47
Appendix
A. Measurement Uncertainty of the ROOTS Model 5 Prover
In meter testing, it is important to understand the realities of measurement accuracy and acceptable
system error. This is commonly referred to as measurement uncertainty. The American National Standard
for Rotary Type Gas Displacement Meters, ANSI B109.3, PART VII, Test Methods and Equipment, 7.2
Measurement Reference Base states: "The final authority for all standards of measurement in the United
States is the National Institute of Standards and Technology (NIST) …" Part VII, 7.5.3 Accuracy addresses
"Accuracy of Test Standards," "Uncertainties of Observations," and "Uncertainties in Method of Test." This
often over-looked subject is an important component of measurement standards, test equipment and test
methods. Part VII, Test Methods and Equipment, 7.6 Calibration of Meter Testing Systems, 7.6.1 General,
second paragraph states: "Meter testing systems shall be calibrated when first installed, and following
alterations, damage or repairs which might effect accuracy. To assure that the accuracy of the meter testing
systems is maintained on a continuous basis, a daily leakage test shall be made and a periodic accuracy
indication with a test meter of known accuracy shall be made. If the test results differ by more than ± 0.5
percent from the test meter accuracy, the cause of the error shall be determined and necessary corrections
made prior to reuse of the system."
Secondary test devices such as Bell or Piston type provers traceable to NIST are recognized by the Natural
Gas Industry as having an accuracy of ± 0.25%. These provers are used to establish the accuracy presets
for a transfer prover Master Meter – a tertiary test device (third removed from NIST traceability).
The accumulated errors of all three standards, as well as the errors of the associated equipment must
be considered when analyzing meter test results.
What does this mean? A test could result in an indicated meter accuracy of 101.25% and still be
considered accurate and within specification. How? By adding in the uncertainty of the prover. A prover
testing at the high end of its accuracy band (± 0.55%) and a meter testing at the higher end of its
specification (i.e., 100.7%) would have a combined theoretical error of + 1.25% (100.55 X 100.7 + 100).
As recommended by ANSI, a test meter of known accuracy (Reference Meter) should be used in
problem identification and resolution. A factory-certified accuracy curve is supplied with each ROOTS
reference meter. To characterize the performance of the Transfer Prover System, a reference meter should
be tested on the prover and the test results then plotted and/or compared against the historical baseline
data. This is best method for a quick prover operational check and verification of repeatability.
Errors associated with measurement of pressure, temperature and flow rate contribute to the overall
uncertainty of the testing process. Assuming that all factory recommended test procedures are followed,
the test environment is within the specified tolerances, and all associated equipment is in proper working
order - the overall uncertainty of the Prover system can exceed the system design or specified accuracy.
A significant reduction in uncertainty can be accomplished by: (1) calibrating the equipment under the
conditions that it will be used during testing; (2) ensuring that the test points do not correspond to a flow
condition that would cause resonance; and (3) by using larger test volumes.
It is extremely important that the meter to be tested is stored (soaked) for a minimum of eight hours
in the stable environment in which the testing will be conducted. Meter soaking and test environment
stability allow the probe well temperature to equalize to the temperature of the air flowing through
the meter. Significant increases in uncertainty (greater errors) can occur if the meter is not properly
conditioned prior to testing or if the test environment is not stable during testing. During field testing,
consideration must be given for the induced errors due to unstable or averaged temperatures during a
test sequence.
The ISO Guide method for calculating measurement uncertainty should be used to determine the total
overall uncertainty.
