30th International North Sea Flow Measurement Workshop

23 – 26 October 2012

Gas Chromatography – Conditional Based Monitoring and

Live Uncertainty Calculation
Anwar Sutan, i-Vigilant Technologies
Paul Daniel, i-Vigilant Technologies

1 INTRODUCTION

On-line gas chromatography is now commonplace in the North Sea and is frequently used
within Fiscal and Custody Transfer measurement systems. The uncertainty of the analysis
from the gas chromatograph (GC) is of the utmost importance with the resultant analysis
frequently at the core of economic transactions [1].

In recent years there has been a lot of interest in condition based monitoring (CBM) and in-
situ verification of measurement devices. However, perhaps surprisingly, it would be
reasonable to state that although the correct functioning of the GC is critical in today’s
measurement systems, comparatively little attention has been paid to verification and
monitoring strategies [2, 3, 4, 5].

The modern gas chromatograph is an extremely repeatable device however there remains
several fundamental issues. For example, if the valve timing in the GC is wrong or drifts over
time a systematic error can ensue. A recent study performed by the author identified a
situation where such an error was present, and if it remained undetected, would have resulted
in an on-going error in the calorific value of up to 1.4%. This error would equate to a value of
more than £300,000 per month based on a production of 100 mmscfd.

Under the assumption of suitable sampling and conditioning the uncertainty of the GC
measurement is generally driven by the repeatability of the GC and the quality of the
(certified) working reference mixture (WRM). Various methods presently exist which may be
used to obtain the GC repeatability. ISO 10723 [6] describes a method of performance
evaluation using multiple calibration gas compositions to obtain the linearity of the GC as well
as its repeatability. However, this method is only valid as long as the GC maintains the
performance characteristics measured on the day of the test. ASTM D1945 [7] provides a
standard test method for the analysis of gas with a GC using stated levels for the expected
repeatability and reproducibility. ISO 6974 [8], following similar principles detailed in ISO
10723, describes data processing for the tailored analysis of natural gas with the aim of
defining the uncertainty in the mole fractions of the components measured. However there
remains little practical guidance and, to the Authors belief, no implementation of CBM
systems for GCs.

This paper presents a novel philosophy for monitoring the health of the GC whilst also
providing an on-line estimate of the reproducibility of each mole fraction and overall
uncertainty in the Calorific Value (CV) and density. The method has been demonstrated
using data from a GC presently in use on a platform in the North Sea.

1
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

2 GC MAINTENANCE

2.1 Current Practice

Within the UK North Sea the following steps are generally taken to ensure that the
performance of the GC is maintained at a satisfactory level:

1. Periodic health check by a GC specialist.

2. ISO 10723 multilevel calibration to determine the GC linearity and measurement bias.
3. Regular calibration using a certified working reference mixture.
4. Periodic repeatability check using the tolerances provided by ASTM 1945D.

It is common for all these activities to be performed by different parties. The result is a
number of different reports that need to be combined and reviewed as a whole to obtain the
full picture of the GC performance.

The following sections provide very brief overview of the available methods and standards
presently in common use.

2.1.1 ISO 10723

ISO 10723 is a standard relating to the performance evaluation for online analytical systems.
It has gained particular prevalence following the publication of the EUETS measurement and
reporting guidelines [9] where annually repeated validation by EN ISO 17025:2005 [10]
accredited laboratory was specified.

A number of test gases are used to establish:

• The ability of the GC to measure the components specified.

• The repeatability of the measurement of each component over their specified ranges.
• The relationship between response and concentration of each component over their
specified ranges.

2.1.2 ASTM D 1945

This defines standard test method for analysis of natural gas by gas chromatography. Of
particular interest to this paper are the sections that detail repeatability and reproducibility.

Repeatability is defined within the standard as the difference between two successive results
obtained by the same operator with the same apparatus under constant operating conditions
on identical test materials. Table 1 details the defined maximum tolerances considered
acceptable for repeatability as stated in the standard.

Mol % Lookup Tolerance %

0 to 0.1 Mol% 0.01 %
0.1 to 1 Mol% 0.04 %
1 to 5 Mol% 0.07 %
5 to 10 Mol% 0.08 %
Over 10 Mol% 0.10 %

Table 1. ASTM D 1945 repeatability limits

A similar table exists for the reproducibility.

2
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

2.1.3 Other available methods

• ISO 6974 provides guidelines for the analysis of natural gas with the aim of determining
the mole fractions of the principal components. The standard is split in to five sections,
with Part 2 describing the determination of the measuring system characteristics and the
statistical approach to data handling and error calculation with the aim of defining the
uncertainties in the component mole fractions.

