IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 38, NO.

3, AUGUST 1991

A Method for Quality Precoordination in a

zyxwvutsr 245

Quality Assurance Information System

Cheickna Sylla and Bay Arinze

Abstract-The paper introduces a new concept of an informa- conflicts, and make decisions can directly affect the produc-
tion center for planning and controlling the quality of products tivity and efficiency of the manufacturing operations [24].
and processes in a factory. This center is called the Quality To function, societies have evolved languages, communi-
Assurance Information System (QAIS), and its role is to tie
together through quality related functions, in design, manufac- cation systems, transportation techniques, and rules of behav-
turing, distribution, sales and other factory management sys- ior. Similarly, today’s factory production systems have com-
tems to optimize overall factory performance. This paper views munication and transportation networks and rules of behav-
the quality functions of QAIS as carried out through coordina- ior. They have defined objectives, and have devised ways of
tion and precoordination activities. While coordination is a well coordinating their activities, and measuring progress toward
known concept, precoordination is new. It aims at organizing achieving these objectives. In today’s highly competitive
the participation of all subsystems in the factory network in the
planning of new product development, and in providing world, quality and productivity are the central issues. For the
methodologies for system evaluation and trade-off analysis for larger business system to increase quality and productivity,
better solutions and integration. This paper presents a formal the management and control systems of the various internal
approach for dealing with precoordination activities. It proposes macrosocieties require interconnection. The concept of the
a methodology for the quantitative evaluation and selection of optimization of the factory activities expressed by this inter-
complex systems satisfying desired product or process quality
requirements from a number of the factory’s subsystems. Practi- connection is called computer integrated manufacturing or
cal steps on how to implement precoordination are presented CIM [ 111. Clearly, the concept of CIM centers operating
and the methodology is illustrated through a case example. together on common manufacturing processes are rapidly
Keywords-Quality assurance; information systems; precoor- giving shape to our factories of the future. However, it is
dination; selection methodology. also becoming apparent that as these systems interact with
each other, the same problems experienced by expanding
societies recur. For instance, languages become irreconcil-
INTRODUCTION able, communication systems become incompatible, and be-
havior becomes governed by different rules. This, in sum-
I N recent years, international economic competition has mary, is today’s CIM challenge [13].
increased at an unprecedented rate. As a result, further
innovations in computer design, information processing, au-
Currently, the new ways of doing business competitively
imply that manufacturers must create new products almost
tomation technologies, and manufacturing processes are be-
annually and/or design clearly measurable qualitative im-
ing conceived to push for integration across all functional
provements in old ones. In addition, management must con-
areas in the search for further productivity gains. The techni-

zyxwvutsrqpo
tinuously plan and design parallel changes in its approach to
cal literature indicates that the use of computer technologies,
planning manufacturing activities to reduce costs relative to
such a computer aided design (CAD) and computer aided
competition. In this domain, research efforts by Feigenbaum
manufacturing (CAM) can help in attaining a significant
[8] and others have introduced the concept of company-wide
competitive edge [18]. These in turn, are leading up to new quality assurance. This systems approach to manufacturing is
concepts in, and the implementation of, totally automated largely centered around information management, which is
factories [131.
increasingly recognized as an important determinant for suc-
Today, the factory itself can be viewed as a society [111. It
cess in current highly complex computer integrated manufac-
comprises diverse workers with an overall goal, but with
turing environments.
differing and sometimes conflicting sub-goals and incentives.
This paper investigates the issue of information planning
Hence the relationships between these people and the re-
for Quality Assurance Information System (QAIS). The QAIS

zyxwvutsrqponm
sources they control can be very complex. Furthermore, the
concept is one of a centralized information unit with the
speed and accuracy with which they communicate, resolve object of planning and coordinating quality control activities

zyxwvuts
Manuscript received October 23, 1989. The review of this paper was
processed by Department Editor J. R. Evans.
C. Sylla is with the School of Industrial Management, New Jersey
Institute of Technology, Newark, NJ 07102.
B. Arinze is with the Department of Management, College of Business
within the CIM network and with all other external depart-
ments related to the factory operations. As such, it ties
together quality achievement-related functions such as de-
sign, manufacturing, and factory management systems, to
plan, optimize and monitor overall CIM network perform-
and Administration, Drexel University, Philadelphia, PA 19104.
IEEE Log Number 9144562.

zyxwvut
ance. These coordination functions include the following
activities among others within the QAIS framework:

0018-9391/91/0800-0245$01.~0 1991 IEEE

246 zyxwvutsrqponm
zyxwvutsrqp IEEE TRANSACTIONS ON ENGINEENNG MANAGEMENT, VOL. 38, NO. 3, AUGUST 1991

1) organizing the participation of all subsystems in the

CIM network in new product development or purchase
decision;
2) Groviding systems analysis methodologies to collect and

zyxwvutsrqpon
disseminate information generated during the manufac-
turing operations for the monitoring and improvement
of all processes.

zyxwvutsrqp
This research addresses the above issues. It deals with two
related Q A I S design problems: 1) the analysis of basic con- PRODUCTION PROCESS

zyxwvuts
PLANNING PUNNING
cepts and fundamental structures to formalize the activities of
coordination in the purchase or design of new product (termed c
zyxwvuts
precoordination), and 2) the analysis of information for coor-
dinating quality related activities to make quality predictions
from process factors. The coverage of these issues in the
technical literature is limited (see for instance Furukawa et
FACTORY
OUALIN
ENGINEERING 6
MANUFACTURING
DATA BASE

c
al. [9]).
The paper is organized as follows. First, the QAIS concept 1
INVENTORY
is introduced and justified. A discussion of the issue of MANAGEMENT

precaordination and its role in planning product and process

quality follows. Next, we develop formalisms for the preco-
UPPLIERS
+ i
ordination problem, and then illustrate the use of these
formalisms via a case example of a new product selection
decision. The summary of results and suggestions for further
research are given in the concluding section. Fig. 1 .
1 COST PLANNING 6 CONTROL

Framework for QAIS.

