Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Heat transfer characteristics in film cooling applications Licu, Dragos N. 1998

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1998-345815.pdf [ 13.02MB ]
Metadata
JSON: 831-1.0080946.json
JSON-LD: 831-1.0080946-ld.json
RDF/XML (Pretty): 831-1.0080946-rdf.xml
RDF/JSON: 831-1.0080946-rdf.json
Turtle: 831-1.0080946-turtle.txt
N-Triples: 831-1.0080946-rdf-ntriples.txt
Original Record: 831-1.0080946-source.json
Full Text
831-1.0080946-fulltext.txt
Citation
831-1.0080946.ris

Full Text

H E A T T R A N S F E R C H A R A C T E R I S T I C S IN F I L M  COOLING  APPLICATIONS By Dragos N. Lieu M. Sc. (Faculty of Aircraft) Polytechnic Institute of Bucharest  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY  in THE FACULTY OF GRADUATE STUDIES MECHANICAL ENGINEERING We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA October 1998 © Dragos N. Lieu, 1998  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission.  Mechanical Engineering The University of British Columbia 2324 Main Mall Vancouver, Canada V6T 1Z4  Date:  Abstract The leading edge region of gas turbine blades and vanes experiences high thermal and mechanical stresses and has to be properly cooled. E x t e r n a l cooling of the leading edge region is typically achieved by a film cooling technique. A n investigation into the film cooling effectiveness of three different large scale leading edge geometries is presented i n this study. One of the geometries investigated represents an original design and is an example of an improved film cooling layout. A l l geometries used have four rows of cooling holes placed symmetrically about the geometrical leading edge, but the layout of the cooling holes is different from one leading edge geometry to another. A broad range of variables is considered including mass flow ratio, coolant density, and jet Reynolds number. F i l m cooling effectiveness measurements were made i n a low speed wind tunnel environment using a flame ionization detector technique and the mass/heat transfer analogy. These measurements significantly extend the insight into the effects of hole geometry on the film cooling characteristics of the leading edge of turbine blades and provide new data for design purposes. The effect of geometry is more important for the case of double row injection where spanwise-averaged film cooling effectiveness is improved by the use of compound angle holes. The spanwise-averaged film cooling effectiveness is higher at lower mass flow ratios and decreases typically as the mass flow ratio increases. A t higher mass flow ratios, the newly designed leading edge geometry produces higher spanwise-averaged film cooling effectiveness than the other two geometries investigated thus providing the necessary backflow margin at operating conditions more relevant to gas turbine use. For the case of single row injection, the effects of geometry scale reasonably well when the  ii  local mass flow ratio is used i n the analysis of the spanwise-averaged film cooling effectiveness immediately downstream of the injection holes. T h e local momentum flux ratio is a more appropriate scaling parameter when coolants w i t h different densities are used. A film cooling effectiveness correlation was also developed for one of the geometries investigated based on an area-averaged film cooling effectiveness and on a newly defined blowing parameter. This correlation accounts implicitly for the particular geometrical layout used and explicitly for the m a i n injection parameters investigated. T h e results can be now more directly used i n existing design procedures. A new experimental technique based on wide-band liquid crystal thermography and transient one-dimensional heat conduction has been developed and implemented. T h e technique combines a real-time, true colour imaging system w i t h the use of a wide-band liquid crystal and multiple event sampling for the simultaneous determination of the film cooling effectiveness and heat transfer coefficient from one transient test. A comparison of different image capture techniques is also presented and computer codes are developed for data processing. For a test case of compound angle square jets i n a crossflow, very good agreement was obtained between the film cooling effectiveness calculated from the transient heat transfer experiments and the film cooling effectiveness measured i n isothermal mass transfer experiments using a flame ionization detector technique. This new approach has been developed as a major part of this thesis and represents a significant contribution to the use of liquid crystal thermography i n film cooling applications.  iii  Contents Abstract  ii  List of Figures  viii  List of Symbols  xi  Acknowledgements  xv  1  2  Introduction  1  1.1  Generalities  1  1.2  Background  2  1.3  Structural and T h e r m a l Considerations  4  1.4  Cooling Schemes  6  1.4.1  Overview  6  1.4.2  Leading Edge Cooling  7  1.5  Concepts and Definitions  9  1.6  Literature Review  13  1.7  Objectives and Scope  18  Measurements of Leading Edge Film Cooling Effectiveness Using a Flame Ionization Detector  21  2.1  E x p e r i m e n t a l Arrangements  22  2.1.1  22  M o d e l and W i n d Tunnel Setup  iv  2.2  2.1.2  Baseline Leading Edge Geometry  23  2.1.3  C o m p o u n d Angle - C A Geometry  24  2.1.4  Improved Leading Edge Geometry  26  2.1.4.1  Background Information  26  2.1.4.2  P V C Geometry  28  E x p e r i m e n t a l Procedures  30  2.2.1  The Heat and Mass Transfer Analogy  30  2.2.2  The F l a m e Ionization Detector  31  2.2.3  F I D Calibration  2.2.4 2.3  2.4  .  32  2.2.3.1  Calibration of the A i r - A i r M i x t u r e  32  2.2.3.2  Calibration of the A i r - C 0 M i x t u r e  33  2  T h e Sampling Procedure  36  E x p e r i m e n t a l Results  37  2.3.1  Baseline Geometry  38  2.3.1.1  Double R o w Injection  38  2.3.1.2  F i l m Cooling Effectiveness Correlation - Double R o w Injection 43  2.3.1.3  Single R o w Injection - Constant Jet Reynolds N u m b e r  2.3.1.4  Single R o w Injection - The Effect of Jet Reynolds N u m b e r .  . . .  45 53  2.3.2  C A Geometry  56  2.3.3  P V C Geometry  58  2.3.3.1  Double R o w Injection  58  2.3.3.2  Single R o w Injection  64  Summary  66  Liquid Crystal Thermography  68  3.1  Thermochromic L i q u i d Crystals  68  3.2  Imaging System  75  3.2.1  Illumination Sources  75  3.2.2  Image Acquisition  79  3.2.3  3.2.4  4  5  3.2.2.1  Colour Photography  79  3.2.2.2  V i d e o Capturing  81  Recording and Storage Devices  85  3.2.3.1  Video Tape Recorders - V i d e o Tapes  86  3.2.3.2  Digitizers - H a r d Drives  88  Imaging System Specifications  92  3.3  Tristimulus Colour Space  93  3.4  Calibration of the L i q u i d C r y s t a l  97  3.5  Summary  102  Compound Angle Jets in Crossflow  103  4.1  Transient Heat Transfer  103  4.2  Jet Temperature D a t a Acquisition  109  4.3  W i n d Tunnel Setup  110  4.4  Solution Algorithms  114  4.5  The T i m e Event Pairs  117  4.6  Experimental Results  119  4.6.1  General Trends  120  4.6.2  Mass F l o w R a t i o M = 0 . 5  123  4.6.3  Mass Flow Ratio M = 1 . 0  127  4.6.4  Mass F l o w Ratio M = 1 . 5  130  4.7  Uncertainty Analysis  133  4.8  Summary  140  Conclusions and Recommendations  143  5.1  Overall Summary  143  5.2  Leading Edge F I D Measurements  145  5.2.1  Conclusions  145  5.2.2  Recommendations  147  vi  5.3  L i q u i d C r y s t a l Thermography  148  5.3.1  Conclusions  148  5.3.2  Recommendations  149  A Colour Space Conversion  151  B Wind Tunnel Setup  153  Bibliography  154  vii  List of Figures 1.1  Two-spool turbofan engine schematic  3  1.2  F i l m cooled airfoil section  8  2.1  W i n d tunnel turbine blade model  23  2.2  P l e n u m arrangement and removable leading edge insert  24  2.3  Baseline geometry  25  2.4  Geometrical nomenclature  27  2.5  P V C geometry - outside view  29  2.6  Mass fraction calibration for the air-(702 mixture  35  2.7  Double row injection arrangement  38  2.8  Baseline geometry: air, double row injection, periodicity check  39  2.9  Baseline geometry: air and CO2, double row injection  41  2.10 Baseline geometry: air and CO2, double row injection, / = 0.6  42  2.11 F i l m cooling effectiveness correlation: baseline geometry, double row injection.  45  2.12 Single row injection arrangement  47  2.13 Baseline geometry - row 6\ cooling: effect of mass flow ratio  48  2.14 Baseline geometry - row 62 cooling: effect of mass flow ratio  48  2.15 Baseline geometry - row $i cooling: effect of local momentum flux ratio. . . .  50  2.16 Baseline geometry - row 0 cooling: effect of local momentum flux ratio. . . .  50  2  2.17 Baseline geometry - single row air injection: effect of local mass flow ratio.  .  51  2.18 Baseline geometry - single row CO2 injection: effect of local mass flow ratio.  52  2.19 Baseline geometry - row 0\ air cooling: x/(7rD/A) = 0.32  54  viii  2.20 Baseline geometry - row 6\ air cooling: x/(irD/4) — 0.64 2.21 Baseline geometry - row 62 air cooling: x/(irD/4:) = 0.64 2.22 Baseline geometry - row 9 air cooling: X/(-KD/4) 2  = 1.27  54 55 55  2.23 Spanwise averaged 77 for baseline and C A geometries  57  2.24 Comparison of baseline and C A geometries: fj, double row air injection. . . .  58  2.25 Injection arrangement for the P V C geometry  60  2.26 Comparison of P V C and Baseline geometries: air, double row injection. . . .  61  2.27 Comparison of P V C and C A geometries: air, double row injection  61  2.28 P V C and Baseline geometries: air, double row cooling at M = 1.0  62  2.29 P V C , Baseline, and C A geometries - fj: air, double row cooling  63  2.30 P V C geometry: single row cooling, effect of M  64  2.31 P V C geometry: single row cooling;f effect of Me  65  3.1  Effect of the viewing angle on the observed colour of a chiral nematic liquid crystal  72  3.2  Colour/temperature response of a thermochromic liquid crystal  73  3.3  Spectrum of electromagnetic radiation  76  3.4  B l a c k b o d y radiation  77  3.5  O p t i m a r c lamp technical specifications  78  3.6  T y p i c a l histograms and statistics for colour photography  80  3.7  Charge-coupled device - C C D  82  3.8  Colour separation optical system  85  3.9  T y p i c a l histograms and statistics for S - V H S tape recorded images  88  3.10 T y p i c a l histograms and statistics for S - V H S P A R - d i s k recorded images. . . .  90  3.11 B l o c k diagram of the imaging system  93  3.12 A d d i t i v e system with R e d , Green, and B l u e primaries  94  3.13 T h e double hexcone H S L colour space  96  3.14 T L C calibration setup  98  3.15 T L C colour response w i t h temperature  99  ix  3.16 Statistics for real-time true colour recording  100  3.17 T L C calibration accuracy  101  4.1  General three-temperature arrangement - nomenclature  104  4.2  W i n d tunnel arrangement  112  4.3  Non-dimensional wall temperature profiles  118  4.4  Contours of film cooling effectiveness  121  4.5  Contours of heat transfer coefficient - Stanton number  122  4.6  Comparison of 77 obtained by the T L C technique w i t h the F I D data of F i n d lay [17], M = 0 . 5  124  4.7  T i m e history of h and 77, M = 0 . 5  125  4.8  Comparison of 77 obtained by the T L C technique w i t h the F I D data of F i n d -  f  lay [17], M = 1 . 0 4.9  128  T i m e history of h and 77, M = 1 . 0  129  f  4.10 Comparison of 77 obtained by the T L C technique w i t h the F I D data of F i n d lay [17], M = 1 . 5  131  4.11 T i m e history of h and 77, M = 1 . 5  132  4.12 M a p of the uncertainty i n film cooling effectiveness data  138  4.13 M a p of the uncertainty i n heat transfer coefficient data  139  f  x  List of Symbols C  Concentration, mass fraction  C  Pressure coefficient ( = f  D  Diameter of the leading edge  DAB  Mass diffusivity  E  F l a m e ionization detector output voltage  Gr  Grashof number ^=  I  Jet-to-crossflow momentum flux ratio ^=  le  L o c a l momentum flux ratio ^— y^fa^  L  Characteristic length  Le  Lewis number (= -I - = -FT—)  M M$  Jet-to-crossflow mass flow ratio ( = ) \ Poo Coo / L o c a l mass flow ratio ^= ^j ^j  N  N u m b e r of data points i n a sample  Pr  P r a n d t l number (= ^)  Q  F l o w rate  Re  Reynolds number (=  S  Spanwise distance covered by the sampling probes  S  Spanwise spacing of holes i n adjacent rows  SLE  Spanwise spacing of holes i n adjacent rows about the stagnation line  5y  Spanwise spacing of holes i n a row  Sc  Schmidt number (= -FT—)  v  ar  a L  ^  9 -  )  [Jr~ °°^ T  for a perfect gas ^ P 2 P  ^2  ^  2  v  Fr  u ' AB  e  ^)  XI  St  Stanton number ( = —fr-)  T  Temperature  U  M e a n velocity component  C, Cp  Specific heat  d  Diameter of the jet; size of a square jet  9  Gravitational acceleration; calibration function, a i r - C O ^ mixture  h'  Heat transfer coefficient i n a two-temperature situation - no cooling jets  ti  Heat transfer coefficient i n a two-temperature situation - w i t h cooling jets  hf  F i l m heat transfer coefficient  Kad  Radiation heat transfer coefficient (= eo-(T + T ,)(T^ +  k  T h e r m a l conductivity  I  Length of the cooling holes  q  Heat rate  t  T i m e ; thickness of the leading edge  X  Streamwise coordinate  y  Spanwise coordinate  z  Coordinate perpendicular to the surface  0  w  oc  Greek A  F i n i t e change h t 2  1—exp  0  Calculated dimensionless temperature 1—exp  f  a  pek  erfc  pck  erfc  *  Measured dimensionless temperature ( = r"[tt)-r~)  a  T h e r m a l diffusivity; angle on the 0 curve  8  Angle measured w i t h a spanwise line on the surface; incremental change  ST  Temperature increment ( = T (t,z) — T^) w  Measurement uncertainty; emissivity  xii  1  (= ^~^ — c ZcZ) T  C  •n  A d i a b a t i c film cooling effectiveness  fj  Spanwise-averaged film cooling effectiveness.  fj  Area-averaged film cooling effectiveness  8  Angle measured from the geometrical leading edge  A  Wavelength  v  K i n e m a t i c viscosity (momentum diffusivity)  2  _  + [^]  i  i  Blowing parameter ( = ^  ^  - l)  p  Density  a  Standard deviation; Stefan-Boltzmann constant  r  T i m e increment  ip  Angle measured with a streamwise line through the hole centre  Subscripts 0  Refers to default output voltage  1  Refers to row of holes closest to stagnation line  2  Refers to coolant conditions; row of holes farther from stagnation line  H  Refers to the Hue colour component  L  Refers to a characteristic length  a, b  Refer to time events  aw  A d i a b a t i c wall  i  Refers to initial conditions; index for data point i n a sample  cond  Refers to heat conduction  conv  Refers to heat convection  rad  Refers to heat radiation  ref  Reference conditions  w  Refers to wall conditions  77  Refers to film cooling effectiveness  xiii  00  Refers to free stream conditions  Superscripts 1  Index for the sampling probes  xiv  Acknowledgements I first of a l l wish to thank m y research supervisors, D r . M a r t h a Salcudean and D r . Ian S. Gartshore, for their enduring support throughout this undertaking. T h e financial contribution of P r a t t & W h i t n e y Canada and N S E R C is fully appreciated. I also wish to thank M r . Russell B o o t h and M r . Steve Giffin from Thermographic Measurements L t d . for manufacturing and generously supplying the liquid crystal used i n this study. T h e assistance of J o h n Inch from Powertech Labs Inc. w i t h special coatings tools and equipment is gratefully recognized. I was also priviledged to share m y time at U B C w i t h a great group of people, and I thank m y colleagues D r . M a t t h e w Findlay, M i c h a e l Savage, Petros Lappas, Sidney C h a n , and D r . Peter M t u i for an unforgettable experience. Last, but not least, I thank m y immediate family for their encouragement and patiance during the completion of this work.  xv  Chapter 1 Introduction 1.1  Generalities  Gas turbine technology has had a tremendous impact on the development of modern society, beginning w i t h the first practical applications as early as 1939.  T h a t year, a 4 M W gas  turbine power generating unit was installed at Neuchatel, Switzerland, and the first aircraft powered by a turbojet flew i n Germany [29] [33]. In 1995, 1.29 billion people used gas turbine powered airplanes as means of travel, and the forecast of the International A i r Transport Association ( I A T A ) predicts that worldwide air travel will more than double over the next 20 years, growing at an annual rate of at least 6%. A i r cargo is also projected to grow at over 6% annually [3]. W i t h the deregulation of the electric utility industry, gas turbines are also becoming a key player i n the electric power generation. T h e combined cycle gas turbine power plants operate today at thermal efficiencies very close to 60%, w i t h simple cycle gas turbines operating at more than 40% thermal efficiency; they have come a long way i n the span of seven decades from the 18% efficiency of the first electric power gas turbine [29]. Such performance makes the gas turbine increasingly important i n electric power generation. A t the heart of such a power plant is a gas turbine driving an electric generator w i t h the hot exhaust gas being used i n a heat recovery steam generator. A steam turbine coupled w i t h an electric generator  1  can be added to the system, thus providing additional electricity, or the generated  steam  can be used as such for various utility processes. The term cogeneration is widely used for such an arrangement. Other major applications where aeroderivative gas turbines are used are marine propulsion, pumping i n oil production platforms and for natural gas pipelines. Gas turbines are particularly attractive for use i n offshore or remote oil and gas fields because they can use fuel directly from the source, deliver high power per unit weight, and typically require low maintenance. W i t h such a wide range of applications, the gas turbine industry has always been highly competitive. It encompasses research and development i n many different engineering fields including aerodynamics, heat transfer, material sciences, structures and vibrations, cooling techniques, control theory, and instrumentation technologies. Aircraft propulsion remains the most exciting and demanding application of gas turbines.  T h e combination of hos-  tile thermal environment, high material stresses, and differential transient thermal stresses makes the design for long life extremely complex and difficult. N e w designs, materials, and manufacturing techniques have emerged over the last decades, steadily increasing the performance of the aircraft gas turbines. Current research programs a i m to double the power of military aircraft engines by the end of the century while increasing reliability, and reducing repair costs and maintenance downtime [50]. The goals set by N A S A for civil aviation are equally challenging, aiming to reduce the cost of air travel and travel time by a factor of two, emissions by a factor of five, and noise levels by a factor of four, a l l within the next 20 years [43].  1.2  Background  Gas turbines are air-breathing machines, operating on the basis of the B r a y t o n cycle. A l though the detailed layout design is different for land based units than for aircraft propulsion units, the principles of operation of the gas generator are the same for both applications.  2  Bypass nozzle  Figure 1.1: Two-spool turbofan engine schematic T h e major components of the gas generator are the inlet, compressor, combustor, turbine, and nozzle. A more complex layout is the two-spool turbofan engine schematic presented as an example i n F i g . 1.1, configuration widely used for aircraft propulsion. A i r from the atmosphere is guided through an aerodynamic shaped inlet into a first fan stage. Here, the air stream is divided into a core flow and an outer flow, or bypass flow. T h e core air flow i n further compressed i n a multistage compressor. T h e compressed core air is then introduced into a combustion chamber where fuel is injected and the mixture undergoes a quasi-isobaric combustion. T h e core hot gas expands i n a multistage turbine producing shaft work to drive the compressor and fan, and then further accelerates i n a nozzle. T h e outer stream typically flows around the core engine casing. T h e forward thrust is then computed as a summation of the jet propulsive forces generated from each air stream. A l l components have to be designed for m a x i m u m efficiency and they also function i n an integrated manner. M a t c h i n g of so many different component characteristics at off-design operation is one of the most challenging problems i n engineering. It is well understood that a modification to any of the components will affect the overall performance of the engine. Detailed analysis of various aircraft gas turbine cycles has been done elsewhere [23], [40]. E v e n  3  the simplest analysis accounts for the interaction of a large number of design parameters [10]. T h e large commercial aircraft engines use higher bypass ratios for higher propulsion efficiency, higher cycle pressure ratios ( C P R ) and higher turbine entry temperatures ( T E T ) for increased thermal efficiency [10], [23], [40]. H i g h propulsion efficiency is thus achieved by increasing the total engine airflow for a given net thrust, while high thermal efficiency is largely associated w i t h the performance of the high pressure compressor and turbine system. A test study done by Rolls Royce for the development of the R B 2 1 1 three-spool turbofan engine [54], showed that an increase of 20% i n T E T alone represents only a 5% reduction i n the specific fuel consumption ( S F C ) , at fixed component efficiency. Further reduction i n S F C can only be achieved by increasing both T E T and C P R . Considering that no loss of component efficiency is acceptable, this translates into a serious design and technological challenge.  1.3  Structural and Thermal Considerations  T h e need to increase the T E T for high thermal efficiency has led to intensive research i n the field of material sciences and cooling techniques. T h e steady increase of the T E T was made possible by the development of new materials, coatings, and improved cooling systems. H i g h pressure turbine blades, especially the first stage nozzle guide vanes and rotor blades, are experiencing high temperatures (1600 K to 1800 K are not uncommon at the entrance to the first stage vanes), and high structural loads (rotor blades i n particular). T h e blades are basically immersed i n the hot gases and receive heat from these gases. These blades typically operate very close to their melting point temperature, and sometimes there are regions on the blades exposed to temperatures higher than their melting point. In order to maintain the structural integrity and fatigue life of the blades, heat has to be removed by means of cooling of the blades. A i r bled from one of the compressor stages is typically used for cooling the blades. T h e blade material used i n advanced gas turbines is clearly of crucial importance. Single  4  crystal alloys are now used for the high pressure turbine blades together w i t h advanced thermal barrier coatings, thus providing the necessary strength, oxidation and corrosion resistant properties needed for extracting the energy from the very hostile environment. A rotor blade experiences large centrifugal forces decreasing from the blade root to the blade t i p . In addition, there are bending forces, generated by the expansion of the hot gases, which create bending moments about the blade root. For these reasons the rotor blades are typically larger at the root and progressively thinner towards the t i p . T h e failure of a turbine rotor blade can easily produce a very serious event; the broken pieces can easily penetrate the engine casing and cut through fuel or oil pipes. T h e design must therefore produce blades which will be highly reliable over their predicted lifetime. The nozzle guide vanes (stator blades) experience less structural stress than the rotor blades. T h e y are clamped at both ends, so the stresses on the vane platform sections will be less than for the rotor blade root sections. There will be radial stresses on the vanes, since they are integrated parts of the engine structural support system. T h e loads generated on the engine shafts axe usually transferred through flanges to the nozzle guide vanes, and from here to the engine mounting pins and thence to the aircraft structure. F r o m a thermal point of view, the stators have to withstand the hottest combustor streak, w i t h a non-uniform gas temperature profile, both radially and circumferentially, at the combustor exit. T h e m a x i m u m spanwise gas temperature distribution tends to peak at about half the blade span leading edge. T h e exact location of the m a x i m u m temperature on the blade surface will vary w i t h different engine operating conditions. The thermal impact on the rotor blades is typically less severe than on the stator blades. The rotation of the rotor assembly leads to a relative temperature of the rotor blade surface less than that experienced by the nozzle guide vanes. T h e r m a l stresses i n the blade material, for both stator and rotor blades, are also caused by transient differential temperature between the blade core and the blade leading edge during take-off, landing, or sudden acceleration/deceleration maneuvers. These stresses can lead to low-cycle fatigue ( L C F ) i n the blade material, and specific research has been conducted to  5  understand and predict this phenomenon [22]. T h e combination of the described mechanical and thermal loads on a turbine blade identifies the mid-span leading edge and the trailing edge regions as being critical for the integrity of the blade over its operational lifetime. Special attention is focused i n the design process on reducing the stresses i n these regions. Advanced cooling techniques are now used to reduce the thermal stresses everywhere i n the blade and particularly i n the leading edge and trailing edge regions.  1.4 1.4.1  Cooling Schemes Overview  For the high T E T used i n today's aircraft engines, the nozzle guide vanes and rotor blades at the forward portion of the turbine need to be cooled i n order to preserve the structural integrity and fatigue life of the airfoil. A modern cooling scheme employs b o t h internal and external cooling of the blades and vanes. Cooling air is fed from the root of the rotor blades through interior passages to remove as much heat from the walls as possible, and discharged through discrete holes placed at predetermined locations on the airfoil surface. In the case of the stator blades, the cooling air can be introduced at both ends of the vanes, flowed through interior passages and ultimately discharged through the film cooling holes onto the airfoil surface. T h e cooling air, ejected from the film cooling holes, extends over the airfoil surface to provide a protective film layer which impedes the hot gases from contacting the surface of the airfoil, thus decreasing the heat transfer to the airfoil. This is known as the film cooling technique. T h e interior cooling passages have to be designed to have high heat transfer coefficients, so that the cooling air can remove the m a x i m u m heat from the blade. A t the same time, the pressure drop i n the interior channels should be minimized to allow for the cooling air to be ejected onto the blade surface i n regions of adverse pressure gradients, as is the case for the leading edge and the pressure side surface. T h e film cooling holes, on the other hand, have  6  to be designed to generate an effective film layer to reduce the external heat transfer rates. T h e amount of the cooling air is limited from considerations of gas turbine cycle thermodynamic efficiency. M i x i n g of the film cooling jets w i t h the m a i n flow causes b o t h stagnation temperature and pressure losses. Therefore, there is an incentive to use as m u c h internal cooling as possible and to minimize the amount of disturbance to the m a i n external gas flow from the cooling jets. There are also losses due to the heat extraction from the blade walls. Since the air used to cool the blades is bled from one of the compressor stages, there is the associated lost work i f this air is not used i n a combustion process. T h e use of high pressure cooling air cannot be avoided, i n particular for cooling the leading edge region of blades and vanes. T h e greater the amount of such high pressure cooling air that is needed, the higher the associated losses i n available work will be. Accordingly, for each T E T there is a determined amount of usable cooling air beyond which the benefits of increased T E T are lost. Such analyses have been done i n [14], [31]. It was found that the use of film cooling improves performance significantly, and that an increase of the overall cycle efficiency is directly related to an increase i n blade operating temperatures. T h e key to improved efficiency lies therefore i n the use of advanced high temperature materials, to increase the allowable blade operating temperatures, and more efficient internal and external cooling schemes, to reduce the cooling flow for given T E T and blade operating temperature.  A reduction by a factor of two of the cooling flow w i t h i n the next 20 years,  down to 10% from the 23% of the 1996 baseline engine, is actually one of the goals set for future large subsonic engines [43].  1.4.2  Leading Edge Cooling  Cooling of the leading edge of blades and vanes is a challenging problem. This region of the blade experiences high thermal and mechanical stresses and needs to be properly cooled. T h e interior of the airfoil leading edge typically shows a designated passage that behaves as a common plenum for the film cooling holes. This passage can be abutted by an impingement plate so that the cooling air, flowing through the interior passages, can be discharged into  7  Figure 1.2: F i l m cooled airfoil section the leading edge cooling passage through impingement onto the inner leading edge wall. T h i s technique increases the heat transfer at the interior of the leading edge wall, thus removing more heat from this region. Such an arrangement is shown i n F i g . 1.2. T h e leading edge region cannot be adequately cooled by internal cooling only.  The  pressurized air from the leading edge passage is also ejected into the main hot gas boundary layer through discrete cooling holes, often arranged i n multiple rows on either side of the stagnation line. T h e leading edge wall is thus cooled by convection through the film cooling holes as well. T h e film layer has to protect the airfoil for some region downstream of the cooling holes. T h e leading edge film cooling holes are placed i n the immediate proximity of the aerodynamic stagnation line. H i g h pressure cooling air is required i n the leading edge passage to make possible the ejection of the cooling air into the high pressure mainstream. In order to provide good cooling of the airfoil surface while minimizing the amount of cooling air used, the designer has to consider a l l the geometrical and flow field parameters of importance for the cooling process. This is a difficult task.  8  1.5  Concepts a n dDefinitions  In a gas turbine environment, the m a x i m u m allowable metal temperatures are known and set by the appropriate selection of materials and coatings. T h e m a x i m u m hot gas temperatures could be set by considerations of high thermal efficiency and long service life of the high pressure turbine vanes and blades. There is a compromise between high turbine i n let temperatures a n d cooling air requirements, since the efficiency and thrust are reduced w i t h increasing amounts of cooling air. T h e design requirement could then be to determine the m i n i m u m coolant flow and the related heat transfer rates to maintain the metal temperatures below critical levels. T h e internal cooling system must take up the external heat transfer rates i n their entirety, but the following analysis focuses on the external cooling only.  T h e general formulation of the convective heat transfer, at an arbitrary location on a film cooled surface of interest, is expressed by E q n . 1.1 as:  q = h (T w  f  - T f)  w  (1.1)  re  where q is the convective heat flux [ W / m ] , hf is the heat transfer coefficient i n the presence 2  w  of the film layer [ W / ( m K ) ] , T is the local surface temperature, and T j is a reference or 2  w  re  driving temperature, which may vary w i t h position. T h e selection of the reference temperature has a direct impact on the evaluation of the heat transfer coefficient. In this study, a well behaved heat transfer coefficient is sought. If the limiting case of an adiabatic (perfectly insulated) surface is considered, the heat flux is zero and the reference temperature must be equal to the resulting adiabatic surface temperature. T h e adiabatic wall temperature can thus be used as the reference, or driving temperature. T h e heat transfer rate equation then becomes:  q = hf(T - T ) w  w  aw  (1.2)  For a film cooled surface, the adiabatic wall temperature distribution is generally un-  9  known and a function not only of the geometry and mainstream and cooling jets flow fields, but also of the temperatures of the cooling jets and crossflow and the degree of mixing between them. This is known i n the literature as the three-temperature problem, namely referring to the temperature of the mainstream and the reference temperature (T ) aw  (Too), temperature  of the cooling jets (T2),  at some location downstream of the injection holes. In  this case both the mainstream and the jet flows are present, w i t h different temperatures i n the mainstream than i n the cooling jets. In this study therefore, film cooling applications are also referred to as three-temperature problems, and use of the three-temperature analysis is understood to be indicative of a film cooling process. If the flow properties can be assumed constant, the use of the adiabatic wall temperature i n E q n . (1.2) ensures that the heat transfer coefficient is independent of the temperature difference (T  w  — T ) aw  and is a function only of the geometry and character of the flow field:  q w  /  rp  rp  J-w  J-ad  (1.3)  T h e heat flux is also naturally zero when the temperature difference goes to zero w i t h no restriction on the heat transfer coefficient. If any other temperature difference is selected (that is, the reference temperature is not the adiabatic wall temperature), the wall heat flux and the temperature difference do not go to zero simultaneously. T h e heat transfer coefficient then has to be zero when the heat flux is zero, and goes to either plus or minus infinity when the temperature difference moves through zero.  