Efficient Extraction of Atomization Processes from High-Fidelity Simulations

TL;DR

This paper introduces a method to extract and analyze atomization processes from large high-fidelity simulation datasets, aiding the development of efficient low-fidelity models for spray prediction.

Contribution

It presents a droplet physics extraction technique and a Neo4j database approach to analyze atomization in high-fidelity simulations, which was previously limited by data size.

Findings

Robust quantitative description of atomization process

Detailed local flow field data for model development

Efficient analysis of large simulation datasets

Abstract

Understanding the process of primary and secondary atomization in liquid jets is crucial in describing spray distribution and droplet geometry for industrial applications and is essential in the development of physics-based low-fidelity atomization models that can quickly predict these sprays. Significant advances in numerical modelling and computational resources allow research groups to conduct detailed numerical simulations and accurately predict the physics of atomization. These simulations can produce hundreds of terabytes of data. The substantial size of these data sets limits researchers' ability to analyze them. Consequently, the process of a coherent liquid core breaking into droplets has not been analyzed in simulation results even though a complete description of the jet dynamics exists. The present work applies a droplet physics extraction technique to high-fidelity simulations to track breakup events and data associated with the local flow. The data on the atomization process are stored in a Neo4j graphical database providing an easily accessible format. Results provide a robust, quantitative description of the process of atomization and the details on the local flow field will be useful in the development of low-fidelity atomization models.