ISO 2859-0:1995
(Main)Sampling procedures for inspection by attributes - Part 0: Introduction to the ISO 2859 attribute sampling system
Sampling procedures for inspection by attributes - Part 0: Introduction to the ISO 2859 attribute sampling system
Explains the terms used in acceptance sampling, describes the various schemes and plans, gives practical advice on sampling inspection and discusses some of the theoretical aspects. This edition cancels and replaces ISO 2859.
Règles d'échantillonnage pour les contrôles par attributs — Partie 0: Introduction au système d'échantillonnage par attributs de l'ISO 2859
Sampling procedures for inspection by attributes - Part 0: Introduction to the ISO 2859 attribute sampling system
General Information
Relations
Frequently Asked Questions
ISO 2859-0:1995 is a standard published by the International Organization for Standardization (ISO). Its full title is "Sampling procedures for inspection by attributes - Part 0: Introduction to the ISO 2859 attribute sampling system". This standard covers: Explains the terms used in acceptance sampling, describes the various schemes and plans, gives practical advice on sampling inspection and discusses some of the theoretical aspects. This edition cancels and replaces ISO 2859.
Explains the terms used in acceptance sampling, describes the various schemes and plans, gives practical advice on sampling inspection and discusses some of the theoretical aspects. This edition cancels and replaces ISO 2859.
ISO 2859-0:1995 is classified under the following ICS (International Classification for Standards) categories: 03.120.30 - Application of statistical methods. The ICS classification helps identify the subject area and facilitates finding related standards.
ISO 2859-0:1995 has the following relationships with other standards: It is inter standard links to ISO 2859-10:2006. Understanding these relationships helps ensure you are using the most current and applicable version of the standard.
You can purchase ISO 2859-0:1995 directly from iTeh Standards. The document is available in PDF format and is delivered instantly after payment. Add the standard to your cart and complete the secure checkout process. iTeh Standards is an authorized distributor of ISO standards.
Standards Content (Sample)
INTERNATIONAL
Is0
STANDARD 2859-O
First edition
1995-08-01
Sampling procedures for inspection by
attributes -
Part 0:
Introduction to the IS0 2859 attribute
sampling system
R&g/es d ’khantillonnage pour /es contr6les par attributs -
Pat-tie 0: Introduction au systeme d ’khantillonnage par attributs de
I ’ISO 2859
Reference number
IS0 2859-0:1995(E)
IS0 2859-O: 1995( E)
Contents
Page
General .
Section 1
...........................................................................................
1.1 Scope
..................................................................
1.2 Normative references
.................................................................................... 1
1.3 Definitions
......... 2
Section 2 General introduction to acceptance sampling
....................................................
2.1 Aim of sampling inspection
.............................................................
2.2 Acceptance sampling
2.2.1 Statistical sampling .
.................................................................
2.2.2 Ad hoc sampling
................................................................
2.2.3 100 % inspection
................................................ 3
2.2.4 Other “sampling” practices
......... 3
2.3 Choosing between attributes and variables inspection
2.4 Lot inspection .
........................................................................................
2.4.1 Lot
................................................................................
2.4.2 Lot size
....................................... 4
2.5 Sequence or isolated lot inspection
2.51 Lot-by-lot inspection .
2.5.2 Isolated lot inspection . 5
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.6 Acceptable quality level (AM
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.6.1 Description
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.6.2 Setting an AQL . .
2.7 Process average . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.8 Limiting quality (LQ)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.9 Normal and tightened inspection
2.10 Reduced inspection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*. 6
0 IS0 1995
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced
or utilized in any form or by any means, electronic or mechanical, including photocopying and
microfilm, without permission in writing from the publisher.
International Organization for Standardization
Case Postale 56 l CH-1211 Geneve 20 l Switzerland
Printed in Switzerland
ii
0 IS0 IS0 2859-O: 1995( E)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.11 Switching rules
AOQL
2.12 Average outgoing quality (AOQ) and its limit (’
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.13 Item; unit of product
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Nonconformity and nonconforming item
2.14
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.14.1 Failure to conform
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.14.2 Nomenclature
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.14.3 Classification of nonconformities
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.15 Critical nonconformities
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.16 Curtailment of inspection
. . . . . . . . . . . . . . . . .a. 14
2.17 Disposal of unacceptable lots
2.18 Single sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.19 Operating characteristic (OC) curves
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.20 Double sampling
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.21 Multiple sampling
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.22 Sequential sampling
2.23 Skip-lot sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.24 Comparison of single, double, multiple and sequential
sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.24.1 Equivalent plans
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.24.2 Average sample size
2.25 Drawing of samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.26 Sampling plans, schemes and systems
2.27 Distributional characteristics (binomial, Poisson and
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
hypergeometric)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Section 3 The IS0 2859-l system
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1 Description of IS0 2859-l
3.2 Preparing a specification for use in conjunction with IS0
2859-l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Classification of nonconformities and nonconforming items 28
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Lots
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Meaning of inspection level 32
3.6 Setting an inspection level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
0 IS0
IS0 2859-O: 1995(E)
....................................................................
3.7 Preferred AQLs
.............................................................
3.8 Non-preferred AQLs
..................................................................... 38
3.9 Setting an AQL
. . 39
3.10 Drawing a sampling plan from the tables in IS0 2859-l
.............................................................. 40
3.11 Normal inspection
..........................................................
3.12 Tightened inspection
............................................... 45
3.13 Switching rules - Example
.......................... 47
3.14 Methods for reducing the sampling risks
............................................................
3.15 Reduced inspection
............................................ 51
3.16 Double and multiple sampling
................................... 52
3.17 Limiting quality and the isolated lot
......................................................................
3.18 Sample sizes
..........................................
3.19 Operating characteristic curves
............................... 53
3.20 The AOQL table (see 2.12 for theory)
........................................................................
3.21 Nomograms
iv
Foreword
IS0 (the International Organization for Standardization) is a worldwide
federation of national standards bodies (IS0 member bodies). The work
of preparing International Standards is normally carried out through IS0
technical committees. Each member body interested in a subject for
which a technical committee has been established has the right to be
represented on that committee. International organizations, governmental
and non-governmental, in liaison with ISO, also take part in the work. IS0
collaborates closely with the International Electrotechnical Commission
(IEC) on all matters of electrotechnical standardization.
Draft International Standards adopted by the technical committees are
circulated to the member bodies for voting. Publication as an International
Standard requires approval by at least 75 % of the member bodies casting
a vote.
International Standard IS0 2859-O was prepared by Technical Committee
lSO/TC 69, Applications of statistical methods, Subcommittee SC 5, Ac-
ceptance sampling.
This first edition of IS0 2859-O cancels and replaces IS0 2859:1974 and
Addendum 1 :I 977.
IS0 2859 consi sts of the following parts under the general title Sampling
I
procedures for
inspection by attributes:
- Part 0: Introduction to the IS0 2859 attribute sampling system
- Part I: Sampling plans indexed by acceptable quantity level (AQL)
for lot-by-lot inspection
- Part 2: Sampling plans indexed by limiting quality (LQ) for isolated lot
inspection
- Part 3: Skip-lot sampling procedures
V
0 IS0
IS0 2859-O: 1995(E)
Introduction
This general introduction to sampling inspection describes the attribute
sampling schemes set forth in parts 1 to 3 of IS0 2859 and in IS0 8422.
This introduction treats the subject of sampling inspection by attributes in
a general way; it introduces the essential operating procedures and the
ways the schemes were designed to be used. To understand fully the
concepts and their applications, it would be helpful to consult IS0 2859-1,
IS0 2859-2, IS0 2859-3, IS0 8422 and lSO/rR 8550.
The individual parts of these International Standards extend this introduc-
tory explanation to more specific uses of the procedures which are ap-
propriate for the particular part or standard.
It is emphasized that IS0 2859-l provides sampling schemes indexed by
AQL. The quality measure used can be percent nonconforming or the
number of nonconformities per 100 items. IS0 2859-l was developed
primarily for the inspection of a continuing series of lots all originating
from the same source, as in this situation adequate protection (of the
maximum process average percent nonconforming) is possible by use of
the switching rules (i.e. from normal to tightened inspection) should a
certain (limiting) number of unacceptable lots be found in a short series
of successive lots.
IS0 2859-2 provides sampling plans arranged for use when individual or
isolated lots are to be sampled. These sampling plans are in many in-
stances identical to those in IS0 2859-l. All the tables of sampling plans
in IS0 2859-2 include information regarding the quality level required to
assure a high probability of lot acceptance. It is recommended that
IS0 2859-2 rather than IS0 2859-l be used for individual or isolated lots.
IS0 2859-3 provides skip-lot procedures for use when the process quality
is markedly superior to the AQL for a defined long period of delivery or
observation. When the quality level is in this state of excellence, it is
sometimes more economical to use IS0 2859-3 than to use the reduced
sampling procedure of IS0 2859-l. Like IS0 2859-1, IS0 2859-3 is appli-
cable to a continuing series of lots from a single source.
IS0 8422 provides a method of establishing sequential sampling plans of
discriminatory power essentially equivalent to that of individual plans of
IS0 2859-l and IS0 2859-2.
A complementary system of sampling plans for inspection by variables,
also indexed by AQL, is provided by IS0 3951 :I 989, Sampling procedures
and charts for inspection by variables for percent nonconforming and by
IS0 8423: 1991, Sequential sampling plans for inspection by variables for
percent nonconforming (known standard deviation).
NOTE 1 Use of the masculine gender in this part of IS0 2859 is not meant to
exclude the feminine gender where applied to persons. Similarly, use of the
singular does not exclude the plural (and vice versa) when the sense allows.
INTERNATIONAL STANDARD 0 IS0 IS0 2859-O: 1995( E)
Sampling procedures for inspection by attributes -
Part 0:
Introduction to the IS0 2859 attribute sampling system
Section 1: General
part of IS0 2859 are encouraged to investigate the
1.1 Scope
possibility of applying the most recent editions of the
standards indicated below. Members of IEC and IS0
This part of IS0 2859 explains the terms used in ac-
maintain registers of currently valid International
ceptance sampling, describes the various schemes
Standards.
and plans, gives practical advice on sampling in-
spection and discusses some of the theoretical as-
IS0 2859-l : 1989, Sampling procedures for inspection
pects.
Part 1: Sampling plans indexed by
by attributes -
quality level (AQL) for lot-by-lot
Section 2 gives general information on methods of acceptable
acceptance sampling inspection with particular refer- inspection.
ence to the sampling procedures and tables for in-
IS0 2859-2: 1985, Sampling procedures for inspection
spection by attributes given in parts 1, 2 and 3 of
by attributes - Part 2: Sampling plans indexed by
IS0 2859 and in IS0 8422.
limiting quality (LQ) for isolated lot inspection.
Section 3 extends the introduction to acceptance
sampling given in Section 2 and amplifies the intro- IS0 2859-3: 1991, Sampling procedures for inspection
ductory text and instructions contained in IS0 2859-1,
by attributes - Part 3: Skip-lot sampling procedures.
by giving detailed comments and examples to assist
in using the method of sampling inspection that con- IS0 8422:1991, Sequential sampling plans for in-
spection by attributes.
stitutes the IS0 2859-l sampling system.
lSO/TR 8550:1994, Guide for the selection of an ac-
ceptance sampling system, scheme or plan for in-
1.2 Normative references
spection of discrete items in lots.
The following standards contain provisions which,
through reference in this text, constitute provisions
1.3 Definitions
of this part of IS0 2859. At the time of publication, the
editions indicated were valid. All standards are subject
For the purposes of this part of IS0 2859, the defi-
to revision, and parties to agreements based on this
nitions given in IS0 2859-l and IS0 2859-3 apply.
IS0 2859-O: 1995(E)
Section 2: General introduction to acceptance sampling
a) sampling based on experience with the product,
2.1 Aim of sampling inspection
the process, the supplier and the consumer (see
2.2.1);
A major aim of acceptance sampling inspection is to
see that the producer submits lots at a quality which
b) ad hoc sampling, for example the inspection of a
is at or better than a mutually agreed level, so that the
fixed percentage, or occasional random checks
consumer receives lots of a quality that is acceptable.
(see 2.2.2);
The producer may use these sampling procedures to
c) 100 % inspection (see 2.2.3);
assure that the quality level will be acceptable to the
consumer. In all these procedures, it has to be re-
d) other “sampling” practices (see 2.2.4).
cognized that the financial resources are not unlimi-
ted. The cost of the article has to reflect the cost of
inspection as well as the cost of production.
2.2.1 Statistical sampling
A real effort should be made to ensure that a system
Sampling based on experience with the product, the
is devised that clearly places responsibility for quality
process, the producer and the consumer can be stat-
with the producer. Inspection can appear to divert the
istically evaluated.
responsibility for quality from the producer to the in-
spector. This may happen whenever there is a feeling
An example is the procedure described in IS0 2859-l
that the inspector is there to sort things out, so that,
which uses a set of switching rules. When quality is
within limits, what happens in production will be
very good, it is possible to go to reduced inspection.
caught by inspection. This feeling is completely mis-
This provides a procedure where, if smaller samples
placed and may result in hard work, high cost and
are used, the producer ’s risk is reduced but the con-
poor quality for the consumer and the producer. The
sumer ’s risk is increased. If experience is good, this
inspector has no means of inserting quality into a
is justifiable, particularly when the process average
product if the producer has not done so.
has been consistently smaller than the acceptable
quality level (AQL) specified. When the process aver-
age over at least IO lots has been very much smaller
than the AQL, some consumers resort to skip-lot
procedures (see IS0 2859-3). This can be even more
2.2 Acceptance sampling
economical than the reduced inspection described in
IS0 2859-l.
Acceptance sampling inspection has the merit of put-
ting the responsibility for quality fairly and squarely In some instances, particularly when routine or non-
critical items are involved, some consumers may feel
where it belongs - with the producer. The inspector
safe in resorting to the practice of inspecting small
is no longer regarded as the person who sorts things
samples of the product and, provided there are zero
out. The producer has to see that the quality of the
nonconforming items, accepting the lot. For example,
product is right, otherwise there will be much trouble
with a sample size of eight this is equivalent to the
and expense with unacceptable lots. Sampling in-
spection can and should lead to less inspection work, small lot single sampling plans with an AQL of I,5 %
lower cost and good quality for the consumer. normal, or 0,65 % reduced inspection. See tables II-A
and II-C in IS0 2859-1:1989.
The sampling inspection schemes of parts 1, 2 and 3
Conversely, in IS0 2859-1, when two out of five suc-
of IS0 2859 and of IS0 8422 provide for
cessive lots fail acceptance, normal inspection is dis-
quantification of the risks of accepting unsatisfactory
continued, and tightened inspection is instituted.
product (known as “consumer ’s risk ”) and the risks
Once tightened inspection has been instituted, normal
of not accepting satisfactory product ( “producer ’s
inspection is not restored until five successive lots
risk ”) and for choosing a plan that allows no more risk
have been accepted on tightened inspection. This re-
than is acceptable.
quirement is intentionally severe, because evidence
In addition to the IS0 sampling plans which are based
of unacceptable quality has been found. The producer
on the mathematical theory of probability, there are
then forfeits the right to the benefit of the doubt. If,
several other practices:
while operating on tightened inspection, the cumulat-
0 IS0
IS0 2859-O: 1995( E)
ive number of lots not accepted on original tightened statistical control of his production process and is
inspection reaches five, inspection by sampling forthright in his evaluation of what is being shipped.
should be discontinued until there is evidence that
If, in a particular situation, information is available of
corrective action has been taken and has been effec-
the true costs of the mistaken non-acceptance of
tive. See 2.11.
good articles and the acceptance of bad ones, and if
from long experience it is known how often lots of
2.2.2 Ad hoc sampling
any given quality are presented, this may be one of
the occasions when compromise is not desirable. It
Ad hoc sampling is not to be recommended as it will
may be possible to calculate a more efficient scheme
lead to uncalculated risks that may be unjustifiably
on the basis of the economic information available.
high; furthermore, there is no formal basis for either
the acceptance or non-acceptance of the lot.
2.3 Choosing between attributes and
2.2.3 100 % inspection
variables inspection
100 % inspection can be a formidable task unless the
The attributes method of inspection consists of
100 % inspection is performed with automatic test
examining an item, or characteristics of an item, and
equipment. In addition, it is not always successful,
classifying the item as “conforming” or “noncon-
particularly when a large number of items have one
forming ”. The action to be taken is decided by
or more characteristics that are marginal dimen-
counting the number of nonconforming items or the
sionally, in appearance or in performance (close to or
number of nonconformities found in a random sam-
concentrated about a tolerance or limit of appearance
ple.
or performance). Under these conditions, sorting by
manual or automatic methods is likely to classify
The variables method starts with selecting a sample
some conforming items as nonconforming, and vice
of a number of items and measuring dimensions or
versa. In addition, 100 % testing by manual, visual or
characteristics so that information is available not only
automatic methods can be unsatisfactory. It can
on whether a dimension, for example, is within certain
sometimes degenerate into superficial 100 % in-
limits but on the actual value of the dimension. The
spection when, in fact, sufficient money, time and
decision whether or not to accept a lot is made on the
staff are not available. 100 % inspection is not viable
basis of calculations of the average and the variability
if the inspection method necessitates destructive
of the measurements in accordance with the pro-
testing.
cedures of IS0 3951 or IS0 8423.
It has to be understood, however, that 100 % in-
Provided certain assumptions are true, the variables
spection may form a necessary part of the inspection
method has the advantage of requiring a smaller
process for both the consumer and the producer.
sample size than the attributes method to attain a
There are situations in which it cannot be avoided, for
given degree of protection against incorrect decisions.
example when inspecting for critical nonconformities,
Also it provides more information as to whether qual-
as will be clear from a study of this part of IS0 2859.
ity is being adversely affected by process mean, pro-
Some types of nonconformity are so important that
cess variability or both. The attributes method has the
every item has to be examined when tests are non-
advantage that it is more robust (not subject to as-
destructive. When the tests are destructive, some risk
sumptions of distributional shape) and that it is sim-
has to be accepted. (See 2.15).
pler to use. The larger sample sizes and the increased
costs associated with using attribute sampling meth-
2.2.4 Other “sampling” practices
ods may be justifiable for these reasons.