This is the case when the mainstream temperature is used as the driving  temperature i n the heat rate equation, as pointed out i n the past [11]. Previous studies have often tried to avoid the complications of measuring the adiabatic wall temperature or ensuring adiabatic wall conditions. A clear distinction has to be made i n such situations between the film heat transfer coefficient hj defined by E q n . (1.2) and the heat transfer coefficient h defined by: 0  q = h' (T - Too) w  0  10  w  (1.4)  In the absence of the film layer T  would be equivalent to the mainstream temperature  aw  Too, and for high speed flows the recovery temperature should be used. E q u a t i o n (1.4) thus describes the convective heat transfer rates i n a two-temperature problem when there is a temperature difference between the mainstream and the surface. In this case, i n the absence of any cooling jets, T  w  ficient ti  and Too are the only temperatures involved. T h e heat transfer coef-  at any location on the surface of interest is then determined by measurement of  0  the two temperatures, T  w  and Too, and of the heat flux q . w  In general, heat transfer coeffi-  cients (ti ) defined on the basis of local wall and mainstream temperatures are dependent on 0  upstream wall temperature distribution even i n an essentially unchanging velocity field. For example, i n the case of a two-dimensional flat plate laminar boundary layer, ti was found 0  to be some 36% different when calculated on the basis of constant wall temperature or constant wall heat flux [4]. However, these differences are only about 4-5% for a fully turbulent flat plate boundary layer [48]. It is therefore generally assumed that, for fully turbulent boundary layers, the values of heat transfer coefficients are essentially independent of the exact upstream wall temperature distribution, provided the flow field remains unchanged and the wall temperature gradients close to the point of measurement are small [4], [11]. This assumption is also made i n the present thesis. To further illustrate the complexity of the convection heat transfer coefficient, consider the situation when both the mainstream and the jet flows are present, and both are at the same temperature. T h e very presence of the cooling jets can enhance the turbulent transport especially near the injection holes, the consequence being i n this case a higher heat transfer coefficient (ti) than that obtained without the presence of the jet cooling flow (ti ). W i t h 0  the temperatures being equal i n the mainstream and i n the cooling jets, the heat transfer rate is expressed i n a form similar to that i n E q n . (1.4):  q = h'(T -T ) w  w  00  (1.5)  where Too = T = T f, w i t h the heat transfer coefficient ti generally different from the heat 2  re  transfer coefficient for the case of no injection h' . Such a formulation might be applicable i n 0  11  the analysis of some internal cooling arrangements, but hardly useful when the cooling jets are present, and the temperatures are different i n the two stream flows. Since this study is concerned with the external cooling only, the adiabatic wall temperature represents the limiting value of the wall temperature when the heat transfer rate is zero, according to E q n . 1.2. A non-dimensional representation of the adiabatic wall temperature is the film cooling effectiveness 77, denned by the equation:  (1.6) For constant property flows, the energy equation is linear i n temperature so that the film cooling effectiveness is a function only of the generated flow field and of the position on the surface. T h e film cooling effectiveness varies from zero to unity, from locations far downstream where the adiabatic wall temperature approaches the mainstream temperature {T  aw  = T^,  to injection hole locations where the adiabatic wall temperature approaches the  temperature of the jet flow (T  aw  = T2) respectively.  T h e present study focuses on low speed, constant property flows, where the temperature field is uncoupled from the flow field and the energy equation is thus linear i n temperature. W i t h i n these assumptions, the definition of the heat transfer coefficient hj by E q n . 1.2, and the definition of the film cooling effectiveness 77 by E q n . 1.6 provide two well behaved parameters to completely describe the three-temperature situation on a surface of interest. F r o m an experimental point of view, the linearity of the energy equation for small temperature differences implies that only the temperature difference between the cooling jets and the mainstream is of importance and not the absolute values of the two temperatures. E x p e r i ments can thus be performed w i t h heated jets rather than cooled jets to obtain the desired temperature difference between the jet flow and the mainstream flow, giving flexibility to a particular experimental setup.  12  1.6  Literature Review  A comprehensive review of the film cooling technique, analysis and experimental studies, was reported by Goldstein [19]. This study summarizes the expertise i n the field of film cooling up to 1970s, almost exclusively including flat plate arrangements. T h e study concluded that for three-dimensional flows, much more experimental work was needed to account for the effects of various parameters such as hole geometry, density difference between cooling jets and mainstream, and the effect of the interaction of individual jets on the adiabatic wall temperature distribution. T w o parameters are typically used i n the literature to quantify the discharge of the coolant into the mainstream, the mass flow ratio M or the momentum flux ratio I defined by:  M = ^ -  (1.7)  Quite often, the actual heat transfer process was simulated by an analogous mass transfer process, so that measurements of r? could be made directly at the wall using mass fractions (concentrations) rather then temperatures [19]:  v = -^—TT-  I-) 1 9  T h e use of mass transfer analogy to heat transfer has the experimental advantage that large temperature differences can be readily simulated by using gases of different densities. T h e analogy is well established for constant property flows and was frequently used. Shadid and Eckert [47] extended the analogy to processes w i t h temperature and concentrationdependent fluid properties. T h e y concluded that the analogy is well satisfied for film cooling and impingement cooling especially for a binary mixture of CO2 and air used i n a mass transfer experiment, and for no heat/mass flux into the solid wall.  13  There is extensive literature dealing w i t h film cooling on flat surfaces, i n various configurations and arrangements. However, since the present study focuses on simulated leading edge geometries, only similar leading edge studies will be reviewed here. A semi-cylindrical leading edge geometry was investigated by Salcudean et al. [44] using the heat/mass transfer analogy and a flame ionization detector technique . T h e geometry consisted of four rows of spanwise inclined fully staggered holes, symmetrically placed about the stagnation line, and fed from a common plenum. Measurements were made i n a large subsonic w i n d tunnel, and results of the flow division between rows of cooling holes and film cooling effectiveness data were reported. A i r was used as coolant. T h e authors found a lower limit of the overall mass flow ratio below which flow reversal occurs i n the rows of cooling holes closest to the stagnation fine. F i l m cooling effectiveness values were highest at lower mass flow ratios, close to the cutoff value for the first rows of cooling holes, and decreased abruptly w i t h both increasing and decreasing mass flow ratio. Salcudean et al. concluded that alternate designs should be investigated to obtain more uniform cooling over a wider range of mass flow ratios, and to increase the mass flow ratio for m a x i m u m 77, to avoid flow reversal i n the rows of cooling holes closest to the stagnation line. T h e uncertainty i n 77 was estimated at ± 0 . 0 5 at one hole diameter downstream of ejection, and the overall uncertainty i n the measurement of mass flow ratios was estimated to be less than 4%. T h e effect of the relative stagger of the rows of holes on 77 was addressed by Salcudean et al. [45]. T h e method and the experimental setup used were similar to that i n [44], and measurements of 77 were taken at different mass flow ratios for different hole pattern arrangements. T h e row of holes closest to the stagnation line was displaced i n the spanwise direction w i t h respect to the row of holes further away from the stagnation fine, from a fully staggered hole pattern to an in-line hole pattern. T h e authors found that there was an o p t i m u m mass flow ratio for highest 77 downstream of the second row of cooling holes, for each relative position of the rows of holes. A fully staggered arrangement was better at lower mass flow ratios, while relatively higher spanwise-averaged 77 was obtained for the in-line hole pattern and for a wide variety of mass flow ratios. T h e authors concluded that the in-line hole arrangement  14  was superior to the staggered arrangement as far as 77 is concerned. M i c k and M a y l e [37] reported measurements of hj and 77 for a semi-cylindrical leading edge geometry containing three rows of spanwise inclined fully staggered holes. T h e model was placed i n a low speed, open circuit wind tunnel, while a secondary system supplied air from a common plenum to the film cooling holes. The model surface was fitted w i t h stainless steel heater strips and the surface temperature distribution was measured using thermocouples. T h e film cooling effectiveness values were calculated from a corrected one-dimensional heat balance on an element of the surface, since the test surface was not adiabatic. T h e heat transfer test used a constant heat flux boundary condition and hj was calculated from an energy balance on a heated surface element that included the effects of surface radiation and heat conduction through the wall. Immediately downstream of the cooling holes the spanwise-averaged 77 was highest at the lowest mass flow ratio tested, and decreased w i t h increasing mass flow ratio. Further downstream of the injection sites, highest values of spanwise-averaged 77 were found for the intermediate mass flow ratio tested.  Near the  cooling holes, hj was found to be three times the values without injection, and slightly downstream of the holes hj dropped to values approximately 10% higher than the values without injection. M i c k and M a y l e also concluded that the regions of high effectiveness do not necessarily correspond to regions of high heat transfer coefficient, and this was more evident w i t h increasing mass flow ratio. T h e uncertainty i n both 77 and hj data was estimated at 3.5% downstream of the leading edge, and was as high as 17% at locations around and immediately downstream of the cooling holes. Mehendale and H a n [32] studied the effect of high mainstream turbulence on 77 and hj on a semi-cylindrical leading edge geometry, similar to that i n [37] and [44]. Four rows of fully staggered spanwise inclined cooling holes were placed symmetrically about the stagnation fine, and measurements were also made i n a low speed wind tunnel. A method similar to that employed i n [37] was used to calculate hj and 77. The uncertainty i n both parameters was estimated at around 15% i n the region around the holes, and at less than 5% downstream of the round leading edge. The authors concluded that 77 decreases w i t h increasing mass  15  flow ratio, and decreases significantly w i t h increasing mainstream turbulence at lower mass flow ratios. T h e film effectiveness was less sensitive to increased mainstream turbulence at higher values of the mass flow ratio. T h e heat transfer coefficient w i t h film injection was found to increase w i t h increasing mass flow ratio, the effect being more pronounced i n the leading edge region than further downstream. For the intermediate mass flow ratio tested, the leading edge hf was found to increase with increasing mainstream turbulence, and no systematic effect of the mainstream turbulence on hf was found for the lower and higher mass flow ratios tested. O u et al. [39] used the same technique and geometry as i n [32] to investigate the effect of high mainstream turbulence for the case of injection from individual rows of cooling holes. T h e authors reported that 77 decreases w i t h increasing mass flow ratio for b o t h cases of injection through the rows of holes closest to the stagnation fine, as well as for injection through the rows of holes further away from the stagnation fine. Increasing mainstream turbulence decreased 77 for both cases but the effect was more pronounced for the rows of holes closest to the stagnation fine. T h e leading edge hf increased w i t h increasing mass flow ratio, and the mainstream turbulence level generally increased hf for the rows of holes closest to the stagnation line, while increased hf was clearly observed for the case of injection through the rows of holes further away from the stagnation fine. O u and H a n [38] followed up the work described i n [32], using inclined film slots to replace the original round cooling holes. The scope and the technique used are basically the same as i n [32]. T h e results and conclusions were also consistent with those reported i n [32]. T h e effect of periodic passing wakes on 77 was addressed by Funazaki et al. [18] for a leading edge geometry similar to that used i n [37], [32], and [44]. T h e surface temperature was measured by thermocouples embedded i n the model surface. T h e method used i n [37] was employed for calculation of 77. The uncertainty i n 77 was estimated at about 6% over the leading edge region, and less than 12% for the regions further downstream of the injection holes. T h e authors identified an o p t i m u m mass flow ratio for almost full coverage of the surface and relatively high 77. Periodic wake passage was found to deteriorate the spanwise  16  77, while increased mainstream turbulence had no systematic effect on the values of 77. T h e surface temperature was also visualized by using a thermochromic liquid crystal, but no quantitative data was inferred from this data.  Thermochromic liquid crystals ( T L C ) can be used to monitor surface temperature distributions, as an alternative to more conventional thermal instrumentation.  Hippensteele  et al. [24] used narrow band liquid crystals on M y l a r sheets to determine external heat transfer coefficients along the midchord of a scaled turbine blade model. T h e heat transfer coefficients were calculated from the liquid crystal monitored surface temperatures using an energy balance on the airfoil model. T h e uncertainty i n the heat transfer coefficient was estimated at ± 6 . 2 % . Chfford et al. [8] used low temperature phase change paints to record surface temperatures i n transient convective heating experiments. Heat transfer coefficients were then calculated for internal blade cooling passages using a one-dimensional conduction formulation w i t h appropriate boundary conditions. Heat transfer coefficients were assumed constant w i t h time and insensitive to the upstream wall temperature effect.  This latter  assumption was based on test cases where heat transfer coefficients were calculated for different upstream wall temperature conditions. Metzger and Larson [36] used similar coatings i n transient heat transfer experiments to record surface temperatures.  Heat transfer coef-  ficients were then calculated from the one-dimensional response of a semi-infinite m e d i u m suddenly exposed to a step change i n the convected fluid temperature. T h e uncertainty i n local Nusselt number was estimated at ± 1 2 % for a 95 percent confidence level. T h e use of thermochromic liquid crystals to record surface temperatures is more common i n the literature as applied to two-temperature situations. Such analyses were also done by C a m c i et al. [5], von Wolfersdorf et al. [52], W a n g et al. [53], and C h y u et al. [7] to name a few. T h e more complicated three-temperature situation, as applied to film cooling arrangements, was addressed by Vedula and Metzger [51] for the case of flat plate air injection through streamwise inclined holes, and E k k a d et al. [12] [13] for the case of flat plate air and  17  C02 injection through compound angle holes. F i l m cooling effectiveness and heat transfer coefficient were determined from two related transient tests using a method similar to that described i n [36]. Relatively narrow band liquid crystals were sprayed onto the surfaces of interest to determine local temperatures.  T h e uncertainty i n 77 was estimated i n [51] at  ± 1 2 % and that i n hf at ± 8 % , and the estimated uncertainty i n [12] was ± 6 . 5 % for hf and ± 8 . 5 % for 77.  In summary, very few studies have addressed the heat transfer characteristics (77 and hf) i n the leading edge region of turbine blades. In some of these studies, the heat and mass transfer analogy was used to obtain 77 from an isothermal mass transfer experiment, while the other studies used steady-state heat transfer experiments w i t h constant surface heat flux to obtain hf, and with constant temperature secondary gas to obtain 77. A couple of studies [51], [12] [13] reported values of 77 and hf i n three-temperature situations using the transient one-dimensional conduction response of a semi-infinite wall.  In these latter  cases, relatively narrow-band liquid crystals were used to monitor the time required for the surface temperature to reach predetermined values set by the colour change of the liquid crystal surface coating; two related tests were needed i n these studies to obtain the required information for the simultaneous calculation of 77 and hf.  This transient technique using  narrow-band liquid crystals i n film cooling applications has the disadvantage that two separate tests are needed to obtain the heat transfer characteristics, and improper choice of the liquid crystal colour change temperatures can lead to large errors i n 77 and hf [26].  1.7  Objectives and Scope  Cooled turbine blades and vanes are very expensive components since they require advanced high temperature materials and coatings, involve complicated cooling schemes, are hard to manufacture, and require special inspection techniques. T h e design of a high performance turbine blade is i n itself a very costly undertaking. M u l t i p l e developments may have to be  18  made i n order to correct deficiencies of a baseline design. E a c h design modification has to be finally certified through expensive and thorough rig testing. Due to the complexity of the modern turbine blade cooling systems, and the large number of flow field and geometrical parameters involved, accurate calculations of 77 and hf over the regions of interest are yet to be made. T h e turbine designer relies heavily on experience and modeling. One of the limiting factors i n using higher turbine entry temperatures is thermal fatigue i n the leading edge of turbine blades and vanes. Reduced thermal stresses i n this region translate into longer blade rupture life, which means fewer blade replacements during the engine lifetime, and big financial benefits for the airline operators.  Improvements i n the  design and performance of leading edge cooling could lead to either reduced cooling air for a given T E T , or the T E T could be increased for given amounts of cooling air, b o t h situations being beneficial for the overall engine performance. T h e literature still contains few studies of leading edge film cooling, the general agreement being that more research is required to analyse and quantify the effects of the many flow and geometrical parameters involved. T h e scope of the present work is to investigate several issues related to the performance assessment and film cooling design of the leading edge of turbine stator blades, or vanes. Specifically, the objectives of the present experimental investigation are as follows: 1. To make a comparative assessment of two different leading edge geometries based on film cooling effectiveness measurements using a flame ionization detector technique. A broad range of variables will be considered including Reynolds number, mass flow ratio, coolant density. 2. To explore the effects of geometry i n a new leading edge design, and to compare the film cooling effectiveness results measured using a flame ionization detector technique from this geometry w i t h the results from the other two leading edge geometries. 3. To create simple correlations of film cooling effectiveness for some of the investigated cases, so that the measured results can be used more directly i n existing design proce-  19  dures. 4. To develop, implement, and validate a transient heat transfer technique based on wideband liquid crystal thermography. This technique uses the surface temperature history, which is calculated from the digitally recorded true-colour information during a single transient test, for the simultaneous determination of 77 and hf i n  three-temperature  situations.  T h e subsequent chapters of this study will provide details of the experimental setup, theory, and methodology used to achieve these goals.  Chapter 2 presents a comparative  investigation of three leading edge geometries based on a flame ionization detector technique i n isothermal mass transfer experiments. Chapter 3 is dedicated to a new technique based on wide-band liquid crystal thermography. Chapter 4 presents an application of wide-band liquid crystal thermography and a comparison of film cooling effectiveness data calculated from the liquid crystal method w i t h film cooling effectiveness data measured w i t h a flame ionization detector technique. In Chapter 5 the conclusions of this study are presented and recommendations for future work are made.  20  Chapter 2 Measurements of Leading Edge Film Cooling Effectiveness Using a Flame Ionization Detector This section describes the measurement of film cooling effectiveness for three leading edge geometries using the mass/heat transfer analogy i n fluids w i t h constant properties.  The  flame ionization technique and the wind tunnel setup used i n the experiments are presented. To simulate the effect of different temperatures, and thus different densities, i n the cooling jets and the mainstream i n an adiabatic process, either air or GO2 was used as coolant i n the corresponding isothermal mass transfer process. The density of the CO2 is approximately 1.5 times higher than the density of air thus simulating an absolute temperature ratio of about 1.5. A c t u a l temperature ratios i n the leading edge of an operating turbine may be higher, but the present experiments at least give trends of film cooling effectiveness as coolant density increases relative to mainstream density. A correlation was developed for the film cooling effectiveness data from the leading edge region of one of the geometries used and results are presented.  A n example of an improved coolable leading edge of a gas turbine  blade is also shown i n this section. This new leading edge geometry achieves higher film cooling effectiveness at mass flow ratios sufficiently high to provide a safe backflow margin.  21  This is accomplished by appropriate distribution of film cooling holes both i n a spanwise and chordwise manner within given structural constraints. T h e experimental investigation is carried out at zero angle of attack w i t h air as coolant medium at different mass flow ratios. T h e experimental investigation described i n this section was undertaken to provide basic insight into the film cooling injection through multiple rows of holes i n the leading edge of turbine blades. T h e final leading edge design was discussed with P r a t t Sz W h i t n e y Canada.  2.1 2.1.1  Experimental Arrangements Model and Wind Tunnel Setup  T h e experiments were performed i n the Aerodynamics Laboratory at the University of B r i t i s h C o l u m b i a using a large open-circuit blower-type wind tunnel. T h e mainstream was driven by a 2.44 m diameter variable pitch 16-blade fan powered by a 125 H P motor. T h e test section was 2.44 m wide by 1.66 m high and was located downstream of a 4:1 area contraction section. T h e m a x i m u m tunnel air speed was above 20 m / s and was measured by P i t o t tube. One of the wind tunnel side walls was made of clear plexiglass for observation purposes. A large scale turbine blade model was manufactured from acrylic plastic and had a semicylindrical leading edge with a diameter D of 127 m m , and an afterbody w i t h parallel top and bottom sides 1.19 m long, and overall chordwise length of 2.28 m .  T h e model did  not follow the shape of a typical turbine blade except for the leading edge region where the measurements were taken. T h e model was mounted horizontally and spanned the test section at zero incidence so that the frontal blockage of the tunnel was 7.9%. To reduce or eliminate spanwise flow i n separation bubbles, fences were placed on the model at spanwise positions ± 0 . 9 5 m from the tunnel centreline. T h e overall arrangement of the model i n the wind tunnel test section follows that i n [44] and is shown i n F i g . 2.1. Turbulence intensities and the uniformity of mean velocities i n the wind tunnel test section were measured i n [44] at the location of the model's leading edge but without the presence of the model i n the test section. Turbulence intensities were less than ± 1 % and mean velocities were uniform to  22  Figure 2.1: W i n d tunnel turbine blade model. within ± 1 % . The central region of the model was hollow and accommodated a plenum attached to a removable leading edge insert which contained the film cooling holes. T h e plenum was free to slide inside the model's afterbody for removal and attachment of the leading edge insert. This modular construction allowed for fast replacement of one leading edge insert w i t h another, so that different film cooling hole patterns could be easily investigated without any modification to the overall wind tunnel setup. T h e plenum was fed w i t h coolant gas (either air or C0 ) 2  contaminated w i t h trace amounts of propane through a 25 m m delivery  hose near the trailing edge of the model. T h e plenum also contained a flow baffle and 2 fine screens for uniform delivery of the coolant gas to a perforated impingement plate attached to the leading edge insert, and thence to the film cooling holes located i n the removable insert. Details of the plenum arrangement are shown i n F i g . 2.2.  2.1.2  Baseline Leading Edge Geometry  The baseline geometry used i n the present study follows the description i n [44]. T h e model had a semi-cylindrical leading edge w i t h a diameter D of 127 m m and a uniform thickness t of 25.4 m m . Four rows of cooling holes, located symmetrically at &i — ± 1 5 ° and 0 = ± 4 4 ° 2  from the stagnation line, were machined w i t h an inclination of S = 30° to a spanwise fine on  23  Screens  /  Flow baffle  insert Figure 2.2: P l e n u m arrangement and removable leading edge insert. the leading edge surface. T h e row of holes at ± 1 5 ° is referred to as row #i, and the row of holes at ± 4 4 ° is referred to as row 0 . E a c h row consisted of 7 holes w i t h a diameter d of 12.7 2  m m located 4 hole diameters apart i n the spanwise direction, or S — Ad. O n each side of R  the stagnation line, the two adjacent rows of holes were staggered i n the spanwise direction at a distance S  AT  a fully  of 2 hole diameters. This particular hole pattern was also referred to as  staggered arrangement. Perfect symmetry was achieved across the stagnation line so  that SLE = Orf, this hole arrangement being also referred to as an in-line  pattern. Due to  the 30° hole inclination and the 25.4 m m thickness of the leading edge wall, the hole length along its centreline was £ = Ad. Details of the baseline geometry are shown i n F i g . 2.3.  2.1.3  Compound Angle - CA Geometry  A removable leading edge insert was machined w i t h a compound angle hole pattern and this geometry is referred to as the C A geometry. T h e outer diameter and the thickness of the leading edge as well as the diameter of the holes were the same as for the baseline geometry (D — 127 m m , t = 25.4 m m , d = 12.7 m m ) .  A l l the other design parameters such as  number and position of the rows of holes, hole location and hole inclination w i t h respect to the spanwise direction were also the same as for the baseline geometry.  24  25  T h e C A geometry was different from the baseline geometry only w i t h respect to the inclination of the film cooling holes. In addition to the S = 30° angle w i t h a spanwise line on the surface, the holes for the C A geometry were also machined w i t h an inclination of = 33° w i t h a streamwise line passing through the centre of the holes (refer to F i g 2.4 on page 27 for the definition of ip). In this configuration, a group of 4 holes intersected each other at the inner side of the leading edge i n a rather complicated pattern. This arrangement was a feature of this particular C A geometry and it should not be assumed that cooling holes intersect each other i n real blade design. T h e common plenum of each group of 4 holes changed the flow entrance conditions for the C A geometry as compared to the entrance conditions for the baseline geometry. T h e hole length along its centreline was different from the holes i n row 6\ to the holes i n row 9 , and different from the baseline geometry case. For the C A geometry, £ = 5d 2  for the holes i n row #i, and £ = 7.bd for the holes i n row 0 . 2  2.1.4  Improved Leading Edge Geometry  2.1.4.1  Background Information  Some of the issues regarding the cooling of the leading edge region of turbine blades have been reviewed i n section 1.4.2 on page 7 of this study, and a typical cooling arrangement was shown i n F i g . 1.2. T h e cavity i n the most forward region of the blade, shown i n F i g . 1.2, extends radially over the entire spanwise length of the blade and provides the necessary coolant flow to the film cooling holes i n the leading edge of the blade. In this particular arrangement, the leading edge cavity is continuously supplied w i t h coolant flow through holes i n the inner wall enclosing the cavity. T h e coolant from the leading edge cavity is discharged through multiple rows of holes into the hot gas boundary layer to protect the skin of the blade for some region downstream of the injection sites. T h e coolant flow i n the leading edge cavity must be therefore at higher pressure than the pressure of the core hot gases which can also vary radially along the blade span. T h e  26  Figure 2.4: Geometrical nomenclature. exact pressure and temperature distributions around the leading edge region of the blade are therefore required as inputs i n the design of the blade cooling system. A potentially serious problem occurs i f the coolant supply pressure i n the leading edge cavity drops below the levels required to maintain a uniform coolant flow through a l l the discharge holes i n the leading edge of the blade. In this situation, the mainstream hot gas can penetrate through the film cooling holes. This phenomenon is known as backflow and can drastically reduce the operating life of the blade or even cause the blade failure. T h e design of the cooling supply system to the leading edge cavity is not addressed i n this study, but it is understood that a uniform cooling flow is required through the holes at the leading edge of the blade. In order to keep a safe backflow margin, a high enough coolant pressure must be maintained i n the leading edge cavity and this translates into higher overall mass flow ratios through the film cooling holes. The coolant removes heat from the leading edge wall by convection through the film cooling holes and then interacts with the hot gas boundary layer to form a protective film layer which reduces the heat transfer rates at critical points on the leading edge outside  27  surface.  T h e arrangement  of the film cooling holes around the leading edge region and  the geometrical characteristics of the array of cooling holes are critical for maximizing the effectiveness of any particular cooling scheme. A strong interaction between the cooling jets and the crossflow can cause significant aerodynamic penalties w i t h direct consequences on the overall efficiency of the gas turbine. T h e change i n the relative position of the aerodynamic stagnation line w i t h respect to the film cooling holes at various engine operating regimes also has to be considered.  In a particular arrangement, the film cooling holes could be  placed about the aerodynamic stagnation line corresponding to the most demanding engine operating conditions, but for any variation i n the position of the aerodynamic stagnation line no backflow is allowed and the coolant still must be discharged through holes on either side of the stagnation line. A good coverage of the leading edge region, both on the inner side and on the outer side of the leading edge wall, is typically achieved w i t h multiple rows of cooling holes disposed at specified angles 9 around the leading edge. T h e cooling holes are spaced at a distance S apart i n either row, could be spaced at a distance S R  AR  apart i n adjacent rows on either  side of the stagnation line, and could be also spaced at a distance SLE apart i n the adjacent rows about the stagnation fine. These geometrical parameters were defined i n F i g . 2.3 on page 25 for the particular case of the baseline geometry. T h e cooling holes could be generally placed at an angle S w i t h a spanwise fine on the surface and at an angle ip w i t h a streamwise fine passing through the centre of the holes. T h e compound angle ip is defined i n F i g . 2.4. M o r e complicated cooling hole exit geometries can be obtained.  In a l l cases, structural  considerations prevail so that the cooling holes are not allowed to intersect each other and a critical wall thickness has to be maintained between holes.  2.1.4.2  P V C Geometry  T h e development of the present P V C geometry included design modifications to different geometrical parameters such as row stagger about the geometrical leading edge SLE-, stagger of adjacent rows S , AT  and hole compound angle ip [30]. The leading edge thickness to hole  28  Figure 2.5: PVC geometry - outside view. diameter ratio t/d [42], the inclination angle with a spanwise line on the surface 8, the hole pitch i n a row of holes S , and the position of the rows of cooling holes denned by angles Oi R  and 6 were fixed i n the design process. The parameters SLE, S , S , t/d, 8, 6i and 6 were 2  AR  R  2  denned i n F i g . 2.3 on page 25, and ip is defined i n F i g . 2.4. It is understood that i n this particular arrangement, the cooling holes shown i n F i g . 2.4 are neither placed i n the same streamwise plane perpendicular to the leading edge, nor could they be defined entirely i n such a plane due to the spanwise angle 8^0. geometry features 4 rows of cooling holes placed at ±0° and ±0  2  leading edge [42].  