Steady-State Multiphase Flow—Past, Present, and Future, with a Perspective on Flow Assurance

Abstract

A monumental amount of research and development work has been invested in multiphase flow modeling over the past 50 years. Yet, many challenges remain as we incorporate additional phases, account for exotic fluids, and push our simulation tools to their limits in the desire to optimize even more-complex production systems fraught with difficult flow assurance issues. A visual history of the evolution of steady-state multiphase flow models will be presented, leading to the current "state-of-the-art". Looking toward the future, it is important that the models be advanced to address areas of greatest uncertainty and align with trends in field development strategies. The authors' views of the top five areas of reseearch and development (R&D) necessary for these purposes will be presented.

An improved method for applying the lockhart–martinelli correlation to three‐phase gas–liquid–solid horizontal pipeline flows

“…Ultimately, though, reliable computer simulations of three‐phase G/L/S horizontal pipeline flows are not yet possible unless two of the three phases can be considered to form a single‐phase, homogeneous flow . The primary issue, which is problematic even in the much more highly developed (and more intensely studied) fields of G/L and L/S (slurry) flows, is the empiricism inherent in the so‐called ‘closure relations’ needed to solve combined continuity and momentum equations for each phase . Many computational fluid dynamics (CFD) applications for G/L pipeline flows, for example require one to specify the flow regime before the model can solve for the local, averaged concentrations and/or velocities of each phase .…”

“…The primary issue, which is problematic even in the much more highly developed (and more intensely studied) fields of G/L and L/S (slurry) flows, is the empiricism inherent in the so‐called ‘closure relations’ needed to solve combined continuity and momentum equations for each phase . Many computational fluid dynamics (CFD) applications for G/L pipeline flows, for example require one to specify the flow regime before the model can solve for the local, averaged concentrations and/or velocities of each phase . Only recently have flow‐regime independent G/L pipeline flow models been developed, but even these have very specific limitations .…”

“…[9] Many computational fluid dynamics (CFD) applications for G/L pipeline flows, for example require one to specify the flow regime before the model can solve for the local, averaged concentrations and/or velocities of each phase. [9,10] Only recently have flow-regime independent G/L pipeline flow models been developed, but even these have very specific limitations. [11] For these models to be extended, experiments are required to formulate and/or validate the required closure relations.…”

On gray-box modeling for virtual flow metering

“…Mechanistic models are based on prior knowledge about the internal structure of a process. Physical, first principle laws such as mass, energy, and momentum balance equations together with empirical closure relations, are utilized to describe the relationship between the process input, internal, and output variables (Shippen, 2012). Due to the extensive amount of prior knowledge utilized in model development, mechanistic models are transparent and have a high degree of interpretability.…”

Data driven methodology for model selection in flow pattern prediction

“…Firstly, there are direct methods based on data analysis that, considering different sets of variables, can estimate the flow type. Taking into account that flow patterns depend on parameters such as pipe inclination, diameter and length, physical properties of the phases, and superficial velocities (Shippen and Bailey, 2012), many machine learning approaches have been developed in the last years to identify flow patterns (e.g. Xie et al (2004); Al-Naser et al (2016); Amaya-Gómez et al (2019)).…”