• ISO 6976 [11] is a standard detailing a calculation method for the determination of
calorific values, density, relative density and wobbe index from composition. Within ISO
6976 precision is expressed as a calculated physical property value that results solely
from random errors in analytical procedure in terms of repeatability and reproducibility.
The standard defines equations for both repeatability and reproducibility.

3 THE PROPOSED METHOD

The common practice of performing distinct tests by different vendors at different times has
led to the realisation that there are distinct advantages associated with collating all the
information from the GC maintenance activities in to one place, combining all the data and
results and presenting it in a clear, concise and easily interpreted manner. This information,
when used in conjunction with a number of additional novel techniques, results in a condition
based monitoring system with an expert system capability providing early detection of
potential GC failures coupled with the ability to assist with the determination of the required
course of actions to rectify any identified problems.

The proposed method consists of the following:

1. Acquisition of GC analysis parameters.

2. Analysis of events timing.
3. Analysis of individual calibration data prior to use as a foot-print.
4. Analysis of each calibration and comparison with the foot-print data.
5. Analysis of historical data.
6. Acquisition of working reference mixture uncertainty.
7. Analysis of repeatability.
8. Extended reproducibility assessment providing the basis for a live uncertainty
calculation.
9. Estimation of the bias based on the ISO 10723 performance evaluation method.

The result of this is an expert system which can provide a live estimate of the uncertainty
incorporating the uncertainty of the working reference mixture used for routine calibration, the
repeatability and the long term reproducibility. An estimate of the uncertainty of other
calculated parameters such as density and calorific value may also be determined.

Knowledge of the live operating uncertainty not only provides assurance that the
measurement is meeting the required specification, but may also be used to determine when
maintenance of the GC needs to be performed.

Steps one to five are covered by section 3.1. The remaining items are discussed with a
working example using live data from an operational GC.

3.1 Foot-printing and analysis of historical data

A previous publication in 2009 described a method to analyse three-column gas

chromatographs using the correlation between the component’s molecular weights and their
response factors [2].
®
This method lead to a condition based monitoring philosophy, GCAS , which utilises the
techniques described within the paper to ensure that the status of a GC is healthy prior to
foot-printing. Figure 1 shows a typical calibration data page.

3
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

®
Figure 1. GCAS Calibration data page

The term “foot-print”” in this context is used to mean a healthy GC status which is captured
and used as a reference against which future calibrations are compared.

Examples of the historical data analysis based on the foot-print data are:

• Response factor deviation control chart.

chart
• Retention time deviation control chart.
chart
• Analysis of the quality of the correlation of the log-log
log log plot of molecular weight and
response factor.

An example chart is shown in Figure 2,, which shows the deviation of the response factor for
methane from the foot-print.

Figure 2.. Deviation of response factor chart - Methane

4
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

3.2 Uncertainty of Working Reference

Refer Mixture

The working reference mixture uncertainty is normally provided in the UKAS certification
accompanying the mixture cylinder. In this example the working
rking reference mixture,
mixture with its
associated uncertainty, is given in Table 2.

Absolute Expanded Standard

Component WRM (%mol/mol) Uncertainty Relative Relative
(k=2) Uncertainty (%) Uncertainty
Methane 75.605 0.05 0.0661 0.03305
Nitrogen 0.5 0.01 2.0000 1
carbon dioxide 1.784 0.011 0.6166 0.3083
Ethane 12.301 0.08 0.6504 0.3252
Propane 6.938 0.022 0.3171 0.15855
i-butane 0.8926 0.005 0.5602 0.2801
n-butane 1.6218 0.0094 0.5796 0.2898
neo-pentane 0.1026 0.0022 2.1442 1.0721
iso-pentane 0.2259 0.0021 0.9296 0.4648
n-pentane 0.1973 0.0024 1.2164 0.6082
n-hexane 0.0991 0.0017 1.7154 0.8577

Table 2.. Working Reference Mixture relative uncertainty

3.3 Uncertainty from the GC repeatability

The repeatability of each component may be established as per ISO 10723 as a function of
concentration (Figure 3).. The standard deviation of the response for each component
co is
expressed as:
∗ ∗ ∗

is standard deviation
∗
a, b, c, d are the coefficients of linear regression of on
∗
is un-normalised concentration of component

