of Rib Pitch-to-Height ratio on K for Ribbed Side Wall.

10 15 20
2
4
6
8
10
12
14
16
Effect of P/e on C for a Ribbed Duct. Re = 30 000
P/e
C
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 8 Effect of Rib Pitch-to-Height ratio on C for Ribbed Side Wall.

105

10 15 20
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Effect of P/e on Efficiency, eff=K/C for a Ribbed Duct. Re = 30 000
P/e
e
f
f
=
K
/
C
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 9 Effect of Rib Pitch-to-Height Ratio on eff for Ribbed Side Wall.

5 10 15 20
1.7
1.75
1.8
1.85
1.9
1.95
2
2.05
2.1
2.15
2.2
Effect of P/e on K for a Ribbed Duct. Re = 30 000
P/e
K

Figure 10 Effect of Rib Pitch-to-Height Ratio for Han et al. 1978. W/H=12, e/D
h
=0.08.
106
Appendix F Effect of Aspect Ratio
Effect of aspect ratio
0 0.5 1 1.5 2 2.5 3 3.5 4
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
Effect of W/H on K for a Ribbed Duct. Re = 30 000
W/H
K
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 11 K as a Function of W/H, Case A.

0 0.5 1 1.5 2 2.5 3 3.5 4
0
2
4
6
8
10
12
14
16
18
20
Effect of W/H on C for a Ribbed Duct. Re = 30 000
W/H
C
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 12 C as a Function of W/H, Case A.


107
0 0.5 1 1.5 2 2.5 3 3.5 4
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Effect of W/H on Efficiency, eff=K/C for a Ribbed Duct. Re = 30 000
W/H
e
f
f
=
K
/
C
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 13 eff as a Function of W/H, Case A.
108
Appendix G Effect of Rib Height
Effect of e/Dh
0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
Effect of e/Dh on K for a Ribbed Duct. Re = 30 000
e/Dh
K
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 14 K as a Function of e/D
h
, Case A.
0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075
2
4
6
8
10
12
14
16
Effect of e/Dh on C for a Ribbed Duct. Re = 30 000
e/Dh
C
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 15 C as a Function of e/D
h
, Case A.

109

0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Effect of e/Dh on Efficiency, eff=K/C for a Ribbed Duct. Re = 30 000
e/Dh
e
f
f
=
K
/
C
Han 1988
Han & Park 1988
Han et. al 1989
Han 1984, OBS! Nu
average
Han et al. 1985
Q3D

Figure 16 eff as a Function of e/D
h
, Case A.

110
Appendix H Q3D Correlations
The value of A
s
depends on both rib angle and channel aspect ratio. Therefore it is not
possible to explain the calculation of A
s
without use of a function called Sfun from the
code. Appendix F contains the the equations from Q3D, function Sfun and the vectors
xbs, cbs, xfs and cfs that are used for calculation of A
s
. The equations are presented in
the reduced area form.

A
s
is calculated by function Sfun, see below. The vectors xbs, cbs, xfs and cfs that are
used for calculation of A
s
are displayed in Table1. The Q3D equations below are
presented in the reduced area form.

Rib angle of 90
( )
( ) ( )
( ) e H W
e H D e W
f
n
cbs xbs H W Sfun
A
n
A
h
s
s
r
r
+
+ +
=
=
=
=
=
2
2 03 . 0 65 . 1 03 . 0 001231 . 1
871764 . 0
100
, , , 5
77 . 0
053 . 0

(84)
Rib angle of 75
( )
( ) ( ) ( ) ( ) ( )
( ) e H W
e H D e H W W
f
n
cfs xfs H W Sfun A
n
H W
A
h
s
s
r
r
+
+ +
=
=
=
=
|
|
.
|
\
|
=
2
2 03 . 0 65 . 1 03 . 0 127 . 0 127 . 1 554446 . 1
763682 . 0
, , , 4 04369951 . 0
688383 . 0
4
143712 . 0
077 . 0

(85)
Rib angle of 60
( )
( ) ( ) ( ) ( ) ( )
( ) e H W
e H D e H W W
f
n
cfs xfs H W Sfun A
n
H W
A
h
s
s
r
r
+
+ +
=
=
=
=
|
|
.
|
\
|
=
2
2 03 . 0 65 . 1 03 . 0 127 . 0 127 . 1 524792 . 1
736073 . 0
, , , 4 05741897 . 0
644134 . 0
4
20802 . 0
154 . 0

(86)

111
Rib angle of 45
( )
( ) ( ) ( ) ( ) ( )
( ) e H W
e H D e H W W
f
n
cfs xfs H W Sfun A
n
H W
A
h
s
s
r
r
+
+ +
=
=
=
=
|
|
.
|
\
|
=
2
2 03 . 0 65 . 1 03 . 0 127 . 0 127 . 1 032169 . 1
743917 . 0
, , , 4 05153355 . 0
620173 . 00
4
229176 . 0
231 . 0
(87)

Rib angle of 30
( )
( ) ( ) ( ) ( ) ( )
( ) e H W
e H D e H W W
f
n
cfs xfs H W Sfun A
n
H W
A
h
s
s
r
r
+
+ +
=
=
=
=
|
|
.
|
\
|
=
2
2 03 . 0 65 . 1 03 . 0 127 . 0 127 . 1 539795 . 0
8047405 . 0
, , , 4 02649813 . 0
601749 . 0
4
23931 . 0
308 . 0

(88)

Smooth duct
03 . 0
8 . 0
018 . 0
8 . 0
018 . 0
=
=
=
=
=
f
n
A
n
A
s
s
r
r

(89)

112
Sfun takes 4 variables, N, X, Y and C and returns a number. N and X are numbers and
Y and C are vectors.

The function Sfun in Visual Basic code:

Function Sfun(N, X, Y, C) As Double
Dim a As Double
Dim mySfun As Double

mySfun = C(N + 1) + C(N + 2) * X

For i = 1 To N
If X > Y(i) Then
a = X - Y(i)
mySfun = mySfun + C(i) * a * a * a
Else
Exit For
End If
Next i

