Tracking the fiery path of particles: UF leads combustion research

A research team led by the University of Florida is using artificial intelligence and high-performance computing to explore swarms of airborne particles to solve combustible mysteries, a study that may lead to more efficient engines and better combustion in power plants.

What can we learn from fast-and-furious particles burning in a group? 

Armed with a $11 million cooperative agreement award from the U.S. Department of Energy’s National Nuclear Security Administration, or NNSA, researchers from UF’s Herbert Wertheim College of Engineering, Purdue University and Georgia Tech are examining the physics of how particles and droplets burn as a group, which also can lead to improved safety measures.  

“It can improve safety, including dust explosions in mines and silos,” said the project’s principal investigator, Sivaramakrishnan Balachandar, UF’s Newton C. Ebaugh professor in Mechanical and Aerospace Engineering, or MAE. 

In simple terms, the team is developing powerful computer models that can predict what's happening in a complex cloud of particles moving very fast, as well as how they're interacting with each other and their surroundings. 

“In this project we will be modeling a swarm of burning particles and droplets, such as in a jet combustion engine or in a coal-mine dust explosion,” he said. 

With this five-year project, scientists with UF, Georgia Tech and Purdue plan to establish a Center for Multiscale Modeling of Multiphase Combustion at UF. The project stretches across different departments at UF’s Herbert Wertheim College of Engineering, as well as Georgia Tech and Purdue. 

The National Nuclear Security Administration will task scientists to solve complex problems using the world's largest supercomputers. Here, the team will run AI-assisted simulations on exascale computing platforms that are more powerful than most supercomputers and designed to tackle complex scientific and engineering problems that require immense computational power.  

“We propose solving the problem of phase change and group combustion of a dense cloud of particles under high-speed and detonation conditions at an unprecedented level of physical detail with a state-of-the-art multiscale modeling framework that combines the powers of scientific computing and machine learning,” the researchers wrote in their proposal to the NNSA’s Predictive Science Academic Alliance Program. 

Phase change is the transition of matter — solids to liquids to gas, for example. 

“We will develop better AI-informed group combustion models,” Balachandar said. “These models can be used by NNSA Labs (Los Alamos, Lawrence Livermore and Sandia National Laboratories) in their simulations of varying energy, environment and national security problems.” 

Those models, he said, may also be used by private companies such as General Electric and Pratt & Whitney.  

Swarms of particles burn differently than individual particles. All burning droplets and particles compete and fight for oxygen, but they also help each other by heating the environment, Balachandar said. 

“By focusing on the fundamentals and better understanding and modeling of group combustion, we plan to impact a wide variety of applications,” he added. “When you see a volcanic eruption, for example, you have a hot gas carrying trillions of little particles. Something blows up, and it hurls a lot of dust and things like that. Sometimes you want to mitigate it, sometimes you want to use it to our advantage.” 

In addition to Balachandar, the team consists of James Fairbanks, Ph.D., UF MAE; Alina Zare, UF Electrical and Computer Engineering, or ECE; Joel Harley, Ph.D., UF ECE; Ryan Houim, Ph.D., UF MAE; Thomas L. Jackson, Ph.D., UF MAE; Nam-Ho Kim, Ph.D., UF MAE;  Spencer Bryngelson, Ph.D., Georgia Tech; and Daniel Guildenbecher, Ph.D., Purdue.