zyxwvu
I
SHIPPING h
WAREHOUSINa

THE QAIS CONCEPT

JUSTIFYING
Q A I S is a major information center for the management of
the factory production operations. Its central tasks consist of
aiding in the acquisition and development of new products, in
1
I
Computer
the monitoring of daily factory operations, and in facilitating
\lehlde. Controllers
product performance analysis in the market place. It accom-
plishes the production management functions by studying the
related operations, and directing by exception, supervisory
Computer
attention to productivity, reliability, quality, product flow,
and material flow problems before they become intractable. Controller Controller

In addition, it acts as a logical connector to the CAD and

CAM systems at lower levels ’in the manufacturing hierarchy
of CIM, and to management information systems and deci- Equlpment
i..-.-]
Equip"

sion support systems (MIS/DSS) at higher levels in the I I -w

I I
hierarchy.
Q A I S plans and supervises the required activities during
Local Area Network lor Operatlons Control
new product development and seeks the best composite solu-
tions whenever possible. This function of Q A I S is called

zyxw
precoordination. It involves coordinating the participation of
representatives from all the subsystems in the planning for
the acquisition and development of a new product, and in synms
choosing the best feasible manufacturing solutions (based on
Fig. 2. Q A I S in typical distributedfactory CIM network.
the established criteria). Q A I S may also be used to perform
product performance analysis in the marketplace via the
collection of field data, customer satisfaction ratings, com- separable activities if high level quality is to be achieved. In
plaints and warranty adjustments, and also via research on addition, workstations will soon become available which
competitors, data on patents and claims, etc. A conceptual integrate business planning functions with design and manu-
framework of QAIS for a factory system is illustrated in Fig. facturing processes [181. Furthermore, network management
1. The typical location of such an information center as part structures will simultaneously decrease design cycle time and
of a distributed factory information network is illustrated in tremendously increase the amount, variety and accuracy of
Fig. 2. technical information sharing (including large data bases with
Currently, design and manufacturing can no longer be graphic data items). In such an environment, the ability to
SYLLA AND ARJNZE: QUALITY PRECOORDINATION IN QUALITY ASSURANCE INFORMATION SYSTEM zyxwvu 241

zyxwvutsrqp
respond to the customer’s needs and requirements will de-
pend upon the effectiveness of QAIS to coordinate the factory
manufacturing activities to promote and support customers’
satisfaction [3].
As an example of different approaches to dealing with
customer problems, compare the standard approach to a
product field failure with the approach that may be based on
the QAIS concept. In the standard approach, the objective is
ing. Indeed, traditional methods do not deal with the most
fundamental issue of integrated manufacturing systems,
namely, the problem of information sharing across the fac-
tory network. With the traditional methods, interrelating the
data and knowledge involved with the different manufactur-
ing functions may be carried out only partially, and the
generalization to non-manufacturing services will be difficult.
This is a serious gap in many systems today [13].
to remedy the failure as quickly and as cost-effectively as The above difficulties have resulted in many examples of
possible. This may require repair and replacement. In the independent information models being used by various func-
QAIS approach, one objective is still to rectify the immediate tions in the factory, and these problems are increasing in

zyxw
problem. However, another equally important objective is to complexity. Commercially available CAD models do not
gather the information generated in the process in a pre- contain sufficient information to drive most primary produc-
planned manner, and to learn as much as possible from the tion and quality control processes. In addition, despite many
failure, to reduce the incidence of future failures and to available MIS/DSS, it is still very difficult to correlate
improve the future field maintenance. To know exactly what accounting data with manufacturing processes; furthermore,
part failed and why. In addition, to obtain answers to the material requirements planning (MRP) , computer aided proc-
following important question among others, to serve for ess planners (CAPP) and many other production information
improvement company-wide. What was the correct solution? packages are still not directly linked to the shop floor they
How did the field data help indicate what was wrong? What control [3]. Unfortunately, to deal with these difficulties,
changes in product design manufacturing procedures, recom- much of the work today is striving to interface various
mended utilization, and maintenance procedures, or purchase software packages rather than addressing the fundamental
decision, could have prevented or postponed the failure? issues of integration.
How much did customer utilization deviate from the recom- Within QAIS, integration refers to the means of managing
mended product usage and why? Did field data indicate that and linking information related to the quality achievement
spare parts were adequately and easily in hand? How helpful functions of various segments of the company in order to
was the customer’s description so that appropriate and timely coordinate, correlate, and communicate it in a timely manner
service response could be sent? for the company’s overall benefit. To achieve integration,
Thus with QAIS, each field failure contingency may be QAIS should use structured information analysis to enable
treated as an opportunity for learning. With such an informa- computerized tools to be applied to the validation, verifica-
tion system on hand, every repair person’s knowledge is tion and documentation of various precoordination and moni-
improved, and as a result, the next time similar product toring activities. Hence the concepts of QAIS and precoordi-
symptoms are encountered, it will take less time, effort, and nation used here are new, and their impact has not yet been
resources to remedy the problem. Without QAIS, the impetus fully analyzed in the technical literature. The next section
will not exist for institutional learning. Indeed, without it in discusses the problem of precoordination for QAIS.
place, maximization of learning benefits from any manufac-
turing service function such as field service will be impossi- FRECOORDINATION
IN A QUALITY
ASSURANCE
ble or at best, difficult. For example, with QAIS, even units INFORMATIONSYSTEM
to be replaced or discarded would be subjected to a “post- QAIS acts as a controller of the feedback-loop and as an
mortem” study and the information compiled would be made observation window from which manufacturing activities are
available to product and process designers, and other related coordinated. In this role, QAIS performs its control cycle
decision makers for incorporation in the modifications of new through five steps: observation, evalution, diagnosis, deci-
product design, or purchase decision. sion-making and implementation. Conceptual frameworks
To design QAIS, we must therefore be aware of the data similar to QAIS such as the concept of the quality control
requirements that every subsystem within the factory. may window (QCW) have been discussed in reference [ 5 ] . How-
place upon the design, and the related data structures. This ever, these concepts are made towards the day-to-day product
means that early in any design, both intra- and inter-systems planning and control issues. Precoordination, on the other
communication strategies must be established and the infor- hand, arises out of the desire to respond to market demands
mation required by each subsystem defined. Clearly, a fac- and to develop or acquire defect-free products that are tai-
tory cannot evolve optimally without this level of planning, lored to customers’ needs and preferences in the shortest
which may be provided by QAIS. amount of time. Thus, the emphasis is on the problems of
QAIS is based on the desire to promote and support the preproduction planning and systems selection and/or integra-
highest quality by preventive control methods and through tion.
efficient use of the information derived throughout the whole One such problem includes solving the differences in views
factory. QAIS is a promising concept because a factory is a or choices among subsystems. Thus, the aim is to have QAIS
complex human-machine environment, and as such, tradi- appropriately organized for such a goal. Here, the term
tional data collection and processing methods inhibit the “subsystem” is used to denote a step in the processes of
successful implementation of the data unification and process- product development and implementation, e.g., product plan-
~ zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

