EEMUA 175 - Code of Practice, Cal & Valid Process Analysers - 2013

THE ENGINEERING EQUIPMENT AND MATERIALS USERS' ASSOCIATION
EEMUA Code of Practice
Calibration and validation

of process analysers
PUBLICATION 175
Edition 2
Copyright © 2013 The Engineering Equipment and Materials Users' Association.

A company limited by guarantee. Registered in England. Company number 477838.
ISBN 978 0 85931 197 7

Imprint reference 09-2013
Registered and Trading Address

63 Mark Lane London EC3R 7NQ
Telephone: +44 (0)20 7488 0801
Fax: +44 (0)20 7488 3499
E-mail: sales@eemua.org
Website: www.eemua.org
EEMUA Publication 175 – Calibration and validation of process analysers © EEMUA
ENGINEERING EQUIPMENT AND MATERIALS USERS' ASSOCIATION
About EEMUA
The Engineering Equipment and Materials Users’ Association is usually known by

its initials, EEMUA. EEMUA is governed by a Council of members. This Council is
responsible for the policies, financing, membership, and strategy of the
Association. Any organisation that is a substantial user or purchaser of
engineering products is eligible to apply to become a member of EEMUA.
Membership is subject to the approval of Council.
EEMUA Aims
EEMUA aims to improve the safety, environmental and operating performance of

industrial facilities in the most cost-effective way, pursuing and promoting
leadership in industrial asset management by sharing engineering experiences and
expertise, and by the active, enlightened promotion of the distinct interests of
users of engineering products. These aims are achieved by various means,
including:
– Providing the organisation within which networking, information sharing and

collaboration among users on non-competitive technical matters can take
place.
– Influencing the way written regulations are interpreted and applied in practice.
– Promoting and presenting members’ views, and encouraging the application of
good sound engineering practices.
– Developing user guides, specifications, training and competency schemes.
– Facilitating members’ participation in national and international standards
making.
– Influencing relevant national and European legislation and regulation.
The income and resources of EEMUA are applied solely towards the promotion of
the aims of the Association. No profit is transferred by way of dividend, bonus or
otherwise to members. The work of the Association is mainly carried out by
members’ representatives on EEMUA Technical Committees, Working Groups, and
Forums, supported by EEMUA staff.
Incorporation, Membership and Other Information
For the purposes of conducting its business, EEMUA is incorporated in England and
Wales as a company, limited by guarantee, not having share capital. All usual
aspects of company management and business administration are dealt with by
the EEMUA executive directors, including policy implementation and leadership of
the executive team (the staff).
A list of EEMUA publications for sale is given at the end of this Publication. The full
list is also on the Association's website, including details of on-line shopping
facilities.
To enquire about corporate Membership, write to enquiries@eemua.org or call +44

(0)20 7488 0801.
ii
ABOUT THIS PUBLICATION
Legal Aspects
All rights, title and interest in this Publication belong to EEMUA. All rights are
reserved. No part of this Publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means: electronic, mechanical,
photocopying, recording or other, without the express prior written agreement of
the EEMUA Chief Executive. Infringement of copyright is illegal and also reduces
the Association's income thereby jeopardising its ability to fund the production of
future publications.
It has been assumed in the preparation of this Publication that: the user ensures
selection of those parts of its contents appropriate to the intended application; and
that such selection and application is correctly carried out by appropriately
qualified and competent persons for whose guidance this Publication has been
prepared. EEMUA and the individual members of the Work Group that prepared
this EEMUA Publication do not, and indeed cannot, make any representation or
give any warranty or guarantee in connection with material contained herein.
EEMUA and the individual members of the Work Group that prepared this EEMUA
Publication expressly disclaim any liability or responsibility for damage or loss
resulting from the use or reliance on this Publication. Any recommendations
contained herein are based on the most authoritative information available at the
time of writing and on current good engineering practice, but it is essential for the
user to take account of pertinent subsequent developments and/or legislation.
Any person who encounters an inaccuracy or ambiguity when making use of this
Publication is asked to notify EEMUA without delay so that the matter may be
investigated and appropriate action taken.
Consultation and Feedback
EEMUA encourages constructive comments on this Publication from both Members

of the Association and others.
Comments should be sent on the standard feedback form, a copy of which is

provided towards the end of this Publication. An electronic version of the feedback
form is available from EEMUA - e-mail your request for a copy to
enquiries@eemua.org, telephone your request to +44 (0)20 7488 0801, or
download the form from the EEMUA website at www.eemua.org. Submit
comments by e-mail (preferred) or post.
Comments are considered by the relevant EEMUA Technical Committees and may
be incorporated in future editions of this Publication. New editions are publicised
on the EEMUA website
iii
iv
Contents
Preface .................................................................................................... ix
1. Scope: calibration and validation .......................................................... 1
2. Definitions ............................................................................................ 3
2.1 Equipment ......................................................................................... 3
2.2 Measurement ..................................................................................... 3
2.3 Calibration and validation .................................................................... 4
2.4 Statistical terms ................................................................................. 5
3. General considerations: calibration and validation................................ 9
4. Preconditions for analyser calibration and validation ......................... 11
4.1 General ........................................................................................... 11
4.2 Factory Acceptance Testing (FAT) ....................................................... 11
4.3 Site installation ................................................................................ 11
4.4 Site Acceptance Testing (SAT) ............................................................ 12
4.5 Site precision ................................................................................... 12
4.6 Read-out instrumentation .................................................................. 12
5. Principles of initial calibration, correlation and initial validation......... 13
5.1 Initial calibration and correlation ......................................................... 13
5.2 Probationary/initial validation ............................................................. 14
5.3 Analyser result and accepted reference value comparison methods. ......... 14
6. Principles of routine validation and calibration adjustment ................ 15
6.1 General ........................................................................................... 15
6.2 Analyser result and reference test/primary test method comparison
techniques ............................................................................................ 16
6.2.1 Reference sample method ........................................................ 16
6.2.2 Paired sample method ............................................................. 16
6.3 Routine validation (scheduled check of continuing performance) .............. 17
6.4 Calibration adjustment (adjustment, if required, to re-establish analyser
results within performance limits)............................................................. 17
6.5 Routine validation and calibration adjustment intervals .......................... 18
6.6 Number of result values required for routine validation and calibration
adjustment ........................................................................................... 19
6.7 Routine validation and calibration adjustment of model based/chemometric
process analysers................................................................................... 20
7. Comparison methods: reference sample method and paired sample
method ................................................................................................... 21
7.1 Comparing analyser and reference method results................................. 21
7.2 Routine validation and calibration adjustment using the reference sample
method ................................................................................................ 21
7.2.1 Advantages and limitations ...................................................... 21
7.2.2 Choice of reference samples ..................................................... 22
7.2.3 Testing of reference samples .................................................... 23
7.3 Routine validation and calibration adjustment using the paired sample
method ................................................................................................ 23
7.3.1 Advantages and limitations ...................................................... 23
7.3.2 Collecting paired sample results ................................................ 25
8. Evaluation of routine validation and calibration adjustment results: use
of statistical tools and control charts ...................................................... 27
8.1 General ........................................................................................... 27
8.2 Use of statistical tools ....................................................................... 27
8.3 Routine validation and calibration adjustment: use of control charts......... 27
8.3.1 CUSUM approach .................................................................... 29
8.4 Control chart limits ........................................................................... 30
v
8.5 Control chart limits derived from analyser reference sample or paired sample
laboratory data ...................................................................................... 31
Appendix 1: Example interpretation of control chart readings ................ 33
Appendix 2: Determination of control chart limits by measuring standard
deviations of differences ......................................................................... 35
References and Bibliography .................................................................. 37
vi
Figures
Figure 1 Example of a control charting diagram……………..…………………..28
Figure 2 Examples of analyser results in a CUSUM chart………………………29
Figure 3 Example of accurately distributed control chart reading………….33
Figure 4 Example of biased control chart reading…………………….............33
Figure 5 Example of drifting control chart reading………………………………33
Figure 6 Example control chart reading; value outside warning limit…….34
Figure 7 Example determination of control chart limits by measuring
standard deviations…………………………………………………………………………36
Tables
Table 1 Sample validation uncertainty results…………………………………….19
Table 2 Example of the control chart limits for a distillation analyser.…..35
vii
viii
Preface
This Code of Practice has been prepared by the EEMUA On-Line Analysers
Committee. It replaces the First Edition of this Publication, previously named
‘Calibration and checking of process analysers’, which in turn replaced the
following historical IP Codes of Practice relating to analysers: IP 340/92, 341/82,
347/82, 348/82, 349/82 and 353/82.
Other associated EEMUA publications are:
EEMUA 138 - Design and installation of on-line analyser systems.
EEMUA 226 - Design and installation of on-line analyser systems - a guide to

technical enquiry and bid evaluation.
EEMUA 187 - Analyser systems - a guide to maintenance management.
The word analyser has been used throughout this Publication to refer to
instruments, variously known as on-line analysers, process stream analysers,
process analysers, quality analysers, quality measuring instruments and process
quality monitors.
ix
x
1. Scope: calibration and validation

This Publication is a general guide to the principles and methods used for the
calibration and validation (verification) of process (on-line) analysers. It covers
the principles of initial calibration and correlation, in the factory or on the plant,
and initial validation. The detailed guidance is focussed on continued (routine)
validation (in-service checking) and calibration adjustment (only if required, based
on appropriate validation results) including the use of control charts.
The content and presentation of this Publication recognises that the effort available
for routine validation and calibration adjustment is limited. The simplified
approaches recommended are therefore intended to require only a reasonable
expenditure of effort in order to maintain assurance that the initial calibration and
correlation of the analyser to the reference method is still valid.
In some cases the simplified approaches will produce procedures, which although
adequate, may be less rigorous than could be evolved on purely technical and
theoretical grounds. The approaches should be amended to ensure that they meet
the individual business needs of the manufacturing or other process to which they
are applied.
It should be noted that this Publication is primarily intended for use by those
directly involved with supplying, using and maintaining on-line/process analysers.
These individuals are generally not specialists in quality measurement principles
and the associated statistical calculations. Therefore, as far as possible, this
Publication avoids the explicit use of statistical methods and favours descriptions
and explanations of a more practical nature. In order to remain consistent with
the theoretical approach these practical approaches are assumed to exhibit
random statistical behaviour with a normal/Gaussian distribution.
This Publication is not intended to cover Continuous Emissions Monitoring Systems