T h e PVC  about the geometrical  T h e holes i n row 6i and row 0 are machined at an angle ^ i and ip 2  2  respectively between the hole centreline and the local spanwise plane normal to the blade surface.  A l l holes are machined at an angle 8 w i t h a spanwise line on the surface.  29  The  outside layout of the cooling holes is shown i n F i g . 2.5 and detailed technical specifications were provided to P r a t t & W h i t n e y Canada, a principal sponsor of this research [30].  2.2 2.2.1  Experimental Procedures The Heat and Mass Transfer Analogy  Measurements of film cooling effectiveness (denned by E q n . 1.6 on page 12) i n a heat transfer process could be made by measuring temperature at an adiabatic wall, i n the coolant and i n the mainstream (T , aw  T , and 2  respectively). It is often difficult to ensure adiabatic  boundary conditions experimentally, but it is relatively easy to make the wall impermeable i n a mass transfer process. If the corresponding non-dimensional transport equations have the same form, the same mechanism governs the transport of momentum, heat, and mass for a given configuration and the processes are said to be analogous. This is also known as the Reynolds analogy. T h e analogy between heat and mass transfer processes proved very useful i n film cooling applications since measurements of adiabatic wall temperatures are basically analogous w i t h measurements of isothermal impermeable wall concentrations. Rather than having the jets at a different temperature (and hence, at a different density) from the mainstream i n a heat transfer process, a gas of different composition is injected isothermally i n a mass transfer process.  A s discussed by Goldstein [19], the film cooling effectiveness can be then calcu-  lated based on measured concentrations at an impermeable wall, i n the coolant and i n the mainstream (C , C , and w  2  respectively) according to E q n . 1.9 on page 13.  Details of the derivation of the analogy between heat and mass transfer may be found i n [46], and details of the similarity of the velocity, thermal and concentration boundary layers may be found i n [25]. T h e restrictions associated w i t h using the analogy were analysed i n [47]. For the low speed constant property fluid situation, the heat and mass transfer analogy depends on having geometrical similarity and the P r a n d t l number (Pr) equal to the Schmidt number (Sc), or the Lewis number (Le = Sc/Pr)  30  equal to one. T h e P r a n d t l number  is the ratio of momentum diffusivity v to thermal diffusivity a , a n d the Schmidt number is the ratio of momentum diffusivity to mass diffusivity DAB- I n the case of fully turbulent boundary layers, both the turbulent Lewis number and the molecular Lewis number should be unity for the analogy to be complete. These conditions are satisfied for gases a n d i n film cooling applications w i t h adiabatic and impermeable walls respectively, since the temperature a n d concentration gradients at the wall are zero i n such situations a n d thus the influence of molecular Pr a n d Sc is small.  2.2.2  The Flame Ionization Detector  A system used for measuring concentration values i n turbulent flow was described by Fackrell [15], [16]. This system was designed to measure very low concentration levels, where conventional laboratory techniques could not be used. T h e technique uses a flame ionization detector ( F I D ) that is sensitive to a hydrocarbon trace gas. T h e F I D consists of a hydrogen and air flame which burns i n an insulated flame chamber. A voltage is applied across the chamber a n d the introduction of a hydrocarbon trace gas into the flame, such as propane, leads to the production of ions. T h e small current which can pass through the contaminated gas is amplified to a suitable voltage output by an electronic amplifier. T h e analog signal can then be sampled by a data acquisition board for direct storage into a digital computer (PC). Theoretically, the F I D response is linear w i t h the mass flow rate of hydrocarbons entering the flame chamber such that the output signal should be E = aCpQ, where a is a constant, C is the concentration, p is the density of the hydrocarbon gas, a n d Q is the flow rate through the chamber. However, the efficiency of the ionization process is a more complicated function of flow rate, such that the output signal is actually E = aCpf(Q) but is still linear w i t h concentration. Tests were done w i t h different calibration gases, a n d it was found that the F I D response is indeed linear with concentration for values up to 2000 p p m , a n d that the system was sensitive to concentration values as low as 1-2 p p m of propane [15]. For the present experimental investigation, propane was used as a trace gas a n d it was  31  thoroughly mixed w i t h the coolant flow (which was either air or CO2).  The maximum  concentration of propane i n the plenum varied from one test to another, but it was never higher than 1000 ppmv (parts per million by volume). T h e very high sensitivity of the F I D system to low levels of concentration was particularly useful i n the present experimental procedure.  T h e use of very low trace gas concentrations did not affect the restrictions  imposed by the use of the heat and mass transfer analogy, i n particular the effect of the trace gas on the mass diffusivity coefficient DAB were assumed to be negligible.  2.2.3  F I D Calibration  2.2.3.1  Calibration of the Air-Air Mixture  W h e n air is used as coolant, gas of the same composition is present both i n the jets and in the mainstream. In this case, the system needs to be calibrated only for two reference concentrations, and the linear response of the F I D is used to calculate other concentrations. Sampling of the propane contaminated flow was done through a series of needle probes as described i n section 2.2.4. T h e calibration procedure used is described i n the following.  1. T h e system was first calibrated for " 0 % " concentration of propane. This was achieved by disconnecting the propane feed from the plenum flow. T h e default output voltage EQ was recorded for this situation. This value should be the same for a l l needle probes. 2. T h e propane feed was connected to the plenum flow which was sampled to determine the highest possible, or "100%", concentration of propane.  T h e needle probes were  then individually calibrated for their "100%" reference point and the corresponding output voltage E \  0 0 %  was recorded, where i = 1 to 12 represents the number of the  needle probes being calibrated. 3. One needle probe was installed inside the plenum and was used to monitor the "100%" reference point i n the plenum at the beginning of each sampling sequence.  A linear  interpolation was used to adjust this reference level for variations within ± 2 . 5 % of the  32  trace gas concentration i n the plenum. If reference levels fell outside the ± 2 . 5 % limits, the test was aborted. T h e other 11 needle probes were disconnected from the plenum flow and were exposed to the mainstream at predetermined locations on the leading edge. For each sampling probe, linear interpolation was used between the two reference calibration points to determine the local concentration at the wall according to E q n . 2 . 1 :  ^100% ~  ^°  Since only the plenum flow was contaminated with the trace gas, C = 1 and Coo — 0 i n 2  E q n . 1.9 so that the film cooling effectiveness was calculated directly from:  77 =  2.2.3.2 W h e n C0  Calibration of the A i r - C 0 2  2  C  (2.2)  w  Mixture  is used as coolant, gases of different composition are present i n the jets and  mainstream flow respectively. Downstream of the injection sites, the C0  2  jet flow (which is  contaminated with small amounts of propane) is mixed with the mainstream air flow. In this case, a special calibration of the a i r - C 0 2 mixture is required and the procedure is described i n the following. 1. A separate setup was used for calibration of the a i r - C 0 mixture. T h e setup included 2  a metered air feed, a metered C0  2  feed connected to the C0  2  feed connected to an in-line heater, and a propane  feed. T h e in-fine heater was used to maintain the C0  2  at the same temperature as the air flow. T h e air and the propane contaminated  flow C0  2  were then thoroughly mixed i n a specially designed mixing box, and directed through a chamber where a sampling needle probe was installed. 2. W i t h the C0  2  and propane flow disconnected, air flowed through the sampling chamber  and the default voltage EQ was recorded. 33  3. W i t h the air flow disconnected, a known amount of C 0 was contaminated w i t h a trace 2  amount of propane and directed through the sampling chamber. T h e corresponding F I D voltage was recorded as E % • T h e C 0 WQ  2  contaminated flow was maintained at  this rate for the rest of the calibration procedure. 4. W i t h the C0  2  contaminated flow fixed at a known rate, the air flow rate was varied i n  incremental steps and the mixture was directed through the sampling chamber. T h e corresponding F I D output voltage E was then recorded for each known mass fraction of CO2 i n the mixture. T h e calibration data set was analysed and a fourth order polynomial was fitted through the data points. T h e above procedure was repeated for different C0  2  flow rates and different  amounts of propane. T h e repeatability of the calibration curve was extremely good, w i t h variations less than ± 2 % . This procedure allowed a check to be made on b o t h the linearity of the F I D response w i t h the small amounts of propane used, and the independence of the F I D response to the amount of C0  2  gas present i n the plenum. T h e average calibration curve for  the a i r - C 0 mixture is presented i n F i g . 2.6 on a mass fraction basis. This calibration curve 2  was then used i n the experiments to calculate the actual CO2 concentration at the wall from the recorded raw F I D output voltage E. T h e m a x i m u m uncertainty of the calibration curve was estimated at less than ± 2 % . T h e calibration of the " 0 % " and "100%" reference points followed the procedure described in section 2.2.3.1. This time, the plenum flow consisted of C0  2  gas contaminated w i t h a  trace amount of propane. T h e exact procedure is described next. 1. W i t h the CO2 feed disconnected, air flowed through the plenum and the reference voltage EQ was recorded. This value should be the same for a l l needle probes. 2. T h e air feed was disconnected from the plenum and the propane contaminated C 0  2  feed  was connected instead. The needle probes were fed from this plenum flow exclusively, and were individually calibrated for their "100%" reference point. T h e corresponding  34  0.0  0.2  0.4  0.6  0.8  10 .  • E )/(E 0  %CO2ptenum  - E) 0  Figure 2.6: Mass fraction calibration for the a i r - C C ^ mixture. output voltage E\  QQ%  was recorded, where i = 1 to 12 represents the number of the  needle probes being calibrated. 3. One needle probe was installed inside the plenum and used to monitor the "100%" reference point i n the plenum at the beginning of each sampling sequence. A linear interpolation was used to adjust this reference level for variations w i t h i n ± 2 . 5 % of the trace gas concentration i n the plenum. A g a i n , i f the reference voltage showed variations larger than ± 2 . 5 % , the test was terminated. T h e other 11 needle probes were disconnected from the plenum flow and were exposed to the mainstream at predetermined locations on the leading edge.  Here too, only the plenum flow was contaminated w i t h the trace gas, so that and C  2  = 0  = 1. T h e film cooling effectiveness was consequently calculated from the raw F I D  35  output voltage E w i t h the use of the a i r - C C ^ mixture calibration curve as:  n = c, = A^W  (2.3)  ^ ^100% ~ ^0 )  where g is the calibration function shown i n F i g . 2.6.  2.2.4  The Sampling Procedure  Measurements of film cooling effectiveness were made by sampling the mixture of contaminated coolant and mainstream gas very close to the wall and immediately downstream of the injection sites. Sampling was achieved through a series of 11 very fine needle probes w i t h outer diameter of 0.5 m m and spaced d/2 = 6.35 m m apart lying directly on the surface of the model, thus covering a spanwise distance of 5rf = 63.5 m m on the leading edge. T h e sampling probes were mounted on a specially designed rake that kept the probes parallel to the crossflow and at predetermined locations on the leading edge. T h e rake was attached to a suitable traverse mechanism driven by step motors to allow the accurate positioning of the needle probes on the leading edge. T h e positioning error of the probes was ± 1 m m . T h e gas mixture was sampled through these probes and sent through small tubes to a Scanivalve and thence to the flame ionization detector which measured the concentrations w i t h an estimated uncertainty of ± 1 % . A dedicated P C and software were available for data acquisition from the F I D . A computer program controlled the location of the rake, the switching and time delay between sampling probes, and the processing of the amplified F I D output voltage. T h e sampling flow rates were kept constant during the experiment by using a suitable p u m p and a settling chamber as described i n [16]. A time delay of 15 s was used before sampling from each probe to permit the cleaning of the tubes of the residual mixture from a previous sample sequence. A total of 2000 data points sampled at 130 H z were averaged to calculate 77 at each needle probe location. T h e estimated film cooling effectiveness uncertainty i n [44] accounted for errors i n flow rate measurement (less than ± 2 % ) , flame ionization detector sensitivity (less  36  than ± 1 % ) , sampling at 0.25 m m above the wall (less than ± 3 % ) ; for the present measurements, 77 was measured w i t h an uncertainty of less than 2.5% for air injection and less than 3% for CO2 injection.  2.3  Experimental Results  For the baseline geometry, measurements of 77 were taken using either air or C 0  2  as coolant  gas. T h e effect of double row and single row injection was investigated. T h e mass flow ratios (defined by E q n . 1.7 on page 13) were adjusted by modifying the mainstream velocity while the jet Reynolds number was kept constant at a value of Re  2  = U20I/U2 — 4200. In the  case of double row injection, five different overall mass flow ratios were used corresponding to M =0.50, 0.64, 0.78, 0.97 and 1.20 respectively. A correlation for the film cooling effectiveness data was determined i n this case. In the case of single row injection, the mass flow ratios used varied between approximately M = 0.21 and M = 1.55. T h e effect of the jet Reynolds number was investigated for the case of single row injection of air. For the C A geometry, measurements of 77 were taken for the case of double row injection of air at a constant jet Reynolds number of i ? e = 4200. In a l l cases, the reported results 2  represent spanwise-averaged values of film cooling effectiveness (77) obtained by integration of local values of 77 over a spanwise distance S = 4d. For the P V C Geometry, measurements of 77 were made for the case of multiple row and single row injection using air as coolant. For the case of multiple row injection, five different mass flow ratios were used corresponding to M = 0.6, 0.8, 1.0, 1.2, and 1.4 respectively. For the case of single row injection, measurements of 77 were made at fixed locations immediately downstream of the film cooling holes, and using mass flow ratios ranging from M = 0.25 to M = 1.2. For a l l cases, the jet Reynolds number was kept constant at a value of Re  2  =  U d/v = 4200 and different mass flow ratios were achieved by modifying the mainstream 2  2  velocity.  37  FID sampling probes  Row 8 .  Row 0  8  f  Row 6 2  0 0  ROW  e  2  T  Figure 2.7: Double row injection arrangement.  2.3.1  Baseline G e o m e t r y  2.3.1.1  Double Row Injection  For this case, 2 injection holes were opened i n each row while a l l others were obstructed using a t h i n tape on the outside of the leading edge. Due to symmetry, measurements were taken only on one side of the leading edge stagnation line, beginning immediately downstream of the second row of cooling holes. Figure 2.7 shows the hole pattern and the fixed spanwise location of the sampling probes. T h e relative direction of coolant injection is also shown in F i g . 2.7 as well as the direction of the mainstream flow. T h e streamwise location of the F I D sampling probes was automatically adjusted during the experiments by moving the rake w i t h the F I D probes i n a direction perpendicular to the stagnation line as shown i n F i g . 2.7. This configuration could be considered representative of an infinite row i f rj values are periodic over the pair of two holes downstream of which measurements were taken.  This  means that r\ values from the sampling probes number 9, 10, and 11 should be the same as TJ  38  I  I  I I I  I  I I  I I  I  I I  I  1  1 1  1  I ' 1  1  1  I  1  M = 0 . 5 X/(TCD/4)=1 . 0 3  2  3 1  1  4 1  1  1  1  1  5 1  1  1  1  1  1  6 1  1  1  7  8  9 FID  1 0 11 probe 1.0  1  1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 '  M=1.2  0.9  1 1 ' 1 1 1 1 '  1 1 1  I I  i  1  x/(itD/4)=1.03  0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0  . i . . . .  1 2  i . . . .  3  i  4  5  6  7  8  9 10 11 FID probe  Figure 2.8: Baseline geometry: air, double row injection, periodicity check.  39  values from the sampling probes number 1, 2, and 3 respectively. This was actually the case during the experiments: F i g . 2.8 presents spanwise distributions of film cooling effectiveness for the lower and higher mass flow ratio used at two streamwise locations placed immediately downstream of the cooling holes and immediately downstream of the cylindrical leading edge on the flat afterbody. Overall, very good periodicity was observed for the streamwise planes passing through at least the sampling probes number 9, 10, a n d 1 a n d 2 respectively for a l l mass flow ratios used i n the experiments. Therefore, it was assumed that the arrangement shown i n F i g . 2.7 was representative of an infinite row. T h e values of 77 from probes 1 to 9, between the periodicity planes, were then used to calculate 77 according to E q n . 2.4:  _ 1  0-5(T7I +r? ) + Ei=|>7i 9  =  =  8  A direct comparison of 77 for the case of air and CO2 injection is shown i n F i g . 2.9. T h e streamwise position was measured from the stagnation line so that x/(irD/4) = 0 represents the stagnation line and x/(irD/4) = 1 represents the end of the semi-cylindrical leading edge and the beginning of the flat afterbody. Streamwise locations x/(7r£)/4) < 1 are therefore indicative of regions on the round leading edge while streamwise locations  X/(TTD/A)  >  1 are  indicative of regions on the flat afterbody. T h e overall mass flow ratio through the cooling holes was calculated based o n the total discharge area, the density of the coolant a n d mainstream gases, a n d the average coolant velocity for a l l holes U . T h e average coolant velocity for a l l holes was calculated from the 2  condition that the jet Reynolds number be a constant, and this velocity was used to calculate the jet to mainstream mass flow ratios M. For the case of air injection and for a l l mass flow ratios, 77 is highest at the location closest to the injection holes a n d decreases immediately downstream. For the rest of the circular leading edge region 77 is fairly constant. A sudden increase i n 77 is observed at the measurement location on the afterbody closest to the leading edge for the intermediate mass flow ratios of M =0.64, 0.78, a n d 0.97. Overall, 77 decreases w i t h increasing mass flow ratio being highest at M = 0.50 a n d lowest at M = 1.20. O n average, a 50% reduction i n 77 is observed at a l l streamwise locations when the mass flow  40  I  I I  I  I I  I — 1 1 1 1  1 1 I I  : solid symbols  J  <: hollow symbols 2  .  M = 0 . 5 0  *  M = 0 . 6 4 1=0.41 ;  ,  M=0.78  1=0.61  •  M=0.97  1=0.94  •  M=1.20  1=1.44  •  M = 0 . 5 0  1=0.17-  A  M = 0 . 6 4  1=0.28  V  M=0.78  <5  M=0.97  o  M=1.20  i .... " " 0 . 5  0.6  0.7  0.8  0.9  1.0  1.1  1.2  1.3  1.4  i  .  .  1.5  .  1=0.25:  1=0.40  '.  1=0.62: 1=0.95  .  i  .  .  .  ; .  1.6  1.7  X/(TCD/4)  Figure 2.9: Baseline geometry: air and CO2, double row injection. ratio is increased from M = 0.78 to M = 0.97. The injection of CO2 coolant, whose density is about 1.5 times that of air, produced higher values of 77 for all mass flow ratios used. The momentum flux ratio / (defined by E q n . 1.8 on page 13) is the parameter mostly used when comparing 77 values from injection of gases of different densities. E v e n for the same momentum flux ratios, 77 is higher for the case of CO2 injection except for / = 0.6 where the results are similar for b o t h air and C02This latter situation is shown i n detail i n F i g . 2.10. It can be seen from this figure that similar 77 for air and CO2 cooling were produced by similar area coverage of the cooling jets. The same trend of decreasing 77 with increasing mass flow ratio was observed for injection of CO2 as for injection of air. T h e highest spanwise-averaged film cooling effectiveness value was around 77 = 0.5 for the case of CO2 injection at M = 0.5. The fact that higher film cooling effectiveness was achieved at such low mass flow ratio is not necessarily useful i n a real gas turbine application. Since a l l holes at the leading edge are usually fed from a common plenum, the pressure distribution around the leading edge w i l l  41  c\jir>cOT-*tr-T-Ttr>-o o o o o o o o o o o  © d b d b o o c  CM  o> b o o o _ SE <  O)  ~  d  00  ^ o  o °>  2  O  II  c-i O  a 5 m 6 Q 6  6  CO  6  CM  6  O  o 6  M  5  d Q d  d  CO  d  CM  d  "5.  Figure 2.10:  Baseline geometry: air and C 0 , double row injection, I 2  42  force the discharge from the holes i n row #i, closest to the stagnation line, to be less than the discharge from the holes i n row 6 . For the present baseline geometry, complete cutoff 2  of the flow from the holes i n row 6\ occurred at M = 0.45 for the case of C0  2  injection and  at M = 0.32 for the case of air injection [44]. For injection at mass flow ratios below these cutoff values, backflow occurs through the row of cooling holes closest to the stagnation line jeopardizing the structural integrity of the blade. A safe backflow margin has to be maintained for a l l engine operating conditions and this requires the use of higher mass flow ratios.  2.3.1.2  Film Cooling Effectiveness Correlation - Double Row Injection  There are numerous parameters which affect the film cooling performance at the leading edge of turbine blades. For a given leading edge geometry, some of these factors are: position and number of rows of cooling holes, hole spacing i n each row, hole spacing i n adjacent rows, geometry of the holes, jet to mainstream mass flow/momentum flux ratios, jet to mainstream temperature differences, mainstream turbulence, pressure gradients, relative mainstream to blade angle of attack.  E v e n a brief analysis of the effect of each of these parameters on  film cooling effectiveness requires an extensive research program. Design optimization often requires that modifications be made simultaneously to more than one parameter.  Conse-  quently, a very large amount of data needs to be analysed at each step i n the design and experimentation procedures. In these conditions, the film cooling effectiveness data become useful only i f presented i n the form of correlations based on the parameters used. These correlations can be more easily implemented i n existing design procedures than the raw data, and have the significant advantage that the process can be automated based on continuous feedback. In general, correlations are extensively used i n engineering applications and they are even more useful i n the complicated conditions found i n gas turbines. For the present study, one of the goals was to develop appropriate correlations for the film cooling effectiveness based on variable coolant density data.  T h e baseline geometry  was chosen for investigation-. The film cooling effectiveness data for this case was already  43  presented i n F i g . 2.9. For both cases of air and CO2 injection, the spanwise-averaged film cooling effectiveness was fairly constant for most of the leading edge region except for the downstream location closest to the film cooling holes x/(7r£)/4) = 0.57. A t this location and for the case of air injection, the values of fj are typically higher than farther downstream. For the case of CO2 injection, the same trend was expected immediately downstream of the film cooling holes so that measurements of 77 were only taken at streamwise locations X/(TTD/4)  > 0.64. In a gas turbine application, the region closest to the film cooling holes  benefits from the convection cooling through the cooling holes. Therefore, the focus must be on the leading edge region farther from the cooling holes where lower values of 77 are expected. For the baseline geometry, this region corresponds to streamwise locations 0.64 < X / ( T T D / 4 ) < 1.0.  For this region, an area-averaged film cooling effectiveness i n spanwise and streamwise directions was introduced to represent the m i n i m u m effectiveness that can be expected for the leading edge of the baseline geometry. This parameter is further referred to as 77 and is denned by E q n . 2.5:  A correlation was developed for 77 and the results are presented i n F i g . 2.11. A t h i r d order polynomial fits the data with an error of less than 6.7% for the case of air injection and  7.2% for CO2 injection respectively.  The independent variable is represented by a  blowing parameter denned by E q n . 2.6:  1.41  P2  - 1  PocUoo  (2-6)  where £ reduces to the mass flow ratio M when p = Poo (air injection into air). 2  It is understood that the correlation i n F i g . 2.11 applies explicitly to the baseline geometry for values of the blowing parameter between £ = 0.25 and £ = 1.2 at constant jet Reynolds number of  Re = 4200, and for streamwise locations 0.64 < 2  44  X/(TCD/A)  <  1.0. T h e correlation  0.6  O.o •  0.0  '  1  1  i  I  i  i  i  0.5  i  I  i  i  1  1  1.0  ^=M  1.5 +  (p /pJ-  1 / 1 4 1  2  -1  Figure 2.11: F i l m cooling effectiveness correlation: baseline geometry, double row injection. i n F i g . 2.11 accounts implicitly for the geometrical layout of the baseline geometry w i t h respect to the position of the rows of cooling holes, their relative stagger, and the hole pitch i n a row of cooling holes. M o r e experiments are needed using different geometrical layouts to obtain a more complete correlation where the variation of fj, or possibly the variation of fj itself, is expressed explicitly as a function of the various geometrical parameters involved.  2.3.1.3  Single R o w Injection - Constant J e t R e y n o l d s N u m b e r  This section presents the results for the case of injection through the holes i n either row 8\ or row 9 . Perfect symmetry was maintained about the stagnation line; a l l the holes i n row 2  6 were taped and 4 holes were opened i n row 0\ for the case of first row cooling, a n d vice 2  versa a l l the holes i n row 6\ were taped and 4 holes were opened i n row 6 for the case of 2  second row cooling. T h e hole layout used and the fixed spanwise location of the sampling  45  probes is shown i n F i g . 2.12. Only the holes that were opened are shown along w i t h the relative direction of coolant injection and mainstream flow. A g a i n , the rake w i t h the F I D sampling probes can only move i n a streamwise direction perpendicular to the stagnation line. T w o streamwise locations were investigated i n each case. T h e jet Reynolds number was kept constant at a value of Re  = 4200 and different mass flow ratios were achieved by  2  modifying the mainstream velocity. In the case of injection through the holes i n row 9\, the effect of the mass flow ratio is illustrated i n F i g . 2.13. The values of fj are higher for C0  2  injection than for air injection  for a l l mass flow ratios used, and decrease w i t h increasing mass flow ratio. This effect is more pronounced at the location closest to the film cooling holes (x/(irD/4)  = 0.32) where fj  decreases more rapidly w i t h increasing mass flow ratio. For a l l mass flow ratios, fj is higher at the streamwise location closest to the row of cooling holes than at the location farther downstream. Values of fj should have a peak for some value of the mass flow ratio M > 0 since fj = 0 for M = 0. T h e exact values of the peak fj could not be determined i n this case due to limitations of mainstream velocity so that lower mass flow ratios could not be produced. Figure 2.14 shows the effect of mass flow ratio at two different streamwise locations for the case of injection through the holes i n row 9 . 2  T h e location  X/(7TJD/4)  = 0.64 was  immediately downstream of the cooling holes and the other streamwise location was farther downstream on the flat afterbody. Peak values of fj are evident at both streamwise locations and are similar for both air and C0 . 2  These peak values of fj occur at lower mass flow ratio  for air and at higher mass flow ratio for C0 , 2  as expected since the air coolant has a higher  injection momentum at the same M than the C0  2  coolant. T h e values of fj are higher at the  streamwise location closest to the injection holes than farther downstream for a wide range of M for both air and C0  2  cooling, but the differences become smaller w i t h increasing mass  flow ratio. For air cooling, the values of fj are eventually the same at the two streamwise locations for M > 1. T h e peak values of fj are some 5 0 % higher at the location closest to the cooling holes than at the location on the flat afterbody. Overall, the use of the C0  2  46  coolant  ure 2.12: Single row injection arrangement.  47  0.6  •  i  i i i  i  1 11  -i  0.5  0.4 -  i  i  1  1 1 1 1 r-  i  ^  A I R x/(7iD/4)=0.32  *  C 0 x/(rtD/4)=0.32  «  A I R x/(jtD/4)=0.64  o  C 0 x/(7tD/4)=0.64  2  2  0.3  0.2  0.1  0.0 0.0  0.2  0.4  0.6  0.8  1.0  1.2  1.4  1.6 M  Figure 2.13: Baseline geometry - row 9i cooling: effect of mass flow ratio. 0.6  0.5 -  0.4  *  A I R X/(JID/4)=0.64  *  C 0 x/(rcD/4)=0.64 2  *  A I R X/(JCD/4)=1 .27  o  C 0 X/(TCD/4)=1.27 2  0.3  0.2  0.1  0.0 "0.0  0.2  0.4  J i 0.8  0.6  i  i  i  i i i_ 1.0  1.2  1.4  1.6 M  Figure 2.14: Baseline geometry - row 0 cooling: effect of mass flow ratio. 2  48  is more effective at higher M. For mass flow ratios greater than about 0.68, fj is higher for ( 7 0 2 injection than for air injection at both streamwise locations investigated. The above analysis, based on the overall mass flow ratio, is consistent w i t h the general approach from published film cooling literature. Nevertheless, since the exact location of the rows of cooling holes on the semi-cylindrical leading edge is known, a local mass flow ratio or momentum flux ratio can be calculated based on the local velocity outside the boundary layer just upstream of the cooling holes. T h e local velocity approaching the holes at 15° and 44° on the cylindrical surface can be calculated from the pressure coefficient measured at these positions. T h e usual pressure coefficient is defined by E q n . 2.7 for this low-speed flow:  2 P  C  -  l  TJ2  0  2^00^00  -  V- ) 1  7J2  1  oo  u  where 9 is the angle around the cylinder measured from the stagnation line, pg and Ug are the static pressure and the local velocity outside the boundary layer respectively. T h e pressure coefficient is about C  p  = 0.77 for the holes at 9 — 15° and about C  p  = —0.5 for the holes  at 9 = 44° [20]. T h e local momentum flux ratio Ig can then be determined from the overall mass flow ratio M , the pressure coefficient C , and the coolant to mainstream density ratio p  Pi I Poo according to E q n . 2.8:  P2UI pooUi  P2U2  2, Ul  poo  PooU  M  2  (2.8)  a  The results from F i g . 2.13 and F i g . 2.14 are replotted i n terms of the local momentum flux ratio i n F i g . 2.15 and F i g . 2.16 respectively. For the case of injection through the holes at 15°, the spanwise-averaged film cooling effectiveness is still higher at the same local momentum flux ratio for CO2 injection than for air injection. For the case of injection through the holes at 4 4 ° , fj is the same at lower 1$ for b o t h air and CO2 injection. T h e peak values of fj still occur at lower Ig for air injection than for CO2 injection, but appear much closer when plotted i n terms of local momentum flux ratio than i n terms of overall mass flow ratio. Immediately downstream of the cooling holes, the peak  49  0.5  -i—i—i—r—  r  1  1  i  0.4  •a a> u> 2  *  A I R x/(rcD/4)=0.32  ^  C 0  «  A I R x/(jtD/4)=0.64  o  C O , X/(TID/4)=0.64  2  X/(TTD/4)=0.32  0.3  (0  4> (0  5 c (0  0.2  Q.  0)  0.1  0.0 0.0  i i  -i 2.0  1.0  3.0  4.0  5.0  i_  6.0  8.0  7.0  9.0  I = (P uV( u ) 2  6  P  e  Figure 2.15: Baseline geometry - row 9\ cooling: effect of local momentum flux ratio. 0.5  -1—I—I-  0.4  TJ 0) O) 2  —  AIR  —  C 0  X/(TXD/4)=0.64 X/(JID/4)=0.64  2  *  AIR  o  C O , X / ( T C D / 4 ) = 1 .27  x/(rcD/4)=1.27  0.3  (0 Q) W  1 CO  0.2  Q.  0.1  0.0_ 0.0  '  0.2  0.4  0.6  0.8  •  1  1.0  1.2  1.4  l = (p U ) / ( p U ) 2  e  2  2  Figure 2.16: Baseline geometry - row 6 cooling: effect of local momentum flux ratio. 2  50  e  0.6  '  1 1  1  I  1 1 1 1  I  I  I  I  I  I  I  I  I  I  I I I I I I I  I I  1 1 1 1  I  1 1 1 1  I  I  1  I  I  I  I  I  I  I  I I  Air, row 9,, x/(7tD/4) = 0.32 Air, row 9 , X / ( J I D / 4 ) = 0 . 6 4 2  0.5  •o  CD O)  0.4  E «  0.3  §.  0.2  a co '3 c  (/)  0.1  0.0  1  0.0  0.2  0.4  • ' • •  0.6  1  1  1  ' •  0.8  1  1  1.0  • ' •  ' • • • •  1  1.2  1.4  1.6  1.8  2.0  '  2.2  2.4  2.6  2.8  3.0  M = U /U e  2  e  Figure 2.17: Baseline geometry - single row air injection: effect of local mass flow ratio. fj occurs at about Ig = 0.15 for air injection and at about Ig = 0.20 for C0  2  values of local momentum flux ratio greater than about Ig = 0.2, C0  2  injection. For  injection provides  better fj than air injection. A l t h o u g h the local momentum flux ratio does not completely correlate the spanwiseaveraged film cooling effectiveness for the case of single row injection, some trends were better represented i n terms of Ig than i n terms of M. It is therefore recommended that this approach be pursued for future analysis of single row injection w i t h variable density coolant. Immediately downstream of the rows of cooling holes, the spanwise-averaged film cooling effectiveness results are expressed i n terms of the local mass flow ratio Mg i n F i g . 2.17 and F i g . 2.18 for air injection into air and for C0  2  injection into air respectively. T h e local mass  flow ratio is defined by E q n . 2.9 w i t h the use of the pressure coefficient defined by E q n . 