This paper outlines a technique for filtering elevation profile data for multiphase pipeline calculations. When gas and liquids flow together in a pipeline, slip between the phases results in the accumulation of liquid in uphill sections of the pipeline. This liquid holdup causes the pressure losses due to hydrostatic head in the uphill sections of the line to be greater than the pressure recovery in downhill sections of the line. Consequently, it is necessary to use detailed elevation profiles when using software to simulate flow in a multiphase pipeline. Extremely detailed elevation profiles (e.g. the result of electronic processing of aerial photography) are becoming more commonly available, but can result in pipeline models with long calculation times, large memory usage, and large output files. This paper presents a filtering technique that removes redundant points, reducing the computing requirements of a model without significant impact on the integrity of the results.
The procedure outlined in this paper was evaluated using measured data from existing pipeline systems. In one such system, the number of points in the elevation profile was reduced by 83% with a corresponding 91% reduction in the time taken to perform the calculations and less than a 0.25% change in the calculated pressure losses and only a 0.2% change in the calculated liquid holdup. Results for other systems that were evaluated were comparable. Trials conducted with as few as 3.2% of the original number of points showed no more than a 2.2% change in the calculated pressure losses and no more than an 8.4% change in the calculated liquid holdup. The size of the resulting output data files was reduced by 64–96%. Filtering elevation data profiles using the procedure outlined in this paper should be worthwhile in almost any multiphase pipeline modelling where extremely detailed elevation profiles are available.
Multiphase pipeline modelling with detailed elevation profile data is often assumed to be a time-consuming process, taxing for even today's powerful desktop computers. The procedure outlined in this paper can dramatically improve the efficiency of the modelling process.