International Society for Industrial Process Tomography

Deep Learning Approach for Oil Flow Rate Estimation in Two-Phase Flow using Dual-Modal Electrical Tomography

M Ziaul Arif¹*, Marko Vauhkonen¹

¹ Department of Technical Physics, University of Eastern Finland, 70211 Kuopio, Finland

^*Email: muhammar@uef.fi

ABSTRACT

Difficulties in accurately estimating phase flow rates in multiphase flow systems have posed significant challenges in various industries. This study presents a novel approach that combines electromagnetic flow tomography (EMFT) and electrical tomography (ET) with machine learning technique, specifically deep neural network (DNN), to predict the oil flow rate in a two-phase oil-water flow. The two-phase oil-water flow conditions are generated with varying flow regime using COMSOL software, providing a realistic representation of the flow. Furthermore, the dual-modal system measurement data are simulated using a dense finite element mesh, ensuring realistic measurements, and serving as input for the DNN. The results demonstrate promising potential of the proposed approach in accurately estimating the oil flow rate in complex two-phase flow systems. By combining the strengths of dual-modal EMFT and ET along with the DNN approach, this study offers valuable insights for guiding experimental studies and represents a significant step forward in overcoming challenges related to flow rate estimation in multiphase flow problems. Ultimately, this approach holds great potential for benefiting industries reliant on precise flow rate measurements in multiphase flow systems.

Keywords: Deep neural-network, Dual-modal imaging, Electrical tomography, Electromagnetic flow tomography, two-phase flow

Industrial Application: Petroleum and processing industries.