An algorithmic/heuristic methodology for the optimum utilization of technological innovations as applied to diamond coating

MISSING IMAGE

Material Information

Title:
An algorithmic/heuristic methodology for the optimum utilization of technological innovations as applied to diamond coating
Physical Description:
xvii, 275 leaves : ill. ; 29 cm.
Language:
English
Creator:
Stanfill, Richard Keith
Publication Date:

Subjects

Subjects / Keywords:
Mechanical Engineering thesis, Ph. D
Dissertations, Academic -- Mechanical Engineering -- UF
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1995.
Bibliography:
Includes bibliographical references (leaves 269-274).
Statement of Responsibility:
by Richard Keith Stanfill.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002059453
oclc - 33857598
notis - AKP7516
System ID:
AA00009040:00001


This item is only available as the following downloads:


Full Text











AN ALGORITHMIC/HEURISTIC METHODOLOGY FOR THE OPTIMUM
UTILIZATION OF TECHNOLOGICAL INNOVATIONS AS APPLIED TO
DIAMOND COATING














By

RICHARD KEITH STANFILL


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1995















ACKNOWLEDGEMENTS


It is my privilege to acknowledge the help and guidance received from many individu-

als during my doctoral studies. I am indebted to my chairman, Professor Ali Seireg, for his

patience, encouragement and consistently invaluable advice. His vision and mastery of

mechanical systems are a constant source of inspiration. I would like to thank Dr. Jack

Elzinga for developing my understanding of the manufacturing system, serving on my

doctoral committee and being a good coach. Dr. Carl Crane deserves great thanks for

being an excellent supervisor for my graduate assistantships in the SUCCEED program.

Dr. Crane's hands-off management style gave me great latitude while developing and

delivering the Design and Build curriculum-a luxury few research/teaching assistants are

allowed. I owe a thank you to Dr. Doug Dankel and Dr. Scott Smith for serving on my

committee and being excellent teaching role models. I would also like to acknowledge the

help from Dr. Rajiv Singh for sharing his resources and providing me with details of the

diamond coating process.

Thanks also go to Mr. Jose Rodriguez for taking the leap from industry with me. Jose's

vision and collaboration have led to many successful outcomes, chief among them our

Design and Build course. I would like to thank Ms. Angela Maria Ventura-Medyk for all

her help during the Design and Build program. Without her perseverance, we never would

have gotten the program running. I would also like to acknowledge the help from the

office staff in the Center for Intelligent Machines and Robotics, the Department of

Mechanical Engineering and the Department of Electrical Engineering.

And lastly, I would like to thank my wife Linda for all her loving support, encourage-

ment, patience and proof-reading prowess during my doctoral studies.















TABLE OF CONTENTS




ACKNOW LEDGEM ENTS ...................................................................................... ii

LIST OF TABLES .................................................................................................... vii

LIST OF FIGURES .................................................................................................. xii

ABSTRACT ........................................................................................................... xvii

CHAPTERS

1 INTRODUCTION ...................................................................................... 1

Research M motivation .................................................................................... 1
Importance: Technology as a Worldwide Competitive Product ............ 2
Barriers to Technology Transfer ........................................................... 3
Framework of Proposed M ethodology ........................................................ 3
Plausible Application Selection ............................................................. 3
Algorithmic Analyses of Plausible Applications .................................. 5
Cost/benefit Analysis ............................................................................. 5
Outline of the Reported Study ..................................................................... 6
"New Technology" Knowledge Base ................................................... 6
"Needs and Applications" Knowledge Base.......................................... 6
Inference Method for Matching Technology to Needs and Applications 7
Algorithmic M ethodology ..................................................................... 8
Cost/benefit M odeling........................................................................... 9
Considered Example for Laboratory Process Technology Transfer ............ 9
M ultilayer Diamond Thin Film Coating ............................................... 10
Tribological considerations................................................................... 10
Preview of Subsequent Chapters ................................................................ 10

2 LITERATURE REVIEW ......................................................................... 12

Technology Transfer ................................................................................... 12
Coating Technology .................................................................................... 15
Tribology ..................................................................................................... 16
M machine Elements....................................................................................... 17
Knowledge-Based Systems......................................................................... 17









Attribute Matching...................................................................................... 18
Software Engineering.................................................................................. 18

3 THE THERMAL PROBLEM ....................................................................... 20

Transient Heat Transfer ............................................................................... 20
A Heat Source With a Hertzian Distribution Moving Over a Layered
Semi-Infinite Solid.......................................................................... 21
Extension to Multi-Layered Semi-Infinite Solid................................... 24
Thermal Stress Considerations ................................................................... 30
Thermal Stress Relationships................................................................ 31
Predicting Fatigue Life Debit Due to Thermal Stresses ....................... 35

4 PROCESS ATTRIBUTES ......................................................................... 39

Properties of Diamond ................................................................................ 39
Pre-Coating Surface Treatment................................................................... 40
Chemical Vapor Deposition of Diamond.................................................... 42
Cost Modeling............................................................................................. 45
Near-Term Inputs .................................................................................. 46
Long-Term Inputs ................................................................................. 46
Other Cost Factors ............................................................................... 47
Near-Term Outputs ............................................................................... 47
Long-Term Outputs............................................................................... 48

5 POTENTIAL APPLICATIONS ................................................................ 51

Application to Gears ................................................................................... 51
Gear Nomenclature............................................................................... 51
Governing Equations for Gear Pairs..................................................... 52
Minimizing Contact Temperature in Gear System Design......................... 53
Design of a Single Reduction Gear Set....................................................... 55
Example Single Reduction Gear Set.................................................... 58
Comparison of Carbon Steel and Stainless Steel Gears........................ 58

6 THE MATCHING PROCESS ................................................................... 61

The Matching Process Role in the Technology Utilization Methodology..... 61
The Structure of the Developed Matching Process................................... .. 61
Creation of a Structured Attribute Representation................................ 63
Matching Atomic Values............................................................... ........ 63
Rating Discovery-Application Matches................................................ 64
Prototype System Development........................................................... 64
Road Map for the Remaining Matching Process Discussion................ 65
Elements of the Matching Process............................. ............................. 65
Technology Attribute Hierarchy Development .......... ................. .... 65
Object-Oriented Implementation ....................................... ................... 69









The M watching Engine.................................................................................. 74
M watching Process Overview ................................................................. 75
Strategies for M atching......................................................................... 78
CLIPS Expert System Shell........................................................................ 81
Defining Classes.................................................................................... 84
Defining Instances................................................................................. 88
Defining Rules ...................................................................................... 93
Defining Knowledge Base Queries....................................................... 95
Loading CLIPS Constructs................................................................... 97

7 APPLICATION OF THE QUALITATIVE MATCHING STRATEGY ..... 101

The Implementation Steps of the Developed Process................................. 101
Construction of the Prototype Application ........................................... 101
Populating the Knowledge Bases.......................................................... 103
Executing the System ............................................................................ 103
Building Instances of the Technology Classes............................................ 105
Decisions Common to DISCOVERY and APPLICATION Instances..... 105
Using the Technology Attribute Class Templates................................. 106
DISCOVERY Instances........................................................................ 111
APPLICATION Instances..................................................................... 117
Building Rules ............................................................................................ 124
Decomposing Qualitative Requirements .............................................. 124
Qualitative M watching ............................................................................ 126
Sample Execution of the Qualitative Matching Engine.............................. 127
Overview of the Process ....................................................................... 127
Loading the Knowledge Base Constructs............................................. 129
Executing the system ............................................................................ 131

8 THE QUANTITATIVE PROCESS ........................................................... 140

Elements of the Quantitative Process ......................................................... 140
Algorithmic M odule.............................................................................. 140
Cost/Benefit Analysis............................................................................ 144
Ranking of Applications ....................................................................... 145
Application of the Quantitative M watching .................................................. 147
Qualitative Analysis M odule Inputs...................................................... 147
Results................................................................................................... 149

9 SUMMARY, CONCLUSIONS AND RECOMMENDATIONS ........... 155

Summary........................................... .... ................... ...................................... 155
Conclusions ................................................................................. ........ ........... 155
Recommendations............................... .. .............................................. ..... 156









APPENDICES

A RELEVANT TRIBOLOGICAL EQUATIONS ........................................ 159

B PROGRAM LISTINGS ................................................................................ 166

C KNOWLEDGE BASE OUTPUT FILES .................................................. 214

D TECHNOLOGY ATTRIBUTE CLASS TEMPLATES ........................... 222

REFEREN CES ......................................................................................................... 269

BIOGRAPHICAL SKETCH .................................................................................... 275















LIST OF TABLES


Table aR

1-1. Potential applications knowledge base .................................................... 7

4-1. Properties of diamond.............................................................................. 40

4-2. Deposition parameter space used in the growth of diamond films by
microwave-plasma-assisted CVD ............................................................ 45

4-3. Near-term inputs. ...................................................................................... 47

4-4. Long-term inputs...................................................................................... 48

4-5. External cost factors................................................................................. 49

4-6. Process gas costs in the near and long term (1992 $/m3)..................... 49

4-7. Near-term outputs...................................................................................... 50

4-8. Long-term outputs ................................................................................ ....... 50

6-1. Subclasses defined within the Physical properties class for a sample
knowledge base ........................................................................................... 67

6-2. Subclasses defined within the Process properties class for a sample
knowledge base .......................................................................................... 68

6-3. Combinatorial explosion for the rules-search-for-facts method .............. 80

6-4. COOL query system functions......................... ..................................... 97

7-1. Sample phylums and associated technologies ........................................... 106

7-2. MELT-POINT class template................................................................ 107

7-3. DIAMOND-CVD-MELT-POINT instance slot values......................... 109

7-4. Work sheet for gathering technology physical properties........................... 110

7-5. Work sheet for DIAMOND-CVD thermal-physical properties............... 111









7-6. Work sheet for DIAMOND-CVD mechanical-physical properties......... 112

7-7. Work sheet for DIAMOND-CVD environmental-physical properties ..... 112

7-8. Work sheet for DIAMOND-CVD empirical dimensionless relationships.. 114

7-9. Work sheet for DIAMOND-CVD empirical dimensional relationships..... 115

7-10. Qualitative requirements for GEARS ...................................................... 118

7-11. Work sheet for GEARS design parameters.............................................. 118

7-12. Work sheet for GEARS empirical relationships ...................................... 120

7-13. Work sheet for GEARS empirical relationships, and temperature and
stress prediction. ....................................................................................... 121

7-14. Work sheet for GEARS constants ............................................................ 122

7-15. Work sheet for HI-LOAD- ROLLER-BRG thermal-physical properties... 122

7-16. Work sheet for HI-LOAD- ROLLER-BRG mechanical-physical
properties.................................................................................................. 123

7-17. Work sheet for HI-LOAD-ROLLER-BRG environmental-physical
properties........................................................... ....................................... 124

8-1. Design parameters for technology match ................................................ 148

8-2. Design bounding parameters for technology match ................................ 149

D-1. Slot descriptions for technology classes inheriting characteristics from
TECHNOLOGY, LEVEL-1-ATTRIBUTE through LEVEL-4-
ATTRIBUTE, and TECHNOLOGY-PROPERTY................................... 223

D-2. Slot descriptions for classes inheriting characteristics from
GOVERNING-EQUATION.................................................................... 224

D-3. Slot descriptions for classes inheriting characteristics from miscellaneous
parent classes ........................................................................................... 224

D-4. TECHNOLOGY class template............................................................... 225

D-5. LEVEL-1-ATTRIBUTE class template................................................... 225

D-6. LEVEL-2-ATTRIBUTE class template................................................... 225

D-7. LEVEL-3-ATTRIBUTE class template................................................... 226









D-8. LEVEL-4-ATTRIBUTE class template...................................................... 226

D-9. TECHNOLOGY-PROPERTY class template............................................. 226

D-10. MFG-PROCESS-METHOD-LIST class template .................................. 227

D-11. MATERIAL-ATTRIBUTE class template.................................................. 227

D-12. MANUFACTURING-ATTRIBUTE class template............................... 227

D-13. OTHER-ATTRIBUTE class template......................................................... 228

D-14. GOVERNING-EQUATION class template................................................ 228

D-15. PHYSICAL-PROPERTY class template.................................................... 228

D-16. ECONOMIC-PROPERTY class template ............................................... 229

D-17. OTHER-PROPERTY class template ....................................................... 229

D-18. PROCESS-PROPERTY class template ................................................... 229

D-19. THERMAL-PROPERTY class template .................................................... 230

D-20. MECHANICAL-PROPERTY class template.......................................... 230

D-21. ENVIRONMENTAL-PROPERTY class template .................................. 231

D-22. PROCESS-ECONOMICS class template................................................... 231

D-23. PROCESS-PARAMETER class template .................................................. 232

D-24. PROCESS-METHOD class template ...................................................... 232

D-25. PROCESS-LIMITS class template.......................................................... 233

D-26. EMPIRICAL-RELATIONSHIP class template.................................... 233

D-27. OBJECTIVE-FUNCTION class template............................................... 234

D-28. DESIGN-PARAMETER class template..................................................... 234

D-29. CONSTANT class template........................................................................ 235

D-30. QUALITATIVE-REQTS class template.................................................... 235

D-31. THERMAL-CONDUCTIVITY class template ............... .................... 236

D-32. SPECIFIC-HEAT class template ............................................................. 237









D-33. THERMAL-EXPANSION class template............................................... 238.