248 zyxwvutsrqponmlkji IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 38, NO. 3, AUGUST 1991

ning, product design, process design, production, sales and tion modeling problems which must be examined within the
services, etc. For instance, a difference in view may arise QAIS framework.
when the production function determines that the desired
quality cannot be realized according to the product design MODELING
THE PRECOORDINATION
PROBLEM
specifications. We have learned from experience that these The main activities of precoordination in the QAIS can be
types of problem arise most often when quality planning and summarized as follows:
evaluation are done sequentially [4], [9], [18]. Clearly, this is
1) The desired quality projected in the market place and
the prevailing situation in most current quality control sys-
the trends of technology are clarified through steps
tems, where quality implementation is shaped gradually by
(a)-(c) seen below, and the quality goal that the factory
trial and error procedures.
intends to achieve is summarized in total planning by
In a more efficient quality assurance system, quality is
phases.
related to the engineering techniques to be used in each
a) Definition of the market needs by the sales depart-
subsystem before the final product quality is determined.
ment.
Indeed, by identifying and coordinating the best available
b) Preliminary research (i.e., collection of field data,
techniques (if any), and by establishing that the desired
competitors’ positions and patents) mainly by the
quality goal is attainable, the prospects of achieving a shorter
design engineering department.
product development or acquisition cycle can be improved.
c) Research of claims within the factory by the QAIS.
Furthermore, if in the subsequent following analyses, the
2) The causal relationships of quality and engineering
quality goal is judged unattainable, it may then be possible to
techniques through all steps, from product planning to
determine alternative engineering techniques that could be
sales, are utilized by all the departments. These depart-
successfully developed, and the necessary information for
ments contrive to form unified tables called quality
supporting the required efforts. This is, in essence, the
tables (or production engineering tables). The best tech-
contribution of precoordination.
nique of the whole system is clarified by study of
As a result of precoordination, all efforts will be concur-
available techniques within each subsystem and through
rently supportive, and the decision makers in charge of each
investigation of patents. Following this, the quality goal
subsystem will be able to participate in product planning,
is compared with the predicted product quality, and the
each providing related information so that the goal attainment
attainability of the goal is evaluated during the formula-
activities can be successfully carried out [22].
tion of individual product planning.
The concepts of concurrent engineering were first intro-
3) If the quality goal is judged unattainable, the technical
duced in Japan a number of years ago, [lo]. However, only a
bottlenecks required for goal attainment are identified.
few manufacturing and aerospace companies have begun
Preliminary investigation and experiments are then car-
limited implementation of this approach in the United States.
ried out to see if these may be overcome. It may be
Coverage of this area in the technical literature is, therefore,
decided thereafter to use the best currently attainable
still very limited. This is because the activities of precoordi-
quality or to adopt or develop new techniques required
nation are new to the technical community and as such, are
to achieve the optimum goal. At the end of this phase,
not prescribed formally in the field of quality assurance.
the techniques that should be developed may be summa-
Furthermore, it is difficult to apply usual systems theory to
rized in a bill of development.
the development or selection of new product quality charac-
4) When the quality goal is judged attainable, individual
teristics due to the following reasons:
product planning is carried out at each subsystem level,
1) There generally exist limiting technical bottlenecks to and the plan is summarized in a technical document
overcome in new product development. (called the summary of planning). The quality goal of
2) We must take into consideration the relationships be- each subsystem, i.e., the steps from product design to
tween quality characteristics and technical factors, sales, are clarified at this stage [3].

zyx
which are not yet generally established. These concepts can be analyzed using basic mathematical
3) The global mathematical structures which characterize properties well established in set theory. For instance, a
quality assurance systems (outside of statistical field) modified hierarchical model proposed by DeMarco [7], can
are still weakly defined, and consequently do not pro- be used to represent the factory quality assurance system as
vide a unified application of useful concepts [151, [23]. described below.
The quality assurance system can be assumed to consist of
Therefore, in order to establish any amount of control over
the activities of precoordination it is necessary to: 1) define
the underlying mathematical structures which characterize
this domain, 2) articulate the important relationships among
them, 3) identify the applicable models, and then 4) provide a
methodology to optimize this process. The subsequent sec-
tions provide further descriptions of the modeling of the
activities of precoordination, and elaborates on the informa-
zyxwv
P processes, D decision makers, and a coordinator D o .
Since quality is formed through processes, and since the
applicable techniques are determined by decision makers, the
processes and decisions can be divided into groups
{PI,P2;**, P,} and { D l , D , ; . . , D d ) , respectively.
Therefore, Piand Dj can be regarded as “component parts”
of the system, which correspond to a stage in the quality
assurance process, i.e., the quality design stage. The coordi-
SYLLA AND ARINZE: QUALITY PRECOORDINATION IN QUALITY ASSURANCE INFORMATION SYSTEM

nator usually arranges meetings in which representatives of

zyxwvut
For each set of attributes A i , a requirement set (i.e.,
249

each subsystem or department take part, and coordinates the

zyxwv
resource and technical requirements)

zyxwvutsrqpo
interactions of these subsystems. A formal structure of these
coordination activities can be established to build models for
decision inference. is used such that r i j represents the amount required of the
The approach used in this paper is based on the develop- attribute aii for problem p i .
ment of the figure of merits similar to a procedure initially Thus, for the set of problem P , we arrange the require-
proposed by Sage [20] and modified by Ahmed and Darwish ment sets in the R matrix of n x I dimensions, where the

zyxwvu
[l].The figure of merit, adopted here, is a desired value that ith row represents the requirement set of the ith problem,
combines both problem and system specifications to permit i.e., each row comprises a row vector R i .
the evaluation and selection of the most appropriate system.
This is chosen from a set of candidate systems suitable for a
wide range of applications for which the system can be used.
As such, the figure of merit uses factors such as quality
R= 1 r21 zyxwvutsrq
I
. . . .r22
. . . . *. r.2.n. .
...........

