Advancing Two-Phase Flow Research with Computer Vision and Generative Neural Networks

with Yang Liu,
Associate Professor, Nuclear Engineering Program
Mechanical Engineering Department, Virginia Tech

Sep 6, 2024 10:00 AM
Blacksburg (440 Goodwin), Arlington (6-051 VTRC)

Register

Understanding and accurately predicting two-phase flow dynamics is crucial for the design and safety of a wide range of nuclear reactors. However, complex phenomena such as bubble nucleation, bubble break-up, coalescence, and flow regime transitions present significant challenges in both modeling and experimentation. This talk will present recent research efforts using computer vision and machine learning to advance our understanding of these challenges. The first part of the talk will focus on the development of a high-speed imaging system designed to measure two-phase flow parameters. This system integrates various computer vision techniques, including edge detection, image segmentation, 3-D reconstruction, and motion tracking. These techniques can effectively extract valuable data from high-speed two-phase flow images and enable the visualization of the real-time 3-D interfacial structure of these flows. Additionally, two generative neural networks will be introduced that were developed to create realistic synthetic images. The first network, called Bubble Generative Adversarial Networks (BubGAN), is designed to generate isolated bubbles. It is conditioned on bubble feature parameters and offers full control over bubble properties. Realistic bubbly flow images can be assembled using the generated bubbles and relevant computer vision tools. The second network, Bubbly Flow Generative Adversarial Networks (BF-GAN), is designed to generate bubbly flow images through inputs of superficial gas and liquid velocities. These generative tools can provide effective and large-scale benchmarking and training data for image processing models, which, in turn, can further advance our understanding of two-phase flow dynamics.

Yang Liu is an Associate Professor in the Nuclear Engineering Program within the Mechanical Engineering Department at Virginia Tech. His expertise includes two-phase flow, reactor thermal-hydraulics, and reactor safety. Throughout his career, he has successfully led a wide range of projects, from fundamental two-phase flow studies to flow-induced vibrations, two-phase flow instabilities, and the design of passive small modular reactors. Recently, he has developed several advanced instrumentation systems for multiphase flow measurement, including a multi-camera high-speed imaging system, a compact and fast X-ray densitometry system, and advanced multi-sensor conductivity probes. Dr. Liu has authored or co-authored more than 100 research articles in peer-reviewed journals and conference proceedings. He is the recipient of multiple best-paper awards from international organizations and conferences, including the Japanese Society of Mechanical Engineers (JSME, 2015), the International Topical Meeting on Nuclear Reactor Thermal Hydraulics, Operation and Safety (NUTHOS, 2018), and Advances in Thermal Hydraulics (ATH, 2018 and 2022).