A Geometric and Topological Analysis of Atomization from Large-Scale Simulations

Brendan Christensen (Mechanical & Industrial Engineering Dept, MSU)

Abstract: 

Atomization is the process of liquid breaking up into small droplets. This phenomenon is crucial in many fields, such as fuel injection, spray painting, pharmaceutical manufacturing, and firefighting. However, understanding the underlying physics remains a challenge. Advances in computational power have made large-scale simulations of atomization possible, but the high cost and vast quantity of data generated have limited new insights. This work aims to address these challenges by using the Atomization Simulation Statistics Extraction Tool (ASSET) to efficiently track and analyze liquid breakup events during simulations. ASSET identifies and records key data, such as the gas-liquid interface geometry and local flow conditions, at the moments of breakup. These data are stored in a compact database that offers a new perspective on atomization. The present work aims to assign topological data-markers to these databases of extracted geometries. The goal is to use these databases as the foundation for a reduced-order atomization model that can predict the breakup of a given droplet without fully resolving the physics.

Applications of the Weighted Euler Characteristic Transform in Atomization Simulations

Alex McCleary (Computer Science Dept., MSU)

Abstract: 

In this talk, we explore the Weighted Euler Characteristic Transform (WECT), a general-purpose topological invariant that provides a rich framework for data vectorization. We will focus on an application of the WECT in the context of atomization simulations, where simulated data is represented as triangular meshes. By applying the WECT to these meshes, we obtain a vector representation of the data, which we use to construct a vector database that allows us to efficiently “simulate the simulations”—predicting simulation outcomes without rerunning computationally expensive simulations. Additionally, we will introduce a more efficient algorithm for the computation of the WECT, optimizing its use for large-scale, high-resolution data. This talk will highlight both the flexibility of the WECT as a tool across different types of data and its practical applications in improving simulation workflows.