STATISTICAL METHODOLOGIES FOR QUALITY AND PRODUCTION IMPROVEMENT
| Published date | 01 September 1988 |
| DOI | https://doi.org/10.1108/eb057523 |
| Date | 01 September 1988 |
| Pages | 21-24 |
| Author | T.N. Goh |
| Subject Matter | Economics,Information & knowledge management,Management science & operations |
STATISTICAL METHODOLOGIES
FOR QUALITY AND PRODUCTION
IMPROVEMENT
by
T.N.
Goh
Industrial and Systems Engineering Department, National University of
Singapore
Introduction
"It is a capital mistake to theorise before one has the data" — so says Sherlock Holmes, summarising in a
nutshell the first principle of all empirical work. "Data" is that which arises from observed information, and
in a narrower sense is taken to refer to recorded numbers carrying such information. This management awareness
article discusses how data can be used for quality and productivity improvement, quite apart from the usual
behavioural and technical approaches familiar to most managers and technical personnel.
With the premise that efficient techniques of information collection, analysis and interpretation are essential
to management decisions on matters related to quality and productivity, explanations will be presented in
a non-mathematical
language
of the role of statistics in quality and productivity studies, with particular emphasis
on the active approach to empirical studies — seeking strategic changes and optimised performances rather
than maintenance of the status quo. The practical implications of the application of statistical methodologies
will be highlighted, together with some practical examples.
Conventional Approaches to the Control
of Quality
The concepts and techniques of quality control in
manufacturing have evolved over the years and are still
developing as the demand on quality increases. The
earliest ideas of "control" lie in product inspection: if
a manufactured product fails to meet certain quality
requirements, it is rejected; only those that meet the
requirements are released to the users. Because 100
per cent inspection is often impractical, various
sampling plans have been devised, all based on the
concept of detection of products that are non-
conforming to specifications.
It is quite obvious that the detection approach has one
serious drawback: by the time non-conforming products
are produced and detected, the damage has already
been done. The function of a sampling plan lies in its
ability to limit the spread of such damage to the product
users,
not its ability to salvage the production. Thus a
better philosophy is that of damage prevention, i.e.
aiming at not producing non-conforming products in
the first place. To achieve this, the focus of control has
to be shifted from the product to the process that
generates the product. The manufacturing process
should be adjusted or halted once non-conforming
products are produced or are about to be produced;
this is the central idea behind process control.
Quality Control by Statistical Principles
There is no assurance that subjective intervention of
a process will achieve the control purpose satisfactorily.
Over-reaction to fluctuations in sampled product
property may lead to frequent and unnecessary
stoppages in production or excessive adjustments in
process parameters, while untimely reactions would
result in frequent runs of non-conforming product. The
solution to such problems depends on the ability of the
quality engineer to distinguish between systematic and
random shifts in product characteristics and process
performance; thus theories of probability and statistics
become essential, and their applications have through
the years been developed into what is now commonly
referred to as statistical process control: process
capability studies and Shewhart control chart
procedures are common examples of statistical process
control applications[1].
Linkage between Statistics and
Productivity
Today, no manufacturing activities can claim to have
the quality of the output under control without the
support of statistical evidence. In fact, one can take a
broad view of the relationship between statistics and
productivity by considering the primary concerns the
various categories of technical personnel involved in the
creation of a manufactured product have:
(1) Fundamental research scientists and engineers
are interested in whether and why a certain
incident occurs.
(2) Applied research engineers are interested in how
a desired incident occurs.
(3) Development engineers are interested in making
the incident occur in a desired or specified way.
IMDS
September/October
1988
21
To continue reading
Request your trial