Bridging The Gap:Vision-Inspired Two-Phase Heat Transfer Analysis

ByYoungjoon Suh Doctor of Philosophy in Mechanical and Aerospace Engineering University of California, Irvine, 2022 Professor Yoonjin Won, Chair

AbstractLiquid-vapor phase-change phenomena have been critical to maintaining sustainable and habitable environments on Earth for countless millennia, and are continuing to play central roles in present-day’s industries with ever growing presence. Among different types of phase-change processes, boiling and condensation are two of the most widely used in both domestic and industrial applications. Central to the mechanistic understanding of the thermofluidic processes governing the phase-change phenomena is the rapid and high-fidelity extraction of interpretable physical descriptors from the highly-transient nucleation behaviors. However, extracting quantifiable measures out of dynamic objects with conventional imaging technologies poses a challenge to researchers. This thesis focuses on addressing the fundamentally weak connection between phase-change heat and mass transfer and nucleation statistics available in visual data streams. We outline core ideas of current artificial intelligence (AI) technologies connected to thermal energy science to illustrate how they can be used to push the limit of our knowledge boundaries about boiling and condensation physics. The comprehensive review offers insight into the role of recent advances in AI and computer vision in advancing modern boiling and condensation research. Based on foundational literature analysis and problem definition, the remainder of the thesis proposes various AI-based solutions for connecting visual data streams with heat and mass transfer performances at the device and system level. First, we introduce a data-driven learning framework that correlates high-quality imaging on dynamic bubbles with associated boiling curves. The framework leverages cutting-edge machine learning models including convolutional neural networks and object detection algorithms to automatically extract both hierarchical and physics-based features. By training on these features, our model learns physical boiling laws that statistically describe the manner in which bubbles nucleate, coalesce, and depart under boiling conditions, enabling in situ boiling curve prediction with a mean error of 6%. Our framework offers an automated, learning-based, alternative to conventional boiling heat transfer metrology. Next, we demonstrate an intelligent vision-based framework called Vision Inspired Online Nuclei Tracker (VISION-iT), which unites classical thermofluidic imaging techniques with deep learning to fundamentally address the challenge of extracting high-fidelity interpretable physical descriptors for the highly-transient two-phase processes. We introduce and discuss the detailed construction, algorithms, and optimization guidelines of individual modules so that the framework can easily be adjusted to custom datasets. The concepts and procedures that we propose is transferable, and thus can benefit a broader audience dealing with similar problems. Finally, VISION-iT is deployed in practical phase-change heat transfer analysis applications. For boiling applications, the combined efforts of materials design, deep learning techniques, and data-driven approach shed light on the mechanistic relationship between vapor/liquid pathways, bubble statistics, and phase change performance. For condensation applications, the data-centric analysis enabled by VISION-iT conclusively shows that contrary to classical understanding, the overall condensation performance is governed by a key trade-off between heat transfer rate per individual droplet and droplet population density. Our vision-based approach presents a powerful tool for the study of not only phase-change processes but also any nucleation-based process within and beyond the thermal science community through the harnessing of big data.