GP 31-02 25 April 2005

Document No.
GP 31-02
Applicability Group
Date 25 April 2005
Guidance on Practice for

Analyser Validation
GP 31-02
BP GROUP
ENGINEERING TECHNICAL PRACTICES
25 April 2005 GP 31-02
Guidance on Practice for Analyser Validation
Foreword
This is the first issue of Engineering Technical Practice (ETP) GP 31-02. This Guidance on Practice
(GP) is based on parts of heritage documents from the merged BP companies as follows:
Amoco (ACES)
A AR-FBD-PA-P Architectural—Facility Buildings—Process Analyser Shelters—Supply
Specification
A PC-PA-00-E Process Control—Process Analysers—Engineering Specification
A PC-PA-00-G Process Control—Process Analysers—Guide
A PC-PA-00-P Process Control—Process Analysers—Supply Specification
ARCO (APCES)
Std 312-92 Analysers
British Petroleum (RPSE)

RP 30-02 Instrumentation and Control, Selection and use of Measurement
Instrumentation
GES RP30-2/4 Analyser Sampling System Specification
GEMS/31-001 Guide to Analyser and Instrument Cabinet Design
Copyright  2005,
2004, BP Group. All rights reserved. The information contained in this
document is subject to the terms and conditions of the agreement or contract under which
the document was supplied to the recipient’s organization. None of the information
contained in this document shall be disclosed outside the recipient’s own organization
without the prior written permission of Manager,
Director ofStandards,
Engineering,
BP Group,
BP Group,
unless
unless
the terms
the of
such agreement
terms of such agreement
or contract
or expressly
contract expressly
allow. allow.
Downloaded Date: 6/17/2008 10:42:53 PM

The latest update of this document is located in the BP ETP and Projects Library
Page 2 of 12
25 April 2005 GP 31-02
Table of Contents
Foreword .............................................................................................................................................. 2
1. Scope.......................................................................................................................................... 4
2. Terms and definitions ................................................................................................................. 4
3. Functional requirements of validation application...................................................................... 4
3.1. General............................................................................................................................ 4
3.2. Interfaces with other systems ......................................................................................... 5
3.3. Statistical calculations ..................................................................................................... 6
3.4. User interfaces ................................................................................................................ 6
3.5. Reports and key performance indicators........................................................................ 7
3.6. Operating systems .......................................................................................................... 7
3.7. Use of on-line analysers with validation facilities ........................................................... 8
4. People and processes associated with quality measurement................................................... 9
5. Pre-requisites.............................................................................................................................. 9
5.1. Review of on-line analysers ............................................................................................ 9
5.2. Sampling points............................................................................................................. 10
5.3. Validation of laboratory analysers................................................................................. 10
Bibliography ....................................................................................................................................... 12

Page 3 of 12
25 April 2005 GP 31-02
1. Scope
This GP provides guidance for the validation process of on-line analyser systems.
2. Terms and definitions
For the purposes of this GP (GIS), the following terms and definitions apply:
On-line analyser
Field instruments or inferential calculations that provide real-time indications of process stream
properties.
Distributed Control System (DCS)

Process control system.
Process information system is connected to the DCS and includes bi-directional communication. It
also acts as a process data historian and serves information to users and software applications.
Laboratory analyser
An instrument that is not connected to the process that can analyse discrete samples and provide
measurements that are traceable and compliant with international standards.
Inferential calculations
Calculations that utilise single or multiple on-line process measurements to estimate properties of
process streams. The calculation could include linear or non-linear models.
Process control applications use the results of on-line analysers and manipulate process variables in
real-time in order to achieve the desired process conditions and production targets.
Statistical Process Control (SPC)

