Development of a Data-driven Framework for High-fidelity Boiling Heat Transfer Prediction
Pool boiling is an efficient heat transfer mechanism with strong potential for thermal management applications. However, accurate prediction remains challenging because of the complex dynamics of bubble nucleation, growth, and departure. This research develops a data-driven framework for pool boiling analysis by combining artificial intelligence with fluid mechanics.
The work consists of two parts. The first part applies image recognition, object tracking, and numerical simulations to extract bubble dynamics data from a microscopic perspective and analyze bubble cluster behavior. The second part focuses on the development of Agent-Based Boiling Analytics (ABBA), a lightweight, data-driven simulation tool for pool boiling. Built in MATLAB, ABBA uses an agent-based approach to model bubble nucleation, growth, coalescence, and departure without solving full fluid dynamics equations. Bubble behaviors are represented using empirical correlations, and preliminary simulations have shown good agreement with classical predictions of heat transfer and bubble dynamics.
Highlights:
1. Machine learning-based image recognition and object tracking extracted statistically meaningful bubble dynamics from pool boiling datasets.
2. The effect of surface characteristics on macroscopic heat transfer was interpreted from microscopic bubble dynamics.
3. A data-driven, agent-based framework captured complex bubble and heat transfer behavior with low computational cost.
Development of a Data-driven Framework for High-fidelity Boiling Heat Transfer Prediction
Pool boiling is an effective heat transfer mechanism, widely studied for its ability to achieve high heat fluxes while maintaining excellent energy efficiency. These features make pool boiling a promising solution for complex thermal management challenges in modern industrial applications. With rapid advancements in technologies such as IoT and AI, cooling demands for high-performance computing systems, including data centers, have increased significantly. As a result, there is a growing demand for efficient heat dissipation methods to maintain optimal temperatures and ensure the stable operation of advanced servers and electronic devices. Efficient thermal management is essential not only for reducing energy consumption but also for improving equipment reliability and lifespan. Recent developments in engineered surface structures have shown great potential for enhancing boiling heat transfer and addressing rising cooling needs. This study focuses on optimizing the design of biphilic surfaces —including coating types and heterogeneous wettability pattern geometries—by integrating experiments with a machine learning approach (Bayesian optimization). By optimizing these parameters to enhance the nucleate heat transfer coefficient and critical heat flux.
Highlights:
1. Machine-learning-based optimization of biphilic surface geometry for enhanced boiling heat transfer.
2. High-speed imaging analysis of bubble dynamics to reveal the effects of hydrophilic–hydrophobic pattern distribution.
Numerical Study of Subsurface Evaporation and Wetting Behaviors in Nucleate Boiling
Complex multi-bubble dynamics play an important role in nucleate boiling. However, few studies can explain the mechanism behind it clearly. A unified theoretical model has not yet been established because of the highly coupled multi-physics nature of the boiling process, and current studies still rely heavily on empirical or semi-empirical correlations. Therefore, to understand it better, both single-bubble and multiple-bubble dynamics are investigated in this work. The heat transfer characteristics of nucleate boiling are governed by complex multi-bubble dynamics. Compared with single-bubble cases, multi-bubble simulations are more representative of realistic boiling conditions and play a crucial role in heat transfer enhancement, dry spot formation, and the development of CHF. In this study, a numerical investigation of multi-bubble dynamics in nucleate boiling is conducted based on the OpenFOAM framework using the volume-of-fluid (VOF) method, which accurately captures gas–liquid interface evolution and provides detailed insight into microscale interfacial behavior and local thermal characteristics.
Highlights:
1. The simulation of multi-bubble dynamics is conducted via OpenFOAM.
2. The mechanism of multi-bubble interaction in nucleate boiling is investigated.
3. The effect of the distance between the cavities on bubble dynamics is determined.
Data-driven Elucidation of Heat Transfer Mechanisms of Heterogeneous Boiling
Boiling heat transfer, which is arguably the most efficient heat transfer scheme, is critical to thermal management of power-intensive applications such as data centers and nuclear reactors. As the surface heat flux increases, heterogenous boiling goes through the distinctive regimes of isolated bubbles, slugs and columns (fully developed) before triggering the so-called boiling crisis (i.e., the dryout condition), underneath which are interacting processes of bubble nucleation, growth and release. It is the latter that contributes the most to the inherent complexity of boiling heat transfer, which still lacks an accurate physical description. In this work, we aim at developing a comprehensive understanding of the relationship between bubble behavior (three-phase contact line dynamics) in nucleate boiling and physical and chemical surface characteristics using a neural-network-based approach.
Multiscale Modeling of Complex Phase-change Systems
For a multiphase fluid systems such as heterogeneous boiling, new unknown interfacial physics abound as one moves from macroscale to microscale. The presence of solid surface introduces even more complexity, whose full resolution calls for an ‘organic’ modeling approach that facilitates transfer of information across different time and length scales. This research consists of three interdependent numerical efforts to elucidate heat and mass transport in pool boiling: macroscopic full-scale direct numerical simulation of nucleate boiling, mesoscopic diffuse-interface account of contact-line dynamics, and microscopic molecular dynamics description of heterogenous nucleation. The results are expected to shed new light on the underlying multi scale nature of phase-change phenomena.
Previous projects (prior to 2020)
Copyright©2020-2026 Interphase TRansport Lab, the University of Tsukuba, Japan. All rights reserved.
ITR Lab 