(CEMS). These systems have calibration, validation, on-going quality assurance
(routine validation and calibration adjustment) and the associated statistical tools
defined by other standards, e.g. BS EN 14181 - Stationary source emissions:
quality assurance of automated measuring systems(1).
This Publication is not intended to cover automatic or semi-automatic instruments

used mainly in laboratories to analyse discrete samples. However, the principles
and techniques described can also be applied to laboratory instruments.
For consistency with BS EN 60359 - Electrical and electronic measurement

equipment expression of performance(2), this Publication treats the definitions of
validation and verification as equivalent.
1
2
2. Definitions
The following glossary aims to add further definition to the terms used in the text
in the context of their use when describing activities associated with process
analysers.
The definitions are grouped by a usage category (additional external references

are shown in brackets). Similar words or phrases are listed with a common
description.
2.1. Equipment
Analyser / on-line analyser / process analyser / process stream analyser

/ quality analysers / Quality Measurement Instruments (QMI). An
instrument used to measure continuously and automatically a particular property
or chemical composition of a process stream. The input to the instrument is a
sample or portion of the process stream, either introduced continuously or at
specific points in the analyser operating cycle. The sample is conditioned in a
particular way, for example; flow, pressure, temperature or cleanliness. The
output is a transmitted result or reading displayed on suitable read out
instrumentation such as a control system screen or other operator interface. The
analyser is normally located permanently within the process area.
Analyser system / total analyser system (ASTM D3764(3)). The complete

instrumentation system, including sample extraction and return, sample
conditioning, analytical measurement, and result display/readout/operator
interface. The analyser system typically comprises the sampling system, the
analyser, the read-out instrumentation and all associated piping, wiring and data
communications and operator displays.
Analyser time constant. For a continuous analyser this is the time interval
between the initial response and the time to reach 63 per cent of the final output
value with a step change in the measured property. For cyclically sampling
analysers the time constant is the analyser cycle time.
Readout instrumentation. The device used to accept the signal transmitted

from the analyser and display it for use in process operation or other decision
making. Displayed as a measured property or concentration in the accepted units
of that property or concentration.
2.2 Measurement
Absolute property. Where the measurement result is independent of the type of

equipment and the method used for its measurement. The result from the
analyser and the reference/standard test should be identical.
Analyser range / measuring range / range (IEC 60050-311(4)). The

smallest and largest values that can be represented as an analyser result. The
range is defined by two values of the measurement within which the limits of
uncertainty of the measuring instrument are specified.
Analyser result / analyser system result (ASTM D3764)/ result/ result

value. The measured property or concentration reading, in the measurement
3
units of that property or concentration displayed or recorded by the analyser and

transmitted to the read out instrumentation.
Bias (ASTM E456(5)). The difference between the analyser results and the
accepted reference value. A consistent offset between the analyser result and the
accepted reference value.
Dead band (IEC 60050-311). The finite range of values within which a
variation of the input variable does not produce any measurable change in the
output variable.
Direct measurement (BS EN 60359). The value of the analyser result is

detected directly, without the need for other calculations from a relationship
between the measurement and the property being measured.
Drift (BS EN 61207-1(6), IEC 60050-311). Change in the indication of a

measuring instrument, generally slow, continuous, not necessarily in the same
direction and not related to a change in the properties or composition of the
sample being measured.
Empirical property. Where the measurement is only definable in terms of the

reference/standard test. The result from the analyser and the reference/standard
test may differ, and the analyser should be calibrated to agree with the reference
test result.
Indirect measurement (BS EN 60359). The value of the analyser result is

obtained indirectly using a calculation that represents a known mathematical
relationship combined with a directly measured/detected property. This can
include chemometric models based on a statistical or mathematical analysis of
multiple reference samples to develop a calibration model for the analyser based
on several measurement parameters, e.g. multiple wavelengths for a
spectrometer.
Linearity (IEC 60050-311). The ability of a measuring instrument to provide an

indication having a linear relationship with a defined quantity.
Property. The characteristic or compositional parameter that is required to be

indicated by the analyser result.
Relative sensitivity (IEC 60050-311). Change in the indication divided by the

corresponding change in the value of the property or concentration measured.
Sensitivity. The value of the smallest measurement that can be made and seen
on the read out instrumentation.
2.3 Calibration and validation
Accepted reference value (ASTM D3764) / reference / reference test /

validation reference value. The numerical value of the particular property
reference material against which the analyser result is compared. A value that
serves as an agreed reference for comparison, which can be an established
theoretical value based on scientific principles, and assigned more certified value
based on experimental work of some national or international organisation, or a
4
consensus value based on collaborative experimental work under the auspices of a

scientific or engineering group.
Analyser calibration / calibration (ASTM E456, BS EN 60359). The process,

or set of operations that establishes a relationship between an instrument
indication and a known standard value under specified conditions.
Calibration adjustment. The process of adjusting the results of an analyser to

match the accepted reference value of a reference sample presented to it. This is
usually done after the reason for the exceptional difference between the analyser
result and the accepted reference value has been investigated and any faults
corrected, or based on a set of laboratory cross check sample results.
Certified reference material / reference sample / validation reference

material (ASTM D3764). A material having an accepted reference value
assigned for the property or composition of interest.
Checking / continual validation (ASTM D3764) / on-going validation /

routine validation / verification. The assurance process by which the bias and
precision performance determined during initial validation are shown to be
maintained. The presentation of a reference sample to an analyser in order to
determine the difference between the analyser result and the accepted reference
value. A routine procedure to monitor the performance of the analyser over an
extended period using control charts. No adjustment to the calibration of the
analyser is made unless the control charts indicate a real change has occurred.
Correlation (ASTM D3764). Derivation of the mathematical relationship

between the analyser measurement output and the accepted reference values of
the reference samples.
Initial calibration. Statistically based calibration to establish the relationship

between the analyser results across its entire range using a significant number of
reference materials. To accurately establish the relationship between the analyser
and the corresponding standard test result or certified reference material.
Initial validation / probationary validation (ASTM D3764). Validation

performed using a range of reference materials that were not part of the initial
calibration. Not part of the initial calibration to determine normal expected
variation (precision), for use in future routine validations.
Primary test method (ASTM D3764) / reference method / standard test.

Test results produced from an ASTM or other established standard test method
that are accepted as the reference measure of property or composition.
Validation. The presentation of a reference sample to an analyser of a reference

sample or by using a laboratory cross check to determine the difference between
the analyser result and the accepted reference value. To confirm that initial
calibration has not changed and is still valid.
2.4 Statistical terms
Accuracy (ASTM E456). The closeness of agreement between a test result and
its accepted reference value.
5
Analyser calibration uncertainty. The expanded uncertainty or quantification

of the cumulative errors within the measurement, calibration, and validation
processes, including the uncertainty associated with laboratory results and those
of certified standard materials expressed in the measured output units. Also, at a
stated confidence level (e.g. 95%, two standard deviations, etc.).
Analyser equipment uncertainty. The uncertainty or quantification of the

cumulative errors associated with the analyser result as stated by the equipment
supplier/manufacturer. Expressed in the measured output units, and at a stated
confidence level (e.g. 95%, two standard deviations, etc.).
Analyser precision / analyser site precision. The closeness of agreement

between a sequence of analyser results obtained using the same analyser and
same sample within a short interval of time under normal site operating
conditions. A quantitative measure of the normal expected variability associated
with the same analyser in a given location obtaining successive repeat results on
identical reference material. It can be defined as that difference between two such
single results that would only be exceeded in the long run in one case in twenty.
Equivalent to plus or minus two standard deviations or the measurement
uncertainty with a 95% confidence level.
Analyser result standard deviation / analyser standard deviation / sample

standard deviation (ASTM E456). Representation of the spread of readings
(values/results) over a limited number of observations, and how an individual
reading may differ from the average value of a much greater number of
observations. Calculated as the square root of the sum of squared differences of a
variable from its mean, or accepted reference value, divided by the number of
observations minus one.
Analyser uncertainty component. A source of error in a test result to which is

attached a particular uncertainty. The uncertainty is expressed as two standard
deviations at 95% confidence.
Average / mean (ASTM E456). An average or expected value of a

characteristic in an entire population or for all of samples/results. Sum of all of
the observed values in the population divided by the number of results in the
population.
Certified reference material uncertainty / certified reference method

uncertainty / laboratory uncertainty. The overall uncertainty or quantification
of the cumulative errors within the laboratory results and those of the certified
standard materials used in its calibration. Expressed in the measured output units
and at a stated confidence level (e.g. 95%, two standard deviations, etc.).
Confidence interval (BS EN 60359). An interval (confidence interval)

expressed as a half width (plus or minus value) with a coverage factor of two
times the uncertainty component (two standard deviations). Equivalent to a level
of confidence of 95%.
Control chart (ASTM E456) / statistical control chart. A chart on which is

plotted a statistical measure of a value versus time of sampling, along with limits
based on the statistical distribution of that measure to indicate how much, and the
type of variation that is inherent in the measurement process.
6
Control limits (ASTM E456). A set of boundaries on a control chart that are
used as criteria for signalling the need for action or judging whether a set of data
does or does not indicate a state of statistical control. If all values are within the
control chart control limits this indicates a state of statistical process control.
Normally three standard deviations above and below the control chart centre line,
equivalent to a level of confidence of 99%. Only 1 result in 100 is expected to
exceed the limit.
Deviation (BS EN 60359) / difference. / error. The difference between a

single analyser result indication and the laboratory reference test result (or
certified reference material value) when sampling an identical test material.
Analyser result minus the accepted reference value.
Estimated average / estimated mean. Average or expected value of a

characteristic in a limited number of samples/results from within a larger
population of all results. Sum of the result values divided by the number of
samples/results.
Level of confidence (ASTM E456). A value expressed as a percentage

representing the probability associated with a confidence interval. Probability of a
measured value being within stated confidence limits.
Lower control limit (ASTM E456). Minimum value boundary of the control
chart statistic that indicates statistical control.
Lower warning limit. Minimum value boundary of the control chart statistic that
indicates statistical control.
Measurement uncertainty / overall analyser uncertainty. The sum of all of

uncertainty components affecting an analyser result. Uncertainty expressed in the
units of measurement. The extent of the confidence limits. Confidence limits are
generally plus or minus one confidence interval. Generally two standard
deviations, equivalent to a level of confidence of 95%. The level of confidence
should be recorded along with any quoted uncertainty value.
Normal (Gaussian) distribution (NPL G11). Result values from individual

readings, measuring the same property or concentration, are most likely to be
near the average value of a significant number of such readings. The highest and
lowest values are the least likely to occur.
Performance action limit. Condition derived from the control chart that requires
action. A value exceeds the Upper or Lower Control limit or a series of values
exhibit a pattern that indicates drift, bias, or a particular recognised analyser fault.
Precision (ASTM E456). The closeness of agreement between independent test

results obtained under stipulated conditions. The degree of agreement between
measurements of the same property or concentration expressed in terms of the
dispersion of test results around their arithmetic mean.
Random error (ASTM E456). A part of the error that in the course of a number
of test results remains constant or varies in an unpredictable way.
Repeatability (ASTM E456) / repeatability conditions (ASTM E456).