Various sampling systems exist but only those avail-
It should be noted that go, no-go gauging is faster and
able as International Standards will be considered in
requires less skill than measurement.
detail in this part of IS0 2859. This should not be
taken as meaning that the others are unimportant; it Both methods have advantages and typical fields of
is merely that the main purpose of this part of application.
IS0 2859 is to help people to use parts 1, 2 or 3 of
Although occasionally reference is made to IS0 3951
IS0 2859 or IS0 8422.
and IS0 8423 in subsequent clauses, variables
In many instances, consumers do not perform any
schemes, as such, are not considered further in this
regular sampling but rely on their experience and
part of IS0 2859. IS0 3951 and IS0 8423 include
supporting evidence that the producer is maintaining guidance on their use.
0 IS0
IS0 2859-O: 1995(E)
smaller proportion of the lot than with small lots for
2.4 Lot inspection
the same AQL.
This “large lot” policy should not be overdone, how-
ever. If making up a large lot necessitates putting to-
2.4.1 Lot
gether smaller lots that could have remained separate,
For the purposes of IS0 2859, IS0 8422, IS0 3951 then a large lot is advantageous only if the smaller lots
and IS0 8423, items are offered for acceptance in are of a similar quality. If there is likely to be any
groups, not on a single item basis. Each group of substantial difference between the qualities of the
items is called a lot. smaller lots, then it is much better to keep them
separate. For this reason, lots should consist of items
Each lot should, as far as is practicable, consist of
of product produced under essentially the same con-
items manufactured under essentially the same con-
ditions.
ditions during one time period. This is of the utmost
importance if the acceptable quality level concept is Examples of the formation of lots are given in 3.4.
adopted and there are a series of lots to be delivered. More information on the lot size/sample size relation
is given in lSO/TR 8550.
If two or more sources of supply are mixed, the
presence of a large number of nonconforming items
from one of the sources can result in non-acceptance 2.5 Sequence or isolated lot inspection
of the product from all the sources. Conversely,
product of marginal quality from one source can be
2.5.1 Lot-by-lot inspection
masked by mixing with product from sources of ex-
cellent quality.
Lot-by-lot inspection is the inspection of product sub-
mitted in a series of lots.
From each lot a sample is drawn and inspected. Un-
der attributes inspection, each lot is classified as ac-
If a sequence of lots is to be offered for acceptance
ceptable or unacceptable on the basis of the number
at the time of production, the inspection results from
of nonconforming items or nonconformities found in
the preceding lots can be avaible before the later lots
the sample. Each successive lot is therefore dealt
are made. It is therefore possible that the inspection
with as a more or less independent unit (although the
performed can beneficially influence the quality of
rules for sentencing may sometimes vary according
subsequent production. The lots should be submitted
to the results from preceding lots).
and inspected in the same sequence as they are
should be made
manufactured and inspection
For single lots offered in isolation, see 2.5.2.
promptly. Information obtained from a lot may indicate
that the process appears to have deteriorated. The
information obtained from several lots in sequence
2.4.2 Lot size
can be used to invoke a switching procedure which
requires the use of a more rigorous sampling pro-
in
The responsible authority (see 3.10
cedure in the event that the process deteriorates. This
IS0 2859-l :1989) has the right to specify what is to
is important because, in the long run, it provides the
be the lot size, but it clearly makes sense that, where
best protection a consumer has against poor quality.
possible, this should be done in consultation with the
If the quality remains poor, then under the more rig-
producer, so that a quantity that is mutually con-
orous sampling practice more lots will be returned to
venient may be chosen. Certainly, specifying the lot
the vendor for sorting. This tighter sampling increases
size (and other parameters of the sampling plan)
the producer ’s risk of having an acceptable lot judged
should never be done in ignorance of the production
unacceptable. The identification of possible deterio-
process. It is not essential that an inviolable quantity
ration in product quality is a signal to initiate corrective
should be chosen. Sometimes variation may be al-
action.
lowed, although it will nearly always be desirable that
upper and lower limits of the lot size should be
If the quality is very much better than that agreed
specified.
upon, the consumer may, with the permission of the
responsible authority, elect to adopt reduced or skip-
From the sampling inspection point of view, there is
lot sampling.
an advantage in large lots, as from a large lot it is
IS0 2859-1, IS0 2859-3, IS0 8422, IS0 3951 and
economical to take a large sample, thereby achieving
better discrimination between good lots and bad IS0 8423 are designed principally for use with a se-
ones. With large lots, the required sample size is a quence of lots.
0 IS0
IS0 2859-O: 1995( E)
acceptable. It is always better to have no noncon-
2.5.2 Isolated lot inspection
forming items than any percentage whatever, and the
Inspection may sometimes be performed on an iso- more the percentage can be reduced below the AQL
lated lot, just a few isolated lots, or on stored lots at the better. This reduction improves the probability that
a time when production has been finished. Under
each lot is accepted.
these circumstances, there is insufficient opportunity
for the switching rules to be invoked and hence to
2.6.2 Setting an AQL
influence the quality to be offered.
In setting an AQL, it has to be remembered that the
If a single lot is to be delivered, then it is helpful to
AQL provides an indication of the quality that is re-
know whether the lot is one of many similar lots de-
quired in production. The producer is being asked to
livered to other consumers and consists of material
produce lots of an average quality better than the
from a controlled process or whether it is a mixed lot
AQL. On the one hand, this quality has to be reason-
containing items from different processes and differ-
ably attainable, whilst on the other hand it has to be
ent times. (See also lSO/TR 8550 and IS0 2859-2).
a reasonable quality from the consumer ’s point of
view. Frequently this will mean a compromise be-
Whereas IS0 2859-l and IS0 2859-3 call for estab-
tween the quality the consumer would like and the
lishing the AQL value and the inspection level in ad-
quality he can afford, for the tighter the requirement
vance, IS0 2859-2 requires the establishment of the
the more difficult it may be for the production to meet
limiting quality (LO). In order to provide appropriate
it, and the more expensive may be the inspection to
producer and consumer protection when lots are
ensure that it is met.
sampled under the limiting-quality procedure, infor-
mation is needed as to whether the lot came from a
A properly designed and controlled process may be
continuing series of acceptable lots, or is a mixed lot,
capable of producing product with a smaller percent-
consisting of product made on different production
age nonconforming than the AQL value. When a bet-
lines and/or different dates.
ter process average is obtainable from a process, the
cost of production plus the cost of inspection will be
The tables in IS0 2859-2 are designed principally for
lower for the better quality.
use with isolated lots.
The primary consideration has to be the consumer ’s
2.6 Acceptable quality level (AQL)
requirement, but it is necessary to make sure that the
consumer is being realistic and is not demanding
something tighter than is really needed. It is necess-
2.6.1 Description
ary to take into account how the items in question are
to be used and the consequences of a failure. If the
The acceptable quality level is used as an indexing
items are to be available in large numbers and the
device in the tables of IS0 2859-1, in IS0 3951, and
failure is simply a failure to assemble so that the
in some of the tables of IS0 8422 and IS0 8423.
nonconforming item can be put aside and another
When using these AQL-indexed sampling plans, in-
used in its place, a relatively generous AQL may be
spection lots taken from a process whose quality is
tolerable. If, on the other hand, a failure is going to
equal to or better than the AQL will be accepted most
cause a failure to function of an expensive and im-
of the time.
portant piece of equipment at a time and place where
a replacement of the nonconforming item cannot be
When a continuing series of lots is considered, the
made, a tighter AQL will be required.
AQL is a quality level which for the purposes of sam-
pling inspection is the limit of a satisfactory process
More information and guidance on setting an AQL is
average.
given in 3.9 and in lSO/TR 8550.
The AQL is a chosen borderline between what will be
considered acceptable as a process average, and
2.7 Process average
what will not. As such, it in no way describes a sam-
pling plan, but is a requirement of what the production
The process average is the average quality submitted
should be like, and is a useful quantity to consider in
over a series of lots, resubmitted lots being excluded.
defining a tolerable process.
It is particularly important to realize that, in contrast
The fact that an AQL is specified should not be taken
to the AQL, the AOQL (see 2.12) or the LQ (see
to imply that a percentage of nonconforming items
2.8), the process average is not something that can
up to the specified value is wanted, or is completely
be calculated or chosen, or is a property of a particular
0 IS0
IS0 2859-O: 1995(E)
sampling plan. The process average relates to what is Limiting quality plans are used primarily for isolated
actually produced, irrespective of what inspection is lots. When the product is in manufacture and there
are a series of lots being produced, the procedures
performed.
of IS0 2859-l are more appropriate.
Generally, the estimation of a process average is not
an essential part of a sampling scheme. However, the
2.9 Normal and tightened inspection
process average is important in its own right. Both the
inspector and the producer are interested not only in
An AQL, it will be remembered, is the borderline in
the lot-by-lot decisions but also in the long-term pic-
the quality scale between the good and the bad when
ture of the quality of production.
a sequence of lots is inspected. When the AQL has
been specified for any particular product, the ideal
It is, therefore, desirable to keep a record of the
would be to have a system whereby lots could be al-
overall estimated process average being achieved
ways accepted when their quality was better than the
because this gives a useful measure of quality and is
AQL and always not accepted when worse than the
also invaluable information for those who have to de-
AQL. This ideal is not attainable with any sampling
cide what sampling plans should be adopted when
plan.
similar products are being designed and made in the
future.
To meet the requirements of both the producer and
the consumer, some compromise is needed, and the
Special rules need to be observed where the sampling
device adopted in IS0 2859-l and IS0 8422 is to join
is of the double or multiple form. Only the results of
normal inspection with tightened inspection; i.e. two
the first sample in double and multiple sampling
sampling plans are specified for any given situation,
should be used to estimate the process average.
together with rules for determining when to switch
from one to the other and when to switch back again.
Occasionally a recommendation is made that abnor-
mal results should be excluded. This is a dangerous
Normal inspection is designed to protect the producer
practice that should be used very sparingly, if at all.
against having a high proportion of lots not accepted
The only time this practice may safely be adopted is
when quality is better than the AQL. In fact, the pro-
if the abnormal results are known to be due to a
ducer is being given the benefit of any doubt that
specific cause which is known to have been elimin-
arises due to sampling variability.
ated. Even then it is good practice to quote figures
which include and which exclude these abnormal re-
But the consumer needs protection too, and this is
sults to indicate that these nonconformities did exist.
achieved by arranging that the producer is not given
the benefit of the doubt blindly and invariably, but only
Separate process averages have to be estimated in
for as long as he proves worthy of it. If at any time the
the case of multiple characteristics or multiple AQL
sampling results show that his process average is
classes.
probably worse than the AQL, he forfeits his right to
the benefit of the doubt (that is, his right to normal
inspection), and tightened inspection is instituted to
2.8 Limiting quality (LQ)
protect the consumer.
Limiting quality is an indexing device used in
Further details with examples are contained in 3.11
IS0 2859-2. When a lot is considered in isolation, LQ
and 3.12.
is a quality level in percent nonconforming (or non-
conformities per 100 items) which for the purposes
2.10 Reduced inspection
of sampling inspection is limited to a low probability
of acceptance. This small probability of acceptance is
Sometimes there is evidence that the product quality
called the “consumer ’s risk ”.
is consistently better than the AQL. Where this hap-
Specifying a limiting quality is in fact the specification
pens and there is reason to believe that good pro-
of a quality that is not wanted! To have lots regularly
duction will continue, sampling inspection no longer
prove to be acceptable, the fraction of nonconforming
serves the purpose of segregating the good lots from
items has to be much smaller than the LQ (usually the bad ones. However, inspection cannot be dis-
less than a quarter of the LQ).
pensed with altogether, as a warning is needed if the
production quality worsens.
IS0 2859-2 provides procedures for the application of
LQ sampling plans. These sampling plans and tables
In these circumstances, considerable savings can be
are for the most part consistent with the sampling
made if so desired by using the reduced-inspection
plans used in IS0 2859-l. sampling plans described in IS0 2859-l or the skip-lot
0 IS0
IS0 2859-O: 1995( E)
sampling plans described in IS0 2859-3. The special There is one further safeguard for the consumer. This
rules for allowing the use of these plans, if permitted is the rule that acceptance inspection should be dis-
by the responsible authority, are described in continued, pending action to improve the quality, if
IS0 2859-1, IS0 2859-3 and also in Section 3 of this the cumulative number of lots not accepted in a se-
part of IS0 2859. quence of consecutive lots on original tightened in-
spection reaches five. This is a most important
Reduced inspection is further discussed, with exam-
principle; if the quality is bad, action is needed, and
ples, in 3.15.
the inspector has to be entitled to refuse to inspect
any further lots until he has evidence that suitable
action has been taken.
An example is given in 3.13.
2.11 Switching rules
Subclause 2.9 introduced normal inspection and
2.12 Average outgoing quality (AOQ)
tightened inspection and their purpose. This sub-
and its limit (AOQL)
clause discusses the switching rules by means of
which the decision is taken to change from normal to
As with the AQL concept, the concept of average
tightened inspection or back again when using
outgoing quality and its limit is only meaningful when
IS0 2859-l.
a long sequence of lots is submitted to a defined
system of sampling inspection, e.g. in accordance
If the actual value of the quality being offered by the
with the provisions of IS0 2859-l. When the number
producer were known, the knowledge would be used
of nonconforming items in the sample is equal to or
to sentence the lots instead of submitting them to
less than the acceptance number, the lot will be ac-
acceptance inspection. As the actual quality is never
cepted. Conversely, when the number of noncon-
known, the best that can be done is to use the
forming items in the sample is equal to or greater than
knowledge that is available, i.e. the sampling in-
the rejection number, the lot will not be accepted.
spection results themselves.
When the supply (or source) process operates at a
As normal inspection is designed to accept nearly all
process average close to the specified AQL, most of
the lots offered, provided that the quality is at least
the lots will be accepted. Provided that process qual-
as good as the AQL, it follows that if a high proportion
ity is constant and non-accepted lots are discarded
of lots is not being accepted, the quality cannot be as
rather than rectified, the effect of sampling on the
good as the AQL. The question is: “What proportion
quality is nil.
of non-acceptance is high enough to be convincing?”
In some instances, particularly when the transfer is
A rule is required that will give reasonably quick re-
between departments rather than companies, the re-
action if quality becomes worse than the AQL, while
sult of a lot failing to pass sampling inspection is that
having a low probability of calling, in error, for tight-
the lot is 100 % inspected and the nonconforming
ened inspection when the quality is really better than
items removed (and perhaps replaced with conform-
the AQL.
ing items). This is termed “rectifying inspection ”.
The rule used is that tightened inspection has to be
When lots are submitted to rectifying inspection, the
used for the following lots as soon as two out of any
lot is either accepted with no further inspection or,
five or fewer successive lots on original inspection
when the sample indicates non-acceptance, all the
have not been accepted. The qualification “on original
items in the lot are inspected and nonconforming
inspection”
means that if lots are not accepted but
items discarded or replaced by conforming items. In
resubmitted after rectification, these resubmitted lots
the first case, the outgoing quality is, for practical
are not counted for switching-rule purposes.
purposes, the same as the incoming quality; in the
Once tightened inspection has been instituted, it re-
second case, all items conform to the specification.
mains in force for every lot until five successive lots
Even though the incoming quality may be constant at
have been accepted on tightened inspection, then
p (fraction nonconforming) the process average, the
normal inspection is restored. This requirement is
outgoing quality will vary from lot to lot, taking either
quite a severe one, as acceptance on tightened in-
the value p or zero depending on whether the lot is
spection is more difficult than on normal inspection,
accepted on the sample result or is subjected to rec-
but once there is evidence that quality worse than the
tifying inspection. It is possible, however, to think of
AQL has been produced, the producer ’s right to the
the average of these outgoing qualities over a long run
benefit of the doubt cannot be restored until it is safe
in which the incoming quality was constant at p. This
to do so.
average of the outgoing quality will clearly not be
0 IS0
IS0 2859-O: 1995(E)
greater than p and, where a large proportion of lots is Performi
...
SLOVENSKI STANDARD
01-september-1996
Sampling procedures for inspection by attributes - Part 0: Introduction to the ISO
2859 attribute sampling system
Sampling procedures for inspection by attributes - Part 0: Introduction to the ISO 2859
attribute sampling system
Ta slovenski standard je istoveten z:
ICS:
03.120.30 8SRUDEDVWDWLVWLþQLKPHWRG Application of statistical
methods
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.
INTERNATIONAL
Is0
STANDARD 2859-O
First edition
1995-08-01
Sampling procedures for inspection by
attributes -
Part 0:
Introduction to the IS0 2859 attribute
sampling system
R&g/es d ’khantillonnage pour /es contr6les par attributs -
Pat-tie 0: Introduction au systeme d ’khantillonnage par attributs de
I ’ISO 2859
Reference number
IS0 2859-0:1995(E)
IS0 2859-O: 1995( E)
Contents
Page
General .
Section 1
...........................................................................................
1.1 Scope
..................................................................
1.2 Normative references
.................................................................................... 1
1.3 Definitions
......... 2
Section 2 General introduction to acceptance sampling
....................................................
2.1 Aim of sampling inspection
.............................................................
2.2 Acceptance sampling
2.2.1 Statistical sampling .
.................................................................
2.2.2 Ad hoc sampling
................................................................
2.2.3 100 % inspection
................................................ 3
2.2.4 Other “sampling” practices
......... 3
2.3 Choosing between attributes and variables inspection
2.4 Lot inspection .
........................................................................................
2.4.1 Lot
................................................................................
2.4.2 Lot size
....................................... 4
2.5 Sequence or isolated lot inspection
2.51 Lot-by-lot inspection .
2.5.2 Isolated lot inspection . 5
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.6 Acceptable quality level (AM
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.6.1 Description
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.6.2 Setting an AQL . .
2.7 Process average . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.8 Limiting quality (LQ)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.9 Normal and tightened inspection
2.10 Reduced inspection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*. 6
0 IS0 1995
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced
or utilized in any form or by any means, electronic or mechanical, including photocopying and
microfilm, without permission in writing from the publisher.
