Study of identification of global flow regime in a long pipeline transportation system

Accurate recognition of flow regimes is essential for the analysis of behavior and operation of two-phase flow systems in industrial processes. This paper studies the feasibility of a method for objective recognition of global flow regimes using differential pressure and artificial neural networks. Three main global flow regimes are classified based on differential pressure of the riser. Five statistical parameters are inputted into neural networks classifiers, and good recognition rates of global flow regimes are achieved. With increase of feature parameters, recognition rates of global flow regimes increase, and five selected feature parameters are sufficient to achieve good recognition rates. Recognition rates of two categories are generally higher than those of four categories, and they are found to increase with sample lengths. Average recognition rates of four categories are higher than 94.3% if sample lengths are longer than 240 s and reach almost 100% when sample lengths are sufficient long. [Display omitted] •Experiments of air-water flow in a long-distance pipeline-riser are carried out.•Differential pressures of the riser reveal different trends in various flow regimes.•Flow regime recognition is performed by differential pressure and neural network.•Influences of feature number and sample length on recognition rate are analyzed.