An expert system approach to quality control


An expert system approach to quality control

E.P. Paladini*

Departmento de Engenharia de produção e Sistemas, Universidade Federal de Santa Catarina, CP 476, Trindade, 88010-970, Florianopolis, SC, Brazil

Abstract

This paper describes the basic actions proposed for quality inspection. These actions involve structuring a Decision Supporting Expert
System (DSES) to help with those decisions related to the preliminary activities for inspection development —most of them relating to
determining of the need or convenience of carrying out the inspection itself. Once the opportunity to carry it out is defined, the expert system
helps the user to select the type of inspection to adopt from amongst: (1) automatic or sensorial inspection; (2) inspection by samples or
complete; (3) acceptance or rectifying and, in the most relevant module, (4) inspection by attributes or by variables. The complementary
documentation of the DSES contains the directions to operate it, the rules and qualifiers that make up the system, as well as the results
achieved through its experimental implementation. q 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Expert system; Quality control; Total quality management

1. Introduction

Having the current concept of quality in mind (and the
context of Total Quality Management itself), the feasibility
process of basic quality control actions is nowadays consid-
ered to take place within a well-defined structure, known as
quality evaluation system.

Inspection, in turn, is the most important activity in the
quality evaluation system of an industrial process. When
correctly developed, the inspection makes possible to
carry out a precise analysis of how the process operates
and serves as a basis for a set of decisions that directly affect
it, such as corrective and preventive actions which must be
complied with in order to guarantee acceptable quality
levels.

Quality inspection has got a number of effective techni-
ques with a wide range of applicability. Since there are
several techniques available, there are many options for
those who intend to devise an inspection process. Never-
theless, this situation poses difficulties as to the correct use
of the different techniques, since most of such techniques
have their own utilization particularities and yield results
valid only within certain contexts.

The need of choosing the most adequate technique for
each situation shows that it is necessary to organize the
information related to quality inspection, or else the whole
quality evaluation process could be seriously compromised.
The present paper highlights this question, which is

considered to be a relevant restriction to the perfect use of
quality evaluation. A more efficient way of optimizing qual-
ity evaluation development is proposed.

We dedicate special attention to quality inspection by
attributes, a much more difficult area when it comes to
making the correct decision about quality evaluation of
services, products and processes.

The literature about quality inspection is plentiful and
easily accessible. More often than not, classical texts on
quality control deal with this subject by focusing on proce-
dures of sampling for acceptance, whereas rectifying
inspection is rather rarely analyzed. The development of
the theoretical models of the Operating Characteristic
(OC) Curve associated with basic concepts, such as Accep-
table Quality Level (AQL), Lot Tolerance Percent Defec-
tive (LTPD) and producer’s and consumer’s risk, is always
presented as a means of motivating the analysis of sampling
schemes or sampling plans and inducing the procedures
which aim at structuring such plans and making feasible
their performance analysis, mostly in terms of reliability
and costs. Tables suggesting models of sampling plans are
just as relevant and are present as well in almost all books of
the area. The following authors can be cited as classical in
this field: Besterfield (1990), Charboneau and Webster
(1997), Juran (1999), and Montgomery (1998), to name a
few. Texts of renowned authors, such as Dodge and Deming
can be found in more specialized journals. A collection of
articles by Dodge about inspection by attributes was made
available by the American publication called Journal of
Quality Technology in 1977 (see Dodge, 1977). As for the
work of Deming, articles on this subject written by him can

Expert Systems with Applications 18 (2000) 133–151PERGAMON

Expert Systems
with Applications

0957-4174/99/$ - see front matter q 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 7 - 4 1 7 4 ( 9 9 ) 0 0 0 5 9 - 7

www.elsevier.com/locate/eswa

* Tel.: 155-48-331-7000; fax: 155-48-331-7075.
E-mail address: paladini@iaccess.com.br (E.P. Paladini).



be found in the same journal published in 1985 (Papadakis,
1985).

The types of inspection are manifold, each of them bear-
ing their particularities, specific purposes and restricted
adequacy to certain contexts. Such is the case of sampling
inspection or complete inspection, as well as inspection by
attributes or by variables, for instance.

The choice of the correct type of inspection is extremely
relevant for quality evaluation, since the adequacy of the
inspection process to the context where it takes place has
been considered in many situations and its importance has
always been stressed.