Sfun = mySfun
End Function

Table1 Vectors used to Calculate A
s
.
xbs cbs xfs cfs
0.5 0.054143 0.5 -0.100696
1 .040695 0 1 0.18197
1 -0.145305 2 -0.096736
2 0.062164 4 0.015463
4 -0.011698 1.07578
1.099973 -0.066973
0.093911

113
Appendix I Multipass Results
Ribbed 1
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
x 10
4
60
70
80
90
100
110
120
130
140
150
160
Nusselt Number for Ribbed 1
Reynolds Number
N
u
s
s
e
l
t

N
u
m
b
e
r
Ribbed 1,4 ribbed walls
Han et al. 1978
Ribbed 1,2 ribbed walls

Figure 17 Nu Calculated for W=120 mm, H=10 mm, P=8mm, e=1.5 mm and alfa=90 deg.

1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
x 10
4
0.1015
0.102
0.1025
0.103
0.1035
0.104
0.1045
0.105
0.1055
Friction Factor for Ribbed 1
Reynolds Number
F
r
i
c
t
i
o
n

F
a
c
t
o
r
Ribbed 1
Han et al. 1978

Figure 18 f Calculated for W=120 mm, H=10 mm, P=8mm, e=1.5 mm and alfa=90 deg.

114

Ribbed 2
1 2 3 4 5 6 7 8
x 10
4
50
100
150
200
250
300
350
400
450
Reynolds Number
N
u
s
s
e
l
t

N
u
m
b
e
r
Ribbed 2
Han & Park 1988
Ribbed 2, corrected

Figure 19 Nu Calculated for W=5 mm, H=5 mm, P=3.15mm, e=0.315 mm and alfa=90 deg.

1 2 3 4 5 6 7 8
x 10
4
0.01
0.015
0.02
0.025
0.03
0.035
0.04
Reynolds Number
F
r
i
c
t
i
o
n

F
a
c
t
o
r
Ribbed 2
Han & Park 1988

Figure 20 f Calculated for W=5 mm, H=5 mm, P=3.15mm, e=0.315 mm and alfa=90 deg.

115
Ribbed 3
1 2 3 4 5 6 7 8
x 10
4
50
100
150
200
250
300
350
400
450
500
Reynolds Number
N
u
s
s
e
l
t

N
u
m
b
e
r
Ribbed 3
Ribbed 3, corrected
Han et al. 1989


1 2 3 4 5 6 7 8
x 10
4
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0.11
Reynolds Number
F
r
i
c
t
i
o
n

F
a
c
t
o
r
Ribbed 3
Ribbed 3, corrected
Han et al. 1989


116
Ribbed 4
1 2 3 4 5 6 7 8
x 10
4
50
100
150
200
250
300
350
400
450
Reynolds Number
N
u
s
s
e
l
t

N
u
m
b
e
r
Ribbed 4
Ribbed 4, corrected
Han & Park 1988
Han et al. 1989

1 2 3 4 5 6 7 8
x 10
4
0.01
0.015
0.02
0.025
0.03
0.035
0.04
Reynolds Number
F
r
i
c
t
i
o
n

F
a
c
t
o
r
Ribbed 4
Han & Park 1988
Han et al. 1989


117
Appendix J Interesting Articles
Further information about rib turbulated cooling can be found in the articles below.

Author Title Literature
Hong & Hsieh Heat Transfer and Friction Factor
Measurements in Ducts with
Staggered and Inline Ribs
Journal of Heat Tranfer,
Vol. 115, February 1993

Kiml, Mochizuki &
Murata
Effects of Rib Arrangements on
Heat Transfer and Flow Behavior
in a Rectangular Rib-Roughened
Passage: Application to Cooling
of Gas Turbine Blade Trailing
Edge
Journal of Heat Transfer,
Vol. 123, August 2001

Hwang Heat Transfer-Friction
Characteristic Comparison in
Rectangular Ducts with Slit and
Solid Ribs Mounted on One Wall

Chandra, Fontenot
& Han
Effect of Rib Profiles on
Turbulent Channel Flow Heat
Transfer
Thermophysics, Vol. 12,
No. 1: Technical Notes,
1997

Taslim & Spring Effects of Turbulator Profile and
Spacing on Heat Transfer and
Friction in a Channel
Journal of Thermophysics
and Heat Transfer, Vol. 8,
No. 3, July-Sept. 1994

Acharaya, Eliades
& Nikitopoulos
Heat Transfer Enhancements in
Rotating Two-Pass Coolant
Channel With Profiled Ribs: Part
1- Average Results
Journal of
Turbomachinery, Vol.
123, Januray 2001

Han, Zhang & Lee Augmented Heat Transfer in
Square Channels with Parallell,
Crossed, and V-shaped Angled
Ribs

Korotky & Taslim Rib Heat Transfer Coefficient
Measurements in a Rib-
Roughened Square Passage
Journal of
120, April 1998

Taslim &
Wadsworth
An Experimental Investigation of
the Rib Surface-Average Heat
Transfer Coefficient in a Rib-
Roughened Square Passage
Journal of
119, April 1997
Appendix 2 - Matrix Cooling

119
120
Appendix A Zinc Test
Figure 25 describes the basics of a zink test.

Figure 25 Zink Test Facility [4.3].

Following 6 steps describe how the tests were carried out.

1. The test object is connected to air supply, and is provided with pressure taps
for static pressure measurements. A container with pure zinc is heated in a
furnace to a temperature slightly higher than zinc crystallization temperature.
2. The test object is lowered into the container, which still is in the furnace, and
stays there until temperature equalization between object wall and melt is
reached. The test is monitored during this time to prevent metal crust
formation on the object walls at the initial time of its submersion into the melt.
3. The container with the objects is taken out of the furnace. Cooling of the melt
and object down to the crystallization temperature of zinc takes place.
4. When the crystallization point is reached for the melt and the test object, the
test object is exposed to air blowing through its cooling channels which cools

121
the object walls intensively. A metal crust is formed on the surface of the test
object.
5. After a certain time, the blowing of cooling air is stopped. The test object is
taken out of the container and cooled down to room temperature. The
container is replaced in the furnace.
6. The metal crust of the object is removed and marks along the object contour
are made. Cut along the marks are made.