Precision obtained where independent test results are obtained with the same
7
method on identical test items, in the same laboratory by the same operators,
using the same equipment, within short intervals of time.
Sample (ASTM E456). A group of observations or test results, taken from a

larger collection of observations or test results, which serves to provide
information that may be used as a basis for making a decision concerning the
larger collection.
Standard deviation (NPL G11(7), ASTM E456). Representation of the spread of

readings (values/results) over a number of observations, and how an individual
reading may differ from the average value of all of those readings. Calculated as
the square root of the sum of squared differences of a variable from its mean, or
accepted reference value divided by the number of observations.
State of statistical control (ASTM E456). A process condition when only

inherent random causes are operating on the process.
Statistical process control (ASTM E456). A set of techniques for improving

the quality of process output by reducing variability through the use of one or
more control charts and a corrective action strategy used to bring the process back
in to a state of statistical control.
Systematic error (ASTM E456). A part of the error that in the course of a
number of test results remains constant or varies in a predictable way.
Uncertainty (ASTM E456). An indication of the magnitude of error associated

with a value that takes into account both systematic errors and random errors
associated with the measurement or test process.
Uncertainty component (ASTM E456). A source of error in a test result to

which is attached a standard uncertainty. A standard uncertainty is expressed as
one standard deviation of the estimated value of the material being measured.
Upper control limit (ASTM E456). Maximum value boundary of the control
chart statistic that indicates statistical control.
Upper warning limit. Maximum value boundary of the control chart statistic that
indicates statistical control.
Warning limit (ASTM E456). Boundaries on a control chart that are used as
criteria for signalling the potential need for action or judging whether a set of data
indicates a change in its state of statistical control. Set beneath the control limits.
Normally two standard deviations above and below the control chart centre line.
Equivalent to a level of confidence of 95%. Only 1 result in 20 is expected to
exceed the limit.
8
3. General considerations: calibration and validation

The quality of process products is generally specified in terms of the primary test
methods/standard laboratory tests laid down by such bodies as the ISO/EN,
Energy Institute, ASTM, DIN and AFNOR. Such tests are well established, of
known precision and are widely accepted. However, product quality assurance and
control based solely on laboratory testing is now largely incompatible with the
process industry's methods of production and the scale and speed of modern
operations.
Process analysers have been developed and applied to measure continuously the
quality of material within or drawn directly from a flowing line. The benefits are
very considerable in terms of reduced sampling and testing time, improved
precision, an increase in the number of tests on each batch of product, the
opportunity for automatic control of quality, and reduction in laboratory testing
effort.
Of necessity, a process analyser differs from the standard test apparatus and even
the basic measurement technique may be different. For absolute properties made
by direct measurements (e.g. density, viscosity, etc.) the result obtained should
be independent of the type of equipment and the method used for its
measurement. Thus, on average the analyser and the standard test should give
identical results provided that due account is taken of any difference in the
precision of the two methods.
Empirical properties, made by indirect measurement (e.g. flash point and octane
number) are only definable in terms of the standard test. As the analyser is
different in some degree, it cannot produce independent meaningful values and
should be calibrated to the standard test.
Whatever the property measured and the type of analyser, users should be able to
assure themselves, and demonstrate to other interested parties that the analyser
result is the same within acceptable limits as would be obtained from the relevant
standard test.
The methodology used in this Publication is consistent with ASTM D3764 -

Standard practice for validation of the performance of process stream analyser
systems and ASTM D6299 - Applying statistical quality assurance and control
charting techniques to evaluate analytical measurement system performance(8).
In the terms of ASTM D3764 there are four key activity stages associated with
establishing the measurement performance of a process analyser:
 Initial calibration. To demonstrate that the manufacturer’s specification

for the analyser measurement system has been met. Initial calibration
accurately establishes the relationship between the analyser and the
corresponding standard test results over its full range, and includes setting
up the analyser zero and span. It may be performed at the factory or on
site, before or after installation on the plant.
 Correlation of initial calibration. To establish and apply to the

analyser’s measurement system the mathematical relationship between the
primary test method and the analyser measurement response.
9
 Probationary (initial) validation. To establish that there is agreement

between results from a further independent set of samples and the analyser
results. These are samples that were not used to establish the initial
calibration.
 Continual (on-going/routine) validation. Monitoring of the differences

between primary test method and analyser results using control chart
monitoring. Establishment of statistically based tolerances or control limits.
Validation, also referred to as in-service checking is specifically required
whilst the analyser is in service. It is to confirm that the initial calibration
has not changed and/or is valid for the particular stream being analysed at
the time. Routine validation generally involves only one point within the
analyser range. Choice of this point may be particularly important when the
analyser is used for custody-transfer measurements. Routine validation is
a scheduled procedure using control charts for monitoring the analyser's
performance over the longer term, largely to detect drift or other faults
before these become serious.
If the control limits for the statistical control chart defined for each analyser are
exceeded, a fault/equipment failure is discovered, or confidence in the analyser
results is lost, a calibration adjustment is usually carried out. The calibration is
done after the repair or other correction has been completed and the analyser is
returned to its fully working condition.
Analyser routine validation and calibration adjustment is a continuing process.

The effort necessary to achieve and sustain confidence in analyser results should
not be underestimated. It can usually only be justified by the benefits accruing
from the use made of the analyser results. The use of long term monitoring of
analyser performance may therefore only be justified for analysers used on
applications where a significant value or loss avoidance benefits are attributed to
their reliable operation.
Over the years a variety of routine validation and calibration adjustment

procedures have been evolved and successfully applied. This practice draws upon
the accumulated knowledge and experience in an attempt to provide guidance on
acceptable practices for the routine validation and calibration adjustment of
process analysers.
10
4. Preconditions for analyser calibration and validation

4.1 General
Reliable and accurate calibration is usually only achieved using analysers which:
 have been carefully selected for the duty;

 are well installed and maintained;
 are operating in accordance with specified conditions;
 are presented with a representative sample; and
 are known to be capable of stable and reliable operation.
This Publication is not directly concerned with the selection and installation of
analysers, however a summary of the key aspects that may affect validation and
calibration performance are covered in this Chapter.
4.2 Factory Acceptance Testing (FAT)
Analysers should be acceptance tested at the supplier’s factory as fully as possible

before shipment. A Factory Acceptance Test (FAT) certificate should be obtained.
Correction of faults and any modifications can most easily be done at the factory.
Factory acceptance testing gives the user initial assurance about the reliability and
quality of construction of the equipment.
The test conditions and samples should simulate as closely as practicable the
expected site conditions. Particular attention should be paid to the relationship
between the full scale range of analyser results, the expected normal process
levels, and the chosen routine validation and routine calibration levels.
Actual plant samples from the process stream should be used at the FAT if
practical. Due to difficulties in the transportation of samples and the availability to
the supplier of suitable certified reference materials or laboratory analysis, it may
be necessary to do the initial calibration and acceptance of the analyser on site at
as part of a Site Acceptance Test (SAT).
Evidence of the performance of the analyser under standard test conditions should
be obtained. In particular figures for analyser repeatability over 24 hours in terms
of the analyser result standard deviation. Details should also be recorded for the
uncertainty components/contributions of the analyser and the validation reference
materials used and any other relevant factors, e.g. temperature and atmospheric
pressure.
Rigorous calibration and correlation as defined by the ASTM practices is usually not
justified at factory due to the number and diversity of reference samples that are
required and the need for special facilities to maintain stable sample and
environmental reference conditions for long periods of time.
4.3 Site installation
General guidance regarding site installation, safety and sampling systems is

available in various publications, including EEMUA Publication 138(9) and API
RP555(10). The analyser manufacturer's recommendations should be observed,
with special attention given to the ambient temperature limits, pressure,
temperature and cleanliness of samples entering the analyser.
11
The sample presented to the analyser by the sample system should be

representative of the process material of interest and not alter in any way the
property or composition being measured.
Where possible, facilities should be provided to introduce a validation reference

material or test sample into the analyser. Facilities should also be provided for
withdrawing a sample of the stream flowing through the analyser easily and
quickly without effecting analyser operation.
4.4 Site Acceptance Testing (SAT)
Before any routine validation or calibration adjustment is performed on the plant

installation, it is essential to establish that the analyser system is operating
satisfactorily.
The analyser hardware should be confirmed as functioning to the manufacturer’s

specification. A repeat of part or all of the factory acceptance tests should be
undertaken to demonstrate that the analyser is functioning as expected. For
established installations or analyser types this may only consist of a calibration
adjustment.
The total analyser system, from sample take off to result display, should be
reliable. What constitutes adequate reliability is largely a matter of judgement by
the user. However, the system should only be regarded as satisfactory if it does
not require any kind of intervention to remain within its acceptable limits more
frequently than once per week. Examples for intervention might be viewing
operation conditions/software/alarms, adjusting flow rates, cleaning filters, etc.
4.5 Site precision
The site precision (normal expected performance) of the installed analyser should
be established over a period of a few weeks using a constant reference sample
injected periodically into the analyser or by taking laboratory cross reference
samples. The site precision is then used to determine if the analyser is beginning
to develop a bias (drift) or becoming less stable by on-going regular validation
checks using the reference sample or laboratory cross checks. This is assessed
using control charts as described in the following Sections.
4.6 Read-out instrumentation
The performance of the total analyser system should be validated through to the
operational (normally remote) display. In this context the analyser result means
the reading as it is normally displayed and acted upon such as on a control system
screen or other process operator interface.
All signal converters, input cards, data transmission, and display devices should
also be periodically checked. For example, they should be calibrated and checked
for sensitivity, deadband and linearity, digital resolution, etc., in accordance with
the manufacturer's instructions. Their contribution to the overall uncertainty
should be recorded and included in any evaluation of the overall analyser system
performance. The analyser system result remote display should also show the
result units and name of the property or composition being measured.
12
5. Principles of initial calibration, correlation and initial