International Organization for Standardization
Case Postale 56 l CH-1211 Geneve 20 l Switzerland
Printed in Switzerland
ii
0 IS0 IS0 2859-O: 1995( E)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.11 Switching rules
AOQL
2.12 Average outgoing quality (AOQ) and its limit (’
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.13 Item; unit of product
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Nonconformity and nonconforming item
2.14
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.14.1 Failure to conform
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.14.2 Nomenclature
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.14.3 Classification of nonconformities
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.15 Critical nonconformities
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.16 Curtailment of inspection
. . . . . . . . . . . . . . . . .a. 14
2.17 Disposal of unacceptable lots
2.18 Single sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.19 Operating characteristic (OC) curves
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.20 Double sampling
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.21 Multiple sampling
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.22 Sequential sampling
2.23 Skip-lot sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.24 Comparison of single, double, multiple and sequential
sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.24.1 Equivalent plans
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.24.2 Average sample size
2.25 Drawing of samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.26 Sampling plans, schemes and systems
2.27 Distributional characteristics (binomial, Poisson and
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
hypergeometric)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Section 3 The IS0 2859-l system
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1 Description of IS0 2859-l
3.2 Preparing a specification for use in conjunction with IS0
2859-l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Classification of nonconformities and nonconforming items 28
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Lots
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Meaning of inspection level 32
3.6 Setting an inspection level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
0 IS0
IS0 2859-O: 1995(E)
....................................................................
3.7 Preferred AQLs
.............................................................
3.8 Non-preferred AQLs
..................................................................... 38
3.9 Setting an AQL
. . 39
3.10 Drawing a sampling plan from the tables in IS0 2859-l
.............................................................. 40
3.11 Normal inspection
..........................................................
3.12 Tightened inspection
............................................... 45
3.13 Switching rules - Example
.......................... 47
3.14 Methods for reducing the sampling risks
............................................................
3.15 Reduced inspection
............................................ 51
3.16 Double and multiple sampling
................................... 52
3.17 Limiting quality and the isolated lot
......................................................................
3.18 Sample sizes
..........................................
3.19 Operating characteristic curves
............................... 53
3.20 The AOQL table (see 2.12 for theory)
........................................................................
3.21 Nomograms
iv
Foreword
IS0 (the International Organization for Standardization) is a worldwide
federation of national standards bodies (IS0 member bodies). The work
of preparing International Standards is normally carried out through IS0
technical committees. Each member body interested in a subject for
which a technical committee has been established has the right to be
represented on that committee. International organizations, governmental
and non-governmental, in liaison with ISO, also take part in the work. IS0
collaborates closely with the International Electrotechnical Commission
(IEC) on all matters of electrotechnical standardization.
Draft International Standards adopted by the technical committees are
circulated to the member bodies for voting. Publication as an International
Standard requires approval by at least 75 % of the member bodies casting
a vote.
International Standard IS0 2859-O was prepared by Technical Committee
lSO/TC 69, Applications of statistical methods, Subcommittee SC 5, Ac-
ceptance sampling.
This first edition of IS0 2859-O cancels and replaces IS0 2859:1974 and
Addendum 1 :I 977.
IS0 2859 consi sts of the following parts under the general title Sampling
I
procedures for
inspection by attributes:
- Part 0: Introduction to the IS0 2859 attribute sampling system
- Part I: Sampling plans indexed by acceptable quantity level (AQL)
for lot-by-lot inspection
- Part 2: Sampling plans indexed by limiting quality (LQ) for isolated lot
inspection
- Part 3: Skip-lot sampling procedures
V
0 IS0
IS0 2859-O: 1995(E)
Introduction
This general introduction to sampling inspection describes the attribute
sampling schemes set forth in parts 1 to 3 of IS0 2859 and in IS0 8422.
This introduction treats the subject of sampling inspection by attributes in
a general way; it introduces the essential operating procedures and the
ways the schemes were designed to be used. To understand fully the
concepts and their applications, it would be helpful to consult IS0 2859-1,
IS0 2859-2, IS0 2859-3, IS0 8422 and lSO/rR 8550.
The individual parts of these International Standards extend this introduc-
tory explanation to more specific uses of the procedures which are ap-
propriate for the particular part or standard.
It is emphasized that IS0 2859-l provides sampling schemes indexed by
AQL. The quality measure used can be percent nonconforming or the
number of nonconformities per 100 items. IS0 2859-l was developed
primarily for the inspection of a continuing series of lots all originating
from the same source, as in this situation adequate protection (of the
maximum process average percent nonconforming) is possible by use of
the switching rules (i.e. from normal to tightened inspection) should a
certain (limiting) number of unacceptable lots be found in a short series
of successive lots.
IS0 2859-2 provides sampling plans arranged for use when individual or
isolated lots are to be sampled. These sampling plans are in many in-
stances identical to those in IS0 2859-l. All the tables of sampling plans
in IS0 2859-2 include information regarding the quality level required to
assure a high probability of lot acceptance. It is recommended that
IS0 2859-2 rather than IS0 2859-l be used for individual or isolated lots.
IS0 2859-3 provides skip-lot procedures for use when the process quality
is markedly superior to the AQL for a defined long period of delivery or
observation. When the quality level is in this state of excellence, it is
sometimes more economical to use IS0 2859-3 than to use the reduced
sampling procedure of IS0 2859-l. Like IS0 2859-1, IS0 2859-3 is appli-
cable to a continuing series of lots from a single source.
IS0 8422 provides a method of establishing sequential sampling plans of
discriminatory power essentially equivalent to that of individual plans of
IS0 2859-l and IS0 2859-2.
A complementary system of sampling plans for inspection by variables,
also indexed by AQL, is provided by IS0 3951 :I 989, Sampling procedures
and charts for inspection by variables for percent nonconforming and by
IS0 8423: 1991, Sequential sampling plans for inspection by variables for
percent nonconforming (known standard deviation).
NOTE 1 Use of the masculine gender in this part of IS0 2859 is not meant to
exclude the feminine gender where applied to persons. Similarly, use of the
singular does not exclude the plural (and vice versa) when the sense allows.
INTERNATIONAL STANDARD 0 IS0 IS0 2859-O: 1995( E)
Sampling procedures for inspection by attributes -
Part 0:
Introduction to the IS0 2859 attribute sampling system
Section 1: General
part of IS0 2859 are encouraged to investigate the
1.1 Scope
possibility of applying the most recent editions of the
standards indicated below. Members of IEC and IS0
This part of IS0 2859 explains the terms used in ac-
maintain registers of currently valid International
ceptance sampling, describes the various schemes
Standards.
and plans, gives practical advice on sampling in-
spection and discusses some of the theoretical as-
IS0 2859-l : 1989, Sampling procedures for inspection
pects.
Part 1: Sampling plans indexed by
by attributes -
quality level (AQL) for lot-by-lot
Section 2 gives general information on methods of acceptable
acceptance sampling inspection with particular refer- inspection.
ence to the sampling procedures and tables for in-
IS0 2859-2: 1985, Sampling procedures for inspection
spection by attributes given in parts 1, 2 and 3 of
by attributes - Part 2: Sampling plans indexed by
IS0 2859 and in IS0 8422.
limiting quality (LQ) for isolated lot inspection.
Section 3 extends the introduction to acceptance
sampling given in Section 2 and amplifies the intro- IS0 2859-3: 1991, Sampling procedures for inspection
ductory text and instructions contained in IS0 2859-1,
by attributes - Part 3: Skip-lot sampling procedures.
by giving detailed comments and examples to assist
in using the method of sampling inspection that con- IS0 8422:1991, Sequential sampling plans for in-
spection by attributes.
stitutes the IS0 2859-l sampling system.
lSO/TR 8550:1994, Guide for the selection of an ac-
ceptance sampling system, scheme or plan for in-
1.2 Normative references
spection of discrete items in lots.
The following standards contain provisions which,
through reference in this text, constitute provisions
1.3 Definitions
of this part of IS0 2859. At the time of publication, the
editions indicated were valid. All standards are subject
For the purposes of this part of IS0 2859, the defi-
to revision, and parties to agreements based on this
nitions given in IS0 2859-l and IS0 2859-3 apply.
IS0 2859-O: 1995(E)
Section 2: General introduction to acceptance sampling
a) sampling based on experience with the product,
2.1 Aim of sampling inspection
the process, the supplier and the consumer (see
2.2.1);
A major aim of acceptance sampling inspection is to
see that the producer submits lots at a quality which
b) ad hoc sampling, for example the inspection of a
is at or better than a mutually agreed level, so that the
fixed percentage, or occasional random checks
consumer receives lots of a quality that is acceptable.
(see 2.2.2);
The producer may use these sampling procedures to
c) 100 % inspection (see 2.2.3);
assure that the quality level will be acceptable to the
consumer. In all these procedures, it has to be re-
d) other “sampling” practices (see 2.2.4).
cognized that the financial resources are not unlimi-
ted. The cost of the article has to reflect the cost of
inspection as well as the cost of production.
2.2.1 Statistical sampling
A real effort should be made to ensure that a system
Sampling based on experience with the product, the
is devised that clearly places responsibility for quality
process, the producer and the consumer can be stat-
with the producer. Inspection can appear to divert the
istically evaluated.
responsibility for quality from the producer to the in-
spector. This may happen whenever there is a feeling
An example is the procedure described in IS0 2859-l
that the inspector is there to sort things out, so that,
which uses a set of switching rules. When quality is
within limits, what happens in production will be
very good, it is possible to go to reduced inspection.
caught by inspection. This feeling is completely mis-
This provides a procedure where, if smaller samples
placed and may result in hard work, high cost and
are used, the producer ’s risk is reduced but the con-
poor quality for the consumer and the producer. The
sumer ’s risk is increased. If experience is good, this
inspector has no means of inserting quality into a
is justifiable, particularly when the process average
product if the producer has not done so.
has been consistently smaller than the acceptable
quality level (AQL) specified. When the process aver-
age over at least IO lots has been very much smaller
than the AQL, some consumers resort to skip-lot
procedures (see IS0 2859-3). This can be even more
2.2 Acceptance sampling
economical than the reduced inspection described in
IS0 2859-l.
Acceptance sampling inspection has the merit of put-
ting the responsibility for quality fairly and squarely In some instances, particularly when routine or non-
critical items are involved, some consumers may feel
where it belongs - with the producer. The inspector
safe in resorting to the practice of inspecting small
is no longer regarded as the person who sorts things
samples of the product and, provided there are zero
out. The producer has to see that the quality of the
nonconforming items, accepting the lot. For example,
product is right, otherwise there will be much trouble
with a sample size of eight this is equivalent to the
and expense with unacceptable lots. Sampling in-
spection can and should lead to less inspection work, small lot single sampling plans with an AQL of I,5 %
lower cost and good quality for the consumer. normal, or 0,65 % reduced inspection. See tables II-A
and II-C in IS0 2859-1:1989.
The sampling inspection schemes of parts 1, 2 and 3
Conversely, in IS0 2859-1, when two out of five suc-
of IS0 2859 and of IS0 8422 provide for
cessive lots fail acceptance, normal inspection is dis-
quantification of the risks of accepting unsatisfactory
continued, and tightened inspection is instituted.
product (known as “consumer ’s risk ”) and the risks
Once tightened inspection has been instituted, normal
of not accepting satisfactory product ( “producer ’s
inspection is not restored until five successive lots
risk ”) and for choosing a plan that allows no more risk
have been accepted on tightened inspection. This re-
than is acceptable.
quirement is intentionally severe, because evidence
In addition to the IS0 sampling plans which are based
of unacceptable quality has been found. The producer
on the mathematical theory of probability, there are
then forfeits the right to the benefit of the doubt. If,
several other practices:
while operating on tightened inspection, the cumulat-
0 IS0
IS0 2859-O: 1995( E)
ive number of lots not accepted on original tightened statistical control of his production process and is
inspection reaches five, inspection by sampling forthright in his evaluation of what is being shipped.
should be discontinued until there is evidence that
If, in a particular situation, information is available of
corrective action has been taken and has been effec-
the true costs of the mistaken non-acceptance of
tive. See 2.11.
good articles and the acceptance of bad ones, and if
from long experience it is known how often lots of
2.2.2 Ad hoc sampling
any given quality are presented, this may be one of
the occasions when compromise is not desirable. It
Ad hoc sampling is not to be recommended as it will
may be possible to calculate a more efficient scheme
lead to uncalculated risks that may be unjustifiably
on the basis of the economic information available.
high; furthermore, there is no formal basis for either
the acceptance or non-acceptance of the lot.
2.3 Choosing between attributes and
2.2.3 100 % inspection
variables inspection
100 % inspection can be a formidable task unless the
The attributes method of inspection consists of
100 % inspection is performed with automatic test
examining an item, or characteristics of an item, and
equipment. In addition, it is not always successful,
classifying the item as “conforming” or “noncon-
particularly when a large number of items have one
forming ”. The action to be taken is decided by
or more characteristics that are marginal dimen-
counting the number of nonconforming items or the
sionally, in appearance or in performance (close to or
number of nonconformities found in a random sam-
concentrated about a tolerance or limit of appearance
ple.
or performance). Under these conditions, sorting by
manual or automatic methods is likely to classify
The variables method starts with selecting a sample
some conforming items as nonconforming, and vice
of a number of items and measuring dimensions or
versa. In addition, 100 % testing by manual, visual or
characteristics so that information is available not only
automatic methods can be unsatisfactory. It can
on whether a dimension, for example, is within certain
sometimes degenerate into superficial 100 % in-
limits but on the actual value of the dimension. The
spection when, in fact, sufficient money, time and
decision whether or not to accept a lot is made on the
staff are not available. 100 % inspection is not viable
basis of calculations of the average and the variability
if the inspection method necessitates destructive
of the measurements in accordance with the pro-
testing.
cedures of IS0 3951 or IS0 8423.
It has to be understood, however, that 100 % in-
Provided certain assumptions are true, the variables
spection may form a necessary part of the inspection
method has the advantage of requiring a smaller
process for both the consumer and the producer.
sample size than the attributes method to attain a
There are situations in which it cannot be avoided, for
given degree of protection against incorrect decisions.
example when inspecting for critical nonconformities,
Also it provides more information as to whether qual-
as will be clear from a study of this part of IS0 2859.
ity is being adversely affected by process mean, pro-
Some types of nonconformity are so important that
cess variability or both. The attributes method has the
every item has to be examined when tests are non-
advantage that it is more robust (not subject to as-
destructive. When the tests are destructive, some risk
sumptions of distributional shape) and that it is sim-
has to be accepted. (See 2.15).
pler to use. The larger sample sizes and the increased
costs associated with using attribute sampling meth-
2.2.4 Other “sampling” practices
ods may be justifiable for these reasons.
Various sampling systems exist but only those avail-
It should be noted that go, no-go gauging is faster and
able as International Standards will be considered in
requires less skill than measurement.
detail in this part of IS0 2859. This should not be
taken as meaning that the others are unimportant; it Both methods have advantages and typical fields of
is merely that the main purpose of this part of application.
IS0 2859 is to help people to use parts 1, 2 or 3 of
Although occasionally reference is made to IS0 3951
IS0 2859 or IS0 8422.
and IS0 8423 in subsequent clauses, variables
In many instances, consumers do not perform any
schemes, as such, are not considered further in this
regular sampling but rely on their experience and
part of IS0 2859. IS0 3951 and IS0 8423 include
supporting evidence that the producer is maintaining guidance on their use.
0 IS0
IS0 2859-O: 1995(E)
smaller proportion of the lot than with small lots for
2.4 Lot inspection
the same AQL.
This “large lot” policy should not be overdone, how-
ever. If making up a large lot necessitates putting to-
2.4.1 Lot
gether smaller lots that could have remained separate,
For the purposes of IS0 2859, IS0 8422, IS0 3951 then a large lot is advantageous only if the smaller lots
and IS0 8423, items are offered for acceptance in are of a similar quality. If there is likely to be any
groups, not on a single item basis. Each group of substantial difference between the qualities of the
items is called a lot. smaller lots, then it is much better to keep them
separate. For this reason, lots should consist of items
Each lot should, as far as is practicable, consist of
of product produced under essentially the same con-
items manufactured under essentially the same con-
ditions.
ditions during one time period. This is of the utmost
importance if the acceptable quality level concept is Examples of the formation of lots are given in 3.4.
adopted and there are a series of lots to be delivered. More information on the lot size/sample size relation
is given in lSO/TR 8550.
If two or more sources of supply are mixed, the
presence of a large number of nonconforming items
from one of the sources can result in non-acceptance 2.5 Sequence or isolated lot inspection
of the product from all the sources. Conversely,
product of marginal quality from one source can be
2.5.1 Lot-by-lot inspection
masked by mixing with product from sources of ex-
cellent quality.
Lot-by-lot inspection is the inspection of product sub-
mitted in a series of lots.
From each lot a sample is drawn and inspected. Un-
der attributes inspection, each lot is classified as ac-
If a sequence of lots is to be offered for acceptance
ceptable or unacceptable on the basis of the number
at the time of production, the inspection results from
of nonconforming items or nonconformities found in
the preceding lots can be avaible before the later lots
the sample. Each successive lot is therefore dealt
are made. It is therefore possible that the inspection
with as a more or less independent unit (although the
performed can beneficially influence the quality of
rules for sentencing may sometimes vary according
subsequent production. The lots should be submitted
to the results from preceding lots).
and inspected in the same sequence as they are
should be made
manufactured and inspection
For single lots offered in isolation, see 2.5.2.
promptly. Information obtained from a lot may indicate
that the process appears to have deteriorated. The
information obtained from several lots in sequence
2.4.2 Lot size
can be used to invoke a switching procedure which
requires the use of a more rigorous sampling pro-
in
The responsible authority (see 3.10
cedure in the event that the process deteriorates. This
IS0 2859-l :1989) has the right to specify what is to
is important because, in the long run, it provides the
be the lot size, but it clearly makes sense that, where
best protection a consumer has against poor quality.
possible, this should be done in consultation with the
If the quality remains poor, then under the more rig-
producer, so that a quantity that is mutually con-
orous sampling practice more lots will be returned to
venient may be chosen. Certainly, specifying the lot
the vendor for sorting. This tighter sampling increases
size (and other parameters of the sampling plan)
the producer ’s risk of having an acceptable lot judged
should never be done in ignorance of the production
unacceptable. The identification of possible deterio-
process. It is not essential that an inviolable quantity
ration in product quality is a signal to initiate corrective
should be chosen. Sometimes variation may be al-
action.
lowed, although it will nearly always be desirable that
upper and lower limits of the lot size should be
If the quality is very much better than that agreed
specified.
upon, the consumer may, with the permission of the
responsible authority, elect to adopt reduced or skip-
From the sampling inspection point of view, there is
lot sampling.
an advantage in large lots, as from a large lot it is
IS0 2859-1, IS0 2859-3, IS0 8422, IS0 3951 and
economical to take a large sample, thereby achieving
better discrimination between good lots and bad IS0 8423 are designed principally for use with a se-
ones. With large lots, the required sample size is a quence of lots.