Each test is performed three to five times, to eliminate errors. Crust thickness
measurements are made. The heat flow determination error does not exceed 3 % to 8
%. [4.3] The overall heat transfer is evaluated from the mass of the zinc crust.
Measuring the mass is more accurate than measuring the thickness.

122
Appendix B Correlations and Ranges of [4.2]
Heat Transfer on Base Shell
The ranges for which Nagogas correlations are valid are displayed in Table 2.

Table 2 Ranges for heat transfer correlations.
Re
d
3000-60 000
T
w
/T
f
[K] 1.4-2.2
[] 0-70
d [mm] 1.5-3
L/W 0.96-5.2
0.5-1
W/d 6-34
P [mm] 1.05-2
l/d 12-67
Initial Section
55 . 0
4 . 0 8 . 0
Pr Re 0289 . 0

|
|
.
|
\
|
=
f
w
x x
T
T
Nu (90)
55 . 0
4 . 0 8 . 0
Pr Re 0361 . 0

|
|
.
|
\
|
=
f
w
l l
T
T
Nu (91)
K
T
T
Nu
f
w
d d

|
|
.
|
\
|
=
55 . 0
4 . 0 8 . 0
Pr Re 023 . 0 (92)
( )
36 . 0
4
45 . 0
10 Re 63 . 0 1

|
.
|
\
|
+ =
d
d
l
K (93)
2 . 0
Re 43 . 0

=
x
f (94)
d
x
d x
= Re Re (95)
d
l
d l
= Re Re (96)

Basic Section
55 . 0
4 . 0
Pr Re

|
|
.
|
\
|
=
f
w n
x x
T
T
n A Nu (97)
55 . 0
4 . 0
Pr Re

|
|
.
|
\
|
=
f
w n
l l
T
T
A Nu (98)
( ) ( ) 2 sin 77 . 12 1 0361 . 0
2
+ = A (99)
|
|
.
|
\
|
+
(
(
|
.
|
\
|

= 1 1 1
4
204 . 0 8 . 0
2
n
(100)
( ) ( ) ( ) ( ) ( )
m
x x
Re 2 sin 77 . 12 1 1 1 274 . 1 204 . 0 1
2 2
+ + =
(101)


123
( ) | | 1 1 274 . 1 163 . 0
2
= m (102)

( ) | |
m
l l
K Re 2 sin 77 . 12 1
2
+ = (103)*
1
1
1
Re 43 . 0

|
.
|
\
|
=
k
n
d
d
x
C f (104)
) 2 ( sin 28 1
3
1
+ = C (105)
(
(
|
.
|
\
|

+ = 1 1
4
163 . 0 8 . 0
2
k
(106)
| |
m
l l
C Re ) 2 ( sin 28 1
3
+ = (107)*
(
(
|
.
|
\
|

= 1 1
4
163 . 0
2
m
(108)
* In [4.2], Re
x
is used, which is probably a misprint, since the average enhancement
factor cannot be calculated with the local Re.

Side Bounds
21 . 0
2 1
55 . 0
4 . 0 715 . 0
Pr Re 165 . 0

|
|
.
|
\
|
= D
T
T
Nu
f
w
d ST
(109)
11 . 0
1

=
C
F (110)
4 . 0
2

=
CD
F (111)

Local hydraulic losses for flow turn near the matrix wall -
m

57 . 0 2
48 . 1

=
C CD m
F F (112)

Index of Energy Efficiency
( )
( )
2
3
) 2 sin( 77 . 12 1
) 2 sin( 28 1
+
+
= =
l
l
K
C

(113)

Mechanisms of Heat Transfer and Friction
w
x
w
w
|
|
.
|
\
|
=

arctan
(114)
AV
x
x
w
w
w
,
,

=
(115)
76 . 0
18 . 1 tan

= (116)
= 30

( ) x
in
=

04 . 0 exp
,

(117)
3 . 0
,
=
in

(118)
1
,
9 . 0 1

+ = x w
x

(119)
= 45

( ) x
in
=

04 . 0 exp
,

(120)
124
43 . 0
,
=
in

(121)
1
,
2 . 1 1

+ = x w
x

(122)

Basic Section
( )
8 . 0
,
8 . 0
79 . 0
0 , ,
57 . 0 1
x x x
w Nu Nu
=
+ =

(123)
55 . 0
4 . 0 8 . 0
0 ,
Pr Re 0289 . 0

=
|
|
.
|
\
|
=
f
w
x x
T
T
Nu

(124)
( )
8 . 1
,
8 . 0
2
0
42 . 3 6 . 0
x
w
=
+ =

(125)
2 . 0
0
Re 43 . 0

=
=
x

(126)

125
Appendix C Correlations and Ranges of [4.5]
Correlations for the basic section as presented in Shukin and Nagoga [4.5] are
displayed below.
( ) | |
55 . 0
4 . 0 2
Pr Re 2 sin 77 . 12 1 0361 . 0

|
|
.
|
\
|
+ =
f
w n
x x
T
T
n Nu (127)
( ) | |
55 . 0
4 . 0 2
Pr Re 2 sin 77 . 12 1 0361 . 0

|
|
.
|
\
|
+ =
f
w n
l l
T
T
Nu (128)
(
(
|
.
|
\
|

+ = 1 1
4
2 . 0 1 8 . 0
2
n
(129)
( ) ( ) | |
8 . 0 2
Re 2 sin 77 . 12 1 8 . 0

+ =
n
x x
n (130)

( ) | |
8 . 0 2
Re 2 sin 77 . 12 1

+ =
n
l l
K (131)

| |
8 . 0 3
Re ) 2 ( sin 28 1

+ =
n
l l
C (132)

| |
1 3
Re ) 2 ( sin 28 1 43 . 0

+ =
n
l
f (133)

126
Appendix D Results for a Typical Turbine Blade
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
x 10
4
0
2000
4000
6000
8000
10000
12000
14000
Re
d
A
l
f
a
d
in Basic Section near Channel Inlet, x/d = 0.83
Nagoga average
Nagoga local
Shukin & Nagoga local
Shukin & Nagoga Average

Figure 26 Alfa in Example for Typical Turbine Blade.
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
x 10
4
1
1.5
2
2.5
3
3.5
4
Re
d
C
l
Friction Enhancement Factor C
l
as a Function of Re
d
, Basic Section
C
l
Nagoga
C
l
Shukin & Nagoga

Figure 27 C
l
in Example for Typical Turbine Blade, x/d = 0.83.

