Date of Award

January 2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical Engineering

First Advisor

Clement Tang

Abstract

The heat transfer process in multiphase flow systems becomes essential forindustrial applications and when phase change occurs. The heat transfer coefficients become more efficient during boiling and condensation processes when compared to single-phase fluid flow. The design of thermal systems requires knowledge about these coefficients together with their essential parameters. This study evaluates flow boiling heat transfer coefficients for low-global warming potential refrigerants which include pure ethanol and propane (R-290) and R- 600 and R-600A. The experimental database included more than 25 independent studies that covered various operating conditions and channel sizes and fluid properties. Existing empirical correlations were evaluated, revealing limitations in their predictive accuracy. To address these shortcomings, a dual approach was employed which included running two-dimensional CFD simulations with the Volume of Fluid model through ANSYS Fluent and machine learning with symbolic regression. The 2D simulation results demonstrated strong agreement with experimental data thus validating their reliability and cost-effectiveness. Moreover, symbolic regression, a machine learning technique was uniquely applied in this field to develop 2 new predictive equations which are Ethanol and hydrocarbons-based correlation. These correlations provide 2.4 – 8 % MAE respectively providing interpretable formulas compared to conventional artificial neural networks approaches. The results provide accurate and reliable predictive models for engineers and researchers working with eco-friendly refrigerants and help improving the future design of sustainable thermal systems

Download

Share

COinS