Two-phase oil and water flow pattern identification in vertical pipes applying long short-term memory networks

Abstract

Accurate biphasic flow pattern recognition is essential in the design of coatings for the oil and gas sector because it enables engineers to create materials that are tailored to specific flow conditions. This results in enhanced corrosion protection, erosion resistance, flow efficiency, and overall performance of equipment and infrastructure in the challenging environments of the oil and gas industry. The development of flow maps has been based on empirical correlations that incorporate characteristics such as superficial velocities, volume fractions, and physical properties such as the density and viscosity of the analyzed substances. In addition, geometric parameters such as the inclination and the internal diameter of the pipes are considered. However, due to the difficult working conditions on offshore platforms and the limitations in monitoring internal flow patterns, technological advances have been implemented to improve this process using artificial intelligence techniques. In this context, this study proposes using a long-term memory (LSTM) recurrent neural network to predict the flow patterns generated in vertical pipes. This LSTM network was trained and validated using data obtained from a literature database. The results obtained showed that the model has a prediction error of less than 1%. These technological advances represent an important step towards optimizing the flow pattern identification process in the hydrocarbons industry. By leveraging the capabilities of artificial intelligence, more accurate and reliable forecasts can be obtained, enabling informed decisions and improving the efficiency and safety of operations.