GRAPH-MATRIX-BASED AUTOMATED TOLERANCE ANALYSIS

AND SETUP PLANNING IN COMPUTER-AIDED

PROCESS PLANNING

by

EDWARD ENHAO LIN, B.S.E., M.S.E.

A DISSERTATION

IN

INDUSTRIAL ENGINEERING

Submitted to the Graduate Faculty

of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of

DOCTOR OF PHILOSOPHY

Approved

Chairperson of the Committee

f \ W ^

Accepted

Dearr^f the Graduate/School

August, 2001
 ACKNOWLEDGMENTS

I would like to express my sincere gratitude to Dr. Hong-Chao Zhang, Dr.

William J. B. Oldham, Dr. James L. Smith, Dr. Ronald Pigott, and Dr. Yung-Nien Yang
for serving in my dissertation committee and for their invaluable guidance and help.
I especially appreciate my advisor, Dr. Hong-Chao Zhang, for his considerate
guidance and support throughout my Ph.D. study. To me. Dr. Zhang is not only a
knowledgeable advisor but also a dear friend. He provided me various opportunities for
academic activities and professional training, propelling my current study as well as
fiature career. Without his help, many of my achievements in the last three years would
be impossible.
I would like to thank colleagues and friends: Dr. Samuel Huang, Dr. Steven Yu,
Dr. Huitian Lu, Dr. Susan Lu, Ms. Kai Jin, Mr. Adib Khan, Mr. Greg Hall, Ms. Kay
Vugrin, Dr. Xueqing Qian, Ms. Yan Zhang and many others for their constructive
discussions in the research. I am grateful to Professor Bingyuan Xue, Professor Jun Ni.
Professor Dejin Hu, Professor Junqi Yan, Professor Shen Dong, Professor De-an Wan,
and some gentlemen in the National Education Committee of China for their continuous
encouragement.
My special gratitude goes to my parents and family for their endless love.
The research was supported by National Science Foundation under the grant
contract DMI-9714029. The research was also partially supported by 1999 Summer
Dissertation Research Award from the Graduate School of Texas Tech University. Their
support has catalyzed the progress of this research.

11
 TABLE OF CONTENTS

ACKNOWLEDGMENTS ii
ABSTRACT vi
LIST OF TABLES vii
LIST OF FIGURES viii
CHAPTER
1 INTRODUCTION 1
1.1 Tolerancing, Setup Planning, and Graph Theory 1
1.2 Problem Statement 3
1.3 Methodology 6
1.4 Research Objectives 8
2 LITERATURE REVIEW 9
2.1 Basic Concepts on Tolerance Analysis 9
2.1.1 Variation Versus Tolerance 9
2.1.2 Standard Definitions on Dimensioning and Tolerancing 11
2.1.3 Tolerance Zone 12
2.1.4 Principle of Tolerance Independency 16
2.2 Tolerance Chart and Dimension Chain 18
2.2.1 Tolerance Chart 18
2.2.2 Dimension Chain 23
2.3 Computer-Aided Process Planning (CAPP) 26
2.3.1 Tasks Involved in Process Planning 26
2.3.2 Setup Planning 27
2.3.3 Fixture Planning 31
2.4 Graph Theory and Algorithms 35
2.4.1 Basic Concepts on Graph Theory 35
2.4.2 Matrices of a Graph 38
2.4.3 Graph Algorithms 39
3 TOLERANCE ANALYSIS 41
iii
 3.1 Basis of Tolerance Analysis 41
3.1.1 Tolerance Zone 41
3.1.2 Characteristic Length and Tolerance Factor 42
3.1.3 Manufacturing Errors Classification 45
3.1.4 Setup Methods 52
3.2 ToleranceA'^ariation Stackup Analysis 52
3.2.1 Dimensional ToleranceA^ariation Stackup Analysis 53
3.2.2 Geometrical ToleranceA^ariation Stackup Analysis 58
3.3 Tolerance Stackup Analysis for the Three Setup Methods 70
3.3.1 Dimensional Errors Analysis 71
3.3.2 Geometrical Errors Analysis 73
3.3.3 Comprehensive Analysis 74
4 TOLERANCE GRAPH AND MATRICES 77
4.1 Tolerance Graph 77
4.1.1 The Definition of "Tolerance Graph" and Its Related Concepts 77
4.1.2 The Application of Tolerance Graph in This Study 81
4.2 Matrices 84
4.2.1 Tolerance Matrix 84
4.2.2 Tool Approach Direction (TAD) Matrix 86
4.2.3 Machining-Feature Matrix 87
4.2.4 The Transfer and Operation on Matrices 88
5 ALGORITM DEVELOPMENT AND ILLUSTRATION 92
5.1 Principles and Rules of Setup Planning 92
5.2 The Setup-Planning Model 97
5.3 The Setup-Planning Algorithm 99
5.4 Illustration 102
5.4.1 Case Study 1: Setup Planning of a Prismatic Part 102
5.4.2 Case Study 2: Setup Planning of a Rotational Part 119
6 IMPLEMENTATION AND VALIDATION 130
6.1 The GATO Setup Planner 130

iv
 6.2 Validation Discussion 135
7 CONCLUSIONS 146
7.1 Conclusions 146
7.2 Contributions 147
7.3 Suggestions for Further Research 148
REFERENCES 150
APPENDIX
A: THE SETUP-PLANNESTG ALGORITHM 156
B: THE MAP OF THE RULES TO THE ALGORITHM 163
C: THE BRIEF USER MENU 164
D: THE COMPLETE SIMULATION RESULTS 166
 ABSTRACT

This research focuses on exploring a systematic approach to tolerance analysis

and setup plarming in Computer-Aided Process Planning (CAPP).
In this research, the geometrical characteristics of various types of dimensional
and geometrical tolerances are mathematically represented and compared. Tolerance
analysis is based on tolerance zone analysis and aims at mathematical representation. The
concept of tolerance weight is defined and applied in tolerance analysis and setup
planning. It makes the comparison of different types of dimensional and geometrical
tolerances possible. Both dimensional tolerances and geometrical tolerances are analyzed
in one-, two-, and three-dimension for tolerance stackup in setup planning.
A graph-matrix-based and tolerance-oriented algorithm is developed for the setup
planning of both prismatic parts and rotational parts. The basic concepts on "tolerance
graph" are introduced. In a tolerance graph, the faces of a part are represented as vertices,
and the tolerance relations among faces are represented as either directed or undirected
edges. A tolerance graph is essentially a directed graph by orienting undirected edges into
directed edges or representing an undirected edge as two directed edges. The concept of
tolerance graph is used consistently in both tolerance analysis and setup plarming through
the research. Accompanying tolerance graph, tool-approach-direction matrix, machining-
feature matrix, and tolerance matrix are defined and applied in representing and
computerizing the relations of tool approach directions, machining features, and
tolerances in setup planning. The experience principles and rules of setup planning are
summarized, explored, and incorporated in the setup-planning algorithm. A setup-
planning prototype based on the algorithm is implemented in C++ and Microsoft Visual
C++.

VI
 LIST OF TABLES

2-1. Tolerances and tolerance zones 13

3-1. Characteristic length of toleranced features 44
3-2. Geometrical tolerance zones and tolerance factors 46
4-1. Circuits in the tolerance graph 81
5-1. The whole setup planning strategy for the illustrated part 120
5-2, The whole setup planning strategy for the illustrated part 129
6-1. Simulation results: variation stackup of setup plan 1 140
6-2. Simulation results: variation stackup of setup plan 2 141
6-3. The comparision of the two setup plans 142
6-4. The initial data (Units: mm) 143
6-5. The simulation result (Units: mm) 143
B-1. The map of the rules to the algorithm 163
D-1. The alternative valid setup plans 166
D-2. The simulation results (unit: mm) 169

Vll
 LIST OF FIGURES

1-1. Relationship between tightness of tolerance and manufacturing costs 1

1-2. Methodological structure of this study 7
2-1. Classification of geometrical variations 10
2-2. Classification and symbols of geometrical tolerances 10
2-3. Interpretation of a dimensional tolerance 16
2-4. Interpretation of some geometrical tolerances 17
2-5. Specification and operations of apart (Wade, 1983) 21
2-6. Sample tolerance chart for the steel plug (redrawn from Wade, 1983) 22
2-7. Types of dimension chains (redrawn from Wang and Li, 1991) 24
2-8. Incorrect location of horizontal centerline 33
2-9. Correct location of horizontal centerline 33
2-10. An acyclic directed graph 36
2-11. The transformation of an undirected graph into a directed graph 37
2-12. A mixed graph 37
3-1. Typical tolerance zones 41
3-2. The projecting relation of tolerance zones 42
3-3. The relation of tolerance factor and toleranced feature 43
3-4. Locating error and machining error 51
3-5. Dimensional relation of 3 holes in a plane 54
3-6. Dimension chain of c' 54
3-7. Dimension chain of c 55
3-8. Dimensional relation of 3 holes in a plane 56
3-9. Dimensional tolerance stackup 57
3-10. One-dimension geometrical tolerance stackup analysis 58
3-11. Tolerance zone distributions for one-dimension tolerance stackup 59
3-12. A part for 3-dimension tolerance stackup analysis 60
3-13. Tolerance graph of the part 61
3-14. Two-dimension tolerance zone of face B of the part 61
viii
3-15. The effect of the translation of face B 61
3-16. The effect of the rotation of face B 62
3-17. The translation of coordinate system 64
3-18. The rotation of coordinate system 65
3-19. Tolerance zones of locating faces 67
3-20. Manufacturing error analysis for NC machining 71
4-1. Pictorial representation of ^,,y 78
4-2. Two equivalent graphs in calculating vertex degree 79
4-3. Circuits in a tolerance graph 80
4-4. Specifications of apart 82
4-5. Tolerance graph of the part 83
4-6. Another tolerance graph of the part 84
4-7. Drawings of a prismatic part (modified from Huang, 1995) 85
4-8. Initial tolerance matrix W\ 86
4-9. The TAD matrix Ui 87
4-10. The machining-feature matrix Vi 88
4-11. The machining-feature matrix V2 89
4-12. Tolerance matrix W2 89
4-13. W2nV2 90
4-14. Transferred Ui u Vi 91
5-1. The setup-planning model 98
5-2. 3-D views of aprismatic part 103
5-3. Design specifications of the prismatic part 103
5-4. The matrix of machining features 104
5-5. The TAD matrix of the part 105
5-6. The transferred TAD matrix of the part 106
5-7. Initial tolerance matrix 107
5-8. Initial tolerance matrix with comprehensive information 107
5-9. Tolerance graph with initial information of the part 108
5-10. Tolerance graph with each face in a definite TAD set 110

ix
5-11. Tolerance matrix with each face in a definite TAD set 110
5-12. Tolerance graph with interior tolerances deleted or screened 111
5-13. Tolerance matrix with interior tolerances deleted or screened 112
5-14. Tolerance graph with undirected edges oriented 113
5-15. Tolerance matrix with undirected edges oriented 113
5-16. Simplified tolerance graph with undirected edges oriented 114
5-17. Tolerance graph with definite setups 115
5-18. Simplified tolerance graph with definite setups 115
5-19. Tolerance matrix with definite setups 116
5-20. The trees of setup precedence 117
5-21. The tolerance graph of setup precedence 117
5-22. The sequencing of setups 118
5-23. The procedure of setup sequencing 118
5-24. Tolerance graph with the sequence of setups 119
5-25. Machining process of the part 120
5-26. 3-D view of a rotational part 120
5-27. Design specifications of the rotational part 121
5-28. The matrix of machining features 122
5-29. The TAD matrix of the part 122
5-30. The transferred TAD matrix of the part 123
5-31. Initial tolerance matrix 123
5-32. Initial tolerance matrix with comprehensive information 124
5-33. Tolerance Graph with initial information of the part 124
5-34. Tolerance graph with each face in a definite TAD set 125
5-35. Tolerance matrix with each face in a definite TAD set 126
5-36. Tolerance graph with interior tolerances deleted or screened 126
5-37. Tolerance matrix with interior tolerances deleted or screened 127
5-38. Tolerance graph with undirected edges oriented 128
5-39. Tolerance matrix with undirected edges oriented 128
5-40. Simplified tolerance graph with undirected edges oriented 128

X
5-41. Machining process of the part 129
6-1. The GUI structure of the GATO setup planner 130
6-2. The GUI of the GATO setup planner 131
6-3. The initial information of Parti 132
6-4. The setup planning report of Parti 133
6-5. The initial information of Part2 134
6-6. The setup planning report of Part2 135
6-7. The flowchart of the simulation procedure 137
6-8. The simplified drawing of 5-27 138
6-9. The simplified tolerance graph of 6-8 139
6-10. The tolerance graphs of the two setup plans 139
6-11. Column charts of simulation results 141
6-12. The simulation result 144
D-1. Tolerance graphs (setup plans 1 ~ 10) 167
D-2. Tolerance graphs (setup plans 11 ~ 20) 168

XI
 CHAPTER I
INTRODUCTION

1.1 Tolerancing, Setup Planning, and Graph Theory

The economy of any industrialized nation simply cannot afford to squander vast
sums of money represented by the scrap, rework, and debugging time generated by
uncontrolled tolerances. Competition, both internal and external to the country, puts a
premium on quality of design and quality of manufacturing, and this quality in large part
is based on control of tolerances (Wade, 1983).
The relationship between tightness of tolerance and manufacturing cost is shown
in Figure 1-1 (redrawn from Wade, 1983, pp. 2-2).

0.030

0.025

0.020

Tolerances
0.015
(mm)
0.010

0.005

0.000
1 2 3 5 11 17
Relative costs

Figure 1-1. Relationship between tightness of tolerance and manufacturing costs.

The cost of production increases geometrically for uniform incremental tightening

of tolerances. The relationship has been found to be correct in numerous writings by
different specialists in a number of countries. So a systematic approach to tolerancing
will have as one of its main thrusts the maximizing of production tolerance within the
framework of manufacturing concepts (Wade, 1983).
Tolerancing is one of the most significant topics in the manufacturing area.
Continuous study and application of it have been carried out world wide in the last 100
years (Zhang, 1997).
Setup planning is to group the to-be-machined features of a part into definite
setups, sequence the setups, and select locating datum for each setup, so that the part can
be precisely and economically machined in accordance with the design specifications.
Setup planning is the critical bridge between general process planning and
detailed operation planning; it is also the intimate upstream of fixture planning. Setup
planning significantly affects the overall cost and quality of part machining. Setup
planning will just minimize the effect of manufacturing errors instead of controlling the
errors.
Automated setup planning and fixture planning have been studied for more than
10 years (Huang, 1995; King and de Sam Lazaro, 1994; Chen, 1993; Sakurai, 1991;
Menassa and DeVries, 1990; Lee and Haynes, 1987). Geometrical analysis, kinematics
analysis, force analysis, and deformation analysis in setups were extensively studied.
However, tolerance analysis is somehow the least explored issue in setup planning
(Huang and Gu, 1994; Boerma and Kals, 1988).
The graphs in graph theory are simple geometrical figures consisting of points
(vertices) and lines (edges) which connect some of these points; they are sometimes
called "linear graphs." Because of this diagrammatic representation, graphs have been
found extremely useful in modeling systems arising in physical science, engineering,
social science, and economic problems. The fact is that any system involving a binary
relation can be represented by a graph (Chen, 1997).
Wide application of graph theory can be seen in the area of industrial engineering,
such as operation research (OR), production planning and control (Foulds, 1992),
computer-aided design (CAD) (Soenen, 1995). However its application in tolerancing
(Whybrew, 1990; Sermsuti-Anuwat, 1992) and setup planning (Huang, 1995) is rather
rare and primitive. Since binary relation and diagrammatic representation in tolerancing
and setup planning are widespread, the effective application of graph theory in the two
areas is highly expected.
Zhang (1991), Huang (1995), and Mei (1996) have continuously researched on
tolerance analysis and setup planning in the last 9 years, the tolerance stackup analysis
for different setup methods and the graph-matrix approach to setup planning were
explored. This study is based on their pilot work.
In conclusion,
1. Setup plarming is the target of this study;
2. Tolerance analysis is the basis of setup planning, although tolerance analysis
will go a little further than only for setup planning in this study;
3. Graph theory accompanied with matrix theory is the basic mathematical tool
for setup planning. Graph theory also aids tolerance analysis in this study, that
is, it helps to "automate" both tolerance analysis and setup planning.

1.2 Problem Statement

Even in this computer age, many process-planning activities still rely on the
experience of process planners. Process planners obtain their experience from earlier
training as machinists (most typical), from books, or from discussion with colleagues.
This kind of information can be passed from person to person and generation to
generation. However, there are some problems associated with such a planning base
(Chang, 1998):
1. Experience requires a significant period of time to accumulate;
2. Experience represents only approximate, not exact knowledge;
3. Experience is not directly applicable to new process or new systems.
Without exception, tolerance analysis and setup planning, two significant topics in
process planning, still heavily rely on experience today. In comparison with the boom of
Computer-Aided Design (CAD), Computer-Aided Manufacturing (CAM), and
Computer-Aided Process Planning (CAPP) in the last two decades, the computerization
of tolerance analysis and setup planning lags. Experts and their experience still dominate
the area of tolerancing and setup planning, which is one of the reasons why few CAPP
systems have been applied in industry. As essential components of CAPP, automated
tolerance analysis and setup planning urge more research effort to implement Computer-
Aided Manufacturing (CAM).
Tolerance stackup analysis usually only deals with dimensional tolerances in one-
dimension. The resultant tolerance is always the added sum of component tolerances
(Wade, 1983). Analysis and control of dimensional tolerances are relatively matured
compared to those of geometrical tolerances (Bj(l)rke, 1989). The stackup of geometrical
tolerances was usually ignored or replaced by the stackup of dimensional tolerances. In
this paper, both dimensional tolerances and geometrical tolerances will be considered in
one-, two-, and three-dimensions.

Mathematical presentation is a tendency in dimensioning and tolerancing. The

Y14 Committee created the Y14.5.1 Subcommittee in response to a need identified
during a National Science Foundation (NSF) workshop. The workshop report strongly
identified a need for a mathematical definition for the current tolerancing standards. Tom
Charlton coined the phrase "mathematization of tolerances." The phrase meant to add
mathematical rigor to the Y14.5M standard. The response is the present standard, ASME
Y14.5.1 M-1994. This new standard creates explicit definitions for use in such areas as
CAD and CAM (ASME Y14.5.1 M-94, 1994). Voelcker H.B. predicted that one of the
most important advances in geometrical tolerancing would be made over the next decade
is: "One or more generative formulations of geometrical tolerancing will be produced. A
generative formulation will be more general than current practice but should contain the
current GD&T facilities as special cases. A generative formulation should be teachable in
the engineering colleges because it will be based on a small set of underlying
mathematical principles" (Zhang, 1997). Generative formulation of geometrical
tolerancing is one of the objectives in this study.
Setup planning has been studied as a significant subject in CAPP, however it is
still a relatively unexplored research issue (Boerma, 1988; Hayes, 1989; Demey, 1996:
Chen, 1998). Huang (1995, 1998) explored the setup planning of rotational parts.
However, the setup planning of prismatic parts is far more difficult than that of rotational
parts because:
 1. The tolerance stackup in the machining of rotational parts occurs only in axial
direction, there is no tolerance stackup in radial direction. For prismatic parts,
tolerance stackup may occur in X, Y, and Z coordinates in Cartesian
coordinate system.
2. Types and numbers of features of prismatic parts are usually more than that of
rotational parts. The machining constraints and precedence for the features in
prismatic parts are more complicated than those of rotational parts are, and the
setup number of prismatic parts is more than that of rotational parts.
3. There are more types of tolerances, both dimensional and geometrical
tolerances, in prismatic parts.
4. The fixtures for prismatic parts are more various and complicated that those of
rotational parts, and the locating and clamping of prismatic parts are more
complicated.
Undirected graphs were used in setup planning (Huang, 1995; Huang, 1998).
However, their effectiveness and efficiency are limited because:
1. All geometrical tolerances except some form tolerances (Henzold, 1995) have
one or more datums. Since datum features, except black datum features, must
be machined before being used as machining datum (Zhang, 1994), directed
graphs are more explicit for representing these tolerances.
2. Even for tolerances (such as dimensional tolerances and some form
tolerances) without explicit datums in specifications, a machining datum is
usually needed in machining to satisfy the tolerance specifications. That is, the
"directions" of dimensional tolerances and form tolerances have to be decided
at later steps of setup planning.
3. For parts needing multiple setups, the sequence of the setups has to be decided
at last. Hence, directed graphs are preferred to aid sequencing setups.
In conclusion, automated tolerance analysis and setup planning for both prismatic
parts and rotational parts is not well explored, it is the target of this research.
 1.3 Methodology
Tolerance is the total amount a specific feature is permitted to vary; it is the
difference between the maximum and minimum limits (ASME Y14.5M-1994. pp. 3).
Tolerance is illustrated by the section formed by two points, area formed by lines or
circles, or space formed by faces (Henzold, 1995). The section, area, or space is called
the tolerance zone. Tolerance analysis in this study is started from the analysis of
tolerance zones. Differential formulation, geometrical representation, and kinematics
analysis are applied in tolerance analysis based on tolerance zone analysis. Based on
tolerance zone analysis, the concept of "tolerance weight" which corresponds to "weight"
in graph theory is introduced to compare the relative tightness of tolerances.
Applying graph theory, this study introduces a directed-graph approach to the
setup planning of both rotational parts and prismatic parts. The basic concepts of
"tolerance graph" are introduced. In a tolerance graph, the faces of a part are represented
as vertices, and the tolerance relations among faces are represented as edges. A tolerance
graph is essentially a directed graph by orienting undirected edges into directed edges or
representing an undirected edge as two directed edges. The directed-graph approach is of
particular interest in the analysis of engineering problems, linear problems especially,
since the associated directed graphs in many cases can be set up directly by inspection of
the physical system without the necessity of first formulation of the associated equations
(Chen, 1997; BoUobas, 1998). In fact, a systematic methodology based on "tolerance
graph" for machining feature recognizing, tolerancing, tolerance analysis, and setup
planning is promising in CAPP.
A graph or a directed graph is usually presented with an incidence matrix, a
circuit matrix, and a cut matrix in graph theory (Chen, 1997; Diestel, 1997). In addition,
matrices are effective in representing and computerizing relationship of machining
features, tool approach directions (TADs), and setup groups for all faces of a part. Matrix
theory, accompanying graph theory, is also an effective mathematical tool in this study.
No setup-planning algorithm can exist without experience principles and rules of
setup planning. In some sense, automated setup planning is just to computerize
experience principles and rules of setup planning. The summarization and mathematical
representation of principles and rules are the basis of this study, from tolerancing
formulation to setup-planning algorithm development.
In sum, the methodological structure of this study is illustrated in Figure 1-2,
which shows:
1. The goal of this study is to automate tolerance analysis and setup planning in
CAPP.
2. The two major tasks to achieve the goal are setup planning and tolerance
analysis; tolerance analysis is the basis of setup planning, while setup
plarming affects tolerance control.
3. Experience principles and rules are the base of setup planning and tolerance
analysis.
4. Graphs and matrices are the major mathematical tools for tolerance analysis
and setup planning.
5. Tolerance zone analysis is the fundamental approach to the tolerance analysis
in this study.

Goal

Base

Figure 1-2. Methodological structure of this study.

 1.4 Research Objectives
The general objective of this study is to automate tolerance analysis and setup
planning in CAPP, specifically:
1. To explore the geometrical characteristics of various types of tolerances,
mathematically represent and compare various types of tolerances, and
formulate tolerance analysis in setup plarming;
2. To analyze both dimensional tolerances and geometrical tolerances in one-,
two-, and three-dimension for tolerance stackup in setup planning;
3. To develop a systematic methodology for both tolerance analysis and setup
planning;
4. To develop a setup-planning algorithm incorporating experience principles
and rules for both rotational parts and prismatic parts;
5. To develop a prototype of setup planning system, aiming to improve the
performance of CAPP systems.

8
 CHAPTER II
LITERATUE REVIEW

2.1 Basic Concepts on Tolerance Analysis

2.1.1 Variation Versus Tolerance
In some papers and documents, the distinction between "variation" and
"tolerance" is somehow obscure. For example, "variation stackup analysis" in machining
was confused with "tolerance stackup analysis"; "machining variation" was confused
with "machining tolerance". In this study, the distinction between "variation" and
"tolerance" is clarified in according to the following definitions:
Variation is the deviation of a feature (geometrical element, surface, or line) from
its nominal size or shape (Henzold, 1995, pp. 2). Tolerance is the total amount a specific
feature is permitted to vary; it is the difference between the maximum and minimum
limits (ASME Y14.5M-1994, pp. 3). In other words. Variation is the changes that occur
in size and shape of features or parts we encounter. Tolerances are engineering
specifications that define the extremes in variation from ideal or nominal dimensions that
we will allow or accept (Liggett, 1993, pp. 2).
Variation is also referred as deviation or error in some documents. In this study,
the three terms are considered synonymous or exchangeable.
From the above definitions, tolerance stackup deals with the variation limits in
machining, while variation stackup deals with actual variation. In this paper, tolerance
stackup analysis is based on variation stackup analysis. The mathematical formulas of
tolerance stackup and those of variation stackup basically coincide by substituting
variation variables with tolerance variables.
In machining, the geometrical variations of a part originate from the loosens or
error in guides and bearings of the machine tool, deflections of the machine tool or the
workpiece, error in the fixture of the workpiece, hardness deflection or wear, re-chucking
or the workpiece, and so on (Henzold, 1995). The variations are classified as macro-
variations and micro-variations with subdivisions as shown in Figure 2-1.
 The classification and symbols of geometrical tolerances are shown in Figure 2-2
(ASME Y14.5M, 1994; ANSI Y14.5 M-1982).

Size variations
v . ^ , ;, .1...: : : i., - : ;, A

Macro- | _ Form variations

variations |
Orientation variations
\';:--: ,:.:. : ,' - * >>

Geometrical Location variations

variations k,:::':'::'"-::': , .:, , . \

Waviness

Micro- Roughness
variations
Surface discontinuities

edge variations

Figure 2-1. Classification of geometrical variations.

- ( Straightness ^
- ( Profile (form) of lines y
Unrelated geometrical - ( Roundness ^ (^
tolerances
(Form tolerances) Q Flatness ~ ) /~/
-{Profile (form) of surfaces^
- ( Cylindricity~^ /C/

Geometrical -( Angularity } /^
tolerances - ( Orientation -( Parallelism ) / /

-{^ Perpendicularity ) \

C Position J ^
Related geometrical -( Coaxiliality ) (o)
-( Location }-
tolerances
-( Symmetry ^ - ^

Circular
run-out
-( Run-out y
Total
run-out 1/
Figure 2-2. Classification and symbols of geometrical tolerances.

10
 In the area of dimensioning and tolerancing, only macro-variations are treated.
Micro-variations are usually studied in super-precision machining area, which is not
concerned in this study. In accordance to the geometrical variations, there are
dimensional tolerance, form tolerance, orientation tolerance, and location tolerance. The
definitions on these tolerances are in section 2.1.2.