0 IS0
IS0 2859-O: 1995( E)
acceptable. It is always better to have no noncon-
2.5.2 Isolated lot inspection
forming items than any percentage whatever, and the
Inspection may sometimes be performed on an iso- more the percentage can be reduced below the AQL
lated lot, just a few isolated lots, or on stored lots at the better. This reduction improves the probability that
a time when production has been finished. Under
each lot is accepted.
these circumstances, there is insufficient opportunity
for the switching rules to be invoked and hence to
2.6.2 Setting an AQL
influence the quality to be offered.
In setting an AQL, it has to be remembered that the
If a single lot is to be delivered, then it is helpful to
AQL provides an indication of the quality that is re-
know whether the lot is one of many similar lots de-
quired in production. The producer is being asked to
livered to other consumers and consists of material
produce lots of an average quality better than the
from a controlled process or whether it is a mixed lot
AQL. On the one hand, this quality has to be reason-
containing items from different processes and differ-
ably attainable, whilst on the other hand it has to be
ent times. (See also lSO/TR 8550 and IS0 2859-2).
a reasonable quality from the consumer ’s point of
view. Frequently this will mean a compromise be-
Whereas IS0 2859-l and IS0 2859-3 call for estab-
tween the quality the consumer would like and the
lishing the AQL value and the inspection level in ad-
quality he can afford, for the tighter the requirement
vance, IS0 2859-2 requires the establishment of the
the more difficult it may be for the production to meet
limiting quality (LO). In order to provide appropriate
it, and the more expensive may be the inspection to
producer and consumer protection when lots are
ensure that it is met.
sampled under the limiting-quality procedure, infor-
mation is needed as to whether the lot came from a
A properly designed and controlled process may be
continuing series of acceptable lots, or is a mixed lot,
capable of producing product with a smaller percent-
consisting of product made on different production
age nonconforming than the AQL value. When a bet-
lines and/or different dates.
ter process average is obtainable from a process, the
cost of production plus the cost of inspection will be
The tables in IS0 2859-2 are designed principally for
lower for the better quality.
use with isolated lots.
The primary consideration has to be the consumer ’s
2.6 Acceptable quality level (AQL)
requirement, but it is necessary to make sure that the
consumer is being realistic and is not demanding
something tighter than is really needed. It is necess-
2.6.1 Description
ary to take into account how the items in question are
to be used and the consequences of a failure. If the
The acceptable quality level is used as an indexing
items are to be available in large numbers and the
device in the tables of IS0 2859-1, in IS0 3951, and
failure is simply a failure to assemble so that the
in some of the tables of IS0 8422 and IS0 8423.
nonconforming item can be put aside and another
When using these AQL-indexed sampling plans, in-
used in its place, a relatively generous AQL may be
spection lots taken from a process whose quality is
tolerable. If, on the other hand, a failure is going to
equal to or better than the AQL will be accepted most
cause a failure to function of an expensive and im-
of the time.
portant piece of equipment at a time and place where
a replacement of the nonconforming item cannot be
When a continuing series of lots is considered, the
made, a tighter AQL will be required.
AQL is a quality level which for the purposes of sam-
pling inspection is the limit of a satisfactory process
More information and guidance on setting an AQL is
average.
given in 3.9 and in lSO/TR 8550.
The AQL is a chosen borderline between what will be
considered acceptable as a process average, and
2.7 Process average
what will not. As such, it in no way describes a sam-
pling plan, but is a requirement of what the production
The process average is the average quality submitted
should be like, and is a useful quantity to consider in
over a series of lots, resubmitted lots being excluded.
defining a tolerable process.
It is particularly important to realize that, in contrast
The fact that an AQL is specified should not be taken
to the AQL, the AOQL (see 2.12) or the LQ (see
to imply that a percentage of nonconforming items
2.8), the process average is not something that can
up to the specified value is wanted, or is completely
be calculated or chosen, or is a property of a particular
0 IS0
IS0 2859-O: 1995(E)
sampling plan. The process average relates to what is Limiting quality plans are used primarily for isolated
actually produced, irrespective of what inspection is lots. When the product is in manufacture and there
are a series of lots being produced, the procedures
performed.
of IS0 2859-l are more appropriate.
Generally, the estimation of a process average is not
an essential part of a sampling scheme. However, the
2.9 Normal and tightened inspection
process average is important in its own right. Both the
inspector and the producer are interested not only in
An AQL, it will be remembered, is the borderline in
the lot-by-lot decisions but also in the long-term pic-
the quality scale between the good and the bad when
ture of the quality of production.
a sequence of lots is inspected. When the AQL has
been specified for any particular product, the ideal
It is, therefore, desirable to keep a record of the
would be to have a system whereby lots could be al-
overall estimated process average being achieved
ways accepted when their quality was better than the
because this gives a useful measure of quality and is
AQL and always not accepted when worse than the
also invaluable information for those who have to de-
AQL. This ideal is not attainable with any sampling
cide what sampling plans should be adopted when
plan.
similar products are being designed and made in the
future.
To meet the requirements of both the producer and
the consumer, some compromise is needed, and the
Special rules need to be observed where the sampling
device adopted in IS0 2859-l and IS0 8422 is to join
is of the double or multiple form. Only the results of
normal inspection with tightened inspection; i.e. two
the first sample in double and multiple sampling
sampling plans are specified for any given situation,
should be used to estimate the process average.
together with rules for determining when to switch
from one to the other and when to switch back again.
Occasionally a recommendation is made that abnor-
mal results should be excluded. This is a dangerous
Normal inspection is designed to protect the producer
practice that should be used very sparingly, if at all.
against having a high proportion of lots not accepted
The only time this practice may safely be adopted is
when quality is better than the AQL. In fact, the pro-
if the abnormal results are known to be due to a
ducer is being given the benefit of any doubt that
specific cause which is known to have been elimin-
arises due to sampling variability.
ated. Even then it is good practice to quote figures
which include and which exclude these abnormal re-
But the consumer needs protection too, and this is
sults to indicate that these nonconformities did exist.
achieved by arranging that the producer is not given
the benefit of the doubt blindly and invariably, but only
Separate process averages have to be estimated in
for as long as he proves worthy of it. If at any time the
the case of multiple characteristics or multiple AQL
sampling results show that his process average is
classes.
probably worse than the AQL, he forfeits his right to
the benefit of the doubt (that is, his right to normal
inspection), and tightened inspection is instituted to
2.8 Limiting quality (LQ)
protect the consumer.
Limiting quality is an indexing device used in
Further details with examples are contained in 3.11
IS0 2859-2. When a lot is considered in isolation, LQ
and 3.12.
is a quality level in percent nonconforming (or non-
conformities per 100 items) which for the purposes
2.10 Reduced inspection
of sampling inspection is limited to a low probability
of acceptance. This small probability of acceptance is
Sometimes there is evidence that the product quality
called the “consumer ’s risk ”.
is consistently better than the AQL. Where this hap-
Specifying a limiting quality is in fact the specification
pens and there is reason to believe that good pro-
of a quality that is not wanted! To have lots regularly
duction will continue, sampling inspection no longer
prove to be acceptable, the fraction of nonconforming
serves the purpose of segregating the good lots from
items has to be much smaller than the LQ (usually the bad ones. However, inspection cannot be dis-
less than a quarter of the LQ).
pensed with altogether, as a warning is needed if the
production quality worsens.
IS0 2859-2 provides procedures for the application of
LQ sampling plans. These sampling plans and tables
In these circumstances, considerable savings can be
are for the most part consistent with the sampling
made if so desired by using the reduced-inspection
plans used in IS0 2859-l. sampling plans described in IS0 2859-l or the skip-lot
0 IS0
IS0 2859-O: 1995( E)
sampling plans described in IS0 2859-3. The special There is one further safeguard for the consumer. This
rules for allowing the use of these plans, if permitted is the rule that acceptance inspection should be dis-
by the responsible authority, are described in continued, pending action to improve the quality, if
IS0 2859-1, IS0 2859-3 and also in Section 3 of this the cumulative number of lots not accepted in a se-
part of IS0 2859. quence of consecutive lots on original tightened in-
spection reaches five. This is a most important
Reduced inspection is further discussed, with exam-
principle; if the quality is bad, action is needed, and
ples, in 3.15.
the inspector has to be entitled to refuse to inspect
any further lots until he has evidence that suitable
action has been taken.
An example is given in 3.13.
2.11 Switching rules
Subclause 2.9 introduced normal inspection and
2.12 Average outgoing quality (AOQ)
tightened inspection and their purpose. This sub-
and its limit (AOQL)
clause discusses the switching rules by means of
which the decision is taken to change from normal to
As with the AQL concept, the concept of average
tightened inspection or back again when using
outgoing quality and its limit is only meaningful when
IS0 2859-l.
a long sequence of lots is submitted to a defined
system of sampling inspection, e.g. in accordance
If the actual value of the quality being offered by the
with the provisions of IS0 2859-l. When the number
producer were known, the knowledge would be used
of nonconforming items in the sample is equal to or
to sentence the lots instead of submitting them to
less than the acceptance number, the lot will be ac-
acceptance inspection. As the actual quality is never
cepted. Conversely, when the number of noncon-
known, the best that can be done is to use the
forming items in the sample is equal to or greater than
knowledge that is available, i.e. the sampling in-
the rejection number, the lot will not be accepted.
spection results themselves.
When the supply (or source) process operates at a
As normal inspection is designed to accept nearly all
process average close to the specified AQL, most of
the lots offered, provided that the quality is at least
the lots will be accepted. Provided that process qual-
as good as the AQL, it follows that if a high proportion
ity is constant and non-accepted lots are discarded
of lots is not being accepted, the quality cannot be as
rather than rectified, the effect of sampling on the
good as the AQL. The question is: “What proportion
quality is nil.
of non-acceptance is high enough to be convincing?”
In some instances, particularly when the transfer is
A rule is required that will give reasonably quick re-
between departments rather than companies, the re-
action if quality becomes worse than the AQL, while
sult of a lot failing to pass sampling inspection is that
having a low probability of calling, in error, for tight-
the lot is 100 % inspected and the nonconforming
ened inspection when the quality is really better than
items removed (and perhaps replaced with conform-
the AQL.
ing items). This is termed “rectifying inspection ”.
The rule used is that tightened inspection has to be
When lots are submitted to rectifying inspection, the
used for the following lots as soon as two out of any
lot is either accepted with no further inspection or,
five or fewer successive lots on original inspection
when the sample indicates non-acceptance, all the
have not been accepted. The qualification “on original
items in the lot are inspected and nonconforming
inspection”
means that if lots are not accepted but
items discarded or replaced by conforming items. In
resubmitted after rectification, these resubmitted lots
the first case, the outgoing quality is, for practical
are not counted for switching-rule purposes.
purposes, the same as the incoming quality; in the
Once tightened inspection has been instituted, it re-
second case, all items conform to the specification.
mains in force for every lot until five successive lots
Even though the incoming quality may be constant at
have been accepted on tightened inspection, then
p (fraction nonconforming) the process average, the
normal inspection is restored. This requirement is
outgoing quality will vary from lot to lot, taking either
quite a severe one, as acceptance on tightened in-
the value p or zero depending on whether the lot is
spection is more difficult than on normal i
...
NORME
ISO
INTERNATIONALE
2859-O
Première édition
1995-08-01
Règles d’échantillonnage pour les contrôles
par attributs -
Partie 0:
Introduction au système d’échantillonnage par
attributs de I’ISO 2859
Sampling procedures for inspection by attributes -
Part 0: Introduction to the /SO 2859 attribute sampling system
Numéro de référence
BS0 2859-O: 1995(F)
Sommaire
Page
.,.,,. 1
Section 1 Généralités
1.1 Domaine d’application . . . . . . .,.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Références normatives . . . .
........ .................................... ........... 2
1.3 Définitions .
Introduction générale à l’échantillonnage pour
Section 2
acceptation . . . . . . . . . . .*.*. 3
.......................... 3
2.1 Objectifs des contrôles par échantillonnage
2.2 Échantillonnage pour acceptation .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.1 Échantillonnage statistique
............... ......................................... 4
2.2.2 Échantillonnage ad hoc
2.2.3 Contrôle à 100 % . 4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.4 Autres méthodes ((d’échantillonnage))
2.3 Contrôle par attributs ou par mesures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 Contrôle de lots
2.4.1 Lot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4.2 Effectif des lots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.5 Contrôle de séquences de lots ou de lots isolés . . . . . . . . . . . . . . . . . . 6
2.5.1 Contrôle lot par lot . . 6
2.5.2 Contrôle d’un lot isolé . 6
2.6 Niveau de qualité acceptable (NQA) . . . . . . . . . . .a. 6
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.6.1 Description
2.6.2 Établissement d’un NQA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.7 Qualité moyenne d’un processus . . . . . . . . .a. 7
2.8 Qualité limite (QL) . . 7
2.6 Contrôle normal et contrôle renforcé . . . .“. 8
0 ISO 1995
Droits de reproduction réservés. Sauf prescription différente, aucune partie de cette publi-
cation ne peut être reproduite ni utilisée sous quelque forme que ce soit et par aucun pro-
cédé, électronique ou mecanrque, y compris la photocopie et les microfilms sans l’accord
écrit de l’éditeur
Organisation Internationale de normalisation
Case Postale 56 l CH-121 1 Genève 20 l Suisse
Imprimé en Suisse
ii
CJ KO ISO 2859-O: 1995( F)
...................................................................... 8
2.10 Contrôle réduit
..................................... 8
2.11 Règles de modification du contrôle
2.12 Qualité moyenne apres contrôle (QMAC) et limite de qualité
...................................... 9
moyenne après contrôle (LQMAC)
.................................................. 10
2.13 Individu (unité de produit)
2.14 Non-conformité et individu non conforme .
................................................... 11
2.14.1 Absence de conformité
.................................................................. 13
2.14.2 Nomenclature
................................. 13
2.14.3 Classification des non-conformités
2.15 Non-conformités critiques . . . . . . . . .~. 14
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.16 Contrôle tronqué . . . . . . . . . . . . . .
. . . . . . .a.*. 16
2.17 Devenir des lots non acceptab es
2.18 Échantillonnage simple . . . . . . . . . . . . . . . . . . . . . . .*.*.*. 17
...........................................................
2.19 Courbes d’efficacité
...................................................... 19
2.20 Échantillonnage double
2.21 Échantillonnage multiple . 20
2.22 Échantillonnage progressif . 20
.......................................
2.23 Échantillonnage successif partiel 20
2.24 Comparaison des échantillonnages simple, double, multiple et
progressif . 21
2.24.1 Plans équivalents . 21
2.24.2 Effectif moyen d’échantillon . 23
............................................
2.25 Prélèvement des échantillons 25
2.26 Plans, programmes et systèmes d’échantillonnage . . . . . . . . . . 28
2.27 Lois de distribution binomiale, de Poisson et
hypergéométrique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Section 3 Le systhme de IBIS0 2859-1 .*.m 30
3.1 Description de I’ISO 2859-1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 Élaboration de spécifications à utiliser en conjonction avec
I’ISO 2859-l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Classification des non-conformités et des individus non
conformes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Lots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Q3 ISO
ISO 2859-O: 1995(F)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.5 Signification du niveau de contrôle
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.6 Établissement d’un niveau de contrôle
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .-. 40
3.7 Valeurs recommandées du NQA
3.8 Autres valeurs du NQA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.9 Établissement d’un NQA
3.10 Sélection d’un plan d’échantillonnage à partir des tables de I’ISO
28531 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*.
.*. 44
3.11 Contrôle normal . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.12 Contrôle renforcé . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . 50
3.13 Règles de modification du contrôle - Exemple
3.14 Méthodes de réduction des risques associés à
l’échantillonnage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.15 Contrôle réduit
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.16 Échantillonnages double et multiple
3.17 Qualit limite et lot isolé . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.18 Effectifs d’échantillon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.19 Courbes d’efficacité . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.20 Tables LQMAC (voir théorie en 2.12) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.21 Nomogrammes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*. 59
iv
Q3 ISO ISO 2859-O: 1995(F)
Avant-propos
L’ISO (Organisation internationale de normalisation) est une fédération
mondiale d’organismes nationaux de normalisation (comités membres de
I’ISO). L’élaboration des Normes internationales est en général confiée aux
comités techniques de I’ISO. Chaque comité membre intéressé par une
étude a le droit de faire partie du comité technique créé à cet effet. Les
organisations internationales, gouvernementales et non gouvernemen-
tales, en liaison avec I’ISO participent également aux travaux. L’ISO colla-
bore étroitement avec la Commission électrotechnique internationale (CEI)
en ce qui concerne la normalisation électrotechnique.
Les projets de Normes internationales adoptés par les comités techniques
sont soumis aux comités membres pour vote. Leur publication comme
Normes internationales requiert l’approbation de 75 % au moins des co-
mités membres votants.
La Norme internationale ISO 2859-O a été élaborée par le comité techni-
que ISO/TC 69, Application des m&hodes statistiques, sous-comité
SC 5, Échantillonnage en vue d’acceptation.
Cette première édition de I’ISO 2859-O annule et remplace
I’ISO 2859:1974 et I’Additif 1 :1977.
L’ISO 2859 comprend les parties suivantes, présentées sous le titre gé-
néra I Règles d ‘6chan tillonnage pour les contrôles par attributs:
- Partie 0: Introduction au système d’échantillonnage par attributs de
I’ISO 2859
- Partie 1: Plans d’échantillonnage pour les contrôles lot par lot, in-
dexés d’après le niveau de qualité acceptable (NQA)
- Partie 2: Plans dëchan tillonnage pour les contrôles de lots isolés,
indexés d’après la qualité limite (QL)
- Partie 3: Procédures d’échantillonnage successif partiel
0 ISO
ISO 2859-O: 1995(F)
Introduction
La présente Introduction générale aux contrôles par échantillonnage a pour
objet de décrire les programmes d’échantillonnage par attributs traités
dans les parties 1 à 3 de I’ISO 2859 et dans I’ISO 8422. Elle traite des
contrôles par échantillonnage par attributs de façon générale, présente les
principaux modes opératoires et le mode d’emploi prévu des programmes
d’échantillonnage. La compréhension des différents concepts et de leur
application sera facilitée par la lecture de I’ISO 2859-1, de I’ISO 28532,
de I’ISO 28533, de I’ISO 8422 et de l’lSO/TR 8550.