2.8:  mjr Mg =  P2U J7oo — = M— = PcoUg Ug 2  51  ,  M  (2.9)  y/l-C  p  0.0  0.2  0.4  0.6 0.8  1.0  1.2  1.4  1.6  1.8  2.0 2.2 2.4  2.6  2.8 M  Figure 2.18: Baseline geometry - single row C0  2  e  3.0 =  3.2  3.4  (pU) /(pU) 2  e  injection: effect of local mass flow ratio.  For a given geometry, it was expected that local conditions at the point of injection characterise the film cooling effectiveness immediately downstream of the cooling holes, and that the trends be similar regardless of the position of the cooling holes on the cylindrical leading edge surface. T h e results i n F i g . 2.17 and F i g . 2.18 support this assumption and fj values from row 6i match reasonably the values from row 0 immediately downstream of the 2  cooling holes for the same local mass flow ratios. The differences are probably due to the interaction between the jets and the boundary layer at the location of the cooling holes and immediately downstream. T h e actual velocity profile at the exit of the cooling holes might also change w i t h row position around the cylindrical leading edge; since the holes are fed from a common plenum, the flow entrance conditions are not the same for holes placed at different angles around the cylindrical surface. This effect is not captured by the local mass flow ratio but some trends can be clearly observed from F i g . 2.17 and F i g . 2.18 respectively; both figures show a peak fj at relatively low Mg, and the peak fj was the same for b o t h cases of air injection and C0  2  injection w i t h a value around fj = 0.45. T h e peak fj occured at  52  a lower Mg i n the case of air injection than i n the case of C0  2  injection. For both cases,  a further increase of the local mass flow ratio caused a decrease i n 77. For higher values of Mg, the spanwise-averaged film cooling effectiveness was higher for C0  2  injection than for  air injection. The analysis based on the local mass flow ratio Mg could be used i n the future to quantify the effect of single row injection and different hole geometries for a given coolant gas, while the local momentum flux ratio Ig is useful, especially for lower values of Ig, when analysing the effect of variable density coolant for a given hole geometry.  2.3.1.4  Single Row Injection - The Effect of Jet Reynolds Number  The effect of jet Reynolds number on film cooling effectiveness was investigated for the case of single row injection. For each row, measurements of 77 were taken at two streamwise locations downstream of the injection holes using air as coolant. These locations were immediately downstream of the holes i n row #i (x/(irD/4:) — 0.32), immediately downstream of the holes i n row 6 (X/(TTD/4) = 0.64), and on the flat afterbody (X/(TTD/A) 2  = 1.27). For each fixed jet  Reynolds number, the mass flow ratio was adjusted by modifying the mainstream velocity. Figure 2.19 shows spanwise-averaged film cooling effectiveness 77 for air injection through the holes i n row 61 at the streamwise location closest to the holes. There are no significant differences i n the values of 77 over the range of Reynolds numbers investigated, except at the lowest mass flow ratio of M = 0.2. A t this low mass flow ratio, 77 decreases some 30% w i t h increasing jet Reynolds number from Re — 846 to Re = 2540. Overall, 77 decreases 2  2  w i t h increasing mass flow ratio and this trend is more pronounced at lower mass flow ratios M < 0.4. A t the location farther downstream, the effect of jet Reynolds number on 77 is presented i n Figure 2.20. Here too, fj decreases with increasing mass flow ratio for a l l jet Reynolds numbers but the effect is more pronounced at the lower jet Reynolds numbers used. For the lowest mass flow ratio used M = 0.2, 77 drops again significantly w i t h increasing jet Reynolds number from Re = 846 to Re = 2540. 2  2  For the case of air injection through the holes i n row 6 , the effect of jet Reynolds number 2  53  0.6  ~\— —<1  Re = 846 Re =1270 Re =1693 Re =2540 Re =3387 Re =4233  0.5  2  2  2  "g  0.4  2  2 >  2  (0 0.3  2  (A  re Q.  0.2 0.1  0.0 "0.0  •  '  i  I  l  0.2  I  '  '  0.4  0.6  0.8  1.0  1.2  1.4  M  1.6  Figure 2.19: Baseline geometry - row 6\ air cooling: X/(TTD/4) = 0.32. 0.6  -I  -T—IT—--r—-I—-|—  I  I  I  I  I  I  Re = 846 Re.; =1270 Re. _ =1693 Re =2540 Re =3387 Re =4233  0.5  2  2  ^  2  0.4  2  O)  2  §1 a  (0  re Q-  0)  2  2  0.3 0.2  0.1  0.0 0.0  I  0.2  0.4  0.6  I  I  0.8  I  I  I  , 1 —  1.0  1  1,  1.2  1.4  Figure 2.20: Baseline geometry - row 6i air cooling: X/(TTD/4) = 0.64.  54  M  1.6  Figure 2.21: Baseline geometry - row 8 air cooling: x/(irD/4) 2  — 0.64.  0.6  0.5  Re = 846 2  Re =1270 2  3  Re =1693 2  0.4  Re =2540 2  O)  Re =3387  2 I n  2  0.3  ro Q. 0.2 n  0.1  0.0 0.0  0.2  0.4  0.6  0.8  1.0  1.2  1.4  Figure 2.22: Baseline geometry - row 0 air cooling: X/(TTD/4:) = 1.27. 2  55  M  1.6  is presented i n F i g . 2.21 and F i g . 2.22 for streamwise locations immediately downstream of the cooling holes and farther downstream on the flat afterbody respectively. T h e jet Reynolds number clearly affects the peak fj at the streamwise location closest to the cooling holes. T h e peak fj occurs at lower mass flow ratio for lower Re2 and at higher mass flow ratio for higher Ret, as expected.  T h e values of fj are also some 26% higher at the lower jet Reynolds  number of Re2 = 846 than at the higher jet Reynolds number of i ? e  2  = 4233. For mass  flow ratios higher than about M = 0.5, the values of fj are basically independent of the jet Reynolds number and fj decreases w i t h increasing mass flow ratio. A t the location on the flat afterbody, fj is clearly independent of the jet Reynolds numbers used and fj has a peak at about M = 0.6.  2.3.2  C A Geometry  Figure 2.23 shows the comparison of spanwise-averaged film cooling effectiveness for the case of multiple row air injection for the baseline and C A geometries. T h e same procedure and setup described i n section 2.3.1.1 were used to obtain film cooling effectiveness at a constant jet Reynolds number Re = 4200 for the C A geometry. 2  T h e compound angle of the cooling holes provides significantly higher fj for the C A geometry than for the baseline geometry for a l l mass flow ratios investigated. T h e effect is most beneficial especially for regions closer to the cooling holes. For the C A geometry, a peak in fj occurs at the streamwise location X/(TTD/4) = 0.64. T h e spanwise-averaged film cooling effectiveness decreases at locations farther from the cooling holes, and it decreases more rapidly for the lower mass flow ratios than for the higher mass flow ratios. T h e differences i n fj for the two geometries become less significant at locations on the flat afterbody.  The  highest values of fj occur at the lowest mass flow ratios investigated M — 0.64 for the C A geometry and M — 0.50 for the baseline geometry respectively. M o r e significant was the improvement i n fj for the C A geometry at mass flow ratios farther from the cutoff values of row 6\ of cooling holes. For example, fj is on average some 90% higher for the C A geometry than for the baseline geometry at the intermediate mass flow  56  X/(TID/4)  Figure 2.23: Spanwise averaged 77 for baseline and C A geometries. ratio M = 0.97. This trend is maintained even at the highest mass flow ratio investigated of M — 1.2 where 77 is on average some 70% higher for the C A geometry than for the baseline geometry. Figure 2.24 shows the effect of geometry more simply i n terms of the area-averaged  film  cooling effectiveness 77 for the streamwise region between x/(irD/4) = 0.64 and X/(TTD/4) = 1.03, and for a spanwise region S = 4d downstream of the cooling holes.  For the C A  geometry, the analysis based on 77 is not very accurate since the spanwise-averaged film cooling effectiveness clearly decreases with increasing streamwise distance from X/(TTD/4) = 0.64 to X/(TTD/4) = 1.03 for a l l mass flow ratios investigated.  Nevertheless, the use of  77 provides a simple and clear qualitative indication of the overall improved film cooling performance of the C A geometry which is attributed to the compound angle ip = 33° of the cooling holes. This angle represents the only geometrical parameter that was different for the C A geometry than for the baseline geometry where tp = 0 ° .  57  O.o I—<—i—'—'—I—'—'—i—'—I— — —'— — —i— —'— —I—'— — — — 1  0.4  0.6  1  0.8  1  1  1.0  1  1  1  1  1.2  1  1  1.4  M = U /U. 2  Figure 2.24: Comparison of baseline and C A geometries: fj, double row air injection. 2.3.3  P V C Geometry  2.3.3.1  Double Row Injection  Figure 2.25 shows the layout used and the spanwise location of the F I D sampling probes. For this case, 5 injection holes were opened i n each row. T h e sampling probes covered a fixed spanwise region of S = 7.5d and were only allowed to move i n a streamwise direction, that is, perpendicular to the geometrical leading edge, along the fines shown i n F i g . 2.25. This arrangement was chosen to be representative of an infinite row. T h e streamwise location of the sampling probes was automatically adjusted using the same hardware and software described i n section 2.2.4. T h e total streamwise region covered by the sampling probes was from x/(nD/A)  = 0.46 to X/(TTD/4)  = 1.23. T h e values of TJ from a l l probes were used  to calculate the spanwise-averaged film cooling effectiveness fj. T h e measurements were  58  taken on one side of the geometrical leading edge beginning w i t h a location immediately downstream of row 62 of cooling holes. Figure 2.26 shows a comparison of the spanwise-averaged film cooling effectiveness for the P V C and baseline geometries for mass flow ratios M > 0.6. For the P V C geometry, the highest fj was achieved at the lower mass flow ratio M = 0.6 for the streamwise region immediately downstream of the cooling holes, and at M = 0.8 for streamwise regions farther away from the cooling holes. Overall, there are little differences i n the values of fj for the lower mass flow ratios used, M = 0.6 and M = 0.8. A t higher mass flow ratios, fj decreased w i t h increasing mass flow ratio; the same trend of decreasing fj w i t h increasing mass flow ratio was observed previously for both baseline and C A geometries. A t the lower mass flow ratio M = 0.64, the baseline geometry provides only slightly higher fj than the P V C geometry at M  = 0.6.  For higher mass flow ratios, the P V C  geometry provides significantly higher fj than the baseline geometry. For the mass flow ratio M = 1.0, fj for the P V C geometry is on average almost twice the fj values for the baseline geometry. This situation is shown i n more detail i n F i g . 2.28. E v e n at M — 1.4, the values of fj farther away from the cooling holes are higher for the P V C geometry than for the baseline geometry at M = 1.2. Also, for the P V C geometry, the values of fj for mass flow ratios of M = 1.0, 1.2 and 1.4 show little variation with the streamwise coordinate for locations on the cylindrical region of the leading edge. A comparison of fj for the P V C and C A geometries is presented i n F i g . 2.27. A t the lower mass flow ratio M = 0.64, the C A geometry provides higher fj than the P V C geometry at M = 0.6 and M = 0.8. A t M = 0.8, fj is slightly higher for the C A geometry than for the P V C geometry. For all other higher mass flow ratios investigated, the P V C geometry provides higher fj than the C A geometry and the differences are more significant at locations farther away from the cooling holes. The area-averaged film cooling effectiveness can be used to summarize qualitatively the relative performance of the baseline, C A , and P V C geometries. Figure 2.29 shows the variation of fj calculated for a streamwise region from x/(wD/4)  59  = 0.64 to X/(ITD/4)  =  1.03  60  0.8 0.7  -i—i—i—r-  :  0.6 '-  I  0.5  5 CD  0.4  ^ CO CL  0.3  i  1  1  •  1  1  i  '  T  1  ' ' I  1  ' ' '  I '  1  ' '  M=0.6 P V C  -•  M=0.64 Baseline  M=0.8 P V C  -A  M=0.78 Baseline  M=1.0 PVC  -v  M=0.97 Baseline  M=1.2 PVC  -o  M=1.20 Baseline  M=1.4 PVC  CO CD  CO  CO  0.2 0.1 0.0 0.4  0.5  0.6  0.7  0.8  0.9  1.0  1.1  1.2  1.3  X/(JCD/4)  Figure 2.26: Comparison of P V C a n d Baseline geometries: air, double row injection. 0.8  'l " •[1" I I I |  0.7  T—1 1 1 1 1 1 1 1 1 1 1 |  0.5  2 «  0.4  1  1  1  1  i  1  1  1 1  -o  M=0.64CA  — M=0.6 PVC  ^  M=0.78CA  — M=0.8 PVC  -7  M=0.97 CA  — M=1.0 PVC  M=1.20 CA  — M=1.2 PVC  0.6  1  i  % co  ire  0.3  Q. (0  0.2 0.1 0.4 0.0  0.5  0.6  0.7  0.8  0.9  1.0  1.1  1.2  1.3  X/(TID/4)  Figure 2.27: Comparison of P V C a n d C A geometries: air, double row injection.  61  IB 5  W  d o ©  5  t d  CO  «N  d  d  i-; d  O d  CD 5 un d O d  CO  d  d  CN  d  T-_  o  "S.  Figure 2.28: P V C and Baseline geometries: air, double row cooling at M = 1.0.  62  p  o  against the mass flow ratio for the case of double row air injection for the three geometries investigated.  0.6  I  go  i  1  1  1  1  1  1  1  1  0.4  '  1  0.6  1  1  1  1  1  1  1  1  '  1  0.8  1  1  1  1  1  1  1  1  1  1  1  1  1  1  1  1  1  1  -  —  i  1  r  i I i i >—i I i — i i i I i — i — i — i — 1.0  1.2  1.4  1.6  M = U /U 2  M  Figure 2.29: P V C , Baseline, and C A geometries - fj: air, double row cooling.  T h e comparison of fj presented i n F i g . 2.26 and F i g . 2.27 and the comparison of fj presented i n F i g . 2.29 clearly show the improved film cooling performance of the P V C geometry at higher mass flow ratios. This allows the cooling system to operate w i t h good efficiency while providing the necessary backflow margin at different engine operating conditions. T h i s was actually the m a i n objective of the design of the P V C geometry. T h i s also proves that a design procedure must include the effect of different geometrical parameters such as row location, hole geometry and position i n a row of cooling holes, and relative stagger of the row of cooling holes.  63  0.6  0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  1.0  1.1  1.2  1.3 M  Figure 2.30: P V C geometry: single row cooling, effect of M.  2.3.3.2  Single Row Injection  Figure 2.30 presents the effect of the overall mass flow ratio for the case of injection through the holes i n row 8i and row 8 respectively. F i v e holes were opened i n either row 8\ or row 8 2  2  and the corresponding layout of the cooling holes and the spanwise position of the sampling probes were the same as described i n F i g . 2.25.  T h e streamwise location X/(ITD/A)  =  0.266 was immediately downstream of row 8\ of cooling holes and the streamwise location X/(ITD/4)  = 0.52 was immediately downstream of row 8 of cooling holes. 2  For the entire range of mass flow ratios investigated, the spanwise-averaged film cooling effectiveness was higher for the case of injection through the holes i n row 9 than for i n 2  jection through the holes i n row 8\. This behaviour was actually expected since the actual mainstream velocity approaching the holes i n row 9\ is less than that approaching the holes in row 8 just outside of the boundary layer. Therefore, for the same mass flow ratio M , the 2  local mass flow ratio Mg is higher for the holes i n row 8\ than for the holes i n row 8 . W h e n 2  64  0.0  0.2  0.4  0.6  0.8  1.0  1.2  1.4  1.6  1.8  2.0 2.2 2.4  2.6  2.8  3.0  3.2  3.4  M = U /U e  2  e  Figure 2.31: P V C geometry: single row cooling, effect of Mg. the local mass flow ratio Mg is considered, the variation of fj is as shown i n F i g . 2.31. T h e local velocity approaching the holes just outside the boundary layer was found from experimental pressure coefficient data [20] by the method described i n section 2.3.1.3. Figure 2.31 shows clearly that the local conditions at the jet exit are more important than the relative position of the rows of cooling holes. For a given hole geometry, the same fj can be expected just downstream of the cooling holes for the same local coolant discharge regardless of the streamwise position of the cooling holes. T h e fact that the values of fj do not completely overlap at the same Mg is probably mostly due to the slight differences i n the hole geometry which is a feature of the P V C geometry, and to the differences i n the boundary layers approaching the two rows of holes. For this case of single row air injection, fj values from F i g . 2.31 can be compared w i t h fj values for the baseline geometry from F i g . 2.17 on page 51. T h e peak fj occurs at approximately the same local mass flow ratio Mg = 0.3 for both geometries, and the peak value is only slightly higher for the P V C geometry than for the baseline geometry. T h e trends are  65  clearly preserved i n both cases and variations are probably due to the difference i n the exit hole geometry. For many purposes, a single plot of fj versus Mg could be used for a l l hole locations and for these two hole geometries. The effects of hole geometry i n this plot are not very significant.  2.4  Summary  For the case of multiple row injection using the baseline and C A geometries, the use of the higher density C 0  2  coolant provided higher film cooling effectiveness for a l l mass flow ratios  investigated. Higher values of fj always occurred at lower mass flow ratios, and film cooling effectiveness decreased w i t h increasing mass flow ratio for both baseline and C A geometries used. Significant improvement of the blade leading edge film cooling effectiveness was achieved for the C A geometry w i t h the use of compound angle holes. This proves the importance of geometrical parameters i n the design of film cooling systems. For a specified leading edge region of the baseline geometry, the area-averaged film cooling effectiveness data were correlated based on a newly defined blowing parameter. For the case of single row injection using the baseline geometry, higher values of fj were achieved from injection through the holes at ± 4 4 ° than through the holes at ± 1 5 ° . T h e use of the CO2 coolant provided higher film cooling effectiveness than the air coolant particularly at higher local momentum flux ratios. For the baseline geometry, the effect of jet Reynolds number on the spanwise-averaged film cooling effectiveness i n not significant i n the case of air injection from row #1 except at the lowest mass flow ratio investigated of M — 0.2. A t this mass flow ratio, fj decreases significantly w i t h increasing Re 2  For the case of air injection from row Qi, higher peak  values of fj were obtained at the lowest jet Reynolds number used of Re  2  = 846, and the  peak fj decreases w i t h increasing jet Reynolds number and also occurs at higher mass flow ratio. Overall, the jet Reynolds number has no significant effect on fj at higher mass flow  66  ratio. For the case of double row injection using the P V C geometry, the spanwise-averaged film cooling effectiveness was significantly higher than fj values for the previously tested baseline and C A geometries for mass flow ratios M > 0.64 and M > 0.8 respectively. For the case of single row injection using the P V C geometry, the fj results immediately downstream of the injection holes were consistent w i t h the values from the baseline geometry when the local mass flow ratio was used as the scaling parameter.  T h e spanwise-averaged  film cooling effectiveness has a peak at a relatively low local mass flow ratio and fj further decreases w i t h increasing local mass flow ratio. For a given hole geometry, the fj values from injection through holes at arbitrary locations around the cylindrical leading edge can be predicted w i t h reasonable accuracy when local mass flow ratios are considered.  67  Chapter 3 Liquid Crystal Thermography This section describes the measurement of wall temperature distributions based on the colour change of a thermochromic liquid crystal. Various colour imaging systems are analysed and the development of a high performance imaging system is presented. T h e concepts developed here have application i n convective heat transfer studies where a temperature  difference  exists between an arbitrary surface and the surrounding incompressible flow. T h e developed concepts relate also to more general applications where non-intrusive thermal mapping of an arbitrary surface is needed.  3.1  Thermochromic Liquid Crystals  Detailed distributions of the external heat transfer coefficient together w i t h the film cooling effectiveness parameter are necessary i n the design of film cooled turbine vanes and blades. T h e film cooling layer has to protect the airfoil for some region downstream of the injection sites, and thus high resolution mapping of relatively large regions is required. Some of the techniques most widely used for measuring surface heat transfer i n the past, as reviewed i n [36], are not flexible enough for investigating many different geometries w i t h m i n i m u m instrumentation costs. Quite often, significant modifications have to be made i n order to accomodate a new geometry, and serious complications arise when more complicated and  68  irregular surfaces need to be investigated. For example, heat flux gages together w i t h surface thermocouples have been successfully used i n the past for convective heat transfer studies, and are still used i n situations where the geometry is fixed and only the flow conditions and the temperature fields are changing. T h e y provide good local representation of the heat transfer process, but detailed mapping would necessitate multiple probes w i t h accompanying high instrumentation costs. T h e turbine cooling designer on the other hand, requires not only the investigation of a fixed geometry under various flow and temperature field conditions, but also the comparison of many different complicated geometries for identical surrounding conditions. One of the goals of this study is to develop a technically sound and economically viable experimental program to meet the designer's need for flexibility, close to real-time output of the results, and sufficient accuracy to allow sensible integration of the preliminary external heat transfer performance w i t h the assumed available aerodynamics and heat transfer numerical routines. A detailed temperature distribution on the surface of interest would prove sufficient when the analysis follows the principles discussed i n section 4.1 of this study. T h e focus has to be then on choosing a method that will grant high resolution thermal mapping over a large region of interest of possible irregular shape. T w o such methods have been identified from the literature as suitable candidates, one is based on infrared thermography and the other is based on liquid crystal thermography. B o t h techniques are non-intrusive. The surface of interest is coated w i t h high emissivity paints and thermochromic liquid crystals respectively, and the probing device (an infrared camera and a video camera respectively) is placed at some distance from the surface, thus causing no disturbance to the generated  flowfield.  Infrared thermography is based on capturing the radiation from a surface onto a thermal detector i n the probing device scanning the surface, and digitizing the generated signal to store the emitted intensity values i n a computer data acquisition system. Ultimately, a calibration procedure relates the scanned intensity values with the respective local surface temperatures. W i t h proper calibration to accont for the multiple sources of uncertainty (like op-  69  tical aberrations, radiation contributions from neighbouring elements, hysteresis associated w i t h the camera scan-frequency, etc.), and w i t h appropriate image reconstruction techniques, temperature measurements with accuracies as good as 0.3 — 0.4° C can be achieved [34]. A l though attractive, the high costs of the infrared probe makes the technique prohibitively expensive for lower budget research facilities.  T h e alternate liquid crystal thermographic  technique is clearly much more feasible from an economic point of view, and also has the technical potential of producing accurate results as suggested by studies i n the past. T h e concepts and principles associated w i t h this technique will be further documented and developed i n this study.  Thermochromic liquid crystals change colour when subjected to various temperatures. T h e colour change is reversible and this makes them suitable for repeated use. L i q u i d crystals are i n fact organic molecules which exhibit the flow properties of a liquid and some of the optical properties of a crystalline solid. T h e discovery of the liquid crystalline state lies w i t h the research performed separately i n the late nineteenth century by an A u s t r i a n botanist, Freidrich Reinitzer, and a G e r m a n physicist, O t t o Lehmann [60], [9]. It was not u n t i l 1970s that the general public was exposed to this technology through an application related to the display industry. E v e r since, liquid crystal technology has found numerous applications in many industrial and research fields such as medical thermography, non-destructive testing, heat transfer and flow visualisation, battery testing, scientific instrumentation, and of course, various liquid-crystal-based electro-optic display applications, to name only a few. A n increasing amount of dedicated literature i n now available for both the novice and the specialist. A thorough introductory description of the liquid crystals was found i n [9]. A thermochromic liquid crystal is optically active above a certain temperature called the melting point, and below a temperature called the clearing point. T h e melting point temperature for a particular liquid crystal compound represents the temperature at which the crystalline solid will begin to melt, and the clearing point represents the temperature above which the liquid crystal behaves entirely as a liquid losing the optical properties associated  70  w i t h the crystalline solid phase. L i q u i d crystals are more similar to liquids than they are to solids, having only a small amount of additional order. A typical liquid crystal molecule would be elongated i n shape, w i t h some rigidity i n the central region and w i t h the ends somewhat flexible. T h e molecules orient themselves along a certain direction i n space called director, and this is intrinsic to their anisotropy. T h e most common classification of the liquid crystal phases identifies them as smectic (the most solid-like phase), nematic (the degree of molecular order is smaller than i n smectics), and chiral nematic (Greek words for the exhibited twisted structure of this phase). In a multi-layered chiral nematic liquid crystal the molecular director rotates about an axis i n a helical fashion, the molecules i n each layer being on average oriented along the director. A pitch length can be thus defined as the distance i n the layers of molecules for a 360° rotation of the director. W h e n white light is incident on a chiral nematic liquid crystal, most of it will be transmitted w i t h some optical activity through the liquid crystal compound, and only the wavelength equal to the pitch will be reflected parallel to the direction perpendicular to the liquid crystal. A t any other viewing angle the wavelength reflected is less, w i t h the wavelength decreasing as the viewing angle (defined w i t h respect to the direction perpendicular to the liquid crystal) increases, as shown i n F i g . 3.1. W i t h chiral nematic liquid crystal, the longest wavelength for a specific temperature will be therefore observed on directions perpendicular to the liquid crystal, and for any departure from this direction shorter wavelengths (towards blue hues) will be observed. This is an important aspect that experimentalists using these materials have to compensate for through proper spatial calibrations. A compound that will exhibit m i n i m a l colour change w i t h the viewing angle is desirable since it could be used on round surfaces without the need for more complicated spatial calibrations. For m a x i m u m visualisation of the reflected colours, the thermochromic liquid crystal colour change is always viewed against a black absorbing background. T h e intensity of the produced colours is directly dependent on the available fight. Since the pitch i n the liquid crystal is a function of the temperature of the compound, T L C ' s can be manufactured to  71  Viewing angle  Figure 3.1: Effect of the viewing angle on the observed colour of a chiral nematic liquid crystal. exhibit the entire colour spectrum. R e d appears first immediately after the melting point followed by yellow, green, cyan, blue and purple hues as the temperature increases. After the clearing point, the compound becomes clear again and the black colour of the background is observed. A schematic of the temperature-colour relationship is depicted i n F i g . 3.2. T h e displayed colours of the liquid crystal appear i n reverse order when cooling the compound, w i t h violet appearing first immediately below the clearing point, red immediately above the melting point, and back to clear below the melting point.  T h e temperature  range between the melting and clearing points is also known as the thermochromic liquid crystal band.  A band as narrow as 1 ° C is possible and the liquid crystal will selectively  reflect all the colours i n the visible spectrum over such a temperature range.  Wide-band  liquid crystals can also be custom manufactured by combining different narrower band liquid crystals and exhibit the same colour change over a wider temperature range.  T h e viewed  colours would change more slowly over this wider temperature range. Various thermochromic liquid crystal manufacturers can presently produce mixtures i n any temperature between —30°C and 1 5 0 ° C . M i x t u r e s less susceptible to contamination can be presently produced between - 3 0 ° C and about 90° C [60].  72  Temperature • Clearing point  Melting point Blue  Green  Red  wavelength  Figure 3.2: Colour/temperature response of a thermochromic liquid crystal. W h i l e the melting point, or the clearing point, and the temperature band w i l l define the optically active range of the thermochromic liquid crystal, the exact colour/temperature relationship, shown i n F i g . 3.2 for an arbitrary liquid crystal under specified lighting conditions, has to be carefully determined by suitable calibration. This is a mandatory procedure when the thermochromic liquid crystal is used for quantitative mapping of various thermal regions. Thermochromic liquid crystals have already demonstrated their unique characteristics i n heat transfer research as accurate temperature measuring devices. There are numerous qualities that make these liquid crystals a very attractive alternative to more conventional temperature indicators. Their time response to minute changes i n temperature is almost instantaneous due to their small mass, and precise mass production is achieved by formulating the liquid crystals i n large batches to ensure variations of less than ± 0 . 1 ° C w i t h i n the batch. Microencapsulation and chemical coating are two methods used to protect the liquid crystals against contamination and make them extremely durable and easy to handle. Thermochromic liquid crystal inks are mainly used i n gaseous media while microencapsulated slurries can be used for flow studies i n water. The main advantage of using thermochromic liquid crystals is the capability of mapping out large thermal regions w i t h high resolution,  73  a task otherwise difficult to achieve without expensive more conventional thermographic instrumentation. T h e following procedure describes the general use of thermochromic liquid crystals to obtain the temperature distribution on a arbitrary surface of interest: 1. T h e region of interest on the surface is chemically degreased to obtain a solvent free area; 2. A t h i n layer of specially formulated black paint is then applied onto the cleaned surface to act as a base and absorbing background, and is allowed to dry; 3. T h e specified liquid crystal is then applied i n a thin layer on top of the dried black paint by one of the many available methods including screen printing, painting, spraying, and is allowed to dry; 4. If the liquid crystal instrumented surface is to be used over long periods of time and possibly exposed to various contaminating agents, further protection of the liquid crystal can be achieved by chemically coating the region of interest w i t h a specially formulated varnish; 5. Appropriate illumination has to be arranged, and the viewing angle of the instrumented surface set; 6. T h e liquid crystal colour response w i t h temperature has to be accurately determined through proper calibration under the given lighting conditions and viewing angle; 7. Once the calibration function is known, the temperature at any point on the instrumented surface is given by the viewed colour at that point. This procedure can be used i n a transient heat transfer experiment to monitor the surface temperature variation w i t h time. Once the time history of the surface temperature is known, the method described i n the previous section of this study can produce the desired detailed distribution of the surface film heat transfer coefficient hf and the film cooling effectiveness  74  r). T h e next logical step i n the development of the method is to identify appropriate means of recording and storing the colour change of the liquid crystal instrumented surface, and to find a suitable colour-temperature conversion for the particular thermochromic liquid crystal used.  3.2  Imaging System  T h e use of thermochromic liquid crystals as temperature measuring devices depends on their capability to reflect different colours at different temperatures under certain lighting conditions. A n accurate temperature measurement can only be achieved i f the colour change of the liquid crystal is properly recorded, stored and further analysed. This requires the implementation of a colour imaging system to be used i n conjunction w i t h the aerodynamic arrangements and heat transfer hardware specific to film cooling research. T h e imaging system requirements would then be high resolution colour capturing at rates sufficiently high for the transients used i n the heat transfer experiment, transmission and recording of the colour signal with m i n i m u m distortion and i n a form suitable for further computer processing, image display, enhancement and analysis capabilities. A n automated image acquisition - data processing system has to use appropriate computing power and a dedicated computer is a fundamental part of such a system.  3.2.1  Illumination Sources  It is important to recognize that the thermochromic liquid crystal only produces colour by selectively reflecting the incident available light. T h e amount and quality of the ambient light will therefore dictate the quality and intensity of the colours produced. Visible light is a form of radiant energy that propagates, by some theories, i n the form of electromagnetic waves which cover a small region of the electromagnetic spectrum, having wavelengths A between approximately 380 n m to 780 n m . Figure 3.3 shows i n more detail the spectrum of electromagnetic radiation.  