127
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
x 10
4
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
Re
d
K
Heat Transfer Enhancement K as a Function of Re
d
in Basic Section
Nagoga local
Nagoga average
Shukin & Nagoga local
Shukin & Nagoga average

Figure 28 K in Example for Typical Turbine Blade, x/d = 0.83.
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
x 10
4
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Re
d
f
Friction Factor as a Function of Re
d
for x/d=0.83
f Nagoga Basic
f Shukin & Nagoga Basic
f Nagoga Inital

Figure 29 f in Example for Typical Turbine Blade.
128
0 5 10 15 20 25
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000
6500
Alfa as a Function of x/d in Basic Section, Re=15000
x/d
A
l
f
a
Nagoga average
Nagoga local

Figure 30 Alfa in Example for Typical Turbine Blade.
0 5 10 15 20 25
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
2.3
K as a Function of x/d in Basic Section, Re=15000
x/d
N
u
s
s
e
l
t

N
u
m
b
e
r

E
n
h
a
n
c
e
m
e
n
t

F
a
c
t
o
r

K
K
average
K
local

Figure 31 K as a Function of x/d in Example for Typical Turbine Blade.

129
Appendix E Interesting Articles
Further information about matrix cooling can be found in the articles below.

Gorelov, Y. G. Ways of Further Enhancement
the Convective Cooling of
Vortex-Matrix Rotor Blades in
High-Temperature Turbines of
Gas Turbines Engines
Thermal Engineering,
Vol. 51, No 11, 2004

Acharya, Zhou,
Lagrone,
Mahmood &
Bunker
Latticework (Vortex) Cooling
Effectiveness Part 2: Rotating
Channel Experiments
Proceedings of ASME
Turbo Expo 2004,
Power for Land, Sea
and Air, June 14-17,
Vienna, Austria

Appendix 3 Impingement Cooling

131
132
Appendix A Correlations and Ranges
Correlations and ranges for impingement on flat respectively curved target plates are
presented below. The parameter Nu represents the average Nusselt number over an
impingement row.

Flat Target Plate
The average heat transfer over an impingement row in an array is calculated unless
other information is given.
Florschuetz, Truman & Metzger 1981 [5.7]
( ) ( ) | | { }
3 1
Pr 1 Re =
n
j c
m
j
G G d z B A Nu (134)
where
( ) ( ) ( )
z y x
n n
n
n
n
d z d y d x C m and B m A = , ,

Table 3 Parameters for equation ().
Inline Patterns Staggered Patterns
C n
x
n
y
n
z
C n
x
n
y
n
z
A 1.18 -0.944 -0.642 0.169 1.87 -0.771 -0.999 -0.257
m 0.612 0.059 0.032 -0.022 0.571 0.028 0.092 0.039
B 0.437 -0.095 -0.219 0.275 1.03 -0.243 -0.307 0.059
n 0.092 -0.005 0.599 1.04 0.442 0.098 -0.003 0.304
The standard error of deviations in the equation above is estimated to 5.6 % for the
inline pattern and 6.1 % for the staggered pattern.

A simpler, alternative correlation where m and n are independent of geometrical
parameters is described below. It has essentially the same confidence level as the
former equation.
( ) ( ) ( ) ( )
n
j c
n n
n
n
n
G G d z d y d x C Nu Nu
z y x
=1
1
(135)

( ) ( ) ( )
3 1 727 . 0 068 . 0 422 . 0 554 . 0
1
Pr Re 363 . 0 =

j n n
d z d y d x Nu (136)

Table 4 Parameters for the Equations Above.
C n
x
n
y
n
z
n
Inline 0.596 -0.103 -0.380 0.803 0.561
Staggered 1.07 -0.198 -0.406 0.788 0.660
Metzger & Korstad 1972 [5.10]
The correlation is valid for a single row of impingement.
338 . 0 049 . 0
Re 0822 . 0

=
d
M St (137)
Bailey & Bunker 2002 [5.14]
Tests were done with square inline arrays. Limiting case for a sparse array and dense
array are 3 4 and 20 26 number of jet rows, respectively. The first number refers
to the axial direction and the second to the lateral direction.

( ) ( ) ( ) ( )
( ) | |
( ) ( ) ( ) | | d z d x G G
d x
d z d z d x Nu
j c
d d

+
+ + =

28 7 . 13 2 . 61
Re 10 5 . 1 10 3 . 1 10 4 Re
3 . 2 3 . 7 5 . 5 1 . 47
8 4 3

(138)


133
Martin 1977 [5.11]
The correlation is valid for square arrays.

( )
3 2 42 . 0
Re
6 1
2 . 2 1
Pr
+

=
a d z
a
a K Nu
(139)

05 . 0
6
6 . 0
1

(
(
|
|
.
|
\
|
+ =
a
d z
K (140)

2
4
|
|
.
|
\
|
=
n
y
d
a

for a square inline array (141)

2
3 2
|
|
.
|
\
|
=
n
y
d
a

for a square staggered array (142)

Florschuetz & Isoda 1982 [5.21]
The correlations are valid for inline rectangular arrays.

