Machine learning-based heat flux estimation from high-speed video during saturated pool boiling over vertical tube

Area of Science:

• Thermodynamics and Fluid Mechanics
• Nuclear Engineering
• Materials Science

Background:

• Saturated pool boiling is critical for nuclear power plant safety.
• Accurate prediction of heat flux and heat transfer efficiency is vital for equipment design and reliability.
• Boiling involves bubble formation, growth, and departure, influenced by nucleation sites.

Purpose of the Study:

• To develop a machine learning framework for in situ boiling heat flux prediction.
• To correlate high-speed imaging of dynamic bubbles with heat flux.
• To automate and improve the accuracy of boiling heat transfer metrology.

Main Methods:

• Utilized a machine learning approach correlating high-speed imaging with dynamic bubble behavior.
• Employed deep learning models, including convolutional neural networks and object detection algorithms.
• Extracted hierarchical and physics-based features to learn physical boiling laws.

Main Results:

• Achieved an average heat flux prediction error of approximately 6%.
• Maintained an overall classification accuracy of 88% across all heat flux levels.
• The model statistically describes bubble nucleation, coalescence, and departure.

Conclusions:

• The proposed machine learning framework provides an automated, learning-based alternative to conventional methods.
• The approach enables accurate in situ prediction of boiling heat flux.
• This method enhances the understanding and monitoring of boiling phenomena for improved safety and design.