zyxwvu
1.H r12 . . * r l n
goals, quality attainment, best engineering techniques, tech-
nical bottlenecks, and perference ordering. A summary of the For each problem p i in the set P , a matrix Q i = [qijk]
procedure for the quantitative evaluation and selection of of I x n dimension is used to describe the characteristics of
complex solution systems in quality precoordination based on each candidate system such that qijk represents the extent to
the concept of figures of merit is covered in the next section. which the subsystem si, ( j = 1 , 2 ; . * , I ) , possesses at-

zyxw
tribute a i k , ( k = 1 , 2 ; . - , n ) .
QUANTITATIVE EVALUATION AND SELECTION
Thus, for the set of quality related problems P , we obtain
METHODOLOGY IN F’RECOORDINATION
the following three-dimensional quality attainment array: Q i
The precoordination problem is assumed to include the = [eiik], where the indices i = 1 , 2;.*, m , j =
following elements which are assumed defined from the 1 , 2 ; . - , 1 and k = 1,2;.., n are as definedabove.

zyxwvutsr
previous analyses: A figure of merit may be developed for each solution
1) a set of candidate solution systems from which it is subsystem s j , ( j = 1,2, * , l), allowing evaluation and
required to select the most appropriate one; selection of the solution subsystem having the best quality
2) a set of quality related functions for which the system is merit. This figure of merit may be based on the differences
expected to be used; between what the problems require from the system and what
3) a set of attributes used to describe the requirements and the system is capable of providing, subject to constraints on
characteristics for each problem in the set of problems; the specifications, time, money and other resources needed in
4) a set of requirements representing the amount of at- the precoordination activities. This phase will involve the
tributes for the set of problems; decision makers, D , and the coordinator, D o .
5) an indication of the extent to which each of the candi- DEVELOPMENT
OF A FIGURE
OF MERITIN
date solution systems possesses each attribute in the set PRECOORDINATION
ANALYSIS
of attributes; Fig. 3 illustrates the relationships between system si and
6, a to evaluate and the most appropriate problem p i , and Fig. 4 illustrates how their factors may be
system from the set Of systems that are compared. For instance, as seen in this figure, system sj may
suitable for the set of problems. be deficient for some of the requirements of problem pi for
These elements can be combined to provide a ranking of some of the attributes, a i k , ( k = 1,2;.., n), but it may
the candidate systems from which the most suitable system also exceed the requirements for some other attributes. Natu-
can be selected. Thus, let us define the following: rally, different patterns may exist for different problems p i
(i = 1 , 2 , * m).Attributes can be arranged with respect to
a ,

their importance (for instance, attribute u i j is more important

is the set of m quality related problems: than attribute uik for j > k), and the decision makers in
agreement with the coordination can decide on the appropri-
ate weights that represent the importance of the attributes.
is the set of candidate solution systems which could be used Examples of such weights, wik are also seen in Fig. 4.
for the set P and which consists of I proposed systems (i.e., Thus, for each problem p i , the factors q i j k , rik and wik
solutions). can be used to obtain normalized weighted differences for
For each problem Pithere is a set of attributes; each attribute as

Ai = { a i l , a i 2 , * * *ain)
, d u. k. = I qijk - ‘ik I
Wik. (1)
‘ik
which is used to describe the functional requirements and
system characteristics. The elements of the set A i are ar- Note that while we choose the absolute difference, other
ranged with respect to their order of importance, i.e., a i j is expressions of these differences may be approapriate. These
more important than aik for j < k . differences can be summed over for each problem p i to
250

where
Dij =
n
zyxw
zyxwvu
zyxwvutsrqponmlkjih
z

k= 1
I
D;j

4..- rik
'Jkr,
=

y..
Ilk
k= 1

(Wik

=
Piannlng

Product
Deslgn

This equation can be rewritten and simplified as

Process
Deslgn

Manufacturing

Sales 6
Services

Product
Planning

Process
Design

dijk.

-.q i j k
'i k

Thus, Dij measures how well system sj can be used to

51
k= 1

solve problem p i , hence it represents a figure of merit for the

system sj as regard to problem p i .
Since we are interested in the system appropriate for a
Yijk -
P3

P9

1 Wik
IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 38, NO. 3, AUGUST 1991

Definition of Problem Characteristics

Accumuiation of Appropriate Solution

represent the total normalized weighted difference for all the

attributes of this problem as
zyxwvutsrqpon
<$vim
- Ps2
\P6

(2)

(3)
'62k
/

Fig. 3. Illustration of precoordinationproblem.

s3
s4
rll

r14

wide range of quality problems p i ( i = 1,2; -

* , m), we

develop a figure of merit for sj concerning the whole set of

problems p to take the form:
zy
The weight, w i j k ,of attribute a i j k represents the relative
importance of that attribute as judged by management. Stud-
ies reported in the mathematical programming literature,
such as in multicriteria goal programming literature, have
investigated and established some general classes of weights
appropriate in many similar problem situations (see [71,
[17]). (Each class may define an objective measure, or a
preference structure, for a particular subsystem and its deci-
sion maker). In our problem situation, several classes can be
derived to categorize the possible changes of Wjk with at-
tribute a i k . Four possible classes of changes of W ; k with
z
SYLLA AND AFUNZE: QUALITY PRECOORDINATION IN QUALITY ASSURANCE INFORMATION SYSTEM zyxwvu 25 1

requlrements zyxwvutsrqp

zyxwvuts
zyxwvut
zyxwvu
zyxwvu
zyxwvutsrqpon
ai1 a12

attrlbutes of Pi
Bin

Fig. 4. Illustrationof differences between system sj and problem pi.

attribute aik are considered below. In class 1, where the F = {FlyF 2 , * * F,)
*,
weight may be the same for all problems, we have:

zyxwvut
can be computed for all candidate systems using the appropri-
wik= l f o r l s k s n . (5) ate classes for each sj. In the final step, the system corre-

Wik
zyx
In class 2 , wik may be considered to decrease linearly with
k; for instance it may be represented as
k- 1
=1 - -
n-1
l i k s n .