Equation 1. relationship
ionship between standard deviation and concentration

Figure 3.. Repeatability of response as function of component concentration

5
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

The repeatability of each measured component is a function of the repeatability of the

response on the working reference mixture and the repeatability of the response at the
concentration being measured. The calculation of the standard deviation of the sample
mixture is described in both ISO 6974-2 and ISO 10723. The following equation is defined in
ISO 10723:

( ∗) ( ) ( )
∗ = +

and are the instrument responses to component in the sample and standard

( ) and ( ) are the respective standard deviations

∗
is the un-normalised concentration of component

( ∗ ) is the standard deviation of the un-normalised components ∗

∗
Equation 2. Standard deviation of component

Having calculated the standard deviation of each un-normalised component, the standard
deviation of the normalised mole fractions as defined by ISO 6974-2 is given:

1−2 ∗
( )= × ∗
× ( ∗) + ( ∗ )
!

is the normalised mole fraction of component

∗
is the un-normalised mole fraction of component

Equation 3. Standard deviation of the normalised component

3.3.1 Working example

Table 3 shows coefficients relating the mole fraction to standard deviation from a gas
chromatograph presently in use on a North Sea platform.

Coefficients of linear regression

Component
a b c
Methane 329875.9302 0 0
Nitrogen -534.812 6659.9222 0
CO2 24070.3 0 0
Ethane 13096.0458 5547.2357 0
Propane -9075.6688 15604.4898 0
i-Butane 6941.0164 0 0
n-Butane -2236.4662 21367.3294 0
neo-Pentane 437.4404 69901.5654 0
i-Pentane 4301.6545 41299.3758 0
n-Pentane 13469.4126 0 0
Hexane 23779.2293 0 0

Table 3. Coefficients of linear regression

6
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

Table 4 shows the standard deviation of the response of each component in the working
reference mixture and the sample gas.

WRM Std Dev of Sample Gas Std Dev of

Component
(%mol/mol) WRM (%mol/mol) sample gas
Methane 75.605 329876 83.03 329876
Nitrogen 0.5 2795 0.39 2063
CO2 1.784 24070 1.83 24070
Ethane 12.301 81333 8.56 60580
Propane 6.938 99188 4.17 55995
i-Butane 0.8926 6941 0.55 6941
n-Butane 1.6218 32417 1.09 21054
neo-Pentane 0.1026 7609 0.03 2534
i-Pentane 0.2259 13631 0.09 8019
n-Pentane 0.1973 13469 0.18 13469
Hexane 0.0991 23779 0.08 23779

Table 4. Standard deviation of working reference mixture and sample gas

The standard deviation of each component can then be calculated using Equation 2:

Std Dev of Sample Std Dev of un-normalised

Component Std Dev of WRM
Gas component
Methane 0.000487 0.000447 0.000499
Nitrogen 0.000431 0.000406 0.000003
CO2 0.000883 0.000861 0.000022
Ethane 0.000392 0.000418 0.000070
Propane 0.000664 0.000616 0.000063
i-Butane 0.000308 0.000500 0.000005
n-Butane 0.000762 0.000734 0.000017
neo-Pentane 0.003350 0.012461 0.000013
i-Pentane 0.002104 0.003136 0.000009
n-Pentane 0.002290 0.002511 0.000007
Hexane 0.007322 0.009102 0.000012

Table 5. Standard deviation of un-normalised components

Based on the standard deviation of the un-normalised component, the standard deviation of
normalised component can then be determined from Equation 3.

Std Dev of normalised

Component Relative uncertainty
component
Methane 0.000143 0.019006
Nitrogen 0.000004 0.077837
CO2 0.000023 0.131377
Ethane 0.000087 0.071210
Propane 0.000068 0.098314
i-Butane 0.000007 0.077343
n-Butane 0.000019 0.115847
neo-Pentane 0.000013 1.290049
i-Pentane 0.000009 0.380257
n-Pentane 0.000007 0.343006
Hexane 0.000012 1.168120

Table 6. Relative uncertainty of each component based on the GC repeatability

7
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

3.4 Uncertainty from GC reproducibility

GUM [12] defines reproducibility as closeness of the agreement between the results of
measurements of the same measurand carried out under changed conditions of
measurement, including time.

The reproducibility of a GC can be determined by analysing the calibration data from periodic
calibrations. The responses are converted to un-normalised mole percentages based on the
foot-print response factors and then normalised. The standard deviations of the normalised
components are then calculated to obtain the GC reproducibility.

The following presents such an analysis of the reproducibility based on the same data
provided in Table 2.