D-34. THERMAL-DIFFUSIVITY class template............................................ 239

D-35. THERMAL-SHOCK class template........................................................... 240

D-36. USAGE-TEMPERATURE class template.............................................. 241

D-37. MELT-POINT class template...................................................................... 242

D-38. YOUNGS-MODULUS class template.................................................... 243

D-39. SHEAR-MODULUS class template..................................................... 244

D-40. POISSON-RATIO class template ............................................................ 245

D-41. DENSITY class template......................................................................... 246

D-42. FRICTION-COEFFICIENT-DRY class template.................................... 247

D-43. FRICTION-COEFFICIENT-LUB class template.................................... 248

D-44. ROUGHNESS class template.................................................................. 249

D-45. HARDNESS class template........................................................................ 250

D-46. TOUGHNESS class template .................................................................. 251

D-47. CRACK-RESISTANCE class template...................................................... 252

D-48. LUB-FILM-THKNS class template ........................................................ 253

D-49. LUB-COMPATIBILITY class template .................................................. 254

D-50. OXIDATION-RESISTANCE class template........................................ 255

D-51. MATERIAL-COST class template .......................................................... 256

D-52. CAPITAL-EQUIPMENT-COST class template............. ................... 257

D-53. LABOR-COST class template.................................................................... 258

D-54. POWER-COST class template ................................................................ 259

D-55. OVERHEAD-COST class template ........................................................ 260

D-56. PROCESS-PRESSURE class template....................................................... 261

D-57. PROCESS-ATMOSPHERE class template............................................. 262









D-58. PROCESS-SURFACE-TEMPERATURE class template...................... 263

D-59. PROCESS-LIMITS-THKNS class template ........................................... 264

D-60. PROCESS-LIMITS-SPATIAL class template ..................................... 265

D-61. PROCESS-LIMITS-DEPOSITION class template................................. 266

D-62. METHOD-PREPARATION class template.................. ...................... 267

D-63. METHOD-APPLICATION class template................. ....................... 267

D-64. METHOD-FINISHING class template....................................................... 268














LIST OF FIGURES


Figure Pag

1-1. The technology transfer framework for the proposed methodology ......... 4

3-1. Moving semi-infinite solid under a stationary heat source ...................... 22

3-2. Layered semi-infinite solid moving under a stationary heat source with a
Hertzian distribution. ................................................................................ 23

3-3. Maximum temperature rise (C) in AISI 304 substrate (Tss) and diamond
surface layer (Td) moving under a stationary heat source with a Hertzian
distribution............................................................................................... 25

3-4. Maximum temperature rise (C) in AISI 304 substrate (Tss) and silicon
nitride surface layer (Tif) moving under a stationary heat source with a
H ertzian distribution. ................................................................................ 26

3-5. Two-layered semi-infinite solid moving under a stationary heat source
with a Hertzian distribution. ..................................................................... 27

3-6. Dimensionless temperature rise (prior to data manipulation) in substrate
surface and coating layer surface for diamond on AISI 304 stainless steel
with a silicon nitride interface layer......................................................... 28

3-7. Dimensionless temperature rise (following data manipulation) in
substrate surface and coating layer surface for diamond on AISI 304
stainless steel with a silicon nitride interface layer.................................. 29

3-8. Temperature rise (C) in substrate, interface layer and diamond layer for
diamond on AISI 304 stainless steel with a silicon nitride interface layer. 30

3-9. Temperature rise (C) in substrate, interface layer and diamond layer for
diamond on AISI 304 stainless steel with a silicon nitride interface layer
(velocity increase with length of contact constant).................................. 31

3-10. Temperature rise (C) in substrate, interface layer and diamond layer for
diamond on AISI 304 stainless steel with a silicon nitride interface layer
(velocity constant with length of contact increase).................................. 32

3-11. Variables used in calculating thermal stress in multilayer coatings......... 33









3,12. Normal stress (Pa) for various thicknesses (mm) of diamond and
interface layer. Substrate is AISI 304 stainless steel and interface layer is
silicon nitride. Note: hif = hd = ho, a1 = od, 02 = aof and 03 = ss......... 34

3-13. Shear stress (Pa) for various thicknesses (mm) of diamond and interface
layer. Substrate is AISI 304 stainless steel and interface layer is silicon
nitride. Note: hif = hd = ho, T = dif and 2 = Tifss................................. 35

3-14. von Mises stress (Pa) as a function of coating thickness (mm) in
substrate. Substrate is AISI 304 stainless steel and interface layer is
silicon nitride. Note: hif = hd = ho........................................................... 37

3-15. Life improvement in cycles versus thickness of diamond film in mm.
Substrate is AISI 304 stainless steel and interface layer is silicon nitride.
Note: hif = hd = ho and contact stress level is 1000 MPa........................ 38

4-1. Cross sectional view of substrate surface following laser modification,
seeding and film growth........................................................................... 41

4-2. Seeded silicon substrate surface following a short duration (10 minute)
diamond growth period ............................................................................ 42

4-3. Final silicon substrate's surface topology following the complete
diamond film growth cycle. ...................................................................... 43

4-4. Schematic illustration of microwave-plasma CVD apparatus used for the
growth of diamond films.......................................................................... 44

4-5. Sequential stages in technical cost modeling (TCM) .............................. 46

5-1. Temperature rise at the starting point of contact (c = 200)...................... 55

5-2. Nominal thermal and contact stress for standard gear teeth .................... 56

5-3. Nominal thermal and contact stress for standard gear teeth .................... 57

5-4. Comparison of carbon steel and stainless steel gears under the same
operating conditions. a) Temperature rise (F) in the pinion; b) Heat flux
(w att/mi )............................................................................................. ..... 60

6-1. Framework of the algorithmic/heuristic technology evaluation process.... 62

6-2. Structure for the knowledge base technology class hierarchy tree .......... 70

6-3. The quadrilateral hierarchy illustrating multiple inheritance................... 72

6-4. Structure of a class definition................................................................... 74









6-5. Inheritance mechanism for a child class with multiple parent classes ..... 75

6-6. Matching rules with facts using a rules-search-for-facts approach.......... 79

6-7. Comparison of two strategies for matching rules with facts.................... 81

6-8. Instantiation of a typical rule within a Rete network............................... 82

6-9. Instantiation of two rules within a Rete network ..................................... 83

6-10. CLIPS knowledge base development system multiple document interface
as implemented on the Macintosh computing platform........................... 85

6-11. Edit of CLIPS source file and explanation of CLIPS application menus.. 86

6-12. CLIPS defclass construct. ........................................................................ 87

6-13. Definition of the TECHNOLOGY and MATERIAL-ATTRIBUTE classes. 89

6-14. CLIPS make-instance construct............................................................... 90

6-15. Interactive creation of knowledge base DISCOVERY instances in the
CLIPS environm ent. ................................................................................. 91

6-16. CLIPS definstances construct. .................................................................. 92

6-17. Definition of the DIAMOND-CVD-K instance....................................... 92

6-18. Definition of the THERMAL-CONDUCTIVITY class........................... 93

6-19. Results of a print message sent to the DIAMOND-CVD-K instance ...... 94

6-21. CLIPS defrule construct........................................................................... 94

6-20. CLIPS integrated editor window showing example of an external source
file with a definstances construct .................................................... 95

6-22. Definition of the MATCH-REQT-CAPABILITY rule............................. 96

6-23. CLIPS do-for-all-instances query construct............................................. 97

6-24. Instance-set queries built into the action of a rule ................................... 98

6-25. Sample external source files successfully loaded into the knowledge base. 99

6-26. Batch file and sample execution of the batch command within the CLIPS
environm ent. ............................................................................................. 100

7-1. Process for building knowledge base entities for a technology ............... 104









7-2. Definition of a MELT-POINT instance within a definstances construct..... 108

7-3. Resulting instance DIAMOND-CVD-MELT-POINT slot values
following definstances invocation............................................................ 109

7-4. Source file with a definstances construct being built ............................... 110

7-5. Source file "CVD-Diamond instances.CLP" under development within
the CLIPS environment............................................................................ 116

7-6. Pseudo-code description of the DECOMPOSE-CONTACT-STRESS rule. 125

7-7. Loading and execution of qualitative matching process knowledge base
system within the CLIPS environment .................................................... 128

7-8. Skeleton of the CLIPS batch file "Technology KBS.bat."........................ 129

7-9. Opening the batch file from within the CLIPS environment ................... 130

7-10. Launching the batch file directly from the Desktop................................. 131

7-11. Launching CLIPS directly from the Desktop .......................................... 131

7-12. M ain menu display................................................................................... 132

7-13. M atch table menu display. ........................................................................ 133

7-14. Match table display showing the evaluation of two attribute matches....... 133

7-15. Match statistics including the overall score for the DIAMOND-CVD-
COATING/HI-LOAD-ROLLER-BRG match ......................................... 134

7-16. Match statistics including overall score for the DIAMOND-CVD-
COATING/GEARS match ....................................................................... 134

7-17. Menu display for output of analysis equations ........................................ 135

7-18. Output of governing equations for GEARS in the source file "gears.h." ... 136

7-19. Output of design parameters and constants for GEARS in the source files
"gearparams.h" and "geardefines.h," respectively ................................... 137

7-20. Output of governing equations for CVD-DIAMOND-COATING in the
source file "diamond_cvd.h.".................................................................... 138

7-21. Exiting from the menu system ................................................................. 139

8-1. The quantitative/algorithmic analysis and ranking process ..................... 141









8-2. Mathcad document illustrating the use of standard math notation .......... 143

8-3. Mathcad document illustrating its capability for handling built-in units.... 144

8-4. Mathcad document illustrating plotting capability. ................................. 145

8-5. Data flow for the quantitative analysis procedure.................................... 146

8-6. Empirical multilayer dimensionless relationships within Mathcad
docum ent.................................................................................................. 147

8-7. Coating cost normalized to increase in life in the pinion for three gear
ratios holding the ratio of center distance to facewidth at 2:1 ................. 150

8-8. Coating cost normalized to increase in life in the pinion for three gear
ratios holding the ratio of center distance to facewidth at 3:1 ................. 150

8-9. Coating cost normalized to increase in life in the pinion for three gear
ratios holding the ratio of center distance to facewidth at 4:1 ................. 151

8-10. Increase in life in the pinion for three gear ratios holding the ratio of
center distance to facewidth at 2:1........................................................... 151

8-11. Increase in life in the pinion for three gear ratios holding the ratio of
center distance to facewidth at 3:1........................................................... 152

8-12. Increase in life in the pinion for three gear ratios holding the ratio of
center distance to facewidth at 4:1........................................................... 152

8-13. Coating cost for the pinion to meet full life for three gear ratios holding
the ratio of center distance to facewidth at 2:1 ........................................ 153

8-14. Coating cost for the pinion to meet full life for three gear ratios holding
the ratio of center distance to facewidth at 3:1 ........................................ 153

8-15. Coating cost for the pinion to meet full life for three gear ratios holding
the ratio of center distance to facewidth at 4:1 ........................................ 154














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree Doctor of Philosophy

AN ALGORITHMIC/HEURISTIC METHODOLOGY FOR THE OPTIMUM
UTILIZATION OF TECHNOLOGICAL INNOVATIONS AS APPLIED TO DIAMOND
COATING

by

RICHARD KEITH STANFILL

August 1995

Chairman: Dr. Ali A. Seireg
Major Department: Mechanical Engineering

This research proposes a systematic procedure for the effective transfer of laboratory

discoveries to the parametric design of new mechanical products or the improvement of

existing products. The methodology is based on an associative object-oriented knowledge

base that incorporates information on the new technology's features, attributes, cost and

potential use. An expert system matches the new technology with potential applications.

The resulting application matches are then optimized for life-cycle design criteria. Lastly,

cost-benefit analyses of the different product applications are compared and ranked. The

highest ranked application indicates where the greatest return on investment is possible for

further development of the new technology.

A recent University of Florida laboratory discovery for diamond coating of tribologi-

cal surfaces serves as an illustration of the methodology. This process shows promise as a

practical surface treatment for tribological applications. The methodology is demonstrated

for the case of gear pairs with different materials, center distance and gear ratio.


xvii














CHAPTER 1
INTRODUCTION


1.1 Research Motivation

Efficient transfer of laboratory discoveries into product applications is essential for

current and future competitiveness in the world marketplace. This transfer process is very

difficult because it requires intimate knowledge of the available technology as well as the

needs of the marketplace. Market research is a well established industrial activity. The

work in this area, however, is generally undertaken by business-oriented staff with a

casual knowledge of what technological development can accomplish. In other words,

marketing is generally based on market pull without consideration for technology push.

Critical components of this transfer process are the assessment of economic benefit of

a new technology and selection of appropriate product applications. Industry needs a tool

set to assess the financial risk associated with research and development of new technolo-

gies. A formal methodology for a domain-independent laboratory transfer process will

help industry open new markets and develop new or improved products.

Noted Harvard economist John Kenneth Galbraith [Wein87d] sees the market power

in the hands of the producer:

The myth of economics is the sovereign consumer, but the reality of life is the
producer who reaches out to take control, to influence the consumer, and to
establish a mode of life serviceable to the producer. This mode is one in which
the emphasis is on consumer achievement, which has always been at the heart
of producer power (p. 52).

Galbraith further states "Beyond technology comes good design... so one part of the solu-

tion [to rejuvenating our older, more expensive work force] is to emphasize good design,

along with good technology," (p. 51). He also points out that "the engineer brings in to









existence products that have their justification in being unique. And therefore engineers

are to some degree monopolistic... The market can give a special power to the producer

of a unique product," (p. 52).

1.1.1 Importance: Technology as a Worldwide Competitive Product

Efficient technology transfer has become a critical need in revitalizing industry's posi-

tion in world markets. To survive, products must be competitive inside and outside of

national borders. Seymour Baron [Baro90], associate director for applied programs at

Brookhaven National Laboratory, points out:

Indeed, competition is not viewed anymore between General Motors and Ford
or between GE and Westinghouse but rather between our auto industry and the
Japanese auto industry and our high technology versus those in Europe and the
Far East. The market is not a U.S. market but a world market, and technology
is a worldwide competitive product (p. 38).

From a more global perspective, the economic stability of entire nations may rest in suc-

cessful transfer of technology from defense-related research into commercial products.

The emerging independent states from within the former Soviet Union have a wealth of

research talent previously devoted to military activities. The U.S. Department of Energy

national laboratories and various U.S. industry partners have recognized this talent and

have proposed joint programs to assist in redirecting these research institutions toward

productive nonmilitary applications [Hnat93]. Congress has passed legislation-the

FREEDOM Support Act-to provide U.S. aid to "facilitate the conversion of military

technologies, capabilities and defense industries to civilian applications.. and activities

which will assist the independent states to transition to market economies and become

responsible members of the global economic community," (p. 1).









1.1.2 Barriers to Technology Transfer

Seymour Baron [Baro90] outlines the hurdles to technology transfer from the national

laboratories to industry. Prominent among these hurdles is that

laboratories. do research for research's sake and not with any product in
mind. Results are reported openly and discussed fully at conferences. no
cost-benefit analysis to determine whether the research will be profitable. .
Industry... approaches research mainly as short term missions for reasonable
return on investment (p. 39).

This insensitivity to economic issues directly affects the university-lab-to-industry tech-

nology transfer process. The deficiency can be overcome by developing a tool for labora-

tories to perform systematic, quantitative cost/benefit analyses. In order to entice industry

to participate in the technology transfer process, laboratories need the ability to demon-

strate better return on investment (ROI) for their process over traditional approaches.