validation
There are three main activity stages, which are initial calibration, correlation and
probationary (initial) validation.
For consistency with other international standards, initial calibration, correlation

and probationary (initial) validation are references to the procedures described in
ASTM D3764 - Validation of the performance of process stream analyser systems.
These may, where appropriate, also be based on ASTM D6708 - Statistical
assessment and improvement of expected agreement between two test methods
that purport to measure the same property of a material(11).
5.1 Initial calibration and correlation
For process analysers these activities are usually carried out by the equipment
suppliers as part of their product development activity. This may form part of a
factory acceptance test.
In essence, the principles of analyser validation and calibration are

straightforward. The property of interest is measured by the process analyser and
by the corresponding standard test using a test sample of the same material at a
composition level within the range of the property and types of materials of
interest. The results are then compared, and a decision on any follow up action is
taken.
However, for process analysers establishing the scientific and statistical assurance
to validate this process can require a considerable amount of organisation and
resources even if conducted remotely from the process plant.
To establish initial calibration that remains consistent across a large part of the
analyser range involves a number of different samples each of which need to be
analysed many times. This could extend to over ten significantly different samples
analysed to produce 30 results each using both laboratory and analyser
equipment. This is only likely to be justified for the most critical or important
analysers associated with high financial incentives.
Initial calibration and validation at a single point, or level specific validation is

more commonly applied using a single sample chosen to represent a point in the
analyser range close to normal analyser values or a significant process level. The
sample is continuously presented to the analyser to produce a number of results
normally over a period of 24 or 72 hours to establish the initial precision. These
results are used to produce a statistical evaluation of the analyser performance
under the chosen test conditions. The sample may also be presented to the
analyser for a shorter period such as four hours to establish the short term
reputability of the analyser.
Performance across the analyser measurement range commonly relies upon the
equipment manufacturer’s internal calibration function which compensates for any
non-linearity of the particular detection technique used. There should also be
assurance that each individual analyser produced remains consistent with the
extensive calibration and correlation testing done at the time of that particular
model’s product development.
13
Measurement techniques with non-linear calibration curves may require validation

that gives assurance that the shape of the calibration curve has not changed. This
will require more than one reference material at specified composition levels.
They may also require full scale (span) and zero reference levels.
5.2 Probationary/initial validation
For process analysers these activities are usually carried out by the owners of the
equipment during the initial period of operation of a new analyser using plant
samples or specially purchased reference materials.
This may form part of a Site Acceptance Test.
Differences between single samples with their associated accepted reference

values and analyser system results are compared for at least 25-30 separate
samples.
The data is reviewed/pre-treated and a control chart plotted. If the results on the
chart exhibit a state of statistical control the analyser is deemed to have passed
probationary validation.
See Chapter 8 for more detail on the use of statistical tools and control charts.
5.3 Analyser result and accepted reference value comparison

methods
Calibration, correlation, and validation can be done using either a reference

sample/validation reference material or paired-sample/line sample methods.
See Section 6.2 for a description and comparison of the two methods.
14
6. Principles of routine validation and calibration

adjustment
6.1 General
One of the key success factors in maintaining high process operational confidence
in process analyser results is an organised strategy for scheduled, routine
validation and calibration adjustment.
Although in essence validation and calibration are straightforward, in practice it is

often difficult to provide a traceable and well defined calibration with an overall
uncertainty fully described by its uncertainty components, at a stated level of
confidence.
The effort required to achieve this for a number of analysers is likely to be

significant and although beneficial for all analysers the techniques in this
Publication might at first only be justified for the most critical/important analysers.
The techniques should in time be extended to all analysers on a plant/site.
Analysers are fundamentally different from other process instruments in that they
measure the physical or chemical compositional properties of a material (e.g. pour
point or weight percentage) rather than measure its state (e.g. temperature or
pressure).
For some of the properties measured by process analysers there are no absolute
standards available. For these properties it is possible that the uncertainty
components associated with laboratory/primary test methods and certified
reference materials can be equal to or larger than the analyser equipment
uncertainty component as stated by the process analyser manufacturer and/or as
determined by initial validation. This means that the analyser itself, when
operating correctly, may not add significantly to the overall uncertainty of its
result.
The validation/calibration reference material calibration should preferably be the

actual process sample or a close representation of the process sample.
Depending upon the linearity of the measurement technique, calibration and

validation may require single point, multipoint or span and zero reference
levels/values.
The number of samples and composition of the samples needed for routine
validation and calibration adjustment should be determined during the initial
validation and correlation activity stages. Measurements with non-linear
calibration curves may require a routine validation and calibration adjustment that
gives an assurance that the shape of the calibration curve has not changed. This
may require more than one reference material at specified composition levels.
Multiple validation levels may need to be considered to both confirm the analyser’s
measurement performance, close to the full scale range, and also at a decision
level for the process operation, e.g. at a low alarm point.
15
Other important factors which effect routine validation and calibration adjustment
performance are as follows:
 Nature of the sample. Gas or liquid, stability/difficult to handle at

ambient conditions.
 Nature of the property. For example, bulk, specific, trace component,
etc. Some difficult to measure.
 Absolute property. Direct measurement (such as density) or empirical
property (indirect measurement, e.g. flash point). Test conditions/
equivalence to the primary test method.
 Measurement period. Continuous or intermittent (cyclic). Result
lags/delays.
 Single or multi-stream. Analysers with single or multiple measurements
per stream (e.g. single or several components by a chromatograph, single
or several measurement points per stream by a distillation analyser). Each
measurement point/stream may require a separate validation or calibration.
Cross contamination should be minimised.
 Cross sensitivity of measurement. In relation to type of sample or
other properties (e.g. effect of heavy metals on X-ray absorption
instruments).
 Measurement principle. Consider the measurement performance
compared to the primary test method.
 Adjustment. Practicality of zero and span adjustments.
 Calibration. Linear or non-linear, single point, multi point or zero/span.
 Sample composition changes. Differences during normal or plant upset
conditions.
 Validation level chosen versus full scale range. The uncertainty
component of the analyser result will be greater in the lower part of its
range. Manufacturers’ performance values are generally referenced to
measurements made at or close to full scale range (though this is not
always the case).
6.2 Analyser result and reference test/primary test method

comparison techniques
Routine validation and calibration adjustment, comparing analyser results and the
accepted reference values obtained using the reference test/primary test method
can be achieved using either the reference sample/validation reference material or
paired-sample/line sample methods.
6.2.1 Reference sample method
In the reference sample method, material is extracted from the process or

manufactured/mixed and stored. The accepted reference value for the stored
reference sample is obtained using the reference method and the results recorded
on a certificate of analysis. The stored sample is then introduced to the process
analyser. The results are then recorded and compared to the accepted reference
value.
6.2.2 Paired sample method
In the paired sample method the process analyser result is recorded at the same
time as a reference sample is collected and stored. The reference sample should
be identical to the sample flowing through the process analyser. The accepted
16
reference value is then determined for the stored sample using the reference
method and recorded on a certificate of analysis. The accepted reference value is
then compared with the process analyser result obtained at the same time as the
reference sample was taken.
See Chapter 7 for a more detailed comparison of the two methods.
6.3 Routine validation (scheduled check of continuing

performance)
Routine validation is the scheduled comparison of the process analyser result with
the accepted reference value derived from a reference method/primary test
method. Both results are obtained by sampling the same material, at different
times, using the reference sample or paired sample comparison methods described
in Chapter 7.
The performance of the analyser is determined by plotting the difference between

the analyser result and the accepted reference value. This difference is plotted
using control chart techniques. The control charts include predefined performance
warning and action limits.
If the analyser meets the normal operational performance variation defined for it
in the control chart, no adjustment should be made, even if it does not exactly
agree with the assigned reference value.
Note: Adjusting the analyser to agree exactly with the assigned reference when
not outside normal operational variation limits (site precision) makes the analyser
system less precise (more noisy).
If the analyser results comparison falls outside of its performance limits, action is
taken to identify the cause and correct it. After the completion of repairs or the
identification and solution of problems, a calibration adjustment is performed to
re-establish the analyser result within its performance limits.
Note: No assignable cause may be found due to drift.
Control chart techniques are described in Chapter 8.
6.4 Calibration adjustment (adjustment, if required, to re-establish

analyser results within performance limits)
Calibration adjustment is only done if the results of a routine validation indicate

that the difference between the analyser result and the accepted reference value
has gone beyond the performance action limits of the control chart. A significant
failure of the analyser equipment may mean that a routine validation is not
possible before completing the repair or solving the problem.
The analyser result is adjusted to be equal to the accepted reference value, or to

remove an accepted bias value, using reference method results obtained by
sampling the same material, at different times, using the reference sample or
paired sample comparison methods described in Chapter 7.
17
For optimum precision, over the long term, adjustments should only be made to
the analyser results if the routine validation result is beyond its performance action
limits.
For results that are just below performance action limits it is statistically likely that
the next result will move away from, rather than go beyond the limit. A more
detailed study of the routine validation results for previous routine validations
should be made to determine if there are any underlying problems that need to be
investigated. Some examples of this, using control charts (run rules), are shown
in Appendix 1.
6.5 Routine validation and calibration adjustment intervals
The choice of the validation intervals is important to ensure that process analyser
condition information is available on a regular basis to support the chosen
maintenance and operational strategy. The most important/critical analysers will
have a shorter interval between routine validations.
The frequency of these activities will depend upon many factors specific to each
application. These include:
 The type of analyser. Application and business need (criticality).