In class, 3, wik may be considered to be constant up to

k = t, then decrease linearly with k. This may be repre-
sented as
(6)
spending to the smallest value Of 5 for is
to represent the most appropriate system fitting the desired
requirements and applications. Therefore, any system S j ,
having 5 as a figure of merit, is said to be the best solution
system for the quality problem pi if and only if F, is such
that:
F, = min{F,, F , ; . . , FI).

Therefore, the set F can be used to rank the technological

l s k r t i n
Wik = {’1 - - k - t t i k s n (7)
candidate solution systems, and make it possible to evaluate
and select the most appropriate systems in the precoordina-
n-t tion of factory operations and quality control activities.
Class 4 represents a category in which k decreases expo-
nentially such that:
STEPSIN PRECOORDINATION
IMPLEMENTATION

w.rk = 1 Ik 5 n It is important to note that precoordination involves a

(8) series (or phases) of studies in which several problems are
where CY is the exponential coefficient (0 ICY s 1). solved as prerequisites to implementing the methodology
Using wik from any of the above, we can write the final developed in the preceding section. The problems that QAIS
expression of the figure of merit by replacing wik appropri- must deal with in the steps of the precoordination are as
ately in the expression of 5. These expressions are seen in follows:
the appendix for the above classes. Step 1: Acquire internal and external new engineering
Once the figure of merit F, is computed for candidate factors, and market satisfaction variables.
system sj using the selected classes, the set: Step 2: Select planning quality characteristics which are a
 zyxwvutsrqponmlkjih
252

zyxwvutsr
IEEE TRANSACTIONS ON ENGINEENNG MANAGEMENT, VOL. 38, NO. 3, AUGUST 1991

fit to the market needs and are realizable by the factory from
consensus through subsystems. Otherwise, go to step 1.
Step 3: Identify problems and their attributes, and list
system requirements which remain unresolved for the planned
quality related functions.
Step 4: Propose alternative candidate solutions that are
zyxwConfiguration 1
A multi-user system consisting of one (1) Compaq 80386
microcomputer, running at 33 Mhz, with 16Mb of RAM and
320Mb of fixed disk storage, plus:
0 A multi-user operating system;
worthy of further investigation. There should be at least three 0 Eleven (1 1) Wyse 150 dumb terminals;
or more alternatives, enough to cover a broad spectrum of 0 Multi-user application software and development tools;
available technical solutions. If not enough alternatives are 0 One (1) laser printer and two (2) dot-matrix printers.
available, repeat from step 1.
Step 5: Identify appropriate classes of weights and assign Configuration 2
values to the weighting factors, again from consensus with A Local Area network consisting of one (1) Compaq
the decision makers and coordinator. 80386 microcomputer fileserver, (running at 33 Mhz, with
Step 6: Select the best candidate system through the quan- 16Mb of RAM and 320Mb of fixed disk storage) plus:
titative evaluation and selection methodology detailed in the
previous section. 0 Twelve (12) Compaq 80286 microcomputers (1Mb Ram,
The selected alternatives must be realizable and must fit 20Mb Disk, VGA);
the quality goal, otherwise QAIS must lead the effort to 0 Novel1 LAN software with required network cards and
develop or create new engineering techniques, called bottle- cables;
neck engineering techniques for new appropriate solution 0 Server-based and single user software and development
systems [lo], [16]. In the following section, we propose a tools;
case study illustration of steps 4, 5, and 6. One (1) laser printer and two (2) dot-matrix printers.

A CASEILLUSTRATION Configuration 3
To illustrate the use of the concept, an experimental inves- One IBM AS400 minicomputer, with 16Mb of RAM and
tigation was performed to simulate a decision making sce- 800Mb of fixed disk storage; plus
nario for new equipment purchase in the area of computer A multi-user operating system;

zyxw
0
usage. The new investigation included a detailed question- 0 12 Wyse 150 dumb terminals;
naire developed and submitted to 17 users/developers of 0 Multi-user application software and development tools;
information systems working primarily in either manufactur- One (1) laser printer and two (2) dot-matrix printers.
ing, consulting, or teaching/research environments. All of
the users responded. However, only twelve participated in all
activities, which included answering all questions and partici-
pating in follow-up group discussions during trade/off analy-
ses. Among these, four came from manufacturing manage-
ment background working in medium to large industrial
sectors with a substantial number of computer applications.
Another four worked primarily as managerial consultants
zyxw
These systems represent three different product types used in
three main kinds of applications (i.e., problems), P ;
P = {Pp P*, P3)
The problem types considered are as follows:
Application 1: An order processing and inventory man-
agement system for an estimated volume of 50-100 transac-
with several years of experience, and had exposure to a large tions daily, and maintaining computerized inventory and
variety of hardware and software. The final four respondents customer databases of 30 000-50 OOO inventory items for up
possessed extensive computing experience in both service to 2000 customers. In addition, the system would support
industries and academia. It should be noted that while the several hundred enquiries each day on the status of orders,
users possessed varied backgrounds, each had managerial backorders, item availability, etc.
and decision making experience in information system design Application 2: A marketing information reporting system
and acquisition. In addition, all of the respondents had formal (IRS) and decision support system (DSS) with the function of
training in computer hardware and applications, several of supporting the efforts of a marketing department in a large-
them were computer analysts with MBA’s. The respondents sized clothing company. The responsibilities of this group
were from the Philadelphia area industrial sectors. They were would include the use of models to help determine the
gathered at Drexel University for an advanced MIS training marketing mix for the various products, formulate marketing

