N. . ., M

v} 'I «v 'r u

.
3:7. .. . _ .4... .1, .51....
‘ . n 1.. .; '

a- . ..
l.w-."..\... held .'

hu“

. - ”film:
i'
.:

 

 

 

- (r. . ~ .._.
limit; 21",.»
M

 

 

35‘3" .uzzu
5%-

s.

J!
.. ' i‘
éjzf‘gfir'r m 3 ’“n ”‘7. (1‘
V‘ :21th 2:13. ’- uu . t rt ”3 V3; (’5':qu ..
3:... V ‘ WM.- n: W ‘ an _
: . '
I

.3ng .

,3"?
'gfiv’IS-E—fi
'5'?

r a:
“1%.:2u 0W~‘::;!"
. r

«uglyr:
. .‘

”'14,?" pf»: .‘u..,_‘
" I" «v _',. “"rqflfqfi
drum. .,~"=1mr . .

v

p. v
”3.;‘gnlw
If. :- —r-- .--x rm -’;'.3 can? ' J A
mawaya'z; ”"fifl'o'éV-wr - v - ;#.:_:L.,.' ._.
r

C
u? ‘33:“ , r-
6"? $17.?"
.
m3“: w»
. Jaaafi’iflgm“; m)
'33": '" _

~12
, . , .r .._ ' a, 3' ,;_

. . w . a”...
"' " ‘ -. ~*‘ L" 532.3: .. W *rgéggy‘fui"... "ii-.4
w 3; _ . ._...:....;,
rfi- m‘uhugg a. ‘ .

’ . i' .;-;,~-...P.f.f.'5’ ";“’;.":q“.1.:,":_::’ _

4r 1' I «No r'r 7....-. _ a. ,.
'1'. a ‘3" fifisrr .4;qu ‘ y. . w_ H“ I. ';;$§;,.:."
. i u- a- ”-7

..
, . :3": "s!
‘a ,- “3.3.935: 3:35-43:51
H333... -. '3 — "
i”
I: 0' 1

, when: .' :r. ‘
,I' Mr. I'
'6

0'“ ;;”’;:
‘H‘

mm'nrtz

u- .,.
Hovv’nv—JL-my.‘
p.

4 ~ ‘5:- ”1-5

”7.1533:

b 31'1“»,321

r w: .: x. m .. ‘

\:.a)..n - "n'éw' ’ l' J}.
- .1, { » a ."

‘1} ‘th‘. .5 ..’.'. yu‘smviani‘CffiQ‘Su '- {VT-

 

 

lES

iilllllliii

This is to certify that the

thesis entitled

An Expert System Application to the
Inspection Analysis of Paper Printing
Quality in a Package Printing Company

presented by

Yoshinori Ueda

has been accepted towards fulfillment
of the requirements for

M. S . degree inPackaging

Major p sor

%Icfl%gjgj~

Date 914M /(Z /??/
C/ '

0.7639 MS U is an Affirmative Action/Equal Opportunity Institution

F *‘N
nanny i

Michigan State

I University

\— ,l

fl

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

_____________j
DATE DUE DATE DUE DATE DUE

i
_—_—_ll
"i I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l——ll l

MSU In An Affirmative Amen/Equal Opportunity Institution
owns-p1

 

 

 

 

AN EXPERT SYSTEM APPLICATION TO THE INSPECTION ANALYSIS
OF PAPER PRINTING QUALITY IN A PACKAGE PRINTING COMPANY

BY

YOSHINORI UEDA

A THESIS

submitted to
MICHIGAN STATE UNIVERSITY
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
School of Packaging

1991

a57—7ev/

ABSTRACT

AN EXPERT SYSTEM APPLICATION TO THE INSPECTION ANALYSIS
OF PAPER PRINTING QUALITY IN A PACKAGE PRINTING COMPANY

BY

Yoshinori Ueda

The number of printing experts is gradually decreasing and the work
imposed on experts inspecting printing quality is gradually increasing.
These trends could lead to an increase in experts' careless mistakes,
which, in turn, would decrease the overall quality of printed products.
In this research project, an expert system tor inspecting overall print
quality in a package printing company was built and evaluated as a
substitute for the printing experts. The knowledge needed for building
the expert system for evaluating overall print quality was acquired
through discussions with two printing experts and from two literature

sources. The expert system expressed overall print quality as a

numerical value. The expert system and the visual evaluation agreed on
the score for the printing quality.

 

Copyright by
YOSHINORI UEDA

1991

 

 

 

 

 

 

ACKMDWLEDGEIIENTS

I would first like to take this opportunity of expressing my

appreciation to my major professor, Dr. Harold A. Hughes whose helpful
guidance and counsel, has been a constant source of inspiration. I
would also like to thank my committee members, Dr. Hugh E. Lockhart,
Dr. Vincent Bralts, and Dr. Omar Keith Helferich who have offered

valuable suggestions and advice at crucial points in this research.

My thanks go to Mr. Manabu Hirota, and Mr. Toshio Arita of Die Packaging
Corporation who gave me opportunity to learn packaging. My thanks

also go to Mr. Friedrich-Wemer Dossmann of Dossmann Printing
Corporation, Mr. Masaki Ohno of Dainnipon Ink and Chemicals
Corporation, and Mr. Hitoshi Isono of Mitsubishi Heavy Industry
Corporation, and Mr. Katsumi Yamashita of Ueda Printing Corporation
who provided me with a significant information and helme suggestions

during my research project.

I wish to extend my thanks to the many people who have contributed of
their time and energy in the furtherance of this research. I also wish

to extend my gratitude to the many people who have helped and
influenced me during my stay at Michigan State University.

Finally, I want to thank my parents. Yasuyuki and Shinko, and my
brother, Hiroyuki and my grandfather, kenju Ueda who provided a
tremendous support from Japan to help bring this study to a successful

completion.

iii

—-—E-——-r

TABLE OF CONTENTS

LIST OF TABLES Vii
LIST OF FIGURES viii
91169191 Page
I. Introduction 1
A. Significance of the Paper Printing Industry in Japan 1
B. Why IS Quality Control of Paper Printing Needed 3
C. The Current Situation of Quality Control in Japan 4
D. The Expert System for Problem Solving
(Need for the Expert System) 5
E. Problem Statement 6
F. Research Objective 7
ll. Literature Review 8
A. The Expert System Approach 8
1. What is an Expert System ? 8
2. Components of an Expert System 11
3. Knowledge Acquisition ‘ 13
4. Expert System Application in Packaging 14
B. Quality Control in Printing 16
1. Printing Process 16
2. Quality Scale for Color Printing 18
3. Color Reproduction 20
4. Evaluation Method for Color Reproduction 20
Ill. Method 24
A. Introduction 24
B. Expert System Tool Selection 24
1. Selection of LEVEL 5 24

2. Strength of LEVEL 5 26

iv

 

 

Quanta;

C. Knowledge Acquisition

1.

Conversation and Discussion with Experts

1.1 Inspection Analysis of Ueda Printing
Company

1.2 Computerized Quality Assurance System
of Dossmann Printing Company

Literature Review

2.1 Method of Evaluating Printing Quality

2.2 Information for Diagnosis of
Printing Problems

D. Evaluation of Expert System

IV. Building The Expert System

A. Introduction
B. Knowledge Acquisition

1.

2.

9'?!”

How Experts Inspect Printing Products
1.1 Check Lists

Analysis of Quality Assurance

2.1 Color Reproduction

2.2 Printing Register

2.3 After Printing

2.4 Scum

Building the Expert System

Diagnosis of Printing Problems
Evaluation of Color Reproduction

5.1 Measurement of Evaluation Items
5.2 Evaluation Weight and Point
5.3 Overall Score of Printing Quality

C. Development of the Expert System

1. Diagram of the Expert System
V. Evaluation of Expert System
A. Introduction

B. Limitation of the Data

27
28

28
3O
31
31

32
32

 

C. Evaluation Expert System for Color Reproduction
1. Individualized Outputs
2. Evaluation of EXpert System

D. Evaluation by Printing Experts

E. Comparison of the Expert System Results with
Non-Expert Evaluations

VI. Conclusions and Implications

A. Conclusions
1. Summary
2. The Objective

3. Implications for Practical Application and
Further Research

AEEENQIX
A. OVERALL QUALITY POINT BY EXPERT SYSTEM

B. OVERALL QUALITY POINT BY EXPERTS' VISUAL EVALUATION

C. 10 EVALUATION VARIABLES MEASURED BY DENCITOMEI’ ER

AND IMAGE PROCESSING ANALYSIS INSTRUMENT

D. GRAPH OF THE RELATIONSHIP BETWEEN EXPERT SYSTEM

EVALUATION AND EXPERTS' VISUAL EVALUATION

E. VISUAL EVALUATION SOORE TAKEN FROM SAME PERSON

AT MORNING AND AFTERNOON

F. KNOWLEDGE BASE EXAMPLE RULES
FEW

v1

Rage
65
67
68
68
71
73
73
73
73

74

Base
76

77

81

91

97

130

LIST OF TABLES

Iable

1.1

2.1

2.2

4.1

4.2

4.3

5.1

5.2

5.3

5.4

5.5

Total Value Shipment for the Segments of the
printing Industry, the Publishing, the
Manufacturing Industry.

The Tasks Approached by Expert System and
Usage Number.

Purpose of Expert System Development.

Conversion Diagram for Measurement Data :
Soild Density.

A Set of Composite Variables.
Abbreviation and Weight for Each Evaluation Item.

Quality Ratings by the Expert System
Data : Sample of Japanese Tea.

Quality Ratings by the Expert System
Data : Sample of Rice Cracker.

Mean and Standard Deviation of Overall Quality
Point by Experts' Visual Evaluation :
Sample of Japanese Tea and Rice Cracker.

Reliability of Printing Experts : Correlation Coefficient.

Correlation Coefficient Obtained from Overall Quality

Point between Expert System and a Printing Experts.

vii

Base

10

11

53

56

67

67

70

71

72

LIST OF FIGURES

391113

3.1

4.1

4.2

4.3

4.4

4.5

4.6

4.7

5.1

Illustration of the Correlation between Evaluation

Points from the Expert System and Printing Experts.

Decision Rule Example: Printing Register.
Decision Rule Example : Readability of Character.
Decision Rule Example : Pinhole Problem.

Cause and Effect Diagram for Doubling.

Decision Rule Example : Diagnosis of Doubling
Problem.

Decision Rule Example : Relatively Contrast for
Color Reproduction.

Diagram of Expert System.

Ten Variables for Color Reproduction.

viii

35

43

45

47

49

57

60

66

 

CHAPTER I
INTRODUCTION

f r ' ' In in n
The paper printing industry in Japan is experiencing significant change.
Two reasons for this change are : 1) the customer‘s consciousness of
quality paper printing is rising, and 2) the paper printing industry, in

general, is growing.

Customer demand for high quality printing is rising. Customers pay
attention not only to the products but also to the quality of the packages,
including the print quality. It is easy to understand that print quality is a
significant influence on daily life. Ueno et al, in Houso Gijutsu Binran,
stated that print quality is a significant factor in improving purchasing
power. (Ueno, et at. 1983) They also observed that beautiful printing and
the product explanation on the outside surface of a package are useful in
sales promotion and throughout the whole period of transportation,
storage, and shop window display. Therefore, neat and beautifully
printed products which are made of corrugated cardboard and other
materials have been developed, and attention has been paid to each
product’s inside and outside beauty.

The economic scale of the printing industry in Japan is shown in Table 1.1
with comparison data for manufacturing, bookbinding, and publishing.
The table also shows growth rates compared to 1989.

 

 

2

Table 1.1 Total Value Shipment for the segments of a printing industry,
a publishing, and the manufacturing industry (1990)

 

fidu Total amount of shi ment growth ratio (°/o)
SW p ($) 1989 and 1990

 

Manufacture 1 ,710,612,720 0.5 (-)
Paper 19,607,040 3.3
Newspaper 12,857,100 4.6
Publication 1 1 ,330,700 0.0
PRINTING 36,989,970 4.6
Plate Making 3,617,780 0.3
Bookbinder 1 ,438,990 1 .1 (-)
Printing Ink 1,112,650 9.4
Pnnting Machine 4,193,610 5.2

 

Source: Printing Industry, Sugita Sumio, March 30, 1990

It can be seen that the printing, printing ink, and printing machine
industries have had relatively high expansion rates in comparison with
the other industries, including publication, bookbinding and
manufacturing. The expansion in the printing industry has lead, in turn,
to the expansion of printing machine capacity and printing ink production.
Iizuka reports that some printing machines now employ computer
systems and other electronic improvements to maintain the quality of
printing products and thousands of inks with special characteristics are
prepared to meet customer's demands. (Iizuka,1985) It is expected that
the printing industry will continue to expand in the near future. New
products with the high quality printing which customers demand are
coming into existence everyday. However, in order to maintain the high
quality of printed products, the paper box printing industry will have to
make changes. (T akahata, 1987)

3
3 A '0-..1\ o o o '-_,u: ”013:0".

The significance of the paper printing industry was discussed in the

previous section. A primary reason for growth was shown to be the

increased quality consciousness of the customer. Quality of the product

and the package both have to be provided by the supplier to satisfy

customer demands.

The quality of printed products hashad a tendency to rise with the
development of the packaging industry. This is a good trend, because
high quality products sell well in comparison to inferior quality products.
The print quality of a paper box has a direct bearing on the product image.
The appearance of the package in which the product is packed must be of
consistent high quality to appeal to customers. Suppliers avoid shipping
products or packages which have defects because they lower customer
confidence in the supplier. To get good quality products, the package
manufacturer must use a quality control system. However, quality

control of paper box printing is a complicated endeavor and dealing with
quality control is not easy. (Matsuoka, 1987)

Suppliers should avoid shipping products which have defects. In
particular, food and drug products have necessarily strict quality
controls. It is not enough that the letters printed on the drug container
are fine enough that the label can be read. 01 course, it is impossible to
get rid of all sub-standard products. It is important to take measures to
eliminate situations that produce sub-standard products rather than
having to discard sub-standard products. (Daimon, 1986)

f ' I in n
The present quality control system in Japan is based on market response
(customer side). According to the book, Quality Control Terminology in
JISZ 8101, that is the “Japanese Industrial Standard“, quality control is
defined as a method which economically makes products or service
matched to the customers' demands. (Ootsuki, 1990)

The first issue addressed in the quality control system is to confirm that
the machine plate, raw material, printing ink and so forth conform to
the production specification and the original. These confirmations are
the responsibility of the printing machine operator. The operator
compares the original to the specifications and gets approval from the
printing expert. After this process is completed, the actual printing
operation begins. During operation, random samples are taken for
checking register conditions, scum, color drop off and color

reproduction. These inspections compare samples to the specifications.

The checks are made by printing experts using visual inspection. In other
words, the inspection depends on a printing expert's ability to see
problems with the naked eye. However, the inspection done only by
printing experts is not enough. There are many products produced every
day and experts can't complete 100 % of the inspections. In spite of this
situation, the supplier is expected to satisfy customer demands for all
kinds of products. Thus, a quality control system which eliminates and
prevents human error is needed. Detectors, sensors, and numerical
control of color printing are examples of techniques for avoiding human
error. (Matsuoka, 1987)

5

ram 1:1 cl’-'0_ol‘t'l our. 13150107 1,: in:

Experts check printed products visually to evaluate printing quality. This
creates problems because the printing experts are “only human" and may
make careless mistakes when they get tired. (Mizuraku, 1987). Also some
experts depend on intuition when performing their tasks. It is dangerous
to only use expert visual inspection and intuition and this approach must
be avoided. Computerized expert systems show potential for solving this
problem. Expert systems should be able to judge the quality of printed
products, diagnose printing trouble, and give suggestions using knowledge
about printing inspections gathered from human beings (experts ) and
literature. As a result, information about printing quality could be made
available quickly and accurately without careless mistakes and fatigue.
Akita et at. state that problems having the following characteristics are
suitable for an expert system : (Akita, 1988)

1) Experienced knowledge and know how are clear.
2) Rules are complicated but logic is easy.

3) It is difficult for human beings to inspect.

4) The system can be backed up by experts.

The problems of printing quality control fit Akita's model and seem to be
well suited to an expert system. Printing experts know how to inspect
printing quality by evaluating visual impressions. Additionally, the
knowledge necessary for judging the printing quality is so complicated
that it is difficult for a non-expert to do inspections, but problems can be
backed up by experts.

6
There are some specific tasks a computerized expert system for paper
printing quality control could be used for : (Akita, 1988)
1) Judging the total quality of printed products
2) Diagnosing of printed products' quality defects
3) Giving possible reasons and suggestions for correcting the defects

Computerized expert systems could evaluate the overall quality of

printed products and also diagnose specific printing problems such as out
of register, scum, and color reproduction. An expert system could also
give suggestions about and reasons for printing problems. Many kinds of
expert system applications could be created.

LE_ELoblem_Statemem
Suppliers must avoid shipping products which have defects in order to

maintain product quality. Currently, only experts can visually check
samples of paper printing. This situation is expected to cause problems

in the near future for three reasons. The number of experts is gradually
decreasing. The available experts can't take enough time to relax and rest
because of heavy work schedules. Also, the experts have their own
defects such as making careless mistakes, getting tired over time, and
being lazy. These circumstances will lead to an increase in errors,

which in turn will decrease the overall quality of the product. To avoid
these problems, an alternative is needed to substitute for the experts in

the paper printing industry. One approach is to apply an expert system.

MW

There are three main objectives of this research.

1) To identify, quantify, and classify the knowledge which is acquired
from printing experts and literature for paper printing quality inspection.
(This knowledge will be used for developing the knowledge base.)

2) To develop a prototype expert system for evaluating the overall
quality of printing products. (A prototype expert system will be built for
evaluating the overall quality of printed products, by expressing quality
as a numerical figure, and for diagnosing the printed product's defects.)

3) To test and evaluate the expert system for paper printing quality
control. (The goal will be to compare the evaluation obtained from the
prototype expert system with those from experts at a paper printing
company.)

CHAPTER II
LITERATURE REVIEW

W
The purpose of this research was to develop an expert system for

evaluating the overall quality of printed products. Five prototype expert
systems related to the printing industry were developed according to
Akita, but the prototypes have not yet reached practical use. (Akita,
1988) Expert systems, of course, are still a new approach, but it is
anticipated that they will be practical for the printing industry.

The first part of this chapter addresses expert systems in general and
examines how expert systems are being used to address particular
problems. The second part of this chapter examines quality control
systems in the paper printing industry.

W

W

Many books and publications have been written about expert systems.
Currently, an expert system is defined as a computer program that

assists a non-expert to deal with problems which need the interpretation

of experts. The user is able to solve the problem without expert help by
using a computerized expert system. The computerized expert system is a
tool which has stored the knowledge acquired from experts as the
knowledge base. (Akita, 1988) Many expert system applications have been
reported. According to Akita et aI. problems in the field of law, finance,
architecture, agriculture, medicine, and education can be solved well by

9
using computerized expert systems because the professional knowledge in
these fields has been well established. It is predicted that hereafter the
number of experts and workers will gradually decrease as the use of

computerized expert systems increase.

According to Expert System, ten kinds of tasks are well suited for an
expert system: interpretation, prediction, diagnosis, debugging, repair,
monitoring, control, planning, design, and instruction. (Waterrnann,
1987)

1) Interpretation: analyze the results of the experts' evaluation of
particular problems.

2) Prediction: forecast the future situation based on the past as well as
present models.

3) Diagnosis: find the faults in a system according to the interpretation
of potentially noisy data.

4) Debugging: prescribe remedies for diagnosed malfunctions.

5) Repair: prescribe a plan to administer a remedy.

6) Monitoring: continuously interpret signals that set off alarms when
intervention from outside is needed.

7) Control: interpret, predict, and monitor system behavior.

8) Planning: a series of actions intended to reach a target.

9) Design: method of creating an object which satisfies a specific task.
10) Instruction: diagnose, monitor, and control student behavior.

10

When developing an expert system, it is important to first decide what
kinds of tasks the expert system will address. Table 2.1 shows the tasks
addressed by expert systems and the number of systems in use in Japan.