Ces différentes normes ou parties de norme constituent une extension
des principes présentés dans la présente introduction à des usages plus
spécifiques des procédures dont elles traitent respectivement.
L’ISO 28531 décrit des programmes d’échantillonnage indexés d’après le
NQA. La mesure de la qualité utilisée peut être soit le pourcentage d’in-
dividus non conformes soit le nombre de non-conformités par 100 indivi-
dus. A l’origine, I’ISO 28591 était principalement destinée au contrôle
d’une série continue de lots provenant de la même source, puisque, en
effet, l’application des règles de modification du contrôle (c’est-à-dire de
passage du contrôle normal au contrôle renforcé) permet d’assurer une
protection adéquate (en termes de pourcentage moyen maximal d’indivi-
dus non conformes dans un processus) lorsqu’une petite série de lots
successifs apparaît contenir un certain nombre (limite) de lots non accep-
tables.
L’ISO 28532 décrit des plans d’échantillonnage applicables au cas où I’é-
chantillonnage porte sur des lots individuels ou isolés. Ces plans d’échan-
tillonnage sont à beaucoup d’égards identiques à ceux que décrit
I’ISO 2859-l. Toutes les tables associées aux plans d’échantillonnage de
I’ISO 2859-2 contiennent des informations relatives au niveau de qualité
requis pour assurer une probabilité élevée d’acceptation du lot. II est re-
commandé, pour les lots individuels ou isolés, d’utiliser I’ISO 28592 de
préférence à I’ISO 2859-l.
L’ISO 28593 décrit des procédures d’échantillonnage successif partiel
applicables lorsque le niveau de qualité du processus est sensiblement
supérieur au NQA sur une longue durée, définie, de livraison ou d’obser-
vation. Lorsque le niveau de qualité est aussi élevé, il est parfois plus
économique d’appliquer I’ISO 28533 que la procédure de contrôle réduit
de I’ISO 28591. De la même façon que I’ISO 28531, I’ISO 2859-3 est
applicable à une série continue de lots provenant d’une même source.
L’ISO 8422 décrit une méthode d’établissement de plans d’échantillon-
nage progressif ayant un pouvoir de discrimination sensiblement équiva-
lent à celui des différents plans décrits dans I’ISO 2859-1 et I’ISO 28592.
L’ISO 3951: 1989, Règles et tables d’échantillonnage pour les contrôles par
mesures des pourcentages de non conformes et I’ISO 8423:1991, Plans
d’échantillonnage progressif pour le contrôle par mesures des pourcenta-
ges de non conformes (écart-type connu), proposent un système com-
vi
0 ISO
ISO 2859-O: 1995(F)
plémentaire de plans d’échantillonnage pour les contrôles par mesures,
également indexés d’après le NQA.
NOTE 1 L’emploi du genre masculin dans la présente partie de IX0 2859
n’exclut pas le genre féminin lorsqu’il s’applique aux personnes. De même, I’utili-
sation du singulier n’exclut pas le pluriel (et vice versa) lorsque le sens le permet.
VII
Page blanche
ISO 2859-O: 1995(F)
NORME INTERNATIONALE Q ISo
Règles d’échantillonnage pour les contrôles par
attributs -
Partie 0:
Introduction au système d’échantillonnage par attributs de
I’ISO 2859
Section 1: Généralités
1.2 Références normatives
1.1 Domaine d’application
Les normes suivantes contiennent des dispositions
La présente partie de I’ISO 2859 explique les termes
qui, par suite de la référence qui en est faite, consti-
utilisés en échantillonnage pour acceptation, décrit les
tuent des dispositions valables pour la présente partie
différents programmes et plans d’échantillonnage, de I’ISO 2859. Au moment de la publication, les édi-
donne des conseils pratiques sur les contrôles par
tions indiquées étaient en vigueur. Toute norme est
échantillonnage et en discute certains aspects théori-
sujette à révision et les parties prenantes des accords
ques.
fondés sur la présente partie de I’ISO 2859 sont invi-
tées à rechercher la possibilité d’appliquer les éditions
La section 2 donne des informations générales sur les
les plus récentes des normes indiquées ci-après. Les
méthodes de contrôle par échantillonnage pour ac-
membres de la CEI et de I’ISO possèdent le registre
ceptation, avec référence particulière aux règles et
des Normes internationales en vigueur à un moment
tables de contrôle par attributs donnés dans les par-
donné.
ties 1, 2 et 3 de I’ISO 2859 et dans I’ISO 8422.
ISO 2859-l : 1989, Règles d’échantillonnage pour les
La section 3 constitue une extension de l’introduction contrôles attributs - Partie 7: Plans
Par
à l’échantillonnage pour acceptation donnée dans la d’échantillonnage pour les contrôles lot par lot, in-
section 2 et complète cette introduction et les ins- dexés d’après le niveau de qualité acceptable (NQA).
tructions contenues dans I’ISO 2859-l par des com-
mentaires détaillés et des exemples destinés à ISO 2859-2: 1985, Règles d’échantillonnage pour les
faciliter la mise en œuvre de la méthode de contrôle contrôles attributs - Partie 2. Plans
Par
d’échantillonnage pour les contrôles de lots isolés,
par échantillonnage qui constitue le fondement du
indexés d’après la qualité limite (QL).
système d’échantillonnage de I’ISO 2859-l.
fa 60
ISO 2859-O: 1995(F)
d’échantillonnage pour acceptation pour le contrôle
60 28533: 1991, Règles d’échantillonnage pour les
d’unités discrètes en lots.
contrôles par attributs - Partie 3: Procédures
d’échantillonnage successif partiel.
1.3 Définitions
ISO 8422: 1991, Plans d’échantillonnage progressif
pour le contrôle par attributs.
Pour les besoins de la présente partie de I’ISO 2859,
les définitions données dans I’ISO 28531 et I’ISO
ISO/TR 8550:1994, Guide pour la sélection d’un sys-
2859-3 s’appliquent.
tème, d’un programme ou d’un plan
0 ISO
ISO 2859-O: 1995( F)
Section 2: Introduction générale à l’échantillonnage pour acceptation
ceptation d’un produit non satisfaisant (appelé ((risque
2.1 Objectifs des contrôles par
client))) et le risque de non-acceptation d’un produit
échantillonnage
satisfaisant (appelé ((risque fournisseur))), et de choi-
sir un plan d’échantillonnage associé à un risque infé-
L’un des principaux objectifs des contrôles par
rieur ou égal au risque acceptable.
échantillonnage pour acceptation est de vérifier que
le fournisseur livre des lots d’une qualité égale ou
II existe, à côté des plans d’échantillonnage de I’ISO
supérieure à un niveau de qualité convenu mu-
fondes sur la théorie mathématique des probabilités,
tuellement, de sorte que le client reçoive des lots de
plusieurs autres méthodes:
qualité acceptable.
a) échantillonnage fondé sur l’expérience acquise
Le fournisseur peut appliquer les procédures d’é-
sur le produit, le processus, le fournisseur et le
chantillonnage décrites pour s’assurer que le niveau
client (voir 2.2.1);
de qualité sera acceptable pour le client. Dans I’en-
semble de ces procédures, il faut tenir compte du fait
échantillonnage ad hoc, par exemple contrôle d’un
b)
que les ressources financières disponibles ne sont
pourcentage donné, ou vérifications occasion-
pas illimitées. Le coût d’un article doit refléter aussi
nelles au hasard (voir 2.2.2);
bien les coûts de contrôle que les coûts de produc-
tion.
c) contrôle à 100 % (voir 2.2.3);
II convient de faire un réel effort pour concevoir des
autres méthodes «d’échantillonnage» (voir 2.2.4).
dl
systèmes dans lesquels la responsabilite de la qualité
revienne clairement au fournisseur. Les contrôles
2.2.1 Échantillonnage statistique
peuvent en effet sembler déplacer la responsabilité
de la qualité du fournisseur vers le contrôleur, no-
L’échantillonnage fondé sur l’expérience acquise sur
tamment lorsqu’il existe le sentiment que le contrô-
le produit, le processus, le fournisseur et le client peut
leur est là pour resoudre les problèmes de sorte que,
être évalue statistiquement.
dans certaines limites, tout événement affectant la
production sera mis en évidence lors du contrôle.
La procédure décrite dans I’ISO 2859-1, qui utilise un
Cette conception est complètement erronée et peut
ensemble de règles de modification du contrôle, en
conduire à un travail difficile, des coûts élevés et une
est un exemple. Si la qualité est d’un très bon niveau,
qualité médiocre pour le client et le fournisseur. Le
il est possible de passer au contrôle réduit. Cela si-
contrôleur n’a pas les moyens de conférer un bon ni-
gnifie que, si le nombre d’échantillons utilisés est plus
veau de qualité à un produit si le fournisseur ne l’a pas
faible, le risque du fournisseur diminue mais celui du
déjà fait.
client augmente. Si l’expérience est satisfaisante, le
recours peut être justifié, notamment lorsque la qua-
lité moyenne du processus s’est avérée être sensi-
2.2 Échantillonnage pour acceptation
blement plus faible que le niveau de qualité
Le contrôles par échantillonnage pour acceptation ont acceptable (NQA) spécifié. Lorsque la moyenne du
le mérite de faire clairement porter la responsabilité processus sur 10 lots au moins s’est avérée être
sensiblement inférieure au NQA, certains clients pré-
de la qualité sur celui à qui elle revient de fait, le
fournisseur. Le contrôleur n’est plus considéré fèrent recourir à l’échantillonnage successif partiel
comme celui qui résoud les problèmes. Le fournis- (voir ISO 28593). Cette solution peut être encore plus
seur doit veiller à ce que la qualité du produit soit économique que la procédure de contrôle réduit dé-
correcte, sous peine d’avoir à supporter les ennuis et crite dans I’ISO 2859-l.
les dépenses liés à l’existence de lots non accep-
Dans certains cas, notamment lorsque le contrôle
tables. Le contrôle par échantillonnage peut et devrait
porte sur des unités fournies en grande quantité ou
normalement conduire à un allégement des tâches de
non critiques, certains clients peuvent estimer suffi-
contrôle, une réduction des coûts et une amélioration
samment sûre la pratique consistant à contrôler de
de la qualité du produit fourni au client.
petits échantillons du produit et à accepter le lot si
Les programmes de contrôle par échantillonnage dé- aucun individu non conforme n’est observé. Pour un
crits dans les parties 1, 2 et 3 de I’ISO 2859, et dans effectif d’échantillon de 8, par exemple, ceci équivaut
I’ISO 8422 permettent de quantifier le risque d’ac- à appliquer les plans d’échantillonnage simple sur pe-
Q ISO
ISO 2859=0:1995(F)
tits lots correspondant à un NQA de 1,5 % en contrôle I’ISO 2859. Certains défauts ont une telle importance
normal ou de 0,65 % en contrôle réduit (voir tables que, dans le cas d’essais non destructifs, chaque in-
II-A et II-C de I’ISO 2859-1:1989). dividu doit être vérifié. Dans le cas d’essais destruc-
tifs, il faut nécessairement prendre un certain risque
Si, inversement, selon I’ISO 2859-1, deux lots parmi
(voir 2.15) .
cinq lots successifs sont jugés non acceptables, le
contrôle normal est interrompu et le contrôle renforcé
2.2.4 Autres méthodes «d’échantillonnage»
institué. Une fois le contrôle renforce institue, le re-
tour au contrôle normal ne se fera que lorsque cinq
II existe divers systèmes d’échantillonnage, mais
lots successifs auront éte acceptes sur la base du
seuls ceux qui font l’objet de normes internationales
contrôle renforcé. Cette prescription est volontai-
seront considéres en détail dans la présente partie de
rement sévère, car l’existence d’un niveau de qualité
I’ISO 2859. Ceci ne signifie pas que les autres soient
non acceptable a été démontrée. Le fournisseur perd
d’une importance mineure, mais simplement que
donc son droit au bénéfice du doute. Si, au cours du
l’objectif principal de la présente partie de I’ISO 2859
contrôle renforcé, le nombre cumulé de lots non ac-
est de faciliter l’utilisation des parties 1, 2 et 3 de
ceptés sur la base du contrôle renforcé en première
I’ISO 2859, et de I’ISO 8422.
présentation atteint cinq, il convient de suspendre le
contrôle par échantillonnage jusqu’à ce qu’il soit
II arrive très fréquemment que les clients n’effectuent
prouvé que des mesures correctives ont été prises,
pas des échantillonnages réguliers mais se reposent
avec un résultat effectif (voir 2.11).
sur leur expérience et sur les documents prouvant
que le fournisseur assure la maîtrise statistique du
processus de production et évalue honnêtement la
2.2.2 Échantillonnage ad hoc
qualité des expéditions.
Le recours a l’échantillonnage ad hoc est déconseillé
Si, dans un cas particulier, on dispose d’informations
car il conduit a des risques non calculés qui peuvent
sur le coût réel du rejet a tort d’articles conformes et
être inacceptablement élevés; il n’existe par ailleurs
de l’acceptation à tort d’articles non conformes, et si
aucune base formelle d’acceptation ou de rejet du lot.
une longue expérience permet de connaître la fré-
quence avec laquelle des lots de qualité donnée,
2.2.3 Contrôle à 100 %
quelle qu’elle soit, peuvent être presentés, ceci peut
être une occasion d’éviter de recourir au compromis.
La réalisation d’un contrôle à 100 % peut constituer
II peut être possible de calculer un programme plus
une très lourde tâche si le contrôle n’est pas effectué
efficace sur la base des données économiques dont
avec un appareillage d’essai automatique. Les résul-
on dispose.
tats obtenus ne sont en outre pas toujours corrects,
notamment lorsqu’un grand nombre d’individus pos-
2.3 Contrôle par attributs ou par
sèdent du point de vue dimensionnel, de l’aspect ou
des performances une ou plusieurs caractéristiques
marginales (c’est-à-dire voisines d’une tolerance ou li-
mite relative, ou groupées autour de cette limite). La méthode de contrôle par attributs consiste a exa-
Dans ces conditions, le tri manuel ou automatique
miner un individu, ou des caractéristiques d’un indi-
risque de conduire à classer comme non conformes vidu, afin de le classer ((conforme)) ou ((non
certains individus conformes, et inversement. Le conforme)). On décide de l’attitude à adopter en
contrôle à 100 %, qu’il soit manuel, visuel ou auto- fonction du nombre d’individus non conformes ou de
matique, peut par ailleurs être conduit de façon non non-conformités décomptés dans un échantillon au
satisfaisante. II peut parfois dégénérer en contrôle à hasard.
100 % superficiel, en l’absence de moyens financiers,
La méthode de contrôle par mesures consiste à pré-
en temps ou en personnel suffisants. II est enfin
lever un échantillon constitué d’un certain nombre
inapplicable si la méthode de contrôle necessite le
d’individus, et à mesurer certaines dimensions ou
recours à des essais destructifs.
caractéristiques. Les données obtenues ne portent
II est toutefois reconnu que les contrôles à 100 % donc pas uniquement sur le fait qu’une dimension, par
peuvent constituer une phase nécessaire du proces- exemple, est comprise ou non entre certaines limites,
sus de contrôle, à la fois pour le client et pour le mais sur la valeur effective de cette dimension. La
fournisseur. II existe des situations dans lesquelles ils décision d’acceptation ou de non-acceptation d’un lot
sont indispensables, par exemple les contrôles por- est prise sur la base de calculs de la moyenne et de
tant sur des défauts critiques, comme cela apparaît la variabilité des mesures selon les procédures dé-
clairement à la lecture de la présente partie de crites dans I’ISO 3951 ou I’ISO 8423.
Q ISO
ISO 2859-O: 1995(F)
La méthode par mesures présente l’avantage, à Un échantillon est prélevé dans chaque lot et contrôlé,
condition que certaines hypothèses soient vérifiées, chaque lot est ensuite classé comme acceptable ou
.
de requérir un effectif d’échantillon plus petit que la non acceptable sur la base du nombre d’individus non
méthode par attributs pour assurer un niveau de pro- conformes ou de non-conformités trouvés dans
tection donné contre une décision incorrecte. Elle l’échantillon. Chaque lot successif est aonc traité
comme une unite plus ou moins indépendante (bien
fournit également davantage d’informations sur
que les regles de décision puissent parfois varier se-
l’existence d’une éventuelle influence négative sur la
qualité de la moyenne et/ou de la variabilité du pro- lon les résultats obtenus pour les lots précédents).
cessus. La méthode par attributs présente l’avantage
Pour les lots livrés isolement, se référer a 2.5.2.
d’être plus ((robuste)) (non soumise a des hypothèses
de forme de distribution) et plus simple à utiliser. Ces
avantages peuvent justifier le recours à la méthode
par attributs bien qu’elle implique des effectifs d’é-
chantillons, et donc des coûts afférents, plus élevés.
2.4.2 Effectif des lots
II est à noter qu’une évaluation de type tout ou rien
L’autorité responsable (voir ISO 2859-l : 1989, 3.10)
(bon ou mauvais) est plus rapide et demande moins
peut, de droit, spécifier elle-même l’effectif qui doit
de savoir-faire que l’obtention d’une mesure.
être celui du lot, mais il est évidemment préférable,
Les deux méthodes ont leurs propres avantages et
lorsque cela est possible, que cette valeur soit choisie
leurs propres applications typiques.
après consultation du fournisseur afin qu’elle soit
mutuellement satisfaisante. II va de soi que le choix
Bien qu’il soit fait référence à l’occasion à I’ISO 3951
de l’effectif du lot (de même que des autres paramè-
et à I’ISO 8423 dans la suite du texte, les program-
tres du plan d’échantillonnage) ne devrait jamais être
mes d’échantillonnage par mesures ne sont pas dé-
effectué sans tenir compte du processus de produc-
veloppés en tant que tels dans le cadre de la présente
tion. II n’est pas essentiel que la valeur choisie soit
partie de I’ISO 2859. L’ISO 3951 et I’ISO 8423
étroitement fixee. Certaines variations sont admissi-
contiennent des conseils d’utilisation les concernant.
bles, mais il est presque toujours souhaitable que
soient spécifiées les limites inferieure et supérieure
de l’effectif du lot.