zyxwvutsrqpo
course offered by Drexel University. The respondents were campaigns, and monitor current sales effort (together with
asked to assumed the role of decision makers, while the competitive effort).
authors served as coordinators and provided guidance. Application 3: An Office Automation System that is re-
We considered an application in which there were three quired to serve the office needs of a medium-sized depart-
computer configurations (or systems) being considered for ment. These needs include word processing for a wide vari-
selection: ety of documents, data base support for general filing and
indexing of office files, and spreadsheets for simple financial
models to track the department’s budget and expenses.
These configurations are described as follows: We used nine attributes to describe problem and system
SYLLA AND ARINZE: QUALITY PRECOORDINATION IN QUALITY ASSURANCE INFORMATION SYSTEM 253 zyx
zyxwv
requirements, and each decision maker provided quantitative

zyxwvutsrqp
values to estimate of the extent to which each system satisfied
the requirements in each attribute for the three problems. The
VALUES
TABLE I
OF Yji k FOR PROBLEM
pl
Attribute

zyxwvutsrqp
measures supplied were as follows: Svstem 1 2 3 4 5 6 7 8 9
1) system purchase costs; S1 1.19 1.25 0.98 0.99 1.1 1.04 1.02 0.98 0.9
2) system maintenance costs; S2 1.05 1.025 1.11 1.09 1.12 1.1 1.08 1.03 1.09
S3 0.88 0.9 1.13 1.21 0.89 1.1 1 1.16 1.01
3) CPU performance;
4) 1 / 0 performance;
5) system portability
6 ) software availability;
7) system upgradability; TABLE I1
OF Y,,' FOR PROBLEM
VALUES p2
8) system reliability; and
9) availability of vendor support. Attribute
The nine attributes considered are generally accepted in the System 1 2 3 4 5 6 7 8 9
literature on information systems and represent important S1 1.13 1.08 0.91 0.92 1.09 1.03 0.95 0.93 0.97
assessment criteria for computer configurations [ 2 ] , [ 121. To S2 1.08 1.13 1.07 1.06 1.13 1.15 1.03 1.01 1.07
begin with, we conducted a pre-investigation during which S3 0.98 0.87 1.15 1.09 0.90 1.03 1.03 1.09 1.08
stage, each respondent (acting as D.M.) was explained the
issues and steps involved in the precoordination. It was
determined, based on consensus, that it would be difficult to
provide direct quantitative measure for the quality and techni- TABLE 111
cal requirement factors, namely, qijkand rij respectively. It OF Yiik
VALUES p3
FOR PROBLEM

was felt that it would much easier to provide subjective Attribute

estimates for the ratio Y j k of these two factors instead. System 1 2 3 4 5 6 7 8 9
Therefore, questions related to qijkand rij were dropped in ~~ ~~ ~~

SI 1.21 1.08 1.06 1.05 1.13 1.18 1.01 1.03 1.03

favor of questions related to directly estimating Y j j k .Fur- S2 1.08 1.12 1.25 1.15 1.18 1.30 1.20 1.13 1.18
thermore, while this question was phrased differently for

zyxwvuts
S3 0.86 0.78 1.17 1.08 0.77 0.88 0.99 1.00 1.03
each of the nine attributes as seen in the appendix, the
respondents were explained that they have to provide relative
rankings, using scale from one to ten, providing a measure of