 Batch operation. Whether routine validation will be at regular intervals or
related to batch transfers.
 The nature of the process sample. For example, causing a known
gradual deterioration of the analyser performance.
 The nature of the reference material/sample. For example, the
stability of the reference sample material.
Routine validation at bi-weekly or greater intervals can be considered as a starting

point in the absence of any site practices/experience or statistical data.
Subsequent data may indicate that the analyser can be validated on a less
frequent basis.
For a new installation or new analyser type data may initially need to be collected
at shorter intervals, for example, on a daily or weekly basis.
Statistical tools should be used to determine if the initial routine validation interval
is appropriate. See Chapter 8 for more details.
A decision can then be made as to whether a longer interval between routine

validations is tolerable or a shorter interval is required. If the choice of a longer
interval is indicated consideration should be given to the consequences of an
undetected calibration shift between these routine validations.
As a test for the acceptability of an increased interval, for an error detected at a

routine validation/calibration that requires intervention, if there are no other
indications of when the error or fault started to occur, it should be assumed that
the erroneous results have been reported for half of the time between the previous
and current validation/calibration.
In practice, a large calibration shift is generally quickly apparent from incorrect

operation or other evidence. Examples include, signal validity checking in the
control system to look for excessive rate of change, frozen signal and signal
18
outside of the normal operating envelope. Of more concern is the small but
significant shift only detected by specific routine validation. Many analysers have
a self-diagnostic alarm feature, which is also useful to detect failures between
validations.
Automatic validations can be carried out on a frequent basis although the on

stream time of the process analyser will be reduced and usage of reference
materials may become excessive if periods of less than one day are chosen.
6.6 Number of result values required for routine validation and

calibration adjustment
The analyser results and reference method results/accepted reference values,

which together define each validation or calibration point, need to be established
to the level of accuracy required or the best achievable using the resources that
are available.
The number of results or tests conducted, on the same sample, effects how close
the average value of these results is to the true value. As the number of results
increases their average will become increasingly closer to the true value of the
property or composition being measured.
Agreement with the true value, which is the average or mean of a much larger
number of results, improves in proportion to the square root of the number of
results.
For example, if a single analyser result has an overall uncertainty of plus or minus
6 wppm (watt parts per million) at a confidence level of 95%, any single result can
be up to 6 wppm above or below (plus or minus) the true value with a 95%
confidence level. Increasing the number of samples from 1 to 9 will bring the
average of the 9 results to within plus or minus 2 wppm of the true value with a
95% confidence level. This is summarised for other multiples of results in the
following table. For a single result there is an uncertainty U1 = 6 wppm at a 95 %
confidence level.
Table 1 Sample validation uncertainty results
No. of Results (Tests) SQRT*(No. of results (n)) Uncertainty +/- (wppm)

(U1/SQRT*(n))
1 1 6.0
4 2 3.0
9 3 2.0
16 4 1.5
100 10 0.6
The result values used for comparing the analyser results and the reference
method results can be the average of a number of results to improve the
uncertainty value.
Process analyser precision can be better than that of the reference method/
standard test. In certain applications, for example, where the reference sample is
collected over a period of time, the value of a property measured with a process
analyser is the average of a large number of result readings (perhaps over a whole
production batch). In these circumstances the precision of the estimated average
19
value of the analyser is much better than its repeatability. In this case the
precision of calibration (and validation) depends largely upon the precision of the
results derived from reference method.
To meet high levels of calibration precision, or to be sure that the estimated

average of a number of analyser and reference method results are very close to
the true value, using the above principles, it is possible that a very large number
of analyser and reference test results may be needed.
In practice, reference method testing is often expensive in terms of the time and
facilities required, so the number of tests requested for routine validation and
calibration adjustment purposes needs to be limited. Stored analyser reference
samples may be limited in quantity or plant stability uncertainty. Also, access to
technician resources may constrain the time available to collect larger numbers of
analyser results.
As a compromise between the conflicting requirements of precision and effort, four

reference method results for each calibration adjustment or calibration result is
considered acceptable provided that the spread (i.e. the highest minus lowest) of
the four results is within the limits expected from the known precision of the
analyser or reference method. The arithmetic average of the four analyser results
is taken as a satisfactory estimate of the analyser result value. The arithmetic
average of the four reference method results should also be taken. The reference
method applied should be that normally used by the laboratory for the particular
process sample of interest.
6.7 Routine validation and calibration adjustment of model-

based/chemometric process analysers
In principle, model-based, chemometric, predictive, and multivariate analysers

require the same activity stages of initial calibration, correlation and probationary
(initial) validation, followed by routine validation and calibration. These are
described in ASTM D122 - Validation of the performance of multivariate process
infrared spectrophotometers(12). However, due to the complex nature of the
calibration relationship/mathematical function that maps the analyser response to
the result prediction, regularly scheduled activity is generally confined to routine
validation.
If the routine validation comparison between the analyser result and the reference
method results exceeds the control chart action limits detailed checks on the
analyser are usually required. Any faults should be rectified in order to restore the
analyser diagnostic responses to the required levels. For example, using raw
detector responses to single component reference or blank/zero level materials.
Due to the complex calibration relationships that may be present, particularly with
predictive results from one analyser that are constructed from calibrations
involving reference values obtained from more than one reference method, any
calibration adjustment is likely to be carried out by the owner of the chemometric
model and could involve significant additional processes and reference samples.
This may involve the equipment vendor, a laboratory development or research
department.
20
7. Comparison methods: reference sample method and

paired sample method
7.1 Comparing analyser and reference method results
Calibration, correlation, and validation can be done using either or both of the
following methods:
 Reference sample/validation reference material method. A sample,

which has been accurately tested by the primary test method, is injected
into the analyser. Purchased samples from specialist suppliers of certified
reference materials can also be used. Materials prepared by blending other
tested materials may be required if they use methods which allow the
calculation of a quantifiable uncertainty component.
 Paired-sample/line sample method. A number of samples from the

normal process stream being transported to the analyser are drawn off and
tested using the primary test method. See Section 7.3.2 for further details.
The reference sample method can be used with confidence on analysers that have
shown no significant effect from sample type, sensitivity or contaminants. The
principal attraction of this method is that no reference sample testing is required
at the time of the routine validation, the sample property is known precisely in
advance and the whole routine validation procedure can be automated if required.
Subject to a demonstration of sample stability, sufficient reference sample (or
several different samples) can be stored for routine validations over several
months, reducing the total reference testing effort. If reference sample injection is
automated, the routine validation facility could be made available on demand and
could if necessary be operated at the start and end of a batch transfer, for
example.
The paired-sample method should be used if the analyser is known or suspected to

be sensitive to the process sample type or its contaminants where it is not
practical to use, or the sample is not stable. This method can be automated only
to the extent that the sample could be taken automatically but it needs to be
collected and tested by the reference method/standard method.
Each method has certain advantages and limitations and the choice of which to
use, and when, depends upon several factors. For factory/works testing and for
site testing before a plant is running, the reference sample method should be
used. When the process stream is available both methods can be used to their
best advantage. Each method is dealt with more fully below.
7.2 Routine validation and calibration adjustment using the

reference sample method
7.2.1 Advantages and limitations
This method uses reference samples, which have been accurately tested by the
reference method, are injected into the analyser and the corresponding analyser
result value(s) obtained.
21
7.2.1.1 Advantages
 It is the only method applicable when the normal process stream is not
available.
 When suitably stable, the reference sample may be a stored process sample
that has been analysed and certified using the reference method.
 A reference sample may be more easily handled and/or its reference value
may be more accurately measurable than the process stream.
 With a sufficient number of reference samples, the whole range of the
analyser can be calibrated or tested at the user's convenience without the
need to wait for suitable values to occur in the process stream.
 The same reference samples can be used for initial calibration and for
subsequent routine validation and calibration adjustment. It can also be
used for a cross comparison between different analysers.
 It gives improved precision as it is related to the repeatability rather than
the reproducibility of the standard test method.
 A sample can be stored in a suitable container as part of the analyser
sample conditioning system, in the field, to facilitate manual or automatic
validation.
 The analyser calibration is directly linked to the reference method as used
in the site laboratory if a plant sample is used with an assigned reference
value determined by the site laboratory.
7.2.1.2 Limitations
 Often reference samples cannot be identical in all respects to the process

stream (condition and/or composition). For example, they can contain
traces of reactive materials like water and sulphur compounds and traces of
catalyst or other process material.
 Purchased reference sample blends may have base materials from
processes that have trace contaminants that are not present in the
equipment owner’s process. Therefore, it is best to use an actual plant
process stream sample where ever possible.
 Reference samples are normally introduced at a point that does not use the
whole of the sample transport and conditioning system. Laboratory
samples may be taken at a point in the process, which is not conditioned in
the same way.
 The analyser has to be taken off line for the duration of the validation or
calibration.
 The sample may not be stable in the long term, e.g. sulphur compounds
may adhere to the side wall, which will limit the sample life.
7.2.2 Choice of reference samples
The choice and quantity of reference sample material will be dependent on the
type of analyser. However the following points should be considered:
 Material should be stable in bulk under its storage conditions. In some

cases the range of storage temperature may be limited by the material
composition and special storage enclosures may be needed. The material
should be suitable for transfer to the analyser system without deterioration
of the measured property. This usually requires mobile liquids or gases
under pressure. Liquid mixtures with bubble points and vapours with dew
points within or close to the storage temperature range may require special
22
consideration, e.g. headspace pressurisation or heated connection lines or

heated cylinder storage. Reference samples should ideally be readily
available at reasonable cost.
 Sufficient quantity of material should be obtained to allow for fault finding
as well as routine validation and calibration. If automated validation is
used it has the potential to use more material and consideration should be
given to separate containers for automated and routine validations.
 Ideally the reference sample composition should be close to the working
composition of the process stream being sampled. However, if the process
working level is to be at the extreme of the analyser range, it is generally
not a suitable choice of level for calibration. Calibration levels should
ideally be set between 50% and 100% of the full scale range of the
measurement. Manufacturer’s performance values are normally quoted at
100% of the full scale range.
7.2.3 Testing of reference samples
Reference samples should be analysed using recognised laboratory reference

analysis methods. These should be documented and referenced on the certificates
of analysis provided for use with the reference samples. The measurement
uncertainty and confidence limit basis should be included on the certificate.
For certain types of analysers it may not be possible to obtain or produce

reference samples with a process equivalent composition containing the measured
component at a level corresponding to the analyser zero. Alternative materials
may be needed to simulate the zero level and can be particularly useful for routine
validations, e.g. the use of gas oil for viscometer zero or nitrogen for oxygen
measurements in hydrocarbon streams.
Certain reference samples may be prepared by accurate blending by weight more

