PR-015-15608-R01 Sample Probe Insertion Depth Testing
Author(s): Jacob Thorson,Darin George
Abstract/Introduction:
Probes for natural gas sample collection and analysis must extend far enough into the pipeline to avoid contaminants at the pipe wall, but must not be so long that there is a risk of flow induced resonant vibration and failure. PRCI has sponsored a project to determine the minimum probe depth for obtaining a representative single-phase gas sample in flows with small amounts of contaminants. Phase 2 of the project evaluated the effect of probe tip design and insertion length on the fidelity of gas samples from flows carrying small amounts of liquid hydrocarbons.
Download/Purchase
PR-015-15610-R01 Effect of Upstream Piping Configurations on Ultrasonic Meter Bias
Author(s): Hawley, Grant
Abstract/Introduction:
This project investigated the performance of four different brands of commercially available ultrasonic meters installed in an AGA-9 default configuration in two different header designs. The diagnostics from the flow meters were analyzed to determine if they could be used as an indication of bias in the flow measurement. The results from this project were also compared to previous PRCI projects where similar headers were experimentally tested and modeled using CFD. The results of this report are blinded as to meter make/model.
Download/Purchase
PR-015-15610-R02 Effect of Upstream Piping Configurations on Ultrasonic Meter Bias - Unblinded
Author(s): Adam Hawley,Christopher Grant
Abstract/Introduction:
This project investigated the performance of four different brands of commercially available ultrasonic meters installed in an AGA-9 default configuration in two different header designs. The diagnostics from the flow meters were analyzed to determine if they could be used as an indication of bias in the flow measurement. The results from this project were also compared to previous PRCI projects where similar headers were experimentally tested and modeled using CFD.
Download/Purchase
PR-261-15609-R02 Machine Learning Algorithms for Smart Meter Diagnostics - Part III (TR2777)
Author(s): Jeff Crowe
Abstract/Introduction:
Our objective of this work was to investigate exclusively Daniel USMs. Sixty five thousand individual data points were used in MLA development which totaled over 18 hours of USM data from seven experimental data sets generated at three flow facilities. Six disturbance types were investigated (baseline, single elbow, double elbow out of plane, liquid, elbow header, and tee header). All experimental data was labeled with the disturbance type, if any, and deviation from baseline error. The MLA feature set was improved from the 2015 work by using gas flow conditions to compare measured and predicted flow velocities (flow profiles) and adding features that quantify the stability of the USM flow measurement. Supervised clustering and regression algorithms were fit to the labeled USM data and the accuracy of the MLAs was calculated using a cross-validation technique.
Download/Purchase