1.2 Framework of Proposed Methodology

The proposed methodology is a two-stage process. The first stage is a heuristics-driven

knowledge base management system for qualitatively matching new technologies to

potential product applications. The knowledge base management system provides the sec-

ond stage with discovery-application matches. Stage two is an algorithmic module for

quantitative evaluation of the match and for producing cost/benefit analyses. After multi-

ple discovery-application matches are evaluated, a ranking of the applications for the new

technology can be produced. Ultimately, these modules will interact with one another

inside a common, intelligent framework. A conceptual depiction of the system is shown in

Figure 1-1.

1.2.1 Plausible Application Selection

Plausible product applications for emerging technologies are selected within three

interacting modules of a knowledge-based system. A knowledge base containing pertinent

physical and process information for new technologies interacts with a knowledge base

containing product applications-including current limitations and desired capabilities. A









Knowledge Base Management System



Emerging Potential
Technologies Applications

(Discoveries K.B.) (Applications K.B.)



SInferencing
^e Process y"
Matching
Engine




Qualitative Matching

S .. ..Quantitative Analyses
\ /& Ranking
Discovery- -
Application
Matches Algorithmic
Module






Cost/Benefit
Analyses Ranking

t 2. .......
3 .......


n. .......


e 1-1. The technology transfer framework for the proposed methodology.


*-1 a


Figure









third module uses an inferencing strategy to match potential product applications with

new technology processes. The product of these three interacting modules is a ranked list

of plausible product applications.

1.2.1.1 New technology knowledge base

The knowledge base for the new technologies contains a structured representation of

process information. Knowledge regarding the new technologies' features, attributes,

costs, physics, processing equipment involved, etc. is encoded in the knowledge base.

1.2.1.2 Needs and applications knowledge base

Knowledge regarding existing product domains is captured in a needs and applications

knowledge base. This information includes product features, attributes, costs, physics,

current processing constraints, desired capabilities and processing equipment involved.

1.2.1.3 Inference engine/pattern matcher

The inference engine searches the new technologies knowledge base and the needs

and applications knowledge base for potential application matching. The new technology

and application matched pairs are provided in a ranked list. Heuristics in the inferencing

module produce a qualitative ranking.

1.2.2 Algorithmic Analyses of Plausible Applications

Qualitative discovery-application matches are passed to an algorithmic analysis mod-

ule for evaluation. The module gathers relevant functional relationships and design param-

eters for each match. An objective of maximum cost/benefit is formulated and the design

space is bound by user-provided constraints. A simulation-utilizing the functional rela-

tionships gathered previously-is run to generate a series of feasible designs. The results

are then post-processed for cost/benefit.

1.2.3 Cost/benefit Analysis

Before the discovery-application matches can be ranked, the results of the algorithmic

module are post-processed. The results of this activity are a series of graphs relating the









cost of implementing the discovery versus increase in performance characteristics for the

application. These graphs illuminate circumstances within an application where the dis-

covered technology is worth pursuing. After multiple discovery-application matches are

evaluated for cost/benefit, a ranked list can be produced. The ranking suggests the best

investment of research funding for further development of the new technology.

1.3 Outline of the Reported Study

The objective of this study is to develop a technology transfer methodology which

incorporates information about a new technology, potential applications, and a quantita-

tive procedure for evaluating the ranking of the most beneficial use and advantages over

existing processes. The emphasis is on "what can be" rather than "what is." The methodol-

ogy is prototyped in a two-level decision-making system capable of selecting a technol-

ogy-product/application match, evaluating the design of the product and producing a cost/

benefit analysis of the technology transfer process. The final output is a ranked list of new

technology-application matches.

1.3.1 "New Technology" Knowledge Base

The knowledge base for new technology is developed for a recent laboratory discov-

ery that shows great promise for tribological surface applications-multilayer diamond

thin-film coating. Knowledge about the process capabilities, limitations, surface physics,

manufacturing processes involved, cost factors and design parameters is encoded in a pro-

duction system knowledge base.

Although the contents of the new technology knowledge base are domain specific to

the diamond coating process, the underlying knowledge base structure is domain indepen-

dent. This knowledge base is designed with extensibility in mind.

1.3.2 "Needs and Applications" Knowledge Base

A knowledge base containing potential applications of the new technology is devel-

oped. Knowledge about the potential applications is encapsulated in a production system

knowledge base. The knowledge base contains current properties, limitations and process









parameters for the application. Other important information included is desired properties,

manufacturing processes involved, cost factors and design parameters. A skeleton of this

knowledge base is provided in Table 1-1.

Table 1-1. Potential applications knowledge base


Product Domain Soecific Product


High Sliding, Con-
centrated Contacts


high K, p, C
high hardness
low EcaAT
low surface roughness
high thermal shock resistance
large lubricating film thickness


Bearings rolling element low peak coefficient of friction
isothermal/Newto- low AT
nian regime
high temperature
corrosive environment
Brakes frictional high K, p, C
Clutches frictional low EaAT
high thermal shock resistance
high coefficient of friction
stable coefficient of friction
high hardness
long life


The needs and potential applications knowledge base is similar to the new technology

knowledge base structure in that it is domain specific to the various application areas

described above, but the underlying structure is domain independent. This knowledge

base is designed with extensibility in mind. Knowledge regarding new potential applica-

tion domains can be added as it is acquired.

1.3.3 Inference Method for Matching Technology to Needs and Applications

An inference engine provides the capabilities for matching the new technology to

potential applications. A quantitative procedure for ranking application alternatives is

developed. Ranking is important because high payoff applications should be attacked first.









The CLIPS production system expert shell provides a robust pattern-matching capability

that was beneficial in developing a reliable matching and ranking methodology [Giar91].

CLIPS has the capability of representing knowledge in the form of facts, rules and proce-

dures. Tew [Tew87] developed a knowledge-based approach to searching an on-line soft-

ware application database using the EXSYS expert system shell. EXSYS, like CLIPS, is

also based on the production rules approach to knowledge representation.

Regardless of the inference engine used to implement the matching methodology, the

critical exercise is developing a reliable quantitative method. It is also important to note

the separation of the knowledge from its use. As Gonzales and Dankel [Gonz93] point out,

it is this separation that provides the true power of knowledge-based systems:

This separation allows us to develop different applications by having to create
only a new knowledge base for each application. The generic reasoning tech-
nique (i.e., the inference engine) is not modified. For example, the basic trou-
bleshooting approach used in medical diagnosis is similar to that used by a
mechanic in diagnosing a fault in an automobile. Only the domain is different.
Once the generic knowledge is encoded, it can be applied to either domain,
thereby greatly simplifying the development process (p. 1).

The inferencing process used to match new technology to potential product applications is

based on a technique developed by Forgy [Forg82]-the Rete algorithm-for efficient

many-many pattern matching. The Rete algorithm is integral to the CLIPS system used in

developing the knowledge bases for this study.

1.3.4 Algorithmic Methodology

Ultimately, the algorithmic analysis module will be a tightly integrated, interactive

modeling system. The prototype system developed for this study relies heavily on flexible

engineering mathematical modeling tools and spreadsheet applications. This approach

was taken to prove the concept while minimizing the amount of software development.

The simulations are complicated by the use of dimensional data. Handling units in tradi-

tional, from-the-ground-up software development projects requires a great deal of effort

that could be channeled towards larger goals. The engineering modeling tool used in









building the simulations provided a powerful, built-in units-handler that proved indispens-

able.

The spreadsheet approach does have its drawbacks. Among them is a loose system

integration which places a burden on the user to make sure that the proper data gets used

by the right application at the right time. Results are also slow to appear due to increased

interactive demands on the user. A tightly integrated analysis system would solve many of

these problems.

1.3.5 Cost/benefit Modeling

Cost/benefit analysis is used in this study to attach costs to performance gains. The

components of the cost figure may include yearly production volumes, unit costs, interest

rates and time on the market for the product. Rej and Alexander [Rej94] developed a com-

prehensive cost model for a plasma source ion implantation (PSII) facility-a process

similar to that used in applying diamond. The best pricing and marketing possibilities for

the new technology are explored relative to the cost of development and cost of full scale

production. Relevant data for this segment of the study are compiled in the references.

1.4 Considered Example for Laboratory Process Technology Transfer

The diamond coating process currently under development at the University of Florida

is used as the example technology transfer process. Applications to tribological systems

are investigated as sample products. A cost/benefit analysis for the research and develop-

ment effort is performed. A life-cycle cost comparison illustrates the advantages and dis-

advantages of these new application of technology versus traditional approaches.

One reason for this selection is the great potential for this process to generate consider-

able savings to the U.S. economy. A 1981 study by the ASME Research Committee on

Lubrication for the U.S. Department of Commerce [ASME81] estimated that the U.S.

economy loses $60 billion per year due to friction and surface wear.









1.4.1 Multilayer Diamond Thin Film Coating

The process examined in this study for product applications is in early stages of devel-

opment In essence, the process involves creating a very uniform surface roughness on

metallic surfaces using a laser. Since diamond has such a small thermal expansion coeffi-

cient, it is difficult to grow on metallic surfaces where the thermal mismatch is high. The

laser process allows for a larger contact surface area between the diamond film and the

metallic surface. The larger contact surface area is intended to grade the residual stresses

across the diamond film-allowing the film to better adhere to the surface. Since carbon

from the diamond tends to diffuse into the surface of iron-based substrates, it is desirable

to introduce a buffer layer between the diamond film and the substrate. For the purposes of

this study, it is assumed that effective bonding between the diamond film and the substrate

is realizable. Patents have been recently submitted on the laser surface preparation pro-

cess. There is no published information in the literature describing the process.

1,4.2 Tribological considerations

Tribological considerations for multilayered surface treatments include heat partition

and transient temperature effects, friction factors for high slide-to-roll ratios, effect of

lubricant properties on temperature and wear in sliding concentrated contacts, wear resis-

tance, elastohydrodynamics and thermohydrodynamic shear zone thickness.

1.5 Preview of Subsequent Chapters

Chapter 2 reviews the pertinent literature for this multidisciplinary study. Chapter 3

examines the thermal properties of diamond coated substrates. The thermal expansion

mismatch and the formation of graphite layers when coating steel with diamond necessi-

tates the use of a buffer layer. A procedure is developed for determining the transient heat

transfer properties of multilayered substrates. Chapter 3 also develops a method for

approximate evaluation of the nominal thermal stresses and predicting the fatigue life of

diamond coated steel. Process attributes for the diamond coating discovery are examined

in Chapter 4. Cost modeling for the coating process is also developed in Chapter 4.









Potential applications are investigated in Chapter 5. The case illustrations are for dif-

ferent gear applications. The qualitative matching process is developed in Chapter 6. This

methodology relies heavily on the analytical methods presented in Chapters 3, 4 and 5,

and ties together the process attributes presented in Chapter 4 and the potential applica-

tions of Chapter 5. In Chapter 7, a step-by-step procedure is presented for building the

qualitative matching knowledge base structure and for executing a sample matching pro-

cess. Chapter 8 provides a step-by-step procedure for taking the results of the qualitative

matching process and performing necessary quantitative analyses to determine a ranking

of potential applications. Chapter 9 summarizes the findings and the conclusions of this

study, and presents suggestions for further research.















CHAPTER 2
LITERATURE REVIEW

The problem under consideration incorporates information from a variety of disci-

plines. These include technology transfer, coating technology, Tribology and the technol-

ogy of machine elements whose design can be constrained by surface temperature-such

as gears, bearings and clutches/brakes.

Also relevant to this investigation are the disciplines of knowledge-based systems,

attribute matching and software engineering.

2.1 Technology Transfer

Technology transfer from the laboratory to industrial/commercial implementation has

recently received a large amount of public attention. Many state-assisted consortium,

such as Enterprise Florida, are actively soliciting researchers ready to commercialize their

research results. The shrinking Department of Defense budget has forced many contrac-

tors, once solely dependent on the government for funding, to turn to civilian applications

of their technology. Thus technology transfer is of a great interest to both the individual

researcher or small research team and the established high-technology manufacturing

base.

The published literature has many examples of the philosophical issues related to tech-

nology transfer. Among these issues are how to best manage the process, how to best dis-

seminate new ideas for possible transfer and how to best conduct the communication

process between the technology provider and the technology user/implementor. Other

commonly explored issues include the decision making process on choosing between

competing technologies for a particular application. There are few data on the analytical

aspects involved in evaluating potential applications for new discoveries.









Gibson and Niwa [Gibs91] proposed a new research area in "knowledge-based tech-

nology transfer," (p. 503). Their study models and integrates research and theory in the

areas of knowledge-based systems and technology transfer. The researchers found that

knowledge-based systems and technology transfer share a common objective: effective

knowledge transfer. In their reported work, they discuss three historically linear technol-

ogy transfer models based on the knowledge-consulting paradigm. These models all dis-

seminate technological knowledge in one direction-from technology expert to

technology user. The first of these models, the Appropriability Model, was in use during

the late 1940s through the 1950s. This model was based on "thinking good thoughts" and

publishing results, with the belief that good technologies would sell themselves. The Dis-

semination Model, in use during the 1960s and 1970s, emphasized diffusion of informa-

tion transfer of expert knowledge to users, opinion leaders or willing receptors. The

philosophy behind this model is that once the communication linkages are established, the

knowledge will flow. The third and most modem approach, the Knowledge Utilization

Model, emphasized the behavioral issues such as interpersonal communication between

technology researchers and technology receptors, and overcoming organizational barriers

and finding technology transfer facilitators.

The researchers proposed a new two-way approach, the Communication-Based Model

which enhances the exchange of knowledge from the technology provider and require-

ments from the technology user. In this model, technology transfer is a continuous, inter-

active process where idea exchange between provider and user is simultaneous and

continuous until "convergence of knowledge developers, and users in terms of technology

development, acceptance and application," ([Gibs91] p. 504). The Communication-Based

Model uses a knowledge-sharing methodology where communication is between expert

user and expert user. In this way knowledge is decentralized and transferred in both direc-

tions.









Sage [Sage89] reported a systems engineering technology management process meth-

odology supporting the research and development of emerging technologies. The study

presents issues related to the identification of new applications and the evaluation of their

potential. Sage emphasizes the dynamics of market push and pull and says "successful

innovation more frequently is driven by market pull than by technology push," (p. 311).

This report offers a framework for assessing technology potential for further development

from a managerial and societal perspective but does not propose an analytical framework.