2.1.2 Standard Definitions on Dimensioning and Tolerancing

The following basic definitions on tolerance are extracted from ASME Y14.5M-
1994 and ANSI Y14.5M-1982, which coincide in these definitions. These definitions are
abided by uniformly in this research.
Datum: A theoretically exact point, axis, or plane derived from the true
geometrical counterpart of a specific datum feature. A datum is an origin from which the
location or geometrical characteristics of features of a part are established.
Feature: The general term applied to a physical portion of a part, such as a
surface, pin, tab, hole, or slot.
Datum feature: An actual feature of a part that is used to establish a datum.
Dimension: A numerical value expressed in appropriate units of measure and used
to define the size, location, geometrical characteristic, or surface texture of a part feature.
Basic Dimension: A numerical value used to describe the theoretically exact size,
profile, orientation, or location of a feature or datum target. It is the basis from which
permissible variations are established by tolerance on other dimensions, in notes, or in
feature control frames.
Dimensional tolerance: The total amount of a specific dimension is permitted to
vary, it is the difference between the maximum and minimum limits.
Geometrical tolerance: The general term applied to the category of tolerances
used to control form, profile, orientation, location, and run-out.
Form tolerance: Permitted maximum value of the deviation of a feature
(geometrical element, surface, or line) from its nominal form.

11
 Orientation tolerance: Permitted maximum value of the deviation of a feature
from its nominal form and orientation.
Location tolerance: Permitted maximum value of the deviation of a feature
(surface, line, or point) from its nominal location.
Run-out tolerance: It includes partly orientation tolerance (axial run-out
tolerance), and partly location tolerance (radial run-out tolerance).
Datums can have different priority (order of precedence) as follows (Henzold,
1995, pp. 34):
Primary datum: Oriented according to the minimum rock-requirement relative to
the auxiliary datum element;
Secondary datum: Orientated without tilting relative to the primary auxiliary
datum element (only by translation and rotation) according to the minimum rock-
requirement relative to the secondary auxiliary datum element (the secondary auxiliary
datum element is perpendicular to the primary datum elements);
Tertiary datum: Positioned without tilting and rotation relative to the primary and
to the secondary auxiliary datum element only by translation until contact with the
tertiary datum element (the tertiary auxiliary datum element is perpendicular to the
primary and secondary auxiliary datum elements).

2.1.3 Tolerance Zone

According to ASME Y14.5M-1994 and ANSI Y14.5M-1982, tolerance zone is
the zone that confines the variation of a toleranced feature. It is a regular zone formed by
straight lines, circles, planes, cylindrical faces, cone faces, or spherical. The tolerance
zone of each type of tolerances is shown in Table 2-1, Table 2-2, and Table 2-3.
Depending on the toleranced characteristic and depending on the drawing
indication, the tolerance zone is one of the following:
1. Area within a circle.
2. Area between two concentric circles.
3. Area between two equidistant lines or between two parallel straight lines.
4. Space within a sphere.
12
 Table 2-1. Tolerances and tolerance zones.

Type Drawing indication Tolerance zone

O

c3

C/3 0 0.02

O
(U .ti '^
o
u <u
0.02

O
0.02
o
^'
u.uz 1
\-l
<u

GO

S 0 0.02

O
;3
OH

i^ 02

O
7 0 02
^^
c/2 y

13
 Table 2-1. (Continued).
Type Drawing indication Tolerance zone

^ (U

J-^i- S-
on
o

(U

.11 }-i
o ^ 1 0.02 A
A

PH
o
// 0.02 A
I
PH

0 0.02

PH -^

(U .^ (U 1 @ 00.02 A
^' 0 0.02
-H CO

o
F-H o 3
o
o

^ 0.02 A

(U c
g c3

14
 Table 2-1. (Continued).

Type Drawing indication Tolerance zone

(U
o
Oj

2 i^
I ^ 0.1 A-B

(U ik iL
o
;3
o o ^' V
I A
B
US
O

^r
Zy" 0.1 A - B
i 0.1

2 iL ik

^' V
LJ^V^

O o /\ B

;3 0.1
O (U
/" 0.1 A
CD
O
o
(=1
;-!
i k
A ^
OS (U ^'
/ \.
Viz
o I
s <^
B ^ ly 0.1 A
13 k

x ^
cS O ^r
^ ->
/V
f2

CI / 0.1 A
_o
'-t-> fi fi c =
o .S o 0.1
' 1-H
y CD

o
S-H
C-C

o
o r
S 3fl O
^ 0.1 A Coaxial

3
o a i-H
I
o
5 U
Axial cut-section

15
 5. Space within a cylinder.
6. Space between two equidistant faces or between two parallel planes.
7. Space within a parallelepiped.
The concepts on tolerance zone are the basis of tolerance analysis in this study.

2.1.4 Principle of Tolerance Independency

It is rather complicated to consider dimensional tolerance and geometrical
tolerance simultaneously in variation and tolerance analysis. The International
Standardizing Committees ISO/TCI0/SC5 "Technical Drawings, Dimensioning and
Tolerancing" and IS0/TC3 "Limits and Fits" decided to standardize in ISO 8015 the
principle of independency as the fundamental tolerancing principle (Henzold, 1995,
pp.182). It states that:
"Each requirement for dimensional or geometrical tolerancing specified on a
drawing shall be met independently, unless a particular relationship is specified, i.e.
maximum material requirement, least material requirement, or envelope requirement. "
For example in Figure 2-3, the linear dimensional tolerances 20o' controls only

local sizes (two-point measurements) (e.g., wall thickness), but not the form deviations of
the feature. Angular dimensional tolerances, specified in angular units (e.g., 1), control
the general orientation of lines or line elements of surfaces, but not their form deviations.
The general orientation of the line derived from the actual surface is the orientation of the
contacting line of perfect form.

20 20 -0.1 20 -0.1

Drawing indication Possible actual local size

Figure 2-3. Interpretation of a dimensional tolerance.

16
 In the same principle, the interpretation of some geometrical tolerances is showTi
in Figure 2-4 (Henzold, 1995).
When no relationship is specified the geometrical tolerance applies regardless of
feature size, and the two characteristics are treated (inspected) as unrelated requirement.
Consequently the principle of independency demands a separate indication for each
requirement (which may cause a particular checking operation and may be inspected
regardless of other characteristics).

0.2 max. straightn. dev. any cross-section

o 0.1
i k
1
--HI 020 max. size
0 2 0 0.1
020 n"^ max.
0,
0.2 circ. dev.
Tolerancing ISO8015 Interpretation, extreme permissible part

^1500.5 ^150.5 0.4

0.4 A

350-0.6
Drawing indication Interpretation, extreme permissible part

Figure 2-4. Interpretation of some geometrical tolerances.

The principle of dependence (refer to some national standard, e.g., ANSI Y14.5
Rule #1, BS 308, DIN 7176) applies to size and form of isolated features only. It doesn't
apply to orientation or location of related features. It neither applies to angular
dimensions or angular dimensional tolerances (Henzold, 1995, pp. 182). The principle of
independency (ISO 8015) together with general geometrical tolerances (e.g., IS02768)
and the envelope requirement, where applicable, provides the same interpretation of the
drawings worldwide (no misunderstanding due to different national rules). Hence, the
principle of independency is preferred to the principle of dependence in this study.

17
 2.2 Tolerance Chart and Dimension Chain
Tolerance analysis and control are important during manufacturing to ensure that
parts meet design specifications. Analysis and control are very difficult when parts are
complex, design tolerances are very tight, the number of operations needed to make the
parts is large, and changes in machining datums are frequent. These conditions are
common in the aerospace and automotive industries. Tolerance control in manufacturing
involves control of individual processes and control of tolerance stacks from a sequence
of processes (Zhang, 1997, pp. 13-20).

2.2.1 Tolerance Chart

A tolerance chart is a graphic tool used to ensure the orderly and accurate
development of mean sizes and working tolerances required by a new manufacturing
process or to analyze a set of existing mean dimensions and their tolerances to determine
if a part can be made to print. When tolerance stackup problems must be handled, the
easiest, quickest, and most foolproof way is by use of the tolerance chart (Wade, 1983).
Tolerance chart is only buiU after all the initial engineering decisions have been
made concerning the process. These decisions include:
1. The sequence of operations to be performed.
2. The machine selection for each operation, based on its capacities and known
accuracy.
3. The dimensioning patterns for the cuts to be made in each operation.
4. The selection of the locating surface to be used in each operation.
5. The kind and type of tooling to be used in each operation to control
geometrical characteristics such as flatness, parallelism, concentricity,
symmetry, etc.
Once these decisions have been made, and possibly subjected to critiquing by tool
engineers, the master mechanic, etc., a tolerance chart can be constructed to generate the
dimensions and tolerances required by each process cut. Properly constructed, the
tolerance chart will verify that the following criteria for economical production ha\ c been
satisfied:

18
 1. Within the framework of the process/tooling decisions, as much as possible of
the blueprint maximum tolerance has been allocated among the in-process
cuts, which results in the maximum possible tolerance being assigned to each
cut in the process.
2. The minimum and maximum stock removals on secondary cuts are practical
and acceptable to the shop.
3. Every tolerance assigned is equal to and preferably larger than the estimated
process capability for the cut in question. Since the relationship between the
working tolerance and the process capability has a direct bearing on the
frequency of tool changes or adjustments, many companies have in-house
rules that call for the working tolerance to be 1.5 ~ 2.0 times the process
capability value.
During the course of building the tolerance chart, it may becomes obvious that
one or more of the initial process/tooling decisions results in assigning an impossibly
tight tolerance to an in-process dimension. When this happens, it is necessary to change
these decisions to satisfy the criteria for economic production.
Since all these decisions are still in the paper stage, that is, no tooling has yet been
designed, no great time or dollar loss will occur if a process change is required.
The widespread and growing use of NC machining has reduced the extent of the
tolerance stackup control problem by allowing cuts to be machined as shown on the
blueprint dimensioning schemes, by eliminating manual control of machine decisions
affecting the cuts, and by reducing the number of location surface changes and the
attendant fixturing required by non-NC machining. In general, it has also improved size
control and control of geometrical characteristics of part features. However, not all
tolerance stackups are eliminated by using NC machines (Wade, 1983).
Six distinct steps are involved in the construction of a tolerance chart (Wade,
1983):
Step 1: The outline of the part is drawn at the top of the chart. Feature plane lines
and machining cuts are introduced. All solid cuts are indicated in the stock remo\al
column. The appropriate number (N + 1) of feature plane lines corresponding to the N

19
blueprint dimensions are drawn vertically. The extra feature plan line corresponds to the
surface that receives the initial facing cut and serves as a datum for subsequent cuts.
Step 2: Balance dimensions are installed between feature planes generated by
machining cuts which contribute to the tolerance on one or more mean blueprint values.
In addition, blueprint tolerances are budgeted among cuts present in all schematics, due
consideration being given to (1) the process capabilities of the machine tools performing
each cut, and (2) the occurrence of one or more machined cuts in the schematics for
several blueprint dimensions.
Step 3: The tolerances on cuts from the previous step are entered in the chart.
Other chart entries in this step are (1) mean blueprint values, (2) their tolerances, and (3)
line numbers involved for all stock removals on secondary cuts and balance dimensions.
Step 4: Schematics for blueprint dimension are constructed so that the dimension
chains can be identified. The schematics do not include all the machining cuts in the
process plan. Tolerances on cuts not appearing in the schematics do not influence the
blueprint resultant dimensions but contribute to stackup on stock removal tolerances. An
additional set of schematics is constructed for secondary cuts on feature planes where the
stackup of stock removal tolerance must be controlled.
Step 5: Mean stock removals are computed by the careful consideration of the
relevant machine tool/tool/work material system for each constituent cut in the schematic.
Step 6: Mean dimensions are computed for machined lengths and balance
dimensions, starting from the bottom of the chart and working toward the top. A system
of equations involving machining dimensions and balance dimensions is solved through
inspection and backtracking until all known mean dimensions are obtained.
Figure 2-5 shows the blueprint and process plan of a steel plug. A sample
tolerance chart for the steel plug is shown in Figure 2-6.
For rotational parts a single chart per workpiece is sufficient to control tolerances
along the axis of the workpiece. There is no possibility of stackups occurring in the radial
direction; therefore, a second chart is not required. For prismatic parts it is necessary to
control tolerance stackups in three dimensions, hence at least three charts are necessary
for each workpiece. These charts will, in general, not be independent, as some surfaces,

20
 1.900 .0010
-^^1.3OO.OO10 '<'OO10
G 1.000 .0010
X

Y G
A 1.000 .020
2.000 .009
J 3.000 .002
4.000 .005

(a)

OP. 10: FACE-CNT.-TURN-CUT OFF w&s OP 40: GRIND DIAMETER & SHOULDER NORTON

X-i ^ 1 ^ ^-X-

4-A

< J..5,

OP 20: FACE-CNT.-TURN MONARCH OP 50: GRIND DIAMETER & SHOULDER NORTON

< (, ' \
4 8 ,

OP 30: GRIND DIAMETER & SHOULDER NORTON OP 60: GRIND END FACE BU\NCHARD

20
i^^^

(b)

Figure 2-5. Specification and operations of a part (Wade, 1983).

(a) Blueprint dimensions and tolerances for sample part - steel plug,
(b) Workpiece strip layout - operations to machine steel plug

21
 (^1.915 0.020 ) 1.015 0 . 0 2 0
(20)1.900 0.010 ' 1.000+0.010 (To) 1.315 0.020
1 i r n i !<; j - n nirTi .
V U) l.JUO U . 0 1 k ?
(30)1.600 0.010

LINE OPER. REV. MACH. MACHINE TO BALANCE DIM LINES STOCK REM.
NO. NO. NO. USED MEAN TOL. >k E\ () [) I
MEAN TOL. INV. MEAN TOL.
1 10 W&S .979 .003 ^ 1 , 4^ SC LID
2 10 1.994 .003 - 2 ^ ' SC LID
3 10 3.003 .003 SC LID
d 3
4 , 4 , '> 1.009 .006 3-2
5 , 5 2.024 .006 3-1
6 10 4.031 .010 6 SC LID
7 . 7 , 1.028 .013 6-3
8 20 MONA. 1.008 .004 - 8 4 7-8 .020 .017
9 9 , 2.017 .010 4+8
10 10 3.032 .010 5+8
11 11 4.011 .007 3+8
12 30 NORT. 1.000 .001 12 8-12 .008 .005
13 1.017 .011 9-12
, 13 ,
14 2.032 ,011 10-12
, ""^
15 40 NORT. 1.000 .003 >, 15 ^ 13-15 .017 .014
16 16 , 2.000 .004 12+15
17 . 1'' , 1.032 .014 14-15
18 50 NORT. 1.000 .015 17-18 .032 .029
. ^^
19 19 , 3.000 .019 16+18
20 60 4.000 .001 20 z 11-20 .011 .008
21 21 w 3.000 .002 20-12
22 22 X 2.000 .005 20-16
23 . 23 , Y 1.000 .020 20-19
24
25
26
27
28
29
30
31 ;< Ii () I) E
32 BLUEPRINT RESULTANTS
33 4.000 .005 ,z 4.000 .001 20
34 1.000 .020 w 1.000 .020 23
35 2.000 .009 ' X 2.000 .005 22
36 3.000 .002 Y 3.000 .002 21
37

Figure 2-6. Sample tolerance chart for the steel plug (redrawn from Wade, 1983).

and hence tolerances, will appear on more than one chart. The charts must be linked
together through the common surfaces (Whybrew, 1990).

22
 Tolerance chart analysis is still an effective tool for handling the tolerance stackup
problem. Many researchers are interested in automating the task of tolerance chart
analysis. Ahluwalia and Karolin (1986) developed a computer-aided tolerance control
system named CATC. Xiaoqing and Davies (1988) developed an interactive computer-
aided dimensional planning system named CADP.
Weill (1988) developed a computerized tolerancing and dimensioning software to
optimize tolerance ranges in relation to functional requirement and equipment
capabilities. Dong and Soom (1990) introduced an optimal design approach to tolerance
allocation. Ngoi and Fang (1994) developed a Gauss elimination technique for deriving
the operational dimensions and balanced tolerances during tolerance charting.
Graph theory was also applied in tolerance charting. Irani et al. (1989) and Mittal et al.
(1990) proposed a graph-theoretic representation of tolerance chart. Sermsuti-Anuwat
(1992) developed a tolerance-charting algorithm based on graph-theoretic techniques.
The sequence of operations to produce the component is represented as rooted tree that is
a special kind of directed graph. Ji (1993) proposed a tree approach for tolerance
charting, which is also a graph-theoretic approach.

2.2.2 Dimension Chain

A dimension chain is a consecutive series of interrelated dimensions that form a
closed loop. This chain of dimension refers to a single part or a group of parts. The
essential prerequisites for all dimension chains are the continuity of the interrelated
dimensions and the closed-loop contour formed by these dimensions (Wang and Li, 1991,
pp. 86-90).
Dimension chains may be classified, based on their forms, as follows:
1. Dimension chains comprising linear dimensions and parallel links (Figure 2-7
(a)). The majority of dimension chains are of this type.
2. Dimension chains comprising linear dimensions and non-parallel links (Figure
2-7 (b)). However, if all the dimensions are projected on a single axis, this
class of dimension chains can be converted to the first type.

23
 3. Dimension chains with angular dimensions (Figure 2-7 (c)). Because of the
difficulty encountered in making angular measurements, angular dimension
chains are frequently replaced by dimension chains with linear dimensions.
Such dimensions can be checked by means of gages for measuring center-to-
center distance.
4. Dimension chains in spaces in which a series of dimensions do not lie on the
same plane. Such chains are rarely seen in practice and are solved by
projecting all the dimensions onto a single plane.

4 A

' B '' C '

(a) (b) (c)

Figure 2-7. Types of dimension chains (redrawn from Wang and Li, 1991).

Formulas for manipulating dimension chains of type 1 are (Wang and Li, 1991,
pp. 89):
1. Formula for calculating nominal dimensions:

(2-1)

where
C = nominal dimension of the concluding link;
^ / = sum of nominal dimensions of all increasing links;

^ D = sum of nominal dimensions of all decreasing links.

2. Formulas for calculating limiting values of dimensions:

max ~ / i max / . min

(2-2)

24
 Q.=Z^n-Z^- (2-3)
where
Qax'^mm ^ maxlmum and minimum values of concluding link,
respectively;
Z /max 2 J ^min ^ ^^"^ ^^ maxlmum and minimum values of increasing
links, respectively;

Z /^m.v y^ /^r^in ^ sum of maxlmum and minimum values of decreasing

iTiax, ^^^ min **-^

links, respectively.
3. Formulas for calculating upper and lower deviations of dimensions:

A^c=Z^^/-Z^iD (2-4)

where
l^y^,/S.,j^ = upper and lower deviations of concluding link, respectively;
V Ay^ V A^/ = sum of upper and lower deviations of increasing links,
respectively;
V ls.jjj^ V A^^ = sum of upper and lower deviations of decreasing links,

respectively.

4. Formula for calculating tolerance ranges:

^c=I'5cp (2-6)

where
d(^ = tolerance range of the concluding link;
V 5^^^^ = sum of tolerance ranges of all component links.
Tolerance chains can be divided as dimensional chains and geometrical chains.
The concept of geometrical chain has not ever been seen in documents. The interaction of
dimensional tolerances is usually in one direction or on one plane, while the interaction

25
of geometrical tolerances may be in multi-directions or on multi-planes. For the setup
planning of prismatic parts, 3-dimension tolerance interaction is inevitable.

2.3 Computer-Aided Process Planning (CAPP)

Process planning has been defined as "The subsystem responsible for the
conversion of design data to work instruction" (Link, 1976). In detail, process planning is
the function within a manufacturing facility that establishes the processes and process
parameters to be used (as well as those machines capable of performing these processes)
in order to convert a piece-part from its initial form to final form which is predetermined
(usually by a design engineer) in an engineering drawing. Alternatively, process planning
could be defined as the act of preparing detailed work instructions to produce (machine,
weld, assemble, and so on) a part (Chang, Wisk, and Wang, 1998).
The role of process planning is to convert the design specifications into
manufacturing instructions.
The objective of process planning is not only to make the part relationships within
the specified tolerances but also to make them contain fewer manufacturing error sources.

2.3.1 Tasks Involved in Process Planning

Process planning can be divided into two stages: routing planning and machining
operation design. Process routing planning is an overall planning of the manufacturing
process of a part. The objective is to determine the content, and sequence of operations in
a process plan. The following tasks are performed in process routing planning (Wang and
Li, 1991, pp. 99):
1. Analysis of part drawing.
2. Selection of machining methods and machining routes for each part face to be
machined.
3. Determination of sequence of machining operations.
4. Selection of locating references for each machining operation.
5. Arrangement of heat treatment operations.
6. Arrangement of auxiliary operations.

26
 A completely CAPP system would include (Wang and Li, 1991, pp. 207):
1. Design input.
2. Material selection.
3. Process selection.
4. Process sequencing.
5. Machine and tool selection.
6. Intermediate surface determination.
7. Fixture selection.
8. Machining parameter selection.
9. Cost/time estimation.
10. Plan preparation.
11. NC code generation.

2.3.2 Setup Planning

A setup consists of locating and clamping of a workpiece. After locating, the
workpiece is in a correct position relative to the machine tool or workholder. Clamping is
used to fix the workpiece at the located position. The term "setup" has two different
meanings (Wang and Li, 1991, pp. 6):
1. A setup is to locate and clamp a workpiece on a machine tool or workholder
to obtain the correct position relative to the machine tool.
2. A setup is part of an operation performed without reclamping the workpiece.
Setup is usually referred as a unique locating, supporting, and clamping
configuration (Chang, 1998). Setup plarming is closely related to fixture planning,
however fixture planning is not the major concern in this study. In this study:
Setup planning is to group the to-be-machined features of a part into definite
setups, sequence the setups, and select locating datum for each setup, so that the part can
be precisely and economically manufactured in accordance to the design specifications.
Setup planning is the critical bridge between general process planning and
detailed operation planning; it is also the intimate upstream of fixture planning. Setup
planning significantly affects the overall cost and quality of part machining. Setup
27
planning will just minimize the effect of manufacturing errors instead of controlling the
errors.
The purpose of setup is to locate and fix a part in a definite manner on a machine
tool so that machining can take place. The theoretically exact point, axis, or plane used
to locate the part is referred to as a setup datum. When setups and setup datums of a
process plan are not selected properly, a tolerance chart analysis might find that the
process plan cannot guarantee parts to be made within the specified tolerance. Hence,
tolerance information from part design should be taken into consideration in setup
planning (Zhang, Huang, and Mei, 1996).
Some contributions have been made in the area of setup planning.
Boerma and Kals (1988) proposed an approach that uses the tolerance
specification for setup and fixture selection. In this approach, the selection of setups
depends on the tolerance relation between features, the approach directions of the
features involved, and the number and directions of the machine tool axes. A tolerance
factor was developed to compare different tolerances.
Hayes and Wright (1989) used features interaction to constrain operation
sequence and setup sequence for setup planning.
Chang (1990) proposed a heuristic approach in which tool approach directions and
precedence constraints are considered for setup grouping and sequencing.
Zhang, Nee, and Ong (1995) proposed a hybrid approach for setup planning.
Various constrains on setup planning are identified and discussed. Rules and heuristics
are created based on these constraints to determine the bilateral precedence relation
between the machining process of features and tool approach direction. The criterion to
optimize setup candidates is minimum number of setups. Tolerance relations between
features are not mentioned.
Demey, Brussel, and Derache (1996) proposed a method to determine setups for
mechanical workpieces. The method specified the minimal number of setups needed to
machine the workpiece, the machine to be used for each setup, the orientation of the
workpiece with respect to the machine in each setup, the operation to be executed in each

28
setup, and the precedence relations between the setups. The method takes into account
tolerance and precedence relations between the features.
Kim, Oh, and Cho (1996) proposed a system that implemented machining
sequences of reference surfaces and features. For machining of reference surfaces, the
machining sequences of reference surfaces are determined by using decision tables,
which are drawn from relations of part dimension, degree of surface roughness, fixture
type and its capacity, cutting tool's capacity, and experienced planners' knowledge.
Sarma and Wright (1996) proposed a graphic-theoretic model of planning along
with hierarchical prioritization of the objectives in planning. Algorithms are presented for
selecting the access faces for features, minimizing the number of setups, minimizing the
number of tool changes, and merging setup plans with preferable sequencing routines.
Fixturing concerns are ignored during planing, and tolerance relations are not considered.
Champati, Lu, and Lin (1996) proposed the application of one technique in
Artificial Intelligence (Al), i.e., case-based reasoning, to automate the generation of
operation sequences for intersecting features. The following issues are particularly
considered: (1) Representation of operation sequencing cases; (2) Indexing and retrieval
of operation sequencing cases; (3) Adaptation of operation sequencing cases.
Ozturk (1997) proposed a method using feature dependency and production
know-how rules for setup sequencing. Feature dependency is based on dimensional
tolerances, geometrical tolerances, and surfaces quality requirements.
Chen, Zhang, and Nee (1998) reported an approach to setup planning of prismatic
parts using Hopfield neural net coupled with simulated annealing. The approach dealt
with setup planning in two stages, i.e., (1) Sequence all the features of a workpiece
according to geometrical and technological constraints; and (2) Identify setups from the
sequenced features. The algorithm is based on converting the feature-sequencing problem
to a constraint-based travelling salesman problem (TSP). Each feature represents a city.
Setup time due to setup and tool change is incorporated as the "distance" between cities.
Feature precedence and critical tolerance requirement are treated as constraints. The
Hopfield neural net approach for TSP, i.e., energy function is adopted to model this

29
constraint-based TSP where the feature precedence and critical tolerance constraints are
attached as additional penalty ftanctions.
Wu and Chang (1998) proposed an approach that used the tolerance specification
in a feature-based design system to generate setup plans with explicit datum elements.
The number of setups required and the estimated accuracy of the resultant dimensions
rank these setup plans. Position tolerance is analyzed before the setup planning. After
feasible setup plans have been determined, other geometrical tolerances that are affected
by locating surface are examined.
Zhang (1996), Huang (1995), and Mei (1996) have developed a graph-matrix
approach to setup plarming for rotational parts based on tolerance analysis.
Mei (1996) developed a rulebase and conducted a neural work approach of
experiment for tolerance analysis and setup datum selection. The work is based on
comparative study on tolerance analysis versus operational tolerance analysis, manually
operated machining versus numerical controlled (NC) machining, and designs datums
versus machining datums. Principle and rules are obtained to train a back-propagation
neural network to select datums and setups. The study intended to improve quality and
economy of manufacturing by minimizing the effect of the cause of variation of
machining process without controlling the cause itself The study dealt with rotational
parts rather than prismatic parts.
Huang (1995) developed a graph-matrix approach to generate practical setup
plans based on tolerance analysis. The problem of tolerance stackup in NC machining
was analyzed in terms of manufacturing error analysis. Guidelines for setup planning
were then developed based on the analysis. In the study, the design specification of a part
is represented as a graph. The problem of identifying the optimal setup plan is
transformed into a graph search problem. A setup-planning algorithm for rotational parts
was then developed and its efficiency and effectiveness evaluated. The same principle
was then applied to prismatic parts and resulted in a setup-planning algorithm for
prismatic parts.