zyxwvut
how well the candidate systems configurations satisfy in

zyxwvutsr
terms of the ratio of quality and requirements related to costs, TABLE IV
OF WEIGHTS
VALUES wik
performance, and other attributes. The data (responses) pro-
vided for the values of Yijk( i = 1 , 2 , 3 ; j = 1,2,3; k = Attribute
1,2, ,9) for each system and each problem are given in w,' 1 2 3 4 5 6 7 8 9
~~~

Tables I, 11, and III. We consider four classes 1, 2, 3, and 4 Class 1 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
for the analysis and evaluation of the systems at hand. Class2 1.00 0.82 0.82 0.76 0.62 0.78 0.80 0.90 0.92
Class3 1.00 1.00 1.00 0.87 0.75 0.69 0.50 0.42 0.32

zyxwvutsr
The values of the weights wik for each class are computed Class4 1.00 0.50 0.37 0.46 0.40 1.00 1.00 0.90 0.85
using relations for classes 1 to 4 (with CY = 0.8). Examples
of the computed values for these weights are seen in
Table IV, which were also arrived at by the authors with
consensus from respondents. TABLE V
FIGURESOF MERIT
The set of the figures of merits, denoted by F =
{ F l ,F2, F3,I$}, is computed for the four classes 1, 2, 3, Classes
and 4 using the relations seen in the appendix based on the CL 1 CL2 CL3 CLA
estimated values provided by the respondents. These values Multi-User 2.16 1.59 1.78 0.94
are seen in Table V. LAN 3.00 1.89 2.24 0.74
Minicomputer 2.67 1.90 2.21 0.90
Using the values of the set F given in Table V , the
candidate computer configurations (i.e., systems S , , S, , S,)
can be ranked according to any of the selected classes. For
example, in the instance where management is only consider-
ing class 1, then F, is the optimal figure of merit of choice,
and system S , (the multi-user system) embodies the best ate conclusion is possible about an elective choice for the
solution in that regard; the next best is S, (the Minicomputer). current example. We discuss further aspects of this result in
The same analysis can be carried out for each class type. the trade-off analysis.
However, there is no computer configuration which repre- Some other observations are possible at this point. Clearly,
sents the overall best scenario for every class type (i.e., each there is no worst choice overall. Also note that there is at
decision option) for the data on hand. Therefore, no immedi- least one class type where each system performs worst.
254 zyxw
zyxwvutsrqponmlkjihg
zyxwv IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 38, NO. 3 , AUGUST 1991

Furthermore, based on the overall rankings we can infer that involves changing one or more input values for W j k and y j k
there is a tie between the LAN and Minicomputer solutions (i.e., q j j kand r i k )and recomputing new figures of merit as
( S , and S 3 ) . While the LAN solution ranks last in class 1 and seen in Table V.
class 3, it comes 2nd and 1st in classes 2 and 4, respectively. The results of this experimental survey may be understood
The Minicomputer solution, on the other hand, exhibits a by what we offer as the conventional wisdom in computer
more balanced performance, coming second in classes 1, 3, system development. The multi-user configuration ( S , ) is
and 4; and last in class 2. Hence, we see that if the decision gaining wider acceptance as more powerful microcomputer
makers have different class preferences, there would be no hosts e.g., 386 and 486-based machines, often with multiple
system solution alternative that is acceptable to all of them. CPUs, are becoming available in the marketplace. However,
Therefore, a trade-off analysis would be required, and a they are only now being seen as becoming powerful enough
coordinator (or coordinators) from QAIS would lead a search for serious “CPU-bound” applications e.g., for problem 2
for the “best acceptable compromise solution.” The next and similar types of applications serving many concurrent
section briefly describes the process of trade-off analysis, the users. According to several respondents, their low price in
final phase in the quantitative evaluation and selection proce- comparison with other systems, and ease of setup may ex-
dure. plain why this solution came first in 3 out of 4 classes. The
Minicomputer configuration, on the other hand, is definitely
OF THE TRADE-OFF
SUMMARY ANALYSIS
the traditional solution of the three for a sizable number of
In the final phase, the coordinator Do and the decision users, and the evaluations appear to reflect this. Based on
makers Dm must get together to evaluate the best solution attributes such as reliability, CPU and 1 / 0 throughput, one
system resulting from the quantitative decision and selection can readily see the reason for the more balanced score.
methodology, and from consensus based on their own prefer- However, it was indicated that this particular minicomputer
ence structure. Thus, this phase involves group decision-mak- configuration suffers from a much smaller software base than
ing which will be led and controlled by the coordinator, a the other options in addition to its much higher cost.
member of QAIS. The coordinator will attempt to assess and According to the results, the LAN does not seem to be
modify each decision maker’s preference structure (choice of very well accepted by the respondents. However, when asked
weights, classes or requirements) until a mutually agreeable to comment on the LAN, they indicated that they value its
solution is found. While each decision maker is concerned flexibility based on potential centralized and distributed proc-
with the benefit of his subsystem (reflected in his choices, essing modes, the wide availability of software, and system
that is, his utility function), the coordinator focuses on ways robustness. This is reflected in the fact that the LAN does not
to secure best possible compromise solution achieving the lag far behind the other systems, and in fact, ties with the
best acceptable quality attainment and use of resources. The mini-based system in cumulative rankings. In addition, they
process naturally involves trade-off compromises in iterative

zyxwvutsrqpon
also point to the rapid adoption of LANs by organizations,
steps. During each iteration, alternative attributes and weights but highlight the difficulty of installing and maintaining LANs
(or classes) may be modified and a new solution derived for as continuing concerns.
evaluation. The process stops when the coordinator and the
decision makers feel that the differences between their prefer- CONCLUSION
ences are insignificant. Note that this trade-off analysis proce- This paper investigates the problem of precoordination of
dure is well documented in the management science literature quality control activities for QAIS. The concept of QAIS is
(see VI, t 171). based on a strategic information center for planning and
To illustrate this phase, the authors and several respond- coordinating quality control activities within the factory sys-
ents gathered in follow-up discussions to analyze the input tem. As such it ties together through quality related func-
and output on hand. During this phase, the authors assumed tions, design, manufacturing, distribution, sales and other
the role of the coordinator and the respondents assumed the factory management systems to optimize the overall factory
roles of decision makers. It was decided to leave the choice performance.
for classes unchanged as they only reflect the objective QAIS supervision and leadership efforts are achieved
functions, which although very important, remain determinis- through its coordination and precoordination activities. Pre-
tic in the examples on hand. The choice of objective func- coordination is a new concept which is directed at 1) organiz-
tions may affect the final solution, but do not alter the ing the participation of all subsystems in the factory network
tradeoff process. Further discussions were carried out about in the planning of new product development, and 2 ) provid-
the weights and requirements. Each respondent was asked to ing methods for system evaluation and trade-off analysis for
provide convincing justifications for his choice of the values better solutions and integration.
for Y i j k , and to provide modifications as he saw fit. Further- This paper presents a method for quantitative evaluation