An expert system technology evaluation tool developed by Siemens AG was reported

by Reminger [Remi91]. Siemens uses this tool to assess the viability of new technology

for further development. Over 1000 rules were implemented in the system to capture the

technology evaluation model. The tool diminishes the time that technology evaluation and

planning requires. The example cited was the evaluation of 11 technologies in 2 weeks

versus the pre-tool time of one year. The compressed time factor is a source of competitive

advantage for Siemens. No details were provided concerning the system's evaluation strat-

egy.

Luxhoj [Luxh94] compared alternative technology transition strategies using the Ana-

lytic Hierarchy Process (AHP) "as a justification methodology to integrate and evaluate

both quantitative and qualitative factors in this complex, multi-attribute problem domain,"

(p. 81). The problem domain studied was the Federal Aviation Administration's plan to

modernize its Air Traffic Control Towers with advanced technology flight management

systems. Luxhoj concluded that AHP may be successfully applied to evaluate multi-

attribute problems which consist of both quantitative and nonquantitative influence fac-

tors. AHP was developed by Saaty [Saat90] to provide a systematic way to make decisions

in complex, multi-factor situations where decision parameters are interdependent. Many

deficiencies have been reported for AHP and alternative decision support techniques have

been proposed [Ra91].









2.2 Coating Technology

The properties of diamond films and other synthetic diamond materials are collected in

several references. Among the most comprehensive are properties compiled by Field

[Fiel92] and Davis [Davi93]. The work summarized by Field [Fiel92] includes complete

property information in the appendix which is divided into general properties, mechanical

properties, thermal properties, optical and electrical properties, and industrial products.

Yoder [Yode93] summarizes mechanical and elastic properties of diamond and compares

these properties to common engineering materials. The study also includes a look at high

temperature and tribological applications of diamond.

Lux and Haubner [Lux93] discuss various low pressure methodologies for producing

diamond films. They also present some simple cost models for producing diamond-coated

cemented carbide cutting tool inserts using in situ chemical vapor deposition (CVD). The

wear and cutting applications examined in this study rely mainly on the superhard proper-

ties of diamond. The researchers conclude their report with a look at short- and long-term

outlooks for scaling up production of diamond coated substrates and free-standing dia-

mond sheets. Chief among these concerns is quality assurance for diamond films and

multi-layer approaches that will allow different grades of diamond surface to be produced.

Piano and Pinneo [Plan93] discuss goals of CVD diamond research approaches for

diamond property enhancement and future nonelectronic diamond applications. The appli-

cations discussed include the following:

1. mechanical components-such as machine tools, paper mill paper rollers and ceramic
roller bearings;

2. thermal components-such as heat spreaders for high power density devices, packing
substrates for high speed microprocessors and gas turbine combustion chamber liners;

3. optical components-such as x-ray deflector windows, liquid-cooled high power laser
mirrors and as a laser host material.

The researchers include both thermal and superhard properties of diamond films in their

potential application assessment.









Rej and Alexander [Rej94] developed a semiempirical model for the cost of a com-

mercial plasma source ion implantation (PSII) facility. The cost model estimates amor-

tized capital and operating expenses as functions of the surface area throughput T. The

model predicts a reasonably sized PSII facility should be able to treat a surface area of 104

m2 per year at a cost of $0.01 per cm2. Many of the same power and capital equipment

requirements are similar for CVD and PSII. The PSII facility cost model may be a useful

reference for full scale CVD production of diamond films.

2.3 Tribology

Tribology is "the branch of science and technology concerned with interacting sur-

faces in relative motion and with associated matters (as friction, wear, lubrication, and the

design of bearings)," ([Jost91], p. 129). The published literature contains many papers rel-

evant to this study. The main area of importance focuses on sliding concentrated contacts

and the associated heat generation.

O'Donoghue and Cameron [O'Do66] studied friction and temperature in rolling slid-

ing contacts and were able to develop a correlation for the coefficient of friction based on

load, speed, viscosity radius and surface finish of the contacting surfaces. Seireg and Hsue

Seir81 investigated the effects of lubricant properties on temperature and wear in sliding

concentrated contacts. The researchers found that viscosity does not appear to be the sig-

nificant property of the lubricant temperature rise and wear rate. Li and Seireg [Li89]

developed a dimensionless empirical formula for calculating the coefficient of friction in

sliding-rolling steel on steel contacts operating in the thermal regime. The formula pro-

vides a unifying relationship for all the published data. Othman and Seireg [Othm89]

devised an empirical procedure for evaluating the frictional properties in Hertzian contacts

subjected to sinusoidal sliding motion. The researchers found "that the friction-velocity

function can be adequately approximated by an exponential function in the considered

case where the parameters of the function can be readily determined from a multivariate

search which minimizes the errors of the peak response at resonant frequencies," (p. 4).









The Rashid and Seireg [Rash86] study of heat partition and transient temperature dis-

tribution in layered concentrated contacts is the most relevant tribological study for this

dissertation. The researchers developed dimensionless, empirical relationships for predict-

ing maximum temperature rise in layered concentrated contacts. This paper is the basis for

the design relationships developed in this dissertation. Chapter 3 extends the methodology

presented in this paper to handle multilayered concentrated contacts.

2.4 Machine Elements

The case study in this dissertation compares traditional machine element applications

with diamond coated machine elements. General references for machine element design

include Shigley and Mitchell [Shig83] and Orthwein [Orth90]. A specific reference which

integrates thermal considerations in the design of gear pairs is found in Lin and Seireg

[Lin85]. This paper presents a computerized optimization algorithm for the design of gear

systems which selects the design parameters to provide the best balance between the

meshing elements for strength, surface durability, wear resistance and surface tempera-

ture. Formulas presented in this paper are central to the diamond coated gear system

design presented in this dissertation.

2.5 Knowledge-Based Systems

Knowledge-based systems are playing an increasingly important role in the design of

engineered products. Knowledge-based systems come in many flavors, including auto-

mated reasoning machines, neural networks, production systems, blackboard architectures

and expert systems. Artificial Intelligence is a general field that loosely contains all of

these special forms of knowledge representation and management. Winston [Wins92] and

Rich and Knight [Rich91] are references of general applicability to the field of Artificial

Intelligence. These texts contain a thorough overview of many methodologies that com-

prise the field. Expert systems are of more specific interest to this work. Gonzales and

Dankel [Gonz93] provide a good introduction to a variety of expert systems approaches

and include information on societal, ethical and legal ramifications of their use. CLIPS









[Giar91, Giar93] is an object-oriented, rule-based, expert system language developed by

NASA. This dissertation research takes advantage of many of the features in CLIPS to

model the technology transfer methodology.

2.6 Attribute Matching

Critical to the success of selecting product applications for laboratory discoveries is

the ability to effectively and efficiently match attributes common to both. In this case, a set

of rules determines how applications will match discoveries. The rules pattern match on

facts or objects and take appropriate action. Matching single attributes in the pattern of a

rule is trivial. More realistic situations involve multiple rules with multi-attribute patterns.

In this case, a many-many match algorithm coupled with a conflict resolution strategy is

employed to handle the exponential number of match permutations possible.

One efficient many-many match algorithm is Rete1 which was developed by Forgy

[Forg82]. Rete gains efficiency from three major sources (see Rich and Knight [Rich91]):

1. the temporal nature of data;

2. structural similarity in rules;

3. persistence in variable binding.

The Rete works by maintaining a network representation of rule conditions. As rules fire

(match patterns and take action), the state description changes. Some new rules will now

apply and some old ones may no longer apply. Rete only checks those rules effected by the

change in state and thus saves recomputing the entire matching network after each rule fir-

ing. The Rete algorithm is used by the CLIPS system for its matching strategy.

2.7 Software Engineering

Software Engineering is a relatively new field in engineering. It deals with the entire

life-cycle of software development: need recognition, requirements gathering, information



1. rete: an obscure synonym for net






19

modeling, algorithm development, software system design, verification and validation,

documentation and maintenance. This field has gained increasing importance as software

systems have grown increasingly complex. Sommerville [Somm92] provides a compre-

hensive reference that provides an excellent introduction to the field.

Object-Oriented Design (OOD) is a modem software design paradigm that is ideal for

modeling complex software systems. An excellent reference on the subject is the text by

Booch [Booc91]. Booch provides an overview of OOD and incorporates five complete

examples of systems developed using this technique.















CHAPTER 3
THE THERMAL PROBLEM

To predict the behavior of diamond coated components, it is necessary to model the

thermal process in sliding contacts. Many tribological applications involve Hertzian con-

tacts with high slide-to-roll ratios. The goal of the transient heat transfer analysis devel-

oped in this chapter is to determine the maximum temperature rise, thermal gradients, and

thus maximum thermal stress in layered, sliding concentrated contacts. In this chapter,

computationally efficient, simplified, predictive equations are developed for the maximum

temperature rise in diamond coated substrates.

The temperature gradient in the substrate causes a surface stress. Simplified thermal

stress equations are developed in this chapter to determine the magnitude of the surface

stress. These values are used in an empirical equation to determine a life debit due to the

thermal stress.

3.1 Transient Heat Transfer

Closed form solutions for transient heat transfer problems are available for only a few

simple geometries and boundary conditions. Superposition can be used to combine bound-

ary conditions to yield new solutions in some cases. As the number of boundary condi-

tions increases and the geometric part complexity increases, the use of superposition

becomes unwieldy. Computer modeling is the most practical way to predict the thermal

response in these cases. Finite element methods and finite difference methods are two

widely used approaches for determining temperature distribution in transient heat transfer

cases. These methods are very flexible and are commonly used to solve a variety of tran-

sient heat transfer problems.









Although the expense of running finite element and finite difference models has been

steadily decreasing as the power of computer workstations has increased, structuring these

approaches for optimization problems results in computationally expensive efforts. For

design purposes it would be useful to have a computationally efficient method available-

even if it sacrifices accuracy to gain insight into interaction among the problem's parame-

ters. For the purposes of this study, the maximum temperature rise due to transient effects

is more important than the thermal distribution. The following sections describe a closed

form approach for predicting the maximum temperature rise for a few special cases rele-

vant to diamond coated substrates.

3.1.1 A Heat Source With a Hertzian Distribution Moving Over a Layered Semi-Infinite
Solid

Rashid and Seireg [Rash86] developed dimensionless relationships for heat partition

and transient temperature distribution in layered concentrated contacts. The relationships

are based on curve fits of maximum temperature rise in transient heat transfer cases. The

data were generated from three dimensional finite difference models for a variety of geom-

etries and boundary conditions. The computer simulation was able to repeat the analytical

results of Blok [Blok37] and Jaeger [Jaeg42] for the simple case of a heat source moving

over a semi-infinite solid.

3.1.1.1 Heat source moving over a semi-infinite solid

The case of a heat source moving over a semi-infinite solid was used to check the

accuracy of the computer simulation developed by Rashid and Seireg. Blok and Jaeger

obtained the following relationship for the maximum rise in surface temperature via a

series approximation:


7Ts -T = 1.123 (3.1)
CB (iKpCU)

For the same conditions, Rashid and Seireg obtained the following dimensionless result

using a finite difference simulation:
























Figure 3-1. Moving semi-infinite solid under a stationary heat source.

(T,-T8)K p 03(pCU1- (
qt = 1.03 ,. (3.2)

Equations (3.1) and (3.2) are in general agreement with the differences in the constants
attributable to numerical approximation in the computer model. See Figure 3-1.
3.1.1.2 Heat source with a Hertzian distribution moving over layered semi-infinite solid

This case examines a single-layered substrate with a moving heat source of Hertzian
distribution. This heat distribution is equivalent to the heat generated in a contact problem
such as a semi-infinite cylinder rolling and sliding over a semi-infinite planar surface. See
Figure 3-2. Rashid and Seireg obtained the following relationships:


(Ts 0) Ko D 1 04 07 078
( -T g = 1.137( 1 (3.3)(h (
q, 1 1 ho

(T =) Ko I -0.026 (K )0/026 ( ho -180xo
q- -= 1.164- -1 -\e 0 (3.4)
q7 (Ko) J (Ie


























Figure 3-2. Layered semi-infinite solid moving under a stationary heat source with a
Hertzian distribution.

TsK K
1.12 K (3.5)
qt pCUIl


D CU (3.6)


1 0 (3.7)
e 5 K

where

D = the temperature penetration depth at the trailing edge
Ie = the required entry distance for temperature penetration across the film
Tso = the maximum rise in the solid surface temperature for a layered semi-
infinite solid
Ts = the maximum rise in the solid surface temperature for the unlayered
semi-infinite solid with the same heat input.
To = the maximum rise in the layer surface temperature.









3.1.1.3 Illustrative examples

Two examples demonstrate the effects of a conductive layer and an insulative layer on

a stainless steel substrate. Parameters for these cases are set to the following:

qt = 100 Watt/mm
U = 13700 mm/sec
1 = 1.375 mm
ho = 0 to 16 pm.


The stainless steel is AISI 304. The temperature rise in the stainless steel is denoted by Tss.

Conductive layer on stainless steel. The conductive layer applied to the AISI 304 sub-

strate is diamond. The temperature rise in the diamond is denoted by Td. Tss and Td are

plotted versus diamond layer thickness in Figure 3-3. The temperature rise in the diamond

shows a slight increase as the layer thickness increases. The temperature rise in the stain-

less steel decreases approximately 50C for a 16p.m layer thickness.

Insulative layer on stainless steel. The insulative layer applied to the AISI 304 sub-

strate is silicon nitride (Si3N4). The temperature rise in the silicon nitride is denoted by Tfy.

Tss and Ty are plotted versus silicon nitride layer thickness in Figure 3-4. The temperature

rise in the silicon nitride shows a dramatic increase as the layer thickness increases. The

temperature rise in the stainless steel decreases approximately 200C for a 16ptm layer

thickness.

3.1.2 Extension to Multi-Layered Semi-Infinite Solid

The single-layer approach is not adequate for modeling diamond coating on substrates

such as steel. In this case it is necessary to introduce a buffer layer to prevent the forma-

tion of graphite at the diamond/steel interface. The goal here is to extend the single-layer

to two layers without repeating the extensive finite difference modeling approach used by

Rashid and Seireg. An approximate method is presented here and is intended for use as a











Case: U=12700mm/sec, 1=0.250mm


I I I I I I

A


I I I I I I


0 2 4 6 8 10 12 14 16
h oj 103

Maximum temperature rise ('C) in AISI 304 substrate (Tss) and diamond
surface layer (Td) moving under a stationary heat source with a Hertzian
distribution.