precisely than by measurement using the standard test.
For plant samples, unless otherwise specified by the reference test method, the
accepted reference value for use with the reference sample method is the average
of a minimum of four valid results taken over several shifts/days.
7.3 Routine validation and calibration adjustment using the paired

sample method
7.3.1 Advantages and limitations
This method uses samples of the normal process stream, which is also being
continuously sampled by the process analyser. These are drawn off and then
tested in a laboratory using the standard reference method. The corresponding
analyser results for the time when the samples were taken and reference analysis
values are then compared and used for validation or calibration.
7.3.1.1 Advantages
 Nature of process stream. Validation or calibration takes account of any

effects on the analyser measurement arising from the nature of the actual
process stream (e.g. type sensitivity, contaminating elements, etc.).
23
 In service application. The method can always be applied to an analyser

in service, whereas the alternative procedure using reference samples may
not be practicable.
 The analyser calibration. This is directly linked to the reference method
as used in the site laboratory.
7.3.1.2 Limitations
 Process variation. Validation and calibration using this method should

only be regarded as a means of comparison of the process analyser result
to the standard reference method at each process working level at a chosen
time. The process working level may be at the extreme of the analyser
range which is generally not a suitable level for calibration. Calibration
levels should ideally be set between 50% and 100% of the full scale range
of the measurement. Manufacturer’s performance values are normally
quoted at 100% of the full scale range.
The disadvantage of validation using the paired-sample method is that it is

impossible, in principle, to obtain more than one analyser reading
corresponding to a particular sample because the process stream quality
cannot be assumed constant. In practice during normal process operations,
stream quality does not fluctuate rapidly except during plant upsets or
deliberate changes, and any validation or calibration sampling should be
timed to avoid these changing conditions.
The cross check precision is a function of the analyser and laboratory test
precision. This must be reflected in the control chart limits that are used
(based on actual cross check precision established over a period of time);
not the analyser or laboratory method precision.
 Applicability. The paired sampling method can only be applied to a

system in service where the process stream is available. Therefore,
calibration by this method cannot be used for pre-delivery testing or for
calibration before plant start-up.
 Sample transport. The sample needs to be suitable for being manually

transferred in a container to the laboratory. For example, care should be
taken that temperature changes do not alter the sample phase, e.g.
vaporising liquids or condensing vapours. Where needed, container
materials should be treated or cleaned to minimise any effects they may
have on the sample, e.g. surface absorption of low level sulphur
compounds, moisture, reaction with contamination from previous samples
etc. Refer to current ASTM, ISO standards and EI guides for correct
sampling procedures, and refer to associated laboratory test methods for
any specific sampling requirements.
 Requirement for reference method testing. Use of the reference test

method is required whenever this method is applied.
 Access to Laboratory Analysis. For this method to operate, the process

and process analyser teams need to have access to the services of a
suitably equipped laboratory which has sufficient resources to provide
reference analysis on a routine basis and occasionally at short notice to
support maintenance troubleshooting.
24
 Special Cases. There are certain types of analysers, for example those
measuring water in gases, for which the standard method involves
measurement or special sample collection techniques over a long period.
This can be many hours and generally much longer than one analyser time
constant. The property being measured is very likely to change over this
period. In this special case the analyser value should be the average over
the whole period of the standard test and only one standard test result is
practicable for each routine validation. The procedure should be carried out
a number of times to achieve sufficient points for confident calibration and
routine validation. The methods described in this Publication do not cover
such special cases, and users, if in doubt, are advised to discuss the details
with the equipment supplier.
7.3.2 Collecting paired sample results
Sampling should be conducted when the process plant operation and the process
analyser are performing normally. In other words, the operating parameters and
values (sample and service temperatures, pressures and flows) are substantially
those which were used for the initial calibration where this is appropriate.
A time should be chosen when the analyser result readings and other evidence
indicate that the plant is reasonably steady and is expected to remain so for the
duration of the sampling and data collection period.
The time constant/cycle time for process analyser systems is generally between 1
minute and 30 minutes. In practice the analyser readings before and after the one
reading taken as the analyser result value for validation will serve to show the
steadiness of the process and the process analyser during the test period. This
supporting evidence improves confidence in the analyser results obtained.
If a non-cyclical analyser result is normally 'noisy' (i.e. fluctuating more rapidly

than the analyser time constant) then the analyser value is taken as the average
reading over one time constant.
A sufficient quantity of the process stream should be drawn as quickly as possible

from a point that provides a sample representative of the analyser result. Drawing
the sample should not disturb the analyser operation in any way. The sample
should be drawn in such a way, and into a suitable container to ensure no change
in the property to be measured. The analyser result reading corresponding with
each sample is taken at the same time taking into account of any sample transport
and analysis time lag. For cyclic analysers, the reference sample should be drawn
when the analyser is about to take its sample.
The reference sample container should be carefully labelled with analyser identity,
date and time of sampling and any other relevant data, including the plant and
analyser operating conditions. If being used for calibration adjustment, the
accepted value for the paired sample method should be taken from a minimum of
four results over different days. The manual sample point should be close to the
analyser sample take off point or close to the sample transport system.
Note: Occasionally in a laboratory test, careful consideration needs to be given to

which properties you are trying to determine and any effect pre-treatment may
have.
25
26
8. Evaluation of routine validation and calibration

adjustment results: use of statistical tools and
control charts
8.1 General
Once a process analyser is in routine service, its performance should be monitored

on a scheduled basis to confirm its continuing satisfactory operation and to detect
any drift from the initial calibration. The effort required for proper routine
validation should never be underestimated, and failure to provide sufficient
attention to this is likely to lead to a loss in confidence in the analyser results and
a degradation of the associated process benefits or mitigation of process loss
provided by the analyser.
A record of the routine validations that define the performance of an analyser is of

value for maintenance optimisation and is essential to ensure compliance with
Quality Assurance and any regulatory requirements.
The guidance in this Chapter applies to both the reference sample method and the
paired-sample method. The chosen evaluation tools should be further adapted to
suit the user's particular needs and circumstances.
8.2 Use of statistical tools
Numerical statistical tools are available for the analysis of analyser validation
results or to determine if the analyser is still in control and/or the calibration has
drifted. These techniques are applied directly to the data values obtained during
routine validation.
This Publication presents one simplified control chart technique using the principles
of these tools to track process analyser performance.
Other more rigorous techniques can be applied if required. Examples of which are
detailed in standards such as ASTM D6299 - Applying statistical quality assurance
and control charting techniques to evaluate analytical measurement system
performance(8).
8.3 Routine validation and calibration adjustment: use of control

charts
The control chart is an application of Statistical Process Control (SPC). This

approach employs a graphical indication of the difference between the analyser
validation result, the centre line of the control chart (the accepted reference
value), and the relevant control chart warning and control limits.
The purpose of the control chart is to monitor or track minor differences due to
normal expected deviations, which do not require recalibration around the centre
line of the control chart to occur without adjusting the calibration, e.g. span and
zero settings of the analyser.
The causes of the differences, e.g. reference variation, sample handling variations,
temperature, and pressure may cycle naturally and may not drive the difference
outside of the control limits. The control chart limits include these natural
27
variations. Indeed, unnecessary adjustment makes the analyser results more

unstable and more likely to exceed its control limits leading to unnecessary work.
It is usual to set two control chart limits against which different actions can be
taken. Guidance on some of the methods used for setting these limits is covered
in Section 8.4.
The limit closest to the aim line is the ‘warning limit’. The upper warning limit is
the positive side of the aim line and the lower warning limit is to the negative side.
In general when a difference is outside this limit it is a prompt that the analyser
may need calibration adjustment or adjustment in the future after further routine
validations. A shorter routine validation interval may be put in place until the
reason for any change in the analyser performance is understood.
The limit furthest from the aim line is the ‘control limit’. The upper control limit is
the positive side of the aim line and the lower control limit is to the negative side.
In general, when the difference exceeds the control limit it is an indication of
possible system malfunction. It may be required after an investigation, to
establish the reason for the difference and the repair of any analyser equipment
fault.
Only over a number of routine validations, which may be several weeks or months,
can a pattern of results be obtained that will indicate any bias or drift from within
the normal scatter of results. This scatter is inherent within the uncertainties of
the analyser results and the reference method results.
An example of a control chart is shown below in Figure 1:
Difference
- Analyser +ve (upper) control Limit
Result
minus +ve (upper) warning Limit
reference
(Lab)
result
-ve (lower) warning Limit
-ve (lower) control Limit
Figure 1 Example of a control charting diagram
Where bias is encountered, the offset of the difference should remain constant
unless drift is also present.
For bias correction, the control chart aim line can then be re-defined or the
analyser results given an appropriate offset to account for the bias. No further
action is needed if the differences remain within the new control chart control
limits.
28
Examples of interpretation of control chart readings are shown in Appendix 1. An

example of determining control chart limits by measuring standard deviations is
outlined in Appendix 2.
8.3.1 CUSUM approach
Another variant of a control chart uses a statistical method based on a cumulative

calculation of difference. The technique is known as CUSUM and requires two
independent measurements of the same sample repeated over a period of time but
not necessarily on a regular basis. This can be achieved for on-line analysers by
routine reference or laboratory samples with the associated analyser results
recorded at the time of collection for later comparison to the reference
method/laboratory analysis of those samples.
CUSUM is defined as the cumulative sum of the differences between the reference
method/laboratory and the analyser results. The difference is then added to a
running total to form an updated running total.
CUSUM(n) = CUSUM(n -1) + L(n) -A(n)
Where
n = current time paired value
n-1 = previous time paired value
L(n) = current reference laboratory result or identified reference sample certified
value
A(n) = current analyser result
CUSUM(n) = cumulative sum (current value)
CUSUM(n-1) = cumulative sum (previous value)
Results are plotted as shown on the following diagram, Figure 2:
difference
mean line plot
Figure 2 Examples of analyser results in a CUSUM chart.
The horizontal axis is related to time, i.e. the results are plotted in chronological
order. However, the interval between the samples is plotted as a constant
irrespective of the actual time interval.
29
The slope of the plot is the average bias between the analyser result and the
reference method/laboratory result, or reference material certified value.
Interpretation of the plot is based on the following:
 The slope is the bias (+ve or -ve).