30
 Huang (1998) also developed an automated setup planning approach for lathe
machining, where undirected graph theory is adopted and rotational parts are considered.
The problem of setup planning is formulated mathematically with physical constraints.

2.3.3 Fixture Planning

The function of a fixture is to restrain all the six degrees of freedom with
sufficient accuracy and firmness. In addition to determining the position of the
workpiece, the fixture must absorb the tool forces and transmit the necessary torque in
the case of turning operations. The workholding must not cause damage, or deformation
of the workpiece beyond permitted limits. In general, a fixture should (Nee, 1995):
1. Provide accurate and repeatable location of the datum surfaces of the
workpiece with respect to the axes of the machine tool;
2. Resisting motion, deflection, and distortion of the workpiece under the action
of the cutting tool.
In general, a workholding device serves three primary functions: locating,
clamping, and supporting. In addition to the three primary functions, fixtures may also
perform the operations of centralizing and guiding. Where appropriate, as found in
specialized fixtures called jigs, a special guiding system leads the tool to its precise
position relative to the workpiece (Chang, Wysk, and Wang, 1998, pp. 193).
A fixture is usually composed of (Nee, 1995):
1. Locators
A locator is usually a fixed part of a fixture, the purpose of which is to restrict
movement of the workpiece being fixtured.
2. Clamps
A clamp is a moveable part of a fixture, the purpose of which is to provide a
holding force.
3. Supports
A support is a fixed or moveable part of a fixture, the purpose of which is to
prevent workpiece deflection under the action of the imposed cutting forces or clamping
forces.

31
 4. Fixture body
The fixture body is a rigid structure, the purpose of which is to maintain the
corrected spatial relationship between locators, clamps, supports and the machine tool on
which the workpiece is to be processed.
Selection of locating features is influenced by:
1. Tolerance on design dimensions;
2. Form of the raw material;
3. Sequence of operations;
4. Orientation of the workpiece with respect to the machine-tool spindle axis;
5. Spatial relationship of the workpiece, the machine-tool structure and the
machining envelop;
6. Size and shape of available surfaces.
Clamping features must be chosen so that all forces imposed during machining
can be reacted to the machine tool bed through the location and support surfaces.
Selection of clamping features is influenced by:
1. Locating features;
2. Supporting features;
3. Magnitude and direction of machining forces;
4. Spatial relationship of the workpiece, the machine-tool structure an the
machining envelop;
5. Surface finished specified in the design specification;
6. Workpiece strength and stiffness;
7. Type and size of clamps.
Geometrical control is concerned with stability of the workpiece. The position of
the workpiece in the fixture is defined by a number of locators. For good geometrical
control, the workpiece must automatically come into contact with all locators in an
exactly repeatable way despite operator skill. Any rigid workpiece has six degrees of
freedom and twelve directions of motion. Locators stop movement in one direction only,
therefore for complete location exactly six locators are required. Holding forces prevents
motions in the opposite direction.

32
 When requirement of geometrical and dimensional control conflict, precedence
should be given dimensional control.
For example, in Figure 2-8, located in the way any change in workpiece diameter
will cause a change in the position of the horizontal centerline. The 0.1mm tolerance on
the cylindrical diameter means that the horizontal centerline is only located within
0.07mm. Clearly the design specification of 0.05 is impossible with this method of
location. The locators must straddle the centerline they locate. An acceptable alternative
system of locators is shown in Figure 2-9 (Nee, 1995).

20+0.10 50.05

.^/'Ns. .y>-

Figure 2-8. Incorrect location of horizontal centerline.

20+0.10 50.05

Figure 2-9. Correct location of horizontal centerline.

The typical fixture for the setup planning of prismatic part is vise. The traditional
machine vice for part gripping has been around since the inception of milling operations
and until recently had seen only modest development. So popular are machine vices of
one sort or another that it would be highly unusual to find a company involved in the
machining of prismatic parts without such workholding equipment. Their popularit)
stems from the fact that they are able to accommodate a range of parts, from small

33
intricate prismatic shapes to long, slender workpieces - if used in conjunction with an
identical vice (Smith, 1947, pp. 193).
The vise clamping module uses the following procedure to determine the fixturing
method (Chang, 1990):
1. Select from the workpiece boundary model all faces that are oriented
perpendicular to the main approach direction V of the setup {F i = 1. ..., n\
NiX V^O},Ni is the normal of face Fj.
2. Pair faces FPk = (Fj, Ff) with opposite surface normal (M = -A^,) and store
them as candidate clamping faces.
3. Delete from the candidate clamping face pair the ones that have face-
overlapping areaAk less than a certain acceptable value. Face overlapping area
is defined as the intersection of a face projected to the other face along the
normal direction.
4. Calculate a composite weight Wk for each face pair FPk. The composite
includes weights on overlapping area, inverse of the face distance, and
number of features likely to be machined on the auxiliary faces.
5. Select the face pair FP* which has the maximum weight Wk FP* ^^ W* =
max {Wk, k= 1, ... m).
6. If there is a through hole present in the feature set of the setup, calculate the
distance d from the feature to FP*. This distance d is used as the maximum
vise step size.
7. Query the data with the face pair data and d for the appropriate vise jaw.
8. Calculate offset of the workpiece for machining features on the auxiliary
faces.
Cutting force and stability analysis can also be performed. The potential
interference between the fixture and the features to be machined is checked.

34
 2.4 Graph Theory and Algorithms
2.4.1 Basic Concepts on Graph Theorv
A finite graph G = (V, E) consists of two sets: a finite set V of elements called
vertices and a finite set E of elements called edges. Each edge is identified with a pair of
vertices. If the edges of a graph G are identified with ordered pairs of vertices, then G is
called a directed graph. Otherwise G is called an undirected graph. A graph G with
weights associated with its edges is called a weighted graph (Thulasiraman and Swamy,
1992).
In a graph, the symbols Vj, V2, V3, ... are used to represent the vertices and e^, e^,
e^, ... are used to represent the edges. Hence V= {vj, V2, V3, ...} and E = {e^, e2, e^, ...}.
The vertices v, and Vj associated with an edge Cj are called the end vertices of e/. The
edge ei is then denoted as Ci = (v^, v). While the elements ofE are distinct, more than one
edge in E may have the same pair of end vertices. All edges having the same pair of end
vertices are called parallel edges. Further, the end vertices of an edge need not be
distinct. If Ci = (v^-, v^), then the edge Cj is called a self-loop at vertex v^. A graph is called

a simple graph if it has no parallel edges or self-loops. Consider a graph G = {V, E). G'
= (V, E')isa subgraph of G if V and E' are, respectively, subsets of Vand E such that an
edge (v^, Vy) is in F ' only if v^- and Vy are in V.
An edge is said to be incident on its end vertices. Two vertices are adjacent if
they are the end vertices of some edges. The number of edges incident on a vertex v, is
called the degree of the vertex, and is denoted by d{v^. For a directed graph, the in-
degree d'iy^ of v^ is the number of edges incident into v^ and the out-degree d ^(v,) is the
number of edges incident out of v^.
A walk in a graph G = (F, F) is a finite alternating sequence of vertices and edges
VQ, ^1, V], ^2' ^"^k-\^^h ^k beginning and ending with vertices such that v^.j and v,- are the
end vertices of the edge e^, \ < i < k. A walk is open if its end vertices are distinct;

otherwise it is closed.
A walk is a trail if all its edges are distinct. A trail is open if its end vertices are
distinct; otherwise, it is closed.
35
 An open trail is a path if all its vertices are distinct.
A closed trail is a circuit if all its vertices except the end vertices are distinct.
A graph is said to be acyclic if it has no circuits. A tree is a connected acyclic
graph.
A directed walk in a directed graph G = (F, F) is a finite alternating sequence of
vertices and edges VQ, VJ, V2, ..., v^ such that (v^.j , v^), 1 < z < A:, is an edge in G. A

directed walk is open if its end vertices are distinct; otherwise it is closed.
A walk is a directed trail if all its edges are distinct. A directed trail is open if its
end vertices are distinct; otherwise, it is closed.
An open directed trail is a directed path if all its vertices are distinct.
A closed trail is a directed circuit if all its vertices except the end vertices are
distinct.
A directed graph is said to be acyclic if it has no directed circuits. For example,
the directed graph in Figure 2-10 is acyclic.

Figure 2-10. An acyclic directed graph.

A vertex v in a directed graph G is a root of G if there are directed paths from v to

all the remaining vertices of G. A directed graph is a tree if the underlying undirected
graph is a tree. A directed graph G is a directed tree if G is a tree and has a root. A vertex
V in G is called a leaf if d^{v) = 0.

36
 Let Vi and V2 be two mutually disjoint subsets of Fsuch that F= Fi u V2; Then
the set S of all those edges of G having one end vertex in Vi and the other in Vj is called a
cut of G. This is usually denoted by <Vi, V2>.
Sometimes it is desirable to change an undirected graph into a directed one by the
process of duplication: to each edge of the graph, we replace it by a pair of edges with the
same endpoints but with opposite directions. For example, in Figure 2-11, the undirected
graph is changed into directed graph by duplication process (Chen, 1997).

Figure 2-11. The transformation of an undirected graph into a directed graph.

In many situations, it is natural to consider graphs in both directed and undirected

edges; they are referred to as mixed graphs (Chen, 1997). As an example, a city map with
both one-way and two-way streets may be represented by a mixed graph in which the
streets are edges and the street intersections the nodes. Figure 2-12 is an example of a
mixed graph.

Figure 2-12. A mixed graph.

37
2.4.2 Matrices of a Graph
A graph is completely determined by specifying either its adjacency structure or
its incidence structure. These specifications provide far more efficient wa\s of
representing a large or complicated graph than a pictorial representation. Because
computers are more adept at manipulating numbers than at recognizing pictures, it is
standard practice to communicate the specification of a graph to a computer in matrix
form (Foulds, 1992).

Let G = ( F ^ be a graph with no parallel edges. Let F = (vj, vj, ... , v}. The

adjacency matrix M = [m^j] of G is an n x n matrix with m^ defined as follows

(Thulasiraman and Swamy, 1992):
\ if(v,,Vj)eE;
otherwise.

Consider a graph G with n vertices and m edges and having no self-loops. The all-
vertex incidence matrix Ac = [ajj] of G has n rows, one for each vertex, and m columns,
one for each edge. The element ay of Ac is defined as follows:
G is directed
1, if the yth edge is indident on the zth vertex and oriented away from it;
a -1, if the /th edge is indident on the /th vertex and oriented toward it;
u
0, if the yth edge is not indident on the /th vertex.

G is undirected
1, if the yth edge is indident on the /th vertex;
a <
0, if the yth edge is not indident on the /th vertex.

The cut matrix Qc = [^y] of a graph G with m edges has m columns and as many
rows as the number of cuts in G. The element qtj of Qc is defined as follows:
G is directed

1, if the yth edge is in the /th cut and has same oriention with the cut:
q^j = ) - 1 , if the jfh. edge is in the /th cut and has opposite oriention with the cut:
0, if the yth edge is not in the /th cut.

G is undirected
38
 1, if the yth edge is in the /th cut;
^^J =
0, otherwise.

The circuit matrix Be = [by] of a graph G with m edges has m columns and as
many rows as the number of circuits in G. The element by of Be is defined as follows:
G is directed

1, if the yth edge is in the /th circuit and has same oriention with the circuit;
by =\-^, if the yth edge is in the /th circuit and has opposite oriention with the circuit;
0, if the yth edge is not in the /th circuit.
G is undirected
1, if the yth edge is in the /th circuit;
^ =
0, otherwise.

2.4.3 Graph Algorithms

Graphs arise in the study of practical problems. The first step in such studies is to
discover graph-theoretic properties of the problem under consideration that would help us
in the formulation of a method of solution to the problem. Usually solving a problem
involves analysis of a graph or testing a graph for some specified property. Graphs that
arise in real-life problems are usually very large and complicated. Analysis of such
graphs in an efficient manner, therefore, involves the design of efficient computer
algorithm (Thulasiraman and Swamy, 1992).
Some basic graph analysis algorithms are for the following topics (Foulds, 1992):
1. Tree search,
2. Connectivity,
3. Planarity,
4. Spanning trees,
5. Isomorphism, and
6. Fundamental cycles.
Some basic graph optimization algorithms are for the following problems:
1. Shortest paths,
2. Minimal spanning trees,

39
 3. Maximum weight planar subgraphs, and
4. Network flows.
Some intensively explored optimization algorithms are:
1. The greedy algorithm,
2. Coloring, and
3. Matchings.
The details of these algorithms can be referred from McHugh (1990),
Thulasiraman and Swamy (1992), Foulds (1992), Diestel (1997), and Bollobs (1998).

40
 CHAPTER III
TOLERANCE ANALYSIS

3.1 Basis of Tolerance Analysis

3.1.1 Tolerance Zone
The tolerance zone can be seen as limits of feature variation. The tolerance
analysis in this study is based on the analysis of tolerance zones. Chase et al. considered
geometric feature variations in tolerance analysis of mechanical assemblies (Chase,
1996). Henzold (1995) discussed all kinds of tolerance zones in the book. Those
tolerance zones can be summarized as some typical types as shown in Figure 3-1. The
size of tolerance zone is usually 10" ~ 10' of feature size, in the following figures, the
tolerance zone is exaggerated purposely for illustration, t represents tolerance value.

(a) One-dimension tolerance zone

f
It t
O^
(b) Two-dimension tolerance zone

Or,
T
(c) Three-dimension tolerance zone

Figure 3-1. Typical tolerance zones.

As shown in Figure 3-1, there are three types of typical tolerance zones:
(a) One-dimension tolerance zone.
(b) Two-dimension tolerance zones.

41
 (c) Three-dimension tolerance zones.
Dimensional tolerance zone belongs to type (a). Type (b) and type (c) apph' to
geometrical tolerance zones. In a Cartesian coordinate system, three-dimension tolerance
zones can be projected into two-dimension tolerance zones, and two-dimension zones
into one-dimension zones, as shown in Figure 3-2. Most tolerance zones are three-
dimension; however, tolerance chain and tolerance analysis are usually carried out in
two-dimensions or one-dimension.

4 Z

Figure 3-2. The projecting relation of tolerance zones.

3.1.2 Characteristic Length and Tolerance Factor

There are two methods of comparing tolerance tightness: one is using a tolerance
grade table, the other is calculating tolerance factors.
In some countries and some large manufacturing companies, the tightness of
tolerances is defined in grades according to experience and calculation. The tolerances
are graded according to tolerance value, tolerance type, part shape, and part size (Wang,
1983, pp. 119-121). Hence the relative tightness can be decided by indexing the tolerance
grade table.
The comparison of the significance of the tolerance is necessary for respecting
critical tolerance in setup planning. It is rather easy to compare tolerances of same t\ pc.

42
but different types of tolerances cannot be directly compared. Therefore, the values ha\ e
to be converted to non-type-specific values. For the purpose of comparison, a so-called
tolerance factor (T. F.) has been introduced. A tolerance is converted into a tolerance
factor by dividing the tolerance value by the representative length. This length depends
on the type of tolerance and the dimensions of the part. Depending on the type of
tolerance, positioning errors can be composed of rotation and translation errors. But the
errors caused by rotational misalignment are always dominant. Besides that, translation
errors can be compensated by the machine tool controller (Boerma et al., 1988). The
tolerance factor depends on the tolerance value and the characteristic length of the
toleranced feature. In Figure 3-3 (Demey, Brussel, and Derache, 1996), it is shown that
different toleranced features has different characteristic lengths, hence different tolerance
factors.

"Y
(a) characteristic length of a surface (c) characteristic length of a hole

L * // 0.02 A
B // 0.03 B

I
^-l.
TF = L2/0.03 TF = L,/0.02

(b) T. F. with the surface as toleranced feature (d) T. F. with the hole as toleranced feature

Figure 3-3. The relation of tolerance factor and toleranced feature

In this study, we induce tolerance factors of all kinds of tolerances, some

definitions are different from those of Boerma, Demey, and others. In fact, all these
tolerance factors are approximate representations of tolerance tightness, the only purpose
of the tolerance factor concept is to compare the significance of different types of
tolerances. Generally, the tolerance factor is defined as:

43
 t
(3-1)
'~~L

where

Tj: tolerance factor,

t^: tolerance value,
L: characteristic length of the toleranced feature (Table 3-1).

Table 3-1. Characteristic length of toleranced features.

Toleranced feature Characteristic length

The length of the line
C/2 segment.
Curve line
/
The length of the line
segment.
Straight line
The length of the diagonal
line,
L-{1 +w )
O Rectangular face

y 0d The length of the diameter,

L =d
Circular face
I
0^ The longest distance of two
points on the face,
L =D
Irregular face
The length of the diagonal
on
o 1 line of the axial cut section,
L^il^ + cF f^
Cylindrical face
The length of the diagonal
13 0di L -)&d2
h- line of the axial cut section,
L = [{dx-f).5d2f+h^]^'^
cone face

Toleranced features can be lines (including axis) or surfaces. The characteristic

length of a toleranced feature characterizes the relative size of it. The characteristic length
of some typical features is listed in Table 3-1.
44
 The Tf of a dimensional tolerance is the dividing of the tolerance value by the
value of basic dimension. However the 7/ s of geometrical tolerances are more diverse.
The definitions of the 7/s of geometrical tolerances of Table 2-1, Table 2-2, and Table 2-
3 are listed in Table 3-2, Table 3-3, and Table 3-4, respectively.
Generally, for the feature of line, rectangle, circle, and cylinder:

r, = - ^ ^ (3-2)

l=\

where,
7}: tolerance factor;
/: the number of parameters of the feature;
//: length, width, or diameter of the feature (equal to 0 if absent);
t^: tolerance value;
n: the number of component dimensions.

3.1.3 Manufacturing Errors Classification

Product definition must be based upon functional requirements. Function is the
most important thing to consider in tolerancing a component. On the other hand,
manufacturing processes always induce errors, however small, which affect the
dimensions and other key parameters (Liggett, 1993, pp.1). Whybrew and Britton
summarized the sources of errors in a machining process as follows (Zhang, 997, pp. 20):
1. Machine
Clearance between moving parts of slideways and bearings
Geometrical errors in slideways, bearings, and leadscrews
Dynamic stiffness
Resolution of the measuring system
Resolution of the positioning system

45
 Table 3-2. Geometrical tolerance zones and tolerance factors.

Drawing indication Tolerance zone Tolerance factor

Tf=tll
t: tolerance value
/: the nominal length of
the line

o t Tf^tld

0Ti- t: tolerance value

d: the nominal diameter

r^^ ^
Tf^tll
t: tolerance value
/: the nominal Length
I
Tf=tll
SC75t
t: tolerance value
/: the nominal Length of
the longest line
segment on the surface

ii r/=//(/^ + ^ ) 1/2
t: tolerance value
/: the nominal Length
d: the nominal diameter

i Tf=t/il^ + w^y^^
< \ 4 W H
t: tolerance value
/: the nominal Length
w: the nominal width

46
 Table 3-2. (Continued).
Drawing indication Tolerance zone Tolerance factor

z t A Tf-tK.i^ + w-f^
\A/ .^
t: tolerance value
/: the nominal Length
w: the nominal width

Tf=tl{\^ + w^f^
i;^ t: tolerance value
/: the nominal Length
-^ 1 t A
w: the nominal width

Tf=tl{i^ + w^f'^
1 t: tolerance value
/: the nominal Length
w: the nominal width

^1 t |A |B |c i--1^ Tf^tl{l^ + cf-)Ml

0t
I I
t: tolerance value
tn
f-JcS^- /: the nominal Length
A^
^ ^ ^ d: the nominal diameter
@ 0t A
i 0t Tf=t/{l^ + cf) 1/2

0-d - - /: tolerance value

/: the nominal Length
d: the nominal diameter

/ /I 7 > = r / ( / ^ + w^)'^^
A
i
' -y-Ar
-^4^ /: tolerance value
, 1 Z^' ^ - ^ ^ -
/: the nominal Length
^ ^ 0.02 A w: the nominal width

47
 Table 3-2. (Continued).
Drawing indication Tolerance zone Tolerance factor
/ t A B
Tf=tl d

0'd t: tolerance value

B d: the nominal diameter

^ t A-B
^r Tf=tl{f- + c?) Ml
i k k

izrd t: tolerance value

' '
A 1r I3 1: the nominal Length
d: the nominal diameter

Tf=,ld
/" t A /?\ \
0d L. t: tolerance value
d: the nominal diameter
\L d

ik
ik
< ^^ t A Tf=tl d
"0cl t: tolerance value
^
y\ d: the nominal diameter

^ c
/' t A
^ <
i k Tf^2t/{di + d2)
^
^d 0do
^
t: tolerance value
di, di: the nominal
c-c diameters

^ t I A~| Coaxial Tf=t/[(d,-0.5d^y

zones , 2 il/2
+ W ]
0-d- izsd-)
t: tolerance value
4 d], d2: the nominal
W
Axial cut-section diameters
\v: the nominal width

48
 Thermal stability
2. Cutting Tool
Tool wear
Variation of tool size and cutting geometry
Rigidity of the tool and support
Thermal stability
3. Fixture
Variation between duplicate fixtures
Variation in location
Wear and contamination of locating surfaces
Deflection of locators and fixture
Thermal stability
4. Workpiece
Variation in physical and chemical properties
Variation in workpiece size
Rigidity of workpiece
Thermal stability
Stress relaxation
Coolant
Variation of flow
Variation of temperature
Contamination
Degradation
6. Operator
Variations are particularly apt to occur if the finished size is under the
direct control of the operator
7. Environmental conditions
Changes in temperature affect the machine, fixture and tool geometry, and
hysteresis in moving parts (e.g., slide ways and bearings)

49
 8. Process variable
Changes in process variables, such as feed and depth of cut, have a direct
effect on workpiece size and geometrical variation
Each aspect of the above sources deserves specific study in precision
manufacturing. The errors can be classified in two groups: those that are random,
unpredictable, and cannot be controlled, and those that are constant, time dependent or
capable of being controlled. Two examples of the first group are the effects of hysteresis
and random variations of the chip-forming process. The second group includes the effects
of tool and fixture wear over a period of time which can be measured and compensated
for, or predicted by a tool wear management system (Torvinen, 1995) In this study, we
want to explore the influence of setup methods on machining accuracy. Hence all these
sources should be classified in a way easy for the analysis.
To investigate the manufacturing errors involved in the geometrical relations and
dimensional relations obtained using different setup methods, Huang and Zhang (1996)
have made the following assumptions:
1. The part to be machined is located and clamped on the machine table using a
fixture;
2. The fixture is perfectly set up on the machine table;
3. The cutting tool is perfectly accurate and the effect of tool wear is trivial so
that it can be ignored;
4. The effect of workpiece deformation due to clamping force, cutting force,
gravity, and internal stress is trivial so that can be ignored.
According to the above assumptions, Huang and Zhang classified the errors
affecting the accuracy of dimensions into setup error and machine motion error The
assumptions and classification are valid in the manufacturing with ordinary precision.
However in higher precision manufacturing, the errors mentioned in the assumptions may
not be ignored. Just because these errors cannot be ignored, the significance of
improving the accuracy of a machined part through better setup planning is much more
apparent. Good setup planning always cost less than improving the accuracy of hardware.

50
 In this study, all types of error sources are classified according to their influences
on the geometrical positions of locating features and machining features of the on-line
part. Hence, there are two types of errors that are directly related to the accuracy of part
(illustrated in Figure 3-4):
Locating error: the variation between the position of actual datum feature and the
position of ideal datum. After a workpiece has been located and clamped, the setup error
remains as a constant unless the workpiece is removed from the fixture. Therefore, a
locating error is a deterministic error within each setup.
Machining error: the variation between the position of an actual machining
feature and the position of ideal machining feature.

locating machining
error error

cutter
locator
>

actual
dimension

Figure 3-4. Locating error and machining error.

Both locating error and machining error consist of a number of constant errors
and random errors. The tolerance of locating is usually 20 ~ 50 percent of the part
tolerance. This is necessary to maintain the required precision (Luggen, 1991).
Constant errors are added algebraically, while random errors are added
arithmetically. A resultant error can be calculated by the following formula:

A=I,<i',+jI(/5,,)' (3-3)
/=1

where.

51
 A, resultant error,

at, i= 1,2,3, ...,m : weights of constant error components, with signs,

0i, i = 1,2,3, ...,m: constant error components,
^/, / = 1,2, 3, ..., ^2: weights of random error components,
0i, i=\,2,3, ...,n: random error components.
The value of Pi depends on the distribution status of the random error component
and its geometrical relation with the resultant error.
It needs a lot of work to figure the weights and error components in formula (3-3).
However, the exploration of specific sources of the locating error and machining error is
not the focus of this study.

3.1.4 Setup Methods

In NC machining, three setup methods are used so that the geometrical and
dimensional relations between two features are obtained (Zhang, Huang, and Mei, 1996):
1) Setup method /: machining the two features in the same setup;

2) Setup method H: using one feature as setup datum and machine the other;

3) Setup method lU: using intermediate setup datums to machine the two features

in different setups.

The geometrical variations and dimensional variations obtained using different

setup methods are affected by different manufacturing error sources and should be treated
differently.

3.2 ToleranceA^ariation Stackup Analysis

It is rather complicated to consider dimensional tolerance and geometrical
tolerance simultaneously in variation and tolerance analysis. In accordance with the
principle of independency, dimensional tolerance and geometrical tolerances are anaKzed
separately in this section.

52
3.2.1 Dimensional ToleranceA^ariation Stackup Analysis
As shovm in Figure 3-1 (a), the tolerance zone of a dimension is strictly in one-
dimension, hence the formulation of dimensional tolerance stackup is relativeh'
straightforward. Suppose in a space, the relation of a resultant dimension d with its
component dimensions is as following,

d = f{x^,x^,...,Xi,y^,y^,...,y^,z,,z,,...,z^) (3-4)
where,
d: resultant dimension,
X,, / = 1, 2, 3, ..., /: component dimensions in X coordinate,
yj ,j= 1, 2, 3, ...,m: component dimensions in Y coordinate,
Zk,k= 1, 2, 3, ...,n: component dimensions in Z coordinate.
Theoretically, in worst case:

9/ Ax.. 9/ ^y, df
Az,
+ 1 (3-5)
/=1 Sx:. /=1 9F, A:=l dz,

where
Ad: variation of resultant dimension,
Ax Ayj, Azk: variations of component dimensions.
In statistical case:

I f df V -^
Ad =
(=1
Ax, +z ^y. + 1 -^A2, (3-6)
k=\ \^^k J
J

In the following text, both worst case and statistical case are dealt with. Worst
case and statistical case can deduce similar conclusions in qualitative analysis.
For example, 3 holes are to be drilled in a plane with their dimensional relation
shown in Figure 3-5. The horizontal dimensions are omitted for simplifying analysis.