Table 5. Correlations for Cd
|
.
|
\
|
d
z
d
y
d
x
, ,
Equation

Range of G
c
/G
j
(5,4,1)
85 . 0 =
d
C
0 to 0.63

( )
j c d
G G C =
> 0.63
(5,4,2)
85 . 0 =
d
C
0 to 0.83

( )
j c d
G G C = 08 . 1
> 0.83
(5,4,3)
85 . 0 =
d
C
0 to 0.90

( )
j c d
G G C = 11 . 1
> 0.90
(5,8,1)
80 . 0 =
d
C
0 to 0.54

( ) 893 . 0 169 . 0 + =
j c d
G G C
0.54 to 1.5

( )
j c d
G G C =
> 1.5
(10,4,1)
76 . 0 =
d
C
0 to 0.54

( ) 825 . 0 128 . 0 + =
j c d
G G C
0.54 to 1.8

( )
j c d
G G C =
> 1.8
( )
( )
602 . 0
806 . 0
6 . 1
+
=
j c
j c
G G
G G
(143)

c
f Re 24 = for Re
c
< 2000
(144)

25 . 0
Re 079 . 0
c
f = for 2000 < Re
c
< 30 000 (145)

134
20 . 0
Re 046 . 0
c
f = for Re
c
> 30 000 (146)

Obot & Trabold 1987 [5.13]
The correlations are valid for inline arrays. The flow scheme is defined in the left
picture in Figure 31. Individual differences of less than 10 % were obtained for 85 %
of the data, from 80 data points.

Figure 31Cross Flow Definitions (Left) and Values of Exponent n (Right) [5.13].

x
f
n
A
d
z
A Nu |
.
|
\
|
=
8 . 0
0
Re (147)
( )
( )
602 . 0
806 . 0
6 . 1
+
=
j c
j c
G G
G G
(148)
Table 6. Parameters for Obot and Trabold.

Flow Scheme A
0
n x
Minimum 0.863 see Figure 31 0.815
Intermediate 0.484 see Figure 31 0.676
Maximum 0.328 see Figure 31 0.595
Behbahani & Goldstein 1983 [5.15]
The correlation is valid for a staggered, square impingement array. Nusselt number
averaged over the holes array is calculated.

n
n
j
d
x
a Nu

|
.
|
\
|
=
78 . 0
Re
(149)
Table 7. Parameters for Behbahani and Goldstein.

z/d a n
2 0.0918 0.6262
3 0.0966 0.6610
4 0.10 0.6991
5 0.0954 0.7

135
Hglund 1999 [5.8]
The correlation is valid for a staggered array. The standard deviation was 10.2 % of
120 times 10 data points. All measured values were within 27 %. The mean error over
the surface was less than 8.5 %.
(
(
(
(
|
|
.
|
\
|
|
.
|
\
|
|
.
|
\
|
|
.
|
\
|
+
|
|
.
|
\
|
=
f
j
c
e
n
d c
n b
d
j w
c w a
c
G
G
d
z
d
x
C
d
z
d
x
B
T T
T T
A Nu 1 Re Re
(150)

|
|
.
|
\
|
|
|
.
|
\
|
+ + =
3 1
6
Re
10
20000 1 0055 . 0
c h
D
f

(151)

d s d
C Re 10 136 . 1 7112 . 0
6
,
= 0 < G
c
/G
j
< 0.45
(152)

2
3
,
10 379 . 1 1810 . 0 7265 . 0
|
|
.
|
\
|
+ =

j
c
j
c
s d
G
G
G
G
C 0.45 < G
c
/G
j
< 3 (153)

( )
s d d
C g C
,
1 1 =
(154)

d L e g
d L
+ + =

021 . 0 435 . 0 1
606 . 1
(155)
Table 8. Parameters for Hglund.
Total parameter
range
Parameter range
high
Parameter range
low
Geometric
parameter range
All geometries
tested
x
n
/d = 9.75
3.4 < z/d < 9.44

x
n
/d = 6.9
2.5 < z/d < 9.44

x
n
/d = 4.9
1.7 < z/d < 6.7
x
n
/d = 4.9
1.7 < z/d < 6.7

x
n
/d = 3.5
1.25 < z/d < 4.9

x
n
/d = 2.5
1.25 < z/d < 3.4
Range Re
d
25000< Re
d
<100 000 25000< Re
d
<100 000 25000< Re
d
<100 000
A 0.01 0.01 0.0081
B 0.230 0.316 0.219
C 0.60 0.50 0.65
a 0.80 0.80 0.80
b
d
Re 10 6 73 . 0
8
+

d
Re 10 6 73 . 0
8
+

d
Re 10 6 73 . 0
8
+

c -0.756 -0.960 -0.796
d -0.162 -0.165 -0.068
e 0.2 0.2 0
f 0.75 0.86 0.85

136
Wolfersdorf [5.23]
(
(
|
|
.
|
\
|
|
.
|
\
|
|
.
|
\
|
=

24 . 2
17 . 0 87 . 0
67 . 0
33 . 0 1 Re 505 . 0
j
c n
G
G
d
z
d
z
d
x
Nu
(156)

The ranges of Wolfersdorfs tests are taken from [5.8].

Chance 1974 [5.9]
The correlation is valid for a square staggered or inline array.

0146 . 1 3 1
3 2 1
Pr Re
f
n
d
A Nu =
(157)

0835 . 0
561 . 0
f
A n =
(158)

1
= 1.876 (cooling)
2.06 (heating)
(159)

c
I = 236 . 0 0 . 1
2

(160)

0835 . 0
3
561 . 0
f
A =
(161)

d
z
v
v
I
c c
c

=
0 0

(162)

Goldstein & Seol 1990 [5.16]
The correlation is valid for a single impingement row.

( )
4 1
09 . 0
8 . 22
9 . 2
d z
n
a
e
d z d y
Nu

+
=
(163)


137
Ranges:

Reynolds Number
0 10 20 30 40 50 60 70 80 90 100
Re*10
-3

Jet Plate-to-Target Plate distance
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
z/d

Streamwise Hole Spacing
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
x
n
/d

Note: In the correlation of Florshuetz et al. are Inline arrays valid for x
n
/d = 5-15 and
staggered arrays for x
n
/d = 5-10.

