Supercomputing an improved rocket

A new type of rocket engine that could improve aircraft, spacecraft, and other transportation systems is being developed by private and public institutions such as NASA and the Air Force Research Laboratory. Rotating detonation rocket engines (RDRE) have the potential to be more efficient and safer than traditional rocket systems. Supercomputer simulations are helping guide their design. 

“We're investigating using detonation as an alternative type of rocket combustion and studying whether we can sustain it to create enough thrust for spacecraft with an engine that is compact and without moving parts. The study will have very important contributions for space research,” said Kiran Bhaganagar of the Laboratory of Turbulence Sensing and Intelligence Systems, and a professor of mechanical, industrial, and aerospace engineering at The University of Texas at San Antonio and Associate Fellow of the American Institute of Aeronautics and Astronautics. 

The Air Force Research Laboratory awarded Bhaganagar in July 2024 a grant to develop a physics-based numerical computer model of RDRE that will simulate the chemistry of fuel and air interactions in combustion along with representations of the detonation dynamics that can propel the rocket.

Bhaganagar is developing the RDRE numerical model using the Lonestar6 supercomputer at the Texas Advanced Computing Center (TACC). The allocation is awarded through The University of Texas Research Cyberinfrastructure, which provides computational support to researchers at all 14 UT System institutions.

"The computational model is computationally intensive as it captures the fine time and length scales of combustion and turbulence processes. Having access to UTRC resources is extremely valuable for transcending the limits of computational capability and performing such a simulation,"  Bhaganagar said.

The principles of rocket design have stayed true for nearly a century to those of the American inventor Robert H. Goddard, who launched the world’s first liquid fueled rocket in 1926. Modern rockets such as the NASA Artemis—with a mission to explore the Moon for scientific discovery and to learn how to live and work on another world—mix fuel and oxygen to achieve combustion, which burns the mixture in a controlled way to produce thrust. 

Rotating detonation rocket engines are an entirely different beast in that they generate thrust through detonation. Think of a firecracker versus a campfire. Explosions might seem patently bad for rockets. But in an RDRE the detonations are tamed by the shock waves generated in combustion, which in the lab have been shown to circle the engine chamber in a predictable way at supersonic speeds. These combustion waves burn the fuel in the chamber, quickly pumping out more thrust than conventional rockets.

What’s more, the detonations are self-sustaining, automatically opening and closing the flaps where the fuel enters the chamber and using a simpler fuel injection system than what’s found in conventional rockets.

"The key part is to make sure that the shockwave survives,” Bhaganagar explained. “The shockwave continuously loses energy. The chemical reactions feed energy to the shockwave, which helps it survive and sustain the detonation." The smooth, circular chambers of RDRE have found success in the lab. 

“More importantly,” Bhaganagar added, “the chamber is a very small space. Instead of having big engines, RDREs take a small space and there are no moving parts. There are few losses. It's a smart way of creating the correct conditions. That's the main difference here.”

Bhaganagar’s lab is scaling up the smaller experimental models that have produced on order of 5,000 pounds of thrust all the way up to 15,000 pounds or more—something that could propel small spacecraft or aircraft.

“Scaling up thrust is where our computational models will play a big role,” Bhaganagar said. “Once we are confident that they model existing computational and laboratory work, we can develop unique sets of data to help guide experimental design in finding the right conditions to take RDRE to the next level.” 

Turbulence remains one of the last unsolved problems of classical physics. The computational model Bhaganagar is developing is numerical and physics-based. It solves fluid dynamics equations called Navier-Stokes equations, which describe the chaotic, turbulent conditions of the detonation-based engine. 

 “We are working with turbulence, shock waves, and detonation. They are like three beasts, and they interact as turbulence accelerates the detonation,” Bhaganagar explained.

The model needs many nanosecond time steps to get a few seconds of results that yield meaningful characteristics of the detonation waves inside the RDRE.

“The computation is extremely expensive and poses a challenge for us," Bhaganagar said. "That's where TACC comes in. TACC has been valuable for us and for the Texas community of scientists. We are awarded almost one million service units of computational allocations every year. Thanks to the resources provided, we were able to perform a substantial amount of preliminary results leading up the Air Force Research Lab project."

In addition, Bhaganagar is exploring the utility of convolutional neural networks—an image recognition technique of artificial intelligence—to solve the Navier-Stokes partial differential equations (PDE) that are front and center of her work. She published research in the Springer Nature publication Machine Intelligence Research, (February 2025) that developed a configurable U-Net architecture trained to solve multi-scale elliptical PDEs.

The work aims to let AI do some of the heavy lifting of computation without losing accuracy. “We are not working on machine learning yet,” Bhaganagar said. “But in the future the data we generate will be an excellent gold mine for training data sets to improve rocket design.”