A technique to analyse the numerical variance of process variables and measurements.
Validation Sample
A sample injected into an analyser that is certified by a third party but does not calibrate the analyser
i.e. calibration gas cylinder. These samples shall be traceable and compliant with international
standards.
% Analyser availability
Analyser availability = %Time Analyser is available without errors and accurate and the plant requires
it.
Adcon Utilisation
Adcon Utilisation = % Time advance control loop used / %Analyser available.
3. Functional requirements of validation application
3.1. General
a. The applications shall compare equivalent results from on-line and laboratory based
analyses and/or validation samples.
b. The application shall include facilities to ensure that the values being compared have been
adjusted to account for process, instrument and system dynamics and that sample time
Page 4 of 12
25 April 2005 GP 31-02
stamps for on-line and lab sample results and/or validation samples fall within the time
validity range.
c. The application shall execute when a new result from a laboratory analyser and/or
validation sample becomes available.
d. The application shall have the ability to incorporate samples not entered in chronological
order to the lab reporting system up to a maximum time out period.
e. Each result shall be accompanied by status signals for automatic disqualification from the
statistical analysis process and a status signal that can be manually set to allow specified
results to be disregarded and not used to validate the equivalent on-line analyser result.
f. The application shall include a statistical analysis taking into consideration ASTM/IP
standards and analyser performance criteria.
g. The application shall generate an indication of status and the current bias.
h. The current bias may be automatically calculated or manually entered. This feature shall be
an option and defined during configuration of the specific application.
i. The status and current bias shall be archived by the process information system and made
available for use by the quality measurement team to determine the root cause of any
failures in the overall quality measurement system.
The role of this team is further defined in section 4.
j. The bias and status shall allow the automatic updating of biases within process control
applications. However, the bias updating mechanism within process control schemes
should normally be a manual process.
k. Automatic updating of biases shall be considered on a per application basis.
l. Automatic updating of biases shall not be allowed to mask systematic failures in quality
measurement systems.
m. Process control applications shall allow operators to manually override biases and confirm
acceptance of calculated biases via a DCS interface.
n. Process and operation performance monitoring tools.
o. When a status is changed a text message shall be generated that describes the reason for
the change.
p. The on-line analyser status and text message shall be archived.
q. The application shall include a steady-state indicator.
r. The indicator should be set according to the status of measurements that are indicative of
the dynamic stability of the process.
3.2. Interfaces with other systems

a. Input data required shall be automatically sourced.
b. Manual data entry shall be minimised.
c. The application shall have interfaces with the following as a minimum:
1. LIMS (Laboratory Information Management System)
2. DCS
3. Process information system
4. E-mail
5. Maintenance management system (Maximo, SAP etc.).

Page 5 of 12
25 April 2005 GP 31-02
d. Numerical data entries shall be validated using standardised validation routines and
outliers removed.
3.3. Statistical calculations

a. The following statistical properties of the difference between the laboratory and/or
validation samples and on-line analysers shall be calculated:
1. Mean.
2. Standard Deviation.
3. Mean Time Between Failures for the online analyser.
4. Mean Down Time for the online analyser.
5. Bias between the online analyser and the laboratory/validation result.
6. Bias Drift between the online analyser and the laboratory/validation result.
7. Number of outliers.
8. % Analyser availability.
9. % ADCON Utilisation.
10. True Upper control limit.
11. True Lower control limit.
b. These statistics shall be archived using DCS historian or Plant Information systems and
made available for use in trends and control charts.
c. It is recommended that wide control limits (3 sigma or 99% confidence) be used to
minimize the number of outliers that are due to random fluctuations.
d. When outliers do occur they should be investigated.
e. Root causes for outliers should be identified and eliminated via a continuous improvement
process.
f. Statistical calculation methods shall be as defined in Annex A.
3.4. User interfaces

a. The user interface shall be designed and implemented to accommodate the requirements of
process operators and the quality measurement teams.
b. As a minimum the following data shall be archived in the process information system.
1. Current status of on-line analyser
2. Current raw value of on-line analyser
3. Current bias between on-line and laboratory analysers and /or validation result.
4. Current biased value of on-line analyser
5. Historical, numerical values of equivalent laboratory and /or validation result and on-
line analyser results. This data shall include sample timestamps
6. Historical trend of above including SPC limits
7. Steady state status.
c. It shall also be possible for other users to view and trend this data.
d. The different users of the data, excluding the process operators, shall be notified of system
failures via one or more of the following:
1. E-mail
Page 6 of 12
25 April 2005 GP 31-02
2. Printed reports and messages

3. Maintenance management systems.
e. Process operators shall be notified of system failures via a DCS alarm or message.
f. The process operator shall have access to data and displays via the process information
system.
g. The DCS shall include facilities to allow the operator to manually record and timestamp
when laboratory samples have been withdrawn.
h. The withdrawal of samples should be automatically detected, for example via the use of
limit switches on isolation valves.
i. The application shall be capable of informing the process operator when an analyser
measurement is set to the FAULT state. The operator may be informed via normal DCS
operating displays, alarm summary or message summary. The use of alarms should be
minimised and subjected to an alarm review.
j. The analyser team responsible for maintenance shall be informed when an analyser
measurement is set to the FAULT state.
k. The team shall be informed via normal DCS operating displays, process information
system displays and via e-mail.
l. E-mail, FAULT response shall be automatically generated by the application.
3.5. Reports and key performance indicators

a. The application shall be capable of generating reports and displays to allow users to view
current and historical data.
b. The displays and reports shall include:
1. The availability factor calculated as the percentage of time that the on-line analyser
status is good. The user shall be able to change the time-base and the system shall re-
calculate the availability on demand.
2. Current and historical bias values.
3. Current and historical standard deviations.
4. Samples rejected either by the application or manually
5. Maintenance time required to keep analyser on-line and accurate
6. Analyser Bias to laboratory and /or validation result readings
7. Analyser Bias to validation sample reading
8. Bias drift.
3.6. Operating systems

The system shall also be able to provide the following information systems:
a. Operations information system:
1. Provide operators with real-time display relating to analyser status (traffic lights).
2. Display last laboratory value and associated analyser result at time of laboratory
sample.
3. Plot trend of difference between analyser and laboratory results and /or validation
sample (control charts).