Table 2.1 The Tasks Approached by Expert System and Usage Number

 

 

 

 

Characteristic Number
Inte rotation 15
Pr iction ' 25
Diagnosis 64
Desrgn 44
Planning 35
Monitonng 9
Debugging 5
Repair 10
Instruction 18
Control 19
Others 8
Total 25?
(Akita, 1988)

Source : Y. Hirai ; “Current Status of Expert System Developments in
Japan“ , ICOT JOURNAL, No. 15, 198 .

As the table shows, the task of diagnosis is papular in the field of expert
systems. The task of control, on the other hand, has only had

19 uses. Building an expert system dealing with control is difficult and
there are few examples.

On the other hand, many expert systems have been manufactured. These
expert systems, of course, are being made for specific purposes.
lCOT-JIPDEC Artificial Intelligence in Japan surveyed expert system
development by six hundred computer users in the spring of 1986.
Responses were received from 255 users. Table 2.2 shows the purposes
for expert systems development. (Akita, 1988)

1 1
Table 2.2 PURPOSES OF EXPERT SYSTEMS

 

 

 

 

FUR—ROSE Number
1. Reduction of ex rt's hea work 60
2. Education and iscipline o experts 26
3. Improvement and standardization 35
for customer service
4. Systematization of knowledge 36
5. Improvement of productivity 45
6. Understanding of expert system technology 50
7. Others 2
8. No answers 1
TOTAL 255
(Akita, 1988)

Source : Y. Hirai ; “Current Status of Expert System Developments in
Japan“ , ICOT JOURNAL, No. 15, 198

The survey of ICOT-JIPDEC AI center shows that the primary reason for
expert system development was the reduction of experts' heavy work.
The experts' heavy work loads and the lack of experts can be mitigated by
using expert systems.

W
The concept of the expert system was briefly discussed in the previous

section. An expert system is useful when non-experts encounter problems
which need the interpretation of experts. A computerized expert system
has a knowledge base developed from experts, so non-experts can draw the

similar conclusions.

The following section addresses expert system components. Akita et al
point out that an expert system consists of seven components:
(Akita, 1988)

12

1) Dialogue module

2) Knowledge acquisition function

3) Knowledge base and control module
4) Knowledge base and data base

5) Inference engine

6) Work memory

7) Inference process interpretation function.

The dialogue module is a program which is prepared in question form for
interacting with the expert system. (Akita, 1988) The user answers the
questions presented by the expert system.

The knowledge acquisition function is the most important and most
difficult part of developing an expert system. Generally, the knowledge
and know-how acquired from experts are not systematized, but the
knowledge and know-how have to be put systematically into the computer.
This is a key point in determining whether a system will be valuable .
Currently, the knowledge engineer performs this job rather than the

experts.

The knowledge base and control module are parts of the program that
supply knowledge to the inference engine.

The knowledge base and data base are knowledge and data accumulators.
The data is extracted when needed.

13
The inference engine is a computer program which uses knowledge and

rules to reach inferences. Forward and backward chaining are examples of
actions by the inference engine. Work memory is a memory area where the

facts concerning inference are stored for a short time.

The inference process interpretation function is the part of the program

where the conclusions are reached by the inference engine.

IIEE IS I | E "In
Knowledge acquisition is the most difficult stage of the expert system
development process. The success or failure of an expert system can

depend on how effectively the knowledge is acquired from experts.

The person who acquires information and knowledge from experts and
literature is referred to as the knowledge engineer. The knowledge
engineer interviews experts, extracts the main characteristics of specific
problems, and constructs a trial system. A knowledge engineer needs to
be familiar with expert system methodology. Sometimes the expert and
the knowledge engineer are the same person.

There are two types of expert knowledge; literature knowledge and
heuristic knowledge. (Akita, 1988) Literature knowledge is the basic
knowledge level required to be expert. (Badre, 1973) Literature
knowledge is the most important information source‘for constructing an
expert system. Heuristics, on the other hand, is an important form of
knowledge which distinguishes experts from non-experts. Human experts
offer detailed information concerning particular problems to the

14
knowledge engineer during interviews. Generally, the quantity and quality
of heuristic knowledge is important in constructing practical expert
systems. However, current prototype expert systems use less
information from heuristic knowledge than literature knowledge.

(Akita, 1 988)

Finally, heuristic knowledge acquisition involves important difficulties.
One of the most important is the disagreement in the forms of expression.
The form of knowledge from human experts may differ from that
expressed in the program. (Watennann, 1987) The other important
difficulty is the inability of human experts to express themselves.
Waterrnann pointed out that experts may not be good at expressing their
knowledge. He also indicated that the human expert's knowledge often is
inaccurate, incomplete, and untidy. Therefore, it is difficult to extract
heuristic knowledge for use in an expert system. A knowledge engineer
has to take these difficulties into consideration during knowledge
acquisition.

I 4 ' ' 'n P '

The number of expert system applications in packaging are limited.
Packaging Expertise on a Disk, points out that expert system technology
could be applied in packaging research and development applications and
packaging management applications. (T wede, et, al. 1990)

An application in packaging research and development is to design a
package using data about product characteristics such as weight, shape,
dimensions, and the sensitivity of product quality, in addition to data on
logistical and marketing system characteristics such as customer needs,

15
dynamic force during transportation, and handling. Another application is
to predict future events from packaging science. For example, it is
possible to predict shelf life from permeability models, distribution

damage from a damage boundary curve, and mass transport from diffusion
models. Applications in package system management include: control of
package purchasing, quality control of manufacturing operations,
packaging machinery operation and maintenance, and monitoring a

distribution system for damage causes.

W

The Federal-Mogul Corporation has created a logistics expert system by
using a knowledge based technology in inventory management to provide
advice on ordering, rescheduling, and packaging decisions plus elements
of order entry and forecasting. (Keamey,1990) The main purpose of the
expert system in inventory control application is to achieve desired
availability with minimum inventory.

The Federal-Mogul project started in April of 1988 with an outside
facilitator experienced in expert systems and logistics and was completed
in March of 1990. The first phase was to design a concept and project
plan for using the expert system. It required three months. The second
phase was to develop a prototype expert system. The system was
constructed by using the Aion Development System Tool from the Aion
Company. During this phase, the system was constructed, tested, and
evaluated. Six months were required to complete the second phase. The
last phase of the project was to make the system operational for

inventory control. It was completed after 15 months. The system was

16

validated by checking the accuracy of knowledge in the knowledge base
and the correctness of recommendations. Five benefits from this
operational expert system for inventory control have been reported.
Inventory trade-offs became easily identified by individual part number.
The system allowed access to the total inventory throughout multiple
locations. Inventory balance was improved by repackaging for different
customers. Inventory decisions became more consistent. As a result of
this project, inventory productivity rose 10 or 15 percent.

(Kearney, 1990)

HE D II 5 Il' E'I'
This section focuses on literature regarding the printing process and

quality control in the paper printing industry.

".5 l E . l' E
It is necessary to understand the elements of the printing process before
understanding the entire quality control program of paper printing. Four
points are examined for satisfying complete quality control of paper
printing: registering, the pro-printing process, the printing process, and
maintenance after the real printing process. (Japan Graphic Arts
Technology, 1988)

The first important point to consider is adjusting the printing register.

The register is needed to print in the right position on the paper. This
check is very important in ensuring printing quality because all printed
products are defective when register is not correct. Therefore, adjusting

17
the printing register is the first stage of satisfying paper printing quality

standards. There are three kinds of "out of register“ placement: left and
right gap, up and down gap, and bending gap. (JAGAT, 1988) The reason
for out of register can be very complicated. The cause has to be analyzed
and the machine has to be adjusted accurately until the register matches
before the printing process begins. Moving the plate, the cylinder, and

the position of the paper are examples of adjustments. (JAGAT, 1988)

After finishing the registering process, the pre-printing processes should
be checked before proceeding to production. Character readability, odor,
roughness of solid place, curling, doubling, scum, offsetting, motile,
pinhole, and ghost are checked. (JAGAT, 1988) When defects are
detected, printing experts analyze the cause and make adjustments.

After satisfying the pre-printing process standards, production printing

is started. Quality is checked by observation of the printed products,
using random sampling inspection, and by monitoring changes in water and
printing ink. (Offset, 1987) Random sampling inspection is required
during the actual printing process because machine conditions do not
stabilize in the early stage of the actual printing process and the amount
of ink and water applied can change. After the machine conditions have
been fixed, random sampling inspection is still needed to guarantee the
quality. If the machine has to be stopped for any reason, ink and water
conditions can change and cause the loss of a large amount of time.
Frequent random sampling is preferable to avoid these defects.

18
The final stage of the quality control process is machine maintenance.

Maintenance does not directly influence printing quality, but neglecting
this last stage leads to lower quality of subsequent printed products
because of the loss of machine accuracy. The following procedures are
required after printing has been completed: treatment of the plate and
washing of the ink roll, blanket, and impression cylinder. (JAGAT, 1988)

The next section focuses on literature relating to quality control in the
paper printing industry. To illustrate, three methods of quality control
are introduced.

IIEZ Q ms I I El E'l'
Few studies in the literature deal with quality control applied to a color
scale. However, this topic is becoming more popular because quality
control in the printing industry is changing from inspection using printing
experts' visual evaluation to inspection using instruments which produce
numerical data. One of the main reasons for this trend is the low
reliability of experts. (Akita, 1988)

3 -r-H A-lt~1=ill°:’-_ 0 rumor _; ‘lrua

GAT F (Graphic Arts Technical Foundation) in the United States (Ito, 1990)
researched methods of numerical quality control for printed products.
“Starter Get”, a scale which expresses color reproduction numerical
figure, was reported in 1961. (Ito, 1985) The system is being widely used.
Starter Get is very effective for checking the thickness of dot, slur, and
double during the printing process. (Ito, 1985) It is easy to inspect the
conditions by watching the scale of enlarging Starter Get. Also, Starter

19
Get can calculate the resolving power of printing products. The formula

is:

Resolving power = 11.47 [the width of spread for center solid (Ito, 1985)

Starter Get is accurate for judging the conditions of halftone dot, slur,
and double but it is difficult to control printing using the system because

it uses data from a visual inspection.

W

The dot gain scale was designed for control of dot reproductivity in
printing by GATF in 1965. Dot control is very important for controlling
color. The ideal dot is a complete circle. However, the dots produced by
most printing machines are not complete circles. The disparity of the dot
is evaluated by the dot gain scale which consists of three parts: dot gain
scale, slur gauge, and starter get. The dot gain scale was designed for
inspection and yields numerical data. (Ito, 1985)

W

The last example of printing quality color control is the compact color

test strip designed by GATF. This system, a scale of color printing is
controlled one, two or three piled up color. The following five checks can
be performed: 1) the process ink, hue, and photographic density; 2) the
hue of one or two colors and transparency of ink; 3) the thick or thin for
halftone dot and slur and double; 4) the irregularity of dot reproductivity
and the discrepancy of gray balance; 5) the uniformity of ink supply.

(Ito, 1985) The compact test strip evaluation is expressed as a numerical

20
figure using data from a densitometer. By using the compact color test

strip, quality control for color printing becomes easy.
All three quality control scales are needed for reliable color matching.

I I r

A study by the technical committee of the Japanese Society of Printing
Science and Technology evaluated the parameters and methods for quality
control in the color reproduction process. (Isono, et al. 1989) The
committee reported that several abstract impressions such as brightness,
darkness, clearness, unclearness, softness, and hardness are used in the
printing. These abstract impressions are ambiguous and have not been
standardized. The technical committee members point out (Isono, et al.
1989) that only five abstract impressions are needed to evaluate color
printing quality: 1) color 2) gradation 3) sharpness 4) uniformity 5) gloss.

It is difficult to individually evaluate the abstract impressions by human

visual observations.

II,B,4 Evaluation Method M gala: BBQMUQI'QD
MW

Two methods have been reported to evaluate the color quality of printed
products. (Isono, 1989) In one, hue error, degree for ash color, and
efficiency are calculated for sample sheets. These measure evaluate the

relative inferiority to the original. When comparing color between an
original sheet and a printed product, the color of the printed product may
be inferior to the original because of the following : 1) distortion in the
spectrum characteristics of practical printing ink. 2) surface

 

21
characteristic of printing paper. 3) the operating condition of the
printing machine. To measure the degree of inferiority, the three items
are calculated by the following formula:

Hue error = (M-L) I (H-L) * 100 °/o
Ash degree: LIH * 100%
Efficiency = {1 -(L+M) I 2 H } * 100 % (lsono,1989)
where L: the lowest value of divided color density

M: the middle value of divided color density

H: the highest value of divided color density
The ideal ink has hue error =0 %, ash degree = 0 %, and efficiency = 100 %.

A second measurement evaluation for color is achieved by calculating
trapping efficiency; how the ink is applied on the previous ink during
repeated printing. The trapping efficiency is calculated by the following
formula:

Trapping efficiency = (D12)2 - (D1)2 I (Dz)2 * 100 % (Isono, 1989)
where (D1)2 : divided color density for previous ink
(Dz)2 : divided color density for next ink
(D12)2 : divided color density for layered ink
The ideal trapping efficiency is 100 %.

WW

Gradation is recognized as a second factor for controlling color. Gradation
is used for controlling density from the highlight part, where color is

bright and density is low, to the shadow part, where color is dark and

22

density is high. (Isono, et al 1989) Gradation is mainly controlled by the
halftone dot. The important evaluation measurement item for the halftone
dot is dot gain. (Isono, 1989) Dot gain is calculated by the following
formula.

Dot gain = ( Dot area ratio for printing products ) - ( Dot area
ratio for plate or film ) (Isono, 1989)

4 r f
The third factor for controlling printed color is sharpness. The shape
coefficient is used to evaluate the dots outside reproduction and is
calculated by the following formula:

Shape coefficient = (the length of circumference )2] 2m ( area )

Usono,1989)

If the shape coefficient is 1.0, the sharpness condition of the portrait is
perfect.

11.3.4.4 Measurement of Unjtqmmy
The fourth approach to controlling printed color is to measure the

uniformity of the printed products. There are two types of uniformity
evaluation for printing portrait; micro-unifonnity and macro-uniformity.
Micro-unifonnity is a measure of how the ink is applied, and
macro-uniformity is a measure of the consumption patches of ink. (Isono,
et al. 1989) The evaluation for uniformity of the printed products is

23

expressed by standard deviation of density trace measured by micro
densitometer and change coefficient.

Change coefficient = Standard deviation of density trace / average density
* 100 % (Isono, 1989)

iLBALMeasuLememotfiloss

The last factor for controlling the printing portrait is gloss; expressed as
the rate of bumish for printing products. The gloss is expressed as a
numerical quantity by using the photometer, using the following formula:

Gloss = The light quantity of positive reflection [the positive reflection
light quantity of complete mirror * 100 %.

These studies emphasize that objective evaluation is possible using
numerical control and that an objective expression for printing quality
may be very helpful in judging printing quality.

CHAPTER III
METHOD

M21000
II ‘ mm .‘1I'l :n: .. OV:u:W Of! It 0 I 1:m
The primary objective of this research project was to develop and test a
prototype expert system for color reproduction. There have been prior
investigations into the development of a small scale expert systems, but
there are few situations where an expert system has been put into
practical use in the printing industry.

In this chapter, the procedure followed to develop the prototype expert
system is explained. The steps included ; selection of the expert system
tool, knowledge acquisition, and evaluation of the completed expert
system for color reproduction

|l|,B Expert System Tm! salmon

W

More than a dozen expert system shells or tools are now available in the
field of artificial intelligence. The following are some key items to
consider when selecting a particular tool. (Akita, 1988)

1. Price
Since the expert system is still a new approach in the field of packaging
there is constant revision, reformulation, and augmentation. It was not

necessary to select the highest priced tool.

24

 

25
2. Ease of Ieaming

Ease of Ieaming is important, especially if the author is neither a

computer programmer nor computer expert. A highly developed expert
system tool was not needed for this research. Therefore, the ease of
Ieaming was an important consideration.

3. Previous uses of tool

The extent of use was also an important factor in deciding on the optional
tool. Widespread use of a tool indicates that the methods tend to need
less debugging and fewer changes and that it is easy to get information
about strengths and weaknesses.

4. Connection with other systems

When selecting the tool, connection with other software, hardware, and
networks must be considered because the prototype expert system may be
a practical expert system in the near feature. Thus, it was important to
consider this feature.

5. Language dependency

It was important to recognize to what extent the tool itself depends on

the computer language and how long the tool takes to interface, an

important measure of the responsiveness of the expert system.

Since the purpose of this research was to build a prototype expert system
for paper printing quality, and since limited time was available, the
most important factor in choosing the tool was easy of Ieaming.

The tool selected was Level 5/Macintosh, a rule based tool, produced by
lnfonnation Builders.

26

r , r I ' h

Level 5 uses a versatile knowledge representation language called

Production Rule Language (PRL) for development of the knowledge base. In
PRL, knowledge is represented as IF...AND...OR...THEN...ELSE rules, which
contain the factual lnfonnation comprising the domain of the expert

system. (Level 5)

Level 5 had the follows advantages:

1) It was lower-priced.

2) It was easy to learn.

3) It had flexible application.

4) It had linked knowledge base.

The price of the expert system was an important factor in deciding on a
tool or shell so the lower priced Level 5 was selected. Level 5IMacintosh
was an adequate system for building the expert system in this research
project. '

Level 5 offered simple rules with mathematical capabilities. In short,
the ease of Ieaming was a strong factor because the research objective
was to develop the prototype expert system in a short period of time.

Level 5 is flexible and can suit many applications. In most expert
systems, designers must choose between a forward or backward

inference engine. However, Level 5 can use both backward chaining and
forward chaining to reach an inference.

27
The fourth factor was that Level 5 could be linked to HyperCard or Excel.
The linked knowledge bases can communicate with one another
dynamically and update global facts with the engagement of each
knowledge base.

I r n
Knowledge acquisition was the most important activity in the
development of the expert system for printing quality control. To acquire
knowledge it was necessary to analyze the knowledge and determine how
to obtain it from experts and literature. An optimal method of knowledge
acquisition has not been established, so the process for acquiring
knowledge of a specific printing quality problem can be difficult.

An expert's knowledge can be divided into two types of knowledge:
literature knowledge and heuristic knowledge. (Akita, 1988) Literature
knowledge is the minimum factual knowledge required to be an expert.
Heuristics, on the other hand, includes the knowledge needed for
distinguishing between experts and non-experts. Both kinds of knowledge
were necessary to build the prototype expert system. Literature
knowledge is more than 90 % of the total knowledge required for an expert
system, (Akita, 1988) so it was important to ensure complete acquisition
of this type of knowledge.

The researcher, acting as knowledge engineer, obtained knowledge from
two printing publications and from two printing experts. The knowledge

which was obtained is described in a later section.

28

The following sources of knowledge were accessed.
1. Knowledge acquisition from heuristics
1) Inspection of Ueda Printing Company
2) Computerized quality assurance system of Dossmann Printing
Company
2. Knowledge acquisition from literature
1) Printing quality evaluation method for color reproduction

2) Diagnosis of printing problems

I v ' i ' ' h

i P ' ' m
The purpose of the knowledge acquisition in this section was to discover
and collect information to describe how a printing expert checks the
printed product samples during a quality inspection. The information was
gathered through interviews with printing experts.

A printing company in Japan was selected. This company, UEDA Printing 8
Paper Box Company, has an excellent printing expert who was interviewed
for the purpose of acquiring detailed knowledge.

