Date on Master's Thesis/Doctoral Dissertation

5-2022

Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Department

Mechanical Engineering

Degree Program

Mechanical Engineering, PhD

Committee Chair

Lian, Yongsheng

Committee Co-Chair (if applicable)

Brehob, Ellen

Committee Member

Brehob, Ellen

Committee Member

Berson, R. Eric

Committee Member

Park, Sam

Committee Member

Williams, Stuart

Author's Keywords

Nucleate boiling; microgravity; numerical simulations; data driven techniques; machine learning

Abstract

Nucleate boiling is important in nuclear applications and cooling applications under earth gravity conditions. Under reduced gravity or microgravity environment, it is significant too, especially in space exploration applications. Although multiple studies have been performed on nucleate boiling, the effect of gravity on nucleate boiling is not well understood. This dissertation primarily deals with numerical simulations of nucleate boiling using an adaptive Moment-of-Fluid (MoF) method for a single vapor bubble (water vapor or Perfluoro-n-hexane) in saturated liquid for different gravity levels. Results concerning the growth rate of the bubble, specifically the departure diameter and departure time have been provided. The MoF method has been first validated by comparing results with a theoretical solution of vapor bubble growth in superheated liquid without any heat transfer from the wall. Next, bubble growth rate and heat transfer results under earth gravity, reduced gravity, and micro-gravity conditions are reported and they are in good agreement with experiments. A new method is proposed for estimating the bubble diameter at different gravity levels. This method is based on an analysis of empirical data at different gravity values and uses power-series curve fitting to obtain a generalized bubble growth curve irrespective of the gravity value. This method is shown to provide a good estimate of the bubble diameter for a specific gravity value and time. A new hybrid approach is proposed for calculating the contribution of the depletable liquid micro-layer trapped between the vapor bubble and the heater wall for numerical simulations in microgravity conditions is proposed in this work. This technique does not ``model'' the micro-layer, but calculates the contribution of the vapor flux from the micro-layer into the bubble and distributes it over the cells where the micro-layer should be present. The micro-layer is depletable because an evaporation term is part of the equation which maintains the reduction in the thickness of the micro-layer consistent with the behavior reported in experiments. Results for nucleate boiling simulations under micro-gravity conditions are reported using the proposed micro-layer approach in comparison with experiments performed on the International Space Station. Results for bubble growth rate, bubble shape, and heat-flux are in good agreement with experiments and are verified with two different time instants in the bubble life cycle. Additionally, a data-driven model is proposed for the prediction of heat-flux from experimental parameters like wall super-heat, gravity, liquid sub-cooling, etc. Experimental data from multiple experiments under varying conditions for different liquids have been performed to date. Artificial Neural Networks (ANNs) have been used to predict nucleate boiling heat flux by learning from a dataset of twelve experimental parameters across 231 independent samples. An approach to reduce the number of parameters involved is proposed to increase model accuracy. The approach consists of two steps. In the first step, a feature importance study is performed to determine the most significant parameters. Only important features are used in the second step. In the second step, dimensional analysis is performed on these important parameters. Neural network analysis is then conducted based on the dimensionless parameters. The results indicate that the proposed feature importance study and dimensional analysis can significantly improve ANN performance. The results show that model errors based on the reduced dataset are considerably lower than those based on the initial dataset. The study based on other machine learning models also shows the reduced dataset generates better results. The results also show that ANN outperforms other machine learning algorithms and outperforms a well-known boiling correlation equation. The effect of parameters on heatflux has been quantified, and the effect of parameters on different physical sub-processes in nucleate boiling has been analyzed. The effect of parameters on the boiling regimes has also been investigated. Additionally, the feature importance study concludes that wall superheats, gravity and liquid subcooling are the three most significant parameters in the prediction of heat flux for nucleate boiling. The key contributions made in this work are listed below:

  • MoF method simulations for nucleate boiling have been performed. Simulation results in earth gravity, and reduced gravity are in good agreement with experiments. \item A data-driven technique for the prediction of the effect of gravity on bubble growth rate has been proposed.
  • A novel depletable microlayer approach for microgravity is proposed, results for bubble growth rate, bubble shape, and heat-flux are comparable to experiments performed on ISS.
  • A novel data-driven technique has been used for heat flux prediction. ANN outperforms XGB (Extreme Gradient Boosting), RFR (Random Forest Regression), and Rohsenow correlation in heatflux prediction.
  • Dimensional Analysis and Feature Importance techniques help in reducing ANN error from 25.7\% to 9.12\%.
  • Gravity, Wall superheat, and Liquid subcooling are the three most significant parameters in heatflux prediction. Novel results of quantification of parameter contribution in each boiling regime have been reported.

Download

Share

COinS