Page 7 of 12
25 April 2005 GP 31-02
4. Suggest possible error (either labs or analyser) when difference is outside control
limits.
5. Provide overview screens of all analysers (user configurable) and indication of their
status for each site area. E.g. Refinery, crude unit three.
6. Provide operators with a text description of analysis status and advise appropriate
action if difference is outside control limits.
b. Quality measurement team
1. Provide quality measurement team with all of the above plus:
2. Calculated number of outliers in the last n results
3. Calculated bias apparent in the last n results.
4. Calculated identification of bias drift.
5. Calculated identification of control limit increase/decrease.
6. Automated notification of analyser fault/inaccuracy via email.
7. Validation of both analyser and laboratory data
8. % of invalid lab and/or analyser data points
9. Calculated availability of analyser (on-line, accurate and required by process)
10. Calculated ADCON utilization while analyser is ‘ok’
11. Current and historical standard deviation of the difference between analyser and lab
results.
12. Validation of laboratory sample timings.
13. Calculated value of analyser’s mean time between failures and failure rate.
14. Calculated value of analyser’s mean down time
15. Calculated value of analyser’s mean time to repair .
16. Additional trending of process variables to identify links between process and
analyser changes.
17. Full reporting of above.
18. Fully configurable from web page to add/modify/delete analyser-lab comparisons
c. Management information system
1. All of above plus:
2. Full overview of site analysers
3. Overall reporting on availability, accuracy and reliability on an area by area basis.
3.7. Use of on-line analysers with validation facilities

Some on-line analysers, mainly gas chromatographs, include validation facilities
that can be either automatically or manually initiated.
a. The application shall allow the use of these measurements in the validation process. In the
case of gas chromatographs this will provide more meaningful results.
b. All samples must be traceable as per lab samples.

Page 8 of 12
25 April 2005 GP 31-02
4. People and processes associated with quality measurement
a. Each business unit shall include a team responsible for the maintenance of the quality
measurement systems.
b. This team shall develop and lead the implementation of plans to maintain and improve the
performance of quality measurements.
c. Each business unit shall also include a quality measurement performance team. This team
shall include the following generic functions:
1. Leader/champion
2. Laboratory specialists, (chemist and technician)
3. On-line analyser specialists (engineer and technician)
4. Operations personnel
5. Production planning
6. Process control applications
7. Product quality specialist.
d. The following processes shall be in place and championed by the quality measurement
teams:
1. Procedures governing laboratory quality control
2. Analyser validation exception investigation
3. Root cause analysis and defect elimination
4. Laboratory and on-line analyser routine maintenance and calibration schedule
5. Process for identifying laboratory and on-line analyser maintenance frequency
6. On-line analyser performance criteria – on line certification and lab cover suitability
7. Laboratory analyser routine maintenance schedule
8. Analyser selection and implementation standards
9. MOC for analyser implementation
10. Sample techniques
11. Lab testing and data entry charter – elapsed time from sampling
12. Accurate time stamping of samples
13. Analyser central knowledge management
e. The quality measurement team and processes shall ensure that biases between laboratory
and on-line analysers are not tolerated and shall be eliminated.
5. Pre-requisites
Prior to any site implementing the software applications and people processes described above
the following activities shall be completed.
5.1. Review of on-line analysers

a. The installed base of on-line analysers shall be reviewed and confirmed as being reliable
and fit for purpose. This shall include a review of user requirements to ensure that only the
desired measurements are being provided.

Page 9 of 12
25 April 2005 GP 31-02
b. The end user of each on-line analyser measurement shall be identified and their
requirements evaluated.
c. Any known defects associated with the on-line analysers shall be eliminated.
5.2. Sampling points

a. The on-line and laboratory analysers shall measure the same property from the same
sample with the same engineering units. Where this is not the case then the dynamics
between the sampling points used for the laboratory and on-line analysers shall be
identified and used to provide dynamic compensation.
b. Facilities shall be available to allow the accurate time stamping of samples to be analysed
by the laboratory. This should be achieved via:
c. Automatic detection and time-stamping of samples (eg limit switches connected to
isolation valves and interfaced to DCS)
d. Manual time stamping of samples via DCS based operator interface.
5.3. Validation of laboratory analysers

The performance of the laboratory analysers shall meet and be traceable to ASTM/IP
measurement standards.

