A. Saglam, P. M. A. van der Kam, G. J. van Essen, D. H. Hebels
TIME DEPENDENT PERFORMANCE OF TURBINE GAS METERS
Gastransport Services in the Netherlands, uses turbine gas meters to measure the gas flow to local gas distribution companies, large industrial customers and power plants GTS operates approximately 1100 city gate stations, at a delivery pressure of usually 8 bar.
In the Netherlands, the procedures for determining the operational performance of the turbine meters are agreed upon between Gastransport Services and its customers. As a result of this agreement, Gastransport Services inspects the total population of turbine meters by means of the well-known variables-acceptance-sampling. Each year, 60 turbine meters are selected randomly from the total population. Subsequently, the selected turbine meters are calibrated traceable to international standards by Netherlands Measurment Institute (NMi). The information from these recalibrations is used to study the stability of the turbine meters.
Because, through the years a lot of information has been gathered on the turbine meters performance, the presence of statistically significant relationships (correlation) between the performance in time of the meters on the one hand and properties like the construction year, the pressure class, the size, the maximum allowable flow rate and/or the recalibration period of the meter on the other hand, can be investigated. In this presentation, analysis-of-variance and the quadrant-correlation-test are used to test for the presence of such correlations.
The aforementioned five properties were analysed for the presence of correlation at two different calibration pressures and with different definitions of drift to express the performance in time of gas turbine meters. For a specific calibration pressure and a specific type of drift, the performance in time dependents on the recalibration period and the size of the turbine meter. Furthermore, the correlation analysis, showed that a turbine gas meter seems to reach a steady state in time, expressed in terms of shift of the weighted mean error.