There were three goals for this part of the knowledge base. The first goal
was to discover the printing conditions. The printing conditions were the
principal knowledge needed to understand the printing process, the
beginning step for building a prototype expert system. In order to
understand the conditions, questions were asked about four areas:

 

29
1) Product
a. What kind of packages are printed 7 (food, confectionery, medicine,
cosmetics, and other miscellaneous)
b. Are there regulations that apply 7
c. Does the type of product influence the expert?
2) Material
a. What kinds of materials are used 7 (paperboard, E-flute,
thin paper, etc.)
b. Is there a relationship between paper type and printing quality 7
c. Does material type influence procedures which are conducted before
the actual printing process begins 7
3) Ink
a. What kinds of printing inks are used 7 (carton ink, corrugated ink,
or special ink.)
b. Are there specific print quality checks that apply when dealing with
aparticularmedical orfood product?
4) Equipment
a. What kinds of printing machines are used 7
b. What features do the machines have 7
c. How does the coating system work 7
d. What does the expert observe when evaluating coating and printing
quality 7

The second goal of the interview was to locate checldists for inspecting
printing products and to determine the meaning of each check item. In
most printing companies, only the expert checks printing samples
visually to determine whether the sample is satisfactory. Printing

30
experts have the inspection knowledge in their brains. Thus, the second
goal was to discover the inspection items and their meaning from the
printing experts. The following questions were asked.

a. What are the details of the checklists used by printing experts to
examine quality 7

b. Are there more checklists 7

c. Which checklist does the printing expert use 7

e. What is the meaning of each checklist item 7

The third goal of the knowledge acquisition process was to determine how
experts check the printed products. The printing experts do not check the
printing samples by reference to inspection manuals. They have
individualized knowledge and procedures for inspecting printed products.
This part of the knowledge base was considered to be the most critical for
the development of the expert system, and a large amount of time was
committed to discover how the expert performs the inspection. The
following questions were posed:

a. Do printing experts actually inspect printing samples according to
inspection checldists 7

Are there any experiments involved in the inspection process 7

What tests are used 7

Which checldists do printing experts use for the visual inspection 7
Which checldists are most important for evaluating printing quality.

99.0.6

A'l‘. .131. 3i. P -. It i... .1116 513M ' 13 '2' . ll.‘.ll
E . I' Q
The knowledge in this section was obtained from a discussion with Dr.

31
Friedrich Dossmann, president of the Dossmann Printing Company in West
Germany. Dossmann has created several systems for improvement of
quality and productivity including systems for quality assurance by
computers. The knowledge acquired from Dossman has been included in
the knowledge base of the program. (Dossmann, 1990)

The following procedure was used to complete the knowledge acquisition.

Determine the general procedure for quality assurance by computer.
Examine the flow chart.

Acquire the checklist for quality control.

Determine the procedure used by the computer to promote quality

control and quality assurance.

PS‘N.‘

5. The structure for basic data, including basic classification, quality
classification, and defect classification.

III C 2 | 'l | E .
IIIQZI II II le | I’ E' I' Q I!

The third method was knowledge acquisition from the publication OFFSET
PRINTING MACHINE. (Isono, et al, 1989) The printing machine section
personnel conducted on evaluation of printing quality. The results of the
study are summarized briefly in the next section.

The group established a mathematical formula for the evaluation of the
overall quality of printing using the multivalent analysis method. Ten
quality items are transformed in a single formula, expressing the overall
quality of printed products as numerical values. (lsono et, al., 1989)

32
Evaluation for the overall quality of printed products were inspected by
measuring ten evaluation items. The score for color reproduction was
calculated from the numerical values for ten items. The score was a
numerical value with a maximum 100 points.

IIIQZZII l’ l I' . [E'I' E II
This section of the knowledge base was set up to provide lnfonnation to

assist in the diagnosis of printing problems such as out of register, scum,
and curling. Generally, it is difficult to diagnose when printing problems
happen in an offset printing machine. Offset printing machine problems
result because the machine and the process are complicated. (T akayanagi,
1986) This problem was approached by using the cause and effect

diagram.

WELD

The completed prototype expert system was tested to determine if the
results agreed with the expert's evaluation. Only the section for color
reproduction was evaluated because it relied most heavily on the printing
expert. The method was comparison of the expert opinion with
non-expert system result.

Two cases, pictures of Japanese tea and rice crackers, were tested to
determine whether the prototype expert system could be applied
effectiVely. The correlation coefficient was calculated to measure the
relationship between overall quality score obtained from the expert

system and expert's visual evaluation score.

33
1) Sample
Each sample was classified into five quality levels: highest quality, high
quality, medium quality, low quality, and lowest quality. A total of 5
sample products were prepared by printing experts to be highest, high
quality, etc from Japanese tea sample. The five rice cracker samples

were prepared similarly. Data were recorded in tables, as shown below.

A) Japanese Tea #1

Highest High Medium Low lowest
quality quality quality quality quality

 

Sample

 

B) Rice Cracker #2

Highest High Medium Low lowest
quality quality quality quality quality

 

SamMe

 

2) Evaluation by Prototype Expert System

Ten samples, 5 samples of Japanese tea and 5 samples of rice crackers,
were evaluated using the prototype expert system and given an overall
quality point rating for color reproduction. The score was 100 points as

a maximum.

34
3) Evaluation by Printing Expert
Fifty printing experts from five different printing companies; Ueda
Printing & Paper Box Co., Ltd, Dainippon Ink & Chemical, Inc, Sumida
Paper Industry, and Total Packaging Co., Ltd evaluated the same products.
Printing experts checked the printing samples and rated the overall
quality, with a maximum of 10 points. Based on the expert' scores and
the score from the expert system, the correlation coefficient was
calculated. The expert's reliability was also examined by having each
person score two times in the same day (morning and altemoon). The
correlation coefficient was calculated based on the scores.
4) Method
The overall quality point ratings by the experts and the prototype expert
system operated by a non-expert were compared. Based on this data, the
correlation coefficient between these overall quality points was
calculated in order to evaluate the developed expert system. The graph
which follows is an example which illustrates the correlation between
evaluation points from the expert system and printing experts.
The statistically significant test (p<0.05) was applied to the correlation

coefficient.

35
FIGURE 3.1 Illustration of the Correlation between Evaluation Points
from the Expert System and Printing Experts.

a) In the case of Japanese tea products

  
   

 

 

6
(printing pattern = food product)
5
evaluation
by «pens
4
correlation 0.97
3
O correlation 0.923
2
O
1
20 40 60 80 100
evaluation point by expert system

b) In the case of rice cracker products

6
5

evaluation
by experts

(printing pattern = miscellaneous product)

     
 

correlation 0.912
correlation 0.812

 

 

20 40 60 80 100
evaluation point by expert system

CHAPTER IV

BUILDING THE EXPERT SYSTEM
W
W
This chapter describes how the prototype expert system for printing
quality was built using the knowledge acquired from printing experts and
the literature. Level Five for Macintosh was used as the tool, as
discussed previously.

The expert system used knowledge from four sources, two printing
experts, one from Japan and one from West Germany, and two literature

sources.

IIIB IE II E "l'
IIIBIII E II IE'I' Ell

The first knowledge base for building this prototype expert system was
obtained from printing experts who work at the UEDA Printing 8 Paper
Box Company in Japan. The researcher acting as knowledge engineer,
interviewed printing experts to get knowledge which was needed for
building the expert system. The main goal for the first stage was to find
out what types of check lists exist for inspecting printing quality, what
each check item means, and how the printing experts check printing
products based on the check lists.

W
Currently, only printing experts can check samples visually to determine

36

37

whether the quality of the printing product is good or not, because the
printing expert has these checklists stored in his or her memory.

The following 11 check items for quality inspection were identified
during the discussions with the printing experts. These items were the

main categories in the knowledge base.

Is the printing out of register 7

What is the condition of the character and picture on the surface 7
Is there a strange odor from the surface 7

How is the roughness of solid place 7

Is there a curling problem 7

Is there a doubling problem 7

Are there stains, spots, or blots on the printing surface 7

Is there an offsetting problem 7

PPNP’WPPNT‘

Is there a mottling problem 7
. Are there pinholes on the surface 7
. Is there a ghost image on the surface 7

cull—l
do

The overall printing quality inspection is based on these checklists. In
the next section, each checklists is classified into the four categories
which were gathered from the discussion with Mr. Dossmann.

MW
This section of the knowledge base was obtained from the discussion
with Dr. Friedrich Dossmann who is a president of Dossmann Printing

Company in West Germany. He has created numerous systems for

33
improving quality and productivity in paper printing companies. The main
purpose of this discussion was to acquire lnfonnation about quality
classification and defect classification according to the quality

assurance system created by Dossmann.

The knowledge base for quality classification of paper printing was
divided into four categories. The knowledge in these four categories was
acquired from the quality assurance system for paper printing through
the discussion with Dossmann. The four categories were :

1) Color reproduction

2) Printing register

3) Overall quality immediately after printing operation
4) Scumming

Each of these four categories are discussed in the following sections.

W

The first step in inspecting print quality was to check color reproduction.
Color reproduction indicates how well the original color tone is
reproduced. (Isono, 1989) To inspect color reproduction for paper
printing quality, Dossmann arranged the following items into the quality
assurance checking system:

1. Is the color of the printing surface bright or dark 7

2. Is there enough gloss 7

3. Is the color tone uniform 7
4. Is there any doubling phenomenon 7

These items were used when the answer was neither completely true nor
completely false; however, it was difficult for the non-printing experts

39
to describe the color of the printed products. Inspecting color of the

printed products is the most important phase of evaluating paper printing
quality and it had to be done very well. After careful discussion with the
printing expert, the method from the literature, a standardized
mathematical formula for color reproduction, was selected, making it
possible to evaluate color reproduction by one numerical value based on
the measured data.

Each inspection element for color reproduction consisted of the following
ten individual measures: (Isono, 1989)
1) Relative contrast
2) Solid density
3) Saturation
4) Hue error
5) Three piled up color degree
6) Degree for ash color
7) Effective density in halftone dot
8) Environs scumming of halftone dot
9) Dot gain
10) Shape coefficient in halftone dot

A more detailed discussion about color reproduction is included in a
following section.

IIIEZZ E' l' B 'l
The second step in the evaluating of printing quality was to check the

printing register. There are several kinds of out of register defects such
as right and left, up and down, and bending out of register. If the out of
register phenomenon occurs in the printing operation, it is necessary to

adjust the machine using the results of the out of register inspection.

40
President Dossmann expressed the tolerance limit of printing register by

numerical value in the quality assurance system for printing quality.
Previously, the printing expert inspected the printing register

phenomenon visually.

The following tolerance limits for out of register have been included in
the knowledge base of the program.

1) out of register is over 1 mm

2) out of register is between 0.3 mm and 1.0 mm
3) out of register is between 0.3 mm and 0.1 mm
4) out of register is less than 0.1 mm

The decision rule for the printing register will be outlined and discussed
in detail In the following section.

 

Dossmann classified overall quality immediately after printing operation
as the third classification for printing quality. Overall quality also
required skilled printing experts to inspect the quality. The inspection

items were :

1) to check readability of character
2) to check cleamess of picture

3) to check strange odor

4) to check roughness of solid place
5) to check curling condition

6) to check doubling condition

Printing experts had a variety of ways to inspect the overall quality
immediately after a printing operation. It depended on the printing

41
experts impression, whether these inspection items were accepted or

rejected as measures of the printing quality. All the items were
addressed in the knowledge base which is discussed the following
section. These inspection items were not necessarily used at every
printing company for maintaining paper printing quality. This
classification was adopted in order to build the prototype expert system
for printing quality. More efficient classifications are possible.

NW9

The last classification measure for the quality assurance system created
by Dossmann was to check the scumming. Scumming is a phenomenon
which occurs either during the printing operation or alter the printing
operation and happens during delivering, rubbing machine, and some
other substances. Five inspection items for checking scumming were
drawn from the discussion with the printing experts at Ueda printing

company. The following scumming inspection items were considered:

1) check overall scumming (tinting, stable scumming, partial
scumming, line plumps overall, and scumming in a vertical
direction.)

2) check offsetting

3) check mottling condition of surface

4) pinhole problem

5) check ghost phenomenon

Scumming could be quality controlled by inspecting the above items. This
was, of course, not enough to perfectly control scumming; however, the
overall scumming inspection was satisfied by inspecting these items.
Printed products with scumming lose quality so serious inspections were
necessary. The prototype expert system for overall quality of paper

42
printing was built by classifying : color reproduction, printing register,

overall quality and scumming. Decision rules that needed to evaluate
overall quality of paper printing are outlined and discussed in the next

section.

Mil—Wain

As was described in previous sections, the prototype expert system
evaluated four items color reproduction, printing register, overall
quality immediately after the printing operation, and scumming. The
following sections describe how the prototype expert system for printing
quality was organized and developed.

1. Printing Register

The out of printing register is the most serious printing problem. Out of
register means that lines drawing for multi-color printing or repeated
printing shift up and down or right and left. (T akayanagi, 1986) Printing
experts usually check this shift by a magnifier

The limits on out of register were obtained from printing experts. If the
shift of printing register was over 1m, printing experts classified the
product as out of register, and the product had to be diagnosed to
understand the reasons. Therefore, the knowledge base automatically
shifts to the diagnosis section. If the shift of printing register is less
than 0.3mm, the product was defined to not be out of register and the
knowledge base continue through the inspection process.

43
Figure 4.1 is a decision rule example for this section of the knowledge
base. The knowledge base is written in the versatile knowledge
representation language called Production Rule Language. In PRL,
knowledge is represented as IF...AND...OR...THEN...ELSE rules. (Level 5)
The first line is the name of the rule, in this case, “excellent register“.
The second line is a condition. The knowledge base asks the user to state
the size of the shilt in printing register. The operator measures the shift
of the printing register and responds. In this case, the knowledge base
draws the conclusion that the condition of the printing register is

excellent.

FIGURE 4.1 DECISION RULE EXAMPLE: PRINTING REGISTER

RULE Excellent register

IF Shift of printing register is less than 0.1 mm
THEN The condition of printing register is excellent
AND CHAIN readability of character

2. Readability of Character to Ghost Image Problem

As mentioned in a previous section, many abstract expressions such as
brightness, darkness, clearness, uncleamess, softness, and hardness are
used to express the quality of printing. In inspection analysis of paper
printing applications, decisions are often made on the basis of uncertain
or unreliable information. Many check lists include abstract expressions,
such as readability of character, cleamess of picture, strange odor,
roughness of solid place, curling conditions, surface doubling, overall
scum condition, offset condition, surface mottle, and ghost image
problem. The answers to the above inspection items are neither

completely true nor completely false but are believed with a greater or

44
lesser degree of confidence. A confidence factor of 100 indicates
complete confidence that the statement is true, a confidence factor of 0
indicates complete confidence that the statement is false, and a
confidence factor of 50 is interpreted as noncommittal, statement might
be true or false. A confidence factor was applied to evaluate the

abstract expressions, using a scale from 0 0/o to 100 0/o.

The inspection items which were evaluated with confidence factors were:
readability of character, cleamess of picture, strange odor problem,
roughness of solid place, curling conditions, surface doubling, overall
scum condition, offset condition, surface mottle, and ghost image
problem. The minimum degree of confidence required for a fact to
assumed to be true was set at 70 points. If the user expressed a
confidence evaluation of over 70 points, the knowledge base assumed
that the condition for the item was true and was directed to inspect the
next item by the CHAIN command. On the other hand, if the user gave a
confidence factor of less than 70 points, the knowledge base assumed
that the condition for the item was false and automatically proceeded to
the diagnosis section to find out the possible reasons for the condition
and made suggestions for corrective actions for the printing trouble.

A decision rule example for readability of character is shown in Figure

4.2 The tOp line is title, “Readability of character“. The item next to

TITLE is a description of the knowledge base printed on the screen by the
DISPLAY command. The next line is CONFIDENCE ON, a control statement
that turns on confidence prompting for a knowledge base. The next line

is THRESHOLD=70 which sets the minimum degree of confidence required

45
for a fact to be assumed to be true. The next line is a goal outline for the
inference engine. In this case, “the condition of characters on the
printing surface IS WHAT" is a goal outline. The following lines are rules
for readability of character. TEXT express the declaration to be
substituted for the name of a fact in query displays. The last line is END
which defines the end of a knowledge base.

FIGURE 4.2 DECISION RULE EXAMPLE : READABILITY OF CHARACTER

TITLE Readability of Character DISPLAY

*flflflflflflttfiflflflfiflIflflflflflttt*flifltfiflt.**.***fi********t***t*

#2 SECOND CHECK POINT I READABILITY OF CHARACTER

tit.***Ifltflfltttttflfitt*fi*.***fl*****Ctfitfilttttttfitttfittt

THIS SECTION IS GOING TO CHECK THE CONDITION OF
CHARACTER ON THE PRINTING SURFACE.

“CONFIDENCE FACT OR“ IS USED FOR CHECKING THE CONDITION OF
CHARACTER ON THE PRINTING SURFACE BECAUSE THE DECISIONS ARE
OFTEN MADE ON THE BASIS OF UNCERTAIN OR UNRELIABLE INFORMATION.

A “CONFIDENCE FACTOR" OF 100 MEANS THAT THE FACT IS TRUE.
A “CONFIDENCE FACTOR” OF 0 MEANS THAT THE FACT IS FALSE.

3.....3.ti...*fliflfifltflfl.fiflflflfittfiififlfltfltttfltttii.flfltflfiflflfl

PLEASE CLICK <CONTINUE> WHEN YOU ARE READY TO GO ON.
THE PROCEDURE IS THE NEXT SECTION.

comosuce ON
THRESHOLD=7O

1. the condition of characters on the printing surface IS WHAT
RULE readability of character

IF readability of character

THEN the condition of character in printing surface IS ok
AND CHAIN cleamess of picture

46
Fig 4.2 (Con‘t)
ELSE CHAIN diagnosis letter problem
TEXT readability of character
HOW MUCH CONFIDENCE DO YOU HAVE ABOUT READABILITY OF
CHARACTER 7
PLEASE SHOW THE CONFIDENCE POINT BETWEEN 0 AND 100 POINTS,

BY USING THE ABOVE SCALE.

0 50 100
CAN NOT READ MIDDLE CAN READ WELL
END

The knowledge base for the rest of the inspection items can be seen in
Appendix E.

3. Pinhole Problem
The following four classifications were used for pinholes.

critical condition - the diameter of the pinhole is over 4 mm.
major condition - the diameter of the pinhole is between 2 mm and 4 mm.
minor condition - the diameter of the pinhole is between 1mm and 2 mm.

excellent condition - the diameter of the pinhole is less than 1mm.

The printing expert judges the condition of pinholes by using the
classification system listed above. The diameter of pinhole can be
measured by a magnifier.

47
If the pinhole diameter is over 4 mm, the quality of printing is greatly
degraded. The knowledge base proceeds directly to the diagnosis section
to find out the reasons for the critical pinhole condition.

If the pinhole diameter is between 2 mm and 4 mm, it affects the printed
quality of the product to a lesser extent. Products with pinhole diameter
between 2 mm and 4 mm indicate that pinholes may become a critical
condition later in the operation. The knowledge base assumes it should
proceed to the diagnosis section to find out the reasons for the pinhole

problem.

If the diameter of the pinhole, is less than 1 mm, the pinhole condition is
excellent and the knowledge base assumes that this product is good
quality. As a result, the user goes to the next case to check the quality

of paper printing.

If the diameter of the pinhole is between 1 mm and 2 m, it will not
affect printing quality too much. The knowledge base assumes that a
diameter at this limit is ok and goes to the next section to check the
printing condition.

A decision rule example is shown in Figure 4.3.

FIGURE 4.3 DECISION RULE EXAMPLE: PINHOLE PROBLEM

RULE critical condition

IF the pinhole diameter is over 4 mm

THEN The pinhole condition is critical condition
AND DISPLAY quality

48

Fig 4.3 (Con't)
AND CHAIN diagnose pinhole problem
DISPLAY quality

The diameter of pinhole over 4 mm is a critical condition.
The product affects printing quality seriously.
It is necessary to investigate the reasons for this critical condition.

P ' ' Pr
This section of the knowledge base, taken from Offset Printing Machine,
is used to diagnose printing trouble to find out the reasons for and make
suggestions about printing problems such as out of register, scumming
and doubling. (T akayanagi, et al, 1986)

It is difficult to express the reasons for and suggestions for correcting
printing trouble, because many elements, such as chemistry, physics, and
machine mechanism are involved. (T akayanagi, 1986) The cause and
effect diagram shown in Figure 4.4 was taken from Offset Printing
Machine. (T akayanagi, 1986)

49
FIGURE 4.4 CAUSE AND EFFECT DIAGRAM FOR DOUBLING

 

 

 

 

 

 

 

 

_ CYINDER ARRANGEMENT
POOR NAIL ADJUSTMENT
SHOCK BY MACHINERY
,_ HE _________, FORM OF BEARING
DFFERENCE CRASH OF GEAR
MOVEMENT NAIL SHAFT AT
RIGHT AND LEFT RIGHT AND LEFT
‘— VIBRATION OF MACHINE
MACHINE NOT AT
HORIZONTAL
ooueuuc__ mane lggm'gfgwsm
m l
POOR NAIL ADJUSTMENT
POOR NAIL ADJUSTMENT
— SHEET ONE POOR CYLINDER
ADJUSTMENT
R NAIL ENT
~-— EVERY moms?“
COUPLE OF
SHEETS FLAPPING PAPER

The cause and effect diagram was used to identify the reasons for
problems. The cause and effect diagram for each printing quality item
was included in the knowledge base of the program.