I n


design approximation. The results should be verified via a finite element model. Figure 3-

5 illustrates the two-layer transient heat transfer problem.


The following dimensionless relations were developed for computational convenience:


S(T Tso) Ko (Kd
so0 q K



0s q, K)


TK (K ) (Kd
$ q K 0K


(3.8)



(3.9)



(3.10)


250


200 -


150 F-


T di, I

Tsi, 1
^u


100 h


50 -


Figure 3-3.


J











Case: U=12700mm/sec, 1=0.250mm


800




600



- 400




200




0


Figure 3-4.


0 2 4 6 8 10 12 14 16
h 0o-103

Maximum temperature rise (C) in AISI 304 substrate (Tss) and silicon
nitride surface layer (T7f) moving under a stationary heat source with a
Hertzian distribution.


00 = To +
(Kd)
q,


0= T -= 0 -9
so o\qt S SSO


(3.11)



(3.12)


where


Osso = the dimensionless maximum temperature rise in the solid surface rela-
tive to the maximum unlayered surface temperature rise
Oos = the dimensionless maximum temperature rise in the surface layer rela-
tive to the maximum unlayered surface temperature rise
es = the maximum unlayered surface temperature rise
Oso = the dimensionless maximum temperature rise in the solid surface
0o = the dimensionless maximum temperature rise in the surface layer.


















h if p, C, K T

T--*
J__



p, C, K, TB



Figure 3-5. Two-layered semi-infinite solid moving under a stationary heat source with
a Hertzian distribution.


3.1.2.1 Procedure to predict multilayer temperature rise

A numerical example illustrates the simplified procedure used to predict maximum
temperature rise in a multilayered solid. The procedure involves the following steps:

1. For a given velocity, U, and length of contact, 1, calculate the dimensionless tempera-
ture rise, 0o d and Oso if, for the case of a diamond layer on the buffer layer substrate
using equations (3.11) and (3.12). The values for the surface layer thickness, hd,
should vary from 0 to a specified upper limit (0.016 mm was used in this study).

2. The ratio between the surface layer and buffer layer maximum temperature rise is now
determined. The ratio will remain constant, although the temperature values will be
adjusted up or down based on the uncoated steel temperature.

3. For the same U and 1, calculate the dimensionless temperature rise, 00 yand Oso ss, for
the case of a buffer layer on the steel substrate using equations (3.11) and (3.12). For
simplicity, the same values for hif are used as for hd in step 1 above. See Figure 3-6.

4. The maximum dimensionless temperature rise in the buffer layer, 0iy, in the steel sub-
strate, 0ss, and in the diamond layer, 0d can now be determined. First, the steel sub-
strate temperature rise is scaled by the ratio of the buffer layer temperatures. See
equation (3.13). A scaling factor, A, is then determined in equation (3.14). This factor










Case: U=12700mm/sec, 1=0.250mm


h oi-103


Dimensionless temperature rise (prior to data manipulation) in substrate
surface and coating layer surface for diamond on AISI 304 stainless steel
with a silicon nitride interface layer.


scales the diamond layer, buffer layer and steel substrate temperatures to the uncoated
steel substrate temperature, Os ss. This factor insures that the temperature rise predic-
tions all start at the uncoated"steel substrate temperature at hd = hf = 0, and then
decrease or increase as the coating thicknesses increase. Equation (3.15) shows how A
is used to calculate the dimensionless temperature rise for the diamond and buffer lay-
ers (0d and Oif) and the steel substrate (Oss). See Figure 3-7.

5. The actual temperature rise (Tss, T,f, Td) can now be calculated by equation (3.16).
Figure 3-8 shows a typical result for qt = 100 watt/mm2.

For i = 1 to n


S(so_if
ss= Oso ss
o_if


A -
e1


(3.13)




(3.14)


0 so issi, 1
0oifssi, 1


Figure 3-6.











Case: U=12700mm/sec, 1=0.250mm


0 2 4 6 8 10 12 14 16
h i-103

Dimensionless temperature rise (following data manipulation) in substrate
surface and coating layer surface for diamond on AISI 304 stainless steel
with a silicon nitride interface layer.

For i = 1 to n


if so-if
T = AO_.



K.


(3.15)






(3.16)


3.1.2.2 Effects of varying parameters

The effects of varying the sliding velocity of the solid, U, and the width of contact, 1,

are examined in this section. Since the temperature rise in an uncoated substrate is
1
inversely proportional to the square root of the sliding velocity, AToc it is expected

that the temperature rise for the multilayered case will follow suit. Figure 3-9 illustrates

this case. The parameters, except for U, are the same as in Figure 3-8. As predicted, the


3.5


3


2.5


i, 1 2


0 ssi 1 1.5


1


0.5


0


Figure 3-7.










Case: U=12700mm/sec, 1=0.250mm


h o-103

Temperature rise (C) in substrate, interface layer and diamond layer for
diamond on AISI 304 stainless steel with a silicon nitride interface layer.


magnitude of the temperature rise for the substrate and layers dropped with the increase in

sliding velocity.

The effect of increasing 1, the width of contact, is now reviewed. Since I follows the
1
same inverse relationship as U for the unlayered substrate, AT c it is expected that the

magnitude of the temperature rise will decrease. Figure 3-10 illustrates this case. The

parameters, except for 1, are the same as in Figure 3-8. As predicted, the magnitude of the

temperature rise for the substrate and layers dropped with the increase in contact width.

3.2 Thermal Stress Considerations

Thermal stresses can have a detrimental effect on the life of a machine component. In

this section, simplified equations are developed for predicting the magnitude of the ther-

mal stresses in a multilayered coating. The thermal stress in the substrate will be added to


250



200


Tdi, 1 150

T

iTi1 100



50



0


Figure 3-8.










Case: U=50800mm/sec, 1=0.250mm


0 2 4 6 8 10 12 14 16
ho. 103


Temperature rise (C) in substrate, interface layer and diamond layer for
diamond on AISI 304 stainless steel with a silicon nitride interface layer
(velocity increase with length of contact constant).


the contact stress to determine a maximum stress value for calculating a life debit due to

thermal fatigue.

3.2.1 Thermal Stress Relationships

Nominal stress relationships for design purposes developed in the following sections

refer to the diagram in Figure 3-11. This figure defines the variables used in predicting

normal and shear thermal stresses for the case of a multilayer semi-infinite substrate mov-

ing under a stationary heat source with a Hertzian distribution.

3.2.1.1 Normal stresses

The normal thermal stress in an axial beam built in at both ends is proportional to the

increase in temperature and can be expressed as


T di, 1

T i, 1
T Su


Figure 3-9.










Case: U=12700mm/sec, 1=1.375mm


T di, 1

T

T i 1
0u


Figure 3-10.


0 2 4 6 8 10 12 14 16
h o103


Temperature rise (C) in substrate, interface layer and diamond layer for
diamond on AISI 304 stainless steel with a silicon nitride interface layer
(velocity constant with length of contact increase).


o = EaAT. (3.17)

If this equation is used for a model, then we have the following equations to describe the

normal stress in the diamond, interface coating and substrate:


ad = EdctdTd

Oif= EijaifTif (3.18)
Uss = EssssTss

where Td, Tf and Tss are the temperature differentials between each of the layers and its

substrate. Note that equation (3.18) is a nominal relationship for design approximation

only.





































Figure 3-11. Variables used in calculating thermal stress in multilayer coatings.



3.2.1.2 Shear stresses

Shear stress can be determined by dividing the shear force, Fs, by the shear area, As.

The shear force can be approximated by the difference in normal stresses between two

layers times the cross-sectional area or Fs = (a2 01) Ac. The shear stress can now be

written as


F
S
A
S
( 2- 1) Ac (3.19)
A

Now, referring to Figure 3-11, if we substitute A = hoo and As = Io, we have











r = (-0 ) (3.20)


and we can write the following equations for the shear stress between the diamond and

interface layers, and the interface layer and the substrate:


ha
d
dif (if d)

h
Tifss= (ss "if) (3.21)

3.2.1.3 Example

Figures 3-12 and 3-13 continue with the example from section 3.1.2.


Case: U=12700mm/sec, 1=0.250mm


7'108


6*108 -


5"108


01n,1 4.108 -

02n,
3n 1 3108 -


2*108 I-


108 -


Figure 3-12.


0 2 4 6 8 10 12 14 16
h on. 103

Normal stress (Pa) for various thicknesses (mm) of diamond and interface
layer. Substrate is AISI 304 stainless steel and interface layer is silicon
nitride. Note: hif = hd = ho, 01 = ad, 02 = if and (3 = ass.











Case: U=12700mm/sec, 1=0.250mm


5.106



0



-51106


-1107


-1.5*107



-2*107




Figure 3-13.


0 2 4 6 8 10 12 14 16
h 103


Shear stress (Pa) for various thicknesses (mm) of diamond and interface
layer. Substrate is AISI 304 stainless steel and interface layer is silicon
nitride. Note: hiy = hd = ho, T1 = cyf and c2 = Tifss.


3.2.2 Predicting Fatigue Life Debit Due to Thermal Stresses


Increasing temperature subsequently impacts fatigue life. In this section, the impact on

fatigue life due to increasing diamond coating thickness is developed. First a maximum

stress value is computed. Next, the stress value is compared to a stress level that represents

infinite life.

3.2.2.1 von Mises Stress

Combined rolling and sliding contacts will have stress components due to thermal

effects and contact stress. Other components, such as bending stress are ignored. Thermal

stresses are considered to be tensile, while contact stresses are compressive. The thermal

stress component will therefore be considered oanx and the contact stress component will


n, 1

-2n, 1









be considered omi,. The stresses are combined using the von Mises distortion-energy the-

ory (see Shigley and Mitchell [Shig83]) as follows:

G 2 2
o' = oAA A +B (3.22)

where

YA = the maximum normal stress, oma
(B = the minimum normal stress, onin.


3.2.2.2 Bearing life empirical model

The following empirical relationship is widely used to predict the surface durability of

ball bearings:


aon1 = ct (3.23)

where n is the number of cycles and cl is a constant.

3.2.2.3 Example

Continuing with the previous example, the following additional information is intro-

duced:

1. the level of contact stress, oc, is 1000 MPa;

2. n = 107 cycles is infinite life.

The solution is as follows:

1. max = ss

2. min =ac = -1000 MPa

3. calculate a' and plot (see Figure 3-14)

4. using the minimum value of o' (should occur at the thickest coating), determine cl

5. for each value of o', calculate a new life, n, and plot (see Figure 3-15).














Case: U=12700mm/sec, 1=0.250mm


1.5'109






1"109


Sn


5'108


0



Figure 3-14.


0 2 4 6 8 10 12 14 16
ho.103

von Mises stress (Pa) as a function of coating thickness (mm) in substrate.
Substrate is AISI 304 stainless steel and interface layer is silicon nitride.
Note: hi = hd = ho.


It is interesting to note the asymptotic shape to the life curve. It shows clearly that

reducing the surface temperature in sliding concentrated contacts drastically increases life.













Case: U=12700mm/sec, 1=0.250mm


1 107









n cycles 1 106









11105


Figure 3-15.


h on- 103

Life improvement in cycles versus thickness of diamond film in mm.
Substrate is AISI 304 stainless steel and interface layer is silicon nitride.
Note: hf = hd = ho and contact stress level is 1000 MPa.















CHAPTER 4
PROCESS ATTRIBUTES

This chapter discusses the details of the laboratory discovery used in this dissertation

as a case study. The discovery's process attributes are presented from four perspectives.

First, the properties of diamonds and diamond films are discussed. Second, the laser sur-

face preparation process under development at the University of Florida is introduced.

Third, the chemical vapor deposition process used to grow the diamond film is presented.

And finally, the cost aspects of the process are modeled.

4.1 Properties of Diamond

Diamond is an exceptional material. Most of its important properties can be labeled as

extreme. It has the highest hardness, the highest thermal conductivity, highest molar den-

sity and highest sound velocity of any material known. It also possesses the lowest com-

pressibility and bulk modulus of any known material. The thermal expansion coefficient is

also very low and ranks among the lowest of known materials. Diamond is also extremely

inert chemically-affected only by certain acids and chemicals that act as oxidizing agents

at high temperatures.

In the previous chapter, equations were developed for transient heat transfer in slide-

roll scenarios. The heat generated in these cases is due to friction between the sliding sur-

faces. The temperature rise for the developed cases is inversely proportional to the square

root of the product of the surface material's thermal conductivity, specific heat and den-
1
sity, or AT -I Therefore, for transient heat transfer problems where sliding gener-
4Kpc
ates the heat, it is crucial to have Kpc as large as possible in order to minimize thermal

effects.









Table 4-1 summarizes some important properties of diamond. The table was compiled

from Yoder [Yode93], Spear and Dismukes [Spea94] and Field [Fiel92].

Table 4-1. Properties of diamond.

Property Value Units
Hardness 1.0 x 104 kg/mm2
Strength, tensile >1.2 GPa
Strength, compressive >110 GPa
Coefficient of friction (Dynamic) 0.03 Dimensionless
Sound velocity 1.8 x 104 m/s
Density 3.52 g/cm3
Young's modulus 1.22 GPa
Poisson's ratio 0.2 Dimensionless
Thermal expansion coefficient 1.1 x 10-6 K-1
Thermal conductivity 20 W/cm-K
Thermal shock parameter 3.0 x 108 W/m
Specific heat 0.853 J/gm-K


4.2 Pre-Coating Surface Treatment

The laser-based surface treatment under development at the University of Florida has

not been previously introduced in the published literature. The process involves using a

laser to modify a metallic substrate surface. This surface modification produces a uniform

roughness which provides nucleation sites for the diamond coating growth and serves to

increase the surface area of contact between coating and substrate. The increase in contact

surface area improves the adhesion of the diamond layer and allows the interface to grade

the surface stresses, effectively reducing the chances of premature debonding.

The surface treatment shows great promise for promoting diamond film growth on

steel and other metallic surfaces. The dramatic thermal expansion mismatch between

materials such as steel and diamond makes this endeavor extremely difficult. Silicon has

been used successfully as a substrate for this process and results from this work are guid-









ing efforts on other metallic substrates. As an intermediate step, low-a metals -such as

molybdenum-are being tested with the process.

Figure 4-1 shows conceptually how the surface modification appears in cross section.

Note how the diamond "seeds" sit inside the roughness "valleys." The seeds act as nucle-


substrate surface following laser modification









substrate surface seeded with 20[pm diamond particles









substrate surface after CVD diamond film growth









not to scale
Figure 4-1. Cross sectional view of substrate surface following laser modification,
seeding and film growth.