75  Ultraviolet  X rays Gamma rays  10  |C]|Microwaves^>}  '3  10  10'  10  10'  10  >  Radio waves  Thermal radiation  i i I i  10 "  Infra-red  :  10  10  10'  Figure 3.3: Spectrum of electromagnetic radiation. In this context, colour represents the perceptual result of light on a human receiver, and the various devices of viewing colour are patterned on human vision. Since human vision is a very complex process involving physical, physiological and neurological aspects, colour perception is relative and thus basic tools are required for interpreting and quantifying the phenomenon. T h e science of colorimetry is concerned w i t h the nature of human vision and provides a basis for specifying colour. Also, the Commission Internationale de L'Eclairage ( C I E ) [57] is the governing body that sets standards for a l l the aspects of colour. O n l y those elements important for this work will be exposed here.  Different objects appear to be of certain colour by selective reflection and absorption of the incident light. T h e light could have a natural source, the sun, or could come from an artificial source. A n object would appear green i f it absorbs the incident light at higher (red respectively) and shorter (blue respectively) wavelengths and reflects predominantly the light at wavelengths corresponding to green. Similarly, an object appears black i f it absorbs most of the incident visible radiation and white i f it reflects this radiation. W h e n characterising different artificial light sources, comparison w i t h a standard white  76  Temperature  Figure 3.4: Blackbody radiation. light is often used. T h e energy of the standard white light is well balanced across the visible spectrum. T h e spectral distribution of the white light is set by standard.  Here, spectral  distribution refers to wavelength dependent variations. A parameter called Colour Rendering Index ( C R I ) describes the ability of a given illuminant to render the colour appearance of an object i n comparison w i t h the colour of the same object when illuminated by standard natural light. A C R I of 100 was assigned for natural daylight. T h e correlated colour temperature (in units of degrees K e l v i n ) is another parameter widely used to describe a certain illuminant, and is associated w i t h the blackbody concept used i n radiation analysis. A blackbody is an ideal surface that absorbs all incident radiation (more generally from the thermal region of the radiant energy spectrum which includes the visible spectrum), and will emit radiant energy that is function of wavelength when heated. M a n y sources of illumination have at their core a heated object, and this is the reason why a light source is also characterised by the temperature (in degrees Kelvin) of a blackbody that provides the closest visual color match to the particular light source. T h e colour temperature concept is better visualised i n F i g . 3.4. T h e sun can be viewed as a blackbody radiator at temperatures between 5000 K to 7500 K . A low correlated colour temperature is correspondingly assigned for illuminants w i t h most of the visible spectrum at comparatively longer wavelengths (towards red hues), and the colour temperature increases as more radiation appears at shorter wavelengths. For example, a correlated colour temperature of 3200 K is used for precalibration of consumer type video cameras as well as broadcast type cameras. Higher performance and higher cost video cameras would compen-  77  c 250-I g  I  200-1 OPTIMARC'"  1*150.1  (CRI 91; 5200K  if I|  100-  ?i  IS soc  I i  250200150-  HIGH PRESSURE SODIUM (CRI 21; 2250K)  * 100-  Ii IJ  50-  0-  400  500  L 600  700nm  Figure 3.5: Optimarc lamp technical specifications, sate for changing illumination conditions when shooting outdoors or indoors.  In this study, a white light system is used for best visualisation of thermochromic liquid crystal instrumented surfaces. The white light system consists of two 400 W Duro-Test Optimarc high-intensity discharge lamps with each lamp powered through a dedicated enclosed ballast unit.  T h e technical specifications for the Optimarc lamp are a C R I of 91 and a  lumen-per-watt efficiency of 28,000 lumens at 400 watts. The Optimarc lamp was chosen due to its high colour rendering performance.  With  a correlated colour temperature of 5200 K and a spectral distribution matching closely the standard white fight it proved to be the best choice for the application. Figure 3.5 (taken from Duro-Test O p t i m a r c technical specifications [58]) illustrates the light spectral distribution of the O p t i m a r c lamp as compared to the fight spectral distribution of a fluorescent type lamp (high pressure sodium). N a t u r a l daylight spectral distribution is also shown w i t h a white line i n both graphs. The Optimarc lamp is clearly superior to the high pressure sodium lamp especially at shorter wavelengths, and this will benefit the colour rendering i n the blue region  78  which corresponds w i t h higher temperatures on a liquid crystal instrumented surface. This will allow for the use of a wide temperature-band liquid crystal, a rather novel approach i n liquid crystal thermography.  3.2.2  Image Acquisition  Colour images can be viewed using many types of light sensing devices. T h e most suitable device has to be easily integrated into the imaging system while providing high resolution colour rendition. Various sensing devices have been tested and will be briefly described along w i t h their possible use as primary sources i n the present imaging system.  3.2.2.1  Colour Photography  The first technique investigated was colour photography with images captured on 35 m m film. A C o n t a x camera was fitted with appropriate lenses and used to photograph the colour change of a specially designed liquid crystal instrumented calibration surface. A reference coloured strip was always placed i n the field of view. This allowed for the use of a common colour datum among the captured frames. T h e temperature of the calibration surface was monitored w i t h two type K thermocouples. N o colour correction was made during developing of the  films.  The colour photographs were then digitized using a Hewlett Packard ScanJet I I C scanner and the generated 24-bit T A R G A format graphics files were stored i n an I B M compatible P C . A region of interest of 60x40 pixel between the two thermocouples was consistently sampled from every file and statistically investigated to determine the colour component means and standard deviations. T h e intensity for each of the colour components can take 256 = 2 integer values between 0 (corresponding to black, a l l channels) and 255, typical of 8  8-bit/component digitization. T h e colour at each pixel location is completely specified by its red, green and blue intensity values. T h e scatter around the corresponding mean values is represented by the calculated standard deviations. T h e standard deviation is used as a measure of accuracy of the recorded colour information. T h e smaller the standard deviations  79  are, the higher the accuracy of the colour measurement is. It was found that the uncertainty i n colour measurement varies w i t h temperature. F o r later comparison w i t h other image acquisition techniques though, only the best statistics resulted from the investigated colour photographs are reported i n F i g . 3.6. T h e same analysis can be done for any other colour representation, as discussed i n following sections of this study (hue is denned i n section 3.3 and Appendix A ) . For future reference, the mean hue values a n d standard deviations are also reported i n F i g . 3.6.  Component histograms - region of interest is 60x40 pixels 250  200  £  -i—i  i  i  T  i  i  r-  Average value:  Standard deviation:  RED  o  =128.4  =2.67[%]  RED  GREEN= 124.6  < W N = 1-85[%]  BLUE  =18.07  HUE  =41.32  < W  =3.19|%]  <W  =0-811%]  150  S O  0  1 100 50 -  _i  50  100  150  i  L  200  , . 250 Intensity  Figure 3.6: T y p i c a l histograms and statistics for colour photography.  Colour photography was found to produce good resolution images i n the carefully monitored environment. T h e m a i n drawback of the techniques lies with its output format, the colour photographs, that are not i n a form directly suitable for mass storage into a personal computer. T h e colour photographs have to be individually digitized a n d this can be a prohibitively time consuming process when a large amount of data needs to be analysed. Implementation of this technique i n the imaging system would not allow for the required  80  flexibility  and fast output of the processed results. Other methods needed to be investigated  and their performance assessed.  3.2.2.2  Video Capturing  A s an alternative to colour photography, video capturing offers the benefits of increased frame rate and more versatility on the video output format. H i g h resolution and high image quality are also possible with the appropriate choice of the video camera. Real-time video digitization can also be done through dedicated electronics so that information can be quickly stored into a host P C for further processing and manipulation. T h e main video signal standards i n use are N T S C (National Television Standards C o m mittee, used i n N o r t h A m e r i c a , Japan, K o r e a , some L a t i n A m e r i c a n countries), P A L (Phase Alternation by Line, used i n most of the Western European countries, A u s t r a l i a , C h i n a , Africa), and S E C A M (Sequentiel Couleur A v e c Memoire, used i n France and some other French speaking countries and i n most countries from the former communist block). A l though N T S C has a reputation for poor colour accuracy, this is only true for broadcast television. A s a video format, the N T S C format achieves better signal to noise performance than either P A L or S E C A M formats. This further justifies the use of a video camera working i n N T S C video format. E i t h e r an analog or a digital video camera could be used i n the imaging system. T h e choice was made considering the colour image quality and the associated costs. A t the time this work was conducted, a lower performance digital video camera was almost as expensive as a high performance analog video camera. It was then decided to further investigate the use of analog video cameras to capture the colour change of a liquid crystal instrumented surface. Previous research using thermochromic liquid crystals relied almost exclusively on C C D video cameras for colour capturing, due to their compactness and ease of use. Some fundametal aspects of the way these cameras create colour have to be understood so that their performance i n the overall imaging system can be properly evaluated.  81  C C D stands for  Serial register  Figure 3.7: Charge-coupled device - C C D . charge-coupled device and is actually an array of hundreds of thousands of light-sensitive photosites (or pixels) formed on a silicon chip. W h e n light reflected from an object i n the field of view falls on the C C D , the various photosites will be charged differently according to the amount of light they see. Charge-coupling (cycling the applied voltage) is the technique used to transfer the freed electrons from the various photosites line by line into a serial register, and off to an amplifier at the end of the serial register. A schematic of a C C D is shown i n F i g . 3.7. T h e number of pixels, or light-sensing elements, from a C C D should not be confused w i t h the number of pixels, or the resolution, of the output image. There is no direct relationship between the number of pixels on a C C D and the number of pixels i n the viewed image, typically there are significantly more pixels on the C C D than i n the final image. T h e quality and uniformity of the C C D array of pixels , together w i t h the accompanying electronics would make the difference between the consumer type and the professional type of video cameras. T h e most important technical specifications that directly affect the image quality from a video camera are the S / N ratio (Signal-to-Noise ratio) and the horizontal resolution. T h e S / N ratio expresses the amount of useful video information as compared to the  82  amount  of noise within a certain video image. T h e term noise normally characterises the random variation i n a measured or observed quantity. T h e higher the S / N ratio is, the better the image quality is. A S / N ratio of at least 42 db to 45 db is generally required for a consumer type video camera. T h e vertical resolution i n a video image is basically set by the N T S C format at 525 horizontal lines. A measure of the actual resolution i n an image is then expressed by the lines of horizontal resolution, which represent the vertical lines per picture height one would count across a horizontal line on a video image. This is the appropriate way of representing the resolution of the luminance component, while the colour resolution is always less due to the nature of the human eye response (more sensitive to the black/white signal than to the colour signal i n an image).  For example, the V H S (Video H o m e System) videotape  format is characterised by approximately 240 lines of horizontal resolution, and the S - V H S (Super-Video Home System) format can achieve around 400 lines of horizontal resolution, both referring to the luminance (black/white) component. Colour is reproduced i n a video camera through an additive process by mixing three primary colours namely red, green and blue. Processing the individual colour channel output is done by m a t r i x circuits to yield three types of signals, a luminance signal and two chrominance signals. T h e luminance signal, often referred to as the Y channel, is skimmed from a l l three colour channels, and contains a black and white representation of the viewed object. T h e two chrominance channels (or C for chrominance, or chroma) contain a l l the colour information. Furthermore, the luminance ( Y ) and chroma (C) channels could be encoded together i n what is referred to as a composite video signal, or could travel on separate wires thus producing what is referred to as the Y / C video signal, characterised by higher quality video. T h e latter method is at the heart of the S - V H S format, providing increased resolution and less inherent noise. Individual red, green and blue components can also be used i n what is referred to as a component (or R G B ) video signal. This produces the best quality of the colour images since no encoding/decoding circuitry is needed.  83  Extensive tests were conducted using a single chip Panasonic P V - S 7 7 0 - K O m n i M o v i e SV H S camcorder. Single C C D cameras produce colour with the aid of a filter system placed in front of the C C D and the signal is electronically separated into the three primaries red, green and blue. W h e n the S - V H S format is employed, single chip colour cameras can produce good quality images at a relatively low cost, and this was the motivation for testing such a camera. T h e technical specifications of this camera indicate a S / N ratio of 44 db, luminance resolution of 400 fines and colour resolution of 80 lines. T h e output video signal was recorded on tape using a S - V H S video cassette recorder ( V C R ) , and also directly into a P C through a time base corrector/digitizer board.  M o r e will be said about time base  correctors ( T B C ) and digitizers i n the following sections of this study. It was found from these tests that the colour rendition of the video camera/storing device chain was not yet satisfactory for quantitative interpretation of the thermochromic liquid crystal response w i t h temperature. Figure 3.9 on page 88 shows the best statistics of a region 60x40 pixels out off a test video sequence captured w i t h the S - V H S camera and recorded by a S - V H S V C R on S - V H S tape. Figure 3.10 on page 90 shows the best statistics of a region 60x40 pixels out off the video frames captured with the S - V H S colour camera and stored directly into the computer. These sample statistics will be used later for comparison of a l l image acquisition techniques investigated i n this study.  The final image source device selected for the application was a 3 - C C D J V C K Y - F 5 5 B colour video camera that provides a S / N ratio of 60 db, w i t h a luminance resolution of 750 lines and colour resolution of 580 lines for the output R G B video signal. T h e superior performance of this camera is mainly due to the three chip concept, high performance 380,000 pixel C C D s , and improved circuitry and optics. T h e incoming light received by the video camera lens is split by a colour separation optical prism system into the primary red, green and blue colour components that are individually focused on a corresponding C C D . Such an arrangement is illustrated i n F i g . 3.8. The advantage of having three dedicated C C D s , one for each of the colour primaries, is  84  Optical low pass filter  Figure 3.8: Colour separation optical system. fully appreciated when using the red, green, and blue generated signals which have not gone through a m a t r i x stage. M a x i m u m resolution and image quality can only be achieved i f the output R G B video signal is recorded i n the same format, and its characteristics preserved. Using a high performance 3 x C C D camera and then recording the signal onto a consumer type V C R , incapable of accurate storage and reproduction, would defeat the purpose of generating a high quality signal at the source-end of the imaging system chain. T h e method used to record the colour video signal deserves further attention, and it represents the second component of the integrated imaging system.  3.2.3  Recording and Storage Devices  Video capturing w i t h C C D camcorders (consumer type video cameras) and recording onto a video tape using consumer type V C R s , was the method almost exclusively used i n previous studies involving liquid crystal thermography. T h e motivation for using such a system lies w i t h the relatively low cost of the equipment, wide availability, ease of use, and simple integration.  Since the data has to be finally processed w i t h the use of a P C , storing the  images directly into the P C memory or hard disk is also a very attractive option. These  85  possibilities were investigated during the development and integration of the imaging system finally used i n this study.  3.2.3.1  V i d e o Tape Recorders - V i d e o Tapes  V i d e o tapes and video tape recorders ( V T R ) represent the foundation of b o t h professional and commercial video industry. V T R is the more general designation for any type of reel-toreel recorder and includes the popular video cassette recorder or V C R . A generous amount of literature is dedicated to this topic and a good description of various components was found i n [49]. Component V T R s record a l l three colour signals red, green and blue as generated by the video camera thus providing the best quality among the various video tape formats. This is also typically done i n digital format, so that the original video information is much more durable i n time and multiple copies can be produced w i t h no loss of quality. T h e m a i n drawback of using such machines is their high cost and limited availability. If the component V T R s represent the higher quality of the video tape formats, the popular V H S format is at the opposite lower end of the available formats. T h e low cost and the direct availability of consumer V C R s using this format made them the immediate choice of most researchers involved i n the past i n liquid crystal thermography.  Their main disadvantage  is low signal quality and more specifically limited colour signal recording and reproduction capabilities. These machines were simply not designed for accurate colour reproduction as needed i n typical thermochromic liquid crystal visualisation or other colour measurement applications. T h e images reproduced by a good quality V H S machine have to be only as good as a standard human observer would critically tolerate. T h e high frequency input colour information is down-converted by a factor between 5 and 6 before being recorded on tape, and is up-converted by the same factor when the tape is played back. T h e information that is lost during the down-conversion process cannot be recovered, and this is one of the reasons why the colour resolution is much less than the luminance resolution for this video tape format.  Other image quality limitations are directly related to the composite video signal  86  (luminance and chroma channels matrixed together) used as input to the V H S recorder. Separating the chroma information from the luminance information before recording on tape, and recombining the chroma and luminance signals at play back is another source of image degradation associated with the V H S format. T h e S - V H S format is able to achieve higher luminance resolution by keeping the chroma signal separated from the luminance signal ( Y / C signal). T h e colour signal though goes through the same down-conversion and up-conversion procedures on recording and play back respectively. Although the images look much sharper i n S - V H S format, the colour information i n not much better than i n the V H S format. Initial tests were conducted w i t h the S - V H S Panasonic video camera, and a Panasonic AG-1970 P r o L i n e S - V H S V C R was used to record the colour change of a instrumented liquid crystal surface on S - V H S tape. T h e recorded images were then played back on the S - V H S V C R and transferred to the dedicated P C through a Digital Processing Systems P A R - T B C I V digitizer. For comparison w i t h other investigated image acquisition techniques, the best statistics of the colour investigation on a region 60x40 pixels are reproduced i n F i g . 3.9 . T h e uncertainty i n colour measurement using S - V H S video tape recording is comparable w i t h that using colour photography, as expressed by the standard deviation values i n F i g . 3.9 and F i g . 3.6 on page 80 respectively. Higher accuracy is possible and this w i l l be addressed further i n this study. Storing the colour change history on video tapes was the method most widely used i n liquid crystal thermography i n the past.  T h e advantages are low cost, good frame rate,  ease of use and direct hardware availability. T h e main drawbacks are associated w i t h the colour resolution itself, too low for quantitative colour analysis purposes, and the unstable V C R output signal. T h e latter is due to the mechanical complexity of V C R s that makes them unable to play back a tape at exactly the same speed the tape was recorded. This is referred to as the time base error, measured i n video lines, and is a known problem i n video production since V C R s cannot be integrated with other stable sources through a switcher. T i m e base correction is needed to obtain a stable video signal and this is done w i t h a time  87  Component histograms - region of interest is 60x40 pixels 350  300  250  I  1  1  i  I  i  i  I  i  I  I  1  Average value:  Standard deviation:  RED  o  =86.61  RED  =2.56[%]  GREEN= 231.7  OGREEN = 2-25  BLUE  = 246.3  <W  =2.82[%]  HUE  =132  c  = 0.80 [%]  HUE  [%]  § 200 o o  ai x  A 150  100  § 50  % _L  50  100  150  200  Intensity  250  Figure 3 . 9 : T y p i c a l histograms and statistics for S - V H S tape recorded images. base corrector ( T B C ) . T h e unstable V C R output video signal is fed into the T B C , and the stable video signal coming out of the T B C can now be used for integration w i t h other devices. This comes w i t h a high cost since a high performance T B C is an expensive piece of equipment. Since it appeared that at least a S - V H S machine, good editing and time base correction capabilities are needed for preliminary liquid crystal thermographic tests, other options were investigated within the same budget range.  A more direct way of transfering the colour  information into the computer while achieving good colour resolution would provide a more efficient imaging system, absolute requirement for colour measurement applications.  3.2.3.2  Digitizers - Hard Drives  W i t h the increased performance of today's computer bus system and peripherals, bringing five video into the P C environment at affordable costs is only a matter of time. T h e main  88  problem the manufacturers have to overcome is the bandwidth, or the amount of information that can be transferred per unit of time as required by video applications. A s already pointed out, the best colour resolution is achieved when providing three separate channels for additive primaries red, green and blue. Colour systems that provide 8-bit digitization for each of the three components are referred to as true colour systems or 24-bit systems. A typical N T S C frame would require for true colour reproduction 900 K B ( K i l o B y t e s , 640x480x24=7,372,800 bits), and real-time video at 30 fps (frames per second) would consequently require approximately 26 M B (Mega Bytes) to be transferred i n one second. For comparison, the amount of data generated by digital audio i n the very popular C D format is about 150 times smaller (approximately 185 K B per second). Numerous compression methods (or codecs for compression/decompression algorithms) have been developed and research continues into video image compression, a method widely used to incorporate video i n P C environment. Physiological limitations of the human eye prove handy and are fully exploited i n designing new algorithms of data compression. B a s i cally, a l l audiovisual compression methods need to be patterned on human perception, thus removing the less relevant information to such an extent that the quality loss is not noticeable to a human observer. T h e compression methods can be generally split into two major types, lossless and lossy compression, whether information is lost or not. In lossless compression, as the name suggests, only the redundant information is combined and condensed and no loss of quality occurs. T h e compression factors that can be achieved this way are relatively modest and He between 1.5:1 to 3:1. In lossy compression, information that is considered not important for visual perception is actually removed from the video signal, and much higher compression factors can be achieved. Analog video information cannot be stored as such into a digital computer. T h e analog signal has to be digitized first, which means that its shape is discretely sampled and consequently processed as numbers by the computer. T h e piece of hardware that represents the link between the analog signal and the digital computer is referred to as the digitizer, or frame grabber, or capture board. Digitizers can perform much more complicated functions  89  Component histograms - region of interest is 60x40 pixels  350  i—i—i—|—i—i—  300  250  Average value:  Standard deviation:  RED  a  =103.9  RED  =2.33[%]  GREEN= 162.1  <WN  BLUE  =131.4  °BLUE  =2.39f>]  HUE  =105.3  o  = 0.78 [%]  HUE  =  2-44 [%]  § 200 o o  /  s 150 100 50 i  •  i  •  50  200  Intensity  250  Figure 3.10: T y p i c a l histograms and statistics for S - V H S P A R - d i s k recorded images. and can be specifically programmable as well. Tests were conducted w i t h a real-time video capture deck w i t h image storage directly i n the P C on dedicated disk drives. T h e system used was the Personal A n i m a t i o n Recorder ( P A R ) a n d the time base corrector T B C I V , both from Digital Processing Systems Inc. T h e Y / C video signal from the Panasonic S - V H S camera was fed directly into the P A R - T B C I V system which allows for real-time digitization and storage of the video sequence as animation files. Individual video frames were recovered as 24-bit T A R G A files, a n d the same colour investigation process was conducted on a region 60x40 pixels located between two type K thermocouples on the liquid crystal instrumented surface. F o r comparison with the other image acquisition techniques investigated i n this study, the best statistics from a l l recorded test sequences are shown i n F i g . 3.10. It can be seen that the scatter around the mean component values, represented by the corresponding standard deviations, is not much different from the best results obtained with  90  colour photography or S - V H S video tape recording shown i n F i g . 3.6 on page 80 and F i g . 3.9 on page 88 respectively. This can be explained by investigating the P A R - T B C I V system performance on real-time video data transfer.  In order to manage the large amount of  information, the system uses a lossy compression algorithm. In the best case, a compression factor of at least 5:1 is achieved at the expense of high-frequency colour information. This is of course, not much better than the colour down-conversion process used i n video tape recording. T h e reconstructed images look acceptable when viewed by the human eye, but the lost information can be quantified i n the colour measurement process. A system able to transfer i n real-time a l l the required uncompressed R G B video data would represent an improvement of the status quo as far as colour measurement based liquid crystal thermography is concerned. This can be done w i t h digitizers taking full advantage of improved sustained P C I bus transfer rates. P C I stands for Peripheral Component Interconnect and represents the shared line of communication between the host C P U , memory, display boards, frame grabbers, and other peripherals, the heart of traditional P C architecture. A 32-bit wide P C I bus running at 33 M H z can theoretically transfer up to 132 M B / s .  T h e frame grabber finally chosen for the application is a M a t r o x M e t e o r / R G B colour digitizer w i t h transfer rate capabilities of up to 35 M B / s i n N T S C format. This means that a 32-bit colour input data i n the form R G B x (x represents an extra channel) at a resolution of 640x480 pixel and a frame rate of 30 fps can be transferred i n real-time to either the host computer memory or display. The display is achieved real-time using a M a t r o x M i l l e n i u m M G A display board w i t h 4 M B on-board memory. T h e host P C is carefully specified to perform the sustained P C I transfer rates to system memory. After the video sequence is transferred and stored into the host R A M it can be then saved on a hard disk at slower transfer speeds. Direct transfer and storage onto fast dedicated hard drives would have been possible at higher cost.  91  3.2.4  Imaging System Specifications  A new imaging system was developed and implemented for use i n liquid crystal based thermographic applications. T h e imaging system is based on real-time true colour video capturing and recording on computer R A M with sequential storage onto a hard drive. T h e system is characterised by excellent colour rendering, high colour resolution, high quality colour signal, m i n i m u m required hardware, and direct data availability i n digital format for processing and measurement. T h e lighting is produced by a white light system for best colour rendering, the colour capturing is done by a 3 x C C D R G B video camera, the R G B video signal is transferred uncompressed i n real-time into the host P C R A M through a 32-bit colour frame grabber, visualisation is done real-time on computer monitor through a dedicated 4 M B W R A M display board. T h e host P C is an Intel P e n t i u m 200 M H z processor based Windows N T machine w i t h 256 M B of R A M . A schematic of the imaging system is shown i n F i g . 3.11. T h e imaging software used to control the video sequence capturing, file conversion, image processing and analysis is M a t r o x Inspector-32 version 2.1. Scripting is probably the most important feature of the Inspector. Scripting allows the developer to customize a particular application by writing specific programs to automate complicated repetitive tasks.  The  programming language used within the Inspector has close similarities w i t h the more popular C language. Inspector also supports multiple file formats including T A R G A , T I F F , J P E G and more. A n additional software package, M a t r o x Intellicam, can be used as an interactive utility for interfacing various types of cameras.  A specific digitizer configuration format  ( D C F ) file is required for each attached video source. A memory buffer has to be allocated for real-time data transfer upon installation of the imaging software, and can be further modified by calls to the Meteor driver. T h i s procedure proved unreliable and conflicted w i t h the good operation of the overall system. T h i s again shows the importance, sensitivity and complex interaction of every piece of hardware involved. Perfect integration among the various devices is required to achieve the desired performance of the imaging system. T h e memory buffer allocation allows for capturing of  92  3xCCD  DC  R ir  t  Sync  B  Display board  t  i  4>  Triple channel 8-bit A/D  Lumin 8-bit /  « la  M i l l  I  PCI Bus  S  4MB WRAM  >> «3  M I  Hard Drive  I I II II  I  (32bit x 33MHz = 132MB/s)  Figure 3.11: Block diagram of the imaging system. a m a x i m u m 92 frames real-time. Direct transfer to a dedicated partition on the host hard drive can be done at 3 fps only but for a longer period of time. Creating a memory partition to act as a R A M disk is another option and transfers of up to 10 fps can be achieved i n this way. This newly developed imaging system can be used to accurately analyse the colour change history of a liquid crystal instrumented surface. The colour information is only useful i n a three-temperature  experiment i f converted into temperature information. T h e exact rela-  tionship between the observed colour and the temperature of the surface must be identified.  