[2J06-09] Gas-Liquid Two-Phase Flow 2

[2J07] Development of a gas-liquid two-phase flow image analysis method using deep learning based object detection and recognition techniques

*Eguchi Kohei¹, Miwa Shuichiro², Sawa Kazuhiro¹ (1. Hokkaido Univ., 2. UTokyo)

Keywords:Gas-Liquid Two-Phase Flow, Void Fraction, Interfacial Area Concentration, Deep Learning, Object Detection

Instantaneous and high accurate detection of flow parameter like void fraction of two-phase flow, contribute to safe and efficient operation of nuclear plants. Especially, understanding of the gas-liquid interface information in two-phase flow, which typically shows complicated temporal changes, will help developing advanced numerical models or correlation for two-phase flows. In this study, bubble detection model was developed by training object detection algorithm based on deep learning. Then, flow parameters like void fraction were calculated from position and size information of detected gas phase in images, and the estimation method was validated by being compared with other measurement or calculation method. Near the transition from bubbly and slug flow, characteristics of cap bubbles were investigated to analyze their motion.