To get the reproducibility of the GC, the following steps are performed:

1. Take the foot-print data

a. Take the foot-print response factor for each component ("# ,#% )
b. Take the gas composition from the WRM certificate ( ,&'( )
2. Take periodic calibration data (i.e. daily, weekly, etc)
a. Record the response factor from calibration data ("# , ) )
b. Calculate the Peak Area of the calibration (" ,*+, ) from the response factor ("# , ) )
and the composition of working reference mixture ( ,&'( ).

", ) = "# , ) × ,&'(

3. Calculate the un-normalised components ( ∗ ) from the peak area of the calibration (" ,*+, )
and foot-print response factor

" ,*+,
∗
=
"# ,-.

4. Calculate the normalised composition ( )

∗
= × 100
/ !
∗

5. Calculate the standard deviation of each component. N is the number of calibrations that
have been performed following the foot-print calibration

∑5
3 !( ,3 − ̅ )
( )=1
6−1

( )
6. Calculate the relative uncertainty of each component from the standard deviation.
7( ) =
, 89

8
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

The following is an example of the calculation of the normalised composition from the
calibration response factor.

Foot-print Calibration Un-

WRM Calibration Normalised
Component Response Response normalised
(%mol/mol) Peak Area Composition
Factor Factor composition
methane 75.605 10139000 10045000 757427621 74.7044 75.4329
nitrogen 0.5 12710000 12584000 6275226.32 0.4937 0.4990
carbon dioxide 1.784 15179000 14990000 26670868.77 1.7571 1.7774
ethane 12.301 16616000 16432000 201591178.8 12.1324 12.2548
propane 6.938 21522000 21275000 147212451.1 6.8401 6.9150
i-butane 0.8926 24821000 24523000 21830875.87 0.8795 0.8853
n-butane 1.6218 26052000 25750000 41650019.5 1.5987 1.6131
neo-pentane 0.1026 27393000 26785000 2740814.802 0.1001 0.1016
iso-pentane 0.2259 28652000 28414000 6401611.094 0.2234 0.2246
n-pentane 0.1973 29829000 29362000 5777678.864 0.1937 0.1962
n-hexane 0.0991 32514000 32446000 3206826.752 0.0986 0.1001

Table 7. Calculation of normalised composition

The same procedure is then performed for each calibration and the standard deviation
calculated. The standard deviation and relative uncertainty of each component is shown in
Table 8.

Component Standard Deviation Relative Uncertainty

methane 0.0117 0.0155
nitrogen 0.0009 0.1857
carbon dioxide 0.0013 0.0747
ethane 0.0106 0.0861
propane 0.0080 0.1157
i-butane 0.0025 0.2757
n-butane 0.0024 0.1502
neo-pentane 0.0004 0.3456
iso-pentane 0.0006 0.2499
n-pentane 0.0006 0.3280
n-hexane 0.0002 0.2182

Table 8. Standard deviation and relative uncertainty of normalised components

9
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

3.5 Combined Uncertainty

Once the uncertainty from each uncertainty source is calculated, the combined uncertainty
can be determined.

Repeatability Reproducibility
WRM Relative Combined Relative
Component Relative Relative
Uncertainty Uncertainty
Uncertainty Uncertainty
methane 0.03305 0.0190 0.0155 0.0412
nitrogen 1 0.0778 0.1857 1.0534
carbon dioxide 0.3083 0.1314 0.0747 0.3593
ethane 0.3252 0.0712 0.0861 0.3647
propane 0.15855 0.0983 0.1157 0.2738
i-butane 0.2801 0.0773 0.2757 0.5590
n-butane 0.2898 0.1158 0.1502 0.4063
neo-pentane 1.0721 1.2900 0.3456 1.7810
iso-pentane 0.4648 0.3803 0.2499 0.7403
n-pentane 0.6082 0.3430 0.3280 0.9001
n-hexane 0.8577 1.1681 0.2182 1.4977

Table 9. Combined relative uncertainty

Based on the relative uncertainty and relative sensitivity of each component, as well as the
uncertainty from the ISO 6976 calculation [11], the combined expanded uncertainty from the
calibration gas, repeatability, reproducibility, and calculation can be determined. The
expanded uncertainty in this example for CV and standard density is shown in Table 10 and
Table 11 respectively.