The decision rule example for the doubling problem is shown below in
figure 4.5.

50

FIGURE 4.5 DECISION RULE EXAMPLE : DIAGNOSIS OF DOUBLING PROBLEM

RULE right and left for double
IF printing product doubles at right and left
THEN doubling problem for right and left IS ok

AND DISPLAY light and left
DISPLAY right and left
Please check the following items for doubling:

1 move nail shalt at right and left
2 vibration of machine
3 machine not at horizontal level

* Click Continue *

The knowledge base for this doubling problem diagnosis asked the user
about the doubling condition. The user chooses a condition from the

selection menu :

1) printing product doubles with big difference

2) printing product doubles at right and left

3) printing product doubles at incline

4) printing product doubles at every one sheet

5) printing product doubles every couple of sheets

Based on the user's answer, the knowledge base gave possible reasons
and suggestions for the doubling problem. If the user chose big
difference as near doubling condition, the knowledge base gave the
following reasons and suggestions for the doubling problem by using the
DISPLAY command.

5 1
Please check the following items for doubling :

1. cylinder arrangement
2. poor nail adjustment
3. machine part

4. form of bearing

5. gear

The knowledge base for this section was taken from Offset Printing
Machine (T akayanagi, 1986). Critical information for evaluating the
overall quality of color reproduction on printed products is presented in
Offset Printing Machine (T akayanagi, 1986).

There are four stages in the evaluation of the overall quality of color
reproduction: decide the measurement item for overall quality of color
reproduction, get the conversion diagram which converts the
measurement value for evaluation item to the evaluation point, decide
the weight of each evaluation item, that is, the relative importance in
relation to overall printing quality, and get the overall quality point
according to each measurement value.

The quality points for color reproduction were calculated by using the
following formula.
Y = ); WlPi (1)
where Y : Total quality point (100 points as maximum)
Wi : Weight of evaluation item i
Pi : Evaluation point of evaluation item i

 

52
IV . I
To define the quality of printed products, the evaluation items for color
reproduction had to be measured. The following items, found to be valid
and reliable measures (Takayanagi, 1986), were used for this study.

1. Relative Contrast (RC)
2. Solid density (D)
3. Saturation point (A)
4. Hue error . (I3)
5. Three color ('2)
6. Ash color degree (0)
7. Effective density in the halftone dot (DP)
8. Environs scumming of halftone dot (SD)
9. Dotgain (DG)
10. Shape coefficient in halftone dot (SF)

These items do not have meaning individually, but each item is related to
the others.

HEB-52 E l l' E'l llll'll
The measurement value for each item was not important in itself because

the value didn‘t directly interpret the quality of the printed product.
Therefore, the measured values were converted.

IIEEEZI E I l' E'I
The data were standardized and expressed on an eleven-point scale from
0 to 10 points. To find the value, a diagram for converting to the
evaluation point from the measured value was used.

The conversion diagram for solid density is presented in table 4.1. It can
be seen that from 1.59 to 1.66, solid density was worth 9 points and from

53

1.06 to 1.13, solid density was worth 1 point. Each of the 10 items had
its own conversion diagram and all 10 items were converted from the
measured value to evaluation points. The evaluation points were used as

data in the decision rule for overall quality of color reproduction.
TABLE 4.1 CONVERSION DIAGRAM EXAMPLE: SOLID DENSITY

Evaluation point Value for solid density

 

If less than 1116
1 between 1.06 and 1.13
2 between 1.13 and 1.19
3 between 1.19 and 1.26
4 between 1.26 and 1.33
5 between 1.33 and 1.39
6 between 1.39 and 1.46
7 between 1.46 and 1.53
8 between 1.53 and 1.59
9 between 1.59 and 1.66

10 over 1.66

W

The next step was to decide the weight of each evaluation item. The
evaluation items were not equivalent but the ratios affected the total
quality. (T akayanagi et. al, 1986) To estimate the weight of each item,
multiple regression analysis was applied taking the ten evaluation items
as independent variables and visual evaluation as a dependent variable.
However, since the independent variables were highly correlated with
each other, this process violated one of the assumptions of regression
analysis. Therefore, in order to make these ten evaluation variables
statistically independent, principal component analysis was employed
to combine these variables into a smaller set of composite variables
(statistically independent from each other). The result is shown in the
Table 4.2

54
TABLE 4.2 A Set of Composite Variables

P- F _

 

 

 

 

Z1 _

._ 0.37 -0.31 -0.33 0.28 0.39 -0.12 -0.24 0.30 0.39 0.34
22 0.31 0.43 0.40 -0.31 0.06 -0.05 0.40 0.37 0.21 0.33
23 0.05 0.15 0.21 0.37 -0.06 0.84 -0.18 0.22 0.05 0.04

Each component was interpreted as follows;
where 21 = reproduction for halftone dot shape
22 = low and high density at ink transition place of
printing surface
23 = color difference at place where a repeated
printing was taken
Eighty-six percent of total variance was explained by these three
components.

Finally, multiple regression analysis was applied to the three
components (lndependent variables) on the visual evaluation score
(dependent variables) to assess reliability and validity of the evaluation
items for color reproduction measurements. The third component (23)
was not used because it was not a significant indicator of the dependent

variable.

(TE)
(0)
(A)
('3)
('2)
(G)
(DP)
(SD)
(DG)

 

 

(SF)

55
The result follows:
V = 0.183 21 + 0.247 22 - 0.095 (3)

V : visual evaluation result

 

In this case, the multiple regression coefficient was 0.970. (P<0.01)
Weight of the evaluation was calculated using the following formula.

Wi = 10 " (allli + a2l2i) l 2 (allll + a2l2l) (3)

 

WI : weight of evaluation for each evaluation item i
a1, a2 : coefficient of characteristics 21 and 22.
Ill, I2i : coefficient is indicated in the relationship
between evaluation item i and characteristics 21
and 22.

Evaluation weight calculated by the above formula is indicated in Table
4.3.

56
TABLE 4.3 ABBREVIATION and WEIGHT for EACH EVALUATION ITEM

Evaluation Item Sign Weight

1. Relative Contrast (RC) 1.7

2. She e Coefficient (SF) 1.7
in alftone Dot

3. Scumming of (SD) 1.6
Halftone Dot

4. Dot Gain (DG) 1.5

5. Three Piled up (l2) 1.0
Color Degree

6. Effective Density (DP) 0.6
in Halftone Dot

7. Solid Density (D) 0.6

8. Saturation (A) 0.5
Degree for (G) 0.5
Ash Color

10. Hue Error (l3) 0.3

W
The overall quality point for color reproduction was expressed as a

numerical value, calculated using the following formula,

 

Y = Z wiPI (1)
where Y : Total Quality Point Y = 0 .. 100 point
Wi : Weight for evaluation item i
)3 Wi = 10
Pi : Evaluation point for evaluation item i
Pi = 0 .. 10 point

n : The total number of evaluation items

57
The overall quality scale for color reproduction has a maximum of 100
points. If the number of overall quality points is high, the quality of the
printed product is good.

The next section describes the decision rule for overall quality of color
reproduction.

' n I I r
The last part of the knowledge base for the expert system of print
quality was to check the color reproduction, introduced in the previous
section. Color reproduction was expressed by one score based on
measurement data. (Isono, 1987). Figure 4.6 is an example of a decision
rule that determines overall quality points.

The knowledge base asks the user to input the measured value for
relative contrast, decides the evaluation point for relative contrast, and
calculates the total score for the relative contrast. This procedure is
followed each time so it takes place 10 times to get the overall quality
rating for color reproduction. After examining these procedures, the
knowledge base returns the overall quality rating value which becomes
the index for color reproduction.

FIGURE 4.6 DECISION RULE EXAMPLE : RELATIVE CONTRAST

RULE for density < 0.17

IF have the facts

AND relative contrast < 0.17

THEN #1

AND evaluation point for relative contrastz=0

58

Fig 4.6 (Con't)

RULE for density >= 0.17

IF have the facts

AND relative contrast >= 0.17

AND relative contrast < 0.21

THEN #1

AND evaluation point for relative contrast:=1
RULE for density >= 0.21

IF have the facts

AND relative contrast >= 0.21

AND relative contrast < 0.25

THEN #1

AND evaluation point for relative contrast:=2
RULE for density >= 0.25

IF have the facts

AND relative contrast >= 0.25

AND relative contrast < 0.29

THEN #1

AND evaluation point for relative contrast:=3
RULE for density >= 0.29

IF have the facts

AND relative contrast >= 0.29

AND relative contrast < 0.33

THEN #1

AND evaluation point for relative contrast:=4
RULE for density >= 0.33

IF have the facts

AND relative contrast >= 0.33

AND relative contrast < 0.37

THEN #1

AND evaluation point for relative contrast:=5
RUL-E for density >= 0.37

IF have the facts

A'ND

relative contrast >= 0.37

Fig 4.6 (Con't)

AND relative contrast < 0.41

THEN #1

AND evaluation point for relative contrast:=6
RULE for density >= 0.41

IF have the facts

AND relative contrast >= 0.41

AND relative contrast < 0.45 .

THEN #1

AND evaluation point for relative contrast:=7
RULE for density >= 0.45

IF have the facts

AND relative contrast >= 0.45

AND relative contrast < 0.49

THEN #1

AND evaluation point for relative contrast:=8
RULE for density >= 0.49

IF have the facts

AND relative contrast >= 0.49

AND relative contrast < 0.53

THEN #1

AND evaluation point for relative contrast:=9
RULE for density >= 0.53

IF have the facts

AND relative contrast >= 0.53

THEN #1

AND evaluation point for relative contrast:=10
RULE for getting score

IF #1

THEN sub score is \ ok

AND #2

AND

59

sub total score for density error:=1.7‘ evaluation point for

relative contrast

This section presents organization of the expert system. The diagram for
overall quality consists of fourteen main parts which were discussed in
previous sections. The knowledge base asks the user about the condition
of the printing register. If the answer is satisfactory, the knowledge

base continues on to the section for checking print quality. If not, the
knowledge base goes to the diagnosis section to determine the reasons for
being out of register. Once satisfied with the printing condition, the
knowledge base checks the overall quality. The knowledge base asks the
user six questions, and uses the confidence factors to express overall
quality. Then the knowledge base goes to the next section which is
inspecting the scum condition. The scum section consists of 5 questions.
If the condition of scum is satisfied with a high enough confidence factor,
the knowledge base continues on to the most important element of the
which is checking the condition of color reproduction. The color
reproduction item is important because a very skilled person is required

to check the color in the printing operation. However, when using the
expert system, the user only answers the ten questions concerning color
reproduction, and the knowledge base gives the user the overall quality for
color reproduction.

Figure 4.7 is a diagram of this prototype expert system that is shown in
the below.

FIGURE 4.7 Diagram of Overall Quality for Paper Printing

61

FIGURE 4.7 Diagram of Overall Quality for Paper Printing

 

Overall Quality of Printing 1 < START >

 

 

Printing Register I

 

    
   

 

 

 

 

l Readability of Character [ Diagnose Register

poo, Problem
M

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   
 
 

 

p
I Cleamess of Picture I 00' j Diagnose Letter Problem
Gm
[Odor Problem
* Poor
am
I Roughness of Solid Place I I Diagnose Odor Problem I
Poor
an!
I Curling of Paper Diagnose Roughness
* * poo.- Problem

 

 

 

 

Diagnose Curling

I Doubling of Surface
‘ Problem

it...

 

 

 

 

 

62

h

 

 

 

\I/ r \
’ Diagnose Doubling
I Overall Scumming

 

 

 

 

 

 

 

 

 

Problem
r“ POO!
good \Ié Diagnose Scum
l Offsettlng Problem
and 4|. % °°°'

 

 

 

 

 

 

\
| Surface Mottle I Diagnose Offset
’ poo, Problem
M

 

I Pinhole Problem I Diagnose Mottle

 

 

 

poo, Problem
M

 

 

 

 

 

 

 

 

 

 

 

 

I Ghost Image Diagnose Pinhole
- poor Problem
Goal
Diagnose Ghost
Problem
coma REPRODUCTION |

 

 

< Input the Measurement Value >

3|;

l 1) Relative Contrast

 

J

I 2) Solid Density I

 

 

3) Saturation l
4) Hue Error
V

 

 

 

 

63

 

l 5) Degree for Ash Color

I

6) Three Piled Up Color Degree

 

 

 

 

I

 

l 7) Effective Density in Halftone Dot

 

 

l

l 8) Environs Scum of Halftone Dot

 

 

_i_

9) Dot Gain

I

I 10) Shape Coefficient in Halftone Dot

 

 

 

 

 

 

i V
Overall Quality Point
for Color Reproduction

 

 

 

 

CHAPTER V
EXPERT SYSTEM EVALUATION AND RESULTS

Muslim

The final objective was to evaluate the expert system for evaluating
color reproduction, a component of the expert system for paper printing
quality. The printing samples were prepared by Ueda Printing 81 Paper
Box Co., Ltd. in Japan and the instruments were manufactured by
Dainippon Ink and Chemicals, Inc. and Mitsubishi Heavy Industry.

The samples prepared by Ueda Printing Company were normal products
that can be seen at any food shop and super market in Japan. The
samples were selected by the printing supervisor. The printed samples
were measured at Dainippon Chemical and Mitsubishi Heavy Industry to
get data needed for the expert system. The next step was for the
printing experts to perform visual inspections. Fifty printing experts

were chosen for the evaluation. The system was evaluated by
comparing the results from the expert system operated by a non-expert
and with the results developed by printing experts.

E | . 'I l' I II D |
Two printing samples were evaluated for this expert system. The ideal
approach to this evaluation would have been to use as many printing
experts as possible and for each person to perform the evaluation in the
morning and afternoon or on different days to judge the reliability of

the experts. However, because of time limitations and the heavy work

64

65
loads, only 10 printing experts participated in this portion of the

comparison.

If the correlation was high, the prototype expert system for color
reproduction might see practical use in the near future. If the
correlation was low, other factors besides the ten selected

measurement items might be considered.

I r r
The printing samples measured for evaluation were picture patterns
depicting Japanese tea and rice crackers. Printed products which can be
seen at super markets and food Shops were selected to make the
evaluation realistic. Each sample was determined to be “the highest
quality', "high quality', ”medium quality', “low quality', and “the lowest
quality' by a printing expert.

For each sample, the ten different variables that are illustrated in the
figure 5.1. were measured by the densitometer and picture processing
analysis instrument. The data can be found in Appendix C. Each item
measured by the instrument was input to the expert system when
requested by the knowledge base for color reproduction, and then overall
printing quality point for color reproduction was calculated. This
knowledge base was made up of 124 decision rules and 65.

To evaluate the prototype expert system, the ten samples were further
evaluated by the printing experts. Fifty printing experts participated in

the evaluation giving each evaluation a maximum often points. The

66
evaluation of printing quality for color reproduction took place in two
stages: the comparison between the printing experts' scores and the
expert system's scores, and the correlation coefficient between scores

from the expert system and printing experts.

FIGURE 5.1 TEN VARIABLES FOR COLOR REPRODUCTION

1) Relative Contrast - evaluates how closely the printed product
resembles the proof sheet. ‘

2) Solid Density - checks the highest density in the printed product.

3) Saturation - evaluates the degree of color inferiority for the
printed product. The color of the printed product is usually inferior
to the proof sheet.

4) Hue Error - evaluates the degree of color inferiority for the printed
product, similar saturation.

5) Degree for Ash Color - evaluates the degree of color impurity in
comparison to the proof sheet.

6) Three Layered Color Degree - evaluates the degree of color
inferiority for the printed product.

7) Effective Density of halftone Dot - evaluates the halftone dot's
inside density reproduction. The halftone dot’s density profile is
distorted in comparison to the ideal.

8) Environs Scum of Halftone Dot - evaluates the halftone dot's
circumference density reproduction.

9) Dot Gain - evaluates the degree of halftone dot's area in comparison
to the proof. The area of the printed product's dot is larger
than the proofs.

1 0) Shape Coefficient in Halftone Dot - evaluates the degree of halftone
dot's outline reproduction.

MMW
Table 5.1 and 5.2 present the measurement data for each sample. Any

67

item with over 70 points satisfies the minimum quality standard.

TABLE 5.1 QUALITY RATINGS BY THE EXPERT SYSTEM

 

 

 

 

SAMPLE : JAPANESE TEA

#1 #2 #3 #4 #5
Relative Contrast 0.396 0.348 0.376 0.376 0.388
Solid Density 1.207 1.13 1.093 1.18 1.137
Saturation 0.96 0.95 0.85 0.95 0.92
Hue Error 0.207 0.204 0.212 0.208 0.212
Three Piled u 0.24 0.52 0.42 0.46 0.48
Ash Color of ree 14.87% 12.85% 13.57% 14.50% 14.63%
Effective Density 21.83 20.78 20.62 21.97 22.38
Environs Scumming 0 4.22 5.48 2.95 4.29
Dot Gain -1.53 .083 -3.67 -1 .67 -3.17
Shape Coefficient 1.366 1.468 1.327 1.501 .406
TOTAL POINT 78.5 69.1 72.2 72.3 70.7
TABLE 5.2 QUALITY RATINGS BY THE EXPERT SYSTEM
SAMPLE : RICE CRACKER

#1 #2 #3 #4 #5
Relative Contrast 0.318 0.326 0.243 0.236 0.306
Solid Density 1.127 1.037 1.287 1.43 1.48
Saturation 0.93 0.92 1 .05 1 .14 1 .19
Hue Error 0.199 0.206 0.194 0.192 0.198
Three Piled u 0.21 0.35 0.12 0.23 0.2
Ash Color of ree 13.42% 13.09% 13.46% 13.44% 13.86%
Effective Density 22.15 23.24 21.94 21.21 21.87
Environs Scumming 0 2.67 6.52 5.92 12.25
Dot Gain -3.53 1.07 6 13.03 16.97
Shape Coefficient 1.434 1.655 2.574 2.2 1.946
TOTAL POINT 73.8 65.5 56.9 58.1 57.5

68
Table 5.1 and 5.2 present the scores for color reproduction obtained
from two sets of samples. As the table shows, in sample #1 there
was little variation in the overall score between samples. The
difference between the highest quality and the lowest quality was only
9.4 points (the score of the highest quality is 78.5 and the lowest score
is 69.1). Additionally, the overall score between quality level #2 and
quality level #5 was only 3.2. Sample #2, on the other hand, shows
larger differences between samples. The score difference between the
highest quality and the lowest quality was 16.9. These results indicate
that sample #2 would be evaluated more easily by the experts' visual
inspections than sample #1 because of the overall score differences. It
also indicates that the correlation coefficient would be higher for

sample #2 than for sample #1 because of the easy inspection.

Four items, quality level #1, #3, #4, and #5 from the sample #1 had
more than 70 points, the minimum acceptable quality standard, so the
overall quality of these four items was good. On the other hand, only
one printing item, quality level #1 from sample #2, satisfied the

quality standard so the overall quality of the rest of the items, quality
level from #2 to #5, were not good.

IIDEII'IE'I' E |

As discussed in the previous section, the correlation coefficient was
calculated based on the points obtained from the printing experts and
the expert system to evaluate the expert system for color reproduction.
The score obtained from the expert system was explained in the

69
previous section. The next thing to do is to examine the visual

evaluation.