138
Spanwise Hole Spacing
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
y
n
/d

Florschuetz et al. 1981
Metzger & Korstad 1972
Bailey & Bunker 2002
Martin 1977
Florschuetz & Isoda 1982
Obot & Trabold 1987
Behbahani & Goldstein 1983
Hglund 1999
Wolfersdorf
Goldstein & Seol 1990
Turbine Components
Chance 1974
Combustor


139
Curved Target Plate
All correlations are valid for a single impingement row.

Chupp, Helmst, McFadden & Brown 1969 [5.1]
The average absolute deviation was 8.7 %.

2 . 1
27 . 1
6 . 0
7 . 0
Re 63 . 0
|
|
.
|
\
|

|
|
.
|
\
|
=
p n
D
d
y
d
d
z
p n
d
e
D
d
y
d
Nu
(164)

Hrycak 1980 [5.17]
402 . 0
16 . 0
63 . 0 33 . 0
Re Pr 72 . 0
|
|
.
|
\
|
|
.
|
\
|
=

p
n
d
D
d
d
y
Nu (165)

Metzger, Yamashita & Jenkins 1969 [5.18]

27 . 0
52 . 0
Re
2
355 . 0

|
.
|
\
|
=
b
b
l
St (166)

Nagoga 2000 [5.20]
55 . 0
4 . 0 75 . 0
25 . 0
Pr Re 17 . 0

|
|
.
|
\
|
|
.
|
\
|
=
f
w
S
T
T
R
z
R
S
Nu (167)
( ) 1 4
2

=
p n
p
N y
N d
S

(168)

Kopelev 1988 [5.22]
306 . 0
1
4 7 . 0
Re 0984 . 0
|
|
.
|
\
|
=
F
F
z
d
Nu
d
(169)

140
Metzger, Baltzer & Jenkins 1972 [5.19]
The uncertainty of the correlation varied from 3% to 8 %, based on the Kline and
McClintock method.
504 . 0
208 . 0
2
Re 200 . 0

|
.
|
\
|
=
b
l
St
b
for: (170)
67 . 6 = d y
n

( ) 37 2 / 6 d l
6500 Re 1400
b

410 . 0
290 . 0
2
Re 320 . 0

|
.
|
\
|
=
b
l
St
b
for: (171)
33 . 3 = d y
n

( ) 6 . 18 2 / 3 d l
6500 Re 1750
b

475 . 0
386 . 0
2
Re 870 . 0

|
.
|
\
|
=
b
l
St
b
for: (172)
67 . 1 = d y
n

( ) 14 2 / 2 d l
6500 Re 1750
b

434 . 0
392 . 0
2
Re 799 . 0

|
.
|
\
|
=
b
l
St
b
for: (173)
25 . 1 = d y
n

( ) 7 . 15 2 / 2 d l
6500 Re 1750
b


141
Reynolds Number
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Re*10
-3

Chupp et al. 1969
Hrycak 1980
Metzger et al. 1969
Nagoga 2000
Metzger et al 1972
Turbine

Jet Plate-to-Target Plate Distance
1 2 3 4 5 6 7 8 9 10 11 12 13
z/d

Spanwise Hole Spacing
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
y
n
/d

Target Plate Curvature
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
D
p
/d

142
Appendix B Flat Target Plate, Case B
2.5 3 3.5 4 4.5 5 5.5 6 6.5 7
x 10
4
80
100
120
140
160
180
200
220
240
260
Florschuetz, Truman and Metzger - Comparsion
Re
d
Florshuets et al.
Simplified Florshuetz et al.
Florshuetz et al, no crossflow

Figure 32 Comparison of Florschuetz Correlations for Case B, Inline Array.

0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
Re
d
N
u
Nusselt Number for an Inline Array
Florshuetz et al. 1981
Martin 1977
Woldersdorf
Chance 1974
Baileys & Bunker 2002
Obot & Trabold 1987
Goldstein & Seoul 1990

Figure 33 Nu for Inline Array, Case B.

143
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
Re
d
N
u
Nusselt Number for a Staggered Array
Martin 1977
Hglund 1999
Woldersdorf
Chance

Figure 34 Nu for Staggered Array, Case B.
1 2 3 4 5 6 7 8 9 10
x 10
4
0.76
0.78
0.8
0.82
0.84
0.86
Cd as a Function of Re
Re
d
C
d
Hglund 1999
Florschuetz & Isoda 1982

Figure 35 C
d
for Case B.
144
Appendix C Flat Target Plate, Components
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
Re
d
N
u
Martin 1977
Chance 1974
Obot & Trabold 1987

Figure 36 Nu for Vane 1, Back Insert, the Row Upstream.
Florschuetz lies between Martin and Baileys and Bunker in Figure 36.
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
Re
d
N
u
Martin 1977
Chance 1974
Obot & Trabold 1987

Figure 37 Nu for Vane 1, Back Insert, the Row Downstream.


145
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
350
Re
d
N
u
Martin 1977
Hglund 1999
Woldersdorf
Chance

Figure 38 Nu for a Staggered Array, Vane 1, Front Insert, 1 Row.
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
350
Re
d
N
u
Martin 1977
Woldersdorf
Chance 1974
Obot & Trabold 1987

Figure 39 Nu for Inline Array, Vane 1,Front Insert, 1 Row.

146
Appendix D Flat Target Plate, Random Tests
Table 9 Test 1.
z/d x
n
/d y
n
/d N
p
T
w
T
j
G
c
/G
j
1.2 3 3 8 1007 708 0.2

0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
350
400
Re
d
N
u
Martin 1977
Woldersdorf
Chance 1974
Obot & Trabold 1987

Figure 40 Nu for Inline Array.
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
350
400
Re
d
N
u
Martin 1977
Hglund 1999
Woldersdorf
Chance

Figure 41 Nu for Staggered Array.

