Sensitivity analysis of threshold parameters in slug detection algorithms

Abstract

Slug flow is a common flow pattern in pipelines that is often accompanied by undesirable effects like vibrations, pressure loss, and corrosion. Since these effects correlate with slug frequency, various attempts to predict this parameter by empirical or semi-empirical methods have been undertaken in the past. However, significant mismatches between these predictions can be observed. In this work, different slug frequency calculation methods have been applied to simulation data to investigate the sensitivity of threshold parameters that are often used in slug detection algorithms. The findings reveal that the detection of slugs from liquid holdup data is highly sensitive to these thresholds. Aeration of the liquid phase causes the gas–liquid interface to be less distinct and requires an adaption of the thresholds to the degree of aeration. In contrast, slug detection algorithms based on frequency analysis are robust to small deviations of the liquid level but fail to properly discriminate between slugs and waves. Our investigations show that slug frequency strongly depends on the method chosen for the determination of the liquid level. We propose new approaches that are less susceptible to aeration and approximate the liquid level very close to the authors’ human judgment.