3.3  Tristimulus Colour Space  There are three types of light receptors i n the human eye responsible for daylight vision, the cone cells, while a fourth type of receptor, the rod cells, plays a more important role only at very low light levels, or night vision. T h e three colour cone receptor cells each have a different spectral response to the incident radiation that peak i n the red, green, and blue regions of the visible spectrum respectively. Since there are three types of colour photoreceptor cells,  93  Figure 3.12: A d d i t i v e system with R e d , Green, and B l u e primaries. three numerical components are necessary and sufficient to specify any colour, and this is the basis of the trichromatic theory of colour vision. T h e three components are referred to as primaries, or colour stimuli, and represent the coordinates of a tristimulus colour space. T h e requirement for the primaries is that no single one of them can be produced by a combination of the other two. In a tristimulus colour space, any colour can be produced by an additive mixture of accordingly weighted primaries. Several colour coordinate systems have been standardised by C I E to accomodate the needs for colour specification of different practical applications. Colour space conversions are merely linear transformations and are directly available i n C I E specifications. T h e implementation of such linear transformations are also beneficial from a computational point of view, being fast to compute. T h e largest colour range (gamut) is obtained i n additive reproduction using red, green, and blue as primaries. T h i s is at the basis of the C I E ' s colour matching tests that produced the R G B colour space corresponding to monochromatic primaries red (with A = 700 nm), green (with A = 546.1 n m ) , and blue (with A = 435.8 n m ) . E q u a l amounts of red, green, and blue produce white (all colours) at a specified luminance level as shown i n F i g . 3.12. A combination of red and green produces yellow, green plus blue produces cyan, and blue plus red produces magenta. Yellow, cyan, and magenta represent complementary colours to blue, red, and green respectively. Black would represent the lack of any colour. A direct  94  application of the R G B colour space is used i n television for colour display on cathod ray tube ( C R T ) monitors. T h e N T S C ' s R G B colour space is obtained through a linear transformation from the C I E ' s R G B colour space. A system that uses cyan, magenta, and yellow ( C M Y ) as primaries is also widely used i n colour printing. Other colour spaces have been produced to better match the human visual perception of colour. There are three common attributes that describe the human response to light known as hue, saturation and brightness. T h e hue of the light is the attribute that distinguishes the colour itself, for example red, yellow or blue. W h e n the light source is monochromatic (as the C I E ' s R G B primaries) the hue is an indication of the wavelength of the light. T h e saturation is a measure of the purity of a colour, or a measure of the amount of white light added to a pure colour. P u r e colours are also referred to as spectral colours for which the corresponding spectral power distribution is concentrated at one wavelength. Brightness is a more subjective and complex attribute according to which an area appears to emit more or less light. For measurement purposes, the C I E defined another quantity for brightness called luminance and denoted Y . T h e magnitude of luminance is now proportional to physical power and its spectral composition is related to the luminous efficiency of a standard human observer. Yet another quantity is defined by C I E to account for the nonlinear human perceptual response to luminance, called lightness. Lightness can be referred to as the relative brightness since an area would appear to emit more or less fight i n proportion to a similarly illuminated area perceived as white. A fundamental reference for a l l aspects of colour was found i n [55]. In common language brightness, luminance, intensity, and lightness a l l refer to a black and white version of the image, while hue and saturation contain the chrominance information of the image. Several colour models have been developed to accomodate the visual perception of hue, saturation, and brightness. One such system is the H S L colour space, standing for hue, saturation, and lightness as shown i n F i g . 3.13. Lightness is a term normally used i n discussing light reflection and it is used somewhat loosely here. In the double hexcone arrangement, the lightness is represented on the vertical axis  95  i i LIGHTNESS  WHITE  A\\\\ \ \ 1  /// /// // / / //  \\ \ \ \ \\ v  x  / n v>\ \/ \\ o  f  f  \ A I \  W /\ \B\ \ w/ \ \ \ \  i  I I  /  > R  A  / / \ / /  Lu 7  \\ \ \\ \ \\\  BLACK  \\\ \^  HUE  0 '  Figure 3.13: The double hexcone H S L colour space. ranging from 0 for black, at the lower tip, to 1 for white at the upper tip of the double hexcone. Hue is represented by the angle around the vertical axis w i t h red at 0° and the other colours i n the spectrum appear when the double hexcone is traversed counterclockwise. The saturation is simply measured radially from the vertical axis w i t h 0 on the axis and 1 on the surface corresponding to fully saturated colours. For 8-bit digitization the corresponding range will be from 0 to 255 for each colour coordinate.  The R G B to H S L colour space  conversion algorithm is described i n A p p e n d i x A . B o t h R G B to H S L and H S L to R G B colour space conversions are available w i t h the M a t r o x Inspector-32 imaging software. M o r e conversion algorithms can be found i n [41]. For the present application, an appropriate relationship must be found between the temperature of a liquid crystal instrumented surface and either the R G B or H S L colour spaces. This function must be easy to compute and must minimize the error i n temperature over the entire liquid crystal useful colour response. O n l y when the exact form of this relationship is known, for a given liquid crystal setup under known lighting conditions, can the thermochromic liquid crystal become a powerful tool for thermal mapping.  96  3.4  Calibration of the Liquid Crystal  This section presents the calibration of the thermochromic liquid crystal and video camera system used i n the wind tunnel experiments described i n section 4 of this study. T h e wideband liquid crystal used i n this study was custom manufactured by Thermax-Thermographic Measurements Inc. [60]. T h e same lighting conditions and video camera setup must be used for both calibration of the T L C and actual experiments. T h e incident light on the T L C instrumented surface must be uniform and such that the surface would appear without gloss. A pair of white light lamps was placed symmetrically above the liquid crystal instrumented surface as shown i n A p p e n d i x B , and their relative position and distance to the surface was adjusted u n t i l the above conditions were satisfied. N o measurements were taken until the lamps reached their normal operating temperature. Light uniformity was checked by shooting a white piece of paper placed on the surface of interest within the field of view. T h e position of the lamps was adjusted until a uniform colour signal i n the R , G , and B components was achieved across the white subject. T h e white balance of the video camera was adjusted for the actual lighting conditions, and the white paint data was stored i n the video camera's memory. A master white still image was then captured and stored on the hard drive. This procedure ensured that the settings used for T L C calibration were restored for all runs based on that calibration. Possible electromagnetic interference on the video signal from the noisy wind tunnel environment (wind tunnel motor, in-line heater, heater coil, and lamp ballast units as described in section 4.3), as well as possible induced C C D thermal noise, were quantified by shooting a sequence of dark images w i t h the lens covered. After digitization, noisy pixels would apper to have R , G , B values different than 0, on a 0 to 255 range. T h e averaged results showed that on the R channel 99.994% of the pixels were clean, on the G channel 99.971% of the pixels were clean, and on the B channel 98.547% of the pixels were clean. T h e noisiest pixels had a value of 2, on a 0 to 255 range. The results were considered satisfactory and no further noise correction was made on any of the recorded images.  97  TLC  Figure 3.14: T L C calibration setup. A geometrical calibration was also made by measuring the T L C instrumented patch i n physical units. T h e resultant resolution was 3.23 pixel for each millimetre unit length. T h e calibration of the T h e r m a x wide-band T L C [60] was done i n the same w i n d tunnel, and under the same lighting and viewing angle conditions as the actual experiments described i n section 4 of this study. A specially designed hollow rectangular calibration box was used for this purpose. T h e rectangular box was made of aluminum and was insulated i n a 2 inch thick layer of Styrofoam on all but one of its sides. T h e exposed side was machine polished and treated by the method described i n section 3.1. A type K thermocouple connected to an O M E G A O M N I - C A L unit was mounted on the surface before the black paint and the T L C were sprayed using an air brush. To achieve the same T L C layer thickness on the calibration box and on the flat plate test surfaces, the two were simultaneously treated and sprayed. T h e hollow calibration box was connected i n a closed loop to a L A U D A - B R I N K M A N K 4 R refrigerated circulator using distilled water as the circulator agent, as shown i n F i g . 3.14. T h e temperature of the bath can be set w i t h a thermoregulator and the bath heat input rates can also be adjusted by a wattage proportioning electronic relay. A slow temperature increase of the circulator agent was achieved i n this way, allowing for a uniform temperature on the T L C instrumented calibration surface. T h e temperature of the surface was known at a l l times, being digitally displayed by the O M N I - C A L unit. Still images were recorded directly on the dedicated imaging system P C ' s hard drive for each 0.2°C temperature increase from 1 8 ° C to 4 1 ° C A region 60x40 pixels around the thermocouple leads was then sampled from each image and analysed. T h e wide-band T L C ' s colour response w i t h temperature is  98  I  1  1  1  1  I  1  I  I  1  1  ' I  1  1  1  1  I '  1  ' '  I  1  1  1  1  I  Figure 3.15: T L C colour response with temperature. shown i n F i g . 3.15, i n both R G B and H S L colour spaces. T h e melting point of the T L C was between 1 9 ° C and 2 0 ° C , and the clearing point was around 4 1 ° C . N o clear relationship between the colour displayed by the T L C and the corresponding temperature was identified using the R G B colour space. In the derived H S L colour space, more clear dependencies can be identified between the colour components and temperature.  T h e lightness (L) component shows the same type of dependency w i t h  temperature as the R , G , and B signals. T h e lightness signal peaks at around 2 1 ° C and has a monotonic behaviour below and above this temperature. A possible lightness versus temperature relationship is thus identified for the temperature range T > 2 1 ° C B o t h the hue (H) and the saturation (S) components have a m i n i m u m at around 2 0 . 6 ° C , and are monotonic w i t h temperature above this value. B o t h H and S components appear to flatten out at temperature values around 3 0 ° C . A possible hue, or saturation, versus temperature relationship was therefore identified for the temperature range between 2 0 . 6 ° C and 3 0 ° C T h e final choice for the use of hue or saturation was made after investigation of the absolute error i n H , S, L , as a function of temperature, resulting from multiple calibration  99  I  I  I  1  I  I  I  I  I  I  I  I  I  I  I  I  I  I  -i—i—|—i—i—i—i—|—i—i—i—r—|—p-  I  [%]  SATURATION  -l  22  23  24  l  I  l  _L_J_  l  25  26  27  29  28  30  T[°C]  Figure 3.16: Statistics for real-time true colour recording. data sets. T h e T L C calibration was repeated seven times, under the same lighting conditions, for a temperature range between 20°C and 3 0 ° C w i t h a temperature increment of 0 . 2 ° C . Standard deviation values computed from the calibration data set for a region of 60x40 pixels are presented i n F i g . 3.16. T h e uncertainty values corresponding to the hue component are about 50% less than those from colour photography, S V H S tape recording, and S V H S disk recording shown i n F i g . 3.6 on page 80, F i g . 3.9 on page 88, and F i g . 3.10 on page 90 respectively. The absolute error i n any of the H , S, or L components was calculated, for any temperature, as the m a x i m u m variation about the mean value, according to E q n . 3.1:  e  H  = ma [H(T)]- n[H(T)] mean[H(T)\ X  r  x  m  %  (3.1)  The error values from multiple calibrations are plotted against temperature i n F i g . 3.17. The m i n i m u m error was found to correspond to the H component, and a useful temperature  100  Figure 3.17: T L C calibration accuracy. range was indentified between 2 1 . 6 ° C and 3 0 ° C The error i n the H component was between 3.6% and 1% for temperatures between 2 1 . 6 ° C and 2 4 . 1 ° C , dropped between 1% and 0.5% for temperatures between 2 4 . 1 ° C and 2 6 . 6 ° C , and was below 0.5% for temperatures between 2 6 . 6 ° C and 3 0 ° C . T h e hue component was consequently used to obtain the desired colour temperature relationship. A n eighth order polynomial was used to compute the temperature of the T L C instrumented surface based on the calculated hue colour component. A n absolute error i n temperature could be then calculated based on the calculated absolute error i n hue from multiple calibrations, and the result is also shown i n F i g . 3.17 for the temperature range between 2 1 . 6 ° C and 3 0 ° C .  For this temperature interval, the error i n temperature is as low as  ± 0 . 1 ° C and as high as ± 0 . 2 5 ° C . T h e use of multiple calibration data sets produced the needed relationship between the colour displayed by the wide-band T L C instrumented surface and the corresponding surface temperature. T h e hue colour component was used as the independent variable i n an eighth order polynomial calibration function, w i t h temperature as the dependent variable. T h e  101  associated error i n temperature is on average ± 0 . 1 5 5 ° C , or ± 0 . 6 % , for the range between 2 1 . 6 ° C to 3 0 ° C . This compares well w i t h the results from [53] where a temperature error less than 0.3°C was reported for a 1 5 ° C band liquid crystal used i n a two-temperature situation, and using a different hue, saturation, intensity (HSI) colour space transformation. Script programs were developed within the Inspector-32 environment to automate a l l the above procedures. Finally, the calibration function was incorporated i n a software routine and used to map out thermal regions of interest based on resolved pixel hue input data.  3.5  Summary  T h e colour change properties of a wide-band thermochromic liquid crystal were used to determine the temperature of a T L C instrumented surface. A high performance imaging system which includes a white light system, a component colour video camera, a 32-bit colour frame grabber, and a host computer was developed and implemented. This imaging system provides higher accuracy of the colour representation than colour photography, S - V H S (Super-Video Home System) tape recording, and S - V H S disk recording. T h e hue component from the R G B to H S L colour space transformation was used i n the colour calibration of the wide-band thermochromic liquid crystal. T h e uncertainty values of the hue component, which are based on calculated standard deviation for a region of 60x40 pixels, are about 50% less for the new imaging system than for the other imaging systems investigated. Based on the calibrated hue/temperature relationship, the temperature history of the T L C instrumented surface can be determined from the recorded colour change of the liquid crystal.  102  Chapter 4 Compound Angle Jets in Crossflow This section describes the validation of the newly developed imaging system i n a threetemperature application. Integration and development of appropriate software was required to compute the parameters of interest, the film cooling effectiveness (77) and the film heat transfer coefficient (hf).  The geometry is such that the transient heat conduction through  the wall material can be assumed one-dimensional. T h e surface temperature history will then suffice to determine the heat transfer characteristics i n a suitable form. Comparison of film cooling effectiveness data obtained by liquid crystal thermography w i t h data obtained by a flame ionization detector technique by F i n d l a y [17] is used to validate the method.  4.1  Transient Heat Transfer  T h e calculation of the heat transfer coefficient directly from E q n . 1.3 on page 10 requires the measurement of the heat flux q , temperature of the surface T , and temperature of the w  w  adiabatic wall T j. This is impossible to achieve i n a single experiment since measurements on a  an adiabatic surface (T  w  = T d) would provide no information on the heat transfer coefficient. a  T h e flow field i n an arbitrary three-temperature  situation is three-dimensional, and the  resulting adiabatic wall temperature distribution varies significantly on the surface of interest. In the past, the effort was often focused on determination of the adiabatic wall temperature i n  103  OO  Too  Uoo  q  —-  1.  M/ /  'z /u  2  2  Iconv  rad|  T  X  w  1 *lcond  /  Ti  /  - T  Figure 4.1: General three-temperature arrangement - nomenclature. the form of film cooling effectiveness, and this is especially true i f the heat transfer coefficient hf was considered to be close to the value without injection h . T h e comprehensive study by 0  Goldstein [19] addresses this topic with emphasis on two-dimensional film cooling situations. To avoid the complications induced by imposing adiabatic wall conditions, a different method initially proposed by Metzger et al. [35] uses heat transfer data under different isothermal wall conditions to specify the film cooling performance.  T h e two methods are based on  the linearity of the energy equation i n temperature for constant property flows, and thus allow for the result to be obtained by superposition of other solutions. These methods were compared and analysed i n a study by Eckert [11]. M o s t of the previous work was carried out i n low speed wind tunnels w i t h properly scaled models to keep at least the Reynolds number within the range used i n gas turbine applications. T h e solution to the heat transfer problem is much simplified on larger wind tunnel models i f two- or three-dimensional effects are not significant. A one-dimensional formulation would clearly prove advantageous from a computational point of view. A typical three-temperature situation is summarized i n F i g . 4.1 showing a transient temperature profile into a slab. T h e temperature of the jet flow (T ) is taken to be different from the temperature of 2  the mainstream flow  (Too).  For a given geometry and generated flow field, the surface  temperature downstream of the injection sites becomes a function of the location on the surface and time T (x, y , t). T h e conservation of energy requirements at the interface between w  104  the surface and the fluid can be written for both steady-state and transient conditions i n the form:  Qcond = Qconv + Qrad  (4-1)  where the analysis is on a unit area basis [ W / m ] with no heat generation and w i t h q d 2  con  being the heat flux by conduction from the control surface to the wall m e d i u m , q  conv  the  heat flux by convection from the fluid to the control surface, and q d the radiation exchange ra  from the surroundings to the surface. Steady-state conditions would require that a l l parameters be independent of time. This condition is difficult to realize i n a heat transfer experiment w i t h a strong three-dimensional flow field and w i t h a large scale model. Steady-state heat transfer experiments are usually associated w i t h uniform boundary conditions i n the form of either uniform wall heat flux or uniform wall temperature. A s research i n the past has shown [28], [24], departure from uniform conditions is accounted for through heat losses and are always somehow hard to quantify. Reaching steady-state conditions is also a time-consuming procedure, and could prove quite expensive when large wind tunnels are used. A transient heat transfer technique on the other hand, would allow for the time dependence of the temperature field, and the actual experiments could be completed i n very short periods of time. To obtain useful correlations of the desired parameters many runs are needed, and so the transient techniques have become increasingly popular among experimentalists. Such an approach is therefore investigated further i n this study. In a first approximation it is considered that heat losses through radiation i n E q n . 4.1 are small, so that the radiation term can be neglected (q d = 0). This is a reasonable ra  assumption since the temperature difference between the surface and the surroundings is small, potentially less than 2 0 ° C T h e energy balance at a particular location on the surface requires then that the heat flux by convection be equal to the heat flux by conduction through the wall.  There is now coupling between the external heat transfer and the conduction  within the solid wall.  T h e physical properties of the solid wall must then be part of the  105  solution. W i t h the system initially i n thermal equilibrium at temperature T = T j , transient conduction occurs within the solid when a sudden change i n temperature is imposed at the wall. Another simplification is found to be useful here, and the assumption is that there is heat transfer i n the solid wall only i n the positive z-direction shown i n F i g . 4.1, and that the wall is semi-infinite i n that direction. This idealization is actually very useful i n many practical applications where the temperatures within the wall, far from the surface, are not immediately influenced by the change i n the surface conditions. T h e semi-infinite solid formulation is then a convenient approach to the transient heat conduction i f the thermally affected region is less than the wall thickness. This situation can always be controlled by limiting the duration of the experiment. T h e transient one-dimensional heat conduction i n the assumed semi-infinite solid wall is governed by the equation:  where k is the thermal conductivity [ W / ( m K ) ] assumed constant i n the wall, p is the wall density [ K g / m ] , c is the specific heat [ J / ( K g K ) ] assumed to be a constant for the wall 3  material, and T is the temperature distribution that is function of location a n d time T = T(z,t). The transport and thermodynamic properties of the solid wall are often linked i n one important parameter, the thermal diffusivity a with units of [m /s]. The thermal diffusivity 2  is the ratio of the thermal conductivity k to the product pc, the volumetric heat capacity of the wall. For transient heat conduction experiments, materials w i t h low thermal diffusivity are useful since they take longer time to reach a new equilibrium condition. Before attempting to solve E q n . 4.2, initial and boundary conditions have to be specified. The initial condition is:  T(z,t = 0)=Ti  (4.3)  and for the boundary conditions instantaneously applied at t = 0 there are two equations to  106  describe the interior boundary and the surface convection at any time t > 0. T h e interior boundary condition is expressed i n the form:  (4.4)  T ( z = o o , i ) = Ti and the surface convection condition is expressed by the energy balance requirement:  ,dT dz  = hf(T f  - T {t))  re  (4.5)  w  z=0  where hf and T f are constants depending only on the flow field and are independent of the Te  transient temperature of the wall. A closed form solution exists for E q n . 4.2 subject to the boundary conditions (4.3), (4.4), and (4.5) i n the form:  T(z,t)-Ti , f z \ (zh h)ctt\ ( hfVcri) 2, * = erfc — = - exp \ - r + -f^- erfc — = + — T -Ti \2y/c7t) I A; fe J {2y/cd k J f  ;  z  1  2  ref  . . (4.6) ' t  A  {  Equation 4.6 was adapted from [6], [25]. The solution contains the Gaussian complementary error function denned as:  erfc(u) = 1  —  erf(it)  2  t  u  •=. / exp(—w )dw  = 1  y T T  (4-7)  2  Jo  with values between unity and zero, for values of u between zero and infinity respectively. The surface temperature distribution T — T (t) can be simply determined from E q n . 4.6 w  making the appropriate substitution for the spatial coordinate, z = 0. A t any time during the transient and at any particular location on the surface of interest, the following expression relates the film heat transfer coefficient hf and the reference temperature T / = T d with r e  a  the initial equilibrium temperature condition T i , temperature of the surface T , and physical w  properties of the wall material a :  107  T h e form of E q n . 4.8, which describes a three-temperature situation, is very similar to the solution to the more common two-temperature situation [6], [25], where T d = Too and a  the heat transfer coefficient is h as defined by E q n . 1.4 on page 10. T h e major difficulty i n 0  a film cooling application lies i n the fact that the reference, or adiabatic wall, temperature is unknown and is function of the location on the surface, mainstream and jet flow temperatures, and the degree of mixing between them. For each known location on the surface, E q n . 4.8 has therefore two unknowns, the film heat transfer coefficient and the adiabatic wall temperature. T h e latter can be reduced by the use of E q n . 1.6 on page 12 so that the denominator i n the left hand side of E q n . 4.8 now reads:  T  -Ti  ad  = rj(T - To.) + (Too - T<) 2  and the unknowns conveniently become the film heat transfer coefficient hf and the  (4.9) film  cooling effectiveness rj. T h e experiment has to produce only the value of the surface temperature during the transient T (t) w  at a l l desired locations on the surface of interest. T w o  equations i n the form of E q n . 4.8 can be written for each point on the surface and at different times during the transient t = t and t = tb, to provide the two needed relationships for the a  simultaneous calculation of hf and 77 at that location. Theoretically, any two time events can be chosen from the transient history during one test, or two separate tests can be conducted w i t h the same flow field but different flow temperatures to calculate a time independent heat transfer coefficient and film cooling effectiveness. T h e latter technique has been used i n the past [51], [12], [13] i n three-temperature situations, and can be categorized as double sampling or double event procedure. It is noted at this point that a double event procedure provides the m i n i m u m required information for calculating hf and 77 i n a three-temperature situation, while a single event procedure could provide a l l the necessary information for calculation of the heat transfer coefficient ( E q n . 1.4 on page 10) i n a two-temperature situation. In typical jets i n crossflow applications, unsteady vortex structures were often observed downstream of the injection sites [2], [21]. In such situations, the double sampling tech-  108  niques, as used i n the past, cannot lead to an accurate representation of the heat transfer characteristics. M u l t i p l e sampling has to be considered i n these cases, and the full time history of the surface temperature has to be available for data reduction. Locally time averaged hf and 77 could then provide higher accuracy i n the representation of the heat transfer process on the surface of interest. E v e n i n a perfectly steady flow field, the multiple sampling technique can improve the accuracy by noise minimization through the time averaging process. T h e t e r m noise is used here to describe random errors i n temperature measurement.  This study will further focus on a multiple sampling technique related to the transient heat conduction process with convective boundary conditions i n a film cooling, threetemperature situation. This new technique has also the advantage that 77 and hf can be calculated simultaneously during a single test, while two separate tests were needed i n the past [51], [12], [13]. T h e goal is to develop a low cost, efficient method for calculation of the film heat transfer coefficient hf and the film cooling effectiveness 77 i n arbitrary film cooling arrangements. The underlying assumptions include: small temperature differences and constant property flows; small radiation and buoyancy effects; transient one-dimensional heat conduction response within the solid wall for points close to the surface, and a sudden change i n surface conditions. These assumptions will be addressed i n subsequent sections of this study.  4.2  Jet Temperature Data Acquisition  The solution to the transient impulsive one-dimensional heat conduction problem w i t h convective boundary conditions is described by E q n . 4.8 on page 107 for the three-temperature situation. T h e impulsive condition on temperature is required i n either the mainstream or the jet flow. T h e temperature history of the surface of interest (T ) w  is given by the colour  change of the applied liquid crystal, while the temperature of the mainstream (Too) and the temperature of the jet flow (T2) are monitored by thermocouples. In this study, the initial  109  condition (Ti) corresponds to a l l temperatures being equal, T^ = T = T, and the impulsive 2  temperature change is imposed on the jet flow. If a step change i n T2 is not possible, the whole jet temperature history is needed and has to be recorded during the transient. A temperature data acquisition system was implemented consisting of two type K thermocouples connected to a specially designed amplifier/cold j u n c t i o n compensator printed circuit board. T h e circuit was built around two Analog Devices A D 5 9 5 [56] monolithic thermocouple amplifiers w i t h cold j u n c t i o n compensation working i n dual supply operation. T h e A D 5 9 5 output voltage is linear with temperature i n the range of interest from T — 1 0 ° C to T = 40°C at 10 m V / ° C . T h e thermocouple circuit output voltage was differentially applied at the input of a National Instruments L a b - P C + data acquisition board installed i n an I B M compatible 386 P C , to minimize the common-mode noise voltage. T h e voltage carrying wires were also shielded and then grounded to the P C casing. Excellent noise rejection was achieved w i t h this setup. D a t a acquisition was controlled through specialized Labtech Notebook version 8.1 software. Temperatures were sampled at 200 H z to allow for further noise cancellation by appropriate averaging, and were recorded directly onto the P C hard drive for the entire duration of the transient experiment.  4.3  Wind Tunnel Setup  T h e experiments were performed i n the Aerodynamics Laboratory at the University of B r i t i s h Columbia. T h e wind tunnel used was referred to as the B l u e wind tunnel and is an opencircuit forced-draft tunnel. T h e crossflow was driven by a fan coupled to a 3.8 K W D . C . motor. T h e m a x i m u m tunnel air speed was 12 m / s and was measured by P i t o t tube. T h e test section was 1020 m m x 405 m m x 270 m m and was located downstream of a 4:1 area contraction section. T h e wind tunnel side walls were made of clear plexiglass for observation purposes. A boundary layer trip wire was fixed onto the tunnel floor just upstream of the test section to ensure a boundary layer whose thickness was constant across the span of the tunnel and was also steady w i t h a fixed transition point.  110  T h e test section was modified to include a specially designed removable tunnel floor, and a removable non-glare glass tunnel ceiling. T h e removable floor is basically a flat plate made of | inch thick Plastic G sheet w i t h low thermal conductivity properties. T h e physical properties of interest of the Plastic G are density p = 1187.6 K g / m , thermal conductivity 3  k = 0.1872 W / ( m K ) , and specific heat c = 1464 J / ( K g K ) . T h e removable plate contained a row of 7 square jets of size (the symbol d - diameter is used for size, from analogy with a round jet) d — | inch angled at 30° w i t h the tunnel floor and at 45° w i t h the crossflow. O n l y the middle 5 jets were used i n the experiments, and the marginal 2 jets were taped over. The row of jets was located 406 m m from the test section entrance. T h e length of the jet duct was 4 times its diameter, so L/d = 4. T h e jets were spaced 3 jet diameters apart i n the spanwise direction so that S = 3d. A rectangular patch of T  dimensions around 6 d x l 0 d was instrumented w i t h a wide-band liquid crystal, by the method described i n section 3.1 of this study, and covered the region downstream of the middle jet. T h e T h e r m a x wide-band thermochromic liquid crystal [60] was sprayed onto the surface using an air brush. T h e instrumented liquid crystal surface was illuminated w i t h the white light system described i n section 3.2.1. T w o O p t i m a r c lamps were symmetrically placed above the tunnel ceiling, and the incident light onto the instrumented liquid crystal surface came uniformly through the tunnel non-glare glass ceiling. T h e colour video camera was placed outside the tunnel, perpendicular to the centre point of the liquid crystal instrumented patch, such that the camera lens was perfectly alligned with the tunnel ceiling and shot through an opening i n the non-glare glass ceiling. Details of this arrangement are shown i n F i g . 4.2. T h e removable flat plate fitted tight on top of a metal cylindrical-plenum rectangular collar, and the clearance became hermetically sealed when the plate was screwed down onto the collar. T h e r m a l insulation was attached on the plenum side of the removable plate to avoid heating of the plate by the plenum flow. A stove heater coil was placed inside the plenum on top of a metal radial diffuser plate, the electrical wires being fed through a small sealed hole i n the bottom plate of the plenum. A honeycomb was also inserted i n the plenum to prevent a swirling flow pattern inside the plenum. The metal plenum was further wrapped  111  3xCCD white light system  H  lOd  Crossflow  TLC  /o\  A-A  instrumented  !  3d  patch  Plenum  T h e r m a l insulation  JetS  Figure 4.2: W i n d tunnel arrangement. i n thermal insulation along its entire length to minimize the thermal radiation losses from the heated plenum to the surroundings. T h e jet flow was supplied by a rotary screw Quincy compressor, filtered through multiple oil traps, metered w i t h the aid of a rotameter such that the jet Reynolds number was kept constant at a value of Re  2  = 5000. The jet air was heated i n an in-house designed and built  in-line heat exchanger, diverted by a system of two-way direct-lift solenoid valves so that it flowed through the plenum, and finally discharged into the mainstream boundary layer through the row of compound angle ducts. T h e temperature of the jets was recorded at a sample rate of 200 H z by two type K thermocouples w i t h the setup described i n section 4.2. T h e two thermocouples were placed just inside of the two marginal square ducts out of the five square ducts used, so that the flow through the middle jets (used for thermal mapping of the region downstream) was not disturbed. T h e final jet temperature was found by averaging the data recorded from the two thermocouples. A block diagram of the entire setup is shown in Appendix B . W i t h a l l the hardware and software i n place, the experiments proceed i n the following sequence:  112  1. T h e white light system is turned on. T h e air line is opened and the flow rate is set w i t h the use of the installed pressure regulators. T h e wind tunnel motor is turned on and some arbitrary air speed is set for ventilation of the test section area. 2. T h e removable flat plate is unscrewed from the plenum collar and lifted i n the middle of the wind tunnel test section, alligned w i t h the mainstream flow. T h e plenum discharge is partially obstructed by a removable piece of Plastic G sheet, to avoid the flow of the colder mainstream into the plenum. 