A probabilistic approach to oil-water flow patterns in horizontal pipelines is presented. An experimental dataset was compiled, and validated by means of OLGA Multiphase Toolkit 2014.3; moreover, an associated error was calculated comparing the reported hold up with its corresponding simulated value. The registers, whose associated error was lower than a defined tolerable error, were used to generate flow pattern maps under a Bayesian approach. A probability surface was generated for each pattern and a tool was developed in Matlab 2015a, which calculates the probability of belonging to each flow pattern given the superficial velocities of both phases.
Simultaneous flow of two immiscible liquids through a horizontal pipeline is a common phenomenon in the Oil & Gas industry. Nevertheless, the mixture of both phases is a complex problem due to the significant differences of flow behavior under different flow conditions (1). Understanding the behavior of oil-water flow is crucial for a proper design of separations facilities, multiphase pumps, and artificial lift methods related to oil extraction and to understand gas-oil-water three-phase flows (2; 3).
Oil-water flow is characterized by a low density ratio and a viscosity ratio that can vary in many orders of magnitude (4). It is not only affected by inertia, viscous, pressure and interfacial forces, but also by the inclination and wetting characteristics of the pipe, the pre-wetting component and the operation conditions such as temperature and superficial velocities. The interaction of these components leads to specific geometric distributions of the phases, which are known as flow patterns. Taking into account that each flow pattern results in different flow characteristics such as pressure drops, mixture viscosity and water holdup, understanding the behavior of liquid-liquid flow is crucial. Thus, flow pattern maps or simulation tools are required.