Mauricio Figueiredo, Hugo Gutierrez, Ana M. F. Fileti, Nicolas Ratkovich, and Ricardo D.M. Carvalho
Artificial neural networks, Intelligent signal processing, Liquid-gas flows, Flow pattern, Gas volume fraction
In the oil industry, the well stream often consists of hydrocarbons and non-wanted components such as water, carbon dioxide, salts, sulphur, and sand. Knowledge of the multiphase flow pattern and gas volume fraction (GVF) allows for more reliable and safer plant operation. Multiphase flow metering (MFM) tries to provide these and other pieces of information without full phase separation. The drawback of MFM techniques is the need for prior signal calibration. A broader solution could be the use of artificial neural networks (ANNs). The present work discusses the development of ANNs based on dimensionless numbers for the identification of the flow pattern of liquid-gas flows and determination of the GVF. Experimental data were obtained from the literature for flows of viscous liquids and gas. Selection of the dimensionless numbers for input to the ANNs was based on physical reasoning and analysis of the ANNs performance. Finally, a comparison was made between the ANNs predictions and results from the numerical simulations of the flows using OLGA (OiL and GAs) software.