3. T h e power to the in-line heater is turned on, as well as the power to the inside plenum heater coil. A variac is used to adjust the heater power. 4. T h e hot air flow is routed through the metal plenum until a relatively uniform temperature of the plenum metal skin is reached, and the plenum discharge air temperature is around 40° C. T h e discharge temperature is measured by another type K thermocouple connected to an O M E G A O M N I - C A L calibration unit. T h e jet air flow rate is checked and eventually corrected. T h e tunnel air speed is set according to the desired mass flow ratio M = (pU)2/(pU)oo-  T h e time required to achieve these conditions is  anywhere from 60 to 90 minutes. 5. T h e imaging system and the temperature data aquisition system are now initialized. The transient is set for 30 seconds, w i t h colour image capture at 5 fps. A total of 150 frames are captured during each transient.  T h e jet flow is switched by two solenoid  valves and routed to the surroundings. 6. T h e duct exits are covered during this period to avoid entrainment of the plenum air by the mainstream. T h e plenum discharge obstruction plate is removed and the liquid crystal instrumented flat plate is lowered and screwed onto the plenum collar.  The  duct exits are now uncovered and the tunnel lateral window is sealed. 7. T h e image sequence capture and temperature data acquisition software are started simultaneously w i t h the jet flow being switched and routed through the plenum.  113  8. A t the end of the run the data are stored onto hard drives for further processing. T h e heater power is turned off and jet air is directed through the system until it has cooled. A new r u n follows the same procedure and can be repeated after a couple of hours when the instrumented flat plate temperature has cooled to the mainstream temperature. Values of film cooling effectiveness 77 are available from the flame ionization detector technique study by F i n d l a y [17], using the same jet geometry and the same wind tunnel at the University of B r i t i s h Columbia. For direct comparison of 77, the flow conditions and the mass flow rates used i n this validation study match closely the conditions i n [17]. Three mass flow ratios were used M = 0 . 5 , 1.0, and 1.5. In summary, a wind tunnel setup is described for the investigation of three-temperature situations by a transient heat conduction technique. T h e slab temperature was equal to the temperature of the mainstream at the beginning of the transient, and the required boundary condition of sudden temperature change was provided by heating the jet  flow.  T h e jet  Reynolds number was kept constant at a value of Re = 5000 i n a l l experiments, while mass 2  flow ratio adjustments were made by modifying the wind tunnel air speed. D u r i n g the 30 second transients the images were captured at 5 fps and the jet temperature was recorded by two thermocouples at a sampling rate of 200 H z .  4.4  Solution Algorithms  E a c h experimental run generated a sequence of 150 frames contained i n an A u d i o Video Interleave ( A V I ) file. E a c h frame consisted of 32-bit R G B x data for 640x480 pixels. T h e total amount of R A M required to store each sequence file amounts to 640x480x4x150=184,320,000 byte=174 M B . T h e same region of interest ( R O I ) was selected from each frame and covered a streamwise distance of approximately 10 jet diameters and a spanwise distance of 3 jet diameters, or 416x124 pixels. T h e R G B to H S L colour space transformation was applied on the selected R O I and the resolved hue values were stored i n a newly created Truevision T A R G A  114  ( T G A ) [59] binary file. A script program was written to automate this procedure and store the sequence of 150 T G A files on the hard drive. A C routine containing the calibration function was implemented to reduce the computational domain using a 4x4 pixel grid, a n d to record the temperature at each cell based on the averaged hue value over the 16 pixel cell. E a c h cell covered a physical area on the surface of around 1.5 m m . A new file containing 104x31=3224 temperature data points was 2  stored on the hard drive for each of the 150 T G A files. T h e temperature history of the T L C instrumented surface (T (t)) was then known. w  Using the approach described i n section 4.1, the film cooling effectiveness 77 and the heat transfer coefficient hf can be simultaneously obtained from the solution of a system of two non-linear equations i n the form of E q n . 4.8 on page 107. W i t h the setup used i n this study, the mainstream temperature is equal to the initial temperature, so that Tj = Too i n both E q n . 4.8 and E q n . 4.9. For an arbitrary time event pair, t = t and t — ij,, 77 and hf can be a  found from the solution of the following system of equations:  T (t )-Too r?(T - Too) w  .  0  (h)at \  (hfy/at;)  a  1 — exp  2  ~ ^  77(T - Too)  ] erfc f ^  = 1 - exp f^ P  { ^  2  J  {  )  k  J  v  (4.11) ;  If a step change i n the temperature of the jet flow (T ) is achieved, E q n . 4.10 a n d E q n . 4.11 2  combine to give one equation w i t h one unknown, the heat transfer coefficient:  T (t )  - Too _  T (tb)  Tx)  w  w  a  1  6  X  P  ^)erfc[  k  (4.12)  h at  1  2  f  h  1 — exp and 77 can be then calculated by replacing hf i n either E q n . 4.10 or E q n . 4.11. B u t a true step change i n the temperature of the jet flow could not be achieved i n the experiments, so that the temperature of the jets was actually a function of time,  T = T (£). 2  2  This information was  incorporated i n the solution to the transient heat transfer problem, described b y E q n . 4.8,  115  by a method similar to that described i n [36] where the actual temperature rise was obtained from a superposed set of incremental steps. T h e fundamental solution becomes:  T.W - T . - | { l - ~P  ( ^ ) )  e r f c  ( * / y g * ) },(  A  T  l  )  j  (4  ,3)  T h e number of time steps i n E q n . 4.13 was taken to be equal to the number of video frames, so that N=150. A new algorithm was developed i n this study to calculate the values of hf and 77, and this procedure is further described. For any arbitrary time event pair, t — t  a  and t = tb, two equations i n the form of E q n . 4.13 combine to give one non-linear equation for the unknown variable hf i n the form:  E&i{l-erp(^=^)erfcf^^ T (t )  -  T (tb)  — Toe  w  w  a  Tec _  \  I  J  p c k  [  (AT,)*  yV*  £^l-exp(^)erfc(^) W  (4.14)  and 77 can be thus calculated from the wall temperature response at either time event, t = t  a  or t = tb, for example:  r? =  ,  E £ { l - exp  T w  ^  t a )  ~ °° T  ,  (4.15)  (S£=>] erfc  E q u a t i o n 4.14 and E q n . 4.15 are solved numerically for a series of time event pairs and the final values of hf and 77 are obtained by averaging. T h e grid cells on the surface where the temperature change is less than the estimated average error i n temperature of ± 0 . 1 5 5 ° C (see F i g . 3.17 on page 101) are excluded from the calculations. These are regions uncovered by the film cooling jets, so that the film cooling effectiveness is basically zero and the transient heat conduction approach does not apply. A n y attempt to solve E q n . 4.14 and E q n . 4.15 here would produce an error and the program will stop. T h e heat transfer coefficient is also assigned a value of zero i n these regions, so that hf = 0 and 77 = 0 here and these values are not produced by computations.  116  4.5  The Time Event Pairs  A n important issue is to be raised at this point, with respect to the choice of the time event pair, t = t and t = tb. In previous studies of impulsively started transient conduction it was a  assumed that, hf and 77 being theoretically independent of time, any time event pair could be used a n d the choice was one of convenience only. The investigation will proceed for the simplest case of a sudden change i n temperature for which the solution is i n the form of E q n . 4.12, but the argument also applies t o the present situation described by E q n . 4.14. T h e left hand side only of E q n . 4.12 represents data obtained by measurement (T (t) from the resolved local colour and Too measured by w  a thermocouple) and will be further referred to by the symbol  while the right hand side  accounts for the effects of time and physical properties of the material used.  ~~¥ ~  (4.16)  T {t ) - T  b  w  b  a  T i m e is measured indirectly by setting the duration of the transient for both video sequence capturing, at specified frame rate, and jet flow temperature recording, at specified sampling frequency.  T h e right hand side of E q n . 4.12 represents a non-dimensional wall  temperature profile and will be further referred to by the symbol 0:  0  1 — exp  0a  'h)t  a  pck 'h t  erfc  y/pck  2  1 — exp  f  h  pck  erfc  (4.17)  y/pck  Equation 4.17 is graphed i n F i g . 4.3 as a function of hf/y/pck, for different time event pairs relevant to the transients used i n this study; one of the time events was fixed at tb = 30 s and the product s/pck = 571 W s / / ( m K ) for the Plastic G used. 1  2  2  A s a general trend, the slope for a l l curves i n F i g . 4.3 decreases at higher values of hf, and the effect is more pronounced when the two time events are closer. This i n t u r n , has direct implications for the accuracy w i t h which the heat transfer coefficient can be determined. Small experimental errors i n the measurement of $ can produce large errors i n  117  Figure 4.3: Non-dimensional wall temperature profiles. the calculation of hf, especially for higher hf and for closer time events. For example, for the curve t}, = 30 s and t  a  = 20 s, a value of ^ = 0 = 0.96 ± 1% would produce a heat  transfer coefficient hf = 341 ± 93 W / ( m K ) . 2  This is due to the fact that the closer the  time events are, the smaller the change i n the temperature of the surface at any one point is. This effect is magnified by the behaviour of the low thermal diffusivity material used, i n particular the product y/pck acts as a heat transfer coefficient damping factor i n the solution of the transient problem. To further exemplify the errors i n the determination of n, the simpler case described by E q n . 4.10 on page 115 is considered.  For the same situation w i t h t  a  — 20 s and a non-  dimensional temperature ratio determined within ± 1 % that produced a heat transfer coefficient of hf = 3 4 1 ± 9 3 W / ( m K ) , the right hand side of E q n . 4.10 is determined w i t h an error 2  of around ± 5 % only. T h e error i n n can be thus calculated as e = y/(±l%)  2  v  + (±5%)  2  =  ± 5 . 1 % . In this example, the film cooling effectiveness was calculated more accurately than the heat transfer coefficient. In conclusion, the time events i n E q n . 4.14 have to be as far apart as possible to minimize  118  the error i n heat transfer coefficient.  T h e m a x i m u m value for the time t  a  requirement that there be a measurable temperature change at the farthest  is set by the downstream  points i n the region of interest, while the value for if, is set to ensure the semi-infinite wall assumption that the thermally affected region of the wall be less than the wall thickness. T h e m i n i m u m time required for a finite temperature change, at any distance into the wall, can be calculated by assuming that the surface temperature is suddenly increased from Ti = Too to the m a x i m u m measured jet temperature.  This is equivalent to modified  boundary conditions which include infinite heat transfer coefficient (hf = oo) and film cooling effectiveness value of one (n — 1). T h e solution to the transient one-dimensional conduction equation then becomes (adapted from [25]):  r  » ( ' ) -  T  -  erfc  max(T ) - T 2  a  2y/ai.  0  (4.18)  E q u a t i o n 4.18 is solved numerically for an imposed ST = T (t,z) — T^ to give the w  m i n i m u m time required for a ST temperature increase to appear at the specified z coordinate. T h e actual period of time is always longer due to the finite hf,n<l,  and the time required  for the jet temperature to reach its m a x i m u m value during any one experimental r u n . For example, the value tb = 30 s used i n the previous analysis corresponds to the m i n i m u m time it takes to increase the temperature i n the middle of the | inch thick wall (z = d/2 — | inch) by 0 . 2 ° C . T h e m i n i m u m time required for a 0.01°C temperature increase at the plenum side of the wall (or z  max  = d = | inch) is 65 seconds. Therefore, the value rj& = 30 s corresponds  to a thermally affected region of less than half the wall thickness during the actual runs.  4.6  Experimental Results  Three cases were investigated corresponding to the mass flow ratios M = 0 . 5 , 1.0, and 1.5, and the spanwise values of n were compared, at the same downstream locations, w i t h the values from [17].  T h e heat transfer coefficient is reported i n dimensionless form as the  Stanton number, St =  • T h e reported local values of TJ and hf actually represent  119  averaged values over the investigated two adjacent jets, jet number 3 and jet number 4 respectively, shown i n F i g . 4.2 on page 112. T h e reported results account i n this way for possible non-uniformities i n the T L C ' s layer thickness, and large scale unsteadiness of the velocity field [2], [21]. T h e averaging process was done on the recorded colour information of the region of interest. This differs slightly from the approach i n [17] where the jet number 3 only was used i n the investigation. A periodicity check of the velocity field for these two jets was done i n [17] and the results showed that the jets can be considered as part of an infinite row. T h e film cooling effectiveness results from [17] were obtained using a flame ionization detector technique, similar to that described i n section 2 of the present study. L o c a l values of n from [17] represent averages of 2000 data points sampled w i t h i n 15 s. In the present investigation, the time history of the jets was incorporated by using the multiple sampling technique with t covering the first 8 s of the transient i n 0.2 s increments, a  while tb = 30 s for a l l cases. T h e final results were obtained by averaging over a l l the results within the 8 s of computed time history i f a solution was available for at least 2 time event pairs. A m a x i m u m of 2 X 36 data points were thus included i n the averaging process.  4.6.1  General Trends  Figure 4.4 shows contours of rj and F i g . 4.5 shows contours of hf i n dimensionless form. G r i d cells where a solution was attempted but was not found for at least 2 time event pairs, were discarded and are shown as blank cells i n the contour plots. Interestingly enough, a solution was not possible for a l l the pixel cells at a l l time event pairs. A s F i g . 4.3 on page 118 shows, once the time event pair is fixed, a solution is only possible i f the measured value of  intersects the corresponding 0 curve, so that:  * - 0 = 0  (4.19)  T h e contour plots give a very good indication of the respective area coverage for each of the three mass flow ratios. It is clear that the best coverage is achieved i n the M = 0 . 5  120  to  t—I •  DO  n rt  <!  o  tt>  TO  o o o Ei  3>  o p  O  OP  to to  cr  rtO  0)  §  rt-  1  S  ft)  O  &  0  o  H  n  &>  O  0  o  o  Or  n  cw d  case, when the jet is strongly deflected towards the floor by the crossflow. T h e coverage progressively deteriorates for the M = 1 . 0 and M = 1 . 5 cases, w i t h the poorest coverage i n the M = 1 . 5 case. T h e jet seems to separate from the surface at the higher mass flow ratio, therefore providing extremely low protection of the surface. T h e region w i t h higher 77 is also significantly larger i n the M = 0 . 5 case than i n the other two cases. Based on the 77 data alone, the best performance is achieved at M = 0 . 5 for this compound angle configuration. For a l l cases, hf is higher around the sides of the jet, consistent w i t h regions of stronger interaction between the jet, crossflow, and the surface. These regions at the sides of the jet are also characterised by the lowest 77 values. W i t h i n the area covered by the jet there are also regions of higher hf; such a region, located immediately downstream of the duct exit i n the M = 0 . 5 case, shifts towards one side of the jet i n the other two cases w i t h the highest values of the St number i n the M = 1 . 0 case. For a l l cases, regions w i t h lowest hf are present within the area covered by the jet, rather than on its sides, and at some distance downstream of the injection site, as expected. This trend is more obvious i n the M = 0 . 5 case, where there is such a large central region w i t h lowest values of the St number, starting at about 4 diameters downstream of the injection site. T h e heat transfer coefficient is still at high levels as far as 11 diameters downstream around the sides of the jet for the M = 0 . 5 and M = 1 . 0 cases. Another important conclusion that can be drawn from F i g . 4.4 and F i g . 4.5, is that regions w i t h highest 77 do not correspond to regions of lowest hf, and regions w i t h highest hf do not necessarily correspond to regions of lowest 77. This clearly proves the need to determine both 77 and hf to completely specify the film cooling performance on a surface of interest.  4.6.2  Mass Flow Ratio M=0.5  Since the vertical thermal penetration into the wall is less than a half jet diameter everywhere, it was expected that the semi-infinite assumption holds up to locations a half jet diameter around the upstream duct exit corner, where multi-dimensional conduction effects could be  123  Figure 4.6: Comparison of 77 obtained by the T L C technique w i t h the F I D data of F i n d lay [17], M = 0 . 5 .  124  125  important. To further quantify these effects, and to verify the assumptions and the newly developed formulation based on liquid crystal thermography ( T L C ) , comparison is made w i t h film cooling effectiveness data obtained using a flame ionization detector technique [17] (FID). Figure 4.6 compares spanwise values of 77 at different streamwise locations on the surface. A g a i n , a l l values of 77 = 0 for the T L C technique are not computed but assigned, being indicative of regions where the change i n temperature is less than the measured experimental error. Overall, the agreement is very good. E v e n at the streamwise location x/d  = 1, i n  the immediate vicinity of the downstream duct exit, the trends are reproduced by the T L C technique but for a few points around the spanwise location y/d = 1.5, the location of the downstream duct exit corner, as expected. For the other downstream locations, the peak values of 77 and their location are well reproduced. F i l m cooling effectiveness data is not available beyond the streamwise location x/d = 8 from the F I D technique. T h e time history of both 77 and hf is considered i n F i g . 4.7, at a few arbitrarily chosen locations on the surface. It is evident now from these plots, that the average from a l l time events can best characterise the complex three-temperature  situation A single time event  pair sampling, as used i n the past, could not possibly capture all the details of the transient experiment. E x a m i n a t i o n of F i g . 4.7 shows that there is significant scatter i n both hf and 77 parameters at the chosen locations on the surface. In particular, hf appears to be much higher at the beginning of the transient (the time event pair t and tb further apart) and gradually settles a  around a more uniform value when the first event i n the time pair becomes greater than about 4 seconds (t  a  > 4 s). This trend is probably caused by the development of the jet  flow i n the early stages of the transient with the associated stronger perturbation of the crossflow boundary layer. T h e film cooling effectiveness is not affected by this behaviour of hf at small values of t , but appears very variable when corresponding values of hf are a  below around 50 [ W / ( m K ) ] . A definite conclusion cannot be inferred from these arbitrarily 2  sampled locations alone, since they represent only about 0.2% of the calculated pixel cells on  126  the surface. If more details are required at some particular location, the time history at that pixel cell can be recovered easily from the time history files, and the appropriate treatment can be applied based on the observed trends. These fluctuations of both parameters also indicate that lateral conduction effects are indeed negligible, since these would tend to even out the  fluctuations,  so that the local  response can be considered one-dimensional. T h e wide scatter of the higher values of hf is not entirely surprising. Accompanying flow visualisation studies i n [17] [2] clearly show that the generated flow field is turbulent, the jet flow developing into a strong vortical structure w i t h oscillations present i n particular at the interface between the jets and the crossflow. For the case of injection through a row of square jets perpendicular to the crossflow at a jet Reynolds number of Re  2  = 4700, a large scale unsteady flow structure was present and  was estimated to be 30 times larger than typical turbulence structures [2]. For the present situation, the jet penetration of the crossflow boundary layer is also m i n i m u m at this lower mass flow ratio, so that the floor interaction between the jet and the crossflow is the most significant. T h e heat transfer coefficient is strongly dependent on the generated flow field and higher values of hf are indicative of regions of vigorous interaction between the jets and the crossflow. It is mainly i n these regions that the observed flow oscillations would immediately translate into fluctuations of the heat transfer coefficient and probably higher average values of hf. T h e use of the first 8 seconds of the transient for the first event i n the time pair, appears now as a legitimate choice since no deterioration of the results is observed at the later time events. This allows for the use of as many as 2 x 36 time event pairs i n the calculation of both hf and n.  4.6.3  Mass Flow Ratio M=1.0  Figure 4.8 shows the comparison of n for the M = 1 . 0 case. A g a i n , the agreement between the results obtained w i t h the two methods, liquid crystal thermography ( T L C ) and flame ionization detector technique ( F I D ) respectively, is good. A t some of the positions shown  127  Figure 4 . 8 : Comparison of 77 obtained by the T L C technique w i t h the F I D data of F i n d lay [17], M = 1 . 0 .  128  Figure 4.9: T i m e history of hf and 77, M = 1 . 0 .  129  in F i g . 4.8, x/d = 2 and x/d = 3, a shift i n the spanwise location of the jet is apparent. W i t h i n the experimental accuracy, the trends are nevertheless preserved as well as the peak n values. T h e best agreement is actually achieved at the streamwise location closest to the jet exit, x/d = 1, so that the region around the downstream duct exit corner is well represented by the one-dimensional formulation. This can be explained by the findings i n [17], that the jet is mostly deflected towards the upstream side of the duct after negotiating the sharp corner at the duct entrance. This effect is more pronounced at higher mass flow ratios due to the increased vertical momentum of the jet, so that the region i n the duct around the downstream exit corner is not immediately exposed to the jet flow. For this case, the time history of n and hf is shown i n F i g . 4.9 at four arbitrary locations on the surface.  N o definite trend is present i n the time variation of the two parameters,  and the same conclusions presented for the M = 0 . 5 case can be inferred here too regarding the lateral conduction effects and the choice of the time event pairs.  A g a i n , the use of  the multiple sampling technique provides a better representation of the  three-temperature  situation. Overall, there is more scatter i n the values of hf than i n the values of rj except for the location x/d — Z and y/d = 0.5. F i l m cooling effectiveness values appear to  fluctuate  more at lower values of hf, as i n the M = 0 . 5 case. A t some locations, the values of rj appear extremely stable although the scatter i n the corresponding hf values is significant.  4.6.4  Mass Flow Ratio M=1.5  T h e last case used for comparison and validation is M = 1 . 5 , and spanwise distributions of n at different streamwise locations are shown i n F i g . 4.10. The F I D results from [17] are again used for comparison. The agreement is good for this case too. T h e peak n values and their location is also generally reproduced, and very good agreement is again achieved at the spanwise location immediately downstream of the duct exit as i n the M = 1 . 0 case. N o data is produced by the liquid crystal technique at the streamwise x/d = 8 location. N o t enough temperature change was observed at these points on the surface to attempt a solution, so a  130  1.0 0.9  TLCx/d=2.93 i FID x/d=3  0.8  TLC x/d=4.97 FID x/d=5  0.7 0.6 0.5 0.4 0.3 0.2 0.1  1™1V  0.9  Y/d  3.o °vr-v%  TLC x/d=7.9 FID x/d=8  0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 •  —  •  —  »  -  "Eft—ft—iV"W^"^!b Y/d  Figure 4.10: Comparison of 77 obtained by the T L C technique w i t h the F I D data of F i n d lay [17], M = 1 . 5 .  131  Figure 4.11: T i m e history of hf and rj, M = 1 . 5 .  132  value of zero was assigned for all pixel cells i n this situation (77 = 0). A higher jet temperature would have been required to obtain a measurable temperature change farther  downstream  of the injection site when higher mass flow ratios are used and the jet is not immediately deflected towards the floor by the crossflow. T i m e history data is shown i n F i g . 4.11 at four arbitrarily sampled locations on the surface. N o consistent trends are observed, so that the multiple sampling technique proves again to better represent the complicated three-dimensional flow field.  T h e higher values  of hf are also quite scattered while the corresponding 77 values are quite stable. T h e lower values of hf seem to be more stable but the corresponding 77 values then  fluctuate  more  significantly.  4.7  Uncertainty Analysis  T w o of the initial assumptions have yet to be verified, negligible effects due to buoyancy, and negligible radiation loss respectively. T h e buoyancy effects are considered negligible when Gr/Re  2  «  1 [46], where Gr = gL (T 3  - T ) / ( i / T ) is the Grashof number for a perfect 2  w  0 0  0 0  gas. T h e characteristic length L is taken here to be the streamwise length of the liquid crystal instrumented patch, so that L « lOd = 0.127 m . T h e temperature of the surface varies, so that a measure of the m a x i m u m possible loss is obtained from the m a x i m u m recorded wall temperature T  w  ~  30°C.  T h e temperature of the crossflow was around  Too ~  21.9°0.  Also,  the m i n i m u m crossflow velocity is i n the M = 1 . 5 case, [Too = 3.8 m / s . This gives the ratio:  S = ^ =?  1  0 0024< < 1  '  (4 20)  so that the buoyancy effects can be neglected at the lowest crossflow velocity, and they are even less important at the higher crossflow velocity, or lower mass flow ratio. T h e radiation exchange rate between the relatively small liquid crystal instrumented surface and its surroundings can be expressed per unit area as:  133  q  rad  = eo~(T* - T £ )  (4.21)  where e is the surface emissivity, and a = 5.67 x 1 0 W / ( m K ) is the Stefan-Boltzmann - 8  2  4  constant. It is difficult to appreciate the emissivity of the Plastic G / b l a c k p a i n t / l i q u i d crystal instrumented surface so that the m a x i m u m value e = 1 is used here (an estimated value of e = 0.85 was used i n [4]). A direct comparison can be made with the values of hf when the radiation heat transfer coefficient h d is employed: ra  hrad =  e<r(T + T ) ( T ^ + Tl) w  (4.22)  o c  T h e highest temperature of the surface, T « 30°C, was recorded for the M = 0 . 5 case at w  locations around x/d = 1.5 and y/d = 0.8. T h e values of the convective Stanton number at these locations are taken from F i g . 4.5 on page 122 to be around St = 0.036. A radiation Stanton number can be calculated based on h d defined by E q n . 4.22, so that: ra  St^d =  . ri kr  = 0.044 x 1 0 "  (4.23)  2  The radiation is therefore less than 1.2% of the forced convection heat transfer, and for most of the locations on the surface this value is actually much less since T < 30° C. w  This section is further addressing the sources of error i n the determination of hf and 77 by the liquid crystal thermography technique, i n the investigated three-temperature situation. The goal here is to devise a method to estimate the errors i n this complicated case, and to produce uncertainty maps for both hf and 77. T h e methodology outlined i n [1] will be followed whenever possible. Three-temperature situations always require a m i n i m u m of two time events, with the corresponding surface temperatures, to calculate the two unknowns (hf and 77) when the initial and gas temperatures are known. This produces two equations for two unknowns, different from a two-temperature situation when only one time event is required to calculate  134  the heat transfer coefficient i f the initial temperature and the gas temperatures are known. For the two-temperature situation, the heat transfer coefficient is then directly calculated from E q n . 4.6 by measuring the time it takes for the surface temperature (T(z = 0,£)) to reach a prescribed temperature that corresponds to the calibrated colour of a thermochromic liquid crystal. T h e two-temperature situation requires two time events i f the initial surface temperature is unknown. This approach was first introduced by Jones [27] and the appropriate error analysis was addressed by Ireland and Jones [26]. T h e authors found the heat transfer coefficient to be extremely sensitive to the measured time ratio, and to small errors in a non-dimensional temperature, similar to that referred to i n this study as  Ireland and  Jones estimated that a 3.3% uncertainty i n this non-dimensional temperature ratio could produce an uncertainty i n the calculated heat transfer coefficient of more than 22%. For a given test i n their two-temperature situation, i n which narrow band liquid crystals w i t h fixed temperature sensitivities were used, the non-dimensional temperature was fixed but the time ratio varied w i t h the location on the surface. Small errors i n the measurement of this time ratio are also reported to produce large errors i n the heat transfer coefficient, such that a solution was not actually possible everywhere on the surface. Ireland and Jones conclude their analysis of the so called double crystal method indicating that errors i n the heat transfer coefficient could be substantially reduced by accurate measurement of the temperatures involved, and by appropriate choice of crystal temperatures. In the present study, the time is known and the temperature information is correspondingly retrieved based on the recorded colour information, as needed. A n initial time delay of no more than 0.033 s could be included to account for the scanning frequency of the N T S C video camera, and is constant for each test r u n . Such a time delay would only produce an absolute error i n the time event ratio t /tb of less than 0.1% and no correction has been made a  for i t . T h e use of a wide band liquid crystal for thermal mapping of the surface gives the flexibility of choosing the most suitable time event pair combination, and the corresponding non-dimensional temperature ^ , to minimize the errors i n hf and n. T h e experimental uncertainty i n the determination of the surface temperature was addressed i n section 3.4 and  135  quantitative values are shown i n F i g . 3.17 on page 101. T h e uncertainty is a function of the temperature of the surface and the average value of ± 0 . 1 5 5 ° C is further used i n the analysis. The crossflow temperature is measured by a type K thermocouple connected to an O M E G A O M N I - C A L calibration unit so that its absolute value is known within ± 0 . 0 5 ° C . The uncertainty i n hf is related to the uncertainty i n the measured quantity $ and the choice of the time event pair, as shown i n F i g . 4.3 on page 118. T h e closer the time event pair, the greater the uncertainty i n hf especially at higher values of hf.  In fact, values  of hf/y/pck > 1 cannot be calculated with reasonable accuracy when the time events are close. To avoid overflow errors during the actual computations, and to implicitly reduce the error i n the final results, an appropriate upper bound has to be set on hf. This value could come from an error minimization procedure on the final results, and it was set here at 600W/(m K). 2  The highest computed value of hf still carries the largest uncertainty, while the lowest values of hf appear to be computed with the smallest uncertainty. A n estimation of these uncertainty levels will be made for both hf and the associated rj.  The uncertainty i n hf due to the uncertainty i n the measured value of  for each time  event pair, is calculated numerically from the corresponding variation on the 0 curve at each time event pair. T h e slope of the 0 curve, or t a n a , determines the amplification factor on the uncertainty i n hf:  A0  A *  tan a  tan a  = A(h /Jpck) f  (4.24)  at any location where a solution is possible, according to E q n . 4.19 on page 120. The uncertainty associated with the physical properties of the material used is estimated at  'pck/V'pck = ± 2 % . T h e bias errors i n the temperature measurements are assumed to  be totally correlated for each test r u n . T h e uncertainty i n the non-dimensional temperature is calculated from the uncertainty i n the measured temperatures, and propagates [1] through their functional relationship defined by E q n . 