1. low thermal expansion coefficient









ation sites for the formation of the diamond film. This feature allows film growth to occur

at a lower surface temperature than would otherwise be possible.

An electron microscope photograph illustrating a seeded substrate surface after a short

growth duration is given in Figure 4-2. The surface roughness produced by the surface


Figure 4-2.


Seeded silicon substrate surface following a short duration (10 minute)
diamond growth period.


preparation process is evident from the photo, but the diamond seeds make the surface

look rougher than it is. The final surface topology after a complete diamond growth cycle

is shown in Figure 4-3. The uniformity of the surface roughness is much more evident in

this picture.

4.3 Chemical Vapor Deposition of Diamond

Diamond synthesis techniques have been available since the late 1950s (see Busch and

Dismukes [Busc94]). The commercialization of synthetic high-pressure, high-temperature

(HPHT) diamond grit occurred in 1959. This grit has had tremendous use in industrial pol-

ishing, cutting and grinding applications. The HPHT synthesis method essentially mimics




























"a "
;0;0 oIIQ


Figure 4-3. Final silicon substrate's surface topology following the complete diamond
film growth cycle.

nature's way of producing diamond--only at much poorer quality. HPHT methods are

basically only capable of producing grit.

The chemical vapor deposition (CVD) process, a low-pressure synthesis method, was

successfully used to precipitate diamond-on-diamond seed crystals using carbon monox-

ide gas as a source of carbon, in 1952. This method actually predates the HPHT process by

several years, but presented more challenges for commercialization.

CVD diamond growth methods use simpler apparatus less subject to mechanical wear

and promises the production of physical forms of diamond other than powder (HPHT)

(see Moustakas [Mous94]). One of the early drawbacks to CVD methods was the forma-

tion of graphite during diamond nucleation. Many variations have been tried in cleaning

the graphite structures during diamond growth. Introducing hydrogen to the environment

has been effective in "scrubbing" the diamond structures clean from graphite.

Researchers found that heating the substrate surface using a plasma source increased

the diamond growth rate. This method helps decompose the methane gas into carbon









(methane has a high activation energy that caused slow growth rates). An illustration of a
microwave-plasma CVD apparatus is shown in Figure 4-4.

H2 + CH4
Quartz Tube



Substrate Tuner




Microwave
2.45 GHz





Magnetron


Figure 4-4.


Schematic illustration of microwave-plasma CVD apparatus used for the
growth of diamond films ([Mous94] p. 164).


Moustakas reports [Mous94] that microwave-assisted CVD methods are the most
prevalent for diamond film growth. The process avoids contamination of film during

growth and produces a higher plasma density over RF (radio frequency) methods. This
results in higher concentrations of atomic hydrogen and hydrocarbon radicals necessary









for film growth. Since this method concentrates plasma at the center of deposition cham-

ber, it prevents deposition of carbon on the quartz tube walls. Typical parameters for depo-

sition using the microwave-plasma CVD, as reported by Moustakas ([Mous94] p. 163),

are shown in Table 4-2.

Table 4-2. Deposition parameter space used in the growth of diamond films by
microwave-plasma-assisted CVD.

Total Pressure Microwave Substrate
Gas Mixture (torr) Power (W) Temperature (C)
CH4 (0.5-2%)/H2 5-100 100-700 700-1000


HPHT diamond is limited in application to planar surfaces. In this respect, HPHT dia-

mond is no better than the natural diamond grit that it replaces. CVD diamond, however,

ushers diamond applications to an exciting new level. "CVD diamond offers the potential

to deposit large-area, conformable coatings with properties akin to that of natural diamond

([Busc94] p.592)." For the first time, diamond can be used as an engineered material-

synthesized to meet specific topological and performance characteristics.

4.4 Cost Modeling

Cost modeling was used in this study to predict costs for applying the diamond coat-

ing. Busch [Busc94] used a cost modeling technique to assess factors influencing manu-

facturing costs in plastics fabrication processes. Busch and Dismukes applied this method

to determine present and future costs in diamond coating [Busc94]. The cost modeling

presented here is adapted from Busch and Dismukes' comparative assessment of CVD

diamond manufacturing technology and economics. In their study, several CVD

approaches are compared from near-term (3 to 5 years) and long-term (5 to 10 years) per-

spectives. Mature HPHT technology is used for a baseline. The technique used by Busch

and Dismukes in their economic analysis is known as technical cost modeling (TCM). The

sequential stages in TCM are illustrated in Figure 4-5.










Emerging Technologies


Step 1


Step 2


Step 3


Step 4


Step 5


Step 6


Figure 4-5.


Sequential stages in technical cost modeling (TCM) ([Busc94] p. 586).


4.4.1 Near-Term Inputs

Significant near-term input parameters for the TCM are summarized in Table 4-3. The

deposition parameters are expected or most probable values, as defined by Busch and Dis-
mukes [Busc94]. The assumptions apply for deposition of 81 cm2 diamond wafers, each
250 pm thick.

4.4.2 Long-Term Inputs

Table 4-4 assumes that significant technological improvements-feasible within 5 to
10 years-have occurred. The deposition capability includes 730 cm2 diamond wafers
measuring 250 pm thick.


Existing Technologies









Table 4-3. Near-term inputs ([Busc94] p. 593).


Input' Microwave DC Arc Jet
Hydrogen (%) 88.7 99.0
Carbon (%) 10.0 1.0
Oxygen (%) 1.3 NA
Hydrogen recycle rate (%) 0.0 0.0
Substrate area (cm2) 81 81
Substrate cost $0.01 $0.01
Coating thickness (im) 250 250
Deposition rate (pm) 4 25
Direct laborers/station 0.2 0.2
Load/unload laborers/station 1 1
Load/unload time (min) 60 60
Total gas flow rate (sccm) 484 30250
Carbon capture factor (%) 8.0 8.0
Mass deposition rate (g/h) 0.11 0.71
Machine cost ($/station) $350000 $350000
Machine power (kw) 15 90
Cooling water (GPM) 3 10
Building space (ft2/station) 1000 1000
a. Dollar values are from 1992

4.4.3 Other Cost Factors

Other cost factors necessary to complete the analysis are included in Table 4-5. These

figures are applicable to both the short-term and long-term inputs to the TCM process.

Busch and Dismukes obtained the information for Table 4-6 on specific input costs for

process gases by consulting directly with gas vendors and reviewing trade literature.

4.4.4 Near-Term Outputs

Table 4-7 summarizes near-term outputs for microwave and DC arc jet CVD deposi-

tion techniques. The wafer size is 81 cm2 with a 250 p.m thick deposition. The TCM









Table 4-4. Long-term inputs ([Busc94] p. 594).


Inputa Microwave DC Arc Jet
Hydrogen (%) 88.7 98.0
Carbon (%) 10.0 2.0
Oxygen (%) 1.3 NA
Hydrogen recycle rate (%) 0.0 0.0
Substrate area (cm2) 730 730
Substrate cost $0.01 $0.01
Coating thickness (gpm) 250 250
Deposition rate (pm) 15 60
Direct laborers/station 0.1 0.1
Load/unload laborers/station 1 1
Load/unload time (min) 120 120
Total gas flow rate (sccm) 6544 130900
Carbon capture factor (%) 20.0 20.0
Mass deposition rate (g/h) 3.85 15.42
Machine cost ($/station) $550000 $700000
Machine power (kw) 75 350
Cooling water (GPM) 25 60
Building space (ft2/station) 2000 2000
a. Dollar values are from 1992

reports a total-per-carat cost of $61.61 for the microwave deposition and $22.33 for the

DC arc jet deposition. Since there are 0.2 grams per carat, these figures translate to $308

and $112, respectively, on a per gram basis. A cost estimate of $200 per-gram-deposited,

the approximate average of the microwave and DC arc jet costs, was used throughout this

study.

4.4.5 Long-Term Outputs

Table 4-8 summarizes long-term outputs for microwave and DC arc jet CVD deposi-

tion techniques. The wafer size is 730 cm2 with a 250 gpm thick deposition. The TCM

reports a total-per-carat cost of $3.75 ($18.75/gm) for the microwave deposition and $2.42

($12.10/gm) for the DC arc jet deposition. A cost estimate of $15 per-gram-deposited, the









Table 4-5. External cost factors ([Busc94] p. 595).


Direct wages $13.33/h
Indirect salary $50000/year
Indirect: direct labor ratio 0.33
Benefits on wage and salary 35.0%
Working days per year 360
Working hours per day 24/day
Capital recovery rate 10%
Capital recovery period 5 years
Building recovery life 20 years
Working capital period 3 months
Price of electricity $0.100/kWh
Price of natural gas $6.50/MBTU
Price of building space $100/ft2
Price of cooling water $0.03/100 gal
Auxiliary equipment cost 15.0%
Equipment installation cost 35.0%
Maintenance cost 8.0%
a. Dollar values are from 1992
b. U.S. figure; cost does not include cost of land


Table 4-6. Process gas costs in the near and long term (1992 $/m3) ([Busc94] p. 595).


Gas Near Term Long Term
Hs $13.07 $5.39
CH4 $19.50 $6.57
C2Hs $7.06 $2.40
02 $10.76 $1.74


approximate average of the microwave and DC arc jet costs, would therefore be a reason-

able long-term estimate for deposition costs-a reduction in costs by over a factor of 13.









Table 4-7. Near-term outputs ([Busc94] p. 595).


Output Microwave DC Arc Jet
Variable costs
Material cost $25 $238
Labor cost $270 $60
Energy cost $97 $92
Fixed costs
Capital equipment cost $945 $210
Building cost $45 $10
Maintenance cost $450 $100
Overhead labor cost $30 $7
Cost of capital $334 $79
Total cost ($/run) $2196 $796
Total cost ($/cm2) $27.11 $9.83
Total cost ($/carat) $61.61 $22.33

Table 4-8. Long-term outputs ([Busc94] p. 596).

Output Microwave DC Arc Jet
Variable costs
Material cost $39 $177
Labor cost $73 $48
Energy cost $133 $150
Fixed costs
Capital equipment cost $495 $210
Building cost $30 $10
Maintenance cost $246 $100
Overhead labor cost $5 $2
Cost of capital $184 $79
Total cost ($/run) $1204 $776
Total cost ($/cm2) $1.65 $1.06
Total cost ($/carat) $3.75 $2.42















CHAPTER 5
POTENTIAL APPLICATIONS

This chapter explores potential applications for the diamond coating process. Each

application area discussion includes nomenclature, design objectives and design examples

illustrating limitations with current materials. In each of these applications it will be dem-

onstrated that thermal properties of the contacting materials are extremely important. Dia-

mond's extremely high thermal conductivity will prove to be an excellent match for these

applications.

5.1 Application to Gears

Successful gear system design requires balancing design parameters that affect tooth

strength and surface durability. Surface failure results from interaction between wear

mechanisms, contact fatigue and thermal effects. Many gears operate under conditions of

boundary and mixed lubrication. In such cases tooth surface temperature rise is very

important.

5.1.1 Gear Nomenclature

The following nomenclature is used in gear systems-corresponding equation num-

bers are in parenthesis:

ATp = temperature rise on the pinion surface (5.12)
ATg = temperature rise on the gear surface
4 = heat flux per unit contact length (5.1)
f = coefficient of friction
WN = normal force on the surface per unit length
Tq = input torque per inch of facewidth
F = facewidth
Vr = rolling velocity (5.2)









Vs = sliding velocity (5.3)
Vp = instantaneous tangential velocity at pinion contact point (5.4)
Vg = instantaneous tangential velocity at gear contact point (5.5)
(Op = angular velocity of the pinion
MG = gear ratio
) = pressure angle
x = position of contact along the line of action measured from the pinion
base circle
xs = position x at the starting point of contact (5.11)
xe = position x at the end point of contact
a = width of the contact band (5.6)
Re = effective radius of the contacting cylinders (5.7)
K = thermal conductivity of the gear material
p = density of the gear material
c = specific heat of the gear material
Vp = heat partition coefficient for the pinion (assumed to be 0.5 for equal
heat partition between the pinion and the gear)
Ayg = heat partition coefficient for the gear (assumed to be 0.5 for equal heat
partition between the pinion and the gear)
Ee = effective modulus of elasticity for the contacting materials (5.8)
dp = pitch diameter of the pinion (5.9)
Cd = center distance
Wt = tangential load per inch of facewidth
Np = number of teeth in the pinion.

5.1.2 Governing Equations for Gear Pairs

The following equations define relationships between the gear system parameters:


heat flux:


A7,xKpc
4 = (5.1)

2yp
"p


velocity terms:









V +V
Vr 2= g (5.2)

Vs= Vp- V (5.3)

V = coPCdxsinf (5.4)

Cd
V = o (1 -x) sino (5.5)

contact zone width:



a = 2.15 R (5.6)
E e

effective radius:


Re= Cd (1 -x) xsino (5.7)

effective modulus of elasticity:

1
E = (5.8)
-+-
E1 E2

pinion pitch diameter:


C
d = 2 (5.9)
1+Ma
5.2 Minimizing Contact Temperature in Gear System Design

Lin and Seireg [Lin85] proposed an optimum gear system design strategy that mini-

mizes contact temperature while maximizing the load-carrying capacity of the system.

The researchers used the following dimensionless relation for the temperature rise on the

pinion surface:









AT,
AT= 1/4
( 3Ee2 )1/4

p 2K C2
1

= 0.77 (1 +MG)X-- 1 sin 3 (5.10)
G X)x Cos (5.10)
Similar relations were developed for the dimensionless temperature rise in the gear,

AT The relationships for AT* and ATg are functions of the pressure angle (, gear ratio

MG and the contact position x only. Cheng, et al. [Pati77,Wang81], found the minimum

values of the surface temperature occur at the pitch point and the maximum values occur

at the lowest point of contact on the dedendum. Lin and Seireg confirmed this result and

further found that for "standard addendum, the maximum temperature rise calculated by

the simplified formula for mixed lubrication occurs at the pinion addendum (starting point

of contact) for gear ratios greater than one and at the gear addendum (end point of contact)

for gear ratio less than one," ([Lin85], p. 554). For a gear set with standard addendum, the

start point can be expressed as:



MG G COS*
x = p- (5.11)
s ( +MG) sin
The temperature rise at the starting point of contact, AT and AT is illustrated in
Figure 5-1. The graph indicates that for small number of teeth in the pinion, Np, tempera-

ture rise on the surface can be quite high. The intermittent meshing of the teeth causes a

cycle of high transient surface temperature rise followed by rapid cooling. Dooner and

Seireg have shown this thermal shock cycle causes thermal stresses high enough to be

become significant relative to contact stresses for the case of mixed lubrication. Figures 5-
2 and 5-3 compare the nominal thermal and contact stress for standard gear teeth.



















