Abstract

Flow Pattern and Liquid Holdup Prediction in Multiphase Flow by Machine Learning Approach

Author(s): Chandrasekaran S*, Kumar S

In this work, data models are developed by artificial intelligence techniques to predict the flow patterns and to predict liquid hold up during multiphase flow in the drilling phase and production phase of petroleum exploration process. This is significant because early alerting and mitigation of kick, which is a multiphase flow, is important from the safety of the drillers and preventing unwanted catastrophe of life and property. In the production process of petroleum industry, addressing this problem is significant to know about the productivity of the well, design of the transport system and refining process. It is shown that these methods were able to match and improved the prediction performance from multiphase empirical correlations. Different neural network model structures and machine learning classification and regression algorithms were evaluated and optimized to determine the best models that performs prediction accurately. The data for developing the models was collected from experiments in literature where two phase flow of air-kerosene in inclined pipes were studied.