[1H01-07] 数値解析とシミュレーション技術

16:15 〜 16:30

[1H07] BF-GAN: Development of an AI-driven Bubbly Flow Image Generation Model Using Bubbly Generative Adversarial Networks

*WEN ZHOU¹, Shuichiro Miwa¹, Koji Okamoto¹ (1. UTokyo)

キーワード：

Bubbly Flow、Physically Conditioned Deep Learning、Image Generation Model、Generative Adversarial Networks

A generative AI architecture called bubbly flow generative adversarial networks (BF-GAN) is developed, designed to generate realistic and high-quality bubbly flow images through physically conditioned inputs, jg and jf. 140,000 bubbly flow images with physical labels of jg and jf are collected for training data. A multi-scale loss function is then developed, incorporating mismatch loss and feature loss to enhance the generative performance of BF-GAN further. The comparative analysis demonstrate that the BF-GAN can generate realistic and high-quality bubbly flow images with any given jg and jf within the research scope, and these images align with physical properties.