Du point de vue du contrôle par échantillonnage, il est
2.4 Contrôle de lots
avantageux de prendre des lots d’effectif important,
puisqu’il sera alors possible de prélever dans ces lots
des échantillons d’effectif important, donc de mieux
2.4.1 Lot
discriminer les bons et les mauvais lots. Si le lot est
grand, l’échantillon à prélever représente, à NQA égal,
Dans le cadre de I’ISO 2859, de I’ISO 8422, de
une plus petite proportion du lot que si celui-ci est
I’ISO 3951 et de I’ISO 8423, les individus ne sont pas
petit.
soumis au processus de contrôle de façon indivi-
duelle, mais en groupes. Chacun de ces groupes
II faut toutefois rester modéré dans l’emploi de cette
d’individus est appelé un lot.
stratégie des ((grands lots). Si la constitution de
Chaque lot doit, dans la mesure du possible, se com- grands lots nécessite de combiner des lots plus petits
poser d’individus fabriques sensiblement dans les qui auraient très bien pu rester séparés, l’emploi d’un
mêmes conditions et au cours d’une même période. grand lot n’est intéressant que si les petits lots sont
Ceci est de la plus grande importance si le concept tous d’une qualité équivalente. Si l’existence de dif-
férences de qualité importantes entre les petits lots
de niveau de qualité acceptable est adopte et qu’une
série de lots doit être livrée. paraît probable, il est de beaucoup préférable de les
garder séparés. C’est pourquoi il convient d’opérer sur
Si la livraison est constituée d’individus provenant de
des lots constitués d’unités de produit fabriquées
plusieurs sources différentes, la présence d’un grand
dans des conditions sensiblement identiques.
nombre d’individus non conformes provenant de l’une
de ces sources pourra conduire à la non-acceptation
Des exemples de formation des lots sont donnés en
du mélange provenant de l’ensemble des sources.
34 . .
Inversement, la présence d’un produit de qualité
marginale provenant de l’une des sources pourra être Des informations plus détaillées sur le rapport effectif
masquée par le mélange avec le produit de sources du lotleffectif de l’échantillon sont données dans
d’excellente qualité. l’lSO/TR 8550.
‘0 ISO
ISO 2859-O: 1995(F)
Si la livraison ne comporte qu’un seul lot, il peut être
2.5 Contrôle de séquences de lots ou de
utile de savoir si ce lot fait partie d’un ensemble de
lots isolés
lots similaires livrés a d’autres clients et provient d’un
processus maîtrisé, ou s’il s’agit d’un lot mélangé
formé d’individus issus de différents processus et
2.5.1 Contrôle lot par lot
produits à des moments différents (voir également
I’ISO/TR 8550 et I’ISO 28532).
Un contrôle lot par lot est le contrôle d’un produit
fourni sous forme d’une série de lots.
Tandis que I’ISO 28591 et I’ISO 28593 exigent que
la valeur du NQA et le niveau de contrôle soient fixés
Lorsqu’une séquence de lots est soumise a I’accep-
a l’avance, I’ISO 2859-2 demande que soit etablie la
tation au moment de la production, il peut arriver que
qualité limite (QL). Pour assurer un niveau de protec-
les résultats de contrôle concernant les premiers lots
tion approprié du fournisseur et du client lorsque les
soient disponibles avant que les lots suivants aient eté
lots sont échantillonnés selon la procedure de la qua-
fabriqués. II est donc possible d’utiliser les resultats
lité limite, il est nécessaire de savoir si le lot provient
déjà obtenus pour améliorer la qualité du reste de la
d’une série continue de lots acceptables ou s’il s’agit
production. II convient de présenter et de contrôler les
d’un lot mixte constitué d’unités de produit fabriquées
lots suivant leur séquence de fabrication, et de réali-
sur différentes lignes de production et/ou à des dates
ser les contrôles sans délai. Les informations obte-
différentes.
d’un lot peuvent indiquer une
nues à partir
dégradation apparente du niveau de qualité du pro-
Les tables de I’ISO 2859-2 sont principalement desti-
cessus. Les informations obtenues à partir de plu-
nées au contrôle des lots isolés.
sieurs lots successifs peuvent être utilisées pour
mettre en œuvre une procédure de modification du
contrôle exigeant le passage à un plan d’échantillon- 2.6 Niveau de qualité acceptable (NQA)
nage plus sevère lorsque le processus se dégrade.
Cette façon de procéder est importante car elle
2.6.1 Description
constitue le meilleur moyen d’assurer a long terme la
protection optimale du client contre une qualité insuf-
Le niveau de qualité acceptable est utilisé comme
fisante. Si le niveau de qualité reste médiocre, I’appli-
critère d’indexation dans les tables de I’ISO 2859-1,
cation d’un contrôle plus sévère conduira à ce qu’un
dans I’ISO 3951 et dans certaines des tables de
plus grand nombre de lots seront retournes au four-
I’ISO 8422 et de I’ISO 8423.
nisseur pour être tries. Cet accroissement de la se-
vérité de l’échantillonnage se traduit pour le
L’utilisation des plans d’échantillonnage indexés
fournisseur par un plus grand risque de voir juger non
d’après le NQA conduit à l’acceptation, dans presque
acceptable un lot acceptable. L’identification d’une tous les cas, des lots issus d’un processus dont le
dégradation probable de la qualité du produit constitue niveau de qualité est égal ou meilleur que le NQA.
un signal qui devrait le conduire à prendre des mesu-
Si l’on considère une série continue de lots, le NQA
res correctives.
est un niveau de qualité défini, en contrôle par
Si la qualité est très sensiblement meilleure que celle échantillonnage, comme la limite d’une qualité
qui a été convenue, le client peut, avec l’accord de
moyenne acceptable d’un processus.
l’autorité responsable, décider d’adopter un plan d’é-
Le NQA est une borne, que l’on choisit, séparant ce
chantillonnage réduit ou d’échantillonnage successif
qui sera considéré comme une qualité moyenne ac-
partiel.
ceptable et une qualité moyenne non acceptable d’un
L’ISO 2859-1, I’ISO 28533, I’ISO 8422, I’ISO 3951 et
processus. Ce paramètre ne décrit pas le plan d’é-
I’ISO 8423 sont principalement destinées au contrôle chantillonnage, mais définit ce que la production de-
de séquences de lots. vrait être, et constitue une grandeur utile a prendre
en considération pour définir un processus admissi-
ble .
2.5.2 Contrôle d’un lot isolé
Le fait qu’on a spécifié un NQA ne doit pas être
Le contrôle peut parfois être effectué sur un lot isolé, compris comme impliquant qu’on désire un pourcen-
un petit nombre de lots isolés, ou des lots mis en tage d’individus non conformes égal à la valeur spé-
stock à un moment où la production était terminée.
cifiée ou est tout à fait acceptable. II est toujours
Dans ces conditions, il n’y a pas lieu de recourir aux préférable de n’avoir aucun individu non conforme,
règles de modification et d’influencer ainsi la qualité
plutôt qu’un pourcentage quelconque, et plus le
offerte. pourcentage pourra être inférieur au NQA, mieux ce
0 ISO
ISO 2859-O: 1995(F)
sera. Cette réduction augmente la probabilité d’ac-
II est très important de bien comprendre que ce pa-
ceptation de chaque lot. ramètre n’est pas, à la différence du NQA, de la
LQMAC (voir 2.12) ou de la QL (voir 2.81, une valeur
qui peut être calculée ou choisie, ou constitue une
propriété d’un plan d’échantillonnage particulier, mais
2.6.2 Établissement d’un NQA
qu’il se rapporte au produit de fabrication effectif, in-
dépendamment du mode de réalisation des contrôles.
Lorsque l’on fixe la valeur du NQA, il faut toujours
garder en mémoire le fait que celui-ci constitue une
En règle générale, l’estimation de la qualité moyenne
indication du niveau de qualité que doit atteindre la
d’un processus ne constitue pas une partie essen-
production. Ce qui est demandé au fournisseur est de
tielle du programme d’échantillonnnage. Toutefois,
produire des lots de qualité moyenne meilleure que le
cette grandeur présente en elle-même une certaine
NQA. Cette qualité doit d’un côté être raison-
importance. Le contrôleur autant que le fournisseur
nablement possible à atteindre et d’un autre côté
sont concernés non seulement par les résultats des
constituer une qualité raisonnable du point de vue du
contrôles lot par lot, mais aussi par l’évolution a long
client. Ceci implique souvent de trouver un compro-
terme de la qualité de la production.
mis entre le niveau de qualité que le client souhaiterait
et celui qui est dans ses moyens car, plus ses exi-
II est donc souhaitable de conserver un enregis-
gences seront élevées, plus elles seront difficiles à
trement de la qualité moyenne globale estimée du
satisfaire, et plus le coût des contrôles de vérification
processus car elle constitue une excellente mesure
sera élevé.
de la qualité et un élément d’information intéressant
pour ceux qui ont à décider quels plans d’échantillon-
Un processus correctement conçu et maîtrisé peut
nage il convient d’adopter lorsque des produits simi-
permettre la fabrication d’un produit caractérisé par
laires seront conçus et fabriqués dans l’avenir.
un pourcentage d’individus non conformes inférieur
au NQA, Lorsqu’il est possible d’améliorer la qualité
II faut observer des règles spéciales pour les échan-
moyenne d’un processus, les coûts de production et
tillonnages de type double ou multiple. Dans ce cas,
de contrôle associés a cette qualité améliorée seront
il convient de n’utiliser que les résultats portant sur le
réduits.
premier échantillon pour l’estimation de la qualité
moyenne du processus.
Le critère primordial doit être l’exigence du client,
mais il est nécessaire de s’assurer que celle-ci est
II arrive parfois qu’une recommandation d’exclusion
réaliste et que le client ne demande pas une qualité
des résultats aberrants soit formulée. Cette pratique
meilleure que celle qui est réellement nécessaire. II
est dangereuse et il convient de n’y recourir qu’avec
est indispensable de tenir compte de l’utilisation pré-
parcimonie, sinon de l’éviter totalement. Le seul cas
vue des individus considérés, et des conséquences
où elle peut être adoptée sans risque est celui où l’on
d’une éventuelle non-conformité. Si les individus sont
a pu établir que les résultats aberrants résultent d’une
fournis en grand nombre et que la non-conformité ne
cause spécifique, et que celle-ci a été eliminée.
concerne que le montage, de sorte qu’un individu
Même dans ce cas, il est bon d’enregistrer à la fois
non-conforme peut être mis de côté et remplacé par
les chiffres incluant et excluant ces résultats aberrants
un autre, un NQA relativement élevé peut être ad-
afin d’indiquer l’existence de ces non-conformités.
missible. Si, par contre, une non-conformité est sus-
ceptible d’entraîner un disfonctionnement d’une partie
Dans le cas de caractéristiques multiples ou de clas-
coûteuse et importante d’un équipement en un lieu
ses multiples de NQA, il faut estimer séparément les
et a un moment où le remplacement de l’individu non
qualités moyennes de processus correspondantes.
conforme est impossible, il faudra choisir un NQA plus
sévère.
Des informations complémentaires et des indications
2.8 Qualité limite (QL)
sur l’établissement du NQA sont données en 3.9 et
dans l’lSO/TR 8550.
La qualité limite est un critère d’indexation, utilisé
dans I’ISO 28592. Lorsqu’un lot est considéré isolé-
ment, la qualité limite est un niveau de qualité, ex-
primé en pourcentage d’individus non conformes (ou
2.7 Qualité moyenne d’un processus
en nombre de non-conformités par 100 individus), qui,
La qualité moyenne du processus est la qualité
en contrôle par échantillonnage, est associé a une
moyenne d’une série de lots présentés, à l’exclusion
probabilité d’acceptation faible. Cette probabilité d’ac-
des lots proposés en deuxième présentation. ceptation faible est appelée ((risque du client)).
0 ISO
ISO 2859-O: 1995(F)
II faut souligner que la qualité limite ainsi spécifiée est normal) et la procédure de contrôle renforce sera ins-
en fait un niveau de qualité non souhaité. Pour que de protéger le client.
tituée afin
des lots soient régulièrement jugés acceptables, il
D’autres informations, avec des exemples, sont don-
faut que la proportion d’individus non conformes soit
nées à ce sujet en 3.11 et 3.12.
très sensiblement inférieure à la QL (généralement a
un quart de la QL).
2.10 Contrôle réduit
L’ISO 28592 décrit des procédures d’application des
plans d’échantillonnage indexés d’après la QL. Ces
II apparaît parfois au cours des contrôles que la qualité
plans et les tables associées sont en grande partie
du produit est notablement meilleure que le NQA.
concordants avec les plans d’échantillonnage de
Dans ce cas, et s’il existe des raisons de penser que
I’ISO 2859-l.
cette bonne qualité de production sera maintenue, il
n’est plus nécessaire de recourir à des plans d’é-
Les plans indexés d’après la qualité limite sont es-
chantillonnage pour séparer les bons et les mauvais
sentiellement destines au contrôle de lots isolés.
lots. Les contrôles ne peuvent toutefois pas être pu-
Lorsque le produit est en cours de fabrication et
rement et simplement supprimés car ils constituent
qu’une série de lots est produite, les procédures dé-
un moyen nécessaire de détection d’une éventuelle
crites dans I’ISO 28531 sont mieux appropriées.
dégradation de la qualité de la production.
Dans cette situation, il est possible, si on le souhaite,
2.9 Contrôle normal et contrôle renforcé
de réaliser d’importantes économies en passant aux
plans d’échantillonnage réduit de I’ISO 2859-l ou aux
Le NQA, rappelons-le, est la borne séparant les ni-
plans d’échantillonnage successif partiel de
veaux de qualité définis comme satisfaisants et non
I’ISO 2859-3. Les règles spéciales permettant de re-
satisfaisants lors du contrôle d’une série de lots. Une
courir à ces plans, avec l’accord de l’autorité respon-
fois spécifiée une valeur du NQA pour un produit
sable, sont décrites dans I’ISO 28591, et dans
donne quelconque, l’idéal serait de disposer d’un
I’ISO 2859-3 et également dans la section 3 de la
système dans lequel les lots de qualité meilleure que
présente partie de I’ISO 2859.
le NQA seraient toujours acceptés et les lots de qua-
lité moins bonne que le NQA toujours refuses. Aucun
Le contrôle réduit est présenté ci-après, avec des
plan d’échantillonnage ne permet toutefois d’atteindre
exemples, en 3.15.
cet idéal.
Pour satisfaire à la fois aux exigences du fournisseu
...
NORME
Iso
INTERNATIONALE
2859-O
Première édition
1995-08-01
Règles d’échantillonnage pour les contrôles
par attributs -
Partie 0:
Introduction au système d’échantillonnage par
attributs de I’ISO 2859
Sampling procedures for inspection by attributes -
Part 0: Introduction to the /SO 2859 attribute sampling system
Numéro de référence
BS0 2859-O: 1995(F)
ISO 2859-O: 1995( F)
Sommaire
Page
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Section 1 Généralités
1.1 Domaine d’application . . . 1
1.2 Références normatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
.................................. .............................. ........... 2
1.3 Définitions .
Introduction générale à l’échantillonnage pour
Section 2
acceptation . . . . . . . . . . .*.*. 3
.......................... 3
2.1 Objectifs des contrôles par échantillonnage
2.2 Échantillonnage pour acceptation .
................................ .................. 3
2.2.1 Échantillonnage statistique
................................. ....................... 4
2.2.2 Échantillonnage ad hoc
2.2.3 Contrôle à 100 % . . . 4
................................ 4
2.2.4 Autres méthodes ((d’échantillonnage))
2.3 Contrôle par attributs ou par mesures . 4
............................. ............ ............................. 5
2.4 Contrôle de lots
............................ ....................... 5
2.4.1 Lot .
2.4.2 Effectif des lots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.5 Contrôle de séquences de lots ou de lots isolés
. . . . . . . . . . . . . . . . . . 6
2.5.1 Contrôle lot par lot . . 6
Contrôle d’un lot isolé . . . 6
2.5.2
2.6 Niveau de qualité acceptable (NQA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
........................... ................................................
2.6.1 Description 6
2.6.2 Établissement d’un NQA . 7
2.7 Qualité moyenne d’un processus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.8 Qualité limite (QL) . . . 7
Contrôle normal et contrôle renforcé
2.9 . . . .“. . . . . . . . . 8
0 ISO 1995
Droits de reproduction réservés. Sauf prescription différente, aucune partie de cette publi-
cation ne peut être reproduite ni utilisée sous quelque forme que ce soit et par aucun pro-
cédé, électronique ou mecanrque, y compris la photocopie et les microfilms sans l’accord
écrit de l’éditeur.
Organisation Internationale de normalisation
Case Postale 56 l CH-121 1 Genève 20 l Suisse
Imprimé en Suisse
ii
cJ ISO
ISO 2859-O: 1995( F)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.10 Contrôle réduit
..................................... 8
2.11 Règles de modification du contrôle
2.12 Qualité moyenne apres contrôle (QMAC) et limite de qualité
moyenne après contrôle (LQMAC) .
.................................................. 10
2.13 Individu (unité de produit)
. . . . . . . . . . . . . . . . . . . . . . . . . 11
2.14 Non-conformité et individu non conforme
2.14.1 Absence de conformité . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.14.2 Nomenclature
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.14.3 Classification des non-conformités
2.15 Non-conformités critiques . . . . . . . . . .~.
................................................................ 15
2.16 Contrôle tronqué
...................................... 16
2.17 Devenir des lots non acceptables
2.18 Échantillonnage simple .
2.19 Courbes d’efficacité . 17
...................................................... 19
2.20 Échantillonnage double
2.21 Échantillonnage multiple . 20
2.22 Échantillonnage progressif . 20
....................................... 20
2.23 Échantillonnage successif partiel
2.24 Comparaison des échantillonnages simple, double, multiple et
progressif . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.24.1 Plans équivalents
2.24.2 Effectif moyen d’échantillon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.25 Prélèvement des échantillons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.26 Plans, programmes et systèmes d’échantillonnage . . . . . . . . . . 28
2.27 Lois de distribution binomiale, de Poisson et
hypergéométrique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
..*...................................m
Section 3 Le systhme de IBIS0 2859-1 30
3.1 Description de I’ISO 2859-l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 Élaboration de spécifications à utiliser en conjonction avec
I’ISO 2859-l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Classification des non-conformités et des individus non
conformes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Lots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
. . .