4.16, so that: 136  i AY 5  =  f  f  ]  ( -(^)~T~(ta))8T„  2  T  ST (Q W  r *r,(M l  }  2  2  Equation 4.25 can be readily computed for each time event pair and for each pixel cell on the surface, to yield a new value of the heat transfer coefficient, hf/^/pck-\-S(hf/^/pck),  and a  new value of film cooling effectiveness, 77+ £77. This equation also shows that the uncertainty i n the non-dimensional temperature is reduced i f T [tb) » w  Too, T (t ) w  a  >> Too, and i f the  two wall temperatures are close. This latter condition is contrary to the requirement that the two wall temperatures be well separated, as inferred from F i g . 4.3, a conclusion also reached i n [26]. M i n i m u m uncertainty i n the results is then achieved when the three temperatures are i n the relationship T (tb) > T (t ) w  w  a  »  Too, and they are measured as accurately as  possible. The uncertainty i n the physical properties of the material is accounted for, so that the uncertainty i n hf is finally computed as:  Shi  .\ I[ y/pck ^ f J+ I{*Atan]a 'J  (4.26)  v  '  and the uncertainty i n 77 is simply computed and reported as £77/77. This procedure is repeated for each time event pair where a solution was initially obtained. Since the final results are computed as averages over the set of time event pairs, the associated uncertainty is calculated [1] and reported as:  Sn  where TV is the number of time event pairs for which a solution was obtained. Detailed maps of the uncertainty associated with 77 and hf are reproduced i n F i g . 4.12, and F i g . 4.13 respectively. For the M = 0 . 5 case, the uncertainty i n 77 is less than 9.55% everywhere and is as good  137  00  CO  cr £»  on  < n P  o  «  o o o Bt  cr  E.  cl-  o n i-)  P  (3  «•  i-h  o  to  R  6  I—> •  Figure 4.13: Map of the uncertainty in heat transfer coefficient data.  139  as 0.1%, and the uncertainty i n hf is as high as 20.5% and as low as 1%, w i t h typical values between 2% and 10%. For the M = 1 . 0 case, the uncertainty i n n is as high as 15.2% towards one side of the jet, and is as low as 0.23%, with typical values between 1% and 7%. T h e uncertainty i n hf is comparable w i t h that i n the M = 0 . 5 case.  For the M = 1 . 5 case, the  uncertainty i n rj is less than 7% and that i n hf is less than 10% everywhere but for a couple of cells around the downstream duct exit corner where the uncertainty runs as high as 21% for n and 27% for hf. Overall, the uncertainty i n n is less than 7% and the uncertainty i n hf is less than 10% everywhere except for small regions on the surface. T h e uncertainty i n rj is also less than the corresponding uncertainty i n hf. T h e uncertainty i n the results could be further reduced i f a solution is kept only i f its computed uncertainty is below some imposed levels, at selected time event pairs. T h e choice of the time event pairs could be also optimized according to the calculated uncertainty levels: even more time pairs could be considered when the uncertainty of the results is low, and only the time pairs that produce lower uncertainty values could be kept at higher hf.  T h e uncertainty i n the time event averaged results could be then  significantly improved, i n particular for the regions w i t h higher uncertainty levels identified i n figures 4.12 and 4.13.  4.8  Summary  A complex three-temperature situation is present i n compound angle square jets i n crossflow. This particular geometry was used to validate a newly developed transient heat transfer technique based on wide-band liquid crystal thermography.  T h e method allows for the  simultaneous determination of the film heat transfer coefficient (defined by E q n . 1.3 on page 10) and the film cooling effectiveness (defined by E q n . 1.6).  Available film cooling  effectiveness data from [17] is used for comparison. T h e time scale and the sampling rates for the present investigation were different from the study i n [17], and this might account for some of the differences i n the reported results. Nevertheless, the film cooling effectiveness  140  data from the two methods is i n very good agreement, and this provides confidence i n the computed values of the film heat transfer coefficient obtained w i t h the transient liquid crystal technique. T h e heat transfer coefficient is much more difficult to measure, especially i n a complicated three-dimensional turbulent flow field, since it is inherently coupled w i t h the flow and w i t h the conditions i n the boundary layer. T h e solution algorithm for n and hf was developed around the solution to the i m p u l sive transient one-dimensional heat conduction response of a semi-infinite wall, subject to convective boundary conditions expressed by E q n . 4.8 on page 107. T h e required change i n the temperature of the surface is attempted by the use of heated jets. Since the change i n the jet temperature is gradual, the total temperature change is obtained by superposition of incremental temperature steps. This approach has been also used i n the past i n single event sampling techniques. T h e assumption that the surface is suddenly exposed to a uniform temperature field is the most problematic, since the temperature profile is not constant across the jets and the locations further downstream of the injection sites are not immediately exposed to this temperature  field.  In this situation, it appears to be more beneficial to suddenly change the  temperature of the crossflow, such that the new thermal condition would be quickly convected downstream and i n a more uniform fashion. T h e available hardware at the University of B r i t i s h Columbia does not allow for such an approach. Nevertheless, the use of a large scale model to ensure that the temperature pulse is contained well within the wall during the transients employed proved quite successful. W i t h moderately heated jets, to justify the assumption of constant property flow and negligible buoyancy, the entire method provided accurate values of rj (and hence, by inference, oi hf) over almost a l l of the flow region examined (x/d < 10).  T h e novel approach i n this study combines a real-time true colour imaging system for accurate colour rendition, w i t h the use of a custom manufactured wide-band thermochromic liquid crystal and multiple event sampling for more accurate representation of rj and hf. T h e  141  use of a wide-band thermochromic liquid crystal permits monitoring the temperature of the surface over the entire duration of the transients, and the multiple event sampling technique can be further optimized to provide improved accuracy of the results. This technique represents a significant improvement of the single event sampling procedure used i n the past. Not only that the wide-band liquid crystal allows for a l l the information to be recorded during a single transient test, but the time history of n and hf is available to provide a detailed representation of the heat transfer characteristics i n complicated three-dimensional oscillating jets i n crossflow. T h e temperature of the surface is obtained from the high resolution, accurate true colour information, recorded i n real-time with the use of a dedicated imaging system. This development was undertaken following the assessment of different other imaging systems, some of which (video tape recording for example) are often used i n liquid crystal thermography. T h e present implementation uses the hue component from the R G B to H S L colour space transformation to directly quantify the corresponding temperature. Calibration of the wide-band liquid crystal over an even higher temperature range seems possible, and the use of different other colour transformations should be investigated i n the future. W i t h a l l errors now minimized along the imaging system chain, the limiting factor i n the accuracy of the surface temperature measurement becomes the uniformity and consistency of the liquid crystal layer. More advanced techniques and a specialized environment could be used to apply the liquid crystal i n a thin layer on a variety of surfaces. A small improvement i n the uncertainty of the surface temperature measurement proves extremely beneficial to the accuracy of the final results.  142  Chapter 5 Conclusions and Recommendations 5.1  Overall Summary  A n investigation into the film cooling effectiveness of cylindrical leading edge geometries has been conducted. T w o different existing cooling hole geometries were studied and they are referred to as the baseline geometry and the C A geometry respectively. A t h i r d leading edge geometry tested is referred to as the P V C geometry and represents an original design of an improved film cooling layout. In a l l cases the film cooling holes are disposed i n four rows w i t h two rows on each side of the geometrical leading edge. T h e holes are spaced at a certain spanwise distance apart i n each row, and the holes i n adjacent rows on either side of the geometrical leading edge are staggered with respect to each other, as well as about the geometrical leading edge.  F i l m cooling effectiveness measurements were made using a flame ionization detector technique and the mass/heat transfer analogy for fluids w i t h constant properties. T h e effect of density differences between the cooling jets and the mainstream i n an adiabatic heat transfer process was modelled by using either air or C0  2  as coolant gas i n the analogous  isothermal mass transfer process. T h e effect of mass flow ratio was examined for double and single row injection at a constant jet Reynolds number of approximately 4200 for a l l  143  cases. T h e effect of variable jet Reynolds number was studied for single row injection of the baseline geometry. These measurements give insight into the film cooling characteristics at the leading edge of turbine blades and provide new data for design purposes.  A film cooling effectiveness correlation was developed for the baseline geometry and double row injection w i t h air and C0  2  as coolant gases based on an area-averaged film cooling  effectiveness and on a newly defined blowing parameter. This correlation accounts implicitly for the specific geometrical layout of the baseline geometry and can be used i n existing design procedures to estimate the m i n i m u m film cooling performance of this geometry under various operating conditions.  The development, implementation and validation of a transient heat transfer technique using wide-band T L C for thermal mapping were presented. Based on a comparison of different colour image capture techniques made i n this study, a high performance imaging system was specified and implemented. Computer codes were also developed for processing and analysis of real-time true colour information recorded on a host P C ' s R A M . F i l m cooling effectiveness and heat transfer coefficient were simultaneously calculated from one transient test using a multiple sampling technique to account i n more detail for the surface time history. This approach has not been previously reported and is critical for the accurate calculation of the heat transfer characteristics i n a typical film cooling application described by a complex turbulent three-dimensional flow field w i t h large scale unsteadiness.  A flat plate film cooling geometry w i t h compound angle jets was used for validation of the liquid crystal method by comparison of film cooling effectiveness data w i t h the flame ionization data from F i n d l a y [17]. T h e study i n [17] uses a completely different technique based on flame ionization detector sampling i n an isothermal mass transfer experiment. Such a direct comparison of fundamentally different experimental methods is difficult to find i n the published literature. Very good agreement was achieved between the two sets of data  144  and represents a significant contribution of this thesis.  The new liquid crystal technique allows for accurate and cost effective assessment of both the heat transfer coefficient and the film cooling effectiveness i n film cooling applications. T h e experimental costs are greatly reduced over more traditional heat transfer data acquisition methods, and the initial hardware and development costs can be quickly recovered. Postprocessing of the recorded information provides a fast output of results, which allows for great flexibility i n the research program. Extensive experimental investigation of the effect of various design parameters on the heat transfer characteristics of leading edge geometries is now within reach.  T h e method can be applied i n a variety of situations, such as the  investigation of film cooling performance of different geometries, or for other industrial or more fundamental heat transfer research.  5.2 5.2.1  Leading Edge FID Measurements Conclusions  The spanwise-averaged film cooling effectiveness (77) was used to quantify the relative performance of the baseline geometry, the C A geometry, and the P V C geometry.  For the case of double row air injection and constant jet Reynolds number, 77 downstream of row 62 of cooling holes typically decreased with increasing mass flow ratio. A n exception was for the P V C geometry where 77 for M = 0.8 was slightly greater than 77 for M = 0.6 for the streamwise region from X/(TTD/4:)  = 0.58 to X/(ITD/A)  — 0.96. Overall, 77 was signif-  icantly higher for the C A geometry than for the baseline geometry for a l l mass flow ratios investigated from M = 0.5 to M — 1.2. For mass flow ratios below M = 0.8, the C A geometry produced higher fj than the P V C geometry, but for mass flow ratios above M = 0.8 the P V C geometry is superior. A t the higher mass flow ratio M — 1.4 the P V C geometry still produces higher 77 than the baseline geometry at M = 1.2. These results prove that the  145  design goals set i n the development of the P V C geometry have been achieved; a safe backflow margin can be kept by using higher coolant discharge while maintaining a relatively good film cooling effectiveness.  For the case of single row air injection at constant jet Reynolds number, measurements of fj were made for the baseline and P V C geometries. The local mass flow ratio correlates the spanwise-averaged film cooling effectiveness reasonably well immediately downstream of rows placed at different angles from the stagnation line around the cylindrical leading edge. Nevertheless, for the same local mass flow ratios fj was slightly higher immediately downstream of row 62 of cooling holes (farthest from the stagnation line) than immediately downstream of row 0i (closest to the stagnation line). T h e trends i n the variation of fj were similar for the two geometries where a peak fj occured at around Me = 0.3 and fj further decreased w i t h increasing local mass flow ratio. Farther from the cooling holes, for the baseline geometry, fj was higher downstream of row 8 than downstream of row 61 of cooling holes. 2  For the case of CO2 injection and constant jet Reynolds number, F I D measurements of film cooling effectiveness were made for the baseline geometry. For the case of double row injection, fj downstream of row 82 of cooling holes decreased w i t h increasing mass flow ratio from M = 0.5 to M = 1.2. T h e use of CO2 coolant, whose density is about 1.5 times that of air, produced higher fj than injection of air coolant for the same momentum flux ratios.  For the case of single row injection, fj downstream of row 61 of cooling holes was higher for CO2 injection than for air injection for a l l local mass flux ratios used.  Immediately  downstream of row 62 of cooling holes a peak fj occured at a slightly lower local mass flux ratio for air injection than for CO2 injection, but injection of CO2 produced higher fj for higher local mass flux ratios. Farther downstream of row 62 of cooling holes, fj was the same for air and C0  2  injection at lower local mass flux ratios, but CO2 injection again produced  higher fj for higher local mass flux ratios.  146  For the baseline geometry w i t h double row air and CO2 injection at constant jet Reynolds number, the film cooling effectiveness data were correlated based on a newly defined blowing parameter (£) and an area-averaged film cooling effectiveness (fj) which represents the m i n i m u m expected film cooling effectiveness for some region downstream of the cooling holes.  The effect of the jet Reynolds number on fj was investigated for single row air injection at different mass flow ratios using the baseline geometry. Downstream of row 61 of cooling holes, fj decreased with increasing jet Reynolds number for the lowest mass flow ratio used, but no significant effect of the jet Reynolds number was further observed w i t h increasing mass flow ratios. T h e effect of the jet Reynolds number was more significant immediately downstream of row 62 of cooling holes where the peak fj clearly decreased w i t h increasing jet Reynolds number. T h e peak fj occured at a lower mass flow ratio for the lowest jet Reynolds number used and at a higher mass flow ratio for the highest jet Reynolds number used. Overall, no significant effect of the jet Reynolds number was observed at higher mass flow ratios.  5.2.2  Recommendations  Recommendations for future work refer mainly to the P V C geometry. In the present layout, the P V C geometry produced higher spanwise-averaged film cooling effectiveness and better coverage than the baseline and C A geometries for higher mass flow ratios typical of those used i n gas turbines.  1. T h e set of film cooling effectiveness data from air injection should be complemented by CO2 injection data to simulate the use of jets cooler than the mainstream, modelling the real application more closely. 2. T h e new heat transfer technique based on liquid crystal thermography developed i n this study should be applied to the P V C geometry to obtain the complete heat transfer  147  characteristics of this geometry. T h e measured film cooling effectiveness results from the heat transfer technique could be then compared directly w i t h the measured results from the F I D technique. 3. Further development of the P V C geometry should also be considered.  T h e use of  flared holes or other more complicated hole exit geometries should be investigated. The transient heat transfer technique developed i n this study can be readily used on these geometries for detailed assessment of their heat transfer characteristics, namely the film cooling effectiveness and the heat transfer coefficient. 4. Correlations of film cooling effectiveness and heat transfer coefficient must be constructed for a l l cases so that the results could be used directly i n existing design procedures.  5.3  5.3.1  Liquid  Crystal Thermography  Conclusions  The transient heat transfer technique was tested i n a three-temperature application using a flat plate w i t h a row of square compound angle jets. This geometry was previously used i n the F I D study by Findlay [17]. Three different mass flow ratios of M = 0.5, 1.0, and 1.5 were investigated w i t h a constant jet Reynolds number of around 5000. Detailed quantitative comparisons of spanwise film cooling effectiveness profiles were made for a l l mass flow ratios examined. Overall, the film cooling effectiveness calculated from the heat transfer experiments were i n very good agreement with the film cooling effectiveness from the mass transfer experiments used i n [17].  The best coverage was obtained for the lowest mass flow ratio investigated of M = 0.5 when the jet was deflected towards the floor by the crossflow, and the coverage deteriorated progressively w i t h increasing mass flow ratio. A larger region of higher film cooling effective-  148  ness was present within the area covered by the jet for the lower mass flow ratio M = 0.5 than for the higher mass flow ratio M = 1.5. T h e film cooling effectiveness was higher for regions closer to the injection sites and decreased farther downstream of the cooling holes for a l l mass flow ratios used.  T h e Stanton number (St) based on the mainstream conditions was used to represent the variation of the heat transfer coefficient on the film cooled surface. Higher values of the Stanton number were present typically around the sides of the jet for a l l mass flow ratios tested.  T h e highest Stanton number values occured for the mass flow ratio M = 1.0 on  one side of the jet relatively close to the cooling holes and for the mass flow ratio M = 1.5 around the edges of the jet away from the injection sites. A relatively large region of lower Stanton number was present farther downstream of the cooling holes for the mass flow ratio  M  = 0.5. T h e region of lower Stanton number was significantly reduced for the mass flow  ratio M = 1.0 and occured much closer to the injection sites.  For a l l mass flow ratios examined, the regions w i t h highest film cooling effectiveness did not correspond to regions of lowest heat transfer coefficient, but the regions w i t h lowest film cooling effectiveness (which occured on the sides of the jets) did correspond to regions of high heat transfer coefficient. T h e uncertainty associated w i t h both film cooling effectiveness and heat transfer coefficient was typically higher at the sides of the jets farther away from the cooling holes.  5.3.2  Recommendations  A few recommendations for future work refer to further improvement of the transient heat transfer measuring technique and liquid crystal thermography. 1. A comparative assessment of different colour space transformations should be pursued to determine the optimum colour/temperature transformation to extend the useful calibration range of the liquid crystal. This analysis can be done on the present set  149  of calibration data by implementing various colour space transformations. T h e uncertainty levels and the useful calibration range can be determined based on the standard deviation values associated w i t h a specific colour component as described i n this study. This will allow for the efficient use of available very wide-band liquid crystal [60] i n film cooling applications. 2. T h e use of higher image transfer rates should be considered for future experiments to capture the large scale unsteadiness present i n typical applications of jets i n crossflow i n more detail. T h e implementation of higher image transfer rates is expected to be less costly i n the near future considering the present fast improvement i n computer imaging hardware. 3. T h e multiple event sampling technique developed i n this study should be optimized to further reduce the uncertainty of the output heat transfer data. For example, computed results w i t h associated high uncertainty levels and results that are not w i t h i n a few standard deviations from the mean value could be discarded i n the analysis based on selected confidence intervals. 4. T h e new measuring technique can be used for quick evaluation of different leading edge design configurations without the need for intermediate, and very expensive, rig testing.  Modifications and subsequent developments of different geometries can be  directly and immediately assessed.  150  Appendix A Colour Space Conversion R G B to H S L (Foley and V a n D a m ) max = m a x i m u m of R G B m i n = m i n i m u m of R G B 1 = (max + min) / 2 if m a x = m i n , S = 0, H = undefined if 1 < 0.5, S = (max - m i n ) / ( m a x + min) else, S = (max - m i n ) / ( 2 - max - min) delta = m a x - m i n if R = max, H = ( G - B ) / d e l t a if G = max, H = 2 + (B - R ) / d e l t a if B = m a x , H = 4 + ( R - G ) / d e l t a H = H * 60 if H < 0, H = H + 360  H S L to R G B (Foley and V a n D a m ) if L < 0.5, m2 = L * (1 + S) else m2 = L + S - L * S m2 = 2 * L - m2  151  if S = 0 and H = undefined, R = G = B = L R = value(ml,m2,h + 120) G = value ( m l , m 2 , H ) B = value(ml,m2,h - 120) where value(a,b,c) is the following function: value(nl,n2,hue) if hue >360, hue = hue - 360 else i f hue <0, hue = hue + 360 if hue <60, value = n l + (n2 - n l ) * hue / 60 else i f hue <180, value = n2 else i f hue <240, value = n l + (n2 - n l ) * (240 - hue) / 60 else value = n l return value  152  Appendix B Wind Tunnel Setup White light / Q \  3  X  C  C D  /Q\  W  H  I  T  E  U  S  H T  Test section  Honeycomb Plenum Imaging system dedicated PC  >•  Heater • 2-way . solenoid valve  Flowmeter  Jet temperature data acquisition  Pressure regulator  *—®~Lj  k  Air supply  Oil trap  153  Bibliography [1] Abernethy R . B . , Benedict R . P . , Dowdell R . B . , " A S M E Measurement  Uncertainty",  June 1985, J . of Fluids Engineering, Vol.107, pp.161-163. [2] Ajersch P . , "Detailed Measurements on a R o w of Jets i n a Crossflow - W i t h A p p l i c a tions", 1995, M . A . S c . Thesis, Department of Mechanical Engineering, T h e University of B r i t i s h Columbia, Vancouver, B C , Canada. [3] A v i a t i o n Week & Space Technology, December 1996. [4] B u t l e r R . J . , B a u g h n J . W . , "The Effect of the T h e r m a l Boundary C o n d i t i o n on Transient M e t h o d Heat Transfer Measurements on a Flat Plate W i t h a L a m i n a r B o u n d a r y Layer", November 1996, J . of Heat Transfer, Vol.118, pp.831-837. [5] C a m c i C , K i m K . , Hippensteele S . A . , " A New Hue Capturing Technique for the Quantitative Interpretation of L i q u i d C r y s t a l Images Used i n Convective Heat Transfer Studies", October 1992, J . of Turbomachinery, Vol.114, pp.765-775. [6] Carslaw H . S . , Jaeger J . C , "Conduction of Heat i n Solids", 1959, Oxford University Press, 2nd edition. [7] C h y u M . K . , Y u Y . , D i n g H . , Downs J . P . , Soechting F . O . , "Concavity Enhanced Heat Transfer In A n Internal Cooling Passage", A S M E Paper N o . 97-GT-437. [8] Clifford R . J . , Jones T . V . , Dunne S.T., "Techniques For Obtaining Detailed Heat Transfer Coefficient Measurements W i t h i n Gas Turbine Blade A n d Vane Cooling Passages", A S M E Paper N o . 83-GT-58. [9] Collings P . J . , " L i q u i d Crystals: Nature's Delicate Phase of M a t t e r " , 1990, Princeton University Press. [10] C u m p s t y N . A . , "Jet Propulsion", 1997, Cambridge University Press. [11] Eckert E . R . G . , "Analysis of F i l m Cooling and Full-Coverage F i l m Cooling of Gas Turbine Blades", January 1984, J . of Engineering for Gas Turbines and Power, Vol.106, pp.206-213.  154  [12] E k k a d S . V . , Zapata D . , H a n J . C , "Heat Transfer Coefficients Over A F l a t Surface W i t h A i r and C0  2  Injection Through C o m p o u n d Angle Holes Using A Transient L i q u i d  C r y s t a l Image M e t h o d " , 1997, J . of Turbomachinery, Vol.119, pp.580-586. [13] E k k a d S . V . , Zapata D . , H a n J . C , " F i l m Effectiveness Over A F l a t Surface W i t h A i r and CO2 Injection Through C o m p o u n d Angle Holes Using A Transient L i q u i d C r y s t a l Image M e t h o d " , 1997, J . of Turbomachinery, Vol.119, pp.587-593. [14] E l - M a s r i M . A . , " O n Thermodynamics of Gas Turbine Cycles: Part 2 - A M o d e l for Expansion i n Cooled Turbines", January 1986, J . of Engineering for Gas Turbines and Power, Vol.108, pp.151-159. [15] Fackrell J . E . , " A system for turbulent concentration measurements", 1978, J . P h y s . E : Sci. Instrum., Vol.11, pp.1015-1022. [16] Fackrell J . E . , " A flame ionisation detector for measuring fluctuating concentration", 1980, J . P h y s . E : Sci. Intrum., Vol.13, pp.888-893. [17] F i n d l a y M . J . , "Experimental and Computational Investigation of Inclined Jets i n a Crossflow", 1998, P h . D . thesis, Department of Mechanical Engineering, T h e University of B r i t i s h C o l u m b i a , Vancouver, B C , Canada. [18] Funazaki K . , K o y a b u E . , Yamawaki S., "Effect of Periodic Wake Passing on F i l m Effectiveness of Inclined Discrete Cooling Holes A r o u n d the Leading Edge of a B l u n t B o d y " , January 1998, J . of Turbomachinery, Vol.120, pp.70-78. [19] Goldstein R . J . , " F i l m Cooling", 1971, Advances i n Heat Transfer, Vol.7, pp.321-379. [20] Goldstein S. editor, " M o d e r n Developments i n F l u i d Mechanics", 1957, Vol.11, Oxford University Press, A m e n House, London. [21] Haven B . A . , Kurosaka M . , " K i d n e y and anti-kidney vortices i n crossflow jets", 1997, J . of F l u i d Mechanics, Vol.352, pp.27-64. [22] Heiser B . , M a y B . , " U . S . A i r Force Engine Management; a Cradle-to-Grave Process", 1996, Global Gas Turbine News, Vol.36, No.2. [23] H i l l P . G . , Peterson C , "Mechanics and Thermodynamics of Propulsion", 1992, AddisonWesley, 2nd edition. [24] Hippensteele S . A . , Russell L . M . , Torres F . J . , " L o c a l Heat-Transfer Measurements on a Large Scale-Model Turbine Blade Airfoil Using a Composite of a Heater Element and L i q u i d Crystals", October 1985, J . of Engineering for Gas Turbines and Power, Vol.107, pp.953-960.  155  [25] Incropera F . P . , D e W i t t D . P . , "Fundamentals of Heat and Mass Transfer", 1990, J o h n W i l e y & Sons, 3rd edition. [26] Ireland P . T . , Jones T . V . , "Note O n T h e Double C r y s t a l M e t h o d O f Measuring Heat Transfer Coefficient", 1987, University of Oxford, Department of Engineering Science Report No.1710/87, Oxford, U K . [27] Jones T . V . , Hippensteele S . A . , " H i g h resolution heat transfer coefficient maps applicable to compound-curve surfaces using liquid crystal i n a transient wind tunnel", 1987, Developments i n Experimental Techniques i n Heat Transfer and Combustion, H T D - V o l . 7 1 , A S M E publication, N e w Y o r k . [28] Lander R . D . , F i s h R . W . , Suo M . , " E x t e r n a l Heat-Transfer Distribution on F i l m Cooled Turbine Vanes", October 1972, J . Aircraft, Vol.9, No.10, pp.707-714. [29] Langston Lee, "Market Drivers for Electric Power Gas Turbines: Reasons for the R e v olution", 1996, Global Gas Turbine News, Vol.36, N o . 3 . [30] Lieu D . N . , " P V C Geometry Specifications", 1997, Technical Communications to P r a t t & W h i t n e y Canada. [31] Louis J . F . , Hiraoka K . , E l - M a s r i M . A . , " A Comparative Study O f T h e Influence O f Different Turbine Cooling O n Gas Turbine Performance", A S M E Paper N o . 8 3 - G T 180. [32] Mehendale A . B . , H a n J . C , "Influence of H i g h Mainstream Turbulence on Leading Edge F i l m Cooling Heat Transfer", October 1992, J . of Turbomachinery, Vol.114, pp.707-714. [33] M e h e r - H o m j i C B . , "The Development of the Junkers J u m o 004B-The World's First Production Turbojet", October 1997, J . of Engineering for Gas Turbines and Power, Vol.119, pp.783-789. [34] Meinders E . R . , van der Meer T h . H . , Hanjalic K . , Lasance C . J . M . , " A p p l i c a t i o n of Infrared Thermography to the Evaluation of Local Convective Heat Transfer on A r r a y s of C u b i c a l Protrusions", February 1997, Int. J . of Heat and F l u i d F l o w , Vol.18, N o . l , pp.152-159. [35] Metzger D . E . , Carper H . J . , Swank L . R . , "Heat Transfer W i t h F i l m Cooling Near N o n tangential Injection Slots", A p r i l 1968, J . of Engineering for Power, pp.157-163. [36] Metzger D . E . , Larson D . E . , "Use of M e l t i n g Point Surface Coatings for L o c a l Convection Heat Transfer Measurements i n Rectangular Channel Flows W i t h 90-deg Turns", February 1986, J . of Heat Transfer, Vol.108, pp.48-54.  156  M i c k W . J . , M a y l e R . E . , "Stagnation F i l m Cooling and Heat Transfer, Including Its Effect W i t h i n the Hole Pattern", January 1988, J . of Turbomachinery, Vol.110, pp.6672. O u S., H a n J . C , "Influence of Mainstream Turbulence on Leading Edge F i l m Cooling Heat Transfer Through T w o Rows of Inclined F i l m Slots", October 1992, J . of Turbomachinery, Vol.114, pp.724-733. O u S., Mehendale A . B . , H a n J . C , "Influence of H i g h M a i n s t r e a m Turbulence on Leading Edge F i l m Cooling Heat Transfer: Effect of F i l m Hole R o w L o c a t i o n " , October 1992, J . of Turbomachinery, Vol.114, pp.716-723. Pimsner V . , "Motoare Aeroreactoare", 1983, E d i t u r a D i d a c t i c a si Pedagogica, V o l . 1 , romanian edition. Pitas I., " D i g i t a l Image Processing A l g o r i t h m s " , 1993, Prentice H a l l . P r a t t & W h i t n e y Canada, "Private Communications", 1997. Russo C . J . , "Changing The Landscape O f C i v i l A v i a t i o n " , 1997, G l o b a l Gas Turbine News, Vol.37, N o . 3 . Salcudean M . , Gartshore I.S., Zhang K . , M c L e a n I., " A n E x p e r i m e n t a l Study of F i l m Cooling Effectiveness Near the Leading Edge of a Turbine B l a d e " , January 1994, J . of Turbomachinery, Vol.116, pp.71-79. Salcudean M . , Gartshore I.S., Zhang K . , Barnea Y . , "Leading Edge F i l m Cooling O f A Turbine Blade M o d e l Through Single A n d Double R o w Injection: Effects of Coolant Density", A S M E Paper N o . 94-GT-2. Schlichting H . , "Boundary-Layer Theory", 1979, M c G r a w - H i l l Inc., 7th edition. Shadid J . N . , Eckert E . R . G . , "The Mass Transfer Analogy to Heat Transfer i n F l u ids W i t h Temperature-Dependent  Properties", January 1991, J . of Turbomachinery,  Vol.113, pp.27-33. Taylor R . P . , Coleman H . W . , Hosni M . H . , Love P . H . , " T h e r m a l boundary condition effects on heat transfer i n the turbulent incompressible flat plate boundary layer", 1989, Int. J . of Heat Mass Transfer, Vol.32, No.6, pp.1165-1174. U t z P . , "Today's video: equipment, setup, and production", 1992, Englewood Cliffs, Prentice H a l l , N . J . Valenti M . , "Upgrading Jet Turbine Technology", December 1995, Mechanical E n g i neering, pp.56-60.  157  [51] Vedula R . J . , Metzger D . E . , " A M e t h o d for the Simultaneous Determination of L o c a l Effectiveness and Heat Transfer Distribution i n Three-Temperature Convection Situations", 1991, A S M E Paper N o . 91-GT-345. [52] von Wolfersdorf J . , Hoecker R . , Sattelmayer T . , " A H y b r i d Transient Step-Heating Heat Transfer Measurement Technique Using Heater Foils and L i q u i d - C r y s t a l Thermography", M a y 1993, J . of Heat Transfer, Vol.115, pp.319-324. [53] Wang Z . , Ireland P . T . , Jones T . V . , Davenport R . , " A Color Image Processing System for Transient L i q u i d C r y s t a l Heat Transfer Experiments", J u l y 1996, J . of Turbomachinery, Vol.118, pp.421-427. [54] W i l d e G . L . , "The Design and Performance of H i g h Temperature Turbines i n Turbofan Engines", 1977, Tokyo Joint Gas Turbine Congress, pp.194-204. [55] W y s z e c k i G . , Stiles W . S . , "Colour Science - Concepts and Methods, Quantitative D a t a and Formulae", 1982, J o h n W i l e y &; Sons, 2nd edition. [56] Source: Analog Devices Inc., http://www.analog.com/ [57] Source: Commission Internationale de L'Eclairage, h t t p : / / w w w . c i e . c o . a t / c i e / [58] Source: Duro-Test Lighting, http://www.durotest.com/ [59] Source: Truevision T G A File Format Specification, version 2.2, download from the F T P site at h t t p : / / w w w . t r u e v i s i o n . c o m / S u p p o r t / [60] Thermographic Measurements L t d . - Thermax, B a n k House, Neston R o a d , B u r t o n , South W i r r a l L64 5 T A , England; h t t p : / / w w w . t h e r m a x . c o m ; e-mail: tmc@thermax.com  158  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0080946/manifest

Comment

Related Items