There were two goals for the visual evaluation. The first was to collect
data needed to calculate the correlation coefficient between the visual
evaluation and the expert system. The printing experts rated the overall
quality of each of the ten samples by comparing them to the standard
printed sample. The score of the visual evaluation was a maximum of
10 points. The other task was to examine the reliability of the printing
experts. As previously discussed, the reliability of the printing

experts was examined by having the same individuals the ten printed
samples twice in the same day. Fifty people contributed data for
calculating the correlation coefficient between the visual evaluation
and the expert system. Ten of the fifty participated in examining the
reliability of the printing experts. The printing experts who

participated in this evaluation were gathered from the Ueda Printing
Paper Box Co., Ltd., Dainippon Ink 8: Chemical, Inc, Sumida Paper
Industry, and Total Packaging.

W

Table 5.3 presents the summary of the visual evaluation points obtained
from the fifty printing experts. As the table shows, the mean and
standard deviation were calculated based on the visual evaluation
ratings shown in Appendix 8.

Many printing experts gave few high visual points for sample #1 in

comparison to sample #2. In fact, the average rating of each sample

70
was 6.596 for sample #1 and 5.6 for sample #2: In other words, the

overall quality of sample #1 was higher than sample #2. The expert

system and the visual evaluation agreed on the score for this inspection

point.

Most of the printing experts felt that sample #2 was easier to inspect
than sample #1. This may have been because sample #2 had a greater
differences in quality. The data in table 5.3, show that there was more
visual difference between each item in sample #2 than in sample #1. In
summary, the printing experts found quality differences between the

#2 samples similar to the differences shown by the expert system.

Table 5.3 OVERALL QUALITY POINT BY EXPERTS’ VISUAL EVALUATION :
SAMPLE OF JAPANESE TEA AND RICE CRACKER

 

 

 

 

SAMPLE #1 SAMPLE #2

EEVEIT MW W MEAN SD
#1 7.92 1 .482 7.78 1 .682
#2 7.02 1.879 6.9 1.344
#3 5.3 1 .555 4.66 1 .479
#4 6.58 1.642 5.1 1.619
#5 6.16 1.754 3.56 1.643
Average 6.596 1362 5.6T 1.5%
IIDZ Blil'I'l [II E'l' E |

The second purpose of the \n'sual evaluation experiment was to examine
the reliability of the printing experts. This was accomplished by

asking each expert to evaluate the ten printed samples two times in the
same day (morning and afternoon). Table 5.4 presents the results. The
visual evaluation point difference between morning and afternoon can

71

be seen Appendix E. It was important to examine the reliability of the
printing experts before calculating the correlation coefficient between
the expert system and the printing experts. If the reliability of the
printing experts was low, the correlation coefficient between expert
system and the printing experts would have also been expected to be
low. From table 5.4, the reliability of the experts when evaluating
sample #1 was 0.519. For sample #2, the reliability was 0.457 and
for sample #1 plus sample #2, it was 0.490. The data shows that the
reliability of the printing experts was relatively low. In other words,

to some extent, the visual inspection was not very accurate. It should
be emphasized that there were only ten individuals used to examine the
reliability of the printing experts. So, the estimation of the reliability

may be inaccurate because of the small sample size.

Table 5.4 RELIABILITY OF PRINTING EXPERTS

 

CORRELATION COEFFICIENT
Sample Reliability
Sampfih 03—19—
Sample #2 0.457
Sample #1 + #2 0.490

 

The overall quality of color reproduction was evaluated by the expert
system and the printing experts and the data were discussed in the
previous section. The correlation coefficient was calculated using
those data. Table 5.5 shows the correlation coefficient between the
expert system and the actual printing experts.

72

Table 5.5. CORRELATION COEFFICIENT OBTAINED FROM OVERALL
QUALITY POINT BETWEEN THE EXPERT SYSTEM AND THE
PRINTING EXPERT

Correlation Coefficient

 

Sample? 0.258
Sample #2 0.652
Sample #1 + #2 0.525

As the table Shows, the correlation coefficient of sample #1 was low

in comparison to sample #2 and sample #1 plus #2. The following
reasons are possible. First, the reliability of the printing experts was
found to be relatively low. Vlsual inspection skill is not highly reliable.
Second, there were no only small quality differences between samples
#1, as shown Table 5.1 and Table 5.3. It is difficult for the printing
reliably identify differences inspecting samples without big quality
differences. Third, the overall quality of sample #1 was relatively
higher than sample #2. Four items of sample #1 exceeded the quality
standard of 70 points. The printing experts have good ability for
inspecting defective samples. However, it is difficult for them to
evaluate satisfactory quality samples. The correlation coefficient for
sample #2 was relatively high: in other words, the printing exports has
the ability to ascertain differences between low quality samples.

It appears that the expert system may be able to replace printing
experts for inspecting the paper printing quality by setting the quality
threshold of the expert system at 70 points. The printing experts gave a
higher rating to the samples having the overall quality points at 70.

The expert system can be used for inspecting paper printing quality.

CHAPTER VI

CONCLUSIONS AND IMPLICATIONS
WM
Wt!
In the paper printing industry, the number of experts is gradually
decreasing and the heavy work imposed on experts for inspecting paper
printing quality is gradually increasing. An alternate to the printing
expert is required to reduce the experts' heavy work load. An expert
system was built as to substitute for the printing experts. The expert
system evaluated color reproduction of the printed products by
calculating and expressing a single number. The expert system examined
many inspection points to evaluate product quality. The expert system
diagnosed printing problems and made suggestions for solutions. By
using the expert system, an unskilled worker can evaluate printed paper
quality as well as a printing expert. The expert system reduces the
experts' heavy work.

IEIEZ II QI' |'
One objective of this research was to identify, quantify, and classify

the knowledge which could be used to evaluate paper printing quality.
This procedure, knowledge acquisition, required detailed preparation, a
lot of time, and much investigation to succeed. The person who acquires
knowledge must identify, quantify, and classify the lnfonnation. The
lnfonnation becomes the center of the knowledge base and key factor for

achieving the research.

73

The sr
evalu
lniorr

0000

I93;

74

The second objective was to develop a prototype expert system for

evaluating the overall quality of the printed products based on the

lnfonnation obtained from knowledge acquisition. The following

conclusions were drawn.

1. The expert system may substitute for the printing experts.

2. The expert system will reach the same conclusion as the experts.

3. The expert system can evaluate color reproduction of printed
products by express the evaluation as a numerical figure.

4. The expert system can examine subjective inspection items, using
the confidence factor that expresses the user confidence in the fact.

5. The expert system can diagnose printing trouble so that the possible
reasons and suggestions for the printing trouble are expressed.

The third objective was to evaluate the expert system for inspecting

paper printing quality. Due to time and budget constraints, only the

color reproduction section was evaluated. However, color reproduction

requires the most expert knowledge. The following conclusions reached.

1. The reliability of the printing experts was relatively low.

2. It is difficult for printing experts to inspect printed products with
small quality differences.

3. The printing experts are more effective when working on products

 

The expert system was shown to be a possible replacement for printing
experts. From the perspective of practical applications and further
research, the following points are considered.

75
1. The threshold of the overall quality point for color reproduction was

set at over 70 points.

2. To reduce the time to measure each evaluation item, the
densitometer and the picture processing analysis instrument may be
installed in the printing machine so that data is automatically
measured.

3. Because of the time and budget constraints, the researcher could not
classify the subjective items such as cleamess of picture, surface
mottling, and scum problem of the printed product. More work is
needed to classify the subjective items as objective items.

Since expert system development is still a new area, many applications
of expert systems have not been developed. The expert system for
printing quality control has not reached the level of practical application.
Further knowledge acquisition and other development must continue for
the system to reach the level of practical application.

APPENDIX A

OVERALL QUALITY POINT BY EXPERT SYSTEM

 

 

 

 

SAMPLE #1 JAPANESE TEA

#1 #2 #3 #4 #5
Relative Contrast 0.396 0.348 0.376 0.376 0.388
Solid Density 1.207 1.13 1.093 1.18 1.137
Saturation 0.96 0.95 0.85 0.95 0.92
Hue Error 0.207 0.204 0.212 0.208 0.212
Three Piled up 0.24 0.52 0.42 0.46 0.48
Ash Color of Degree 14.87% 12.85% 13.57% 14.50% 14.63%
Effective Density 21.83 20.78 20.62 21.97 22.38
Environs Scumming 0 4.22 5.48 2.95 4.29
Dot Gain -1.53 -0.83 -3.67 -1.67 -3.17
Shape Coefficient 1.366 1.468 1.327 1.501 1.406
Overall Point 78.5 69.1 72.2 72.3 70.7
SAMPLE #2 RICECRACKER

#1 #2 #3 #4 #5
Relative Contrast 0.318 0.326 0.243 0.236 0.306
Solid Density 1.127 1.037 1.287 1.43 1.48
Saturation 0.93 0.92 1.05 1.14 1.19
Hue Error 0.199 0.206 0.194 0.192 0.198
Three Piled up 0.21 0.35 0.12 0.23 0.2
Ash Color of Degree 13.42% 13.09% 13.46% 13.44% 13.86%
Effective Density 22.15 23.24 21.94 21.21 21.87
Environs Scumming 0 2.67 6.52 5.92 12.25
Dot Gain -3.53 1.07 6 13.03 16.97
Shape Coefficient 1.434 1.655 2.574 2.2 1.946
Overall Point 73.8 65.5 56.9 58.1 57.5

76

APPENDIX B

OVERALL QUALITY POINT BY EXPERTS' VISUAL EVALUATION

 

SAMPLE #1 JAPANESE TEA

EXPERTS #1 #2 #3 #4 #5
#001 6 7 5 4 6
#002 7 4 5 9 6
#003 7 8 6 7 8
#004 7 8 5 7 5
#005 7 9 6 8 9
#006 9 9 6 7 8
#007 9 8 4 6 7
#008 10 9 5 7 4
#009 9 8 6 8 7
#010 6 9 5 8 4
#011 6 8 10 4 2
#012 9 7 4 7 6
#013 8 9 7 6 5
#014 10 9 6 8 7
#015 4 5 8 6 2
#016 8 9 6 5 7
#017 9 8 5 4 6
#018 6 6 5 8 7
#019 9 8 5 4 3
#020 8 9 6 7 5
#021 9 5 4 7 8
#022 9 7 5 6 8
#023 7 8 6 9 5
#024 8 6 5 9 7
#025 8 9 7 5 6
#026 9 8 6 7 5
#027 8 6 7 5 9
#028 10 4 2 6 8
#029 8 9 6 7 5
#030 8 9 5 6 7
#031 5 7 6 4 8
#032 9 5 4 7 8
#033 5 8 3 3 5
#034 8 3 2 4 6
#035 7 8 6 9 5
#036 9 7 5 6 8
#037 9 7 6 8 5
#038 8 6 5 9 7
#039 9 6 6 8 7
#040 9 10 6 8 7

77

   

SAME
NE

#04
#04
#04
#04
#04
104
#04
104
504
#05

78

 

SAMPLE #1 JAPANESE TEA

EXPERTS #1 #2 #3 #4 #5
#041 9 5 5 8 6
#042 7 8 3 6 4
#043 7 4 5 9 6
#044 9 5 5 8 6
#045 9 6 6 8 7
#046 9 10 5 6 6
#047 10 4 2 6 8
#048 8 3 2 4 6
#049 4 5 8 6 2
#050 8 6 7 5 9

79

OVERALL QUALITY POINT BY EXPERTS' VISUAL EVALUATION
SAMPLE #2 RICE CRACKER

QUEENS

3".

#2

at
(A)
=11:
A

=11:
01

 

#001
#002
#003
#004
#005
#006
#007
#008 1
#009
#010
#011 1
#012
#013
#014 1
#015
#016
#017
#018
#019
#020
#021
#022
#023
#024
#025
#026
#027
#028
#029
#030
#031 1
#032
#033
#034
#035 1
#036
#037
#038
#039
#040

VGh(DOCVm03°NOOOOO‘DCVQQOOOOmoonmGO$wao#10
00!VOGOQ§¢DN¢DVGNVNVGONOV¢OVVVMV¢O¢DQVOQGVO«h
roars-oralhumamamwammmarehhmwwmmmmmawummmmwmuw
10¢»hOGMGWQQUIOUIMGQOUIUIONUIOOUIO’hOOUImNQQrbNUI#O)¢DNO)

mwwtnSObNSarwt-NM-etse-hummdmdewmheoamwhmmammm

80
SAMPLE #2 RICE CRACKER

ENHENS

=1:
.41

#2

=It:
(0
=1:
.h
:it:
0!

 

#041
#042
#043
#044
#045
#046
#047
#048
#049
#050

caato~4¢>mrooo¢im
cporm-qtsxlmtnslu

d

mmmmwmwarmm
hQCJIPO’VCJt-Rmm
wmthm-‘whh

APPENDIX C

10 EVALUATION VARIABLES MEASURED BY DENCITOMETER AND
IMAGE PROCESSING ANALYSIS INSTRUMENT

1. RELATIVELY CONTRAST

 

 

 

 

SAMPLE #1 JAPANESE TEA

#1 #2 #3 #4 #5
Wall 0.404 0.339 0.311 0.396 0.373
Magenta 0.411 0.354 0.358 0.391 0.411
Yellow 0.379 0.35 0.362 0.367 0.363
Block 0.366 0.349 0.3338 0.351 0.365
Mean 0.3955 0.348 0.376 0.376 0.366
SAMPLE #2 GERBER SENBEI

#1 #2 #3 #4 #5
Cyan 0.267 0.302 0.301 0.262 0.355
Magenta 0.366 0.324 0.174 0.276 0.274
Yellow 0.226 0.221 0.245 0.221 0.229
Block 0.392 0.457 0.252 0.163 0.366
Mean 0.316 0.326 0.243 0.236 0.306

81

2. SOLID DENSITY

82

 

 

 

 

SAMPLE #1 JAPANESE TEA
#1 #2 #3 #4 #5
Cyan 1.14 1.06 1.06 1.06 1.02
Magenta 1.24 1.13 1.06 1.28 1.24
Yellow 1.24 1.2 1.16 1.2 1.15
Mean 1.207 1.13 1.093 1.18 1.137
SAMPLE #2 GERBER SENBEI
#1 #2 #3 #4 #5
Cyan 1.16 0.96 1.43 1.41 1.41
Magenta 1.29 1.11 1.49 1.52 1.64
Yellow 0.93 1.04 0.94 1.36 1.4
Mean 1.127 1.037 1.287 1.43 1.48

83

3. SATURATION

 

 

SAMPLE #1 JAPANESE TEA
#1 #2 #3 #4 #5
0.96 0.95 0.85 0.95 0.92
SAMPLE #2 GERBER SENBEI
#1 #2 #3 #4 #5
0.93 0.92 1.05 1.14 1.19

84

 

 

 

 

4. HUE ERROR
SAMPLE #1 JAPANESE TEA
#1 #2 #3 #4 #5
Cyan 0.168 0.177 0.171 0.166 0.176
Magenta 0.427 0.412 0.438 0.432 0.422
Yellow 0.0258 0.024 0.0258 0.027 0.0377
Mean 0.2069 0.2043 0.2116 0.2083 0.2119
SAMPLE #2 GERBER SEIiBEl
#1 #2 #3 #4 #5
Cyan 0.18 0.176 0.16 0.158 0.174
Magenta 0.398 0.412 0.402 0.402 0.397
Yellow 0.0212 0.0285 0.0212 0.015 0.0222
Mean 0.1997 0.2055 0.1944 0.1917 0.1977

85

5. THREE PILED UP COLOR DEGREE

 

 

 

 

SAMPLE #1 JAPANESE TEA

#1 #2 #3 #4 #5
l-l-L 0.28 0.59 0.48 0.53 0.51
M-L 0.16 0.34 0.25 0.34 0.41
Length 1.2 cm 2.6 cm 2.1 cm 2.3 cm 2.4 cm
Three Pilec 0.24 0.52 0.42 0.46 0.48
up Color
Degree
SAMPLE #2 GERBER SENBEI

#1 #2 #3 #4 #5
l-I-L 0.24 0.4 0.04 0.21 0.23
M-L 0.11 0.22 0.03 0.01 0.12
Length 1.05 cm 1.75 cm 0.6 cm 1.15 cm 1.0 cm
Three Pilec 0.21 0.35 0.12 0.23 0.2
up Color

Degree

86

6. DEGREE FOR ASH COLOR

 

 

 

 

SAMPLE #1 JAPANESE TEA

#1 #2 #3 #4 #5
Cyan 12.04 10.23 11.63 11.76 12.37
Magenta 2.36 2.68 2 1.57 3.15
Yellow 0.854 0.72 0.776 0.811 0.849
Red 2.174 1.399 0.758 1.418 2.21
Green 26.613 19.12 20.97 25 24.17
Blue 45.16 42.98 45.28 46.46 45.04
Mean 14.87% 12.85% 13.57% 14.50% 14.63%
SAMPLE #2 GERBER SENBEI

#1 #2 #3 #4 #5
Cyan 8.257 8.602 8.09 8.696 8.696
Magenta 2.38 0.869 2.79 2.041 1.887
Yellow 0.957 0.857 0.957 0.677 0.667
Red 1.515 2.19 1.43 1.852 2.47
Green 24.55 24.32 26.47 25.36 25.9
Blue 42.86 41.74 41.04 42.03 43.54
Mean 13.42% 13.09% 13.46% 13.44% 13.86%

7. EFFECTIVE DENSITY OF HALFTONE DOT

 

 

 

 

SAMPLE #1 JAPANESE TEA

#1 #2 #3 #4 #5
Cyan 25.97 26.91 25.95 26.83 27.1
Magenta 21.4 21.51 22.86 23.79 23.96
Yellow 24.15 22.84 22.67 25.67 23.99
Block 15.8 11.85 11 11.6 14.38
Mean 21 .83 20.78 20.62 21 .97 22.38
SAMPLE #2 GERBER SEINBEI

#1 #2 #3 #4 #5
Clan 24.19 24.88 25.51 27.63 26.52
Magenta 22.39 22.87 22.17 20.45 25.92
Yellow 25.66 25.29 26.84 22.12 22.29
Block 18.34 19.93 13.24 14.63 12.74
Mean 22.15 23.24 21.94 21 .21 21 .87

88

8. ENVIRONS SCLMMING OF HALFI' ONE DOT

 

 

SAMPLE #1 JAPANESE TEA
#1 #2 #3 #4 #5
0 4.22 5.48 2.95 4.29
SAMPLE #2 GERBER SENBEI
#1 #2 #3 #4 #5
0 2.67 6.52 5.92 12.25

9. DOT GAIN

 

 

 

 

SAMPLE #1 JAPANESETEA

#1 #2 #3 #4 #5
Cyan -3.7 -8.7 -9.9 -7.5 ~6.8
Magenta -0.7 1.7 -3.6 -1.5 -3.5
Yellow -0.2 4.5 2.5 4.2 0.8
Mean -1.53 -0.83 -3.67 -1.67 -3.17
SAMPLE #2 GERBER SEINBEI

#1 #2 #3 #4 #5
Cyan -6.3 -6.2 -3.3 0.4 -3.4
Magenta -3.4 1.7 13.7 11.7 25.3
Yellow -0.9 7.7 7.5 27 29
Mean -3.53 1.07 6 13.03 16.97

10. SHAPE COEFFICIENT OF HALFTONE DOT

 

 

 

 

SAMPLE #1 JAPANESE TEA

#1 #2 #3 #4 #5
Cyan 1.332 1.196 1.208 1.34 1.296
Magenta 1.229 1.476 1.285 1.36 1.299
Yellow 1.537 1.733 1.488 1.802 1.623
Mean 1.366 1.468 1.327 1.501 1.406
SAMPLE #2 GERBER SEINBEI

#1 #2 #3 #4 #5
Cyan 1.409 1.292 1.518 2.074 1.638
Magenta 1.333 1.506 3.184 2.977 2.036
Yellow 1.561 2.167 3.021 1.55 2.163
Mean 1.434 1.655 2.574 2.2 1.946

GRAPH OF THE RELATIONSHIP BETWEEN EXPERT SYSTEM

APPENDIX D

EVALUATION AND EXPERTS' VISUAL EVALUATION

Insual Evaluation Point

 

 

 

 

Sample #1 Japanese Tea
16 2 1 - E - - . . i-
9.

o e
. 6 o e ‘7
‘ V
" e 0 *
6. 8 O r
. O O ’ . -
41 0 9 v °
1 I
3'1 o e O I'
. l
24 O O .
11 .
0 r f . . . . . f
68 70 72 76 78 80

74
Score from Expert System

91

Vlsual Evaluation Point

92

 

Sample #2 Rice Cracker
1n - L - - I A L A t I

d I
9" . . P

1 t
al 0

   

 

 

 

O b
O f
T
t
1+ 0

1 O »

0 v I v I I I I I f I v I 1 r I I fi
56 58 60 62 64 66 68 70 72 74 76

Score from Expert System

Wsual Evaluation Point

Sample:

Japanese Tea and Rice Cracker

 

 

 

 

' 56 I 58 60 62 64

 

I T I ' I ' I ' I

66 i 66 ' {b ' 72 74 76 7a ' so
SamaMmuBmutanmm

Afternoon Visual Evaluation Point

94

Reliability of Printing Experts
Sample #1 Japanese Tea

    
 

O

perfect correlation

 

0 1 2 3 4 5 6 7 8

 

 

 

i I I fir v f I I I l T

q

0+

10
Morning Visual Evaluation Point

Attamoon Visual Evaluation Point

95

Reliability of Printing Experts

 

 

 

 

Sample #2 Rice Cracker
10 ~ . A L A A . : ,

1 i
9. o O o it
84 o o o O .
L o C) O O -
6- o o
5‘ o o 0 ~
4‘ O O o o o o b
34 O o r
21 L

. perfect correlation .
14 .