III
Q3 ISO
ISO 2859-O: 1995(F)
3.5 Signification du niveau de contrôle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.6 Établissement d’un niveau de contrôle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.7 Valeurs recommandées du NQA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*. 40
3.8 Autres valeurs du NQA
3.9 Établissement d’un NQA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.10 Sélection d’un plan d’échantillonnage à partir des tables de NS0
28531 . 43
3.11 Contrôle normal . 44
.................. 49
3.12 Contrôle renforcé .
3.13 . 50
Règles de modification du contrôle - Exemple
3.14
Méthodes de réduction des risques associés à
l’échantillonnage . 51
3.15 Contrôle réduit . 52
3.16 Échantillonnages double et multiple . 55
3.17 Qualite limite et lot isolé . 57
3.18 . 57
Effectifs d’échantillon
3.19
Courbes d’efficacité . 58
3.20
Tables LQMAC (voir théorie en 2.12) . 58
3.21 Nomogrammes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*. 59
Q3 ISO ISO 2859-O: 1995(F)
Avant-propos
L’ISO (Organisation internationale de normalisation) est une fédération
mondiale d’organismes nationaux de normalisation (comités membres de
I’ISO). L’élaboration des Normes internationales est en général confiée aux
comités techniques de I’ISO. Chaque comité membre intéressé par une
étude a le droit de faire partie du comité technique créé à cet effet. Les
organisations internationales, gouvernementales et non gouvernemen-
tales, en liaison avec I’ISO participent également aux travaux. L’ISO colla-
bore étroitement avec la Commission électrotechnique internationale (CEI)
en ce qui concerne la normalisation électrotechnique.
Les projets de Normes internationales adoptés par les comités techniques
sont soumis aux comités membres pour vote. Leur publication comme
Normes internationales requiert l’approbation de 75 % au moins des co-
mités membres votants.
La Norme internationale ISO 2859-O a été élaborée par le comité techni-
que ISO/TC 69, Application des methodes statistiques, sous-comité
SC 5, Échantillonnage en vue d’acceptation.
édition de I’ISO 2859-O annule et remplace
Cette première
I’ISO 2859:1974 et I’Additif 1 :1977.
L’ISO 2859 comprend les parties suivantes, présentées sous le titre gé-
néra I Règles d ‘echan tillonnage pour les contrôles par attributs:
- Partie 0: Introduction au système d’échantillonnage par attributs de
I’ISO 2859
- Partie 1: Plans d’échantillonnage pour les contrôles lot par lot, in-
dexés d’après le niveau de qualité acceptable (NQA)
- Partie 2: Plans dëchan tillonnage pour les contrôles de lots isolés,
indexés d’après la qualité limite (QL)
- Partie 3: Procédures d’échantillonnage successif partiel
V
ISO 2859-O: 1995(F)
Introduction
La présente Introduction générale aux contrôles par échantillonnage a pour
objet de décrire les programmes d’échantillonnage par attributs traités
dans les parties 1 à 3 de I’ISO 2859 et dans I’ISO 8422. Elle traite des
contrôles par échantillonnage par attributs de façon générale, présente les
principaux modes opératoires et le mode d’emploi prévu des programmes
d’échantillonnage. La compréhension des différents concepts et de leur
application sera facilitée par la lecture de I’ISO 2859-1, de I’ISO 28532,
de I’ISO 28533, de I’ISO 8422 et de I’ISO/TR 8550.
Ces différentes normes ou parties de norme constituent une extension
des principes présentés dans la présente introduction à des usages plus
spécifiques des procédures dont elles traitent respectivement.
LIS0 28531 décrit des programmes d’échantillonnage indexés d’après le
NQA. La mesure de la qualité utilisee peut être soit le pourcentage d’in-
dividus non conformes soit le nombre de non-conformités par 100 indivi-
dus. A l’origine, I’ISO 2859-l etait principalement destinée au contrôle
d’une série continue de lots provenant de la même source, puisque, en
effet, l’application des règles de modification du contrôle (c’est-a-dire de
passage du contrôle normal au contrôle renforce) permet d’assurer une
protection adéquate (en termes de pourcentage moyen maximal d’indivi-
dus non conformes dans un processus) lorsqu’une petite série de lots
successifs apparaît contenir un certain nombre (limite) de lots non accep-
tables.
L’ISO 28532 décrit des plans d’échantillonnage applicables au cas où I’é-
chantillonnage porte sur des lots individuels ou isolés. Ces plans d’échan-
tillonnage sont à beaucoup d’égards identiques à ceux que décrit
I’ISO 2859-l. Toutes les tables associees aux plans d’échantillonnage de
I’ISO 2859-2 contiennent des informations relatives au niveau de qualité
requis pour assurer une probabilité élevée d’acceptation du lot. II est re-
commandé, pour les lots individuels ou isolés, d’utiliser I’ISO 2859-2 de
préférence à I’ISO 2859-l.
L’ISO 28593 décrit des procédures d’échantillonnage successif partiel
applicables lorsque le niveau de qualité du processus est sensiblement
supérieur au NQA sur une longue durée, définie, de livraison ou d’obser-
vation. Lorsque le niveau de qualité est aussi élevé, il est parfois plus
économique d’appliquer I’ISO 2859-3 que la procédure de contrôle réduit
de I’ISO 2859-l. De la même façon que I’ISO 2859-1, I’ISO 2859-3 est
applicable à une série continue de lots provenant d’une même source.
L’ISO 8422 décrit une méthode d’établissement de plans d’échantillon-
nage progressif ayant un pouvoir de discrimination sensiblement équiva-
lent à celui des différents plans décrits dans I’ISO 2859-l et I’ISO 28592.
L’ ISO 3951: 1989, Règles et tables d’échantillonnage pour les contrôles par
mesures des pourcentages de non conformes et I’ISO 8423:1991, Plans
d’échantillonnage progressif pour le contrôle par mesures des pourcenta-
ges de non conformes (écart-type connu), proposent un système com-
vi
0 ISO
ISO 2859-O: 1995(F)
plémentaire de plans d’échantillonnage pour les contrôles par mesures,
également indexés d’après le NQA.
NOTE 1 L’emploi du genre masculin dans la présente partie de IX0 2859
n’exclut pas le genre féminin lorsqu’il s’applique aux personnes. De même, I’utili-
sation du singulier n’exclut pas le pluriel (et vice versa) lorsque le sens le permet.
vii
Page blanche
ISO 2859-O: 1995(F)
NORME INTERNATIONALE 0 ISO
Règles d’échantillonnage pour les contrôles par
attributs -
Partie 0:
Introduction au système d’échantillonnage par attributs de
I’ISO 2859
Section 1: Généralités
1.2 Références normatives
1.1 Domaine d’application
Les normes suivantes contiennent des dispositions
La présente partie de I’ISO 2859 explique les termes
qui, par suite de la référence qui en est faite, consti-
utilisés en échantillonnage pour acceptation, décrit les
tuent des dispositions valables pour la présente partie
différents programmes et plans d’échantillonnage,
de I’ISO 2859. Au moment de la publication, les édi-
donne des conseils pratiques sur les contrôles par
tions indiquées étaient en vigueur. Toute norme est
échantillonnage et en discute certains aspects théori-
sujette à révision et les parties prenantes des accords
ques.
fondés sur la présente partie de I’ISO 2859 sont invi-
tées à rechercher la possibilité d’appliquer les éditions
La section 2 donne des informations générales sur les
les plus récentes des normes indiquées ci-aprés. Les
méthodes de contrôle par échantillonnage pour ac-
membres de la CEI et de I’ISO possèdent le registre
ceptation, avec référence particulière aux règles et
des Normes internationales en vigueur à un moment
tables de contrôle par attributs donnés dans les par-
donné.
ties 1, 2 et 3 de I’ISO 2859 et dans I’ISO 8422.
ISO 2859-l : 1989, Règles d’échantillonnage pour les
La section 3 constitue une extension de l’introduction
contrôles attributs - Partie 7: Plans
Par
à l’échantillonnage pour acceptation donnée dans la d’échantillonnage pour les contrôles lot par lot, in-
section 2 et complète cette introduction et les ins- dexés d’après le niveau de qualité acceptable (NQA).
tructions contenues dans I’ISO 2859-l par des com-
mentaires détaillés et des exemples destinés à ISO 2859-2: 1985, Règles d’échantillonnage pour les
faciliter la mise en œuvre de la méthode de contrôle contrôles attributs - Partie 2. Plans
Par
d’échantillonnage pour les contrôles de lots isolés,
par échantillonnage qui constitue le fondement du
indexés d’après la qualité limite (QL).
système d’échantillonnage de I’ISO 2859-l.
fa 60
ISO 2859-O: 1995(F)
60 28533: 1991, Règles d’échantillonnage pour les d ‘échan Monnage pour accepta tien pour le contrôle
contrôles par attributs - Partie 3: Procédures d’unités discrètes en lots.
d’échantillonnage successif partiel.
ISO 8422: 1991, Plans d’échantillonnage progressif 1.3 Définitions
pour le contrôle par attributs.
Pour les besoins de la présente partie de I’ISO 2859,
ISO/TR 8550:1994, Guide pour la sélection d’un sys- les définitions données dans I’ISO 2859-l et I’ISO
tème, d’un programme ou d’un 2859-3 s’appliquent.
plan
0 ISO
ISO 2859-O: 1995( F)
Section 2: Introduction générale à l’échantillonnage pour acceptation
ceptation d’un produit non satisfaisant (appelé ((risque
2.1 Objectifs des contrôles par
client))) et le risque de non-acceptation d’un produit
échantillonnage
satisfaisant (appelé ((risque fournisseur))), et de choi-
sir un plan d’échantillonnage associé à un risque infé-
L’un des principaux objectifs des contrôles par
rieur ou égal au risque acceptable.
échantillonnage pour acceptation est de vérifier que
le fournisseur livre des lots d’une qualité égale ou
II existe, à côté des plans d’échantillonnage de I’ISO
supérieure à un niveau de qualité convenu mu-
fondes sur la théorie mathématique des probabilités,
tuellement, de sorte que le client reçoive des lots de
plusieurs autres méthodes:
qualité acceptable.
échantillonnage fondé sur l’expérience acquise
a)
Le fournisseur peut appliquer les procédures d’é-
sur le produit, le processus, le fournisseur et le
chantillonnage décrites pour s’assurer que le niveau
client (voir 2.2.1);
de qualité sera acceptable pour le client. Dans I’en-
semble de ces procédures, il faut tenir compte du fait
échantillonnage ad hoc, par exemple contrôle d’un
b)
que les ressources financières disponibles ne sont
pourcentage donné, ou vérifications occasion-
pas illimitées. Le coût d’un article doit refléter aussi
nelles au hasard (voir 2.2.2);
bien les coûts de contrôle que les coûts de produc-
tion.
contrôle à 100 % (voir 2.2.3);
d
II convient de faire un réel effort pour concevoir des
autres méthodes «d’échantillonnage» (voir 2.2.4).
dl
systèmes dans lesquels la responsabilite de la qualité
revienne clairement au fournisseur. Les contrôles
2.2.1 Échantillonnage statistique
peuvent en effet sembler déplacer la responsabilité
de la qualité du fournisseur vers le contrôleur, no-
L’échantillonnage fondé sur l’expérience acquise sur
tamment lorsqu’il existe le sentiment que le contrô-
le produit, le processus, le fournisseur et le client peut
leur est là pour resoudre les problèmes de sorte que,
être évalue statistiquement.
dans certaines limites, tout événement affectant la
production sera mis en évidence lors du contrôle.
La procédure décrite dans I’ISO 2859-1, qui utilise un
Cette conception est complètement erronée et peut
ensemble de règles de modification du contrôle, en
conduire à un travail difficile, des coûts élevés et une
est un exemple. Si la qualité est d’un très bon niveau,
qualité médiocre pour le client et le fournisseur. Le
il est possible de passer au contrôle réduit. Cela si-
contrôleur n’a pas les moyens de conférer un bon ni-
gnifie que, si le nombre d’échantillons utilisés est plus
veau de qualité à un produit si le fournisseur ne l’a pas
faible, le risque du fournisseur diminue mais celui du
déjà fait.
client augmente. Si l’expérience est satisfaisante, le
recours peut être justifié, notamment lorsque la qua-
lité moyenne du processus s’est avérée être sensi-
2.2 Échantillonnage pour acceptation
blement plus faible que le niveau de qualité
Le contrôles par échantillonnage pour acceptation ont acceptable (NQA) spécifié. Lorsque la moyenne du
le mérite de faire clairement porter la responsabilité processus sur 10 lots au moins s’est avérée être
sensiblement inférieure au NQA, certains clients pré-
de la qualité sur celui à qui elle revient de fait, le
fournisseur. Le contrôleur n’est plus considéré fèrent recourir à l’échantillonnage successif partiel
comme celui qui résoud les problèmes. Le fournis- (voir ISO 28593). Cette solution peut être encore plus
seur doit veiller à ce que la qualité du produit soit économique que la procédure de contrôle réduit dé-
correcte, sous peine d’avoir à supporter les ennuis et crite dans I’ISO 2859-l.
les dépenses liés à l’existence de lots non accep-
Dans certains cas, notamment lorsque le contrôle
tables. Le contrôle par échantillonnage peut et devrait
porte sur des unités fournies en grande quantité ou
normalement conduire à un allégement des tâches de
non critiques, certains clients peuvent estimer suffi-
contrôle, une réduction des coûts et une amélioration
samment sûre la pratique consistant à contrôler de
de la qualité du produit fourni au client.
petits échantillons du produit et à accepter le lot si
Les programmes de contrôle par échantillonnage dé- aucun individu non conforme n’est observé. Pour un
crits dans les parties 1, 2 et 3 de I’ISO 2859, et dans effectif d’échantillon de 8, par exemple, ceci équivaut
I’ISO 8422 permettent de quantifier le risque d’ac- à appliquer les plans d’échantillonnage simple sur pe-
Q ISO
tits lots correspondant à un NQA de 1,5 % en contrôle I’ISO 2859. Certains défauts ont une telle importance
normal ou de 0,65 % en contrôle réduit (voir tables que, dans le cas d’essais non destructifs, chaque in-
II-A et II-C de I’ISO 2859-1:1989).
dividu doit être vérifié. Dans le cas d’essais destruc-
tifs, il faut nécessairement prendre un certain risque
Si, inversement, selon I’ISO 2859-1, deux lots parmi
(voir 2.15) .
cinq lots successifs sont jugés non acceptables, le
contrôle normal est interrompu et le contrôle renforcé
2.2.4 Autres méthodes «d’échantillonnage»
institué. Une fois le contrôle renforce institue, le re-
tour au contrôle normal ne se fera que lorsque cinq
II existe divers systèmes d’échantillonnage, mais
lots successifs auront éte acceptes sur la base du
seuls ceux qui font l’objet de normes internationales
contrôle renforcé. Cette prescription est volontai-
seront considérés en détail dans la présente partie de
rement sévère, car l’existence d’un niveau de qualité
I’ISO 2859. Ceci ne signifie pas que les autres soient
non acceptable a été démontrée. Le fournisseur perd
d’une importance mineure, mais simplement que
donc son droit au bénéfice du doute. Si, au cours du
l’objectif principal de la présente partie de I’ISO 2859
contrôle renforcé, le nombre cumulé de lots non ac-
est de faciliter l’utilisation des parties 1, 2 et 3 de
ceptés sur la base du contrôle renforcé en première
I’ISO 2859, et de I’ISO 8422.
présentation atteint cinq, il convient de suspendre le
contrôle par échantillonnage jusqu’à ce qu’il soit
II arrive très fréquemment que les clients n’effectuent
prouvé que des mesures correctives ont été prises,
pas des échantillonnages réguliers mais se reposent
avec un résultat effectif (voir 2.11).
sur leur expérience et sur les documents prouvant
que le fournisseur assure la maîtrise statistique du
processus de production et évalue honnêtement la
2.2.2 Échantillonnage ad hoc
qualité des expéditions.
Le recours a l’échantillonnage ad hoc est déconseillé
Si, dans un cas particulier, on dispose d’informations
car il conduit a des risques non calculés qui peuvent
sur le coût réel du rejet à tort d’articles conformes et
être inacceptablement élevés; il n’existe par ailleurs
de l’acceptation à tort d’articles non conformes, et si
aucune base formelle d’acceptation ou de rejet du lot.
une longue expérience permet de connaître la fré-
quence avec laquelle des lots de qualité donnée,
2.2.3 Contrôle à 100 %
quelle qu’elle soit, peuvent être presentés, ceci peut
être une occasion d’éviter de recourir au compromis.
La réalisation d’un contrôle à 100 % peut constituer
II peut être possible de calculer un programme plus
une très lourde tâche si le contrôle n’est pas effectué
efficace sur la base des données économiques dont
avec un appareillage d’essai automatique. Les résul-
on dispose.
tats obtenus ne sont en outre pas toujours corrects,
notamment lorsqu’un grand nombre d’individus pos-
2.3 Contrôle par attributs ou par
sèdent du point de vue dimensionnel, de l’aspect ou
des performances une ou plusieurs caractéristiques
mesures
marginales (c’est-à-dire voisines d’une tolerance ou li-
mite relative, ou groupées autour de cette limite). La méthode de contrôle par attributs consiste a exa-
Dans ces conditions, le tri manuel ou automatique miner un individu, ou des caractéristiques d’un indi-
risque de conduire à classer comme non conformes vidu, afin de le classer ((conforme)) ou ((non
certains individus conformes, et inversement. Le conforme)). On décide de l’attitude à adopter en
contrôle à 100 %, qu’il soit manuel, visuel ou auto- fonction du nombre d’individus non conformes ou de
matique, peut par ailleurs être conduit de façon non
non-conformités décomptés dans un échantillon au
satisfaisante. II peut parfois dégénérer en contrôle à hasard.
