The achievement of the algorithm duo is driven by advances in graphics processors and finds wide application in everything from chemical engineering to nuclear energy.
The flow properties of two-phase gas-liquid systems are critical for a wide range of applications, from chemical engineering to nuclear power and environmental engineering. Predicting and detecting bubbles in such systems remains a particular challenge.
Fang et al. have presented a method to dynamically detect these bubble interfaces in real time. Using the deep learning algorithms Deep Simple Online and Real-Time (DeepSORT) and You Only Look Once (YOLO), they achieved highly accurate bubble detection and real-time tracking even in complex gas-liquid two-phase flow environments, which can be adapted to numerous different use cases.
“The real-time capability of this system represents a significant advance over traditional methods, which often struggle in terms of both speed and accuracy,” said author Yue Feng. “The combination of cutting-edge object detection and tracking algorithms enables immediate analysis and intervention, which is critical for optimizing processes and ensuring safety.”
To solve the problem, the group configured the algorithms to train in parallel on videos of bubble formation in liquids. Such a process required the team to iteratively adjust the algorithms’ parameters and filters to achieve the best performance.
The algorithms automatically identify bubbles at the gas-liquid interface and can accurately delineate their boundaries with precise contours for each bubble. This basic data about the bubbles helps in further analysis of bubble dynamics, including frame-to-frame tracking.
“Hardware improvements, such as improved GPUs and specialized processors, have significantly accelerated our processing capabilities and enabled us to perform more complex calculations in real time,” Feng said.
The group hopes their work will inspire further efforts to monitor gas-liquid interfaces. Next, they plan to expand their analysis from two to three dimensions to improve detection.
Source: “A deep learning-based algorithm for fast tracking and monitoring of gas-liquid two-phase bubble flow,” by Lide Fang, Yiming Lei, Jianan Ning, Jingchi Zhang, and Yue Feng, Physics of liquids (2024). The article is available at https://doi.org/10.1063/5.0222856 .