1 l
o I I I I I I T T I I I I f T I I I

0 1 2 3 4 5 6 7 8 9 10

Morning Visual Evaluation Point

Ailemoon Visual Evaluation Point

96

Reliability of Printing Experts

 

 

 

 

Sample: Japanese Tea and Rice Cracker

10 A A A A A A A A A o

9. o O O 0 CL
. i
8‘ O o O O Q i-
7: o C) C) ’
6. O O O O 0

5‘ o o O o r-
4: O 0 o O o o L
3} C) () C) o t
2: o L
U perfect correlation L
0 . 1 . . r . . . f . . . '
O 1 2 3 5 6 7 9 10

4
methkmfl

Evaluation Point

APPENDIX E

VISUAL EVALUATION SCORE TAKEN FROM SAME PERSON

 

 

AT MORNING AND AFTERNOON
SAMPLE #1 JAPANESE TEA
EVALUATED BY MORNING
EXPERTS #1 #2 #3 #4 #5
#001 6 7 5 4 6
#002 7 4 5 9 6
#003 7 8 6 7 8
#004 7 8 5 7 5
#005 7 9 6 8 9
#006 9 9 6 7 8
#007 9 8 4 6 7
#008 10 9 5 7 4
#009 9 8 6 8 7
#010 6 9 5 8 4
EVALUATEDBYAFTERI‘DON
EXPERTS #1 #2 #3 #4 #5
#001 5 7 6 4 8
#002 5 8 3 3 5
#003 7 8 6 9 5
#004 9 7 6 8 5
#005 9 6 6 8 7
#006 9 10 6 8 7
#007 8 3 2 4 6
#008 9 10 5 6 6
#009 9 5 5 8 6
#010 7 8 3 6 4

97

98

SAMPLE #2 RICE CRACKER
EWMLMWEDENIMNNWNG

ennsns #1

=It:
N
=It:
00

it
.5

 

#001
#002
#003
#004
#005
#006
#007
#008
#009
#010

d
m¢hc3#-N¢D¢D¢i#¢o

airpoouoooooowouts
wammmmwmww

EWMAMWEDEWVWWERNOON

ennsns #1

4!:
N
it
(A,

GO#NUI#QODMOD

RI:
A

=11:
0|

hUIOO#O)UI#UImN

:3:
0|

 

#001
#002
#003
#004
#005
#006
#007
#008
#009
#010

QVO#\I#VOOG
DOOVOUIVVUIN
h‘lUlthOOUIOOO,

wmmhomhqhm

###QOOQOIOD#

APPENDIX F
KNOWLEDGE BASE EXAMPLE RULES

lgaag¢§gggca

Egagvlua Lia-again
hggg‘gga
ggzgeiga

dog
assign!!!
iglgeggi

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

.gl‘aggaia
8583(ggggii
ggggggi
.glia
Fading-38:851.

32.53!
xo-ulaguniuoauuiuiuouiil
8338.803

luau-Ea

gagglguoaziuiuoulgié

1348.31: .Usgilglagvaflua
igogc‘gfl.‘
lanai-.8368. SEQ-U 8
:53
ggag on
I! all!!! Egg-83E .n.
E
Equal... g :5
gig
ESEEEE £83 an
:85
0000...... E I: g i oooooooooooo g g 8 g .0 a
£85833 an
5085"; 2.n
gigiUSSEuEUfl-uii Egg .-

 

U‘g m Egg 2

99

100

382.03 “Sal-8.5.0 a!
8.3-8 £338.03 :9:- I.» I!

3.0 3» ""3 3 moo-“M- on“: a" 33.

.6033 £83.90 a
hog no» 333-3350 3

~33! soon 3 893-3 3 «o I333 30 In.
3.0 c:- 36 II»!- 3 bound-on 83.5 no 332- s

sou-«.8388»..- III

.388 non-«non 38.38 :30 3
~333- uo 8o 3 .333- 338 no 63an I: 3

3.86388333388332-

uouoqoouuoufl. a
38388333833u334

.gaggsagggfig
aggb bugsgggl

.3
asggaggghea. .g
gaugggigg

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

393 3 I": ~33! 3

86363333303

0.5.8!
lounguinu'ng'ooqu!

no 98.333 ova-«33833263933
ulna unoooqnouuoauouio 3:...qu 3
3i. 3.8328833!“

noun-hath 3:15.50 3
3353: 3

83 .328 8. 83o! hSugars-I...
Incas-0163.33.33! 3
33:36:88-0333

noun-33333350!
0.5.8!
«o 3 «3»! son I”: In.“ In."
.3 “II-33.334 noon-n
'03 uo 3on8. cud-38.58.33:

.83-033333350 3
«.339 is

do: «333 you 'i .5333!
In333o6uou3o3 «oi—3.. 3
3:38:88333

hog uo 3:33.: 350 3

no 3 3

do 3 «35 no. In: .5933 '5

~33! no 0.6 no. 3 o 3 no»... 9333 3
u3§ 3 In: '03.: a

'00.; no >33: 35. 0!

a. hinge s

so 3 3%! no. I388 nos-«.8 '6

-033 no 86 you 3 o 3 86330 38‘ 3
«33¢. 3 In: .6033 a

n3“ sou Ina-n .6033 36
3o! 3 .33...- uo «so .n

33.. .3 I388 3.3-8 3..
39, 3 no“! no So .—

3 g

«3035 sou In: sou-«3 3
39! you 3i -033 g

.Bsgigglagvuag

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

5:38 .38.. .333 38.33 3:...

101

0038333 3 koala-033333 3

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

£30. b.1335 n
.uoul. 0338‘ fill 03 a

a. S‘Sufl
o Aggv ads 0

.33330833303333333 3

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

 

loo-no.8 no 3338.8 .530 9!

3. 2.38 I...

no 3 39... .3 I388 8838 3
ugsniagsouinas'uuilu 3
36.1 3 8298 .33.... a

loo-no.8 no 5338-8 350 a!
3. 93.3.. 9!

no 3 38.. .8 .388 8.3.2. I...

n33.9- no 86-3303 330333 313.80 3
39... 3 .388 8.3.2. a

noon-.38 no 3.3.1.8.. 350 O!
3. 338 9...

no 3 39... non .388 .338. .IF

al.» you .603 o 3 33.33 3350. o» 9333 I38»! 3
39-. 3 .388 ~33! a

.8836 no >333... 350 O.-

.. scan a

do 3 39..- uon In: 803.3 'F

non-«Ion no 26 non 3 o 3 039-9 3033 no [3‘3 3
39o. 3 .383 ~33...» nil

863633333350 91
33333 3

3338. Bu '3‘ -033! .

.3333 Eaves-.3 3.83380 33. 3
391. 3 In: sou-Eon a

nouns—3... no 3338.3 350 3

E 2.33 3

no 3 33. non .338. 303.3 IF

al.» non .833 o 3 330 non 33386 33.. noon 3

IIIIII
IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
OOOOOOOOOO

30833332313 a
333n333 a

I
IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
O

.6333» :33.
33333330333333 3

3333033333 v
38930396 3
38333333 n
330319336. 3

c.3339 138.
33333333331336.5363 3

a! 3!. 3 33k. 3 00:33! 3» n
03 33 no 3 o

i 3!- lnouuo no» 3

Ci 3!- no 33333 n

33. 3a .3qu a

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

0. kg:-

oaag 3360
33333333333330.9103 3

86.3.. 3. a
8.13.. 3.8 n
«3.3.. no .38. n

0000000000000000000000000000000000000000000000000000000000000000

102

:3 no 8:030. .505 c...- noooo lad—9. non goon 3
no! 0303. hon 3’. 25: I 2
03603 5n HID E 3
0.3—Ul- 35.‘ ‘3‘. :08 in 3
o

2.3 on 393.9 0

u! :3. 93a- 93 vnoon is cola- n
«no... 00:33 noon v

36.3. no no!- a

nil: inc 03 .509 n
oinoluuuonofiuonoaon3

'3 [33:3 3» duo“. 033.

at;

o hadgv “‘8 o

.33 Golan-Eliot: I. 3 30.3 3 53g :33; 3 3

 

“:31. 0L! 3!. no 5: n-

33. noon .—

ily-:3- 2! noon 3

035 :1: non agnou no 36 «—
onalconn lino- ccu ~3n-u no 35 =
3::- 3!. no .53-r8..- .no. as. 53. 3
a! :2 non 3 in. 3 o

u! :1. II :3. no n... n
.338 ii 32. 3! no not. ..
nae-:10 ulna I.» no '53» 3!. o
.65 3! n

v3.66 :3. no :36 on v

93683 :3. no .10.. con n
9:": :3. no 0.8:. 8n n

3!.- 32 no n3: 3

 

 

 

 

II: 3338 I... only .33.

u Aggy 80‘s o

'3 only III g I! 3
an...» 3 :33 son .3333! 3a! been 3 col! 3 3

 

:30 non 5:309— 330. in: —

 

'3 [3.633 I: 2 033.

hug

ongfivfiuso
'uniullglouiuuqifinSIsouun

in"! Ono—n n
33‘ u. haul! all I ill on» a

'3 [3.530- 3 2 00-3.
on. g

o Aggy flag a

'3o3ulnlogl3iu
333.3333 «350. o» con-33.353.59.383 3

81.8 yin-loo 33 u..- ui... 58o. 3

I3" :3 ~0qu ‘1- 6]; 3n! Mqualou no 35 a
.3 313:0. 39 do!” 003—.

Inuiull'g
33033.336333-33333 an

1
i
a
I
€35
at;

I
i

331 9| 0.5—yin no oilfi- u

6"! 33303005333

303 335' non 003°

«36 I .5334:— .i .3 5303830 3a...-
3! 3 303‘ II 335

no! :3 1' 3‘ £33 fill
ado-n no 3!.- u‘? o

.3332:

i

 

'3 .3333 .5 «no.3 col:

:0 g
. .3!qu :38 .

'uuoolau-Ellg
339.53 03.69 Inna-33.38.3300!“- 3 3

0 Click m1") 0

AAAAAAAAAAAAAAAAAAA-AAAAAAAAAAA AAAA A
-—- — — w v ‘7‘ ————v‘-—-—

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
v—v _ -——---—- v - vw ‘— ‘- -wv— — v-v'vvv-
AAA-AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA-AAAAAAAAAA.
v —- vi—vvv— __-- "--V- w—— ' -—--—--

Lea-cwumocmammmm-aum

Ill (wiltty a! War
I! nun-um, at Ma

 

 

 

AAAAAAAAAAAAAAAAAA AAA A
- v-wv‘vw ' -—-—-

AAAA
AAAA

mmmumu cit-net- hutch-stac- "at
no alt-elm 0! picture

11113 Witty cl Our-czar mm

103

It! can: 410”“ loud ”I.

I
I”
all-om

”WWQMMWMMNM?
mmmmmlfl-omxum.

m nun-bum at War
a anna- m m.

 

104

DJ» 30. 0::- 9' Evading». .n
beau-.35 SII no. wooed—ab. una— uuosu .a

13:3 98:; 0.3-Ionian 13.33.203.02

guuszouliilaganuz .n
I: was 38”.: I.» 3 .n
.31. 3 hog. g 8» .38 can: 4

.gd’as

-393 can .8. go «588 .53: Ill! 0308.-

.Eggsg
Sgasggz.ggai¢a

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

 

 

' 93.53 13300.:

.33:

R‘s-Eel:- m

bangs

animus

gonna-Bus ”ghouls-50.138»?!
0.53 003.3. [33355080 8 '00:!“ I:

13.33.383.53

behind—O!

u. a!

cor—:3 :guo'aun‘ga :3
9053 38¢§gu¢u~n¢au§uouuou¢au¢£u mm
giggsulggga :31

gallon-5953355383513 .“

....................................................................

 

I'I ' I'll]- II"- 'I' II"
111
11111111 1 1 1111111 111 1111111111111111111111111111111111111111

E=8g\§=§9 n.
§8§§\§I Quinn: «0

11 11 11
1111111111111111111111111111111111111111111111111111111111 111 1

 

5338-233! :8..- 38539 IE...

35 you Ina; 03.85 .53: ~05. .13qu 85-85

.uauogizauouuuli'OuQ—h as ”a

.un' 0« '3. 3:; 80 3 as Cyan.

.0033!!- on.

'23:. 3:» no. I!!! 03.30 [3.633 89 .30.! 030.9-

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

105

i

a:

an! E
God on o
. .

.gigga‘
.sggugoligglig

hasn‘t gig‘gaggg
éafixiu'uuir

-38.. .328 385-3 380 I...

.800 g i

so an 00-93- gufiun I: 8 9503. no 53350 I: in.
533 no colt-Ono 3

9303. no once-Ono a

g! a canvas. [.353 It .3 95003 .0 333-80 .5 .u

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

.8; 8
aigglufligagg g:
BEE—gagging 05%

1111111111 1 1111 1111 11111111111 11111111111111 11111 111111

a: «a

.ailsag
.EguSOEEEEEE

raggggagag

gagilguflqunlusg‘uousuuglfl!
£35.33 3

nan-i.
inu§§3i«3§5§isu8«uig .-
cg

.asgigglgv Bag

111111111111111 111111111111111111111111111111111111111111

.gagsgiofiv-gg‘
.EagggFIIOOuBna—uahgt

.igliEEBCMIEE

.sgasgai
g.ag:§8§aiflua

'II’II
11111111111111 1111111111 111 1111111111111111111111111111

106

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

253.85%! .838834:

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

agggsgilaggiflug

‘ggggsg
giaagggegaagnuflr

ga'uggg‘oug an. .agggeissafihu

107

 

hgasgggiahagig

.BEEEIAE:

 

C
056.03 92.53
a «
noel: u o .
55.93 ”H.385 9:0. 56.: SSE E‘s-00‘ .
is? at»: we mail. 32'- .uwuqflmnuuvfi 0
«.50 out... .08.:
a lion 5. u no 8:30. Sly—50‘ «0 Inn-3:...” .n
5% val—O E on . no: on» 3 5:3
.0 >5 :5 snag . 5 453 «o 33:16 1:. .u
3.. o anus—Ia hogan Eon . fixl on» uIPE .
l: magi-u! 53:.- uci fill 03 .9: near: .n
.1.— .5 05.: 33:03. I.» 2.3 . no. any 3 :5 q
. «III on. ‘ug £5- 3. .3“; and 0 g g you
00qu oat-on {:8 IF in 3.08. {:8 IF
.933! 35.8- .II.I¢I 03.0!

.21: 'uiun Ii hon 0i. Ina-Illa .q
.asasggigvuag

.sgggaa
IFSBIFSIBEIa-uEgE-Easr

.ggisgfilg
sgaiggggss-uagli

g§8g\gin§n

~55.— ‘ug 3.33.5. $8.503 RE...

108

.EE. asaggvuflui‘
aggggissga

.Bgiasgggufiug!

I
s
s! E E
8‘ on a
. . .
.aigsa‘
.EgnSOEEEIEi—n

pggaflggggsgsn

53339 .533 pl...
038.. .336 38.5.8 .930 III

:3 g is
faislisgnazsaugglfi
Suva-no [.350 ha

53an 953.6 a

.0813 B's-50580.5": nap-33980.1» .u
oi

.IU °9.DF.flnC-iu-II.DDfl Ill. AID-HEIDOV “Una” IIKId.
III IIIIIII II III II pl v I'll
1141111<4<1<¢4q111<<1114<<41411114141444414114141‘444414441‘4414

.gaslgioauggt
.Eagggioauanggn‘

109

68.98.9533!
u3:IEa a»: 3691;598:8996..- 4.

99-93-5993
-393 35.8.; jail-Educ! Gauges-ll

gagiglggsfi. ..
«alga-glucose}!!! ..
laglpiuoulaxlu; .58.

99.33080 on.
In: 93» no. 1899835998 .5320 9 IF .i 03.9-

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

 

 

ga :30 g

a
no In '1 gun—6 I
9930 In: I i 89 9009!. sun:- 3
'13-. [-3.30 a

hula-liguauoulusi .«

.asei!§!§vuag

 

I'-’ I ll'l"
1111111111111111111111 111 11111111111111111111111111111111111111

.aggsgggag
:gggaiilagggg

.agggg
ggaalfigegaagg

'II” 'I’ '| 111 1 1111111111
1111111111111111111 11 11111-10111011011fl111111111 1 11

1 111111111111
111111 1 1111111111111 11111 1111111111111 111

 

 

'1 minimum/mammalian

 

fml Mamba?!”
mma<mymmmmnmu.
mmummmm.

his?!”
a:
m). o
In a. mum. Incu- lot “In. an hull

my:

 

 

 

mammammmmammmm

mummuoounmnnanmmmmm

l1 amt-pmlmwmm
Inn-locum“.-

maxi (muuvwmmnma.

7111‘ Magnum mm- Rd)-
W mung publ- lot too “I Mum
r z

110

 

g! E: I
=; 33
£5 E

, 5 ' 5

EH 55 *3
5 3:5 g g;
igfifisgg

3 53‘
figiisga‘; .
i a 33555 a g 5: -. E5
43.55: a

. 2' 5

Hi ‘ ,2

5 aklii $3 is g g!
5?. 5532:: E as . .3
i=g‘3 "3"32 ii £3 £5 '§ 3
35"“ ::§3§= 3.3. :3 33 i: E.
5922§s§5.é=3 333 ,z‘ a; 2:? =5
382:3":“i '5 §—~ ‘5: 3 2 '3
"eiazééééé 53% i 33% i 5% i an: i 33%
iiéééfiééfié $35: 3332: £322 $32: $33
°_ “ d“ " 3 fl. ; _ . ‘
5§§§§§§§5§E§ €23; mg: mg £233 Si:
::::5:5:::§: i: Is: 3:!22 3:!!! 3.12: 3.!

111

organ
.0 9328 .3338 3.5553813
. 0.1-«v.8 «on.
....................... slat-8.6!}!!! «82388888»
25 «significant? '¢iui§! don-Jun
1.30.8!5?
«I ”In.“ in«ai .......................................................
................................................................ 3.388
.5330. {no-5300.: :0 31.83.09.305:
................................................................ viola-USI-
in:
Sag
. .1338. .38. 5.98 8. II: lion—3538...:

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

 

 

cg
gsvaunu
inns-o
guano. ..............................................
£3109.-
gn‘uig anal-Iona
E-iuuxu ion-5’3
gnu-3.338.
I. 38 ' Sizezugg ..............................................
....................................................... szsulugz luau-500'
. .......
. c339
pill-Bug m .Aggvuozu.
vol-ill»:
lilo-39.3560! unaliguzsugullosgia
gas
.33 y‘all-533.; 5.85.5! ........................................................

gun g 3": non.
iauaciigugsuiogilfi'h” uguil‘tgfiflflin

112

opts-c .6111". “llty

. 3 Hi
E i 2 9 gigs
5:; i g 32:;
i i -5 i i 2 22::
§§§ E g3 $5? 3 i
g 25" E 9i :33 ~ ' g“ i
:53 532%: E E: E}? I 3‘ 2
3:;agéid3 E 5 1 E33 :3 35%; - -
1km: ! :3 g: 3 gay? mg;
WWI“:- 8 iii 2 his 5
§3§§§§=' 5 Si 3 5 5 . fififitfi efifii -
.f:..:3§:-: a! 9 5:3: 5‘ "'“" z""
333.2? 5 f

0 Click (On-ulna: o

It I!” no 10!:

6 may th- taught a! Inn 0.!
1 Ink .1» can. of mu
0 In: 0! null plate

1: 1m (outta: a! pn- m

13 do. u min

'1... chuck tho £01an It- to: um.
I nema- mm

3 can: but“

I poor quads at H” an 10!:

0 poor Inn 910:.