Page 10 of 12
25 April 2005 GP 31-02
Annex A
Statistical Calculation Methods
System capability calculations are based on the last “n” samples, using the current sample,
where all samples are not outliers. An outlier is anything that fails the UCL and LCL checks and
the system capability checks. The system capability is calculated by obtaining the standard
deviation of the sample set * confidence factor. The confidence factor would typically be 3.0
but shall be configured per on-line analyser.
The Calculations required for ASTM/IP repeatability, ASTM/IP reproducibility, total standard
deviation will be stored in a database.
The bias, standard deviation, means use the last n samples (n can vary between properties)
including the current sample.
Bias = mean (last n {lab – analyser samples})
T = (abs (((mean of the last n {lab – analyser samples})-bias)) * (sq root of sample set size)) /
StDev. UCL(upper control limit) and LCL (lower control limit)calculations use the last n
samples, not including the current sample
UCL = mean (last n {lab – analyser samples}) + ((t test critical value at 95%, 2 tailed using
sample set size-1) * St Deviation). This should not be greater than the reproducibility of the
tests being compared. For physical property analysers this should not be greater than the
reproducibility of the lab test.
LCL = mean (last n {lab – analyser samples}) - ((t test critical value at 95%, 2 tailed using
sample set size-1) * St Deviation). This should not be greater than the repeatability of the tests
being compared. For physical property analysers this should not be greater than the
reproducibility of the lab test.
UCL and LCL shall be initially set at the ASTM/IP standard reproducibility for the test being
undertaken. If no ASTM/IP reproducibility’s are available for the test in question then the UCL
and LCL shall be calculated using +/- the overall tolerance for the analyser system as shown
below:
Overall Tolerance = UC/LC where
UC = CU1*CU2
LC = CL1*CL2
CL1 = 1-(T1/100)
CU1 = 1+(T1/100)
CL2 = 1-(T2/100)
CU2 = 1+(T2/100)
T1 = Analyser repeatability for the particular test/range of analysis
T2 = Calibration standard tolerance
CL1 = Lower analyser confidence
CL2 = calibration standard lower confidence
CUI = Upper analyser confidence
CU2 = Upper calibration standard confidence
Bias only updated if T > t test critical value (at 95%, 2 tailed using sample set size “n”-1) for 2
consecutive sample times. Where “n” is at least 21 samples.

Page 11 of 12
25 April 2005 GP 31-02
Bibliography
[1] ASTM D3764, Standard Practice for Validation of Process Stream Analyzer Systems
[2] ASTM D6122, Standard Practice for Validation of Multivariate Process Infrared Spectrophotometers
[3] ASTM D6299, Standard Practice for Applying Statistical Quality Assurance Techniques to Evaluate
Analytical Measurement System Performance
[4] ASTM D6300, Standard Practice for Determination of Precision and Bias Data for Use in Test
Methods for Petroleum Products and Lubricants
[5] ASTM D6621, Standard Practice for Performance Testing of Process Analysers for Aromatic
Hydrocarbon Materials
[6] ASTM E456, Standard Terminology for Relating to Quality and Statistics
[7] ASTM E1655, Standard Practices for Infrared Multivariate Quantitative Analysis

Page 12 of 12

GP 31-02 25 April 2005

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

GP 31-02 25 April 2005

Uploaded by

Copyright:

Available Formats

Document No.

Guidance on Practice for

British Petroleum (RPSE)

Downloaded Date: 6/17/2008 10:42:53 PM

Downloaded Date: 6/17/2008 10:42:53 PM

2. Terms and definitions

Distributed Control System (DCS)

Statistical Process Control (SPC)

3. Functional requirements of validation application

3.2. Interfaces with other systems

Downloaded Date: 6/17/2008 10:42:53 PM

3.3. Statistical calculations

3.4. User interfaces

2. Printed reports and messages

3.5. Reports and key performance indicators

3.6. Operating systems

Downloaded Date: 6/17/2008 10:42:53 PM

3.7. Use of on-line analysers with validation facilities

Downloaded Date: 6/17/2008 10:42:53 PM

4. People and processes associated with quality measurement

5.1. Review of on-line analysers

Downloaded Date: 6/17/2008 10:42:53 PM

5.2. Sampling points

5.3. Validation of laboratory analysers

Downloaded Date: 6/17/2008 10:42:53 PM

Downloaded Date: 6/17/2008 10:42:53 PM

Downloaded Date: 6/17/2008 10:42:53 PM

You might also like