 A horizontal plot indicates zero bias.
 A vertical discontinuity, but same slope either side, indicates a suspect
reference method/laboratory result or reference material certified value.
 The breadth of the scatter between points about the mean line plot
indicates the uncertainty of analyser results, reference
method/laboratory or reference material certified value comparisons.
This scatter can have warning limits and control limits set using the
same methods as the control chart.
Action limits using CUSUM would include the maximum change in slope (bias)
permissible and what to do if the limits are exceeded.
Control charting can be implemented on automatically collected data and

contained in a suitable plant information system that accesses laboratory, analyser
and process data.
8.4 Control chart limits
Control chart warning and control limits should be based on the actual
performance data of the analyser under plant operating conditions (site precision
conditions) for the reference sample method or the actual cross check performance
data with the reference method/standard test/laboratory test when the paired
sample method is used.
Using other data, e.g. short term repeatability data from the vendor or the
laboratory method, reproducibility may result in control limits that are either too
tight or too wide. It is important that the on-going validations are assessed
against the actual long term performance of the analyser as installed at site.
However the other data may be used to initially set control chart limits when the
analyser is first installed before sufficient operational data is obtained to calculate
the actual true in-plant operational performance limits, or if the analyser
incentives do not justify the work to develop the true operational performance
limits.
The control chart warning limits are normally set, in statistical terms, equal to two
standard deviations of the mean of the difference between the analyser results and
the accepted reference value. This is equivalent to uncertainties quoted at 95%
confidence limits, i.e. only one difference in twenty is expected to fall outside the
warning limit under normal circumstances. Warning limits can be calculated as the
control limit multiplied by 0.75.
The control chart control limits (CL) are normally set, in statistical terms, equal to
three standard deviations, as defined above, which is equivalent to uncertainties
with 99% confidence limits, i.e. only one difference in 100 is expected to fall
outside the control limit under normal circumstances. These limits can be set from
the ‘warning limit’ values multiplied by 1.5.
30
Once the initial settings of the control limits have been determined, verification
routine validations will be required to establish whether the analyser is within
limits (or in control). The frequency of on-going planned maintenance routine
validations should then be decided by interpretation of the results.
8.5 Control chart limits derived from analyser reference sample or

paired sample laboratory data
The standard deviation and hence chart limits should be determined in operation
by initially calibrating the analyser and then performing a minimum of 25-30
routine validations with the reference sample or paired-sample cross checks with
the laboratory without further calibration adjustments. The validations should be
taken over a period of a few weeks at different times of the day and with different
technicians if possible to ensure they capture the normal analyser variations in
plant operating conditions.
When 25-30 routine validations have been completed the calculation of the
estimated standard deviation can be performed. The data should be checked by
initially plotting in time sequence to ensure there are no anomalies or significant
drift during the period the data was collected. The data should then also be
statistically checked for outliers which should be removed from the data set and
for a normal distribution (see ASTM D6299) to confirm the variations are truly
random. The standard deviation can then be calculated using the root mean
square technique and the warning and control limits can then be derived.
The control chart limits should be regularly reviewed and adjusted after every 25-
30 data points are obtained. Twenty five to thirty results are the minimum
considered to be statistically relevant.
31
32
Appendix 1: Example interpretation of control chart

readings
CL = Control Limit WL = Warning Limit
This Chart pattern is ideal showing

Difference: analyser – reference differences distributed above and below
CL the aim line within the warning limits.
Cause: successful system

W
X X X Action: None
Aim Line Time
X X X
W X
CL
Figure 3 Example of accurately distributed control chart reading
This Chart pattern has 6 or more readings

Difference: analyser – reference on one side between the Warning Limit
CL and the Aim Line.
Cause: zero shift/bias indicated

W
Action: adjust zero
Aim Line Time
X X X X
X X X
W
CL
Figure 4 Example of biased control chart reading
This Chart pattern has a general drift

Difference: analyser – reference
towards a warning limit.
CL
Cause: Possible problem with system,
changing zero or span error
W X X
X X Action: Calibration check and monitor until
Aim Line Time fault found.
X X
X
W
CL
Figure 5 Example of drifting control chart reading
33
This Chart pattern has one value outside a

Difference: analyser – reference warning limit.
CL
X Cause: possible sign of a developing problem
with the system.
W
X X X Action: take further reading.
Aim Line Time
X X
X
W
CL
Figure 6 Example control chart reading; value outside warning limit
34
Appendix 2: Determination of control chart limits by

measuring standard deviations of differences
The standard deviation is a measure of the ‘deviation’ of the values of the
individual data points as measured by the analyser compared to the mean value of
the reference measurement. The ‘warning limit’ is then taken as twice the
standard deviation and the ‘control limit’ is taken as three times the standard
deviation.
For example, the control chart limits for a distillation analyser measuring 90%
point would be determined as follows:
Table 2 Example of the control chart limits for a distillation analyser
Result Time Reference Analyser result Difference/ Difference squared

index value deviation or deviation squared
(analyser-
reference)
(°C) (°C) (°C) (°C)
1 0800 144 140.5 -3.5 12.25

2 1200 149 149.5 0.5 0.25
3 1600 146 144.5 -1.5 2.25
4 2000 152 154 2 4
5 0000 154 160.5 6.5 42.25
6 0400 146 144.5 -1.5 2.25
7 0800 148 147 -1 1
8 1200 148 146.5 -1.5 2.25
No. of Mean Sum of Sum of differences

results differences squared
(sum of values/
number of
values)
8 148.4 0.0 66.5
The estimated standard deviation of the sample data is calculated as
(square root of the sum of differences squared, divided by the number of results minus 1)
=SQRT( 66.5/(8-1)) = SQRT(9.5) = estimated standard deviation 3.1
The estimated standard deviation can be used to set the control chart limits
The ‘Warning Limit’ (WL) is equal to 2 x 3.1 = 6.2 °C

The ‘Control Limit’ (CL) is equal to 3 x 3.1 = 9.3 °C
The sum of differences in the table is zero which indicates the analyser is not
displaying any particular bias therefore the ‘aim line’ of the control chart will
represent zero deviation between analyser and reference and no offset adjustment
is needed.
It must be noted that these control chart limits will differ from the limits calculated
from standard reference test uncertainties and analyser system uncertainties as
35
the measured data used includes an element of sample handling and the fact that
the process itself is not steady therefore any non-linearity in the analyser will also
be included. This usually results in wider limits being set than strictly necessary.
The above distillation analyser example would produce a control chart as shown
below.
Note: this example is based on only eight data points but normally the standard
deviation should be calculated based on a minimum of 25-30 data points.
CL = Control Limit WL = Warning Limit
Difference: analyser – reference

+ 9.3 Deg C
CL (3 x ESDev)
X
WL
+ 6.1 Deg C
(2 x ESDev)
X
Aim line Time
X 0 Deg C
X X
X X X
WL +
- 6.1 Deg C
(-2 x ESDev)
CL
- 9.3 Deg C
(-3 x ESDev)
Figure 7 Example determination of control chart limits by measuring standard

deviations
36
References and Bibliography

References
1. BS EN 14181 - Stationary source emissions: quality assurance of automated

measuring systems, British Standards Institution, 2004.
2. BS EN 60359 - Electrical and electronic measurement equipment expression of

performance, British Standards Institution, 2002.
3. ASTM D3764 - Standard practice for validation of the performance of process

stream analyser systems, ASTM International, 2009.
4. IEC 60050-311 - Electrical and electronic measurements – general terms

relating to measurements, International Electrotechnical Commission, 2001.
5. ASTM E456 - Standard terminology relating to quality and statistics, American

Society for Testing and Materials, 2012.
6. BS EN 61207 - Expression of performance of gas analysers, British Standards

Institution, 2010.
7. NPL G11, NPL measurement good practice guide 11 (Issue 2): a beginner’s
guide to uncertainty of measurement, National Physical Laboratory, 2001.
8. ASTM D6299 - Applying statistical quality assurance and control charting

techniques to evaluate analytical measurement system performance, American
Society for Testing and Materials, 2010.
9. EEMUA Publication 138 Design and installation of on-line analyser systems, 2nd
Edition, Engineering and Equipment Materials Users' Association, 2010, ISBN
978 0 85931 174 8.
10. API RP555 - Process analyzers, Third Edition, American Petroleum Institute,
2013.
11. ASTM D6708 - Standard practice for statistical assessment and improvement of
expected agreement between two test methods that purport to measure the
same property of a material, American Society for Testing and Materials, 2008.
12. ASTM D6122 - Validation of the performance of multivariate process infrared

spectrophotometers, American Society for Testing and Materials, 2010.
Bibliography
1. EEMUA Publication 226 Design and installation of on-line analyser systems - a

guide to technical enquiry and bid evaluation, Engineering and Equipment
Materials Users' Association, 2012, ISBN 978 0 85931 184 7.
2. EEMUA Publication 187 Analyser systems - a guide to maintenance

management, the Engineering and Equipment Materials Users' Association,
2000, ISBN 0 85931 137 6.
37
3. BS EN 60746 - Expression of performance of electrochemical analysers, British

Standards Institution, 2003.
4. Technical guidance Note M20: 2012, Quality assurance of continuous emissions

monitoring systems: Application of EN 14181 and BS EN 13284-2, Environment
Agency, 2012.
38
EEMUA Publication: feedback form
An electronic version of this form is available at www.eemua.org/publications.htm
Date feedback provided: EEMUA Publication: 175, Edition 2

(dd/mm/yyyy) EEMUA Code of Practice for the calibration and
validation of process analysers
Feedback provided by: Feedback provided on behalf of which organisation?
Contact name:
Phone: Organisation name:
E-mail: Location:
EEMUA Page Paragraph Type of COMMENTS Proposed change OBSERVATIONS

ref no. Figure/ comment (Please provide alternative OF THE
(EEMUA Table (General/ wording, or propose removal EXECUTIVE
use concerned Technical/ of, or addition to, existing (EEMUA use
only) Editorial) wording.) only) on each
comment
submitted
4 Fig 2 E This is an example No change to Example
only. wording required. only
When you have completed this form with all your comments, please save and e-
mail to info@eemua.org or post a paper copy to: Publications Department,
EEMUA, 63 Mark Lane, London EC3R 7NQ.
EEMUA Publications Catalogue
All EEMUA publications can be purchased on-line. To order a publication, please
contact EEMUA at www.eemua.org, telephone: +44 (0)20 7488 0801, fax: +44
(0)20 7488 3499 or e-mail: sales@eemua.org.
Asset and integrity management
231 The Mechanical Integrity of Plant Containing Hazardous Substances: A