53
 Figure 3-5. Dimensional relation of 3 holes in a plane.

The machining procedure and machining requirements are:

Step 1. Use face A as the machining datum and drill hole 1. The vertical
dimension from hole 1 to face A is a;
Step 2. Use face A and hole 1 as machining datum and drill hole 2. The vertical
dimension from hole 2 to face A is b, the angle from horizontal line to the connecting line
of hole 1 and hole 2 is ^ ;
Step 3. Use face A as machining datum and drill hole 3. The vertical dimension
from face A to hole 3 is c''.
Dimension c and c^are resultant dimensions.
For c'', there exists a dimension chain as shown in Figure 3-6.

Figure 3-6. Dimension chain of c'.

54
 c'=b'-a (3-7)

In worst case.

dd dc'
Ac'= Aa + Ab'
da db' (3-8)
= Aa + Ab'

In statistical case,

AC'^VAO^TA^ (3-9)
The dimension chain of c Ms one-dimension, which is what ^'dimension chain"
used to mean. In one-dimension case, the variation stackup is independent from basic
values of component dimensions.
For c, there exists a dimension chain as shown in Figure 3-7.

Figure 3-7. Dimension chain of c.

b-a (3-10)
c -
sin^

In worst case.

dc dc dc
Ac = Aa-\- Ab + AO
da db dO (3-11)
1 1 {b-a) cos 9
Aa + Ab + AS
sinO sinO sin^6'

In statistical case.

55
 Ac =
r 1 .^ ^ (b-a) cos 6
Aa + ' A*^ + AO (3-12)
sin^ sin^ sin^O

If the 0 specification is replaced by a horizontal dimension d as shown in Figure

3-8, then

Figure 3-8. Dimensional relation of 3 holes in a plane.

c = [d- +{b-ayy (3-13)

dc dc dc
Ac = Aa + Ab +
Ad
da db dd
(3-14)
(b - a)iAa + Ab) + dAd
4d^ +{b-aY

The dimension chain of c is in 2-dimension. As shown in equation (3-12) and

equation (3-14), 2-dimension variation stackup is usually not only dependent on
component variations but also dependent on basic values of component dimensions.
Dimensional tolerance chain is usually in one-dimension when dimensional
tolerances in X, Y, or Z directions of Cartesian coordinate system do not interact w ith
each other, as shown in Figure 3-9.

56
 Zi Az

Figure 3-9. Dimensional tolerance stackup.

In worst case, tolerances in Figure 3-9 are always added up.

In X direction:

X = Xj-{X2-X3) (3-15)

Ax= Axi + Ax2+ Ax3 (3-16)

In 7 direction:

Y=Yi-Y2 (3-17)

Ay = A y; + A Y2 (3-18)

In Z direction:

Z=Zi + Z2 + Z3 (3-19)

Az=Azi + Az2+Az3 (3-20)

57
3.2.2 Geometrical Tolerance/Variation Stackup Analysis

3.2,2,1 One-Dimension Geometrical Tolerance Stackup Analysis

One-dimension geometrical tolerance stackup applies to the situation that
component tolerance types are the same and the basic dimensions do not affect tolerance
stackup. As an example, a part with 5 identical parallel slots is shown in Figure 3-10.
Faces A, B, C, D, and E are setup datum for machining faces B, C, D, E, and F
respectively. The following notations are used for error stackup analysis:
TM, M = A, B, C,...: parallelism of locating face relative to ideal vertical face,
also called locating error,
XM, M = A, B, C, ...: parallelism of machining face relative to ideal vertical face,
also called machining error,
TMN^ M,N=A,B,C,...: parallelism between face M and face N.

Machining
face

Operational
datum

// 1 TAFI M
Design datum // I TAEI A|

Dimensional / Parallelism error

// TAB A zone relative to face A

Figure 3-10. One-dimension geometrical tolerance stackup analysis.

In Figure 3-10, the dimensional tolerance stackup and parallel tolerance stackup
can be shown in the follows,

TAB='^A+^B (3-21)

58
 r.^ =
AC AB -F-TD
B + AC
= T,+Tj,+A+A
B B ^C (3-22)
B C
I

i=A j^B

^^
AD" - ^AC '^'^C ~^^D

= Tj-\-Tn+Tr^
A B C + Jir,+
B An
"C
+X
'D (3-23)
C D

i=A j=B

Similarly,

D E
(3-24)
i=A j=B

E F
TAF = I^/+ S^; (3-25)
i=A J=B

Apparently, in the one-dimension case, the resultant tolerance is always equal to

the sum of component tolerances. The cases of one-dimension geometrical tolerance
stackups are relatively few in practical machining. The tolerance zone distributions of
some typical cases for one-dimension geometrical tolerance stackup are shown in Figure
3-11. The case of Figure 3-10 belongs to case (a) in Figure 3-11.

(a) (b) (c)

Figure 3-11. Tolerance zone distributions for one-dimension tolerance stackup.

59
3.2.2.2 Two-Dimension Geometrical Tolerance Stackup Analysis
For the part shovm in Figure 3-12, suppose the machining precedence of all the
faces is: D => B => C => E => F => A. The machining datums of face B, C, E, F, and A
are D, B, C, E, and F respectively. Assume the machining errors and locating errors are
all within the tolerances, therefore in the following text we substitute the error values
with corresponding tolerance values for error stackup analysis.

/AZ /
/I
/ 1
!__F_!
rbj
^
V F

Front View 3-D View

a ID
1 II tpE 1 K1

1 tEcl q
A L, w

Top View Left View

Figure 3-12. A part for 3-dimension tolerance stackup analysis.

The tolerance graph (which will be discussed in detail in Chapter IV) for the part
is shown in Figure 3-13. The resultant tolerances are illustrated in dot lines, component
tolerances in arrow lines.
Figure 3-14 shows the two-dimension view of the tolerance zone of face B. The
size of tolerance zone is usually 10 - 1 0 " of part size. In the following figures, the
tolerance zone is exaggerated for illustration.
From Figure 3-14, the tolerance zone indicates two possible maximum
movements of the part: horizontally translation of AB and rotation of 6. Suppose using
face B as machining datum for machining face D and face C, and the variation zone of
60
face B is equal to the tolerance zone of face B. The variation of face D is equal to the
translation of face B (as shown in Figure 3-15):

AD = AB (3-26)

I ICBTB
1 U^ID

rtiTi:
r-wTF

Figure 3-13. Tolerance graph of the part.

D Tolerance zone of B
B

- ^ A B ^

Figure 3-14. Two-dimension tolerance zone of face B of the part.

-*\ A o k - -^ A B ^

Figure 3-15. The effect of the translation of face B.

61
 The translation of face B has no effect on the variation of face C. The effect of the
variation of face B on face C is through the rotation of 6, as shown in Figure 3-16.

Figure 3-16. The effect of the rotation of face B.

From Figure 3-16,

^B ^ Ac
tgO (3-27)
L.

A,-^A B (3-28)

The following tolerance stackup analysis is based on the above variation stackup
analysis.
From the tolerance graph in Figure 3-13, tcD has two components as tsD and tcB-
Face B is parallel to face D, according to Formula (3-26), the tolerance stackup of face D
on face B is tBD^ to- Face C is perpendicular to face B, according to Formula (3-26), the
tolerance stackup of face B on face C is tcB- According to Formula (3-28), the tolerance
stackup on face C is

''CD ~ J VBD "*" ^Z) / "*" ^CB (3-29)

In same way.

62
 ^3
/'ED = t +/
J ''CD ^ 'EC

~'1~\^T~\^BD'^''D)'^'CB\'^^EC (3-30)

1 / \ 3
~ ~Y~ \^BD + * D / "^ ~T~ ^BC "^ ^C
^2 ^1

'FD ''ED "^ ^Ff

~I^(f ^i ^u.:^/ ^. _._. (3-31)

VBD "*" ^D / "^ ^ ^CS "*" 'EC "^ ^
^2 ^2

^2
/ = t +t
'AD J 'FD ^ 'AF
2 r 1 / \ 3 T
~" ~j~ ^~j~ ^ BD "*" ^Z) i "^ " ^ ' Cfi "*" * C "*" ^FE J "^ ' / i f

L. . . L. . , (3-32)
~ J \'BD "^ ^D / "*" ^CB "*" , \'EC "*" ^F ) "^ ^AF
^3 ^3

~ " ^ V BZ3 + *^D / "*" " ^ y'EC "*" '^Ff ^ "*" 'AF "*" ^CB

3.2.2.3 Three-Dimension Geometrical Tolerance Stackup Analysis

For three-dimension geometrical tolerance analysis, kinematics analysis of a rigid
body is helpfiil.
Assume the coordinate system OXYZ translates to O'X'Y'Z', the coordinate of O'
in OXYZ is (a, b, c). The position of the fixed point P in Cartesian coordinate system
OXYZ is (x, y, z), the position of point P in Cartesian coordinate system O'X'Y'Z' is (x".
y', z), as shown in Figure 3-18. Then the position of point P in coordinate system
O'X'Y'Z' is

63
 Z'

Figure 3-17. The translation of coordinate system.

X = x'+a
y = y+b (3-33)
z = z^+c

or

^x^ ^x'^ ^a^

y y + (3-34)
\X J K^J v^y

Assume when OXYZ rotates to OX'Y'Z', the rotational angles of OX', OY'. OZ'
relative to OXYZ are:
OX OY OZ
OX' ai Pi Yi
OY' a2 P2 Y2
OZ' as Ps 73

Then the relation of the coordinates in the two coordinate systems is

64
 X = x' COS | + y COS ^2 + z' COS a^
y = x'cos /?! + y cos y^2 + ^'cos /?3
(3-35)
z = x' cos ;ri + >^' COS ;^2 + ^' cos ^3

or
{
^X^ cosofj cosa2 cosa3 \f ^y\

y C0S/?i C0Sy^2 C0Sy^3 y (3-36)

\^j ^COS;KI C0S/2 c o s ; ^ 3 ^ \^j

The variation of the coordinates is

^x-x'^ (
cosa^-l cos 0^2 cosQr3 ^x'^

y-y COS/?i C0Sy^2~l C0S/?3 y (3-37)

^z-z J COS;KI C0SX2 C O S / 3 - I J v^y

Z A
I ^'
73 /

Figure 3-18. The rotation of coordinate system.

Basic transformation of a rigid body includes translation and rotation.

Mathematically, the transformation is represented in matrix form, T. Any point P in a 3-D
Cartesian coordinate frame (OXYZ) is defined by its homogenous coordinates [x, v. r. I].

65
To transform the point into a new point P' with coordinates [x', y', z', 1] in OXYZ fi-ame
(Chang, 1998),

P'=PT (3-38)

where T is a 4x4 matrix.

To transform a point by [a, b, c] on X, Y, and Z axes, denote the translation

matrix as Tt,

~1 0 0 0'
0 1 0 0
7. =
0 0 1 0 (3-39)
a b c \

In homogeneous coordinates, the 4 x 4 transformation matrix for rotating (named

rotation matrix) about the X axis with angle a can be written as

1 0 0 0
0 COS a sin a 0
T
X 0 -sin a cos a 0 (3-40)
0 0 0 1

In homogeneous coordinates, the 4 x 4 transformation matrix for rotating about

the Y axis with angle p can be written as

COSy^ 0 -sin p 0
0 1 0 0
h= siuy^ 0 cos p 0 (3-41)
0 0 0 1

In homogeneous coordinates, the 4 x 4 transformation matrix for rotating about

the Z axis with angle y can be written as

66
 cos;^ sin;' 0 0
-sin;' cos;' 0 0
^z = 0 0 10 (3-42)
0 0 0 1

In setup planing we usually select plane faces or cylindrical faces (either convex
cylindrical faces or concave cylindrical faces) as locating faces. The tolerance zone of a
plane face is usually as (a) or (b) in Figure 3-19. The tolerance zone of a cylindrical face
is usually as (c) or (d) in Figure 3-19.

0d t
(Ci3L__j [__I^
l^_____^^
T
(a) (b)
> Y

On
(0) (d)

Figure 3-19. Tolerance zones of locating faces

The transformation matrixes of Figure 3-19 (a) are:

1 0 0 0"
1 0 0
r. = 00 0 1 0
0 0 t 1_

67
 1 0 0 0
0 cos{arctg ) sin(arc/g ) 0
Tx = b b
0 - sin{arctg ) cos(arctg ) 0
b b
0
0 0 1

t
cos(arctg ) 0 - sin(arctg ) 0
a a
0 1 0 0
Ty = t t
sm{arctg ) 0 cos{arctg ) 0
a a
0 0 0 1

1 0 0 0
0 1 0 0
T^ = (3-43)
0 0 1 0
0 0 0 1

The transformation matrixes of Figure 3-19 (b) are:

1 0 0 0
0 1 0 0
^t- 0 0 1 0
0 0 t \_

1 0 0 0"
0 cos(arctg ) sin{arctg ) 0
d d
^x =
0 - sin{arctg ) cos{arctg ) 0
/7 d
0 () 0 1

68
 cos{arctg ) 0 - sin{arctg ) 0
d d
Ty = 0 1 0 0
t
sm{arctg ) 0 cos(arctg ) 0
d d
0 0 0 1

1 0 0 0
0 1 0 0
Tr, =
0 0 1 0 (3-44)
0 0 0 1

The transformation matrixes of Figure 3-19 (c) and Figure 3-19 (d) are:

1 0 0 0
0 1 0 0
T =
0 0 1 0
0 t t I

1 0 0 0
0 1 0 0
^X = 0 0 1 0
0 0 0 1

cos(arctg -) 0 - sin{arctg -) 0
0 1 0 0
h= t cos(arctg -) 0
sin(arctg -) 0
0 0 0 1

69
 K . t
cos{arctg-) sm{arctg-) 0 0
K . t
T.
-sm{arctg-) cos(arctg-) 0 0 (3-45)
0 0 1 0
0 0 0 1

For three-dimension tolerance stackup analysis, the space position relation of

tolerance zones determines which of the matrices are to be used.
For the above analysis, only the primary datum face is considered for tolerance
stackup analysis. In practical locating and clamping, the secondary datum face and
tertiary datum face should also be considered. The situation will be more complicated.
However, the primary datum face still plays a dominant role, and the analysis method can
be based on the above methods.

3.3 Tolerance Stackup Analysis for the Three Setup Methods

Figure 3-20 illustrates the manufacturing error analysis for a prismatic part
machined on a CNC milling center. Both geometrical tolerance and dimensional
tolerance are studied. Parallelism tolerance is used as the representation of geometrical
tolerance. Figure 3-20 (a) shows the machined part. For the purpose of tolerance analysis,
ti, t2, t-i. Ax, AY, and Az are use to represented geometrical variations and dimensional
variations (rather than tolerances); Figure 3-20 (b) shows the raw stock; Figure 3-20 (c)
shows the first setup and machining; Figure 3-20 (d) shows the second setup and
machining.
In the principle of tolerance independency, geometrical tolerances and
dimensional tolerances are inspected and satisfied respectively. Hence geometrical errors
and dimensional errors are analyzed separately in the following text.

70
 I // I tA ID'

// I fa IC'

Machine

Figure 3-20. Manufacturing error analysis for NC machining

(a) machined part, (b) raw stock (c) the first setup, (d) the second setup.

3.3.1 Dimensional Errors Analysis

The following notations are used for dimensional errors analysis (modified from
Huang, and Zhang, 1996):
i == 1,2, . . . : locating error of each setup (could be negative),
2 ^'

2 /'
J == 0, 1, 2,. ..: machining error of each machining face,

^7, 7 == 0, 1, 2,. ..: basic dimension for each machining,

A,B,. . . : faces of raw stock,
A',B\_ ' : machined faces,
LpQ: the length between surface P and Q.
Prior to machining, the operator will pick a location on the fixture to be the
program zero. In this example, the program zero is O'. When setting the program zero

71
0\ the cutting tool needs to be brought to O' and the distance between the machine
origin O and the program zero O' is recorded. Assume the recorded distance is XQ. due to
the machine motion error, the actual distance would be XQ + 0.5^. Here SQ is still
classified as machining error although there is no actual cutting.
Xis the length between machined surface C and D'. Surface C and D' are both
machined in the first setup in which surface A is used as a setup dattim. Therefore, X is
obtained using setup method I.

= (^2-^2)-(^i-^i) (3-46)

= ( X 2 - X , ) - ( ( ^ , +^2)

The dimension of Xis influenced only by two machining errors, the locating error
has no influence on it. Dimensions obtained using setup method I are denoted as type I
dimensions. Thus, type I dimensions are only influenced by two machining error
components.
In the second setup. A' and B' are machined using D' as setup datum. Y is the
length between machined surface A' and D'. Surface D' is the setup datum when
machining surface A'. Therefore, 7is obtained using setup method II.

Y=T

= (^4-^4)-(^o-^o)-A2 (3-47)

= (^4-^o)-(^0+^4+A2)

The dimension of Y is influenced by one locating error component and two

machining error components. Dimensions obtained using setup method II are denoted as
type II dimensions. Thus, type II dimensions are influenced by one locating error
component and two machining error components.
Z is the length between machined surface B' and C . Surface C is machined
using surface A as a setup datum in the first setup; while surface B' is machined using
72
surface D' as a setup datum in the second setup. Therefore, Z is obtained using setup
method III.

~ -^D'B' " ^C'D'

= [(x3^^3)-(xl<yo)iAJ-[(x,-x,)^(<5,+<?,)] (3-48)

The dimension of Z is influenced by one locating error component and four

machining error components. Dimensions obtained using setup method III are denoted as
type III dimensions. Thus, type III dimensions are influenced by one locating error
component and four machining error components.

3.3.2 Geometrical Errors Analysis

The reference datum of a geometrical error is different from that of a dimensional
error, hence the geometrical errors analysis is different from dimensional error analysis.
For the purpose of illustration, parallelism error is used as the representation of
geometrical error.
The following notations are used for geometrical errors analysis:
TM, M = A, B, C,...: parallelism of a locating face relative to an ideal vertical
face, also called a locating error;
XM, A/= A, B, C, ...: parallelism of machining face relative to ideal vertical face,
also called machining error;
TMN, M, N = A,B,C, parallelism between face Mand face A^.
Surface C ' and D' are both machined in the first setup in which surface A is used
as a setup datum. Therefore, Tcv'^s obtained using setup method I,

TC'D'=XC-+XD- (3-49)

The parallelism of face C and face D' is influenced only by two machining
errors, the locating error has no influence on it. Geometrical errors obtained using setup

73
method I are denoted as type I geometrical errors. Thus, type I geometrical errors are
only influenced by two machining error components.
In the second setup. A' and B' are machined using D' as setup datum. Surface D'
is the setup datum when machine surface A'. Therefore, the parallelism error between A'
and D' is obtained using setup method II.

TA-D' = XA'+ To- (3-50)

The parallelism of face A' and face D' is influenced by one machining error
component and one locating error component. Geometrical errors obtained using setup
method II are denoted as type II geometrical errors. Thus, type II geometrical errors are
influenced by one locating error component and one machining error component.
Surface C is machined using surface A as a setup datum in the first setup; while
surface B' is machined using surface D' as a setup datum in the second setup. Therefore,
the parallelism of face B' and face C' is obtained using setup method III,
TB'C'= TB'D'+ TD'C

^^B'+FD'+XC'+XD' (3-51)

The parallelism of face B^ and face C is influenced by one locating error

component and three machining error components. Geometrical errors obtained using
setup method III are denoted as type III geometrical errors. Thus, type III geometrical
errors are influenced by one locating error component and three machining error
components.

3.3.3 Comprehensive Analysis

Type II errors are obtained when the design datums are used as the setup datums.
They are errors obtained directly, hence no error chain calculations are needed for type II
dimensions. However, type I errors and type III errors are obtained when the design
datums do not coincide with the setup datums. Those are errors obtained indirectly.
According to traditional error stackup analysis (it is usually called tolerance stackup
analysis), they need to be calculated based on an error chain analysis. The error for each
component link within a chain will be added up without regard to which setup method is
74
used. However, the manufacturing error analysis showed that in NC machining type I
error is smaller than type III error. This fact implies that type I errors and type III errors
should be treated differently when applying error stackup analysis.
When using setup method I, the setup error is not included in the final error
obtained. The geometrical relationship of the features machined in the same setup mainly
depends on the geometry built into the machine tool. The dimensional relationship, such
as the distance between two parallel faces, is determined mainly by the accuracy of the
control unit, which is a built-in capability of the machine tool. The geometrical
relationship, such as the parallelism between two parallel faces, is determined mainly by
the static and dynamic accuracy of machine components, which is a built-in characteristic
of the machine tool. Unlike manual machining, no error stackup problem is encountered
for type I dimensions in NC machining. Errors obtained using setup method I consist of
the least manufacturing errors. Whenever possible, this setup method should be used to
facilitate tolerance control.
Setup method II is the most recommended method in the literature (Wilson and
Harvey, 1963; Buckingham, 1954; Doyle, 1953). In manual machining, after taking a
machining cut, the obtained dimension is measured to ensure that it is within the design
specification. The reference feature used for measurement is regarded as the measuring
datum. This is basically a trial-and-error method that is not observed in NC machining.
In NC machining, when setup method II is used, the locating error is included in the
errors obtained. To control the tolerance of the dimensions, the accuracy of setting up the
part has to be considered. Errors obtained using setup method II are usually larger than
those obtained using setup method I in NC machining in precision manufacturing.
However, setup method II is regarded as a good method when two features cannot be
machined in the same setup.
Setup method III is the least desired setup method. In this case, an error chain is
formed for the geometrical relafion obtained. As a result, the errors will stack up. When
the tolerance requirement is tight, this setup method should be avoided. If setup method
III is used to obtain a geometrical relation with tight tolerance, although the technique of
tolerance-chart analysis can be used to calculate the operational dimensions and

75
tolerances, the undesirable stackup of errors might make the manufacturing of the part
impossible even with the best machine tool.
The above discussion showed that error control can be achieved proactively via
setup planning. Features with tight tolerance relationships should be arranged into the
same setup whenever possible. When two features with tight tolerance relationships
cannot be machined in the same setup, they should be machined by setup method II.
Only when the tolerance stackup is within tolerance specification can setup III be used.

76
 CHAPTER IV
TOLERANCE GRAPH AND MATRIXES

4.1 Tolerance Graph

4.1.1 The Definition of "Tolerance Graph" and Its Related Concepts
In this section, some basic concepts of "tolerance graph" are defined according to
traditional graph theory (Thulasiraman and Swamy, 1993). Most concepts in this study
have explicit physical meanings and are not identical to the theoretic ones so that graph
theory can be effectively applied in setup planning.
In this study, by representing the set F of faces of a part as vertices and the set T
of tolerance relations among faces as edges, we define such a graph as a tolerance graph.
G^{F, r). In a tolerance graph with n vertices and m edges,

F=\fMflJx-Jn) (4-1)

T^{t,j,\i,j=\,2,...,n} (4-2)

where/i,/2, f,...fn represent the faces of a part, and t,j represents the tolerance relation
between f a c e / and facej^.
tij is denoted as

kj = (f.,f?> (4-3)

where, /,,; is the tolerance weight (always positive) between/ a n d / , / is the datum of/,
defined as

Characteristic length of toleranced feature (mm)

''' 1000 X tolerance value between / and /^ (mm)

Characteristic length designates the size of a toleranced feature. Tolerance weight

designates the relative tightness of a tolerance specification. The tolerance weights are
the weights of a tolerance graph.
According to the definition of tolerance factor in Chapter III, tolerance factor
TFij (always positive) between/ a n d / is defined as

77
 tolerance value between / and /
f = (4-5)
'-J Characteristic length of the toleranced feature

The tolerance weight of a tolerance graph is 10"^ of the reciprocal of the tolerance
factor,

1
t, i = (4-6)
'' lOOOxT, ^
' i,J

Because of the variety of tolerance types, a tolerance graph G may consist of

ordered pairs of vertices (i.e., tolerances with explicit datum), unordered pairs of vertices
(i.e., tolerances without explicit datum), isolated vertices (i.e., faces without tolerance
relation), and edges incident on identical vertices (i.e., diameter tolerances and form
tolerances). In this study, an undirected edge is treated as two directed edges with
opposite directions and same end vertices at certain stages. Hence the algorithms of
directed graphes can be used. In this sense, a tolerance graph is a directed graph in
essence, and the tolerance factors are the weights of the directed graph.
Because dimensional tolerances and form tolerances usually have no explicit
datum, they can be represented as either tij = (ft, f) or tjj = (f, / ) . Pictorially, tjj = (f, ,JJ)
can be represented as in Figure 4-1. In particular, self-loop edges (form tolerances and
diameter tolerances) are represented as

tu=(f.f.) (4-7)

f, ) ^-^ ( f
(b) t | , r ( f i , f j ) , ( f j , f | H .
(a) [. .={f., f.;)={f.,f .)=t..
t'u

^^^^-l^
(c) t|,r(^.^)

Figure 4-1. Pictorial representation of/,,;.

78
 For a tolerance graph G^(F,T),F^ { / i , / , / , - , / },T= {t,j,\ i,j=\,2,...,n }.,
the adjacency matrix M = [w/^ j] ofGisannxn matrix with W/,y defined as follows:

^U' if t^j = ( / , / . ) , i.e. the zth face is the datum of the yth
face in the tolrance relation;
^u =< (4-8)
0, if there is no tolerance relation between the zth face
and the yth face.

Apparently, for a tolerance relation without explicit datum,

rni,j^kj = (fi,fi) = (fj>f) = tj,i. (4-9)

Because the tolerance graph includes both directed and undirected edges, the
adjacency matrix M = [mj^ j] of G is nonsymmetrical.

The number of edges incident on a vertex/ is called the degree of the vertex, and
is denoted by d(f). The in-degree d'{f^ off is the number of edges incident into/, the
out-degree d^(f) is the number of edges incident out off. An undirected edge is treated
as one incident-in edge and one incident-out edge in calculating the number of degree.
Hence,

d(/d = d'if) + d-if). (4-10)

Figure 4-2. Two equivalent graphs in calculating vertex degree.

For example, in calculating the degree of vertex, the graph in Figure 4-2(a) can be
treated as the graph in Figure 4-2(b), hence
d(fi)-d'{f) + d-if)-2+\-3,

79
 d(fi)-d'{f2) + d\f2)=\ + \=2,
d(f,)-d'(f) + d-{f)=l+2-3.
Because an edge connecting face z and face j is denoted as //, y = (/J . / ) in a
tolerance graph, it is inconvenient to represent circuits in a tolerance graph by 2-
dimension matrix as usual. In this study, the circuits in a tolerance graph are represented
as a 3-dimension circuit. The circuit matrix C = [ck,i,i] of G with n vertices and m circuits
is an m x 77 X Z7 matrix with cj^jj defined as follows:

1, if edge /, j is on and has same orientation with the Ath circuit;

k,i,j
-1, if edge t^ j is on and has opposite orientatiai with the Ath circuit; (4-11)
0, if edge r, j is not on the kth circuit.