:(;<) ∗ =(;<)
Expanded Standard
mole Norm Coverage Relative
Component relative relative
Fraction Mol % factor Sensitivity
uncertainty uncertainty

Methane 0.7561 0.7540 0.0823 2 0.0412 -0.1513 -0.0062

Nitrogen 0.0050 0.0050 2.1067 2 1.0534 -0.0050 -0.0053
CO2 0.0178 0.0178 0.7185 2 0.3593 -0.0178 -0.0064
Ethane 0.1230 0.1227 0.7295 2 0.3647 0.0501 0.0183
Propane 0.0694 0.0692 0.5477 2 0.2738 0.0696 0.0191
i-Butane 0.0089 0.0089 1.1180 2 0.5590 0.0142 0.0079
n-Butane 0.0162 0.0162 0.8125 2 0.4063 0.0259 0.0105
neo-Pentane 0.0010 0.0010 3.5620 2 1.7810 0.0022 0.0040
i-Pentane 0.0023 0.0023 1.4805 2 0.7403 0.0049 0.0037
n-Pentane 0.0020 0.0020 1.8003 2 0.9001 0.0043 0.0039
Hexane 0.0010 0.0010 2.9953 2 1.4977 0.0028 0.0041
ISO 6976 N/A N/A 0.1000 2 0.0500 1.0000 0.0500
Combined Standard Uncertainty 0.0595
Coverage Factor (k) 2
Combined Expanded Uncertainty (%) 0.1190
CV absolute uncertainty (MJ/m3) 0.05620

Table 10. CV combined expanded uncertainty

10
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

:(;<) ∗ =(;<
Expanded Standard
mole Norm Coverage Relative
Component relative relative
Fraction Mol % factor Sensitivity
uncertainty uncertainty
Methane 0.7561 0.7540 0.0823 2 0.0412 0.5563 0.0229
Nitrogen 0.0050 0.0050 2.1067 2 1.0534 0.0064 0.0068
CO2 0.0178 0.0178 0.7185 2 0.3593 0.0361 0.0130
Ethane 0.1230 0.1227 0.7295 2 0.3647 0.1706 0.0622
Propane 0.0694 0.0692 0.5477 2 0.2738 0.1414 0.0387
i-Butane 0.0089 0.0089 1.1180 2 0.5590 0.0240 0.0134
n-Butane 0.0162 0.0162 0.8125 2 0.4063 0.0436 0.0177
neo-Pentane 0.0010 0.0010 3.5620 2 1.7810 0.0034 0.0061
i-Pentane 0.0023 0.0023 1.4805 2 0.7403 0.0075 0.0056
n-Pentane 0.0020 0.0020 1.8003 2 0.9001 0.0066 0.0059
Hexane 0.0010 0.0010 2.9953 2 1.4977 0.0040 0.0059
ISO 6976 N/A N/A 0.1000 2 0.0500 1.0000 0.0500
Combined Standard Uncertainty 0.0961
Coverage Factor (k) 2
Combined Expanded Uncertainty (%) 0.1923
Density absolute uncertainty (Kg/m3) 0.09081

Table 11. Standard density

d - combined expanded uncertainty

3.6 Systematic Error

It is common for GCs too not implement the response functions obtained from the ISO 10723
calibration. Using a single response factor that is assumed to pass through the origin will
introduce a systematic error when the sample gas is of a different composition to the working
reference
ference mixture. The magnitude of the bias on each component can be estimated
estimat from the
response functions.

GUM states that itt is unnecessary to classify components as “random” or “systematic” (or in
any other manner) when evaluating uncertainty because all all components of uncertainty
uncertain are
treated in the same way. It further states that occasionally where a correction for a systematic
effect has not been applied to the reported result of a measurement, no attempt should be
made to take the effect into account by enlarging the “uncertainty” assigned to the result.
Only
nly in very special circumstances should corrections for known significant systematic effects
effect
not be applied to the result of a measurement. Figure 4 shows the bias introduced for
Methane in the present example when using the response factor from a single point
calibration. This highlights the importance of using a suitable working reference
nce mixture.

Figure 4.. Bias introduced when using response factor from single point calibration

11
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

The above uncertainty calculation has not taken into account the bias introduced by operating
away from the working reference mixture calibration point. It is recommended that for
scenarios where significant bias is predicted the response functions from the ISO 10723
multi-level calibration are implemented to remove the known bias. If this is not performed,
then consideration should be taken when expressing the uncertainty of the calorific value.

3.7 Expert System

The methods described above facilitate the determination of the live uncertainty of calculated
quantities such as CV and standard density, as provided in the example. Coupled with the
use of suitable control charts, this allows prescribed limits to be entered, above which
intervention is required. For example a control chart of CV uncertainty can easily be set up
which can clearly demonstrate that the system is operating within its target uncertainty. This
can also quickly alert the operator to problems and to help decide when maintenance is due.

