U-M leads new DOE-funded computational center focused on next-generation hypersonic flight
How much faster could engineering progress with an artificial intelligence oracle that could answer any physics question?
Such a machine is the big picture aim of the newly formed Center for Prediction, Reasoning and Intelligence for Multiphysics Exploration, or C-PRIME, led by the University of Michigan and funded by the U.S. Department of Energy's National Nuclear Security Administration.
While physics is governed by many known equations, it's hard to get from those equations to answers about how real-world objects will behave-for instance, the swirls of fuel and air inside a complex engine or the precise wind resistance over the surface of a vehicle. In theory, it's all knowable, building up from the molecular level, but the calculations are too big to actually perform.
While an AI approach can't attack that problem directly, an AI agent could build physics models based on known equations that it uses to generate trustworthy data. It could then use that data to produce simplified yet accurate models for specific physics problems, which would feed into engineering design of complex devices.
"The notion is that we, as humans, should provide certain concepts we trust-Newton's laws or E=mc^2. The machine then composes more complex ideas from these basic building blocks," said Venkat Raman, director of C-PRIME and the James Arthur Nicholls Collegiate Professor of Engineering.
"Because we trust these building blocks, we can-to a large extent-trust engineering concepts that are composed from them."
However, formally establishing this trust, known as verification and validation, is in itself a complex challenge, which is at the core of the project. The sequences of simulations designed by the AI agents will run on some of the world's largest supercomputers to discover the inner workings of propulsion systems behind hypersonic flight-five times faster than the speed of sound. The team will focus on rotating detonation combustors, which are becoming a critical technology for hypersonic flight.
Rotating detonation combustors can be used for propulsion-in rockets, air-breathing engines or satellite thrusters-or energy conversion, such as in gas turbines that generate electricity. They have the potential to be very efficient, roughly 25% better than conventional combustion, but maintaining their burn is a nuanced endeavor. A series of explosions run around a ring, and the resulting shockwave compresses and ignites the fuel-air mixture at each fuel injection point in sequence.
"AI and hypersonics are critical to national security and U.S. scientific leadership, and we're committed to developing technologies and talent to move both fields forward," said Karen A. Thole, the Robert J. Vlasic Dean of Engineering. "This federal investment enables our researchers to bring together expertise in physics, computer simulation, AI and machine learning to push the boundaries of what's possible and develop tomorrow's AI-savvy workforce in the process."
Student researchers on the project will draw on the University of Michigan's Ph.D. in Scientific Computing-the nation's first, established in 1988-administered by the Michigan Institute for Computational Discovery and Engineering, or MICDE.
The project is divided into five research thrusts:
- Physics and data: This effort covers foundational physics, developing models and honing them with experiments that fill holes in the existing data, with a focus on how materials mix and react. Led by Eric Johnsen, center co-director, professor of mechanical engineering and director of the scientific computing Ph.D. program.
- Verification, validation and uncertainty quantification: With the goal of ensuring the accuracy and reliability of the computer models, this thrust digs into how assumptions and simplifications in the physics models affect predictions. Led by Alex Gorodetsky, associate professor of aerospace engineering.
- Exascale supercomputing architecture: This effort optimizes the models to take full advantage of powerful supercomputers and lays groundwork for building next-generation supercomputers optimized for AI. Led by Reetuparna Das, professor of computer science and engineering.
- Machine learning: This team will develop machine-learning-based tools that will accelerate computation of complex physics, using data generated by autonomously acting AI agents. Led by Karthik Duraisamy, professor of aerospace engineering and director of MICDE.
- AI-based integration: Based on "physics composition"-the formal approach for integrating different physics equations-this team will build the AI agents responsible for coding and simulation. Raman leads this thrust.
In addition, specially designed laboratory experiments will test the accuracy of the AI-based combustor design, to be conducted at U-M by Mirko Gamba, professor of aerospace engineering, and Carolyn Kuranz, professor of nuclear engineering and radiological sciences.
"Through our research, and the education of the next generation of researchers, we have the opportunity to shape the field on a large scale," Johnsen said. "In particular, we need to ensure that our trainees-undergraduate and graduate students and postdoctoral researchers-understand how to leverage AI resources in their research because their success after they leave Michigan will depend on how well they do this."
David Etim, federal program manager in the National Nuclear Security Administration's Office of Advanced Simulation and Computing and Institutional Research & Development, spoke highly of the new center, which is part of the fourth phase of NNSA's Predictive Science Academic Alliance Program.
"This center with its focus on AI-driven solutions for complex physics problems aligns perfectly with PSAAP's mission to advance high-fidelity predictive simulations," Etim said. "We eagerly anticipate the groundbreaking contributions C-PRIME will make in areas critical to national security, particularly in next-generation hypersonic flight and exascale computing, further strengthening the program's impact."
C-PRIME builds on U-M's prominent leadership in computational science and engineering, which is anchored by MICDE. U-M is also home to a $15 million Strategic Partnership and Accelerated Research Collaboration with Los Alamos National Lab, which is coordinated by MICDE and brings together Los Alamos staff scientists and U-M researchers. Additionally, the university is partnering with LANL on a $1.25 billion facility for high-performance computing and AI research in Michigan.
C-PRIME includes a total of 13 co-investigators from U-M across four departments, as well as a co-investigator from Princeton University. Researchers at Sandia, Los Alamos and Lawrence Livermore national laboratories will be collaborating with the center.
Raman is also a professor of aerospace engineering and mechanical engineering. Thole is also a professor of mechanical engineering and aerospace engineering. Duraisamy is also a professor of mechanical engineering and nuclear engineering and radiological sciences.