A directed circuit consists of only edges with same orientation with the circuit.
For example, in the tolerance graph shown in Figure 4-3 (a), there are 4 directed circuits
as shown in Figure 4-3 (b) to Figure 4-3 (e).

(c)C, (d)C, (e)C,

Figure 4-3. Circuits in a tolerance graph.

The four circuits are shown in Table 4-1,

80
 Table 4-1. Circuits in the tolerance graph.

^1,2 ^2,3 ^3,4 ^1 ^4,2 ^4, 1 hA

Ci: 1 1 1 0 0 1 0
C2: 0 1 0 1 1 0 1
C3: 1 1 0 1 0 0 0
C4: 0 1 1 0 1 0 0

A directed path in a tolerance graph G = (F, T) is a finite sequence of distinct

vertices/i, 72, ,/, such that tt.]^, e T, i< / <k+i. A group of faces in a tolerance graph can
also be seen as a grand vertex in a directed path.
In this study, the adjacency matrix is used to represent the information of
tolerances in a part, and each tolerance graph corresponds to an adjacency matrix.
Directed circuits and paths are used for setup sequencing. In setup planning, vertices are
sometimes grouped according to some principles or relations. These grouped vertex sets
are also named as islands. In setup sequencing, an island may be treated as a vertex. One
task of setup planning is to find or build a directed path that connecting all the islands.

4.1.2 The Application of Tolerance Graph in This Study

Tolerance graph can be applied to tolerance charting mentioned in Chapter 111.
For example, assume the dimension chain is in the direction of one axis of the Cartesian
coordinate system. The procedure of dimensional tolerance charting is:
1. Represent faces as vertices / , / , basic dimension between two faces as
undirected edge difuf). Dimensional tolerance between/ a n d / is denoted as
tifi,f) (or transferred to the bilateral format).
2. Identify a circuit, denote it as Ck, In the right-hand rule, set the clockwise as
the direction of the circuit. Orient all the edges.
3. Identify the resultant edge and component edges.
4. If the direction of difufj) is in the positive direction of the coordinate system,
denote the value of d as positive value. If the direction of d{f,. fj) is in the
negative direction of the coordinate system, denote the value of J as negati\e
value.

81
 5. The basic value of the resultant dimension is

Y'^ajj) (4-12)
hj

where d{fi,fj) is resultant dimension with positive or negative symbols.

6. The tolerance of the resultant dimension is

+ Ya^UJj) (4-13)
h]

where t{fi,fj) is always positive.

Tolerance graph can also aid machining error stackup analysis for a part. For
example, for the specification of a part is shown in Figure 4-4,

^A,n ^tA,D

^B,C ^B,C 1 ^C,D 'C,D

B
VA c
/ / / / / / / D

Figure 4-4. Specifications of a part.

Denote
DM, N ' Basic dimension between face Mand face N;

- tu, N ' Bilateral tolerance between face M and face A^;

- dM, N ' Bilateral machining error between face M and face A^;

AM : Locating error by using face M as virtual datum feature;

SM : Machining error of the machining face M.
The setup and machining procedure of the part is as following:

82
 Setup 1: Machining face C and face D;
Setup 2: Using face D as datum, machining face B and face A.
The corresponding tolerance graph of the part is as Figure 4-5.
Face C and face D are machined in a same setup up, that is, they are machined in
setup method I, the error stackup is:

^CD =^C+^D (4-14)

Face D and face A are machined in two different setups, face D is the machining
datum of face A. They are machined in setup method II, the error stackup is:

Setup 1 Setup 2

Figure 4-5. Tolerance graph of the part.

^D^A =^D+^A
(4-15)

Face C and face B are machined in two different setups, neither face is the
machining datum of another face. Hence they are machined in setup method III, the error
stackup is:

(4-16)
= {Sc+S^) + {Aj,+S,)

Generally <i(.^ < J^^ < d^^' that is, setup method I has least error stackup,

while setup method III has largest error stackup.

83
 If in a third setup, face A is used as machining datum for machining an imagined
face E, as shovm in Figure 4-6. From the tolerance graph, the error stackup is.

"D, - dD,A ^dA,E

(4-17)
= (Az) + ^J-^(A,+^^)

Setup 1 Setup 2 Setup 3

Figure 4-6. Another tolerance graph of the part.

That is, the tolerance graph makes tolerance analysis easier.

However, in this study, the most significant application of tolerance graph is in
setup planning, which will be discussed in later on.

4.2 Matrices
In order to illustrate the matrices used in this study, a prismatic part with its
tolerance specifications is designed and shown in Figure 4-7.

4.2.1 Tolerance Matrix

For a tolerance graph G = (F, 7), F = { / I , / , / , . . . , / ].T= {t,,j.\ i,j^\.2 n }.
the tolerance matrix is defined as:

Wk=M,^[m^X (4-18)

where k denotes the ^ h form of the tolerance matrix; since a tolerance matrix will be
frequently transformed in a setup-planning algorithm, the subscript is necessar\ to denote

84
the different forms of it. M= [mj j] is the adjacency matrix of G with m,,j defined as in
equation (4-8).

f2 f4 f5 f6 f16 f3

unit: mm
100.01
00.05-f7
100.05-

0.02

i o

CO
O 02OO.O1
O

* 500.02 150.04
-800.04-

600.02

Or d
H
o

30

-f15-
30 0 - 0 0 . 0 4 - f 15- -f1
-f1-

Figure 4-7. Drawings of a prismatic part (modified from Huang, 1995)

85
 For example, the tolerance relations of the part are represented by the tolerance
matrix shown in Figure 4-8.

11 12 " 13 14 15 IB f7 18 19 110 111 112 113 114 115 116

' 1.56 0.561 1 j i 1 i i
f1
12
1 \
0.67
f3 0.67 1
""1 150 0.10i
15 4.24 0.50 1
16 ............. '" i
w 0.45 I
18 0.67 0.67 0.50 OST^ 1 0.17
19 0.17
110 ;1.25
111 1.251 0.19;

112 0.17
113 0.10
1 1 II 1.00
i 1
114 j 1 !0.19 1.00
1 115 i
5
10.56* J, ^ ^^^
i f16 1^ ' ... - 0.50 j

Figure 4-8. Initial tolerance matrix W\.

4.2.2 Tool Approach Direction (TAD) Matrix

For a tolerance graph G = (F, T),F= {f\,fi,f,...Jn },T= {tij\i,j^ 1,2, ...,Z7 },
the TAD matrix is defined as:

^k =K,;L (4-19)

where k denotes the Ath form of the TAD matrix; Uij is defined as:

1, if zth face and yth face can be machined in a same TAD;

Ut i = 1 (4-20)
' 0, if zth face and yth face can not be machined in a same TAD.

For example, for the setup planning of the part, assume the part is machined on a
3-axis milling center. In according to the coordinate system in Figure 4-7. there exist 6
TADs for machining the part. The six TADs are +X, -X, +Y, -Y, +Z, and -Z. The set of
faces that can be machined from +X, -X, +Y, -Y, +Z, and - Z TADs are denoted as .S'A-,

S-x, 5'+y, iS.K, 5+2, and 5.2 respectively. The face sets of the part are as follows:

86
 S-Z = {f\,fl,f,f,f5,f6],
S+z = {fe, fi, fs, f9,f 10, f\\,fi},
S+Y = {fu},
S.Y= {fn},
S-x - {/is},
S+x = {fie}-
The TAD matrix of the part is shown in Figure 4-9.

! f1 LR_ _13 14 15 16 17 1 18 19 flO f11 112 113 114 115 116

' i '

'i^'i 1 1 ; 1 1 1 J r v\ X 1 i __ [._ j _ _ .
' 12 11 1 1 1 1 1 1
13 11 1 1 1 1 1 1 i ! ; !

14 11 1' 1 1 1 1 j ' 1
15 11 1 1 1 1 1 1 I
16 M 1 1 1 1 /\/ 1 1 1 i "l f \ ]
U 1 1 1 1 1
18 1 1
19 1 1
110 1 1
f11 1 1
f12 " \ I
"1 1 1 1 1 ly'
113
114 J i ,
1
.
i
,
i i
"T"!
(
i f 1 i
f15
J
\^ }
f16 I i ! ^VJ

Figure 4-9. The TAD matrix U\.

4.2.3 Machining-Feature Matrix

For a tolerance graph G = (F, 7), F = ( / i , / , / , . . . , / ],T= {t,,j,\i,j= 1,2 n).
the machining-feature matrix is defined as:

(4-21)

where k denotes the k\h form of the machining-feature matrix; v,^ is defined as:

1, if zth face and yth face are in a same machining feature;

^'-.7 = (4-22
0, if zth face and /th face are not in a same machining feature.

87
 There are three muW-face features in the part shown in Figure 4-7, denoted as
follows.

Fl=\f2,f3},
p2-{fufj5],
P3^ {/9,/l0,/ll,/l2}.
The inifial TAD matrix of the part is shown in Figure 4-10.

f1 '"'W 13 14 15 16 n 18 19 flO 111 'f12 f13 114 115 f16

L J L Ji ^ 1 1 "' s . \

12 1 1 1 1 ^ I
""'13 1 1
L f4 1 1 1 1
15 1 1 "V 1

1 fr
18
f- .,
1 19 1 1 1
110 1 1 1 1 ! 1
111 1 1 1 1
112 1 "T^ 1 1
i
113
f14
115
116 [ ^vw.w* /

Figure 4-10. The machining-feature matrix Vi.

4.2.4 The Transfer and Operation on Matrices

In this study, the matrices are frequently transferred to explicitly illustrate their
physical meanings. For example, Vi can be equally transferred into V2 as shown in Figure
4-11.
Correspondingly, the initial tolerance matrix is transferred into IV2 shown in
Figure 4-12.
Some operations of the matrices in this study are unique because of the principles
of setup plarming. For example, in setup planning, the tolerance relations among faces in
a same machining feature are usually ignored. Accordingly, the intersection of the
machining-feature matrix and the tolerance matrix is defined as follows:

88
 W,nV,=lm,jX\v,^.-l\], (4-23)

n^fz 13 f1 14 15 16 f7 18 19 HO 111 f12 flS 114 115 , 116

12 '11 1 i ^
13 Ii 1j -p
f1
1 j 1 - <!
14 " 1
5 [ 1 1 1 -
16 .
f7
18
1 0 n~~T" ____ - j v
110 1 1 1 1
: 111 1 1 1 1
1 1
f12
: 113
U 1 ,'

1 f14
1 115
116 \ /

Figure 4-11. The machining-feature matrix V2.

12 13 ^ f1 14 15 16 U \ 18 19 110 111 112 113 114 115 116

12 /i I
1
1
0.67 , > \
1 13 '0.67 ' ^
<
r 1.56 0.56
^ 14 j1.50 0.10
15 , ^ ^ 4.2^' 0.50
16 .,.,.L,...__ i i 1
f7 0.45
18 0.67 0.67 1.50 *0.17 0.17
19
t
__}_

0.17r 1
1
: 110 ' 1 1.25 , 1^1 ..1^.^.^.J
1 fii 11.25^ 1
1
0.19
112^ 0.17 1 /
113 0.1 Oi 1.00
114 U J : (0.19 1.00
115 0.56 I ! 0.50
0.50 j
; 116 [

Figure 4-12. Tolerance matrix W2.

89
 By equation 4-23, the intersection of W2 and V2 is shown in figure 4-13.

oTlBTfi w 15 16 i f7 18 19 110 f1l' 112 113 114 115 116

12 0.67 ^ j -^ >
13
fj 1 1
1 0.67
L
11 1 0.56
14 1 0.10
15 1 0.50
16 -,.. b
f7 I 1
1
1
U45 i
i
1
i
18 0.67 0.67; 0.50 !ai7 0.17
s
19 i , .
t
1 ;
110 i 1 :
l
111 j i
i ! 0.19
112 0.17', i 1

113 0.10 1.00
114 0.19 1.00
'""*""'""""'
115 0.56 0.50
116
L,... 1 1
10.50 1

Figure 4-13. W2n V2.

Similarly, after each face is assigned to a definite setup, the tolerance relations
among faces in a same TAD can be ignored, that is, the intersection of tolerance matrix
and TAD matrix is:

W,nU,=[m^jX\u.j-l\], (4-24)

The union of TAD matrix and machining-feature matrix is defined as:

Uk^yk= Kj X ^ij + ("/,; - ^ij )l (4-25)

By equation 4-25, the transferred matrix of the union of W2 and VJ is shown in

Figure 4-14.
More transfers and operations on matrices in setup planning will be introduced in
Chapter V.

90
 12 13 f1 14 15 ^ iB i f7 B 19 flOi f11 i 112j 113; f14 f l S i f l S
r-^ /'I 1 1 .JLi_Li 1^^ s
! 5 1 ! ' \
13 1 1 1 i-^jT-ji^-i 1
11 1 1 1 1 1 1 1
1 14 1 1 1 1 1 1 J
15 1 j_J_ 1 1 1 T '1
1
16 \ l " ^ 1 1 1 1/ V 1 1 1 ^
1 _ . ^ ...^
17 1 1 1
18 1 TTT"'
ZtjL
19
110 .
n^ ! 1
1
1 ""
:] _.
111 1 1 1
112 : \1 1 I 1 ly^

113 4.I -j i- ~ :;D

^ 114
1 f15
ri.
^1 j
i f16 L l-jj

Figure 4-14. Transferred Ui'U Vi.

91
 CHAPTER V
ALGORITM DEVELOPMENT AND ILLUSTRATION

As stated in the Chapter I, the ultimate goal of this research is to develop a setup
planning system (an applicable prototype in this study) based on tolerance analysis.
Based on the study introduced in the last four chapters, this chapter introduces the
algorithm of a setup planning system and the illustration of the algorithm. The validation
of the algorithm and the implementation of the setup-planning prototype will be
discussed in Chapter VI.

5.1 Principles and Rules of Setup Planning

Setup planning is usually an art rather than a science, that is, expert experience is
critical for a good setup planning. To automate setup planning is in fact to minimize the
requirement of experience for a setup planner. Hence, mathematical representing and
incorporating experience principles and rules of setup planning in algorithm is the
essence of automating setup planning.
The following principles and rules are the basis for the development of a setup-
planning algorithm:
#01. Features with fight tolerance relafionship should be arranged into a same
setup whenever possible.
# 02. When two features with tight tolerance relationship cannot be machined in
the same setup, they should be machined by using one of them as the
machining datum of the other, and vice versa.
#03. If the tolerance stackup between two features is within the tolerance
specification, they can be machined in different setups to give priority to
other constraints.
#04. All faces in a same machining feature are usually machined in one
operation, hence the tolerance relations among the faces in a same
machining feature can be ignored in setup planning.

92
 # 05. The tolerances among the faces within a same tool approach direction
(TAD) set can be temporarily ignored (screened) before deciding definite
setups. They may be reconsidered when spilt or regroup TAD sets into
definite setups.
# 06. The principle of datum coincidence: using a same feature as operational
datum, setup datum, measuring datum, and assembly datum to achieve
highest product accuracy.
Notes for # 06:
A datum is a theoretically exact point, axis, or plane derived from the true
geometrical counterpart of a specified datum feature. A datum is the origin from which
the location or geometrical characteristics of features of a part are established.
Datums, based on their applications, can be classified into five categories (ANSI
Y14.5M, 1982):
1. Design datums, which are datums used on the part drawing for the purpose of
dimensioning.
2. Setup datums, which are datums used to locate and setup the part on the
workholder (fixture) or the machine table.
3. Operational datums, which are datums used to determine the way the
machining cuts are taken.
4. Measuring datums, which are datums used for inspection purpose after
machining.
5. Assembly datums, which are datums used to determine the position of a part in
an assembly or the position of an assembly on an assembly machine.
The principle of datum coincidence is a most recommended principle in
machining operations. However, it is developed on the characteristics of manual machine
tools. In CNC machine, the principle is not rigidly observed. For example, in Setup
method I, higher part accuracy may be obtained without having setup datum coincided
with design datum.
# 07. All features sharing the same TAD are initially grouped into one cluster
(the prototype of a setup).

93
 # 08. Minimize the number of clusters.
# 09. Maximize the size of a cluster.
# 10. Minimize tool change (unconsidered currently).
Notes for # 0 7 - # 1 0 :
Feature clustering is based on a commonality which can either be tool approach
directions or tools (Chang, 1990).
There are two levels in the sequence determination: the global level and the setup
level. The global level determines the final cluster and its sequence. The setup level
sequencing determines the feasibility of the features to be machined in the particular
setup. The operation sequence within the setup is also determined.
The global level sequencing is actually done before process selection and tool
selection. The following operations are conducted:
Step 1. Create clusters using tool approach direction as the commonality.
Step 2. Refine clusters using the precedence.
Step 3. Apply heuristic to remove the duplicate features from clusters. A feature
can appear in one cluster only. Try to minimize the number of clusters by maximizing the
size of a cluster.
The setup level sequencing is done after tool selection:
Step 1. Create clusters using tool as the commonality.
Step 2. Refine clusters using the precedence and operating sequence within the
cluster determined.
Step 3. Check the precedence between the clusters, if there is any conflict, split
the cluster and reorder the sequence between the clusters.
By creating clusters using tool as the commonality, the objective of minimizing
tool change can be achieved.
#11. Principle of process concentration: to concentrate as many processes as
possible on a machine within a setup while making process plans.
Notes for # 11:
The process planner should try to select a machine that has high producti\ it> and
multiple process functions, and may also try to concentrate as many processes as possible

94
on the high productivity machine. This strategy may reduce the requirements of
additional machines, the number of operators, the number of steps in the process, and the
route of material handling, as well as the manufacturing lead time. This strategy is called
the principle of processes concentration (Zhang, 1994, pp. 31).
In CIM, the "one-hit machining" is highly recommended on CNC machines,
especially machining centers. "One-hit machining" means to concentrate as many
operations and processes as possible on a setup or on a machine tool. "One-hit
machining" has several major economic benefits (Smith, 1993):
1. It reduces capital outiay on a second machine tool; it significantly cuts down
work-in-progress time;
2. It increases the range of component parts to be machined by the machining
center;
3. It improves the overall machine tool efficiency by increasing the productive
cutting time;
4. Higher quality parts results as the machining center completes the parts at one
setup on the machine;
5. It uses less floor space, and fewer machine tools are needed. Some auxiliary
components are adopted on CNC machine tools to facilitate "one-hit
machining," such as 4-axis twin turret, "driven tooling"' facility,
programmable "C-axis" headstock spindle, adjustable angled heads, four- and
five-axis rotary tables.
# 12. Minimize the number of setups.
Cost of setup comes from the time in handling the part, the waiting time, and the
non-operation on the machine. Large cost savings are possible if the number of setups is
reduced. If design tolerances are relaxed, the process planner can redesign the sequence
of operations to eliminate finishing operations or by producing dimensioned features
indirectly (i.e., increasing tolerance stacks) (Zhang, 1997, pp. 26).
#13. Two or more TAD sets can be treated as a combined TAD set if ihcy can
be machined at a same setup by rotating the fixture without unclamping
the workpiece.

95
 Notes for # 1 3 :
More setups of the workpiece in an operation cause more time and more setup
errors. For this reason, when a workpiece must be machined at different positions on
machine tool, a workholder often is employed. This can change the position of workpiece
without unclamping it (Wang and Li, 1991).
# 14. Milling the slot before drilling the bottom hole.
#15. Avoid machining a deep feature if it is possible by machining other
features first.
Notes for # 15:
For example, in machining two concentric holes with different depths, the hole
with the smaller depth and larger diameter should be drilled before the other. Otherwise,
deep-hole drilling will be required for the smaller-diameter hole.
#16. All rough processing should be done prior to finish processing.
# 17- All major processing should be machined prior to minor processing.
#18. All bright datum should be machined prior to ordinary features.
#19. A black datum is selected only when necessary; Usually a feature with
larger size is selected as black datum for the stability of locating and
clamping.
# 20. All bright datum should be machined prior to use.
Notes for # 1 6 - 2 0 :
A datum that has not been manufactured before is called a black datum. A datum
that has been manufactured before is called a bright datum (Zhang, 1994).
#21. The principle of datum unification: use the same bright datum to
manufacture as many features as possible within a setup.
# 22. Setups with fewer number and larger size of features are usualh
sequenced before those with larger number and smaller size of features if
there is no other precedence restraints.
# 23. Avoid drilling a hole on a slide face that is not perpendicular to the axis of
the hole if possible.

96
 The above principles and rules will be embedded in the setup-planning algorithm
(Appendix B). They could be further enriched during concrete implementation.

5.2 The Setup-Planning Model

The work of algorithm development is not as easy as "computerizing" experience
principles and rules. Rigid effort in tolerance analysis, principles and rules exploration
and summarization, and mathematical and logical reasoning are required for algorithm
development.
In this study: setup planning refers to group the to-be-machined features of a part
into definite setups, sequence the setups, and select locating datum for each setup, so that
the part can be precisely and economically manufactured in accordance to the design
specifications. That is, the tasks of setup planning fall into four aspects:
1. Extract initial information,
2. Group to-be-machined features into setups,
3. Sequence the setups, and
4. Select setup datums.
The four tasks can be further divided into 7 steps in the setup-planning model
shown in Figure 5-1:
1. Extract initial information. The initial information on dimensions and
tolerances of parts, machine tool type and tool approach directions, and fixture
type and general configuration is extracted from data file; the information is
represented in three matrices: tolerance matrix, machining-feature matrix, and
TAD matrix as introduced in Chapter IV.
2. Assign a definite TAD for each face. The general configuration of machine
tool and fixture, the priority of tighter tolerance relations, the principles of
feature clustering, and the principles of feature interaction are used as
constraints in this module.
3. Simplify tolerance relations. All faces in a same machining feature are usualU
machined in a same setup, hence the tolerance relations among the faces in a
same machining feature can be ignored in setup planning. The tolerances

97
 among the faces within a same tool approach direction (TAD) set can be
temporarily ignored (screened) before deciding definite setups. They may be
reconsidered when spift or regroup TAD sets into definite setups.

Data liles
Machine /
Parts Fixture
tool

Extract initial inlormation

Machining
Tolernace
feature TAD matrix
matrix
matrix

Principles of
TADs of machine tool
Assign a definite TAD feature clustering
(Tool Approach Direction)
lor each lace
Principles of
TADs of fixture
feature interaction

Simplify tolerance relations

Priority of tighter Screen tolerance relations
tolerance relations within a same TAD set

Delete tolerance relations Select machining Principles of process

within a machining feature datum lor each lace concentration

Principle of datum Principles of machining

coincidence precedence
Group laces into
definite setups
Principle of datum Priority of TADs with fewer
unification and larger faces

Sequence setups
Others
Priority of datum features
(further consideration)

Priority of Priority of
bright datums
Setup datums selecting black datums

Principles and rules Setup planning modules Principles and rules

Figure 5-1. The setup-planning model.

98
 4. Select machining datum for each face. The principle of datum coincidence,
the principle of datum unification, the priority of bright datums, the principles
of feature clustering, and the principles of feature interaction, the principles of
process concentration, and the principles of machining precedence are used as
constraints in this module.
5. Group faces into definite setups. The priority of tighter tolerance relations and
the principles of machining precedence are used as constraints in this module.
6. Sequence setups. The priority of datum features, the principles of machining
precedence, and the priority of TADs with fewer and larger faces are used as
constraints in this module.
7. Select setup datums. The priority of bright datums and the priority of black
datums are used as constraints in this module.

5.3 The Setup-Planning Algorithm

Apparently, this is no ready-made algorithm for setup planning. A setup-planning
algorithm should incorporate manufacturing principles and rules with mathematical
algorithms in graph theory. The setup planing will be decomposed into several sub-
algorithms. In the bound of physical constraints, the following topics or mathematical
algorithms in graph theory are helpful in the implementation of setup-planning algorithm:
1. Connectivity,
2. Tree search,
3. Circuit search,
4. Path search,
5. Network flows.
The mathematical algorithms on these topics will be developed as basic functional
components and used in certain steps of the setup planning system. In addition to these
topics, some small algorithms such as (1) edge deleting/screening, (2) edge orienting, and
(3) vertices grouping should also be developed as functions according to practical
conditions, which involve the configurafion of the machine tools and the fixtures. In this
study, it is assumed that the prismatic parts are machined on a 3-axis vertical milling

99
center with vice as fixture; the rotational parts are machined on CNC lathe with 3-jaw
chuck as fixture.
In according to Figure 5-1, the setup-planning algorithm is devevoped as follows:
Step 1. Input initial information.
1) Define a coordinate system for the part;
2) For prismatic parts, define six TADs (tool approach directions ) as +X, -X,
+Y, -Y, +Z, and -Z. For rotational parts, define 2 TADs as +X, -X;
3) Group faces that can be machined from +X, -X, +Y, -Y, +Z, and - Z TADs
into S+x, S-x, S+Y, S.y, S+z, and S.z respectively. For rotational parts, group
faces that can be machined from +X, -X TADs into S+x, S.x respectively.
4) Group faces that in same machining features as Fi, F2, FT,,...',

5) Represent the tolerance relations of the part by adjacency matrix M= [mi^ , j .

6) Input ^S+x, S-x, S+Y, S.Y , S+z, S.z , Fi , F2 , F3,..., [m,j].

Step 2. Assign each face to a definite TAD set.
1) Iff has more than one TAD, a n d / has no tolerance relation with other faces.
Assign/ to a TAD set that contains the largest amount of faces;
2) Iff has tolerance relation with only one face, and the face is in one of the
overlapping TAD set, assign/ to the TAD set that contains its adjacent face;
3) Iff has tolerance relations with more than 1 face, and all the faces are in one
of the overlapping TAD sets, assign / to the same TAD set with one of its
adjacent faces that has the tightest tolerance relation with/.
4) Assign/ to a TAD set according to good machining practices;
Step 3. Delete or screen interior tolerance relations.
1) For tp, g = (fp,fq), iffp =fq, ovfp e F , and fg e F ,n= 1,2,3, .... delete tp g\
2) For tp, g = ifp,fg). I f / e So , and/, e So, D ^ +X, -X, +Y, -Y, +Z, and -Z,
screen tp, g.
Step 4. Orient undirected edges into directed edges.
1) Orient undirected edges into directed edges according to process precedence
constraints;

100
 2) For f, if d^{f) = d '( / ) > 2, all of the undirected edges incident on / are
changed into directed edges w i t h / as initial vertex;
3) F o r / , if d^if) = d'{f) < 2, remain the undirected edges. Treat the undirected
edges as directed edges with either one direction.
Step 5. Group faces into definite setups.
1) For ^Di and 5-02 , Dl, D2 = +X, -X, +Y, -Y, +Z, -Z, if ^-DI and 5*02 are
connected by at least two one-way edges which are not uniform in directions
(that is, there is a directed circuit between the two islands), split out as less
vertices as possible as long as there are two islands connected by at least two
one-way edges which are not uniform in directions. Respect tight tolerance
relation and the unity of machining features;
2) Repeat step 1) as long as there are directed circuit among islands;
3) Regroup vertices as long as the number of islands can be reduced.
Step 6. Sequence setups.
1) For any two islands Sm, Sn, if there are only one-way edges connecting them,
the initial island is precedent to the end island;
2) For any two islands Sm, Sn, if there are only two-way edges connecting them,
sequencing islands by the principle that a setup must not have conflicting
datums;
3) Islands with less vertices and less total degrees precede;
4) Sequence all the islands in according to:
a) Datum features are machined before being used as datum;
b) Observe machining feature intersection;
c) Setups with less machining features are machined first;
d) Observe good machining practices.
Step 7. Select setup datum.
The initial vertices of the edges connecting two adjacent islands in a path should
be selected as the datum for the ending island.
Directed graph theory, tolerance analysis, and expert principles of setup planning
are incorporated in the setup-planning algorithm. Since each of the seven steps contains

101
many rules and principles and hence be complex, the algorithm is briefly illustrated step
by step with a case study in the next section. The detailed setup-planning algorithm for
both prismatic parts and rotational parts is listed in the Appendix.

