UC Irvine

The continuous increase in the performance of high-power energy systems calls for thermal management systems capable of dissipating large concentrations of heat. Phase change cooling strategies utilize latent evaporation energy to absorb substantial amounts of heat. Although phase change cooling has considerable potential in thermal management, little research has been done regarding the optimization of these systems. The extreme volatility within phase change processes poses a challenge for researchers to obtain measurements with any reasonable uncertainty. Essential to altering the heat transfer performance is understanding nucleation dynamics, which involves quantifying individual nucleation features as well as their correlation to system properties. The complex dynamics within these systems prevent accurate extraction of nucleation statistics using conventional methods. Machine learning enables researchers to automatically quantify these system dynamics. For instance, a 10 second dataset composed of high-speed boiling images contains at least one million bubble features, where an experiment will exceed millions of bubble features. To accurately extract data from these systems requires an extensive amount of time and an in depth understanding of the underlying physics. Inspired by this challenge, this thesis reports an autonomous vision-based framework capable of measuring nucleation dynamics at a high-temporal, microscopic scale.

We employ a modular framework consisting of an artificial intelligence (AI) based object detection and object tracking to study nucleation dynamics occurring within phase change systems including pool boiling, dropwise condensation, and film wise condensation. Vital nucleation statistics are extracted by tracking individual droplets within phase change systems using sequential high-speed, high-resolution images.