AT* 1.0 2 4
-------AT;






14 18 22 26 30 34 38 42
Np

Figure 5-1. Temperature rise at the starting point of contact (( = 200) ([Doon95]
p. 419).

5.3 Design of a Single Reduction Gear Set

Provided that center distance, Cd, is not pre-selected, a typical design objective would

be that the gear set provides the required speed ratio and carries the given design load with

the minimum volume of gear materials. Lin and Seireg state the problem as follows:

given parameter = MG, (op, Tq, 0, with standard addendum
decision parameters = Np, Cd
constraints = ATp < ATg = 100 F
Obmax ba = 50,000 psi
Omax < Ha = 150,000 psi
xs >0
design region is = 18 < Np < 40
bounded by 55 Cda 45
2
7 1 + MG 2
objective function = minimize V -= 1C MG 2
4 (1 +MG)2


The equations governing the constraints are defined as follows:















400,000








200,000


1000 2000 3000
Load (lIbin.)


Nominal thermal and contact stress for
p. 420).


standard gear teeth ([Doon95]


AT = 0.77f


Ee 2WtCd


(1 +MG)X- 1 sin 4 4

MG (1 -s)cos


2. maximum bending stress:

WN
Yb = 13.5 (1+ MG)
max maximum contact stress:

3. maximum contact stress:


W1Ee
a1, = 0.417 We-
Recos4


Figure 5-2.


(5.12)


(5.13)


(5.14)


































1000 2000 3000
Load (Itin.)


Nominal thermal and
p. 421).


contact stress for standard gear teeth ([Doon95]


4. tangential load per unit length:


Tq (1 +MG)
Cd


400,000








200,000


Figure 5-3.


(5.15)









5.3.1 Example Single Reduction Gear Set

The following numerical example, using a steel gear, is taken from Lin and Seireg:

MG =2
Tq = 30,000 in. lb/in.
(Op = 1750 rpm
f = 0.045
Vp = 0.5
< = 200
K = 26 Btu/hr-ft.-F
P = 490 Ibm/ft.3
c = 0.10 Btu/lbm-F
Ee = 15,000,000 psi.


The design constraints are the same as defined in section 5.3.

An optimization strategy using the steepest gradient and golden-section methods was

used to obtain the following solution:

Np =40
Cd = 36.65 in.


The constraint values are as follows:

ATp = 200.0 F
Obmax = 11.0 kpsi
amax = 58.1 kpsi
x = 0.202.


5.3.2 Comparison of Carbon Steel and Stainless Steel Gears

Many corrosive environments preclude the use of high-strength carbon steel gears. In

these environments it would be beneficial to use stainless steel gears. Unfortunately, stain-

less steels have relatively low thermal conductivity compared to carbon steels. A typical

value for stainless steel thermal conductivity is 16.6 watt whereas a typical carbon steel
m-OC









watt
used in gear applications has a value of 46.7 m- Temperature rise on the gear contact
m-C
surfaces is dependent on the magnitude of Kp c as shown in Equation (5.12).

The difference in temperature rise on the pinion surface is compared for carbon and

stainless steel gears under the same operating conditions for Np ranging from 18 to 40

teeth in Figure 5-4. The allowable temperature rise for this example is 200 OF. Note that

while the heat flux generated in the contact zone for each gear are very close, the tempera-

ture rise values differ by approximately 50 oF for the best case (Np = 40). This temperature

difference limits the choices of the designer when selecting components for this applica-

tion. To maintain the same gear ratio, the center distance between the pinion and the gear

must be adjusted. If center distance is a design constraint, then it may not be possible to

use a stainless steel gear pair. The most important constraint, however, on the use of stain-

less steel in sliding contacts is the tendency for galling. Consequently, the use of an appro-

priate coating layer is mandatory in this case.














1200


1000


800
AT csi

AT ssi

400


200 -


0
15







81.108
8- 108 -



qcs 6"108 -

q SS,
S4108


2*10 -


0 -


Figure 5-4.


20 25 30 35


15 20 25 30 35


Comparison of carbon steel and stainless steel gears under the same operat-
ing conditions.
a) Temperature rise (F) in the pinion; b) Heat flux (watt/m2).


carbon steel properties: Kpc = 1570 lbf.2/F-in.2-sec., E=30xl06 psi

stainless steel properties: Kpc = 680 lbf.2/oF-in.2-sec., E = 30x106 psi

Cd = 10", Wt = 2500 lb./n., N = 1800 rpm, MG = 5, and <) = 200















CHAPTER 6
THE MATCHING PROCESS

6.1 The Matching Process Role in the Technology Utilization Methodology

This study determined that finding suitable applications for new laboratory discoveries

is a two-level process. First, qualitative requirements of potential applications are com-

pared with qualitative capabilities of the new discovery. Once a qualitative attribute match

is established, a quantitative design analysis with cost/benefit evaluation can be per-

formed. Multiple application matches can be ranked based on the results of cost/benefit

analyses. The ranking suggests where the best return on investment can be realized for fur-

ther development of the new technology. The framework for the process is illustrated in

Figure 6-1.

In this chapter, a hierarchical structure is proposed for capturing qualitative and quan-

titative attributes for technologies. These technologies represent either new discoveries or

potential applications. Next, a knowledge base for managing the attribute structure is pre-

sented. The knowledge base, developed in the CLIPS expert system shell, includes capa-

bilities for pattern matching on attributes within a rule-based framework.

A detailed procedure for building the potential applications and emerging technologies

knowledge bases is presented in Chapter 7. A sample run through the qualitative matching

procedure is also included.

6.2 The Structure of the Developed Matching Process

The matching process discussed in this chapter is intended to present the basic

approach for qualitative matching. The implementation, however, is heavily biased to the

demonstrated process. Consequently the matching rules are limited in scope since the









Qualitative Matching
.0

*l' rifi_
o o
'- \z


Quantitative
Analyses
& Ranking


0









objective is to demonstrate the approach rather than the rigorous development of a general

matching scheme.

6.2.1 Creation of a Structured Attribute Representation

Central to the development of the matching strategy was the creation of a structured

representation for the attributes of emerging technologies-discoveries--and their poten-

tial applications. It was necessary to capture a variety of information in this attribute struc-

ture, including the following:

material attributes

manufacturing attributes

equations governing the use of the technology

qualitative requirements for the potential applications.

A common structure was developed to represent a technology---discovery or potential

application. A technology's attributes were decomposed progressively from general cate-

gories such as material attributes down to atomic property values such as thermal expan-

sion coefficient. An atomic property cannot be further broken down into finer

specializations.

Qualitative properties defined for discoveries are considered to be capabilities,

whereas qualitative properties defined for potential applications are considered to be needs

or requirements. This distinction between discovery capabilities and application needs

makes it possible to develop a straightforward strategy for matching applications to dis-

coveries.

6.2.2 Matching Atomic Values

The developed matching strategy compares atomic values of application attributes

with atomic values of discovery attributes. Each atomic match of the discovery-applica-

tion pair enters a match table where the fit between the capability and the need is evaluated









qualitatively. The match is considered strong or weak depending on how closely the crite-

ria correspond.

6.2.3 Rating Discovery-Application Matches

Once all the attributes are matched, it is possible to judge how well the individual

applications fit the discovery. A simplified qualitative rating scheme based on the percent-

age of strong matches relative to total matches for a given discovery-application pair was

implemented. This method worked acceptably for the considered case study, but requires

modification before it can be generalized. It is feasible to develop an attribute weighting

scheme such that capability-need matches can be emphasized or de-emphasized for differ-

ent discovery-application pairs. This capability was not implemented in the prototype

matching system.

6.2.4 Prototype System Development

The developed matching strategy was implemented in a knowledge base system to

demonstrate the technique. The following steps were involved in developing the prototype

system:

1. a structure for the discovery and application attributes was developed and encoded;

2. a rule-driven mechanism based on atomic discovery-application attribute matching
was developed and encoded;

3. a match table structure was developed and implemented to store and evaluate discov-
ery-application attribute matches;

4. an application framework was developed and implemented to manage the matching
process;

5. the knowledge base was populated with discovery and potential application attributes.

Steps 1 and 2 are discussed in this chapter and the remaining steps are developed in the

next chapter.









6.2.5 Road Map for the Remaining Matching Process Discussion

Development and implementation details regarding the technology structure are

detailed in this chapter. The principles behind the rule-driven matching engine are

reviewed and matching engine inference strategies are compared. An introduction to the

knowledge base system used to develop the prototype matching application completes the

chapter. Chapter 7 details the construction of the prototype qualitative matching applica-

tion, the process of populating the knowledge base and discusses the application system's

execution.

6.3 Elements of the Matching Process

6.3.1 Technology Attribute Hierarchy Development

The American Heritage Dictionary of the English Language defines technology as the

following:

1. a. The application of science, especially to industrial or commercial objectives. b. The
scientific method and material used to achieve a commercial or industrial objective.

2. The body of knowledge available to a civilization that is of use in fashioning imple-
ments, practicing manual arts and skills, and extracting or collecting materials
(Anthropology).

Definition 1 applies for the purposes of this study. The challenge here is to decompose a

technology into attributes that can be compared with other technologies to find attribute

matches. This study deals with two types of technology which are broadly labeled as DIS-

COVERY and APPLICATION. A DISCOVERY technology type represents a new labora-

tory discovery. An APPLICATION technology type represents a potential application for a

DISCOVERY. It is useful to have a universal technology attribute hierarchy that can cap-

ture the attributes for both DISCOVERY and APPLICATION technology types.

6.3.1.1 Technology attribute categories

The proposed methodology groups technology attributes into five broad categories.

This study looks at technology transfer from the perspective of turning new discoveries









into products. Therefore, the technology attribute groups proposed have a physical product

objective. This physical product objective is further limited to engineered products. This

class of products is aimed at meeting societal and customer needs and thus precludes the

application to pure aesthetic products. With this qualification or product focus, the tech-

nology attribute categories are material attributes, manufacturing attributes, governing

equations, qualitative requirements and other attributes. These attribute categories are con-

sidered to be second-level attributes. Although the tabulated attributes are specific to the

considered problem, the approach is applicable to any technology under study.

Material attributes. Material attributes are composed of physical properties, economic

properties and other properties. Physical properties deal with how materials in a technol-

ogy interact with the physical world. These interactions may be thermal, mechanical or

environmental. Subclasses within the physical properties attribute class are defined in

Table 6-1. Economic properties may include such areas as cost modeling for the materials

involved. These models refer to raw material costs rather than processing costs. Process-

ing costs more logically fall under the manufacturing attributes branch of the technology

tree. The other properties branch includes miscellaneous material properties not associated

with the previous two categories.

Manufacturing attributes. Manufacturing attributes include any production-related

properties and property classes relevant to the technology. The process subclass is the only

currently defined specialization of the manufacturing attributes class. The process class

includes information such as the economic cost model, parameters of the process, process

methods and process limitations. Subclasses within the attribute class for Process proper-

ties are defined in Table 6-2.

Governing equations. The governing equations subclass captures the empirical rela-

tionships, design parameters and constants used in describing the behavior of the technol-

ogy. Empirical relationships might include predictive equations that relate qualities such

as thermal stress to various attributes or parameters used in analyzing the technology's









Table 6-1. Subclasses defined within the Physical properties class for a sample
knowledge base.

Class
Subclass Description
Thermal class representing how a material's physical proper-
ties interact thermally with the environment
conductivity, K heat transfer proportionality constant
specific heat, C coefficient representing the energy storage capacity
per unit density
expansion coefficient, a coefficient representing the increase in linear material
growth per change in material temperature
diffusivity, adiff measure of heat transport relative to energy storage
shock resistance qualitative value of the thermal shock resistance
usage temperature, Tmx maximum useful material operating temperature
range
Mechanical class representing how a material's physical proper-
ties interact mechanically with the environment
Young's modulus, E proportionality constant relating stress to strain; a
measure of the stiffness or rigidity of a material
shear modulus, G proportionality constant relating stress to strain; a
measure of the stiffness or rigidity of a material
Poisson's ratio, v ratio of lateral to (-)axial strain
density, p mass per unit volume
coefficient of friction kinetic coefficient of friction for the material sliding
(wet), tlub on itself with a lubricant film
coefficient of friction kinetic coefficient of friction for the material sliding
(dry), jdy on itself without a lubricant film
crack resistance qualitative value of the material's crack resistance
roughness, Ra peak-to-valley amplitude
hardness, H Brinell hardness
toughness, Kc fracture toughness or critical stress-intensity factor
Environmental class representing how a material's physical proper-
ties interface with the environment
lub. film thkns., ho range of lubricating thicknesses
lub. compatibility a list of compatible lubricants
oxidation resistance qualitative measure of the material's oxidation resis-
tance









Table 6-2. Subclasses defined within the Process properties class for a sample
knowledge base.

Class
Subclass Description
Economics class representing economics of the manufacturing
process attributes
material cost of materials per unit weight of finished product
labor cost of labor per unit weight of finished product
power cost of power per unit weight of finished product
overhead cost of overhead per unit weight of finished product
capital equipment cost of capital equipment per unit weight of finished
product
Parameters class representing manufacturing process parameters
pressure process operational pressure
atmosphere process atmosphere
surface temperature surface temperature of substrate in process
Methods class representing the manufacturing process methods
preparation preparation/setup methods for process application
application application methods used for process
finishing finishing methods used for process
Limits class representing manufacturing process limits
thickness range of thicknesses that can be applied with the pro-
cess
spatial requirements required space to setup and operate processing equip-
ment
deposition rate thickness/area/volume of material that can be applied
per unit time

design applications. Design parameters specify which variables used in an empirical rela-

tionship that are typically varied during the design process to achieve a suitable result.

Oualitative requirements. The qualitative requirements category only applies to

APPLICATION technology types. This class describes qualitative information relative to

a technology. These qualitative requirements may be further decomposed into specific

attributes as needed for matching. For example, a gear application may require a high

slide-to-roll contact ratio. This need may be decomposed into a series of specific require-









ments, such as a material with high thermal conductivity, low coefficient of friction and

high hardness.

Other attributes. The other attributes category is defined for completeness. Some tech-

nology attributes may not be appropriate for the other second-level attributes. The other

attributes category is a convenient location for such.

