CookBook 08 Determination of Conformance - 09-2018

EUROLAB “Cook Book” – Doc No.
DETERMINATION OF CONFORMANCE WITH SPECIFICATIONS USING MEASUREMENT

UNCERTAINTIES – POSSIBLE STRATEGIES
Introduction
Conformity assessment is a common activity performed in testing, inspection and calibration, required to
assure the compliance of products, materials, services and systems to requirements defined by
standards, regulations, legal frameworks and contract agreements, being defined to establish confidence
for consumers and for the safety and quality of life. Today, it has a major impact on the global economy
as it implies the acceptance and rejection of items directly affecting risk analysis, business decisions,
and reputational and financial costs.
In the evaluation of compliance, which is based on
quantitative results, different scenarios can be
considered, which can be illustrated in 4 case studies
(cases A to D, see Figure 1). In this, cases A and D are
unambiguous as the decisions are not influenced by the
measurement uncertainties. However, in cases B and C
where the measurement uncertainty interval is
overlapping the limit value, careful analysis that should
establish objective criteria (decision rule) is required to
accept results that are outside the tolerance with part of
the measurement uncertainty interval.
Fig. 1: Test results and their measurement
uncertainties in relation to an upper limit value
A general approach to conformity assessment

Decisive for a suitable definition of a decision rule is the question of what should be proved with the
conformity assessment: compliance or non-compliance with a specification or a limit value. Based on the
answer, either the supplier’s risk (α) or the consumer’s risk (β) has to be specified.
Defining a procedure to perform the conformity assessment may be based on the following steps:
a. The specification of a measurand (Y) and the measurement item to be tested.

b. The experimental / analytical results (estimates y of the measurand Y).
c. The measurement standard uncertainty, u(y), and for a certain confidence level, the expanded
measurement uncertainty.
d. The specification of a single tolerance limit (upper or lower) or tolerance interval limits.
e. The definition of the acceptance zone, rejection zone and a guard band assuming a probability
of type I error (supplier’s risk α) or type II error (consumer’s risk β).
f. A decision rule.
The terminology adopted is described in known references, [EURACHEM Guide:2007],

[ASME B89.7.3.1:2001] and [EUROLAB Technical Report 1/2017]. Two of these are particularly
relevant.
• Decision rule: a documented rule that describes how measurement uncertainty will be
allocated with regard to the acceptance or rejection of a product according to its specification
and the result of a measurement.
• Guard band: the magnitude of the offset from the specification limit to the acceptance or
rejection zone boundary.
Rev. 2018 1
EUROLAB “Cook Book” – Doc No. 8
Establishing the decision rule

In the event that regulations or normative standards contain provisions for compliance with specifications
or limit values taking into account measurement uncertainties, these provisions have to be applied. If
such provisions are missing, rules have to be established prior to testing to meet market or safety
requirements.
The international standard ISO 14253:2016 part 1: Decision rules for proving conformance or non-
conformance with specifications distinguishes whether conformance or non-conformance shall be
determined with a high degree of probability. The expanded measurement uncertainty U and a
confidence level of approx. 95% (coverage factor k = 2) is generally considered adequate. Only in
exceptional cases a higher confidence level of e.g. 99% (coverage factor k = 3) is chosen.
The determination of decision criteria should take into account whether the specification is an interval or
a limit (upper or lower), whether guard bands should be considered and, if so, whether they should
reduce or enlarge the acceptance interval. The following Figures illustrate various possibilities (where TU
– tolerance upper limit; GU – guard band upper limit, TL – tolerance lower limit, GL – guard band lower
limit, U(y) – expanded uncertainty of the measurement.
Figure 2 – Example of areas defined for a tolerance interval in order to minimise the consumer’s risk
Figure 3 – Example of areas for the tolerance interval in order to minimise the supplier’s risk
Figure 4 – Guard band for upper tolerance and Figure 5 – Guard band for upper limit and guarded
guarded acceptance defined with a confidence level rejection
of 95%
In the case where guard bands are used, in particular for measurement results with the same
uncertainty, it may be a simple strategy to establish a decision rule by comparing the measurement
results with the acceptance zone limits, where the measured value must be within these acceptance
zone limits, otherwise rejected.
Rev. 2018 2
If the measurement results have different measurement uncertainties, it is recommended to consider an

approach without guard bands.
Figure 6 – Example with single upper tolerance Figure 7 – Example with single lower tolerance
In these cases, the criteria may be established by performing a hypothesis test where fulfilment of the H0
condition implies the decision of acceptance and, otherwise, implies the decision of rejection. Therefore,
assuming the probability of type I error (), the decision rule can be expressed as:
Decision rule
Acceptance, if the hypothesis H0: is true;
Rejection, if the hypothesis H0 is false, .
Expression to test:
The following is a pratical example of application :
Consider a measurement estimate of y = 2,7 mm with a measurement uncertainty of u(y) = 0,2 mm, a
single tolerance upper limit of TU = 3,0 mm, and a specification of conformity (1 – ) of 0,95 (95 %) and
thus assuming a type I error of  = 0,05 (5%).
With the experimental result and the tolerance limit, assuming a normal PDF (Probability Distribution
function), the decision rule will be as follows:
Acceptance, if the hypothesis H0: is true
Rejection, if the hypothesis H0: is false
To estimate the probabilities for the given example, the conformance probability (Pc) need to be
calculated using the general expression for normal PDFs:
(93,3 %) < 0,95
Thus, the hypothesis H0 is false and the decision to be taken is rejection (non-compliant).
Note:
The value of can be obtained by using tables of standard Gaussian PDF or by
software having functions to perform this type of calculations, e.g.:
MS Excel function NORMDIST (x, mean, standard deviation, cumulative), for the case
described above is: NORMDIST(3,0;2,7;0,2;TRUE) and would be the result ( 0,933 ).
Rev. 2018 3
Literature:
• ISO 14253-1: Geometrical product specifications (GPS) - Inspection by measurement of
workpieces and measuring equipment - Part 1: Decision rules for verifying conformity or
nonconformity with specifications (2017)
• EURACHEM Guide: Use of uncertainty information in compliance assessment (2007)
• ASME B89.7.3.1: Guidelines for Decision Rules: Considering Measurement Uncertainty,
Determining Conformance to Specifications (2001)
• EUROLAB Technical Report 1/2017: Decision rules applied to conformity assessment (2017)
• ISO/IEC Guide 98-4, Uncertainty of measurement — Part 4: Role of measurement uncertainty in
conformity assessment (2012)
Rev. 2018 4

CookBook 08 Determination of Conformance - 09-2018

Uploaded by

Copyright:

Available Formats

CookBook 08 Determination of Conformance - 09-2018

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CookBook 08 Determination of Conformance - 09-2018

Uploaded by

Copyright:

Available Formats

EUROLAB “Cook Book” – Doc No.

DETERMINATION OF CONFORMANCE WITH SPECIFICATIONS USING MEASUREMENT

A general approach to conformity assessment

a. The specification of a measurand (Y) and the measurement item to be tested.

The terminology adopted is described in known references, [EURACHEM Guide:2007],

Establishing the decision rule

If the measurement results have different measurement uncertainties, it is recommended to consider an

The following is a pratical example of application :

Rejection, if the hypothesis H0: is false

(93,3 %) < 0,95

You might also like