more, he also had to determine how much change in his input and selection of complex systems satisfying the desired prod-
he was willing to accept if others would require such changes. uct (or process) quality requirements from a number of the
That is, how much he was willing to trade off for his desired factory’s subsystems. This method is based on the develop-
changes. Hence, the values of wik and y i j k , seen in Tables I ment of a figure of merit combining both problem and system
to IV have been derived through a compromise procedure specifications, and permitting the evaluation and selection of
which lasted few iterations. In real practice, an iteration the most appropriate system from a set of candidate systems
 zyxwvut
zyxwvutsrqp
SyLL.4 AND ARINZE QUALITY PRECOORDINATION IN QUALITY ASSURANCE INFORMATION SYSTEM

suitable for a wide range of applications for which the system Class 4
255

can be used.
A case example is given to demonstrate how the method m n

can be used in the evaluation and trade-off analysis for the

selection of a suitable candidate system. Practical steps on
how to implement all phases of precoordination in this sce-
nario are also presented. No example was given to illustrate Any other classes (or choices of objective function) may be
the phases for the participation of all subsystems in the suitably considered for evaluation based on appropriate weight
planning of new product development. However, the method- selection.
ology for the quantitative evaluation and selection remains
essentially the same, and is in the case illustration, subse-
quent steps providing means for organizing the different
subsystems involved in decision-making in new product se- REFERENCES
lection represent the same mechanism for structured integra- S. H. Ahmed and M. G. Darwish, “Etude et evaluation des systemes
informatiques repartis: Application sur quelques algorithmes paralle-
tion between groups, which is also moderated by a coordina- les,” in ISMM Int. Conf. on Mini- and Microcomputer Rec.,
tor. It should be recognized that in both scenarios, organiza- Paris, 1982.
tional culture and procedures will largely dictate the form of B. W. Boehm, Software Engineering Economics, Englewood Cliffs,
NJ: Prentice-Hall, 1981.

zyxwvu
this interaction, and that this may present differences between C. Chang, “The structure of quality information system in a computer
the two suggested areas of application. Presently, how to deal integrated manufacturing environment, ” Computers and Industrial
with the organizational culture and how to set the critical Engineering, vol. 15, pp. 338-343, 1988.
G. B. Davis, “Strategies for information requirements determination,”
success factors for the interaction problem presents a much IBM System Journal, vol. 21, no. 1 , 1982.
needed area for continued research. M. I. Dessouky, S. A. Kapoor, and R. E. Devor, “A methodology
In summary, the paper shows how QAIS can be used to for integrated quality systems,” Journal of Engineering f o r Indus-
try, vol. 109, pp. 241-247, Aug. 1987.
deal with the coordination and precoordination problems in T. DeMarco, Structured Analysis and System Specification. New
manufacturing as well as in business functions of the factory York: Yourdon, Inc., 1979.
system. This focus stems from the knowledge of the impor- P. H. Farquhar, “Utility assessment methods: State of the Art,”
Management Science, vol. 30, no. 1 1 , 1984.
tance of quality and the realization that it is impossible to A. V. Feigenbaum, Total Quality Control. New York: McGraw-Hill,
achieve optimal quality in the absence of total system integra- 1961.
tion. Furthermore, it emerges from the realization that in 0. Furukawa, H. Ikeshoji, and H. Ishizuchu, “A theoretical study on
configurations of quality control systems,” IEEE Trans. Eng.
today’s marketplace, business information systems are strate- Manag., EM-29, no. 4, pp. 128-134, 1982.
gic in that they vitally affect profitability and survival just as 0. Furukawa, M. Kogure, S. Ishizu, The Japan Steel Works, Ltd.,
quality affects productivity and competitiveness. Hence, this Hiroshima Plant, “System approach to QC on job production,”
ASQC Quality Congress Trans., San Francisco, CA, pp. 255-262,

zyxwvutsrqpon
paper deals with quality and business information and maps 1981.
out a way to successfully coordinate them for effective com- M. P. Groover, Automation, Production Systems, and Com-
pany-wide quality assurance. puter-Integrated Manufacturing. Englewood Cliffs, NI: Prentice-
Hall, 1987.
R. V. Head, and M. S. Goff, “Standard benchmarks aid in competi-

zyxwvutsr
APPENDIX tive systems selection,” Journal of Systems Management, pp.
Using wik as given by (5) to (S), we can write the figure of 4-10, Jan. 1979.

zyxwvutsrqpo
C. Hsu, and C. Skevington, “Integration of data and knowledge in
merit 6. which takes the following forms for the classes manufacturing enterprises: a conceptual framework,” Journal of
discussed above: Manufacturing Systems, vol. 6 , no. 4 , pp. 277-284, 1987.
K. Ishikawa, TQC Story. Tokyo, Japan: Kajima Shuppankai, 1981.

zyxwvutsrqpo
zyxwvutsrqponm
Class 1 -, “Quality and standardization: Program for economic success,”
Quality Progress, pp. 16-20, 1984.
m n S. Ishizu, 0. Furukawa, H. Ikeshoji, “A theoretical study on adaptive
Fj= c I Y i j k - 11.
i=l k=l
quality control systems,” International Journal of General Sys-
tems, vol. 10, nos. 2 and 3, pp. 89-104, 1985.
R. L. Keeney and H. Raiffa, Decision with Multiple Objectives,
Preferences and Value Tradeoffs. New York: Wiley, 1976.
Class 2 J. Meredith, “Implementing the Automated Factory,” Journal of
Manufacturing Systems, vol. 6, no. 1 , pp. 1-13, 1987.
~~

M. D. Mesarovic, D. Macko, and Y. Takahara, General Systems

Theory: Mathematical Foundations. New York: Academic, 1970.
A. P. Sage, Methodology f o r Large Scale Systems. New York:
McGraw-Hill, 1977.
C. Sylla, “Quality assurance information requirements planning,”
Class 3 Computers and Industrial Engineering, vol. 15, no. 1-4, pp.
302-306, 1988.
m r t C. Sylla, and K. Toraskar, “Quality, Standardization and Interrela-
tionships: Some Manufacturing Implications,” Manufacturing Re-
view, vol. 3, no. 1 , pp. 6-15, 1990.
F. Takei, “Engineering quality improvement through TQC Activity,”
IEEE Trans. Eng. Manag., vol. 33, no. 2, pp. 92-95, 1986.
M. Tribus, “Managing to Survive in a Competitive Word,” Society
of Automotive Engineers, Detroit, MI, 1983.
256 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 38, NO. 3, AUGUST 1991

Cheickna Sylla was born in 1950 in Kayes, Re- sional journals. He is a member of ORSA and TIMS, IIE, and The Human
zyx
zyxw
public of Mali, West Africa. He received the B.S. Factors Society.
degree in electrical/mechanical engineering in
1975, from the Ecole Nationale D’Ingenieurs, Ba- Bay Arinze received the M.Sc. and Ph.D. degrees
mako, Republic of Mali, and the M.S. and Ph.D. in systems analysis from the London School of
degrees in industrial engineering from SUNY at Economics and Political Science, London, Eng-
Buffalo, NY, in 1980 and 1983, respectively. land.
He is currently an Associate Professor in the He is presently an Assistant Professor of Man-
School of Industrial Management at the New Jer- agement Information Systems in the Department of
sey Institute of Technology, Newark, NJ. He has Management, Drexel University, Philadelphia, PA.
ureviouslv served as a facultv in the Eneineerine His current research interests include knowledge-

zyxwvutsrqponm
Management Program at Drexel University, Philadelphia, PA. His-researci based systems and their applications in operations
interests include applied operations research techniques in manufacturing management, and decision support system design
quality assurance systems, management information systems, and the model- methodologies and applications.
ing of training systems. Dr. Arinze has published articles in Journal of Management Informa-
Dr. Sylla has published articles in IEEE TRANSACTIONS ON SYSTEMS, tion Systems, IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, Com-
MAN,and CYBERNETICS, Human Factors, Manufacturing Review, Com- puters and Industrial Engineering, Industrial Marketing Management,
puter and Industrial Engineering, Decision Sciences and other profes- and the ACM Sigart Special Issue on Knowledge Acquisition.