Let us consider the inspection of raw materials as an
example. An incorrect development of this type of inspec-
tion can give the suppliers the impression that quality is not
important to their customers and of course the suppliers will
not have to regard it as relevant. Such a situation, which can
be seen in practice, is to be found in several texts. Horsnell
(1988), for example, draws attention to this fact, by showing
that the use of inspection by attributes when products enter
the factory is of such usefulness that it transcends inspection
proper, in addition to being an easily applicable technique.
In a broader analysis, one can see that the quality of the
process can be significantly (and positively) affected by
the mere inclusion of inspection in the process (it has
been considered many years ago—see, for instance,
Whittle, 1964).

Quality inspection is widely used nowadays. Nonethe-
less, several problems have been encountered upon using
inspection systems and such problems derive for the most
part from an inadequate application of the prescribed meth-
odology. This inadequacy results from decisions which
ought to be made as regards the way the inspection is to
be carried out.

Such decisions are considered here to be intuitive.
However, they do deserve detailed investigation. We under-
stand that the lack of an adequate methodology for planning
the decision-making process is a fundamental restraint to the
use of quality inspection.

This situation justifies the elaboration of a general Deci-
sion-Making Support System. Due to its peculiarities, the
system was organized under the format of an expert system,
which benefits from the advantages of artificial intelligence
(AI) in order to carry out an evaluation.

It should be noted that since inspection is the most impor-
tant part of Quality Evaluation, the expert system proposed
here plays a fundamental role in the Total Quality Manage-
ment area.

It is easy to see that this paper describes a process of
interaction between users and computers in industrial
companies. The process becomes feasible through the
development and application of an expert system related
to the area of quality control. This paper describes the devel-
oped expert system, its operating methodology and lists the
results obtained after its application in the companies we
have studied. The interaction between the human resources

of the organizations and the computational resources was
intense in the last few years. With the advent of AI this
interaction has acquired new, rather specific characteristics.

2. Artificial intelligence applied to quality evaluation

A brief review of the specific technical literature reveals
that there are many cases of successful applications of Al
techniques to the Quality Evaluation area. Such applications
were helpful in the resolution of relevant problems.

Below are some examples thereof:

(a) Hosni and Elshennavy (1988) reported the develop-
ment of a quality control system based on knowledge,
suitable to specific procedures of inspection by variables
and to the selection of control graphs, both by attributes
and by variables.
(b) Eyada (1990) developed an expert system aimed at
auditing procedures of Quality Assurance involving both
suppliers and products in the process. Some years earlier,
in the same field, Gipe and Jasinski (1986) conducted an
analysis of expert system adequacy to Quality Assurance
problems, showing the feasibility of this application.
Another interesting expert system was developed by
Crawford and Eyada (1989) with the purpose of planning
the allocation of resources for the Quality Assurance
program.
(c) There are several expert systems used to select control
graphs (see Alexander & Jagannathan, 1986; Dagli, 1990;
Dagli & Stacey, 1988). An expert system designed for
Process Control at a general level was developed by
Moore (1995).
(d) Evans and Lindsay (1987) reported the development
of an expert system for Statistic Quality Control, which
not only selects control graphs, but also offers interpreta-
tions of such graphs and provides conclusions about the
control status of a process.
(e) Brink and Mahalingam (1990) developed an expert
system which evaluates quality at manufacturing level, so
as to detect and correct defects occurring during the
productive process.
(f) Pfeifer (1989) reported the development of an expert
system to detect defects during the productive process
and described its successful application in Germany.
(g) An expert system which makes use of pattern recog-
nition in order to ‘visualize’ pieces under inspection was
developed in 1989 in the United States (see Ntuen, Park
& Kim, 1989). This system, called KIMS, was success-
fully tested in a variety of experiments (one of them is
reported by Ntuen, Park & Sohn, 1990), where the perfor-
mance of the KIMS in image recognition tasks is
assessed.
(h) Fard and Sabuncuoglu (1990) developed an expert
system which seeks to select sampling by attributes,
determining which type of sampling is the most adequate
for each case: simple, double or multiple. A project of a

E.P. Paladini / Expert Systems with Applications 18 (2000) 133–151134



https://isiarticles.com/article/4495