Guide to Periodic Examination and Testing
206 Risk Based Inspection: A Guide to Effective Use of the RBI Process
193 Recommendations for the Training, Development and Competency

Assessment of Inspection Personnel
181 A Guide to Risk Based Assessments of In-situ Large Ex 'e' and Ex 'N' Machines
161 Guide to the Selection and Assessment of Silencers and Acoustic Enclosures
159 Users' Guide to the Inspection, Maintenance and Repair of Above-ground

Vertical Cylindrical Steel Storage Tanks
148 Reliability Specification - Model clauses for inclusion in purchasing

specifications for equipment items and packages
Automation and electrical equipment
226 Design and installation of on-line analyser systems - a guide to technical

enquiry and bid evaluation
222 Guide to the Application of IEC 61511 to Safety Instrumented Systems in

the UK Process Industries
214 Toolbox Guide - Electrical Installation, Inspection and Maintenance in

Potentially Explosive Atmospheres
201 Process Plant Control Desks Utilising Human-Computer Interfaces - A Guide

to Design, Operational and Human Interface Issues
191 Alarm Systems - A Guide to Design, Management and Procurement
189 A Guide to Fieldbus Application for the Process Industry
187 Analyser Systems - A Guide to Maintenance Management
186 A Practitioner's Handbook - Electrical Installation, Inspection and

Maintenance in Potentially Explosive Atmospheres
178 A Design Guide for the Electrical Safety of Instruments, Instrument/Control

Panels and Control Systems
175 Code of Practice for Calibration and Validation of Process Analysers

155 Standard Test Method for Comparative Performance of Flammable Gas
Detectors against Poisoning
138 Design and Installation of On-Line Analyser Systems
Materials selection, plant design and construction
224 Quality assurance of safety-critical engineering equipment and materials:

A guide to risk-based procurement
218 Quality requirements for the manufacture/supply of duplex stainless steels
197 Specification for the Fabrication of Non-Primary Structural Steelwork for

Offshore Installations
194 Guidelines for Materials Selection and Corrosion Control for Subsea Oil and
Gas Production Equipment
185 Guide for Hot Tapping on Piping and other Equipment
182 Specification for Integral Block and Bleed Valve Manifolds for Direct
Connection to Pipework
179 A Working Guide for Carbon Steel Equipment in Wet H2S Service
176 Specification for Structural Castings for Use Offshore
158 Construction Specification for Fixed Offshore Structures in the North Sea
154 Guidance to Owners on Demolition of Vertical Cylindrical Steel Storage

Tanks and Storage Spheres
149 Code of Practice for the Identification and Checking of Materials of

Construction in Pressure Systems in Process Plants
147 Recommendations for the Design and Construction of Refrigerated Liquefied

Gas Storage Tanks
146 90/10 Copper Nickel Alloy Piping for Offshore Applications - Specification:
Fittings
Flanges Composite and Solid
Tubes Seamless and Welded
143 Recommendations for Tube End Welding: Tubular Heat Transfer Equipment,
Part 1 - Ferrous Materials
141 Guide to the Use of Noise Procedure Specification
140 Noise Procedure Specification

133 Specification for Underground Armoured Cable Protected against Solvent
Penetration and Corrosive Attack
105 Factory Stairways, Ladders and Handrails (incl: Access Platforms/Ramps)
104 Noise: A Guide to Information required from Equipment Vendors
101 Lifting Points - A Design Guide
Pressure equipment
223 Pressure equipment testing after repair, modifications or re-rating: a guide

to the pressure test waiver
211 Guidance on the specification of pressure vessels
208 Guide to life-cycle management of Pressure Relief Systems
204 Piping and the European Pressure Equipment Directive: Guidance for Plant
Owners/Operators
196 Valve Purchasers' Guide to the European Pressure Equipment Directive
184 Guide to the Isolation of Pressure Relieving Devices
177 Guide to the UK Pressure Systems Safety Regulations (2000)
168 A Guide to the Pressure Testing of In-Service Pressurised Equipment
149 Code of Practice for the Identification and Checking of Materials of

Construction in Pressure Systems in Process Plants
Process pipework, valves and machinery
215 Industrial Cooling Tower Fans and Fin Fans – Guide for Design,
Maintenance and Operation
205 Guide to the Production Testing of Valves for the Energy, Process, Oil and
Gas Industries
204 Piping and the European Pressure Equipment Directive: Guidance for Plant
Owners/Operators
200 Guide to the Specification, Installation and Maintenance of Spring Supports

for Piping
199 On-Line Leak Sealing of Piping - Guide to Safety Considerations
196 Valve Purchasers' Guide to the European Pressure Equipment Directive
192 Guide for the Procurement of Valves for Low Temperature (non-cryogenic) Service
188 Guide for Establishing Operating Periods of Safety Valves

185 Guide for Hot Tapping on Piping and other Equipment
182 Specification for Integral Block and Bleed Valve Manifolds for Direct
Connection to Pipework
164 Seal-less Centrifugal Pumps: Class 1
151 Liquid Ring Vacuum Pumps and Compressors
Storage and containment systems
217 Safe and effective operation of storage tanks for oil movements
213 Emission reduction from oil storage tanks and loading operations
207 Double Concrete Tanks for Liquefied Gas: Guidelines on design,

construction and operation
190 Guide for the Design, Construction and Use of Mounded Horizontal
Cylindrical Bulk Storage Vessels for Pressurised LPG at Ambient
Temperatures
183 Guide for the Prevention of Bottom Leakage from Vertical Cylindrical Steel
Storage Tanks
180 Frangible Roof Joints for Fixed Roof Storage Tanks: Guide for Designers and
Users
159 Users' Guide to the Inspection, Maintenance and Repair of Above-ground

Vertical Cylindrical Steel Storage Tanks
154 Guidance to Owners on Demolition of Vertical Cylindrical Steel Storage

Tanks and Storage Spheres
147 Recommendations for the Design and Construction of Refrigerated Liquefied

Gas Storage Tank

EEMUA 175 - Code of Practice, Cal &amp; Valid Process Analysers - 2013

Uploaded by

Document Information

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

EEMUA 175 - Code of Practice, Cal &amp; Valid Process Analysers - 2013

Uploaded by

Copyright:

Available Formats

THE ENGINEERING EQUIPMENT AND MATERIALS USERS' ASSOCIATION

EEMUA Code of Practice

Calibration and validation

Copyright © 2013 The Engineering Equipment and Materials Users' Association.

ISBN 978 0 85931 197 7

Registered and Trading Address

ENGINEERING EQUIPMENT AND MATERIALS USERS' ASSOCIATION

The Engineering Equipment and Materials Users’ Association is usually known by

EEMUA aims to improve the safety, environmental and operating performance of

– Providing the organisation within which networking, information sharing and

Incorporation, Membership and Other Information

To enquire about corporate Membership, write to enquiries@eemua.org or call +44

ABOUT THIS PUBLICATION

Consultation and Feedback

EEMUA encourages constructive comments on this Publication from both Members

Comments should be sent on the standard feedback form, a copy of which is

Other associated EEMUA publications are:

EEMUA 138 - Design and installation of on-line analyser systems.

EEMUA 226 - Design and installation of on-line analyser systems - a guide to

EEMUA 187 - Analyser systems - a guide to maintenance management.

1. Scope: calibration and validation

This Publication is not intended to cover Continuous Emissions Monitoring Systems

This Publication is not intended to cover automatic or semi-automatic instruments

For consistency with BS EN 60359 - Electrical and electronic measurement

The definitions are grouped by a usage category (additional external references

Analyser / on-line analyser / process analyser / process stream analyser

Analyser system / total analyser system (ASTM D3764(3)). The complete

Readout instrumentation. The device used to accept the signal transmitted

Absolute property. Where the measurement result is independent of the type of

Analyser range / measuring range / range (IEC 60050-311(4)). The

Analyser result / analyser system result (ASTM D3764)/ result/ result

units of that property or concentration displayed or recorded by the analyser and

Direct measurement (BS EN 60359). The value of the analyser result is

Drift (BS EN 61207-1(6), IEC 60050-311). Change in the indication of a

Empirical property. Where the measurement is only definable in terms of the

Indirect measurement (BS EN 60359). The value of the analyser result is

Linearity (IEC 60050-311). The ability of a measuring instrument to provide an

Property. The characteristic or compositional parameter that is required to be

Relative sensitivity (IEC 60050-311). Change in the indication divided by the

2.3 Calibration and validation

Accepted reference value (ASTM D3764) / reference / reference test /

consensus value based on collaborative experimental work under the auspices of a

Analyser calibration / calibration (ASTM E456, BS EN 60359). The process,

Calibration adjustment. The process of adjusting the results of an analyser to

Certified reference material / reference sample / validation reference

Checking / continual validation (ASTM D3764) / on-going validation /

Correlation (ASTM D3764). Derivation of the mathematical relationship

Initial calibration. Statistically based calibration to establish the relationship

Initial validation / probationary validation (ASTM D3764). Validation

Primary test method (ASTM D3764) / reference method / standard test.

Validation. The presentation of a reference sample to an analyser of a reference

2.4 Statistical terms

Analyser calibration uncertainty. The expanded uncertainty or quantification

Analyser equipment uncertainty. The uncertainty or quantification of the

Analyser precision / analyser site precision. The closeness of agreement

Analyser result standard deviation / analyser standard deviation / sample

Analyser uncertainty component. A source of error in a test result to which is

Average / mean (ASTM E456). An average or expected value of a

Certified reference material uncertainty / certified reference method

Confidence interval (BS EN 60359). An interval (confidence interval)

Control chart (ASTM E456) / statistical control chart. A chart on which is

Deviation (BS EN 60359) / difference. / error. The difference between a

Estimated average / estimated mean. Average or expected value of a

Level of confidence (ASTM E456). A value expressed as a percentage

EEMUA 175 - Code of Practice, Cal & Valid Process Analysers - 2013

EEMUA 175 - Code of Practice, Cal & Valid Process Analysers - 2013