Extracting and analyzing atomization physics from high-fidelity simulations

Abstract

Liquids break apart into droplets in countless natural and industrial processes, yet predicting when and how this breakup occurs remains a fundamental challenge. Understanding these mechanisms can lead to more efficient spray systems and deeper insights into environmental flows. However, studying atomization at a fundamental level requires analyzing immense amounts of simulation data, often reaching tens to hundreds of terabytes. To overcome this barrier, this research develops a method for extracting key information from large-scale atomization simulations, significantly reducing data storage and processing requirements while capturing the physics of liquid breakup. This method was applied to two practical cases: the breakup of water and shear-thinning liquids in high-speed gas flows, and the behavior of electrically charged liquid jets. In both cases, the tool enabled the identification of previously unknown breakup mechanisms. Additionally, a technique was developed to extract and vectorize liquid shapes and surrounding flow fields, allowing for more advanced data analysis. This approach will enable fast comparisons of liquid shapes and flow fields, providing a foundation for a reduced-order model. Ultimately, this research establishes a new way to access data from high-fidelity simulations with the intent that the data be used to improving predictive models for engineering and scientific applications. By providing access to data from expensive, large-scale simulations and uncovering insights from the underlying physics, these findings contribute to the advancement of multiphase flow research and the design of next-generation spray systems.