Automated Segmentation and Analysis of High-Speed Video Phase-Detection Data for Boiling Heat Transfer Characterization Using U-Net Convolutional Neural Networks and Uncertainty Quantification

Boiling heat transfer is a complex phenomenon used for cooling and heat management purposes in various industrial applications, such as nuclear reactors. Accurate characterization and understanding of boiling dynamics are essential for the design and optimization of heat transfer systems. High-speed video (HSV) imaging is a powerful tool for capturing the intricate details of boiling processes. However, the manual analysis of HSV data is time-consuming and prone to subjective interpretation. This thesis presents a novel approach for the automated segmentation and analysis of HSV phase-detection images using U-Net Convolutional Neural Networks (CNNs) and uncertainty quantification techniques. The proposed methodology involves the development of specialized U-Net CNN models for segmenting HSV data of boiling phenomena in different fluids, including liquid nitrogen, argon, FC-72, and high-pressure water, under various experimental conditions. The performance of the U-Net models is evaluated and compared with traditional adaptive thresholding techniques. The results demonstrate the superior accuracy and robustness of the U-Net models in identifying and delineating bubbles compared to manual segmentation, particularly in scenarios involving smaller bubbles and complex bubble topologies. To assess the reliability of the calculated boiling metrics, such as contact line density and dry area fraction, a comprehensive uncertainty quantification analysis is also conducted. The impact of discretization errors arising from the pixelation of bubbles is investigated using weighted average percentage relative errors and mean errors under both erosion and dilation conditions. The analysis reveals higher relative uncertainty in contact line density measurements than dry area fraction measurements across all fluids studied. The limitations of the U-Net models in generalizing to other HSV datasets are addressed, emphasizing the need for developing more sophisticated image segmentation models, such as foundation models, that are less sensitive to domain shifts. This is crucial for enabling autonomous experimentation and reducing the reliance on specialized models for each fluid and operating condition. Future research directions are outlined, including the investigation of advanced uncertainty quantification techniques, the development of real-time segmentation and analysis algorithms, the evaluation of uncertainty propagation in heat flux reconstruction, and the extension of the methodology to other multiphase flow phenomena. By addressing these recommendations, the understanding, characterization, and modeling of boiling phenomena can be further enhanced, contributing to the advancement of boiling heat transfer research and the development of improved heat transfer models and correlations. Overall, this thesis presents a comprehensive approach for the automated segmentation and analysis of HSV phase-detection images using U-Net CNNs and uncertainty quantification techniques. The proposed methodology demonstrates significant potential for accurate and reliable characterization of boiling dynamics, paving the way for advanced boiling heat transfer research and the optimization of heat transfer systems in various industrial applications.

Name thesis:degree_name

Master

Department dc:contributor.department

Massachusetts Institute of Technology. Department of Nuclear Science and Engineering

Grantor dc:publisher

Massachusetts Institute of Technology

Year dc:date.issued

2024

Author dc:creator

• Maduabuchi, Chika

Advisor dc:contributor.advisor

• Bucci, Matteo

Statement dc:rights

• In Copyright - Educational Use Permitted
• Copyright retained by author(s)

Licence dc:rights.uri

• https://rightsstatements.org/page/InC-EDU/1.0/

Handle dc:identifier.uri

https://hdl.handle.net/1721.1/155645

OAI identifier oai:identifier

oai:dspace.mit.edu:1721.1/155645