5.4 Illustration
The setup planning of prismatic parts is a complex task. To simplify the task and
emphasize the "tolerance graph" solution, this paper assumes:
1. The prismatic parts are machined on a 3-axis vertical milling center, using
vice as fixture.
2. The rotational parts are machined on an NC lathe, using a 3-jaw chuck as
fixture.
3. All the operations of a feature are finished in one setup (that is, only finish
process is considered in this study).
4. All the features of the part are to be machined.
5. The raw stocks of prismatic parts are in box shape; the raw stocks of rational
parts are in cylinder shape.
6. The part is to be machined as precisely as possible without increasing the
precision requirement of the machine tool.

5.4.1 Case Study 1: Setup Planning of a Prismatic Part

For the convenience of illustrafion, the prismafic part shown in Figure 5-2 is used
as an example to illustrate the "tolerance graph" solution to the setup planning. The
design specifications are shown in Figure 5-3.

5.4.1.1 Input initial information

According to the coordinate system shown in Figure 6-1, there are 6 tool approach
direcfions (TADs) for machining the part. The six TADs are +X, -X, +Y, -Y, +Z, and -Z.
The set of faces that can be machined from +X, -X, +Y, -Y, +Z, and -Z TADs arc
denoted as S+x, S.x, S+Y , S.Y , S+z, and S. z, respectively.

102
 f1
^2

Figure 5-2. 3-D views of a prismatic part.

Units: mm

140 0.03

40 0.02
-80 0 .04-J
^
:^ ^ ^ ^

^ -
120 0.03
160 0.03
200 0.08
-^ \ \ \ . \ -SWv^

Figure 5-3. Design specifications of the prismatic part.

103
 Slope faces such a s / g and/19 are usually machined by rotating tool holder or
fixture in NC machine. In the principle of the smallest rotation of tool holder or fixture,
/18 and/i9 can be assigned to either S+Y or S+z.
Feature is the general term applied to a physical portion of a part, such as a
surface, pin, tab, hole, or slot. A machining feature is a combination of faces that can and
should be machined in one operafion in terms of manufacturing convention. Of course, a
feature may consist of only one face. To be clear, we name a feature consisting of only
one feature as face in this study. The machining features can be extracted fi-om feature-
based designs.
The matrix of machining features is shown in Figure 5-4.

11 12 13 ^ 14 15 ^ 16 17 ^ 18 19 ^ 110 I l l 112 113 114 115 116 117 f18 119

11 [ A
12
_.. .

13
L

14
15
16 1 ! 1 1 1 1 1
17 1 ! 1 1 1 1 1
J

18
19
110 1 1 1 1 1
1 -~ 1 r " r " r _.j_ _ _
111 1 1
112 1 1
r r" " "1
113
1 1 1 1
114 1 1 1j
115 i_ 1 1 1 1 L.1._._1..-.1.i
1
116
117
118
119 ! J

Figure 5-4. The matrix of machining features.

There are five multi-face machining features in the part shown in Figure 5-2.
which are denoted as follows,
^l={/2,/3},
Pl={f6jl},

104
 F^={f,,f9),
^ 4 = {/ll ,/l2 },
^ 5 = {/l3 ,/l4,/l5 }
Since/ a n d / are designated as a same face in tolerance specifications, they are
treated as a machining feature. So a r e / i and/2- AH the faces of a machining feature are
usually machined in a same setup.
The face sets of the part are as follows:
S.x = {f,fi,f]-{f,F,],
S+x={fi,f,f]={F,,fA),
S-Y = { / , / ,/? , / , / ,/lO } = { / ,F2,F3 ,/io },
S+Y = {/lO,/ll ,/l2,/l3 ,/l4,/l5 ,/l8,/l9 } = {/lO , F4 , Fs,fn,f\9],
S.Z = { / 8 , / , / l 6 } ,
5'+Z = {/l7 ,/l8,/l9}-
The TAD matrix of the part is shown in Figure 5-5.

11 13 f4 15 16 17 18 19 110 111 112 113 114 f15 f16 117 118 119
11 ( 1 1
12 1 1
13 1 1
14 1
15 1
16 1
17 1
18 1
19 1
110 1 1 1 1 1 1 1 1
111 1 1 1 1 1 1 1 1
112 1 1 1 1 1 1 1 1
113 1 1 1 1 1 1 1 1
114 1 1 1 1 1 1 1 1
115 1 1 1 1 1 1 1 1
116 1
117 1
118 1 1 1 1 1 1 1 1
119 V 1 1 1 1 1 1 1 1 i

Figure 5-5. The TAD matrix of the part.

105
 By switching the relative positions of rows and columns, the TAD matrix can be
transferred into an equal matrix as shown in Figure 5-6, in which each TAD face set and
machining features are explicitly shown.

.12.. _13, 14 116 18 19 15 16 17 110 111 112 113 114 115 118 119 117
11 fi 1 _1_
12 1 1 1 1 11
13 1 1 .iJ 1
14 1 1 1;
116 1
18 1
19 .i_J
15
16
17
110
111 u
112
113
114
115
118 11
119 ^j
117 V V]
Note: -- TAD lace set - Machining feature

Figure 5-6. The transferred TAD matrix of the part.

The tolerance relations of the part are represented by the tolerance matrix shown
in Figure 5-7. The machining-feature matrix, TAD matrix, and initial tolerance matrix
can be combined as a comprehensive initial tolerance matrix as shown in Figure 5-8.
Every tolerance matrix corresponds to a tolerance graph, and vice versa. The
initial tolerance matrix can be pictorially represented by the tolerance graph shown in
Figure 5-9.

106
 11 12 13 14 15 16 17 18 19 110 111 112 113 114 115 116 117 118 119
f
11 ^ 0 4 5.7 3 0 4.5 0 0 0 0 0 0 0 0 0 0 0 0 2\
12 0 5.7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
13 0 0 5.7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
14 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
15 0 4 4 0 0 0 0 0 0 0 2.5 2.5 0 0 0 0 0 0 0
16 0 0 0 0 0 4 0 0 0 24 0 0 0 0 0 0 0 0 0
f7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4.7 0 0
19 0 0 0 0 0 0 0 0 0 0 5.3 5.3 0 0 0 0 0 0 0
110 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0
111 0 0 0 0 2.5 0 0 0 5.3 0 0 0 0 4 0 0 0 2 2
112 0 0 0 0 2.5 0 0 0 5.3 0 0 0 0 4 0 0 0 2 2
113 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0
114 0 0 0 0 0 0 0 0 0 0 4 4 0 0 0 0 0 0 0
115 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0
116 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2.7 0 0
117 0 4 4 0 0 4.5 0 4.7 0 0 0 0 0 0 0 2.7 31 2.5 2.1
118 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0
119 I 2 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 2

Figure 5-7. Initial tolerance matrix.

11 12 13 14 116 18 19 15 16 f7 110 111 112 113 114 115 118 119 117
11 _4___5.7_^_3_ 4.5 A
I 5.7 I
12 I
1
13 I[ 5.7;
14 3 >
'
116 2.7
18 4.7
19 5.3 5.3
15 2.5 2.5
16 24
f7
110 A-'
111 5.3 2.5 4 2 2
112 5.3 2.5 4 2 2
113
114
115
118 2
119 2 .2-'
117 V 2.7 4.7 4.5 '.2.5 2.1 zyj

Figure 5-8. Initial tolerance matrix with comprehensive information.

107
 Figure 5-9. Tolerance graph with initial information of the part.

5.4.1.2 Assign each face to a definite TAD set

In NC machining, three setup methods are used to obtain the dimension between
two faces within a part (Zhang, Huang, and Mei, 1996):
Setup method I: Machine the two features in the same setup;
Setup method U: Use one feature as the setup datum to machine the other, and
vice versa;
Setup method IIP Use an intermediate setup datum to machine the two features in
different setups.
To assure higher machining precision, we prefer setup method I, secondly setup
method II, setup method III is the last choice. Of course setup methods II and 111 may be
chosen if the tolerance stackup is within tolerance specifications, and the object is to
minimize the number of setups. This paper assumes the part is to be machined as
precisely as possible.

108
 From Figure 5-9, apparently, S.x n 5'+x= { / , / }, S.Y n ^.Z = { / , / }. 'S'.Y n 5'.Y
= {/lo },and S+Y (^ S+z = {/is ,/i9 } These faces need to be assigned to a definite TAD
set. The principles to assign a face/ to a definite face are:
Priority 1: Assign / to the TAD set in which there exists a feature having tight
tolerance relation w i t h / ;
Priority 2: Assign/ to the TAD set that contains maximum features;
Priority 3: Assign/ to a TAD so that good machining practices are observed.
Hence:
S.xnS+x^ {fi,f),
since ^1,2/3 ^^{fufifi,), an&feS.x ,
{/,/}-> S.x.

S.Yr^S.z={f8,f9),
denote Num(S) as the number of faces in a set 5",
since Num(S.Y) > Num (S.z),
{ / , / } ^ S.Y.

S.YnS+Y= {/lo},
since ^10,6 ={/6, f\o}, and fee S.Y ,
{ / l o } ^ '^-Y.

S+Y n 5+2 = { /18 > /l9 },

^17,18 ={/i7,/i8 }, andfneS+z,
however, ^ii,i8={/ii./i8 }, andf\\eS+Y,

since /ii,i8=2.0 < /ivjs^Vso^+2x40V (1000x0.04) = 2.45.

{/l8,/l9}^' S+Z-

The tolerance graph with each face in a definite TAD set is shown in Figure 5-10.
The matrix corresponding to the tolerance graph in Figure 5-10 is shown in Figure 5-11.

109
 Figure 5-10. Tolerance graph with each face in a definite TAD set.

11 12 13 14 116 18 19 15 16 f7 110 111 112 113 114 115 118 119 117
11 f(' 4 5.71 3 4.5 2 A
1
12 5.7 1

13 'v. .5j;.
1
14 3 ---
116 / s
2.7
\
18 4.7
19 5.3 5.3
15 4 4 2.5 2.5
16 4 24
f7
V
110 4-'
111 5.3 2.5 1 4 2 2
112 5.3 2.5 4 2 2
113 2
114 4 4
115 2
118 'v2 /
if
\
119 2 2 2
117 V 4 4 2.7 4.7 4.5 'v2.5 2,1 3^;^

Figure 5-11. Tolerance matrix with each face in a definite TAD set.

110
5.4.1.3 Delete or screen interior tolerance relations
We define tolerance relations within a TAD set as interior tolerance relations.
Obviously, self-loop tolerance relations and tolerance relations within faces of a feature
are all interior tolerance relations.
Since tolerance relations among faces within a same feature and self-loop
tolerances don't infiuence datum selection and setup sequencing, they are deleted at this
stage. However, since features with same tool approach direction are not necessarily
machined in a same setup, we temporarily screen the tolerance relations involving
features that can be machined from the same tool approach.
Let tp, g^ (fp, fq) be an edge of the tolerance graph. The edge tp, g is deleted if/ e
Fn and fg e F , n^ 1, 2, 3, ..., o r / = / ; Otherwise the edge tp g is screened out
(represented by dash lines) i f / e SD , ^nd fg e So , D = +X, -X, +Y, -Y, +Z, and -Z.
Hence, ^2/3,2/3, ^6,6, ^10,10, ^17,17, ^19,19, and /n, 15 in Figure 5-10 are deleted; ^1,2/3, ^10,6, ^14,
11/12 , ^17, 18 and t\i^ 19 in Figure 5-6 are screened. The tolerance graph with interior
tolerances deleted or screened is shown in Figure 5-12.

Figure 5-12. Tolerance graph with interior tolerances deleted or screened.

Ill
 J1___12____13_ 14 116 18 19 15 16 17 110 111 112 113 114 115 118 119 117
11 fi
12
13
14
116
18
19
15
16
17
110
111
112
113
114
115
118
119
117

Figure 5-13. Tolerance matrix with interior tolerances deleted or screened.

In Figure 5-13, all the matrix elements in the dot-line square area are replaced as
zero. Those tolerance relations in the filled square are permanently deleted.

5.4.1.4 Orient undirected edges into directed edges

In machining, every tolerance relation should have a machining datum (position
tolerances may have 2 or 3 datum. Therefore, some two-way edges should be oriented
into one-way edges if their common incident vertex is adjacent to more than 2 vertices in
two-way. Secondly, we hope as many features as possible can be machined without
changing datum. Thirdly, some features can only be generated after the machining of
other features, which should be selected as potential datum. According to these three
principles, we can orient some two-way edges in Figure 5-12 into one-way edges:
Since d'(f) = 4 > 2,
Orient ti^4 and ti^ 19 by selecting/ as datum.
Since ^ ( / y ) = 6 > 2,
Orient t\j^ 8 and ^17 15 by selecfing/17 as datum.

112
 Figure 5-14. Tolerance graph with undirected edges oriented.

J1___12____13_ 14 116 18 19 15 16 f7 110 111 112 113 114 115 118 119 117
11 fi 4.5 ^
12
13
14
116 0
18 0
19 0 0
15 4 4 0 0
16
17
110
111 5.3 2.5 N 2 2
112 5.3 2.5 2 2
113
114
115
118 .0.
119 0
117 4 4 2.7 4.7 4.5

Figure 5-15. Tolerance matrix with undirected edges oriented.

113
 Since ^ ( / i ) + d'ifn) = 5>2,F, = {fn,fn].
Oriental 1/12,5, ^^11/12,9, ^11/12,14, /^n, is,and ^n, 19 by selecting F4 as datum.
Now the tolerance graph evolves into a directed graph as shown in Figure 5-14.
The corresponding matrix of Figure 5-14 is shown in Figure 5-15.
By treafing faces in a same TAD set as a grand vertex, tolerance graph in Figure
5-14 can be simplified into the tolerance graph shown in Figure 5-16.

Figure 5-16. Simplified tolerance graph with undirected edges oriented.

5.4.1.5 Group faces into definite setups

Now, every face has a definite TAD. However, it is not necessary that all the
faces in one TAD be machined in one setup. Because of tolerance requirements and other
machining precedence, for example, a datum face should be machined before it is used as
datum, some TAD sets should be broken into smaller sets.
For example, if two islands are connected by more than two one-way edges which
are opposite in direction, at least one of the islands should be split into smaller islands
until the case no longer exists. Two circuits exist in Figure 5-16:
Denote circuit between S.x and .S-Y as C\: ^.x => ^1,6 => S.Y =^ /s. 2/3 => 5'.x ,
Denote circuit between ^.x and S+z as C2: ^.x => ^i, 19 =^ S+z =^ /17,2/3 => S-x

114
 Since ^.x e Ci, and ^.x e C2, break S-x into S.x^ = {fi,f) = {Fx} and ^-.x" = ( / } .
and there will be no circuit between any two setups. Simultaneously, the formerly
screened edge t\^2i3 is recovered.
Now the tolerance graph with definite setups is shown in Figure 5-17.

Figure 5-17. Tolerance graph with definite setups.

Figure 5-18. Simplified tolerance graph with definite setups.

115
 Tolerance graph in Figure 5-17 can be simplified into the tolerance graph shown
in Figure 5-18. There is no longer directed circuit in the tolerance graph. The
corresponding matrix of Figure 5-17 is shown in Figure 5-19.

11 12 13 14 116 18 19 15 16 f7 110 111 112 113 114 115 118 119 117
11 . 3 - ^ . 5 2 \
12
13 i I
14 0 ;
116 0
18 0
19 0 0
15 0 0
16
17
110
111 5.3 2.5 \ 2
112 5.3 2.5
113
114
115
118 .0.
119 0
117 2.7 4.7 4.5

Figure 5-19. Tolerance matrix with definite setups.

5.4.1.6 Sequence setups

At this stage, treat each definite setup as a grand vertex, "island", the task of
sequencing is to find a reasonable path that traverses all the islands. According to the
principles: (1) a face must be machined before it is used as datum, and (2) setups with
fewer faces or larger faces should be machined before those with more faces or smaller
faces, the setups can be sequenced now. We denote the symbol " ^> " as "precedent to'\
From Figure 5-19, three trees can be extracted as shown in Figure 5-20. Each tree
shows explicit precedence among islands as follows:

S+Y => 5'.Y , S+z ,

s.x => S.Y , s.z , s+z , s+x , s.x .

116
 Figure 5-20. The trees of setup precedence.

The Figure 5-20 can be redrawn as Figure 5-21,

Figure 5-21. The tolerance graph of setup precedence.

The setup precedence can also be shown in the matrix in Figure 5-22.
Now the sequencing task evolves into a simple scheduling problem, which is
solved as follows:
Since Num(S.x^) < Num(S+Y), S.xi => S+Y , designate ^.x^ as No. 1 setup.
Since Num{S+x) < Num(S+Y), S+x => S+Y , designate S+x as No. 2 setup.
Apparently, S+Y is No. 3 setup, and S+z is No. 4 setup.

117
 t 1 _ 12 13 14 116 18 19 15 16 17 110 111 112 113 114 115 118 119 117
11 /l__Q_! 9: A-fi o A
12
13 - ^ , *

14 l__Q_
116
18
19
15
16
17
110
111 5.342.5
112 5.3 2,5
113
114
115 V V
118
119
117 4 4 2.7 4.7 4.5 u
Figure 5-22. The sequencing of setups.

Since Num{S.z) < Num{S.x) < Num(S.Y), S.z => S.x^ => S.Y, designate ^-.z , .S^.x'.
and S.Y as No. 5, No. 6, and No. 7 setups respectively. Hence the sequence of setups is:
S-x => S+x => S+Y ^> S+z ^^ S.z ^> S.x => 'S'-Y.

The above procedure can be shown pictorially as Figure 5-23 and Figure 5-24.

Figure 5-23. The procedure of setup sequencing.

118
 setup 4

'^4 setup 3

Figure 5-24. Tolerance graph with the sequence of setups.

5.4.1.7 Select setup datum

The term setup datum here refers to faces that should be used as candidates for
locating faces for the ongoing setup and machining, they are not necessarily the final
locating faces. The selecting of complete locating faces is a comprehensive topic, which
must be explored in detail as another study. At this stage, datum selection is
straightforward. Generally, the initial vertices of the edges cormecting two adjacent
islands in a path should be selected as the datum for the end island. Hence,/ ( r a w ) , / , /
(raw), {/i,/ii,/i2,},/i7, {/i , / ,/i7} and {/i , F^ , fn) should be selected as datum
candidates for islands S.x , ^^^-x , S+Y , S+z , S.z , S.x\ and S.Y, respectively.
The entire setup plarming strategy is summarized in Table 5-1 and Figure 5-25.

5.4.2 Case Study 2: Setup Planning of a Rotational Part

The rotational part shown in Figure 5-26 is used as a second example to illustrate
the "tolerance graph" solution to the setup plarming. The design specifications are shown
in Figure 5-27.

119
 Table 5-1. The whole setup plarming strategy for the illustrated part.

Setup No. TAD set Machining Features Datum Candidates

1 s.x' /i /(raw)
2 S+x ./4 /i
3 S+Y / l l ,/l2 ,/l3 ,fl4,fl5 / (raw)
4 S+z Fn ,/l8 ,/l9 / l j./ll /l2
5 S.Z ./is /l7
6 S.x' /2 ,/3 / l , . / ,./l7
7 S.Y Fs,fe,fi ,fi ,f9 ,/io / i , F^ ,f\i

. (S

TAD

Figure 5-25. Machining process of the part.

Figure 5-26. 3-D view of a rotational part.

120
 0.01 f9

@ 0.C1 f9
CJ 0.02 200.01 -400.04-

-1100.05- unit: mm

Figure 5-27. Design specifications of the rotational part.

Because the dimensional tolerances in the axial direction of the rotational part
have no effect on the setup planning, they are omitted purposely in Figure 5-27.

5.4.2.1 Input initial information

According to the coordinate system shown in Figure 5-27, the set of faces that can
be machined from -X, +X TADs are denoted as S.x, S+x respectively.
There are three multi-face machining features in the part shown in Figure 5-27.
which are denoted as follows,
^l={/2 , / , / } ,
^2= {/,/},
^3= {/lO , / l l }

The matrix of machining features is shown in Figure 5-28. For easy view, the 0-
value elements of the following matrixes are usually left blank.

121
 f1 f2 f3 f4 f5 f6 f7 f8 f9 f10 f11 f12
f1 r J i_ _ L

f2 1 1 1
f3 1 1 1
f4 1

f5 L
"
1 1 ,
f6 1 1
f7 1 1
r~ ~~ 1r r - - - -r r *<
f8
f9
r - H
f10 1 1
f11
f12 V J

Figure 5-28. The matrix of machining features.

The face sets of the part are as follows:

S+x = {/ , / , / , / , / , / ,/lO ,/ll ,/l2 }= {/ , ^1 , / , / , ^3 ,/l2 }
S.X = { f i , / , / , / , / , / , / , / }= { ^ 1 , / , F2,/,/ },
The TAD matrix of the part is shown in Figure 5-29.

16 f7 18 0 f f12
fi r
f2 1
13 1
f4 1
15 1
f6
f7
f8
19 1
flO 1
f11 1
f12 V 1 i7

Figure 5-29. The TAD matrix of the part.

122
 By switching the relative positions of rows and columns, the TAD matrix can be
transferred into an equal matrix as shown in Figure 5-30, in which each TAD face set and
machining features are explicitly shown.

f6 f7 fS f2 fS f4 fS f9 f10 f11 f12 f1

f6 f/'^ 1 1
^
1
H
1
H
1
.1
1
.1 ^
1 ^ A
f7 ] 1 1 1
fS 11 1 1
f2 j 1 1 1 / ^ I -1 -I
1 <(
1 -1
1 <
1 1 1 1 1 ^
fS 11 1 1
f4 i 1 1 1
fS I 1 1 1
/
f9 M_____1___._1__, 11 11 11 iI i -^
j ^

flO
fii
f12
\
f1 V ^J. 1 1 1 1 1 1 1 .r

Figure 5-30. The transferred TAD matrix of the part.

The tolerance relations of the part are represented by the tolerance matrix shown
in Figure 5-31. The machining-feature matrix, TAD matrix, and initial tolerance matrix
can be combined as a comprehensive initial tolerance matrix as shown in Figure 5-32.

fi f2 fS f4 f5 16 f7 18 f9 flO fii f12

fl
f2
fS
f4
fS
f6
f7
fS 2.2
f9 6.5 8.5
f10
fii 0.75
f12 2.2 0.75 4.257

Figure 5-31. Initial tolerance matrix.

123
 f6 .f7___f8___f2____f3 _f4____f5____f9 flO fii f12 fi
f6 (7" 1 2 ^\ "A
f?
fS _2.2.
f2 1
fS
f4
fS
f9 6.5 8.5
flO
f11 0.75
f12 2.2 0.75 4.25
fi .yJ

Figure 5-32. Initial tolerance matrix with comprehensive information.

Every tolerance matrix corresponds to a tolerance graph, and vice versa. The
initial tolerance matrix can be pictorially represented by the tolerance graph shown in
Figure 5-33.

Figure 5-33. Tolerance Graph with initial information of the part.

5.4.2.2 Assign each face to a definite TAD set

From Figure 5-33, apparently.

124
 S+xr^S.x = {f2,f,fA,f,f9] = {F^,f,f},
ti,n = {/i7,/i8 }, and/i2 e 5'+x ,
however, ^4,5 = { / , / }, a n d / e S.x ,
since /'2,i2 = 1.0 < t^^ = 2.0, a n d / , / e Fi^
(/,/,/} ^ S.x.

^9,1 = {/9,/i2 }, and/12 e 5'+x ,

however, ^9,7= I / , / }, a n d / e 5.x ,
since ^9,12= 8.5 > t9j= 6.5,
{ / } ^ S+x.

since at this stage, Num{S.x) = 6 > A^ww (5'+x) = 5,

{ / } - > 5-x.

The tolerance graph with each face in a definite TAD set is shown in Figure 5-34.
The matrix corresponding to the tolerance graph in Figure 5-34 is shown in Figure
5-35.

Figure 5-34. Tolerance graph with each face in a definite TAD set.

125
 f6 n i8. f2 _f3 f4 f5 f9 f10 f11 f12 fi
f6 1
f7
f8 2.2
f2 1
fS
f4
f5
f9 6.5 8.5
f10
f11 0.75
f12 2.2 1 0.75 4.25
f1 JJ

Figure 5-35. Tolerance matrix with each face in a definite TAD set.

5.4.2.3 Delete or screen interior tolerance relations

Since tolerance relations among faces within a same feature and self-loop
tolerances don't influence datum selection and setup sequencing, they are deleted at this
stage. However, since features with same tool approach are not necessarily machined in a
same setup, we temporarily screen the tolerance relations involving features that can be
machined from the same tool approach direction.

Figure 5-36. Tolerance graph with interior tolerances deleted or screened.

126
 Hence, ^12,12, in Figure 5-34 is deleted; t^^ 12 , ^11,12 , ^4 ,6 - and ^5,8 in Figure 5-34
are screened. The tolerance graph with interior tolerances deleted or screened is shown in
Figure 5-36.

f6 f7 f8 f2 .f_3____f4_ f5 f9 flO f11 f12 fi

Figure 5-37. Tolerance matrix with interior tolerances deleted or screened.

In Figure 5-37, all the matrix elements in the dot-line square area are replaced as
zero. Those tolerance relations in the filled square and diagonal position are permanently
deleted.

5.4.2.4 Orient undirected edges into directed edges

In figure 5-36,
since / ( / 1 2 ) = 2 , / ( / ) = 1 , d^{f) = 1 ,
orient ^12,2 and /12,8 by selecting/12 as datum.
Now the tolerance graph evolves into a directed graph as shown in Figure 5-38.
The corresponding matrix of Figure 5-38 is show in Figure 5-39.
By treating faces in a same TAD set as a grand vertex, tolerance graph in Figure
5-39 can be simplified into the tolerance graph shown in Figure 5-40.