147

Table 10 Test 2.
z/d x
n
/d y
n
/d N
p
T
w
T
j
G
c
/G
j
1.5 3.5 4 9 1007 708 0.4

0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
Re
d
N
u
Martin 1977
Woldersdorf
Chance 1974
Obot & Trabold 1987

0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
350
Re
d
N
u
Martin 1977
Hglund 1999
Woldersdorf
Chance


148
Table 11 Test 3.
z/d x
n
/d y
n
/d N
p
T
w
T
j
G
c
/G
j
1.5 5 8 20 1007 708 0.8
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
Re
d
N
u
Martin 1977
Woldersdorf
Chance 1974
Obot & Trabold 1987


0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
Re
d
N
u
Martin 1977
Hglund 1999
Woldersdorf
Chance



149
Appendix E Flat Target Plate, Effect of Parameters
Table 12 Effect of Various Paramers, Base Case.
z/d x
n
/d y
n
/d N
p
T
w
T
j
G
c
/G
j
2 5 4 15 1073 700 0.7

0.5 1 1.5 2 2.5 3 3.5 4
40
50
60
70
80
90
100
110
120
130
140
x
n
/y
n
N
u
Nu as a Function of x
n
/y
n
Martin 1977
Hglund 1999
Woldersdorf

Figure 46 Effect of Aspect Ratio.
0 2 4 6 8 10 12 14 16
40
60
80
100
120
140
160
180
z/d
N
u
Nu as a Function of z/d
Florshuetz et al 1981,no crossflow
Martin 1977
Hglund 1999
Woldersdorf,fel range
Chance 1974
Obot & Trabold 1987

Figure 47 Effect of Jet Target-to-Impingement Target Distance.
150
For the same case as above, but with zero cross flow, following results were obtained
for effect of z/d.
0 2 4 6 8 10 12 14 16
40
60
80
100
120
140
160
180
z/d
N
u
Florshuetz et al 1981,no crossflow
Martin 1977
Hglund 1999
Chance 1974
Obot & Trabold 1987

Figure 48 Effect of Jet Target-to-Impingement Target Distance without Cross Flow.
1 2 3 4 5 6 7 8 9 10
0
50
100
150
200
250
300
350
400
x
n
/d
N
u
Nu as a Function of x
n
/d
Martin 1977
Hglund 1999
Woldersdorf
Chance 1974
Obot & Trabold 1987

Figure 49 Effect of Streamwise Hole Spacing.

151
1 2 3 4 5 6 7 8 9 10
0
50
100
150
200
250
300
350
y
n
/d
N
u
Nu as a Function of y
n
/d
Martin 1977
Hglund 1999
Chance 1974
Obot & Trabold 1987

Figure 50 Effect of Spanwise Hole Spacing.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
-40
-20
0
20
40
60
80
100
120
140
160
Gc/Gj
N
u
Nu as a Function of G
c
/G
j
Martin 1977
Woldersdorf
Chance 1974
Obot & Trabold
Hglund 1999

Figure 51 Effect of Crossflow.

152
Appendix F Analysis of Correlations
Table 13 Analysis of Correlations.
Staggered array

Correlation: Martin Wolfersdorf
Chance
Hglund

Florschuetz
et al.

Bebhahani
Goldstein
Reference: [5.7] [5.11] [5.8], [5.23] [5.9] [5.15] [5.8]
Considered parameteres:

z/d yes yes yes
yes
yes yes
y
n
/d yes no(square) no yes yes no
x
n
/d yes yes yes yes no yes
T_diff no no no
no
no yes
jetplate
thickness no no no
no
no no
Relevant Range:

Re yes yes yes
yes
no yes
z/d yes partly yes
yes
partly yes
y
n
/d partly partly hardly
partly
partly partly
x
n
/d partly hardly hardly
partly
partly partly
D
p
/d

Sum:

yes 5 3 4 3 2 5
no/hardly 2 4 5 2 4 2
partly/? 2 2 0 4 3 2
sum: 9 9 9 9 9 9


153
Inline array
Correlation: Florschuetz Martin -77 Wolfersdorf Chance Bailey & Florshuetz Obot &
et al. -81 Bunker Isoda Trabold
Reference: [5.7] [5.11] [5.8], [5.23] [5.9] [5.14] [5.21] [5.13]
z/d yes yes yes yes yes yes yes
y
n
/d yes no(square) no yes no yes yes
x
n
/d yes yes yes yes yes yes yes
T_diff no no no no no no no
jetplate
thickness no no no no no no no
Relevant Range:
Re yes yes yes ? yes no no
z/d yes partly yes partly yes yes partly
y
n
/d partly partly hardly partly partly hardly partly
x
n
/d partly hardly hardly partly partly hardly partly
D
p
/d

Sum:
yes 5 3 4 3 4 4 3
no/hardly 2 4 5 2 3 5 2
partly/? 2 2 0 4 2 0 4
sum: 9 9 9 9 9 9 9

154

Single Row
Correlation:
Metzger &
Korstad
Goldstein
& Seol
Reference: [5.10] [5.16]
Re no partly
z/d hardly yes
y
n
/d partly hardly
x
n
/d
D
p
/d

Sum:
yes 2 3
no/hardly 4 3
partly/? 1 1
sum: 7 7


Curved target
plate
Correlation: Chupp Hrycak Metzger Nagoga Metzger Kopelev
etal.
et al.
1969
et al.
1972
Reference: [5.1] [5.17] [5.18] [5.20] [5.19] [5.22]

z/d yes no no yes no no
y
n
/d yes yes no no yes yes
x
n
/d
T_diff no no no yes no no
jetplate
thickness no no no no no no
D
p
/d yes yes yes yes yes yes
Re no hardly no yes no ?
z/d yes partly no yes no ?
y
n
/d hardly yes yes partly yes ?
x
n
/d
Dp/d yes no no partly no ?