100 % superficiel, en l’absence de moyens financiers,
La méthode de contrôle par mesures consiste à pré-
en temps ou en personnel suffisants. II est enfin
lever un échantillon constitué d’un certain nombre
inapplicable si la méthode de contrôle necessite le
d’individus, et à mesurer certaines dimensions ou
recours à des essais destructifs.
caractéristiques. Les données obtenues ne portent
II est toutefois reconnu que les contrôles à 100 % donc pas uniquement sur le fait qu’une dimension, par
peuvent constituer une phase nécessaire du proces- exemple, est comprise ou non entre certaines limites,
sus de contrôle, à la fois pour le client et pour le mais sur la valeur effective de cette dimension. La
fournisseur. II existe des situations dans lesquelles ils décision d’acceptation ou de non-acceptation d’un lot
sont indispensables, par exemple les contrôles por- est prise sur la base de calculs de la moyenne et de
tant sur des défauts critiques, comme cela apparaît
la variabilité des mesures selon les procédures dé-
clairement à la lecture de la présente partie de crites dans I’ISO 3951 ou I’ISO 8423.
Q ISO
ISO 2859-O: 1995(F)
La méthode par mesures présente l’avantage, à Un échantillon est prélevé dans chaque lot et contrôlé,
condition que certaines hypothèses soient vérifiées, chaque lot est ensuite classé comme acceptable ou
.
de requérir un effectif d’échantillon plus petit que la non acceptable sur la base du nombre d’individus non
méthode par attributs pour assurer un niveau de pro- conformes ou de non-conformités trouvés dans
tection donné contre une décision incorrecte. Elle l’échantillon. Chaque lot successif est aonc traité
comme une unite plus ou moins indépendante (bien
fournit également davantage d’informations sur
que les regles de décision puissent parfois varier se-
l’existence d’une éventuelle influence négative sur la
qualité de la moyenne et/ou de la variabilité du pro- lon les résultats obtenus pour les lots précédents).
cessus. La méthode par attributs présente l’avantage
Pour les lots livrés isolement, se référer à 2.5.2.
d’être plus ((robuste)) (non soumise a des hypothèses
de forme de distribution) et plus simple à utiliser. Ces
avantages peuvent justifier le recours à la méthode
par attributs bien qu’elle implique des effectifs d’é-
chantillons, et donc des coûts afférents, plus élevés.
2.4.2 Effectif des lots
II est à noter qu’une évaluation de type tout ou rien
L’autorité responsable (voir ISO 2859-l : 1989, 3.10)
(bon ou mauvais) est plus rapide et demande moins
peut, de droit, spécifier elle-même l’effectif qui doit
de savoir-faire que l’obtention d’une mesure.
être celui du lot, mais il est évidemment préférable,
Les deux méthodes ont leurs propres avantages et
lorsque cela est possible, que cette valeur soit choisie
leurs propres applications typiques.
après consultation du fournisseur afin qu’elle soit
mutuellement satisfaisante. II va de soi que le choix
Bien qu’il soit fait référence à l’occasion à I’ISO 3951
de l’effectif du lot (de même que des autres paramè-
et à I’ISO 8423 dans la suite du texte, les program-
tres du plan d’échantillonnage) ne devrait jamais être
mes d’échantillonnage par mesures ne sont pas dé-
effectué sans tenir compte du processus de produc-
veloppés en tant que tels dans le cadre de la présente
tion. II n’est pas essentiel que la valeur choisie soit
partie de I’ISO 2859. L’ISO 3951 et I’ISO 8423
étroitement fixee. Certaines variations sont admissi-
contiennent des conseils d’utilisation les concernant.
bles, mais il est presque toujours souhaitable que
soient spécifiées les limites inferieure et supérieure
de l’effectif du lot.
Du point de vue du contrôle par échantillonnage, il est
2.4 Contrôle de lots
avantageux de prendre des lots d’effectif important,
puisqu’il sera alors possible de prélever dans ces lots
des échantillons d’effectif important, donc de mieux
2.4.1 Lot
discriminer les bons et les mauvais lots. Si le lot est
grand, l’échantillon à prélever représente, à NQA égal,
Dans le cadre de I’ISO 2859, de I’ISO 8422, de
une plus petite proportion du lot que si celui-ci est
I’ISO 3951 et de I’ISO 8423, les individus ne sont pas
petit.
soumis au processus de contrôle de façon indivi-
duelle, mais en groupes. Chacun de ces groupes
II faut toutefois rester modéré dans l’emploi de cette
d’individus est appelé un lot.
stratégie des ((grands lots). Si la constitution de
Chaque lot doit, dans la mesure du possible, se com- grands lots nécessite de combiner des lots plus petits
poser d’individus fabriques sensiblement dans les qui auraient très bien pu rester séparés, l’emploi d’un
mêmes conditions et au cours d’une même période. grand lot n’est intéressant que si les petits lots sont
Ceci est de la plus grande importance si le concept tous d’une qualité équivalente. Si l’existence de dif-
férences de qualité importantes entre les petits lots
de niveau de qualité acceptable est adopte et qu’une
série de lots doit être livrée. paraît probable, il est de beaucoup préférable de les
garder séparés. C’est pourquoi il convient d’opérer sur
Si la livraison est constituée d’individus provenant de
des lots constitués d’unités de produit fabriquées
plusieurs sources différentes, la présence d’un grand
dans des conditions sensiblement identiques.
nombre d’individus non conformes provenant de l’une
de ces sources pourra conduire à la non-acceptation
Des exemples de formation des lots sont donnés en
du mélange provenant de l’ensemble des sources.
34 . .
Inversement, la présence d’un produit de qualité
marginale provenant de l’une des sources pourra être Des informations plus détaillées sur le rapport effectif
masquée par le mélange avec le produit de sources du lotleffectif de l’échantillon sont données dans
d’excellente qualité. l’lSO/TR 8550.
Q ISO
ISO 2859=0:1995(F)
Si la livraison ne comporte qu’un seul lot, il peut être
2.5 Contrôle de séquences de lots ou de
utile de savoir si ce lot fait partie d’un ensemble de
lots isolés
lots similaires livrés à d’autres clients et provient d’un
processus maîtrisé, ou s’il s’agit d’un lot mélangé
formé d’individus issus de différents processus et
2.5.1 Contrôle lot par lot
produits à des moments différents (voir également
l’lSO/TR 8550 et I’ISO 2859-2).
Un contrôle lot par lot est le contrôle d’un produit
fourni sous forme d’une série de lots.
Tandis que I’ISO 2859-l et I’ISO 2859-3 exigent que
la valeur du NQA et le niveau de contrôle soient fixés
Lorsqu’une séquence de lots est soumise à I’accep-
a l’avance, I’ISO 2859-2 demande que soit établie la
tation au moment de la production, il peut arriver que
qualité limite (QL). Pour assurer un niveau de protec-
les résultats de contrôle concernant les premiers lots
tion approprié du fournisseur et du client lorsque les
soient disponibles avant que les lots suivants aient été
lots sont échantillonnés selon la procedure de la qua-
fabriqués. II est donc possible d’utiliser les résultats
lité limite, il est nécessaire de savoir si le lot provient
déjà obtenus pour améliorer la qualité du reste de la
d’une série continue de lots acceptables ou s’il s’agit
production. II convient de présenter et de contrôler les
d’un lot mixte constitué d’unités de produit fabriquées
lots suivant leur séquence de fabrication, et de réali-
sur différentes lignes de production et/ou à des dates
ser les contrôles sans délai. Les informations obte-
différentes.
nues à partir d’un lot peuvent indiquer une
dégradation apparente du niveau de qualité du pro-
Les tables de I’ISO 2859-2 sont principalement desti-
cessus. Les informations obtenues à partir de plu-
nées au contrôle des lots isolés.
sieurs lots successifs peuvent être utilisées pour
mettre en œuvre une procédure de modification du
contrôle exigeant le passage à un plan d’échantillon-
2.6 Niveau de qualité acceptable (NQA)
nage plus sévère lorsque le processus se dégrade.
Cette façon de procéder est importante car elle
2.6.1 Description
constitue le meilleur moyen d’assurer à long terme la
protection optimale du client contre une qualité insuf-
Le niveau de qualité acceptable est utilisé comme
fisante. Si le niveau de qualité reste médiocre, I’appli-
critère d’indexation dans les tables de I’ISO 2859-1,
cation d’un contrôle plus sévère conduira à ce qu’un
dans I’ISO 3951 et dans certaines des tables de
plus grand nombre de lots seront retournés au four-
I’ISO 8422 et de I’ISO 8423.
nisseur pour être triés. Cet accroissement de la se
vérité de l’échantillonnage se traduit pour le
L’utilisation des plans d’échantillonnage indexés
fournisseur par un plus grand risque de voir juger non
d’après le NQA conduit à l’acceptation, dans presque
acceptable un lot acceptable. L’identification d’une tous les cas, des lots issus d’un processus dont le
dégradation probable de la qualité du produit constitue niveau de qualité est égal ou meilleur que le NQA.
un signal qui devrait le conduire à prendre des mesu-
Si l’on considère une série continue de lots, le NQA
res correctives.
est un niveau de qualité défini, en contrôle par
Si la qualité est très sensiblement meilleure que celle échantillonnage, comme la limite d’une qualité
qui a été convenue, le client peut, avec l’accord de moyenne acceptable d’un processus.
l’autorité responsable, décider d’adopter un plan d’é-
Le NQA est une borne, que l’on choisit, séparant ce
chantillonnage réduit ou d’échantillonnage successif
qui sera considéré comme une qualité moyenne ac-
partiel.
ceptable et une qualité moyenne non acceptable d’un
L’ISO 2859-1, I’ISO 2859-3, I’ISO 8422, I’ISO 3951 et
processus. Ce paramètre ne décrit pas le plan d’é-
I’ISO 8423 sont principalement destinées au contrôle
chantillonnage, mais définit ce que la production de-
de séquences de lots. vrait être, et constitue une grandeur utile à prendre
en considération pour définir un processus admissi-
ble .
2.5.2 Contrôle d’un lot isolé
Le fait qu’on a spécifié un NQA ne doit pas être
Le contrôle peut parfois être effectué sur un lot isolé, compris comme impliquant qu’on désire un pourcen-
un petit nombre de lots isolés, ou des lots mis en tage d’individus non conformes égal à la valeur spé-
stock à un moment où la production était terminée.
cifiée ou est tout à fait acceptable. II est toujours
Dans ces conditions, il n’y a pas lieu de recourir aux
préférable de n’avoir aucun individu non conforme,
règles de modification et d’influencer ainsi la qualité
plutôt qu’un pourcentage quelconque, et plus le
offerte. pourcentage pourra être inférieur au NQA, mieux ce
Q ISO
ISO 2859-O: 1995(F)
sera. Cette réduction augmente la probabilité d’ac-
II est très important de bien comprendre que ce pa-
ceptation de chaque lot. ramètre n’est pas, à la différence du NQA, de la
LQMAC (voir 2.12) ou de la QL (voir 2.81, une valeur
qui peut être calculée ou choisie, ou constitue une
propriété d’un plan d’échantillonnage particulier, mais
2.6.2 Établissement d’un NQA
qu’il se rapporte au produit de fabrication effectif, in-
dépendamment du mode de réalisation des contrôles.
Lorsque l’on fixe la valeur du NQA, il faut toujours
garder en mémoire le fait que celui-ci constitue une
En règle générale, l’estimation de la qualité moyenne
indication du niveau de qualité que doit atteindre la
d’un processus ne constitue pas une partie essen-
production. Ce qui est demandé au fournisseur est de
tielle du programme d’échantillonnnage. Toutefois,
produire des lots de qualité moyenne meilleure que le
cette grandeur présente en elle-même une certaine
NQA. Cette qualité doit d’un côté être raison-
importance. Le contrôleur autant que le fournisseur
nablement possible à atteindre et d’un autre côté
sont concernés non seulement par les résultats des
constituer une qualité raisonnable du point de vue du
contrôles lot par lot, mais aussi par l’évolution a long
client. Ceci implique souvent de trouver un compro-
terme de la qualité de la production.
mis entre le niveau de qualité que le client souhaiterait
et celui qui est dans ses moyens car, plus ses exi-
II est donc souhaitable de conserver un enregis-
gences seront élevées, plus elles seront difficiles à
trement de la qualité moyenne globale estimée du
satisfaire, et plus le coût des contrôles de vérification
processus car elle constitue une excellente mesure
sera élevé.
de la qualité et un élément d’information intéressant
pour ceux qui ont à décider quels plans d’échantillon-
Un processus correctement conçu et maîtrisé peut
nage il convient d’adopter lorsque des produits simi-
permettre la fabrication d’un produit caractérisé par
laires seront conçus et fabriqués dans l’avenir.
un pourcentage d’individus non conformes inférieur
au NQA, Lorsqu’il est possible d’améliorer la qualité
II faut observer des règles spéciales pour les échan-
moyenne d’un processus, les coûts de production et
tillonnages de type double ou multiple. Dans ce cas,
de contrôle associés a cette qualité améliorée seront
il convient de n’utiliser que les résultats portant sur le
réduits.
premier échantillon pour l’estimation de la qualité
moyenne du processus.
Le critère primordial doit être l’exigence du client,
mais il est nécessaire de s’assurer que celle-ci est
II arrive parfois qu’une recommandation d’exclusion
réaliste et que le client ne demande pas une qualité
des résultats aberrants soit formulée. Cette pratique
meilleure que celle qui est réellement nécessaire. II
est dangereuse et il convient de n’y recourir qu’avec
est indispensable de tenir compte de l’utilisation pré-
parcimonie, sinon de l’éviter totalement. Le seul cas
vue des individus considérés, et des conséquences
où elle peut être adoptée sans risque est celui où l’on
d’une éventuelle non-conformité. Si les individus sont
a pu établir que les résultats aberrants résultent d’une
fournis en grand nombre et que la non-conformité ne
cause spécifique, et que celle-ci a été éliminée.
concerne que le montage, de sorte qu’un individu
Même dans ce cas, il est bon d’enregistrer à la fois
non-conforme peut être mis de côté et remplacé par
les chiffres incluant et excluant ces résultats aberrants
un autre, un NQA relativement élevé peut être ad-
afin d’indiquer l’existence de ces non-conformités.
missible. Si, par contre, une non-conformité est sus-
ceptible d’entraîner un disfonctionnement d’une partie
Dans le cas de caractéristiques multiples ou de clas-
coûteuse et importante d’un équipement en un lieu
ses multiples de NQA, il faut estimer séparément les
et a un moment où le remplacement de l’individu non
qualités moyennes de processus correspondantes.
conforme est impossible, il faudra choisir un NQA plus
sévère.
Des informations complémentaires et des indications
2.8 Qualité limite (QL)
sur l’établissement du NQA sont données en 3.9 et
dans l’lSO/TR 8550.
La qualité limite est un critère d’indexation, utilisé
dans I’ISO 2859-2. Lorsqu’un lot est considéré isolé-
ment, la qualité limite est un niveau de qualité, ex-
primé en pourcentage d’individus non conformes (ou
2.7 Qualité moyenne d’un processus
en nombre de non-conformités par 100 individus), qui,
La qualité moyenne du processus est la qualité
en contrôle par échantillonnage, est associé a une
moyenne d’une série de lots présentés, à l’exclusion
probabilité d’acceptation faible. Cette probabilité d’ac-
des lots proposés en deuxième présentation. ceptation faible est appelée ((risque du client)).
Q ISO
ISO 2859-O: 1995(F)
II faut souligner que la qualité limite ainsi spécifiée est normal) et la procédure de contrôle renforcé sera ins-
en fait un niveau de qualité non souhaité. Pour que
tituée afin de protéger le client.
des lots soient régulièrement jugés acceptables, il
D’autres informations, avec des exemples, sont don-
faut que la proportion d’individus non conformes soit
nées à ce sujet en 3.11 et 3.12.
très sensiblement inférieure à la QL (généralement a
un quart de la QL).
2.10 Contrôle réduit
L*IS0 2859-2 décrit des procédures d’application des
plans d’échantillonnage indexés d’après la QL. Ces
II apparaît parfois au cours des contrôles que la qualité
plans et les tables associées sont en grande partie
du produit est notablement meilleure que le NQA.
concordants avec les plans d’échantillonnage de
Dans ce cas, et s’il existe des raisons de penser que
I’ISO 2859-1.
cette bonne qualité de production sera maintenue, il
n’est plus nécessaire de recourir à des plans d’é-
Les plans indexés d’après la qualité limite sont es-
chantillonnage pour séparer les bons et les mauvais
sentiellement destinés au contrôle de lots isolés.
lots. Les contrôles ne peuvent toutefois pas être pu-
Lorsque le produit est en cours de fabrication et
rement et simplement supprimés car ils constituent
qu’une série de lots est produite, les procédures dé-
un moyen nécessaire de détection d’une éventuelle
crites dans I’ISO 28531 sont mieux appropriées.
dégradation de la qualité de la production.
Dans cette situation, il est possible, si on le souhaite,
2.9 Contrôle normal et contrôle renforcé
de réaliser d’importantes économies en passant aux
plans d’échantillonnage réduit de I’ISO 2859-l ou aux
Le NQA, rappelons-le, est la borne séparant les ni-
plans d’échantillonnage successif partiel de
veaux de qualité définis comme satisfaisants et non
I’ISO 2859-3. Les règles spéciales permettant de re-
satisfaisants lors du contrôle d’une série de lots. Une
courir à ces plans, avec l’accord de l’autorité respon-
fois spécifiée une valeur du NQA pour un produit
sable, sont décrites dans I’ISO 2859-1, et dans
donne quelconque, l’idéal serait de disposer d’un
I’ISO 2859-3 et également dans la section 3 de la
système dans lequel les lots de qualité meilleure que
présente partie de I’ISO 2859.
le NQA seraient toujours acceptés et les lots de qua-
lité moins bonne que le NQA toujours refusés. Aucun
Le contrôle réduit est présenté ci-après, avec des
plan d’échantillonnage ne permet toutefois d’atteindre
exemples, en 3.15.
cet idéal.
Pour satisfaire à la fois aux exigences du fournisseur
2.11 Règles de modification du contrôle
et du client, il faut recourir a une solution de compro-
mis, et le dispositif adopté à cet effet dans Les concepts de contrôle normal et de contrôle ren-
I’ISO 2859-1 et I’ISO 8422 consiste à définir un
forcé, ainsi que les objectifs de ces types de contrôle,
contrôle normal et un contrôle renforcé, avec spéci-
ont été pr
...














Questions, Comments and Discussion
Ask us and Technical Secretary will try to provide an answer. You can facilitate discussion about the standard in here.
Loading comments...