127
 Figure 5-38. Tolerance graph with undirected edges oriented.

f6 f7 f8 f2 f3 f4 f5 f9 flO fii f12 fi

f6 [/ A
f7
f8
f2
fS
f4
f5
f9
flO
fii
f12
fi JJ

Figure 5-39. Tolerance matrix with undirected edges oriented.

(5^~^
Figure 5-40. Simplified tolerance graph with undirected edges oriented.

128
5.4.2.5 Group faces into definite setups
Since there is no directed circuit in Figure 5-40, the two TAD sets are exactly the
two setups need to be sequenced.

5.4.2.6 Sequence setups

From Figure 5-40,
S+x => S.x .

5.4.2.7 Select setup datum

At this stage, datum selecfion is straightforward. Generally, the initial vertices of
the edges connecting two adjacent islands in a path should be selected as the datum for
the end island. H e n c e , / (raw), and/n should be selected as setup datum for islands .S+x .
and ^-x respectively.
Now the entire setup planning strategy is summarized in Table 5-2.

Table 5-2. The whole setup planning strategy for the illustrated part.

Setup No. TAD set Machining features Datum

1 S+x / , / , ./lO, / l l , / l 2 / (raw)
2 S.x fl, f, f, f, fe, fl, f f9,f\l

The machining process is shown in Figure 5-41.

TAD K
/

Setup 1 Setup 2

Figure 5-41. Machining process of the part.

129
 CHAPTER VI
IMPLEMENTATION AND VALIDATION

6.1 The GATO Setup Planner

In this research, the setup-planning algorithm for both prismatic parts and
rotational parts was implemented in C++. A graph-aided and tolerance-oriented (GATO)
setup-planning prototype was developed in Microsoft Visual C++ . A graphical user
interface (GUI) was designed for the interaction of the users and the setup-planning
algorithm. The structure of the interface is shown in Figure 6-1.

Help Information
Initial Selection
1. Tutorial
1. Select part 2. Warning message
2. Select machine tool 3. Running message
3. Select fixture 4. Running status

\ /
Matrix View
Part View 1. TAD matrix
1. First 3-D view / \ 2. IVIachining feature matrix
2. Second 3-D view 3. Tolerance matrix
3. Drawings view 4. Setup matrix
..u.ti\"::'iiiiL"c

Running Style
-/VE*
Report Process
1. Step by step 1. Generate report
2. All at once 2. Open, customize, save, or
3. Running controls print report

Figure 6-1. The GUI structure of the GATO setup planner.

The GUI has 6 basic functional blocks:

1. The inifial selecfion block. The user is asked to select the part, the machine
tool, and the fixture for setup planning.

130
 2. The part view block. The user can choose to view the 3-D views and the
specificafion drawings of the selected part.
3. The running style block. The user can choose to run the setup planning either
step by step or all at once.
4. The matrix view. The user can choose to view the TAD matrix, machining
matrix, tolerance matrix, and setup matrix during setup planning.
5. The report process block. The user can generate, open, customize, save, or
print the setup planning report.
6. The help information block. The GUI provides a help document for the user to
learn using the system, a static text box and a process status bar for the user to
monitor the setup planning process, and on-line fauh-proof message.
The initial user interface of the GATO setup planner is shown in Figure 6-2.

WkTiO. Se(u( f'iNtMier

file yiew Ryn Reoort Help

initial selection Matrix View

Select Pait:
r '. ' .-jrei'ij'pMs!!! (^ I '"'ancet-'ls'iiv . ,1
d
FaceSFace
_J ^^_\^_\_i_\_\_\^_\_\_\ _I_U -
Select Machine Tool: t p..-.-..-

1 d
Select fixture:

1 d
...-..

Part View - 1 i
(^ First 3-D View
u ^
f~ Second 3_D View Setup Matrix:

(~ S pecification D rawings View Setup\Face

_J_J^ J _ J ^ ^ _ J _ J J _ 1 _ J _ J ^ ^ ^ _ J ^ ^ -
Show the Part View

Running Style -

(^ Step By Step
^1 1 ^
C All Al Once
IRease select Part, Machine Tool, and Fixture before start the setup planning!

Start Setup Planning Exit

! g B S t a i l | Jl[j GATO Setup Planner | 8:00 PM

Figure 6-2. The GUI of the GATO setup planner.

131
 A brief user menu is included in Appendix C.
For the example of secfion 5.4.1 in Chapter V. selecting the inifial information as
shown in Figure 6-3, the setup planning report generated b\ the system is shown in
Figure 6-4. For the example of secfion 5.4.2 in Chapter V, selecting the initial
information as shown in Figure 6-5, the setup planning report generated by the s> stem is
shown in Figure 6-6. The automatically generated results are the same as the anahtic
results in section 5.4 of Chapter V.

mWSMwaPkmnet
ite yiew Ryn Recort Help
IIQ
Initial Selection Matrix View ^
C TAD Matrix C Machining Feature Matrix (^ Tolerance Matrix Show Matrix
Select Part:
Parti

Select Machine lool:

|3-Axis Vertical Milling Center ^

Select Fixture:
Vice

Part View -

(^ Fust 3-D View

C Second 3_D View

C Specification Drawings View

Show the Part View

Running Style

(^ Step By Step

r All At Once
jStep 1: Inpxjt initial information.
II
Start Setup Planning I Next Step I Ger'ig'ate ('epotf Exit

igQ S tail ^ GAT 0 S etup Plannei [^s[: i ^ r a ^ 10-.08PM

Figure 6-3. The inifial informafion of Parti.

132
 GATO Selup
Reporti - Notepad
M3 HH^
file View Ryn
FDe dit Search Help
Initial Selection THE SETUP PLftNNING REPORT OF PftRT-1
Show\^a^ux
Select Part: Part type: Prismatic
I Parti
Setup No. 1
Select Machine Machining faces: F1 ,
13-Axis Vertical
Candidate datuns: Select a large raw Face in the opposite TAD.
Setup No. 2
Select fixture: Machining faces: f^ ,
[Vice
Candidate datuns: f1 ,

Setup No. 3
Machining faces: f11, f12, f13, fUt, f15.
Candidate datuns: Select a large raw face in the opposite TOD.
Part View 20 :o
20 20
(^ First 3-D Vi Setup No. U
Machining faces: f17, f18, f19. Ld
C Second 3_ Candidate datuns: f1 , f11, f12.

C S pecificatir Setup No. 5 18 19 jg A

Machining faces: f16. aS
Candidate datuns: f17.
Show tf
Setup No. 6 1 1
Machining faces: f2 , f3 ,
Running Style
Candidate datuns: f1 , f5 , f17.
(^ Step By St.
Setup No. 7
C All At Once Machining faces: f5 , f6 , f7 , f8 , f9 , f10.
Candidate datuns: f1 , f11, f12, f17.

btatt betupt-'lanning lienerate Meport -UT #

install I ^ GATO Setup Planner | Tjy Microsoft Word - GATOJm... [ 1 ] RepoiH - Notepad ^ 4 ^ E 3 ( i J 10:18 PM

Figure 6-4. The setup planning report of Parti

133
File ifiew Ryn Refiort Help

iniuai aeieciion - Matrix View

Select Part:
|Part2
d
Select Machine l o o t
1 CNC Lathe
d
Select fixture:
13x1 aw Chuck
d
Part View

(f Rrst 3-D View

C Second 3_D View Setup Matrix:

C Specification Drawings View SetupVface 1 2
Sx 1
S+x 1 1
Show the Part View
^^^^H
Running Style

(^ Step By Step

r All At Once
jStep 1: Input initial information. III

Start Setup Planning Next Step Ge'iera^e Repoi* Exit

gQStail /tfr GATO Selup Planner ^ y Microsoft Word-G AT 0_ Reporti - Notepad { ^ i ; j [ , i ^ E B ^ ' 10:21 PM

Figure 6-5. The initial information of Part2.

134
L2:BiA10M9mi^Umm ^HS
file yiew Ryn Refiort Help

Initial Selection Matrix View

C TAD Matrix C Machining Feature Matrix (* Tolerance Matrix Show Matrix
Select Part:
FaceKFace 1 2 8 10 11 12
jPart2
"3
Select Machine Tool:
Repo[l2 - Notepad Rl
|Ct^C Lathe File dit S.eaich Help

Select Fixture:
THE SETUP PLANNING REPORT OF PflRT-2
|3\Jaw Chuck Part type: Rotational

Setup No. 1
Machining faces: f1 , f9 , f10, f11, f12.
Part View Candidate datuns: Select a large raw face in the opposite TftD.
(^ Fust 3-D View Setup No. 2
Machining faces: f2 , f3 , fit . f5 , f6 , f7 , f8 ,
^ Second 3_D View Candidate datuns: f9 , f12.
C Specification Drawing:
STir
Sx
Show the Part View

Running Style

(^ Step By Step

C All At Once
jStep 7: Select setup datum. The setup planning is successful. Congratulations!

Start Setup Planning Ue:'.' Step Generate Report Exit

iggStaitl (I^GATO Setup Planner | ^ y Microsoft V/ord GAT... | 1 ] Reporti -Notepad |V) Repoit2 - Notepad \M'^~'mL'^^ 10:25 PM

Figure 6-6. The setup planning report of Part2.

6.2 Validafion Discussion

It is difficult to evaluate a setup planning system because (1) they is no standard
criteria for the evaluation, (2) setup planning is only a part of CAPP which itself is hard
to evaluate. In industrial terms, automatic setup planning is still a complicated and under
development area. In this study, some assumptions on machine tool, fixture, and part are
set for the focus of the academic invesfigation. The GATO setup planner is just a setup-
planning prototype rather than an industrial product. Hence the setup-planning prototype
should be evaluated from an academic view rather than an industrial view.
In the context of setup planning, the process of determining the correctness of the
prototype typically consists of two separate parts: verification and validation. Verification

135
is the process of determining whether the prototype operates as intended. Through the
verificafion process, we try to find and remove errors in the implementation of the
prototype. The activity is commonly referred to as debugging the prototype.
As shown in the last section, for the illustration examples of Chapter V, the
automatic setup planning resuhs by the GATO setup planner are identical with those of
the analystic resuhs, which verifies that the algorithm has been correctly implemented in
the prototype.
Validation is the process of reaching an acceptable level of confidence that the
inferences drawn from the prototype are correct and applicable in practice.
Validafion of a setup planning is very hard because of the lack of standard
criteria. Experimentafion is one of the most acceptable validation methods in engineering.
Unfortunately, an experiment method is very expensive, even impossible for a setup
planning system because (1) at least hundreds of parts should be machined, (2) dozens of
constant conditions involving machine tools, machining parameters, environmental
conditions, and operators should be controlled, (3) precision measurement equipment is
required. Since the three requirements are impractical, simulation is taken as an
alternative method.
The tolerance/variation stackup analysis for alternative setup plans is the basic
approach for validation in this research. Since one of the most important objective of this
research is to control tolerance through setup planning, the purpose of simulation is to
show that the setup plan generated by the proposed prototype can reduce variation
stackup for critical dimensions and geometric shapes. The flowchart of the simulation
procedure is shown in Figure 6-7.
As mentioned in section 3.1.3, the manufacturing errors are classified into two
general classes in this research, that is, locating errors and machining errors. These errors
have their typical distributions. It is known from experience that the machining errors
possess distributions varying between a normal distribution and a rectangular distribution
(BjOrke, 1989). Both types of errors can be generated by a simulation program. The
variation stackup analysis for all the setup plans is then conducted. Statistical case
analysis will be executed. If the proposed prototype is \alid. the variation stackup of the

136
critical dimensions and geometric shapes in the generated setup plan should be the least
among all the setup plans. Otherwise, the prototype should be suspect and modified.

Part design

Setup planing prototype

Alternate setup plans

besides the generated plan
Setup plan generated by the
setup-planning prototype

Simulation of locating errors and

machining errors in NC machining

Modify Variation stackup analysis

for all the setup plans

No

Yes

No

Conclusion

Figure 6-7. The flowchart of the simulation procedure.

137
 For different parts or the same part with different tolerance specifications, the
setup plans will vary. Hence the simulation procedure may be repeated many times until
a stable conclusion can be drawn.
Take the dimensional variation stackup of the part in section 5.4.2 as an example,
the simplified drawing of the part is shown in Figure 6-8. Only the faces related to the
dimensional chain in the horizontal axis are marked in the figure.

f8

-X +x
<1- -O

150.02

200.01 -400.04-
unit: mm
1100.05-

Figure 6-8. The simplified drawing of Figure 5-27.

As illustrated in section 5.4.1,

{f, fA } e S'+x n S.x.
From Figure 6-6, the GATO setup planner assigns f2 and f^ to be machined in
S_j^. An alternative setup planning method to assigns/2 andf to S^y. In according to the
tolerance analysis method introduced in section 4.1.2, denote

- di,j: Bilateral dimensional variation between/ and/;

138
 Am : Locating error by using/^ as actual datum feature;
Sn : Machining error of the machining/.
The corresponding tolerance graph of Figure 6-8 is shown in Figure 6-9, which is
simplified from Figure 5-33. The corresponding tolerance graphs of the two setup plans
are shown in Figure 6-10.

Figure 6-9. The simplified tolerance graph of Figure 6-8.

Setup Plan 1 (generated) Setup Plan 2 (alternative)

Figure 6-10. The tolerance graphs of the two setup plans.

139
 To make the simulation valid, the following rules should be observed:
1. The sample number (part number) is large enough. This study assumes that
the part number is 1000;
2. The larger the size of a machining feature, the bigger its machining error.
According to the dimensional specifications shown in Figure 5-27, this study
assumes that

S2:5^:5e:S^:Sn ^12 = 10.0 : 15.0 : 15.0 : 65.0 : 45.0 : 85.0

= 2 . 0 : 3 . 0 : 3 . 0 : 13.0: 15.0: 17.0;
3. The larger the size of a datum feature, the smaller its locating error. According
to the dimensional specification shown in Figure 5-27, this study assumes that
ZI4 : ^6 : ^8 : ^11 : ^12 = 95.0 : 95.0 : 65.0 : 45 : 85.0
= 19.0: 19.0: 13.0:9.0: 17.0;
4. The locating error is generally 1.0 ~ 2.0 times larger than machining error;
5. The smaller the basic dimension, the more accurate the machining result.
The simulation results are shown in Table 6-1 and Table 6-2. The comparision of
the two setup plans are shown in Figure 6-11 and Table 6-3. The silulation results of all
the alternative set plans are shown in Appendix D

Table 6-1. Simulation results: variation stackup of setup plan 1.

Dimensional Component Worst-Case Statistical-Case

Variation Errors Value (mm) Value (mm)

<^4, 6 A^, 5^, 5(y, ^ 0.00666 0.00414

^6,8 Se, Si 0.01644 0.01330

^2, 12 S\2, S2 0.02823 0.02618

<^8, 12 A\2, Si 0.05134 0.03801

^11, 12 S\\, S\2 0.02888 0.01975

(Z</)'" 0.13155 0.05209

140
 Table 6-2. Simulation results: variation stackup of setup plan 2.

Dimensional Component Worst-Case Statistical-Case

Variation Errors Value (mm) Value (mm)
^4,6 Zlg, ^ 4 , < ^ , ^ 0.05393 0.03635
^6,8 4 , Si 0.01616 0.01334
d2, 12 S\2, Si 0.01933 0.01722
d%, 12 A\i,Si 0.03941 0.02907
d\\, 12 S\\, S\2 0.02820 0.01968

0.15703 0.05503

Setup Plan 1

0.05 J/
0.04
0.03
D tolerance
0.02
variation
._
0.01
0
d4,6 cl6,8 c12,12 d8,12 d11,12

Setup Plan 2

D tolerance
I variation
L

d4, 6 d6, 8 d2, 12 d8, 12 d11, 12

Figure 6-11. Column charts of simulation results.

141
 Table 6-3. The comparision of the two setup plans.

Dimensional Tolerance
Setup Plan 1 Setup Plan 2
variation, di,j value, ti,j
Statistical-Case ^ij - ^u Statistical-Case ^,y - ^',7

Value (mm) t.j Value (mm) ^U

^4,6 0.01 0.00414 58.60% 0.03635 -263.50%

<^6, 8 0.04 0.01330 66.75% 0.01334 66.65%
<^2, 12 0.03 0.02618 12.73% 0.01722 42.6%
<^8, 12 0.05 0.03801 23.98% 0.02907 41.86%
<^11, 12 0.02 0.01975 1.25% 0.01968 1.6%

From Table 6-3, we can see that the setup plan 2 cannot satisfy the dimensional
tolerance requirement between/ and/. Furthermore, the setup plan 2 has about one time
larger gap between/ and/n (so as between/ and/), than that of the setup plan 1, which
is a "waste" of the machining capacity. Hence, the setup plan 1 generated by the GATO
setup planner is superior to the alternative setup plan.
For the stackup of geometrical tolerances, the example in section 3.2.2.2 of
Chapter III is a suitable case to illustrate the validity of the setup-planning algorithm
developed in this study.
In the simulation procedure, the initial data shown in Table 6-4 were generated for
the machining of the part shown in section 3.2.2.2.
In table 6-4, denote:
'MN
Geometrical machining error between face M and face N, which
corresponds to the geometrical tolerance shown in Figure 3-12.
E^^: The stackup geometrical error between face M and face N, which
corresponds to the geometrical tolerance shown in Figure 3-12.
The formulas of tolerance analysis in section 3.2.2.2 of Chapter III were used in
the simulation, the simulation result is shown in Table 6-5 and Figure 6-12.

142
 Table 6-4. The initial data (Units: mm).

Parameter Value
U 500.00
Ll 300.00
L3 400.00
^DD 0.01

^BD 0.01

^CB 0.02

^EC 0.03

^FE 0.03

^AF 0.02

Table 6-5. The simulation result (Units: mm).

Worst Case Statistical Case

-'CD 0.04500 0.03219

EED 0.09000 0.05223

'FD 0.12000 0.06017

EAD 0.11000 0.06335

From common sense, it is natural that ECD < EED < EFD < EAD with the increasing
number of component errors. However, from the output of the worst case, there is one
exception as EAD<EFD- The reason of the "exception" is as follows:
With the calculated value of tpD (i-e., E^i^ ) from formula (3-26) and (3-28):

143
 E = -^FD *" ^AF
^AD L3
300
X 0.1200+ 0.02
400
-0.1100
<Ej,j^ =0.1200

E Worst Case
E
"Statistical Case

E_CD E_ED E_FD E AD

Stackup Error

Figure 6-12. The simulation result.

That is, the geometrical error stackup is affected by the basic dimensions of
related features, the stackup error is not the straight sum of component errors.
It can be seen that, in both of the simulation examples, the statistical-case results
are more convincible for analysis. Hence in the simulation of tolerance/variation stackup,
the statistical-case method is preferred to the worst-case method.
The above two simulation examples demonstrate that the tolerance stackup
analysis for both dimensional tolerances and geometrical tolerances in Chapter III is
valid. Hence the principles on selecting setup methods applied in the development of the
setup-planning algorithm in Chapter V are effective for tolerance control. The algorithm
has ensured the major objective of this research
Another concern of validation is the soundness of assumptions made in this
research. In the case studies of Chapter V, it is assumed that the prismatic parts are
machined on a 3-axis vertical milling center using a vice as the fixture. This assumption

144
makes the developed algorithm more generic. If 4-axis or 5-axis machine tools and
fixtures with holders that can be rotated are used, the essential difference obtained is the
number of TADs in the machining is decreased. Since the developed algorithm in this
research can solve the setup planning problems with largest number of TADs (6 TADs),
it can easily be adapted to the cases with less TADs. The illustrated capability of the
algorithm in solving the setup planning of rotational parts (2 TADs) in Chapter V
validated this assumption.
A second significant assumption made in this research is that all the operations of
a feature are finished in one setup, that is, only the finish processes are considered in this
study. This assumption makes the research focus on scientific investigation. Since the
tolerances are mainly assured in the finish processes, the finish machining of the
workpiece is most critical in setup planning in terms of tolerance control. Furthermore,
the algorithm can be modified to involve the raw processes and intermediate processes.
Of courses, in industrial applications, the raw processes and intermediate processes
should also be considered, which is not the focus of this study.
In conclusion, the two major assumptions in this research are valid and necessary
for the exploration of a systematic methodology in this research.

145
 CHAPTER VI
CONCLUSIONS

7.1 Conclusions
In this dissertation, the problems existing in the area of tolerance analysis and
setup planning were stated; literature on tolerance, setup planning, and the application of
graph theory was reviewed; mathematical tolerance analysis for both geometrical and
dimensional tolerances in one-, two-, and three-dimensions was investigated; a graph-
matrix-based setup-planning algorithm was developed; a setup-planning prototype
incorporating the setup-planning algorithm was built.
Tolerance is a complicated while important topic in CAPP. Its study and
application in CAPP, especially in setup plarming, is not mature. The tolerance analysis is
usually confined in dimensional tolerances in one dimension. This study shows that the
geometrical tolerance analysis in two dimensions or three dimensions are much more
challenging, and it is inevitable in the study on the setup planning of prismatic parts. The
impact of basic dimensions/sizes of parts in 2-D and 3-D tolerance stackup is inevitable.
The key point to include both dimensional tolerances and geometrical tolerances in the
tolerance analysis of setup planning is to compare and transfer different types of
tolerances. The tolerance zone and tolerance weight approach explored in this stud> is
demonstrated to be effective in the task. The tolerance zone and tolerance weight
approach also makes the mathematical representation and analysis of tolerances easier
and more accurate.
Many factors must be considered in setup plarming, however tolerance is one of
the most significant and cannot be ignored. Tolerance analysis helps determine the setup
planning strategies. On the other hand a good setup planning strategy can effectiveK
control tolerance. Tolerance analysis and setup planning should be executed in the early
stage of CAPP so that tolerance control can be achieved without increasing the hardw are
requirement of the machine tool and fixture. A good setup planning must assure the
critical dimensional and geometrical tolerances, and release the non-critical tolerances at

146
the same time. That is one of the motivations that tolerance analysis and setup planing are
investigated together in this dissertation.
The generative setup planning system is still rare in industry. The major reason
exists in the variety of manufacturing conditions and part characteristics. A "true"
generative setup planning system is unpractical. However a modifiable setup planning is
always desirable. Since there are a lot of principles and rules in setup planning, a
systematic approach to incorporate them in setup planning is necessary. The graph-matrix
approach applied in this study is effective and flexible. Accompanying graphs, tool-
approach-direction matrix, machining-feature matrix, and tolerance matrix are defined
and applied in representing and computerizing the relations of tool approach directions,
machining features, and tolerances in setup planning. The principles and rules can be
represented and manipulated in the graphs and matrices. The directed-graph approach
explored in this study is more appropriate and effective than undirected-graph approach
when datum selecting and setup sequencing are concerned. By dividing the setup-
planning algorithm into 7 steps, the various principles and rules are allocated to separate
steps, and each step has its corresponding graphs and matrices. The directed-graph
approach is also demonstrated to be applicable in tolerancing and tolerance analysis.
The setup-planning algorithm in this study was originally implemented in C++. It
was finally incorporated in the setup-planning prototype developed in Microsoft Visual
C++ . Without exception, the assumptions listed in the Chapter V are applied in the
development of the prototype, which has open architecture for fixture development.

7.2 Contributions
The major contributions of this research are as follows:
1. Tolerance analysis is based on tolerance zone analysis and aims at
mathematical representation. The concept of tolerance weight is defined and
applied in tolerance analysis and setup planning. The tolerance zone and
tolerance weight approach exposes the essence and relation of various types of
tolerances, makes the comparison and transformation of different types of
tolerances possible;

147
 2. Both dimensional tolerances and geometrical tolerances are analyzed in one-,
two-, and three-dimension for tolerance stackup in setup planning. The impact
of basic dimensions/sizes of parts on tolerance stakup is investigated;
3. The concept of tolerance graph is defined and used consistenth' in both
tolerance analysis and setup planning in the study. Tolerance relations are
represented as both directed graphs and undirected graphs so that various
types of tolerances are appropriately illustrated, and the datum selecting and
setup sequencing become more convenient. Accompanying graphs, tolerance
matrix, tool-approach-direction matrix, and machining-feature matrix are
defined and applied in representing and computerizing the relations of
tolerances, tool approach directions, and machining features in setup planning;
4. A graph-matrix-based and tolerance-oriented algorithm is developed for the
setup planning of both prismatic parts and rotational parts. A setup-planning
prototype based on the algorithm is developed.

7.3 Suggestions for Further Research

The study of tolerance analysis and setup planning in this research is relatively
thorough as a dissertation work. However, a mature setup-planning system based on the
investigated approach and the final goal of tolerance control requires more effort. The
following aspects exposed by this research deserve fiarther investigation:
1. The relation of tolerance design and setup planning. This research shows that
the setup planning can proactively control tolerances, it is possible to design
tolerance concurrently with setup planning so that critical tolerances are
assured while non-critical tolerances are loosened. As show in Table 6-3,
some tolerances may be too tight while others too loose according a specific
set plan, which means the machine capacity can be more economicalh
explored.
2. The analysis of geometrical tolerances is still primitive in this research due to
its complexity. Further investigation of the stackup of geometrical tolerances
in the setup planning and the machining of prismatic parts is desired.

148
 3. The developed setup-plarming algorithm in this research provides an effecti\e
approach to setup planning, however it needs further development in two
directions: firstly, the assumptions on machine tool, fixture, and part should be
decreased to make the algorithm more general; secondly, more principles, and
rules should be incorporated in the algorithm so that it is more accurate.
Similarly, the GATO setup planner is just a prototype at its current status, its
capability can be extended.
4. The developed algorithm has taken tolerance as major concern so that the
generated setup plans carmot guarantee the optimal solution. In ftirther
development, other objectives such as the minimum number of setups and
other conditions such as multi-axis machine tools and fixtures should be
considered. The issue of optimization (on cost and quality) should be further
studied.
In conclusion, automatic tolerance analysis and setup planning are traditional but
underveloped topics in manufacturing, continuous endeavors should be invested in them.

149
 REFERENCES

Ahluwalia, R. S., and Karolin, A. V., 1986, "CATC: A Computer Aided Tolerance
Control System," Journal of Manufacturing Systems, Vol. 3, pp. 153-160.

ANSI Y14.5M, 1982, Dimensioning and Tolerancing, New York, NY: ASME United
Engineering Center.

ASME Y14.5M, 1994, Dimensioning and Tolerancing, (Revision of ANSI Y14.5M-1982

(R1988), New York, NY: American Society of Mechanical Engineers.

ASME, ANSI Y14.5.1 M-94, 1994b, Mathematical Definition of Dimensioning and

Tolerancing principles. New York, NY: American Society of Mechanical
Engineers.

Bass, L., Clements, P., 1998, Software Architecture in Practice, Reading, MA: Addison
Wesley Longman, Inc.

Boerma, J. R., and Kals, H. J. J., 1988, "FIXES, a system for automatic selection of set-
ups and design of fixXmcs,'' Annals of the CIRP, Vol. 37, No. 1, pp. 443-446.

Bollobas, Bela, 1998, Modern Graph Theory, New York, NY: Springer-Verlag Inc.

