Machine Learning Assisted Experimental Characterization of Bubble Dynamics in Gas-Solid Fluidized Beds.

Area of Science:

• Fluid dynamics
• Chemical engineering
• Machine learning applications

Background:

• Accurate bubble identification is crucial for understanding gas-solid fluidized beds.
• Traditional methods face challenges with varying illumination and focus, impacting hydrodynamic study reproducibility.
• Oscillating fluidized beds present complex bubble dynamics that are difficult to analyze.

Purpose of the Study:

• To develop and validate a machine learning-assisted image segmentation method for automatic bubble identification in gas-solid fluidized beds.
• To enhance the accuracy and consistency of bubble tracking and analysis.
• To apply the method to challenging oscillating fluidized beds and identify new flow characteristics.

Main Methods:

• Utilized a machine learning (ML) model for binary image segmentation to identify bubbles.
• Developed an in-house Lagrangian tracking technique to monitor bubble evolution.
• Validated the ML-assisted segmentation and tracking across diverse operational conditions and particle sizes, including oscillating beds.

Main Results:

• Achieved 98.75% accuracy in bubble recognition, filtering out common sources of uncertainty.
• Successfully captured complex bubbling dynamics and subtle changes in velocity and size distributions.
• Identified new features of oscillating beds, linking bubble morphology and flow stability to operational parameters.

Conclusions:

• The ML-assisted method offers an efficient, standardized, and repeatable tool for hydrodynamic studies in fluidized beds.
• The technique demonstrates versatility and effectiveness across various particle sizes and operational conditions.
• This approach has potential for broader application in other multiphase flow systems.
• The study provides new insights into the hydrodynamics of oscillating fluidized beds.