Development of a Machine Learning-Based Model for Cryogenic Film Boiling Flow Pattern Prediction During Transfer Line Chill-Down Operation

Abstract Efficient transfer of vapor-free liquid cryogen in low-temperature applications hinges on achieving thermal equilibrium during transfer line chill-down, a phase characterized by intricate two-phase heat and mass transfer phenomena. This study introduces a novel machine learning-based model tailored to identify flow patterns that are Inverted Annular Film Boiling (IAFB), Inverted Slug Film Boiling (ISFB), and Dispersed Flow Film Boiling (DFFB)—in cryogenic quench flow boiling. Utilizing three distinct datasets, the research evaluates model performance, highlighting significant input variables. Initial accuracy improvements, from 15% to 98.5%, underscore the impact of feature selection and dataset composition changes. A non-dimensional approach achieves 90.2% accuracy, showcasing promising advancements in predictive modeling for cryogenic systems. Future research will focus on refining non-dimensional models and exploring additional feature transformations to enhance robustness and performance further.