S In: an out

0 your nu u an! out

no you put. alm'o odjm date.

0
m .31

113

IS...- Iso 33.539 .33 mil

8.3383 .990 9!

so 2 83.3- .5338 8. lilac 3325 33an .5 IE
:82... 3

in 3325 a

so. an 83.5. Dav-3.3.13 In! and: noggin-o can .u

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

.EEE
ggagegnuagng

 

IIII"IIIII III I I IIIIIIII'III'IIII II IIII II IIIII IIII
‘11‘I“{““1‘111‘44{4{111‘11‘111444{{1‘I‘11‘11‘114‘1411‘4‘I11‘1‘11‘1

 

IIIIII III IIIII II IIIIIIIIIIIIIIII I II IIII II IIIIIIIIII-
““‘11411‘1111{1“‘1 414144“1“111‘1{11I{‘111“14111“‘111‘1{4‘114

. .383 88»... 8:58 8. lila- 28-5 IE...

:9 n. 83% la!!- 35 I.“ I...
53983 «33»... 0.3.5» a on: 3 21'6- La
83033 «.3qu» in a I: i a

.53-83 :30 O!

a. some 2

no on 53983 .533 .536 9.16. .IF
:1: .I on: 3 i 3

3:26 .I 8.: 9.16. a

0.5.3 g
iuuiuuggagi!
gill-5.3.5158 B
usual-33in:

.5300qu g i

no 5.3 a

so an 53% .533 I33 Bid. in.
.23.. a 3.: 3 in a

can. a on: [5.6. III

a. g s
soaauuogggaggai!
53% ’35-. 0.0-I [and .308 [3.6. 3
533-9 [-353 I396 ‘ug in a
53% .5053 ‘9. [and In! 1a
.33; 95% 0.9. [:36 'ng in .n
.33; 3% .55 in a.“

so: .5835. Id In: on! .«

IV 3

giggligig
B‘Egglgaié

.aseggglngv “as

. on. 913 .83.; .5059. laid

.3660 I6. I on gal-no 80 [3'60 .093 2.: 9.15:3 IF

.533 .533 0.0-I
no 8:988 .533»: '53 .550 ooh—.68 3 ‘133 g

.Eggsg
ggagggenegfiuaglg

0000000000000000000000000000000000000000000000000000000000000000000

114

8:8 .- 65.... 53:8 3 .3 a. in! 3 u. . 8 338
5.33am a. :3... flies... .53
.n 032. .85 38-» a. A
.328 an cons-o 3 3 a an! an col-Io 3 a“ a ................................ c ............ 1.30..
00000000000000000000000000000000 ’d 8 3 8‘ a“. .0.
.................. .3333. Dan—.188 n6
. c an. «080 033. . E 310 In. «.0
............................... ‘uaullo on. g 09: 3' [38:0- 03. 9500.3» noon «.0
......................................... 3:1: 3a.!- 3 a." 3 3 o
n. g
.3»!- .0 5330.0 .88 «A
............................... Pun-.4339 33.... ii .3... .833 .3 I 32389 noon 3.»
#3330833! anal. .85 ad Site-3833 n
830833043 3n n. 3! 3.5 «.0
38.338» n4 3?: igloo-303 « a
3'3 3 I85 «3
E i. 8» 1n and; n n
3598283.“ 5.383353.“
3»! .330.- 88 ad 908.88» a a
'33 an .Uuuo... unqual- n.n .............................. {Jud ........ ”313..—
g in i a.” 8 030 O a. 3 0..
.83-89333; ‘3 a 330 ulllth—Cuiguouk
Bug—us 3.30 10 s udfl ga'u A.“ oooooooooooooooooooooooooooooooooooooooooooooooooooo
~00!- 00 103 «.3 no 2“
adv-ado. 8 8n .I.. Cadil— u.« 33 “as o

2.3030. 3353 an 9038 3 3 a

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

 

 

. .83-:3. in. 100.3 00': undo nan-:5. [3.3.6 .
.9.qu ..... 93 1533:. .u SI. .3323 IF '39- ..3 50.339 noon u." .nd .33 no» a;
IIIIIIIIIIIIIIIIIIIIIIIIIIIII a a
................... fififififl.§..............m .5... fins? 3 isflvafl 3
23.2.. gal-3.8.3:: n 3983358 3: .
803303.16. 0.»
........................................ 8588.33... all In I. :0 ll no»! .5889 «a
E‘s-8.8333338..- . 3R3»!- Ilu a.»
O I. O
.3... .. .3... a... 53...... .. ages... .5 x .3?“ 3 3...... s u... .
E no 5333 I: 9. .
0005.02» non-o o» 9333 [3.3-Io. 3 con-lo 3 3 n 3 I6 uni-Cal. II. n." 09"” ”5.33 M.“
8‘00; :30. u. '33! a.“ 3 r 3 3 0d N 8’ 3 ‘3 Id dd u
83E 520 a. .53.: .4 ........
.8931... £28 an lanai n4 .. ...........................
gunk .0010”: u. .33.- ~.u gidI a. 3 in.
001‘ no! a. I3“ 3.. 93.3930 and 3 O3. .33 31530. In.
Sign. 0» n'uluou '3‘»! sh 3 Id 3 a .................................................................... mm ...... 3
£30.53. £80 3 00.3.
6033-50 9.0 33330 or» 3' 33:3 08-. .. ............ c .33. 83qu
. .... rum ............................. 5 a! sauna:
........................... 3281.»: «III-5 al.-:3 .88 .6 3L «afghan: u."
3 2.38 .3338 a no .38 «a .5 who [all 1»
30% i colic o u n
. 82-3.! 3133.. .6 . 3
35.33.!“- . £63532 .6
. 53.88 .6 ulna-«Ilsa 9n
5:833. .33 I!» 3"“ .6 a: _ o.»
«J 5 ago 3.... .
I: lung “.2: .588 3.3!! al.: I]! 1. girl-N“ 88 n."
0 33—33833 goo-fills» «A

115

J. T
digital—B.-
.528.Ololi§§3.§l!

bgggigsgag

i 303:0 in.

.34.! 888 385-8 .590 a
3...... a... ski... hawks... has
I.- a
no 3 .3 I335 33339 3
I38: 388.! a

g! 3 an! 333 33.3 3 3033.5 no [33139 30 .u

 

1111111111111111 111 11111111 111 1 111111111111111 111111 111 111

8823.8. a”. .gaglgiouoaggc

.3». I»! gonna“ ”gnarl-kofiun . .33 .9: 33383 B .5536. 4

«final-5.1: 1. 35355.33 Egg
Inkling-.53.: nun

coagufiuoflun n." groan-5.332.338.3953!!!—

3§I~35883333 a 3588353333338
338:: £1 a;

lug-503338330838 an: 1 .EBEIBHIDBBB-ulflulg
8:83.833: .

...................................................................

0502‘. i a i“. ..................................................................
33303333333.‘

000000000000000000000000000000000000000000000000000000000000

116

 

0000050 no 030' g g

00 503 0!

do 3 08300030 no. Isa 03000. 05 '3.
0530030 03 .8000» 0 3 30300090 05:!!- 3
.038. .5308! a

000! no 0303 as g

00 g s

so a [28:3 8. .838 8988 I... .0.
05300.00 you .8000.- 0 0g 83300 50058 a
In: 0530030 3

‘9: 3 Canon 3 g

0. g s

:0 3 050000000 03 I300: 030000 010 I...
.5300qu you 80000 0 0‘ 3 3 000030 Ill! 3
-398 9:30.80 a

0009!! go 030' g g

a. g 02

do 3 gun-00:0 03 I800!— 0~300on 05 in.

9.3000qu 3. .8000» 0 0d 00! 3 03040» 0.35-II 3
In: 05300000 a

000“ so 030' 50 g

a. go '0

so In [.3000qu no. 05000.. 031005 05 .F
soul-1:0 0330 0 0. 3 g

.5300qu you .8000.— 0 3 056' lay-3.3 Lu
00330 05303.0 In!

0000.50 00 039' g g

a. g g

:0 3 [3303.0 you '80.» 03:08 05 I...
00:00 :0 ~00qu «fl .8 0. an s

0530300 03 .80! 0 3 0.53 [.3500 Lu
la: 05:00qu 3

0530030 00. 30000 03000. 05 .n

8 3

.53-33 8. all! 35-80 I: 053?!

.Bsaigélgvgg

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

5839253508.
:ggggglgssgggg

 

""I||""| II'|I |"--I’I’ I"|"'Iy' -I |-|I|I " -II III
Al‘11“““““‘1““‘1“1““‘{“1{“‘ “““‘1““‘{‘““‘1“{1“

 

'iI’I'DIII III’IIIIII I|||DD|I ’IIIII DDDIF 'D’||F| I’i
{{‘411‘11‘41‘I‘141‘l41111114l4‘441114‘4‘ ‘1{{1‘1““‘1‘111111{1I41

0 Aggv luau 0

.luuigig
l00u§uui300§§3§0uuu

5

00.543503

.Iblus‘ai'SUaun
81883-5333 a
500:»! n
339000.000. .-

................... Bflnmrw r... ...l: 816:8 I: «II II:

.
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

0 ail-«Ev Jude 0

.‘u‘iiuln'iigg an 03055—0
93000 5300.0 .0830! 5300 k. 00,00 0‘ an

E ink. 08. 0
030000. 0535.3 00 100a 0d uu a
300300.00 05! 0a 3 a
0033 0:000 0033!: 0a 3 u

0 aggv ”do o

.'3 0.6. 000:» II 50005 '0 0a 0005 3 .025
«no 000 «HI 53300.0 .05: 3050. an 90!! 0‘ an

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

Sagacity. 0
0.3000053838001033 3

9
0
E 00a 0« 3 a

£0 0‘ 3 0

n

a

II 0033' 003001: 0a 3 a

”unwantUnthuhluu-

‘uluuunuunxr

m "£1.

0
tannonoroo -

 

 

AAAAAA‘AAAAAAAA‘AAAA AAAAA‘AA‘AAA‘AAAAA‘AA AAAAAAAAAAAAA‘A‘
'-‘ -'—-— ‘v'--"'_." -— w-‘v-‘v- —---—---
AAAA‘AAAAAAAAAAAAAAAAA ‘-AAAA--‘AAAAAA‘--AAAAA-AA AAAAAAAA
w- v v— - ' —-—- v — --'-' '--

II.
am
.1-
an.

Itnun-dhyunuuqunuuluoUtxtuunlou-n
manna-nuuunthIu-I.
Oaut«huum>0

 

'huodutflblflhmhlk-u

etumnhuumhudllll

:.uucuduu-flnlunu

bra-um.maudnwmumunuuuumlt

m at}. d "1000

I? In“.

117

a-Iuneuuuu-unuuunulumnuuamuaxnd:

an ”Mo
IlhuMIdu'uuolmnh'umw-

flit-muodnnnun

flaunmutunldMHm'um-

00 Ian. 00 In! £0 gut Ilnlllttclttan

havIMn

33

umuu'unmao

ii
.3“: g
3%?
2.4-3a»

E5435 3

o
H C .d 0
I.“ COOP 0“.

 

118

 

 

3 a 5: 5 5

33:3!35535‘3 5 5 : :

5:555: 5333?: ’5 35 55 3; 3;

§ : ' 5 5 55 ~«

g 33!: 3 E! 5 Egg; g 535: .3533”: g; 53; 35 E5
1 3g: 55 E : 553 3 E 3533 ”5‘333~3 .’ 553 3 5 :55
3 5 5 a g 3 3 3 I 5333 23335333 ‘3 “ 3 3 ”i3
5:- 535:53g33: '5... 5.55535: 3: ’5 53?: E33: 5
55:: :55 3k: 33;; (3.3.5.5? 3.. 353.5. 3.5;: 5
5:55:35 355 :55555 53535555 3353535 5555:: 5:55:53
3%: 35 :5 5 3:35 3 “:5 ' 53:: :5: :35: 535::

:3: : 555 “5“5“5“5 5 ~ = ~ = ~

3 35!! 3 3235323: 3:!!! 3:!!! 3:!!5! 3:5!!! 3
i E g E E E 3 E . .

: : . 3 55:5 33:5; :3:
g: 3:: E:- , 35:: :3 35:5: 35:
“ 555 E 555 5 iii
333 33 33 55 5 53.5355: 23::- :5:
553 :3 5.3 =3 3 553 3335i: :5: :5:
5‘= 5:' 55' 35 3 5 535?" 3 . :5: 55 5 . 5::
{a I; l3 S :3: is a": 5* =9”!
i=5 5:! 333 35 353 iii? 5 3 55; 3i g i §§§ 8
'3' . 35:5 . 3535 . 3" . :3: 5,3 5 3 55: ’5 3 3 -:3: 5
35353335533533}! :5: :32: 5:3 :5
xi: 3 :fzgg axing; gang: 3 -‘- “'*" ' ' a °‘° *~ ' a '3- -
3.33 53.3 «3.3 338 g E
:!!!! 3:2!!! 3:!!!! 3:!!! n-“ 3

 

 

 

119

I!» 8:8 In «Ii .

33. no 533» a

£6 .0 calls! 0..» film. a

£6 fill-sno- o

i3?! 328 n

filly-:80 '33 you ”Shoal .5 null o

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

I
no 253
lieu—Boy «38 838
I.» n '33 no. Ila-cu I3 lulu. n
cacao-83!!!!" 'Stguflnwiugn
gigaozilltafi¢§lufiofi . .
..........H.Ivmmwwr..mwm.umm ....... mm.“ ........... mm 330' no. '0‘ gauOu I: 3 in.
v. 3.8 O.

'uug On. an

OOOOOOOOOOOOOOOOOO0.0.0.0...OOOOOOOOOOOOOOOOOOOOOOOO.
- ---A-A-------A---AAA-A--AA‘A-A---------AAA--AA-AA--
--_ --—- _——-- - -_ --- _--_ ——---'

 

 

--A----AAAA AAAAA-AA-AA-A-AAAAAAAAA-AAAAAAAAAAAAA-AAAAAAAA-
—- v w — _——
-

mwttudotnno

m at: ten mum

III-Ell ll III35IIIIU13IDI'IIICI lflll ID! IIICI III OI

'NIIIIIIIEI 'IOUII' I. UIID InltflIICIIIfl 'lllflll
D

I

I

0.00..
.11

m1-

m1. lcu tun :- 1. a
god

am!-

mat-tun!
monastic
0 use: claim 0

 

120

um
M. m!-
mu:
Moll
1
(can:
Lalo-ct

:3
35335": 5:
33% ‘ I ‘ ,
”3 i833333 33333 33333 33: 3
:3!! 3:33! 3:355 3:!!! 3

’3
3
. 3
3"}:
5 5 5 ‘33
3 1"5 :5 33 35‘
'a' '3 3 :3
'5? ‘3 33 33 55
5 35 ‘3
3 3
g

gal-:udmuum-
flzmumuw-
£0.11”:

I

“Glam“.
DIM"

121

.seggsegagv Bag
.sggnsgggsfiagfiu

. .afilfigsggnguagg
gal-EH: gfigigueggigbu

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

.60 on a... aid 3!
.0093 :5 I.» 0.6-In in v:- uoil—A a... 3.3 I.» [-238

don-3’80 Cu. E Gaga sou 133.00.... 3.633 .5.
Jails I: so u:- oonuu.‘ 3.3 0» Ian:-

auuou o. ‘3 .103. no»! «0 «3&0»- I.» has 3 I.»
£028 no >33... .uld unal- .. g :5. In: 3

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

to a slug 3923 you '3». goo-Boa low
506 and». g

303 on» 2.1. mu

'u!‘ 0.39. a

can: on» 031 g

a. 5.3 s

so an :9an you 5:..." u:- 80! I.» .F
0383'. 3 con—a 0:0. 2.» >330... 9!

:33 05503 I: 5 9:600 3933 no 8309 an; a
sing 03.1.:- a

g! 3 In: 393.3 sou -9.»- 333 .u
'«g 03.39 g

.Bsegggiagv Bag

.gghggag
Egggggagssggguufi

.‘g‘ghgsg
ggaagggagflnfig

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

9|
.9! 3..
g
In! 58a. .59!
can ha a
. . .

 

.giggag
.Esugoligglig

nagggggagsg

gain!

a. in 0.3!!
noun-outage: 5.258938 .85”!!- 3
inguinal

glau§§'«§unsjuofltuosnalolfl .u

GEE.

.Bseiaglgv Bag

.§=E£§i°83§§04
.E=E=§.i848-§§04

.igaaggsue
Eglganiiggggflng

.g‘g
gaggggagaagflg

AAAA AAAAAAA AAAAAAAAAAAAAAAAAAAAA AAAAAA AAAAAAAA AAAAA
_ v — v v- ' vvv—wu—vvv ———v-' v—-
A AAAAAAAA AAAAAAAAAAAAAAAAAAAAA A AAAAAAAAAAAAAAAAAAAAAAAAAA
— v v v v—
-

 

 

 

AAAAA

122

m. ch. Iona-tn mm. m can-loud (or and"

mm”
“ml-“Min...

lug-m than
or"

"Mutant-puck

”termini

ascertain

ml.
1
tau

2%
"5 53%-
13%; 33323
4 3:!!! i=3!!! u-“

-AA--- -AA---A-------A------A--------------A---A- ---A-------------
-_- _ --- - --_ - - -— — - ---
’III.. Cl!!! (OUIIIIUI) 'l'. IUD II. III”. 'U’GD’CI.

 

 

“batmunlnthI-me-mmwmm
Mloummlywduuumlydmm

I. '1’: canny a mum

mu

aa;a;;s ; £55.

123

 

 

 

 

 

 

5; ii iii. i
. is ‘54 4 H g :
g :5 3E5 3
E; IE5! E I _ § "
ii; a l‘ s gs: i
Ni ii: i'a' Sis =5 .
a gig e 3 lg" Ea 5:5 5 3 E
iii 5'; 3:35;, IEEE 5 a . §§
gii 35"“ .2” s 5 55 ‘ i “5 i
iiiga ’53 5: i ssss i is izsig
s i siiigiis :EEE: si iiiii i iiiiiiiii
Ei=5 355 g giigiasasi " gt§§fi i 35
-- 25 33-3. 5 i233..i:ii. a
a in} 255 . g 5222 2222222‘ °
_ - 2222222292 4 a: as: assassi i:
5 E s. is: 2:.
iiiM iii»; $333.41? a
. E; 535 § 5; ifiig a! :3: ° ii 3‘
:3 giggigigii £5 iiiigiiifiizii iii iiii iiii ii
5g? g iiaaiaiiii E g f. g iiiziimi‘f 5&3 g 553% g 5533 a: 53
a s gasssasssss- g Ess 55525555555 5: igi=a igi2=a igii=a 9;
a!!! 5:229:222222! i=2!!! i=2sssssezssiss 3:!!! i=9!!! 3=ssis i:

124

0.5000 g

. a. IF

8 a v and! 030. g
n

83 v haul. .OU “I.“

!.P. dlng g

.31... g

a... I. 0.9.330 3.303.. g
. .uUOu .lu :1 Ln
3 o 1 ball. .0. III

.1... g

. a. IF

3 o v 3% bugaugg g

a. o I. 0.3.80 325.3... g
30.. .5. 2'..-

...o I. >33 .0. In...

oni- OI
. a. In

.4 o v 0.9.5 hind-«O. g
a. o I. vain-o 3.303.. g
.00.. .8» 2.... Lu

96 n. 3...... .8. a

:1... g

. a.
mm o v 3.3.8 32,330.“
a. o I. 0...: 3.303.. g
30.. I.» 2.... .u
a... I. and! you a

.1... 8

. 3 IF

«4 o v van-I5 3.5.3.. g
5 o la 0.9.5 hiugng g
.003 I.» .2... .u

3.. I. >31! .8. 3

al.... !