It is expected that the use of the correlation parameters from the log-log chart combined with
control charts will be able to provide significant assistance in determining the cause and
hence solutions to any problems. At this level, the method is approaching that on an expert
system, which may loosely be defined as a computer system that emulates the decision-
making ability of a human expert.

4 CONCLUSION

This paper has described methods that may be used for monitoring the performance of gas
chromatographs. Data commonly available from the WRM certificate, ISO 10723 multilevel
calibrations, and long term reproducibility can be utilised to provide live uncertainty
calculations giving assurance that GC performance is within prescribed limits throughout the
year.

The principle is equally applicable to calculations such as line density based on

compressibility from AGA 8 [13] and any other compositional based calculations. The
application of the method is expected to help to maintain the GC within its specified
requirements throughout the year.

12
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

5 NOTATION

GC Gas Chromatograph curves of the measuring

CBM Conditional Based Monitoring system.
CV Calorific Value WRM Working Reference Mixture.
EUETS European Union Emission Mixture which is used as a
Trading Scheme working standard for regular

, , ,
k Coverage factor calibration of the measuring

regression of on ∗
Coefficients of linear system
GUM Guide to the expression of

∗
"# ,#%
standard deviation uncertainty in measurement
Un-normalised concentration of Foot-print response factor for

"# ,
component component
)
Normalised concentration of Calibration data response
( ) ∗
component
factor for component
normalised components ∗ ,&'(
Standard deviation of the un-
Mole fraction of component in

",
WRM
)
Instrument responses to
component in the sample Peak Area from calibration of
Instrument responses to component

( )
component in the standard N Number of calibrations that
Standard deviations of have been performed following

( ) ( )
component in the sample the foot-print calibration
Standard deviations of Relative sensitivity coefficient

( ) 7( )
component in the standard of component mole fraction
Standard deviation of the Relative uncertainty of

>
normalised mole fractions component mole fraction
Number of components mmscfd Million standard cubic feet per
analysed day
CRM Certified Reference Mixture.
Mixture which is used for the
determination of the response

13
 30th International North Sea Flow Measurement Workshop
23 – 26 October 2012

6 REFERENCES

[1] S. Fosse, R. Sakariassen, and F. Teknikk, “Online gas chromatography : A technical

and historical overview - design and maintenance advices to achieve an accurate end result,”
North Sea Flow Measurement Workshop, 2010.

[2] A. Sutan, C. Johnson, and J. Laidlaw, “Three columns gas chromatograph analysis
using correlation between component’s molecular weight and its response factor,” North Sea
Flow Measurement Workshop, 2009.

[3] D. Pettigrw, “Advances in on-line gas chromatography in the natural gas industry,”
North Sea Flow Measurement Workshop, 2000.

[4] C. Cowper and R. Mounce, “Evaluation of on-line chromatograph performance,”

North Sea Flow Measurement Workshop, 1997.

[5] M. Crane, R. Beaty, and W. Lake, “Calibration of gas chromatagraph for improved
performance,” North Sea Flow Measurement Workshop, 1995.

[6] BS EN ISO 10723, “Natural gas - performance evaluation for on-line analytical
systems,” BS EN ISO, 2002.

[7] ASTM D1945, “Standard test method for analysis of natural gas by gas
chromatography,” ASTM, 2001.

[8] BS EN ISO 6974-2, “Natural gas - Determination of composition with defined

uncertainty by gas chromatography - Part 2: Measuring-system characteristics and statistics
for processing of data,” BS EN ISO, 2002.

[9] Commission Decision 2007 589 EC, “Commission decision. establishing guidelines
for the monitoring and reporting of greenhouse gas emissions pursuant to directive
2003/87/ec of the european parliament and of the council,” Official Journal of the European
Union, May 2007.

[10] ISO/IEC 17025, “General requirements for the competence of testing and calibration
laboratories,” ISO, 2005.

[11] ISO 6976, “Natural Gas - Calculation of calorific values, density, relative density and
Wobbe index from composition,” ISO, no. Technical Corrigendum 3, 1996.

[12] OIML G 1-100, “Evaluation of measurement data - guide to the expression of

uncertainty in measurement,” OIML, 2010.

[13] AGA 8, “Compressibility and supercompressibility of natural gas and other related
hydrocarbon gases (2nd printing july 1994),” AGA, 1992.

14