Presentation Information
[2E14]Development of Validation Technology for Detailed Two-Phase Flow Simulation Codes to Optimize Innovative Reactor Design(2) Development Plan of Bubble Detection and Void Fraction Estimation Techniques Using Deep Learning and Numerical Analysis
*Shinichiro Uesawa1, Natsuki Hiramatsu2, Koji Ono2, Takahiro Maeshima3, Makoto Okada3, Hirobumi Tomita3, Kunitomo Aoki3, Shinichi Wako3, Hiroyuki Yoshida1 (1. JAEA, 2. HST, 3. ITIC)
Keywords:
Visualization,Wire mesh sensor,Deep learning,Numerical analysis
For quantitative validation of detailed two-phase flow simulations, measurement methods that can obtain instantaneous and local gas-liquid interface information are needed. In this study, the gas-liquid interface information of dispersed bubble flow is obtained by image processing. By using qualitatively validated simulation results as training data for an image recognition technique based on deep learning, we will develop the technology for bubbly flow with high bubble density. For churn flows, by using the two-phase flow simulation and electrostatic simulation results around a wire-mesh sensor (WMS) as training data, we will obtain the gas-liquid interface information with higher resolution than cross-points of the WMS. In this presentation, we report our development plans for these techniques.