Sum:
yes 5 3 2 5 3 2
no/hardly 4 5 7 2 6 3
partly/? 0 1 2 0 4
sum: 9 9 9 9 9 9


155
Appendix G Curved Target Plate, Components
Table 16 Curved Target Plate Component.
z/d y
n
/d N
p
T
w
T
j
2 4.5 20 1000 613

0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
50
100
150
200
250
300
350
Re
b
N
u
Nu as a Function of Re
b
Chupp et al. 1969
Hrycak
Q3D
Metzger et al. 1969
Nagoga
Kopelev 1988
Metzger et al. 1972

Figure 52 Nu for Vane 2, LE, Middle Row.

0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
x 10
4
Re
b
A
l
f
a
b
X: 2.8e+004
Y: 5280
Chupp et al. 1969
Hrycak
Q3D
Metzger et al. 1969
Nagoga
Kopelev 1988
Metzger et al. 1972

Figure 53 Alfa for Vane 2, LE, Middle Row.
156
Appendix H Curved Target Plate, Random Tests
Table 17 Test 1.
z/d y
n
/d N
p
T
w
T
j
1.25 2 20 900 600
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
0.5
1
1.5
2
2.5
x 10
4
Re
b
A
l
f
a
b
Chupp et al. 1969
Hrycak
Q3D
Metzger et al. 1969
Nagoga
Kopelev 1988
Metzger et al. 1972

Figure 54 Nu for a Random Test.

Table 18 Test 2.
z/d y
n
/d N
p
T
w
T
j
4 6 25 1200 600
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
2000
4000
6000
8000
10000
12000
14000
16000
Re
b
A
l
f
a
b
Chupp et al. 1969
Hrycak
Q3D
Metzger et al. 1969
Nagoga
Kopelev 1988
Metzger et al. 1972

Figure 55 Alfa for a Random Test.

157
Table 19 Test 3.
z/d y
n
/d N
p
T
w
T
j
2 5 12 1000 600
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
0.5
1
1.5
2
2.5
x 10
4
Re
b
A
l
f
a
b
Chupp et al. 1969
Hrycak
Q3D
Metzger et al. 1969
Nagoga
Kopelev 1988
Metzger et al. 1972


Table 20 Test 4.
z/d y
n
/d N
p
T
w
T
j
1 3 18 1100 550
0 1 2 3 4 5 6 7 8 9 10
x 10
4
0
0.5
1
1.5
2
2.5
3
x 10
4
Re
b
A
l
f
a
b
Chupp et al. 1969
Hrycak
Q3D
Metzger et al. 1969
Nagoga
Kopelev 1988
Metzger et al. 1972

158
Appendix I Curved Target Plate, Effect of Parameters
Table 21 Base Case for Test of Effect of Different Parameters.
z/d y
n
/d N
p
T
w
T
j
D
p
/d
2 4 15 1000 600 10
1 2 3 4 5 6 7 8 9 10
5
10
15
20
25
30
35
40
45
50
55
60
y
n
/d
N
u
Nu as a Function of y
n
/d
Chupp et al. 1969
Hrycak 1980
Metzger,Yamashita & Jenkins 1969
Nagoga 2000
Metzger, Baltzer & Jenkins 1972
Kopelev 1988

Figure 58 Effect of Hole Spacing for Case D.
0 2 4 6 8 10 12 14 16 18 20
0
10
20
30
40
50
60
70
z/d
N
u
Chupp et al. 1969
Hrycak 1980
Nagoga 2000
Kopelev 1988

Figure 59 Effect of Jet Plate-to-Target Plate Distance for Case D.


159
0 5 10 15 20 25 30 35 40
0
10
20
30
40
50
60
70
D
p
/d
N
u
Nu as a Function of D
p
/d
Chupp et al. 1969
Hrycak 1980
Nagoga 2000
Kopelev 1988

Figure 60 Effect of Target Plate Curvature for Case D. Florschuetz et al. is also Displayed for
Comparison with a Correlation for Inline Arrays and Flat Target Plate, x
n
/d = 5, G
c
/G
j
= 0.5.

160
Appendix J Interesting Articles
Further information about impingement cooling can be found in the articles below.
Persson Experimental Investigation of
Impingement Cooling with
Turbulators or Surface
Enlarging Elements
Thesis Work at Siemens
2000

Van Treuren, Wang,
Ireland, Jones &
Kohler
Comparison and Prediction of
Local and Average Heat
Transfer Coefficients under
and Array of Inline and
Staggered Impinging Jets
ASME 96-GT-163

Hollworth & Berry Heat Transfer from Arrays of
Impinging Jets with Large Jet-
to-Jet Spacing
May 1978, vol 100

Tabakoff & Clevenger Gas Turbine Blade Heat
Transfer Augmentation by
Impingement of Air Jets
Having Various
Configurations
Journal of Engineering for
Power, January 1972

Zuckerman & Lior Impingement Heat Transfer:
Correlations & Numerical
Modeling
Vol. 127, May 2005

Huber & Viskanta Effect of Jet-Jet Spacing on
Convective Heat Transfer to
Confined , Impinging Arrays
of Axisymmetric Air Jets
Int. J. Heat Mass Transfer,
Vol. 37, No 18, 1994

Gillespie, Wang,
Ireland & Kohler
Full Surface Local Heat
Transfer Coefficient
Measurements in a Model of
an Integrally Cast
Impingement Cooling
Geometry
Journal of Turbomachinery,
Vol. 120, January 1988

San & Lai Optimum Jet-to-Jet Spacing
of Heat Tranfer for Staggered
Arrays of Impinging Air Jets
Int. J. Heat Mass Transfer,
44, 2001

Gau & Chung Surface Curvature Effect on
Slot-Air-Jet Impingement
Cooling Flow and Heat
Transfer Process
Vol. 113, Novemeber 1991

Siemens Sodenberg

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Siemens Sodenberg

Uploaded by

Copyright:

Available Formats

Heat Transfer Correlations for Gas Turbine Cooling

was calculated and included in the comparisons. Also,

K between the models does

K is found in Table 1 (60)

K (SR) 1.00 2.20 3.40 2.90 1.25

K (SSR) - 4.4 3.9 2.6 1.3

K is found in Table 1 (71)

You might also like