Dear Mr. Khan, Ultimately the ability to perform rigorous on-line real-time pipeline leak detection will rest upon low uncertainty calibration.  Once metering systems can be calibrated with low uncertainty, one can determine if they are suitable for the flowrates (which they are frequently not) and characterize their behavior.  If the meters aren't accurate, but are nominally repeatable, then they can be characterized in software for improved accuracy. The lowest detectable point-to-point leakage/theft in distribution mains then becomes ~2.5*(calibration uncertainty + metering repeatability) -- unless the metering is calibrated systems are calibrated in series simultaneously. Once distribution main loss is known, then one can consider large data set analysis methods to attempt to determine leakage in branch lines,  While I have little experience or specific/detailed knowledge in the area of end-of-line (inaccurate) metering, this will obviously obviously involve pairwise comparison of flowrates over a given time period, comparison of slopes, etc. If one considers the distribution main to be analogous to that shown in Figure 1 of the natural gas leak detection paper, and branch line configuration to be similar to Figure 2, we would have the basis for a more rigorous pipeline leak detection.  Unfortunately most distribution main metering is poor, and end-of-line metering really terrible.  So measurement systems really have to be improved -- via calibration -- if anything meaningful is to be achieved.  (The last page of the full version of my large WATER flow calibration facility proposal includes a letter from US-NIST regarding the design of a large NATURAL GAS metering calibration facility.) I can provide more voluminous data via direct email, if you want further information. Arnold

Published by

Dear Mr. Khan,

Ultimately the ability to perform rigorous on-line real-time pipeline leak detection will rest upon low uncertainty calibration.  Once metering systems can be calibrated with low uncertainty, one can determine if they are suitable for the flowrates (which they are frequently not) and characterize their behavior.  If the meters aren't accurate, but are nominally repeatable, then they can be characterized in software for improved accuracy.

The lowest detectable point-to-point leakage/theft in distribution mains then becomes ~2.5*(calibration uncertainty + metering repeatability) -- unless the metering is calibrated systems are calibrated in series simultaneously.

Once distribution main loss is known, then one can consider large data set analysis methods to attempt to determine leakage in branch lines,  While I have little experience or specific/detailed knowledge in the area of end-of-line (inaccurate) metering, this will obviously obviously involve pairwise comparison of flowrates over a given time period, comparison of slopes, etc.

If one considers the distribution main to be analogous to that shown in Figure 1 of the natural gas leak detection paper, and branch line configuration to be similar to Figure 2, we would have the basis for a more rigorous pipeline leak detection.  Unfortunately most distribution main metering is poor, and end-of-line metering really terrible.  So measurement systems really have to be improved -- via calibration -- if anything meaningful is to be achieved. 

(The last page of the full version of my large WATER flow calibration facility proposal includes a letter from US-NIST regarding the design of a large NATURAL GAS metering calibration facility.)

I can provide more voluminous data via direct email, if you want further information.

Arnold