Gas-Liquid Two-Phase Flow Image Analysis Using Bubble Detection Algorithm

Abstract

Internal gas-liquid two-phase flow is commonly observed in various engineering disciplines. One of its unique characteristics is the presence of flow regime, and its identification is crucial for improving the accuracy of thermal-hydraulic analysis codes as well as securing the integrity of piping systems. Furthermore, transition of two-phase flow regime takes place gradually, and is highly unsteady phenomenon, objective flow regime identification near the transition region still remains challenging to this day. In the present study, we developed the novel flow regime identification method based on the state-of-the-art AI technique which is fully automated and capable of detecting bubble characteristics from high-speed images at high accuracy. The present tool was developed based on a machine learning algorithm, which can quickly detect flow characteristics and capture individual bubble positions from given images. Furthermore, it is possible to calculate major two-phase flow parameters such as void fraction and bubble rise velocity using the detection results. By utilizing the current tool, instant objective two-phase flow feature extraction is now possible, and our results showed promising performance by coupling the state-of-the-art AI technique with conventional thermal-hydraulics.