Bj0rke, 0yvind, 1989, Computer-aided tolerancing, second edition. New York, NY:
ASME Press.

Champati, S., Lu, W. F., and Lin, A. C , 1996, "Automated Operation Sequencing in
Intelligent Process Planning: A Case-Based Reasoning Approach," The
International journal, advanced manufacturing technology. Vol. 12, No. 1, pp.
21.

Chase, K. W., Gao, J., Magleby, S. P., and Sorensen, C. D., 1996, "Including Geometric
Feature Variations in Tolerance Analysis of Mechanical Assemblies," Inst. Ind.
Eng. Trans., HE transactions. Vol. 28, No. 10, pp. 795-807.

Chang, T. C , 1990, Expert process planning for manufacturing, Reading, MA: Addison-
Wesley Pub. Co.

Chang, T. C , Wysk R. A., and Wang H. P., 1998, Computer-aided Manufacturing,

Englewood Cliffs, N.J. : Prentice Hall.

150
Chen, C. L. P., 1993, "Setup Generation and Feature Sequencing Using Unsupervised
Learning Algorithm," Proceedings of the 1993 NSF Design and Manufacturing
Systems Conference, Dearborn, MI: SME, Vol. 1, pp. 981-986.

Chen, J., Zhang, Y. F., and Nee, A. Y. C , 1998, "Setup planning using Hopfield net and
simulated annealing," International journal of production research. Vol. 36.
No.4,pp. 981.

Chen, W. K., 1997, Graph Theory and Its Engineering Applications, Singapore, World
Scientific Publishing Co. Pte. Ltd., pp. 140.

Chemg, John G, Shao, Xin-Yu, Chen, Yubao, and Sferro, Peter R., 1998, "Feature-based
part modeling and process planning for rapid response manufacturing,"'
Computers & industrial engineering. Vol. 34, No. 2 pp.515.

Demey, S., Brussel, H. Van, and Derache, H., 1996, "Determining set-ups for mechanical
workpieces," Robotics and computer-integrated manufacturing. Vol. 12, No. 2,
pp. 195.

Diestel, R., 1997, Graph Theory, New York, NY: Springer-Verlag Inc.

Dong, J., and Vijayan, S., 1997, "Features extraction with the consideration of
manufacturing processes," International journal of production research. Vol. 35.
No. 8, pp. 2135.

Dong, Z., and Soom, A., 1990, "Automatic Optimal Tolerance Design for Related
Dimension Chains," Manufacturing Review, Vol. 3, No. 4, pp. 262-271.

Eary D. F., and Johnson G. E., 1962, Process Engineering for Manufacturing,
Englewood Cliffs, NJ: Prentice-hall.

Edward Enhao Lin and Hong-Chao Zhang, 2000, "Theoretical Tolerance Stackup
Analysis Based on Tolerance Zone Analysis," the International Journal of
Advanced Manufacturing Technology, 1999, accepted.

Foulds, L. R., Jr., 1992, Graph Theory Application, New York, NY: Springer-Verlag, pp.
97-103.

Fuller, James Edward, edited by Grabowski, Ralph., 1996, Using AutoCAD release IS for
Windows. Albany, N.Y. : Autodesk Press.

Henzold, G., 1995, Handbook of geometrical tolerancing: design, manufacturing and

inspection. New York: Wiley.

151
Hayes, C. and Wright, P., 1989, "Automating Process Planning: Using Feature
Interacfions Guide Search," Journal of Manufacturing Systems, Vol. 8, No. 1. pp.
1-15.

Hoffman, E. G., 1996, Setup Reduction Through Effective Workholding, New York, NY:
Industrial Press Inc. pp. 5-16.

Huang, S. H., 1998, "Automated Setup Planning for Lathe Machining,** Journal of
Manufacturing Systems, Vol. 17, No. 3, pp. 196-208.

Huang, S. H., 1995, A graph-matrix approach to setup planning in computer-aided

process planning (CAPP), Ph.D. Dissertation, Texas Tech University.

Huang, S. H., Zhang, H.-C, 1996, "Use of tolerance charts for NC machining,"
Engineering Design and Automation, Vol. 9, No. 1, pp. 91-104.

Huang, Samuel H., Zhang, Hong-Chao, 1996, "Tolerance Analysis in Setup Planning for
Rotational Parts," Journal of manufacturing systems. Vol. 15, No. 5, pp. 340.

Huang, X., and Gu, P., 1994, "Tolerance Analysis in Setup and Fixture Planning for
Precision Machining," Proceedings of the Fourth International Conference on
Computer Integrated Manufacturing and Automation Technology, Rensselaer
Polytechnic Institute, Troy, New York, October 10-12, pp. 298-305.

Irani, S. A., Mittal, R. O., and Lehtihet, E. A., 1989, "Tolerance Chart Optimization,"
International Journal of Production Research, Vol. 27, No. 9, pp. 1531-1552.

Ji, P., 1993b, "A Tree Approach for Tolerance Charting," International Journal of
Production Research, Vol. 31, No. 5, pp. 1023-1033.

Kim, In-Ho, Oh, Jung-Soo, and Cho, Kyu-Kab, 1996, "Computer aided setup planning
for machining processes," Computers & Industrial Engineering, Vol. 13, No. 3-4,
pp. 613.

King, D. A., and de Sam Lazaro, 1994, "Process and Tolerance Considerations in the
Automated Design of Fixtures," Journal of Engineering for Industry, Vol. 116,
pp. 480-486.

Lee, J. D., and Haynes, L. S., 1987, "Finite-Element Analysis of Flexible Fixturing
System," Journal of Engineering for Industry, Vol. 109, pp. 134-139.

Liggett, John V., 1993, Dimensional variation management handbook: a guide for
quality, design, and manufacturing engineers, Englewood Cliffs, N.J.: Prentice
Hall, pp. 2-3.

152
Lin, A. C , Lin, S. Y., and Cheng, S. B., "Extraction of manufacturing features from a
feature-based design model," International Journal of Production Research, Vol.
35, No. 12, pp. 3249.

Lin, Edward Enhao, Zhang Hong-Chao, 1999, "Theoretical Tolerance Stackup Analysis
Based on Tolerance Zone Analysis," the International Journal of Advanced
Manufacturing Technology, accepted.

Link, C. H., 1976, "CAPP-CAM-I Automated process planning system,*' CAPP-CAM-I

Automated process planning system. Proceedings of 1976 NC conference.

Luggen, William W., 1991, Flexible manufacturing cells and systems, Englewood Cliffs,
NJ: Prentice Hall.

Mei, J., Zhang, H., Oldham, William J. B., 1995, "A neural network approach for datum
selection in computer-aided process planning," Computers in Industry, Vol. 27,
pp. 53-64.

Mei, J., 1996, Tolerance Analysis For Setup planning in computer-aided process
planning (CAPP), Ph.D. Dissertation, Texas Tech University, pp. 2-10.

Menassa, R. J., and DeVries, W. R., 1990, "A Design Synthesis and Optimization
Method for Fixtures with Compliant Elements," ASME Symposium on Advances
in Integrated Product Design and Manufacturing, pp. 203-218.

Mittal, R. O., Irani, S. A., and Lehtihet, E. A., 1990, "Tolerance Control in the Machining
of Discrete Components," Journal of Manufacturing Systems, Vol. 9, No. 3, pp.
233-246.

Nee, A. Y. C , Whybrew, K., and Kumar, A. Senthil., 1995, Advanced fixture design for
FMS. London, New York: Springer-Verlag.

Ngoi, B. K. A., and Fang, S. L., 1994, "Computer-Aided Tolerance Charting,"

International Journal of Production Research, Vol. 32, No. 8, pp. 1939-1954.

Ozturk, F., 1997, "The use of machining features in set-up planning fixture design to
interface CAD to CAPP," International journal of vehicle design: the journal of
the International Association for Vehicle Design, Vol. 18, No. 5, pp. 558-573.

Sakurai, H., 1991, "Automatic Setup Planning and Fixture Design for Machining."
Journal of Manufacturing Systems, Vol. 11, No. 1, pp. 30-37.

Sermsuti-Anuwat Y., 1992, Computer-aided Process Planning and Fixture Design

(CAPPED), Ph.D. thesis. University of Canterbury.

153
Sarma, Sanjay E, and Wright, Paul K., 1996, " Algorithms for the Minimization of Setups
and Tool Changes in 'Simply Fixturable' Components in Milling," Journal of
manufacturing systems, Vol. 15, No. 2, pp. 95.

Soenen, R., and Oiling, G. J., 1995, Advanced CAD/CAM Systems, State-of-the-art
Future Trends in Feature Technology, New York, NY: Chapman & Hall.

Smith, Graham T., 1993, CNC Machining Technology, London; New York: Springer-
Verlag.

Torvinen, S., Andersson P.H., and Vihinen, J., 1995, ''Monitoring the accuracy
characteristics of the machinery by using a dynamic measurement approach,"
Proceedings of the 3rd international conference on computer integrated
manufacturing , Singapore, Uuly 11-14, pp.1435-1442.

Tulasiraman, K., and Swamy, M. N. S., 1993, Graphs: Theory and Algorithms, New
York, NY: John Wiley & Sons.

Xiaoqing, T., and Davies, B. J., 1988, "Computer Aided Dimensional Planning,"
International Journal of Production Research, Vol. 26, No. 2, pp. 283-297.

Wade, O. R., 1983, "Tolerance control," Tool and manufacturing engineers handbook: a
reference book for manufacturing engineers, managers, and technicians, 4th ed..
Dearborn, Mich.: Society of Manufacturing Engineers, Vol. 1, Machining, pp. 2-1
to 2-60.

Wang, W., 1985, Interchangeability and Base of Measuring Technology, Harbin, PRC.
Harbin Institute of Technology, pp. 110-130.

Wang Hsu-Pin, Li Jian-Kang, 1991, Computer-Aided Process Planning, New York:

Elsevier Science Publisher B.V..

Weill, R., 1988, "Integrating Dimensioning and Tolerancing in Computer-Aided Process

Plarming," Robotics and Computer Integrated Manufacturing, Vol. 4, No. 1/2, pp.
41-48.

Wu, H. - C , and Chang, T. - C , 1998, "Automated setup selection of feature-based

process planning," International journal of production research. Vol. 36, No. 3.
pp. 695.

Whybrew, K., Britton, G. A., Robinson, D. F., and Sermsuti-Anuwat, Y.. 1990, "A
Graph-theoretic Approach to Tolerance Charting," International Journal oj
Advanced Manufacturing Technology, Vol. 5, pp. 175-183.

154
Zhang, H. - C , Alting, L., 1994, Computerized manufacturing process planning systems,
1st ed., London; New York: Chapman & Hall.

Zhang, H. - C , and HUQ, M. E., 1992, "Tolerancing technique: the state-of-the art,"
International journal of production research , Vol. 30, No. 9, pp. 2111-2135.

Zhang, H.-C, Huang, S. H., and Mei, J., 1996, "Operational Dimensioning and
Tolerancing in Process Planning: Setup Planning," International Journal of
Production Research, Vol. 34, No. 7, pp. 1841-1858.

Zhang, Hong-chao, 1997, Advanced tolerancing techniques, Wiley series in engineering

design and automation. New York: Wiley.

Zhang, Hong-chao, and Lin, Enhao, 1999, "A Hybrid-Graph Approach to Automatic
Setup Planning In CAPP," Robotics and Computer-Integrated Manufacturing, Vol.
15, pp 89-100.

Zhang, Hong-chao, and Lin, Enhao, 1999, "A 'GATO' Algorithm For the Setup Planning
of Prismatic Parts," North American Manufacturing Research Institute,
Transactions ofNAMRI/SME, Vol. 27, pp. 245-250.

Zhang, Hong-chao, and Lin, Enhao, 1999, "Graph-Aided & Tolerance-Oriented

Automatic Setup Planning For Prismatic Parts In CAPP," 1999 NSF grantee's
conference. Design & Manufacturing Session, CD-ROM, Jan, 1999.

Zhang, Y. F., Nee, A. Y. C , and Ong, S. K., 1995, "A Hybrid Approach for Set-up
Planning," The International Journal of Advanced Manufacturing Technology,
Vol. 10, No. 3, pp. 183.

155
 APPENDIX A: THE SETUP-PLANNING ALGORITHM

The pseudocode of the setup-plarming algorithm for both prismatic parts and
rotational parts is listed in the following text. The C++ codes for the setup-planning
algorithm was developed on the pseudocode algorithm, which concentrates on the seven
major steps in the setup planning. Other sub-algorithms for the implementation of the
GATO setup planner are relatively simple and can be referred in the original codes.

Step 1. Input initial information.

1) Read the part type:
'" PRISMATIC", if the part is of prismatic type;
strPartType =
" ROTATIONAL", if the part is of rotational type.

2) Read the machine-tool type:

f"3 - Axis Vertical Milling Center", for prismatic part;
strMachine = <
["CNC Lathe", for rotational part.
3) Read the fixture type:
f" Vice", for prismatic part;
strFixture i
["3 - Jaw Chuck", for rotational part.
4) Read the TAD matrix:

fl, if zth face and yth face can be machined in a same TAD:
''' \o, if /th face and yth face can not be machined in a same TAD.

5) Read the machining-feature matrix:

F = lf,J
[l, if /th face and yth face are in a same machining feature:
'' ~ [0, if /th face and jth face are not in a same machining feature.

6) Read the tolerance matrix:

156
 'J' if t^j = ( / , / ) , i.e. the /th face is the datum of the yth
t =< face in the tolrance relation;
0, if there is no tolerance relation between /th face and ythface.

7) Extract the number of machining faces in the part from either of the above
three matrices, denote as iFaceNum.
8) Extract the initial setup matrix:

1, if yth face can be machined in the 5, TAD;

'.J =
0, if yth face can not be machined in the S TAD.

where
0, when TAD is-X;
1, when TAD is + X;
2, when TAD is - Y;
3, when TAD is+ Y;
S. = \ 4, when TAD is - Z;
5, when TAD is + Z;
6, split TAD set in - X or + X;
7, split TAD set in - Y or + Y;
8, split TAD set in - Z or + Z;

Apparentiy, for rotational parts, Si,j = 0 when / > 1,

Step 2. Assign each face to a definite TAD set.

1) Search for aface that has more than one TAD:
5
1 J >1
if Us,
;=0

/ has more that one TAD

(record the T A D s / has)
if 5IJ

157
 record the /th TAD S, for/

2) Calculate the number of faces in each TAD set:

iFaceNum
Num(S,)= Zs,,
7=0

3) I f / has no tolerance relation with other faces, assign/ to a TAD set that
contains the largest amount of faces:
iFaceNum iFaceNum
if Z^,,+ ZO.=o
M ;=1

fj^S,

where / . e 5"^ and Num{Sj^) = Max{Num{S^, where / e 5,))

4) I f / has tolerance relation with only one face, and the face is in one of the
overlapped TAD set, assign/ to the TAD set that contains its adjacent face:

if Num(t,j ^ 0.0 II tj, ^ 0.0, for / = 1,2,...iFaceNum) - 1

fj^S,

where f. e 5^ and t,j ^ 0.01| tj, ^0.0, and / G 5"^

5) If/ has tolerance relations with more than 1 face, and all the faces are in one
of the overlapped TAD sets, assign / to the same TAD set with one of its
adjacent faces that has the tightest tolerance relation with/:
if Num(t,J =l\\ tj, = 1,for/ = 1,2,...iFaceNum) > 1

fj^S,

where /^ G 5^ and(r,^ ^ 0.0 || /^, ^ 0.0,for feS^) and Minit,j \\ r^,)

6) Assign/ to a TAD set according to good machining practices:

(Leave blank for fiirther development)
7) Repeat the above 6 sub-steps until all the faces are visited:
fory = 1, 2, 3, ..., iFaceNum, repeat the above 6 sub-steps

Step 3. Delete or screen interior tolerance relations.

1) Delete the self-loop edges and tolerances within a machining feature:

158
 for /? = 1 to iFaceNum do
if /,,, ^ 0.0

tp,P = 0-0

2) Screen the tolerances in a same TAD set:

for / = 1 to iFaceNum do
fory= 1 to iFaceNum do

i f ( r , ^ ^ 0 . 0 ) a n d ( ( \ , =1)&&(5,^ =1),^ = 0,1,2,...,5)

t^j = 999.9(999.9 is a flag for screening)

Step 4. Orient undirected edges into directed edges.

1) Search for undirected edges:
for / = 1 to iFaceNum do
fory= 1 to iFaceNum do
if t,^^. = t^., ^ 0.0

catch t, J and tj, for orienting

2) Select the face that has more tolerance relations with other faces as datum:
if Numit,,^ ^ 0.0,A: = 1,2,...iNumFace) > Num{tjj^ ^ 0.0,A: = 1,2,...iNumFace)

tu = 0-0
if Num(tj 1^ ^ 0.0,k = 1,2,...iNumFace) > Num(t-,^ ^ 0.0, A: = 1,2,...iNumFace)

'., = 0-0
3) If/ and / has same number of tolerance relations with other, select the one
that exists in a TAD with fewer faces:
if Numitjj^ ^ 0.0,A: = 1,2,...iNumFace) = Num(t^j^ ^ 0.0,k = 1,2,...iNumFace)

if Num(S^ ,feSJ> Num(S, f. e S)

if Nu^iSm J,^SJ< Num(S^, / , e S)

159
 '..J = 0.0

4) If the above sub-steps still cannot decide the direction of the undirected edge,
orient it according to process precedence constraints or other good machining
rules:
(arbitrarily orient the edge; leave for further development)
5) Repeat the above 1-4 sub-steps until all the undirected edges are exhausted.

Step 5. Group faces into definite setups.

1) Search for directed circuits among TAD sets
for w = 0 to 4
fox n = m + I Xo 5

i f (3 t^^j ^ 0.0,t, e S^,t^ e S) and{3t,j ^ 0.0,t, e S^,t, e 5 J

catch the directed circuit among S^ and S^

2) Search for the nodes in the directed circuits 5"^ and S^.

3) for a node in the directed circuit S^ or S^

temporarily split the node for the TAD set it resides in

split other nodes share a same machining feature with the split node
split other nodes has tight tolerance relation with the split node
count the number of split nodes
4) finally split out a node that cause minimum other nodes to be split
assign all the split nodes into a new TAD set
5) Repeat step 1-4 sub-setups until there is no directed circuit among TAD sets.
6) Regroup vertices as long as the number of islands can be reduced.

7) Update the setup matrix: ^ l^ij J .

8) Restore the screened tolerances

9) Delete the tolerances in a same TAD set:
for / = 1 to iFaceNum do
fory= 1 to iFaceNum do

160
 i f ( r , , , ^ 0 . 0 ) a n d as,, =\)&&(S,^=l),k = 0,U,...,5)

^u = 0-0

Step 6. Sequence setups.

1) For any two TAD sets (islands), if there are only one-way edges connecting
them, the initial island is precedent to the end island:
for S^ a n d S ^ , m ^ n

if3/G5^and/G5,and/,^^0.0

Sm ^^ S^
else if 3 / G S^ and / G 5', and r^, ^ 0.0

Sn => S^

2) Establish the sub-sequence of TAD sets.

3) Search for the TAD sets without precedent TAD sets layer by layer for roots
to leaves.
4) For any two TAD sets (islands), if there are no one-way edges connecting
them, islands with less vertices precede:
for 5"^ and S^ in a same layer, m^n

ifV/G5^and/^G5,and^,^=0.0

if Num{SJ <Num{S)

Sm =^ S^
else if Num(S^) > Num(S)

s.^s^
else
(assign the precedence ofS^ and 5 arbitrarily; lea\ e for

further development in other principles and rules)

5) Repeat the above 1-4 sub-steps until a definite sequence of T.XD sets is
established.

161
Step 7. Select setup datum.
1) The initial vertices of the edges connecting two adjacent islands in a path
should be selected as the datum for the ending island:
For S^ andS^,mif^n

If S^ => S

If 3 feS^ and f^eS,andt^^^ 0.0

select / as datum candidate of S^

2) Repeat step 1 until the datum candidates for all the TAD sets are selected.
3) If a TAD set has no datum candidates up to now, select a large face in its
opposite TAD set as datum candidate.

162
 APPENDIX B: THE MAP OF THE RULES TO THE ALGORITHM

The map of the principles and rules in section 5.1 and the 7 major steps in the
setup-planning algorithm in section 5.2 is show in Table B-1. The map can be referred for
the further modification and development of the setup-plarming algorithm and prototype.

Table B-1. The map of the rules to the algorithm.

The Step in the Algorithm Rules Applied in the Step

Step 1. Input initial information 7,8,9,10
Step 2. Assign each face to a definite TAD set 1,2,3,7,8,9,10
Step 3. Delete or screen interior tolerance relations 4,5
Step 4. Orient undirected edges into directed edges 6, 11,21
Step 5. Group faces into definite setups 1,2,3, 10, l l , 12 13
Step 6. Sequence setups 14,15,16, 17,18,20,22
Step 7. Select setup datum 18, 19,20,23

163
 APPENDIX C: THE BRIEF USER MENU

The "GATO" setup planner is user friendly. The users can learn using the s> stem
by selecting the "Help" menu and "Tutorial" sub-menu in the prototype to read the
tutorial. A new user can start rurming the system step by step in the following sequence:
1. Select a part.
2. Select a machine tool. If you select a wrong machine tool, the system will
warn you and give you a suggestion.
3. Select a fixture. If you select a wrong fixture, the system will warn you and
give you a suggestion.
4. Click the "Show the Part View" button to view the part. The user can choose
to watch the 3-D views and the specification drawings of the selected part.
5. Select a running style. The user can choose to run the setup planning either
"Step By Step or "All At Once".
6. Click the "Start Setup Planning" button to start setup planning.
7. Select a matrix view. The user can choose to view the TAD matrix, the
machining matrix, the tolerance matrix, and the setup matrix during setup
planning.
8. Click the "Next Step" button if you selected the "Step By Step" rumiing style.
run the setup planning step by step until finished.
9. When the setup planning is finished. A "Congratulations" message shows up.
The "Generate Report" button is activated automatically. Click the "Generate
Report" button to generate the setup-planning report. A message box will pop
up and tell you where and what the report is.
10. Select the "File" menu and its sub-menus to customize, save, or print the setup
planning report.
The system provides fault-proof functions in the formats of warning windows, a
message/hints box, and a running status bar. The users can run the system freeh after
they practice the above 10 steps. They can also use the menus to execute the setup

164
planning. Since the system is fauk-tolerant, the users can "play" the system willfully
without breaking the system.

165
 APPENDIX D: THE COMPLETE SIMULATION RESULTS

For the part shown in Figure 6-8, there are 20 alternative valid semp plans as
shown in Table D-1.

Table D-1. The alternative valid setup plans.

Setup Plan Setup 1 Setup 2

Datum Machining Features Datum Machining Features
1 h f\\,f\l,fl, fA /l2 / , / ,

2 ft, f\\,f\l,fl, fA /l2 / , / ,

3 h f\\,f\l,fl, fA /ll fe, / ,

4 k f\\,f\l,fl, fA /ll / / ,

5 h f\\,f\l,fl, fA / fe, /s,

6 h f\\,f\l,fl, fA fA / , / ,

7 f% f\\,f\l. fn / , f,fl, fA
8 U f\\,f\l. fl fe, fs, ji^ fA
9 fz f\\,f\l. fn fe, f,fi, fA
10 h f\\,f\l^ fn fe, f%,fi, fA
11 fn h, h. fs fn,f\iji' JA
12 fn / , / , fe fn,f\i,fi, fA
13 f / , h. /8 fn,f\i,fi, fA
14 /ll / , / , fe fn,f\iji, fA
15 /4 / , / , /8 fn,f\iji, fA
16 /4 / , / , / fn,f\i,fi, fA
17 fn / , / , / , / f fn,f\i.
18 .fn / , / , / , / fe fn,f\i^
19 /ll / , / , / , / /8 /ii'./i:-

20 /ll / , / , / , / / fnjM-

166
 The tolerance graphs of the corresponding setup plans are shown in Figure D-1
and Figure D-2.

Setup Plan 1 Setup Plan 2 Setup Plan 3

S-x 1 S+

Setup Plan 4 Setup Plan 5 Setup Plan 6

Setup Plan 7 Setup Plan 8 Setup Plan 9

Setup Plan 10

Figure D-1. Tolerance graphs (setup plans 1-10).

167
 Setup Plan 11 Setup Plan 12 Setup Plan 13

Setup Plan 14 Setup Plan 15 Setup Plan 16

Setup Plan 17 Setup Plan 18 Setup Plan 19

Setup Plan 20

Figure D-2. Tolerance graphs (setup plans 11 ~ 20).

Using the tolerance graph method, all the 20 setup plans were simulated. The
statistical-case values of the dimensional variations of the 20 setup plans are listed in
Table D-2. The notations in the table are corresponding to those in section 6.2.
168
 Table D-2. The simulation resuhs (unit: mm).

Setup
dA,e ^6,8 di, 12 ^8,12 <^11.12
Plan
1 0.021817 0.013342 0.017116 0.021401 0.022672 0.043806
2 0.019235 0.013341 0.017117 0.021403 0.022656 0.042579
3 0.016371 0.013342 0.017115 0.021401 0.022673 0.041364
4 0.019235 0.013343 0.017112 0.030000 0.022676 0.047487
5 0.019238 0.013342 0.017113 0.021401 0.022678 0.042579
6 0.019236 0.013346 0.017118 0.021817 0.022681 0.042790
7 0.004243 0.013342 0.017117 0.021401 0.022669 0.038222
8 0.004245 0.013345 0.017112 0.021110 0.022671 0.038223
9 0.004246 0.013341 0.023302 0.021409 0.022678 0.041364
10 0.004240 0.013340 0.027111 0.030001 0.022662 0.048425
11 0.021813 0.013343 0.017117 0.021401 0.022671 0.043807
12 0.019232 0.013332 0.017112 0.021401 0.022679 0.042568
13 0.016379 0.013349 0.017131 0.021402 0.022666 0.041365

14 0.019232 0.013345 0.017116 0.030010 0.022672 0.047487

15 0.019235 0.013339 0.017117 0.021401 0.022659 0.042580

16 0.019232 0.013350 0.017123 0.021817 0.022679 0.042791

17 0.004243 0.013336 0.017117 0.021401 0.022672 0.038230

18 0.004247 0.013342 0.017117 0.021403 0.022668 0.038220

19 0.004241 0.013338 0.023302 0.021401 0.022673 0.041364

20 0.004246 0.013341 0.027121 0.030005 0.022675 0.048420

Since the simulation parameters are random generated, the results are usualh' not
identical among different simulations. However, the same conclusion can be drawn from
different simulations. The simulation resuhs of this appendix are a little smaller than
those in section 6.2 because smaller error parameters were used. Ho\\c\er. the ratio
among alternative setup plans is basically unchanged.

169
 From Table D-2, it is shown that the variation stackup of setup plan generated b\
the "GATO" system is close to the smallest stackup. Taking simulation error and other
setup-planning constraints into consideration, we can claim the advantage of the
generated setup plan over other alternative setup plans.

170