. n. I...

rm 0 v 0...”! >133: g
an o a. 0...: 3.30.7. a.
30.. .1. i .u

a... I. as: ho. a

7.... B!
. 3 IF

an o v y'all. 5.503.. g
a... I. 0.9.380 bug‘s? !
30.. .8. :1 Lu

3.. IA haul! no. a

.1... g

g I.

.n... v wall. 5.503”. s
an... 1 8.350 £2.33.- 9!

o... 1 gall. .3 a

:4 v has 3a.. !
on; I. baa! i=8 2

a... I. .31.. .8 a

on; v had! Ouuo. s
an; O. haul! 0:0. 8

a... I. and! no. 3

«OJ v haul Guns 2
:4 I. had! Guns !

:4 0. haul. an ”I.

3.. v .31... .38 I!
:4 u. 3...... 9:8 a!

on; a. snail. no. a

a... . Pall. 33. I
a... u. .39... 33. O...

a... I. a»... .0. a

and v Avail. 33. g
on; .1 ball. 92.. I

3.. a. 3...... I. a

o... v >33 3a.. 9!
.1. I. bull. 93.. g

.«.u IA id‘s tau 3

:4 v rail. 33. s
2.. I. .25 33. BI

nué IA had! 8c a

:4 v Avail. 33. g
84.x an“ 33. I

a... a. .3! no. a

On.” In 8‘05.- 8' a

‘ui g

n. I

uni v gall-0.. A!

55 I. “30.50.. g

.0 a

5.. I. 53...... no. a

.3... i

a. 'F

5.. v 33.85.. g

2... I. 3:5... 9!

v. .—

2... I. 83.35.. .3 a

Ola! g

0. I

0.6 v gala-.00. !

‘0.“ IA Bdusufll I

o. in

uni IA Eu; IOU a

7:... i

a. II.

and v .330: i

a... la ‘35... g

.0 ha

a... 0. gas... .9. a

OI.‘ g

a. I!

an.“ v sung».- s

an; in Eda-haul. g

0. ha

004 I. £050.. .0. a

«2.... i
a. In.

a... v 83.85.. g

r... I. .3395... a.

o. .u

r... I. 53.5... no. a
an”... a.

.- Ir

9... v 83.55.. g

2.... la I305... g

o. .u

on; .1 53.55.. .9. III

.2... g

on.“ v adv-I50. Bu 3
£008.0613 8

00.4 I. baa! Guns 8

“lat!

ho!
2.0.3.903!
3.91.9.3!!!
noun
3.1.9.3883
Glut!

no!
3..VE.I.S
2.0.15.3!
I...
3.01%.!!!
alaig

on!

«.5 IA India! s
O. bu
«an I. 5.03.! 8U a

0...... g

n. 'F

a... v 83.5... g

00.0 I. 8¢§ g

c. .u

a... In 83.9... .3 III

.3... a.

a. E

a... v 83.33. g
a... X 83.3... a.
v. 5

«I3: 2
o. 'F

...u Usiisugg
2: u. 8.85-8.83...

.05"

06‘ In ”daisy;

«I5: g
o. I

0.3vsfls-‘Bugg
0.31 3301......38

gun

a... 0. 3.8.1.3.!!!

al. 3‘ g
0. IF

u.dflv%isugg
131.381.335.002

gnu

v.3:83i8ua

00;: g
0. IF

d.ufl .5 sun‘sugg

«.3 0a Ida'sua

uuuian.OI—g g
.0 g

a... I. Ella

.I.ig

Ah!
0.6.8.5...qu
3.91.93.03.9-
0..
onélagig

‘_£9_

9....
.n.°vE,-9.
36:8...19
'0.
3.91%!!!

'uifl

on!

00.0vgi9

3.1.9.8289
o.

00.0Iagi!

o... v 85 8.8 .- 9.3n 8!... 8

on... I. 0088 8.8 D. .0... 82a 8 88. .580. 8. 8

on... v 8.8 8.8.8 8
a... I. 88 8.8 8. 8 3... I. 8.8 8.80:. 8

n...v88.88883088
36.888.88.18

3... I. 3.8 8.80:0 .8. 8

o... I. 88 8.8 8 8... 8a 8

8.6a v tau! 8u8u' 8

o... I. 88.. 88 8. 8 0.... I. 8.8 8.8.8 8

o... v .8... 8.8 8 8:. 8 8
63.3.5.7...“ 8

a... I. 8.... 8.8 O. 83. 88a 8
a... I. 8.... 8.8 8n 8

on." IA is! 8¢8~8 8U a

3... v .38 8.80:0 8

”.0 v E 58 8 ‘na. 80 8 Nu...“ la :«i ’d‘fl‘ s

a... 2888.80.82.88
361.888.88.18

a... v 8.8 8 8. 8n... 8 2.... I. 5.8 8.8... 8. 8

n.3v8u888u8

8... v .38 8.80:0 8

o... v 88 8.8 O. .0... 8!.» 8 on... I. 8.8 8.8.8 8

8.0188u898zn8u8
8.0288888u8

a... v 8.888. 888
a... I. 8.8 8 8a 888

a... I. 8u8 8 8a 8

on... I. A.“ 958.8 8n a

a..« v 50". 8a8u' 8
C..." 0. ha¢8 8a8u' 8

o..." I. and! 8u8u' 8n 8

an‘ v 8889‘nuh8u8
o... I. 008 8.8 9 o0... 003.» 8

o... I. 88 8.8 .8. 8

125

a... v 8.8 8 8n .98 8

a... I. 8.8 8 8. 8.... 8
a... I. 88 8 8. 8

am... e 8.8 8.8.8 8

a... v 8.8 8.8 O. 8... 8!.» 8 a. v. I. .38 8.8.8 8
a... I. 8.... 88 O. 8:. 8!.» 8

a... I. 8.... 8.8 8. 8

a...v8.888u888 . I
a... I. 8.8 8 8. 888 8 3 on . .38 8.8.8 8. 8
a... I. 8.8 8 8. 8

no... 8 8.8 838.8 8
on... I. 8.8 8.8.8 8

9.... I. 8.0.8 8.80:0 no. 8

n...v888888=.88

15.8.888u0088 361888.898...0088

a... I. 8.8 88. 0088

F.018Ag88883

8.: v .38 8.8:. 8
8.8 I. 3.8 038:0 8

a... I. 8.8 8.8... 8. 8

36'8888’83688 .
3.9288898...888

a... I. .08 88 8a 8

1.. .8888u888
39.18.888.888

o... I. 8.8 8 8. 8
3... I. 9.8 8.8:. 8

a... v 88 8.8 9 .0... 00!» 8 3... I. 3.8 8.8030 no. 8

92.83.88.888 8.. 2888.80.82.88

1.. I. 8.88 8. 888
1.. I. 8.8 88. 8

«2.8.3.7.... 8
8 z 3 .8. 8. 8:.

8.8 8
a: '8

a... v 8 88 8
a... I. 8 8.8 8
2. .u

o... I. 8 8.8 8

8.8 8

a: In.

0... v 8860 83.8 8
o... I. 8 8.8 8
:- .u

a... I. 888 0.3.8 8

8.8 8
a: I

o... v 8 88 8
o... I. 8 8.8 8
3. .u

o... I. 8 8.8 8

.188

.4. IF

0... v 8 888

. o... I. 8 8.8 8
3. .u

o... I. 8 88 8

«38 8
=- 8r

3. .u
o... I. 8 88 8

.18 8
a: 83.

o... v 8 88 8
a... I. 8 8.8 8
3. Lu

0.... I. 8 88 8

8.8 8

0.. 8:.

a... I. [.8 8.8 8
3. .u

o... I. 8 88 8

3.88.8.8 8
:- 8

8 I 3 .8. 8 8:.
:0 bu

.8. I5»... 8. 8

0‘18 8

a: I

o... v .38 8.8... 8

«.0 .u

3.. v 3.8 8.80:0 8. 8

8.8 8

an. IF

.0... v .38 .38... 8

u... I. .38 8.8.... 8

a: L.

o... I. >38 .38.... 8. 8

126

h: in.

h... v 5‘ as g
a... I. 5c- 80 92
. 0: ha
a an I. 5‘ Us 3

nl.‘ a!

ha. '3.

n.nn v 5'. as g
5.: I. 5‘ 08 a
On. bu

P.0d IA 5‘ 5 a

.1809!

2. IF

a...“ v 5‘ 08 g
“.0" IA 5‘ 3 g
3. a

a... I. 3.- 80 a

7.3092

to!

4.: v 5‘ as g
.6. I. 5‘ 80 g
2...

3.. I. 5.. 8.. a

'.E!

=- '3.

.1: I. 5‘ as s
:0 ha

at: I. 3‘ 3 a

go..dl.§ s
On. 2

i / Id ’5. a In.
In. SH

38. .53.- uou a

.4. a
o.« I. ['60 I33»... 3

.289 g

2.6 v £33.08 a! g
3.5 I. 8.3338 0.3 O!

.3 u.
3.. I. £33.18 no. a!

«I...- a

3. II.

«fin v £73308 0! I

a..." I. H.333 a: g
:0 F

run I. :33: you a

al.... 3

o: In.

8.. . £30318 col. 9!
no.” I. 037.5380 0: a

:- in

no.” I. 3.33.08 to- 3
cl... g

a: In.

a... I. ‘33! a! g

a: mu
8.. I. 8'33: you a

§Q.ul.E g
on. i

f 1 Ca 0! Cl In.
:0 F

Obno- guuou tea a

31‘ g

3. IF

9.9 v 5‘ as g
:- tn

.2. v 3‘ as a

.25!

2. '3

a; v 3.- 80 a!
to I. 3.. «3 a!
3. 3

.6 I. 5‘ 80 a

.35!
so!
Gays-.800!
9.13-3.80...
3...
1.15-.803

16392

2.!

1.. v 5‘ «8 a!

.6 I. 5.. to!

2...

.6 I. 5.. .8 In!

al.-B!

$- .8.

93 v 3‘ as O!
13 I. 5‘ as O!
:- s.

u... I. I... 80 a

7.3 !

."|I""I 'I
““““1“‘1‘1“

.Eagagibgglfllagg

Inv|"|"-l'- I-"'II”"I||'I""|I|""|I |"'--
1lAl‘“““““‘1“““{“‘“““““““‘{““““11

 

00‘ g !
Eoior.ul.uuu 8
00. i

no / 3 8.8.68.- 3.8 no. .52 .313 2326 In. I...

a: fin

2|...- g

:- F

3.. v 8.3:: I: i
on. s.

3.. v :38: no. 3

‘ui !

.3 I

2... . £32389 0511 O!
3.. I. £33308 II.- B!
2- u.

3.. I. ‘33: you III

CI... 8

2. .IF ,
a... v 80.3380 col.- 9!
2... I. 05.3580 8|.- g
:0 Ln
2... I. ‘33: how a

en... a

an. IF

5‘ v :33: 0'. g
o... I. H.333 I: g
a: L.

o... I. £30318 no. all

I... g

an. IF

3.x v £33.80 0.... g
5.« I. ”I33! 0.... g
:- nu

8‘ I. ‘33! you 3

al.... 8

:- IF

3.. . $313 col.- 9!

3‘ I. '33: in g
:- s.

36 I. 8.3:} no. a

0|...- 9!

0.. .F

3‘ v Sirius-8 as g

:6 I. «593308 0'.- g
a: =

.c.« I. 9.7.33.8 son a

0.“! g
on. In.

127

Sold. 1.8.3!

3328 2:833.

.858 8.8 9 no... .85.
:88 .1. noo non-no.

.858 8... . III I!
.5333... . 3n!!!-
Sfillo 2.8. . Ell. 8.8-

7.9.3 5.2.30.0..—

11111111 11 1 1 1 1
1 111 111 11111
11 111111 111111111111111
1111111111111111111

aha. IDVBDIIIDO .N .IPH IU‘I ‘0
.
111111111111 11 "' I’
1111 1111 1111 1
1 1111111111111111111111 111111111 1 " '
1111 11111 1

its}... at...

.53 82 . HI .8

in! IBERI—

ifillo 258.80.

.338 8.8 9 1:. In...
:38 in .8. .958.

.858 I... . II I!

Eons-noun.— . Bug!-

333 33... . Ell. 3.8-

.:s hzugdga
00000000000000000000000000000000000000000000000000000000000000000000000

.Eagagsagg'iagg
no. 253
.oooocoooomfimmmmfi.fim oooooooooooooooooooooooooooooooooooooooooo
3.. so. . at .5.
.i luv-«BI.
:3! 2.38:0.
795.. no.8 0. ‘30 00nd».
:38 so. now cons.
ego-no In: . II I!
Bananas-o. . g
:35 3.0..
.uoonuso 32500.0...

. no.8 0 non coal! 3 .3! IF
«no... on. 83¢. 38: .0... cm“...- uo ”u“! I.» 62».
III! undo 0.80 08 o 2. no lie 9!...

a... . 8.8 . p.366 . 83- 283.8

.. an!» 0.8. 53-93. .5 3338. I'll-I
l3!!- uo 3 co. 9| 0 a . nou'udllo 3:839. .035 2"

Joli-noun in igsagaiFuA‘o‘a
6:8 .. 83....qu law-1538.58» «I... 1.3 I.» .39.!
giisigugtigazglug

silo .3333 E

. .‘Evg .ggiggghu
.§!§.!v§ .aggglriah:

.gglfliga inn-=8

o aisv 10‘s 0

.Shaagsglr
if!!! .ggga.=§§§fiu

.833
BE! .gorgaEE—gihu

1111111111111111 11111111111 11111111111111111111111111111111111111

1111111111111111

II- i ll 'I'I ’||'| I. 'Ipl |IrDI.D.II.”I-III'|".II|'I I'l’l’
11111111111111111 11 111 11111111111 1111111111 11111111111111 1111

 

’DD-’ Fl II ’|I' |’||'I'I’l’|’l' ’Ip' ’ 'i"sr- I. |’.I'Ip|'
11111111111111111111 1 111111111111111111111111111111111111111111111

7333880 col...
3‘ as. . g 58

128

88833.!

.It..3o.i:‘i
\.§.§0..§fi83§§§. I358:

£0.00 0.64. no 3.8.

Siaaigilsiafl.‘ .030.
6" .880
.58... u. 6.3.28.8 8.8.... 0....- .91.
8.90.3: 8.8.0.38 .0 o .830 .1
g: a... u. gull-809.8 no to... 3.00 can In. .00- ii
889.089 8.80.! .00... I'll.

.8.§805880§8.8i
Baggiggua‘igigi

2. .1.“ Baggagagli!
apple-IE

is. 9.. 8.00.308 i. 8.8 0.8086
28... .38.? .8008 I 8.3!...-

.3888...‘
388.888.831.888883
69.8.9.3

8.033.. E
.86. .388... 30...... 0. .05 8.8

3% 0.! n no a»... 3.0. no. ouch-so 08.3.. 1'
30.8 a bid. ill. .880 8.8.8

.4» I. 0.8. 8.»! u. .80.. In... 3.0. a .50.! Ill.

80...! 8:00.! .00... all."

m... . 8.888 2.888.... 8.8..
.58....50..85.8.883.5..83..8.3£.83.8.
.0. .83-35588338888338...
3.83.8...
3.83.818
8.03.81...
“8“. n..d8u
.83-.888 60.18%.80

.EESIualvsqu ..
\zigs Ma 0
u

1.8.5.030. 0» 0.8.000...» on...
{.8800 5.5.5...

.u I! 3803 .883...- 3 .83.: II!!!

83.8.»...- 3 30...;

i 8.850.838... .8 III-Ia.
2.0. agioigoooguoag
gigiggxgoaa.I§%I-uuag
.83083-33.‘
88.888!!!.888|a.58.¢ 85...
6.58.9380
58.888.82.88 .58... 0.88....
moi-0.88.51... 0.5800301
it I. I.o.0.§u0.oo.n 3.0....» ”8.380%
889.588.! 30.5%
.8398588588'3835
annual-anogaiigschSIg
3. .53!!!
Easgllgglig

...........

.8.8u..88
8a....lo!.8.588.83.9.88
88.8.8. .3. 3.98 .83888 85...
.iaufllo
8.8.8.. 3.0... 8.8 0.... 33.8.8.8
. .8.8I.I.I0.885388..93.8Iu 8. 8.88.5.8
8.83.! 8.80.! 003 ii
$831.-..38'8 838585.80
a. «53.809.580.338... .gaiaala
.303; 33.3.8.8.”‘583'28'.’
3. .:
filial-Elbgglfli!

129

.04 3 ‘33! a! .0353 ‘3"! 3 as a an

.‘ul 3 0.3. o
\n.i!u«0 ‘ n In. I '38! i.

.08 you '3 is II Julia 3‘23 i Gala.
in: «Dun-lo launch. an

Fidi. 3% we .93. 08 .333 a on .338.- I'll.

on; «833‘ 30.8-- is. .035 Q's-ll

. £38. Bull-a. :3 .553 56.0.:

38 o. in: no no... 055" you 3»!-
II yuan-n [33‘ I5 can: In. .38
6.3. 3.3-lo [asl-

33.3%
as 3303 g . 83. .83! 33C.
\ .iganguorllui... till»:

.anfi'sghuul
83.33.883.818 saloon-5!...-

REFEFENCES

1.) Akita, K., 1988
mum Electric Study Publishing.

2.) Daimon, T., “Automatic Inspection at Carton Box Manufacture”
MM June 1986

3.) Diana, T., 1990

W, Michigan State University School of
Packaging, Lansing, Michigan

4.) Frederick, Hayes, R., Donald A. W., Douglas B. L., 1986
W. Industry Booke-

5.) Iizuka, Y., 1983

WWW Niikan Industry newspaper
Tokyo, Japan.

6.) Information Builder, Inc. 1989

LuLiMamtosflfielLStudLC-Mde Information Builder. Inc
Broadway, New York, NY 1989

7.) lsI'Iikawa, K., 1989
W, Niika Kiren Books, Tokyo, Japan

8.) lsono I-I., Takayanagi 8., Nimoda T., Toda S., 1983
“Research on Evaluation of Printing Quality"

WWW January 1983
Vol: 20 No 1

9.) lsono H.,1989
'Quality Control“ In Color Reproduction Process“

W Vol: 26 N05 Tokyo Japan

10.) lto Y., 1985
Numerical com] for QQIQE En'ming
Printing Academy Publication, Tokyo, Japan

130

131

11.) James J., Phillips 1987

-1"; 1:m ‘.H|I.-.-I'l - I F.' .1”. ‘ -. b. f ,: I:

W. Master Thesis.
Michigan State University

12.) A.T. Kearney, Inc 1990

Elln E Is Im | 1!! |.! 'I'
Council of Logistics Management, Oak Brook, IL USA.

13.) Matsuoka M., 1985
“Inspection System for Carton Box Manufacture Process“

W February 1985

14.) Mizuraku Y., 1987
“Inspection Method of Paper Box Printing“

WW August No.66
Tokyo, Japan

15.) Ootsuki S., 1990
“Quality Control for Carton Box Manufacture“

WWW January No.95
Tokyo, Japan

16) Takahata H., 1987
“How to Advance Quality Activity without Failure“

WWW
October No.68 1987 Tokyo Japan

17.) Takayanagi S., 1986

Wm.
Japan Printing Newspaper Co.Ltd.,

Tokyo, Japan

18.) Sugita S., 1990

W
Education Co.Ltd., Tokyo, Japan

132

19.) Ueno S., et.al., 1983

WWW.
Nikkan Industry newspaper,
Tokyo, Japan

20.) Watennann.D.A. 1986

Em
Industry library, Tokyo, Japan

"IWilli]III‘IIIIIIIIIIIII“