6.3.1.2 Skeleton for a technology hierarchy

A partial skeleton for a technology hierarchy, or technology tree, is illustrated in Fig-

ure 6-2. Note that not all leaves of the tree are populated. The Technology class is the root

node of the tree and is said to be at level one. As each branch is explored level by level, the

attributes become more specific. Thus the surface temperature leaf at level 5 is more spe-

cific than the process attribute at level 3.

6.3.2 Object-Oriented Implementation

The tree structure developed in the previous section suggests an object-oriented pro-

gramming (OOP) language is appropriate for modeling and implementation. There are

five primary characteristics that an OOP system must possess: abstraction, encapsulation,

inheritance, polymorphism and dynamic binding [Booc91,Somm92,Giar91]. Booch

defines OOP as follows:

Object-oriented programming is a method of implementation in which pro-
grams are organized as cooperative collections of objects, each of which repre-
sents an instance of some class, and whose classes are all members of a
hierarchy of classes united via inheritance relationships [Booc91] (p. 36).

6.3.2.1 OOP system characteristics

Abstraction. An abstraction is a high-level, intuitive means of representing a concept.

We as humans use abstraction to grasp complex concepts. A well-developed abstraction








70




































4.





Cd


0
0


















,l
O
4.)


0
0


e3
0

0


u

bo
0



I-
0
4.
I-
4-



COi

4.)









emphasizes details that are important to the user and suppresses those details that divert

attention or confuse the subject. Booch defines abstraction as follows:

An abstraction denotes the essential characteristics of an object that distinguish
it from all other kinds of objects and thus provide crisply defined conceptual
boundaries, relative to the perspective of the viewer [Booc91] (p. 39).

Encapsulation. Encapsulation, also known as information hiding, is the process of hid-

ing the implementation details of an object using a well-defined interface. Whereas

abstraction concentrates on how an object is viewed from the outside, encapsulation con-

centrates on the inner workings of the object. Booch summarizes encapsulation as follows:

Encapsulation is the process of hiding all of the details of an object that do not
contribute to its essential characteristics [Booc91] (p. 46).

The only access to an encapsulated object is through its message handling system. This

means that it is only possible to "get" data from an object if that object has a message-han-

dler (also known as a method) called "get." This access to objects only through their pre-

defined message-handlers prevents unintended corruption of data and minimizes the

possibility for unwanted side effects.

Inheritance. Inheritance is a mechanism to describe a subclass in terms of one or more

superclasses. Single inheritance systems only permit a subclass (child class) to inherit

characteristics from one parent class. Multiple inheritance systems permit a child class to

inherit characteristics from one or more parent classes. Classes are arranged hierarchically

such that the most general classes are at the top and the most specific at the bottom. Figure

6-2 illustrates a simple single inheritance hierarchy. Figure 6-3 represents a simple multi-

ple inheritance hierarchy for quadrilaterals.

Inheritance also speeds up the development of software systems because it enables

code to be easily reused. Many commercial and public domain object libraries are avail-

able to software developers.










Quadrilateral




Kite Trapezoid





Isosceles
Parallelogram Trapezoid




Rhombus Rectangle


Square




Figure 6-3. The quadrilateral hierarchy illustrating multiple inheritance ([Giar93]
p. 101).

Polymorphism. Polymorphism is the capability for multiple objects to respond to the

same message in different ways. For example, the "+" operator may be defined such that

for a "+" message sent to two number objects, the sum of the numerical values is returned

and that for a "+" message sent to two text string objects, the result returned is the concat-

enation of the strings into a single string. The process of assigning multiple behaviors to

an operator, such as +, -, /, *, >, etc., is referred to as operator overloading. If a function

name is used instead of an operator, then the function is said to be overloaded.

Dynamic binding. Dynamic binding defers selection of a reaction to a message until

run-time. This means that the argument types passed to a function or method do not have

to be known at compile time. Instead, the system will determine the appropriate behavior

to a message when it is received. Dynamic binding works in conjunction with the poly-









morphism concept In the case of the overloaded "+" function described previously, it

would not be necessary to define the argument types passed to "+" at compile-time.

6.3.2.2 OOP advantages

Using an OOP language for developing large software systems has well-documented

advantages [Booc91, Somm92]. A key advantage for the work presented in this study is

the ability to manage a complex system through the use of data abstraction, also known as

information hiding. The inheritance feature of OOP languages is invaluable for capturing

the structure of the technology tree attribute relationships. Another important overall fea-

ture of OOP languages is the simplification of implementing object-oriented designs

(OOD). OOD is a modem software engineering paradigm best used for developing large,

complex software applications.

6.3.2.3 OOP terminology

The following terms are specific to OOP and are used throughout this chapter:

1. class: a template for describing the common attributes (slots) and behaviors of a group
of objects called instances of a class.

2. object: an instance of a class.

3. instance: an object that can only be manipulated with messages.

4. message: the mechanism to manipulate an object.

The technology tree can be modeled as a series of parent-child relationships, where

the parent represents the superclass and the child represents a subclass of a parent. Refer-

ring back to Figure 6-2, Material Attributes is the parent class of Physical Properties, Eco-

nomic Properties and Other Properties.

6.3.2.4 Class structure of the technology tree

Class definition. A class is defined by its data attributes, or slots, and its message-han-

dlers. Objects pass control from one to the next via messages. Message-handlers are an

object's interface to other objects. This structure prevents objects without a specific func-









tion from corrupting the data from another. An example of a message sent to an object

might be a print request. A typical class definition is shown in Figure 6-4.


class name__________
""-----Technology
slots:
slot names --p. technology-name slot values
Stechnology-type
Sphylum I

message-handlers:
-- messages that can be
get sent to the object
put
print



Figure 6-4. Structure of a class definition.


Inheritance mechanism. A child class is defined in terms of one or more parent classes,

which may have in turn inherited properties from their parent classes, etc. Figure 6-5 dem-

onstrates a typical inheritance mechanism. In this case the Material-Attribute class inherits

the slots "technology-name", "technology-type" and "phylum" from the Technology

superclass and the slot "classification" from the Level-1-Attribute class. Note that the

classes are shown connected with "is-a" links. This is read as the Material-Attribute is-a

Technology and the Material-Attribute is-a Level-1-Attribute.

6.4 The Matching Engine

The new technology-application matching engine was developed using an expert sys-

tem. There are three components to the expert system (also referred to as the knowledge-

based system) used in this study. First is the knowledge base, which is composed of rules

(heuristics). Next is the data, which is comprised of facts and instances. And lastly, the

inference engine, which is provided by the expert system shell to apply the knowledge to

the data.














parent class classification paremn class technology-type
message-handlers: phylum
get1put print message-handlers:
get put print
is-a inheritance links



Material-Attribute child class

inherited slots technology-name l
I technology-type
phylum default slot value
classification | MATERIAL "r
message-handlers:
get put print



Figure 6-5. Inheritance mechanism for a child class with multiple parent classes.


This section develops details describing the qualitative matching process. First, an

overview of the process is presented. Next, the expert system shell used to implement the

matching engine is described. And lastly, sample qualitative rules used in the process are

given.

6.4.1 Matching Process Overview

Information regarding new technologies and potential applications is encoded within

the knowledge-base management system in the form of explicit facts, instances of the

technology class hierarchy and rules which manipulate the facts and instances. The match-

ing process searches for fact and instance patterns that match rule conditions. If a rule

matches facts and/or instances, then the rule takes some consequent action that may

include retracting certain facts, asserting new ones, manipulating instances or launching a

procedural algorithm.









When many rules interact with many facts and instances, considerable computation

time may be consumed. A strategy must be employed to determine which rules fire first

and also minimize computational effort each time the state of the knowledge base

changes. One way to decrease the computational effort is to implement an efficient pattern

matching network. The network used in this study is the Rete, which was developed by

Forgy [Forg82] for the OPS5 production system.

6.4.1.1 Facts

Facts are used in an expert system as the data about which rules reason and represent

the current state of the world. The world in this case exists within the confines of the

knowledge-base management system. Facts are transient data within an expert system and

may therefore be asserted, retracted (deleted) or modified. Facts may take the following

form:

(light red)
(signal walk)
(car stop).

The facts listed here might be part of an expert system that directs traffic. When these facts

are manipulated via rules within an inference engine, then the transient nature of facts

within the expert system will be more clearly shown.

6.4.1.2 Instances

Classes only define structure and cannot contain data. Therefore, classes cannot be pat-

tern matched or directly manipulated by the inference engine. Instances, which are objects

based on user-defined classes, are used to create copies of classes that can be manipulated.

Instances may be pattern matched and can be modified via messages. For example, a print

message might be sent to an instance, resulting in a list of the object's contents output to

the system interface. Concrete examples of instances are developed later in this chapter.









6.4.1.3 Rules

Rules are collections of conditions and actions. When patterns within the rule condi-

tions are matched, actions are taken. Rules have a format based on an if-then construct. A

simple rule is based on the following pseudocode:

(Rule
()
()


()

()
()



()
); end of rule.

This rule can be read as "if and and ... then do

, , .. ."

A rule that controls traffic might have the following format:

(Rule traffic-signal
(change-light red green)
(light red)
(car stop)
(signal walk)

(retract (light red))
(assert (light green))
(retract (car stop))
(assert (car go))
(retract (signal walk))
(assert (signal don't-walk))
); end traffic-signal.

Before the rule traf f ic- signal can fire, the following facts must exist in the knowl-

edge base:

(change-light red green)









(light red)
(car stop)
(signal walk).

After the traffic-signal rule fires, the following facts will be in the knowledge

base:

(change-light red green)
(light green)
(car go)
(signal don't-walk).

6.4.2 Strategies for Matching

6.4.2.1 Matching rules with facts

It is the purpose of an inference engine to manage the process of matching rules with

facts. There are several approaches for pattern matching facts with rules. A simplistic

approach is to have rules search for facts1. Figure 6-6 illustrates the rules-search-for-facts

matching process graphically. The advantage of this method is that the computer program-

ming required is simple to understand and the execution process is a straightforward,

sequential activity. The disadvantage is that this method suffers from combinatorial explo-

sion when the number of rules, rule patterns and facts increase. Under this matching

scheme, about 90% of the total system execution time is spent in searching for facts that

satisfy the rules. The balance of the processing time is spent determining which match

should be acted on first (a process called conflict resolution) and completing the actions

specified by the matched rules. It is desirable to significantly decrease the amount of time

spent in the match phase and thereby increase the efficiency of the knowledge base sys-

tem.





1. The discussion in the next two sections is based on lecture notes from the University of
Florida course CAP 6627, Expert Systems, as taught by Dr. Douglas Dankel, II, Spring
term, 1993.





























Figure 6-6. Matching rules with facts using a rules-search-for-facts approach.


The following scenarios show how quickly this method runs into the combinatorial

explosion. The worst case number of searches required to find a single rule/fact pattern

match can be expressed as


Search = Nrules X Mcond Pfacts (6.1)

where

Nsearch = number of pattern match comparisons
Nrules = number of rules
Mcond = average number of patterns per rule
Pfacts = number of facts in the knowledge base.

Several numerical examples of the combinatorial explosion are summarized in Table

6-3. Notice that for the first case, which is unrealistic since it only has one pattern per rule,

it may require as many as 10,000 comparisons to find a single rule/fact match. The third

case, representative of a large system, requires up to 3,000,000 comparisons to find a suit-









able rule/fact match. High-speed, expensive computing hardware cannot practically make

up for the slow response time in such an inefficient system.

Table 6-3. Combinatorial explosion for the rules-search-for-facts method.

Nrules Mcond Pfacts Nsearch
100 1 100 104
100 3 100 3*104
1000 3 1000 3*106


6.4.2.2 Exploiting temporal redundancy

Fortunately, only a small number of facts change with each cycle. In a system with

hundreds of facts, less than one percent of the facts are likely to change with each cycle.

This means that there is a slow rate of change in the conflict set. Since there is little change

from one time period to the next, there is a large amount of redundant data. This condition

is known as temporal redundancy and it can be exploited by having the facts search for

rules rather than the other way around. Figure 6-7 illustrates how this approach can avoid

unnecessary computation. This facts-match-rules strategy is clearly less cumbersome than

the rules-match-facts strategy. The Rete algorithm was developed specifically to take

advantage of this strategy.

6.4.2.3 Rete network

The Rete algorithm is an efficient many-many match procedure. The method involves

the development of a pattern network and a join network. These networks are illustrated in

Figure 6-8 for a sample rule. The Rete gains efficiency from reusing as many patterns

from similar rules as possible. An example of two rules instantiated within a Rete network

is shown in Figure 6-9. It can be seen from this figure how the pattern network reuses as

much information from similar rules as possible. As facts enter the knowledge base, they

are pushed as far as possible through the Rete networks. Facts that satisfy the patterns of
























rules-match-facts strategy


facts that have
changed since
last cycle


Figure 6-7.


facts-match-rules strategy

Comparison of two strategies for matching rules with facts.


the instantiated rules are placed on an agenda where the inference engine decides which

matches to act on first

6.5 CLIPS Expert System Shell

CLIPS2 [Giar91, Giar93] is an expert system tool developed by the Software Technol-

ogy Branch (STB), NASA/Lyndon B. Johnson Space Center. CLIPS is designed to facili-

tate the development of software to model human knowledge or expertise.

2. C Language Implementation Production System










Sample rule:

(rule R-1
(field-1 field-2 field-3)
(field-4 field-5)


...); end rule R-1


Pattern
Network R-1


node for field-1 node for field-4


node for field-2 node for field-5 *


node for field-3 termin node 2


terminal node 1


join for conditions
1&2


join for conditions
(1, 2 &3

Join
Network rule instantiated
(at this point the rule
can be added to the
conflict set)


Instantiation of a typical rule within a Rete network.


Figure 6-8.










~mpkiuks:


(rule R-2
(TI B ?D)
(T2 ?F)
(T3 ?D ?F)

...); end rule R-2


(rule R-3
(Ti B ?D)
(T2 I)
(T3 ?D ?F)

...); end rule R-3


Pattern
Network R


=t


bind .F = I


bin ?D terminal node 2


te / te
terminal node 1


bind?D


bin)?F


terminal node 3


Join
Network


join 2
of R-2


R-2 instantiated


Figure 6-9. Instantiation of two rules within a Rete network.


R-3




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E29EZFAKI_1H527A INGEST_TIME 2012-02-20T23:17:01Z PACKAGE AA00009040_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES