When aerospace vehicles travel at hypersonic speeds, or materials face temperatures too high to replicate in a physical experimental lab, how can engineers be sure their designs will hold up?
That’s the challenge driving a new $16 million cooperative agreement awarded by the National Nuclear Security Administration (NNSA), an agency within the U.S. Department of Energy, to establish the SAGEST Predictive Simulation Center at the University of Virginia (UVA). SAGEST — short for Stochastic Simulations of Ablative Geometries with Error-Learning in Space and Time — will develop simulation tools that give scientists confidence in exploring extreme physical conditions that are too difficult or costly to test directly.
The multi-university Center will be led by UVA and directed by Xinfeng Gao, professor of mechanical and aerospace engineering at UVA’s School of Engineering and Applied Science. Gao is an expert in high-performance computing, computational fluid dynamics and data assimilation, and is known for her work developing advanced algorithms for complex systems.
The core goal of this Center is to build simulations you can trust.
“The core goal of this Center is to build simulations you can trust,” Gao said. “We want to develop predictive technologies that don’t just generate results but give scientists and engineers confidence in those results, particularly under extreme conditions.”
Hypersonics is A Perfect Use Case
Although the SAGEST team will apply its tools to aerospace and hypersonic flight, a domain where vehicles travel at a Mach number greater than 5 (i.e., faster than 4,000 miles per hour), the research is not just about predicting how any one vehicle or system will perform. Rather, hypersonic conditions provide an ideal proving ground for rigorous, adaptive simulation. These systems will be pushed to their limits, making them a valuable test case whose lessons can translate to other complex fields such as energy, medicine, materials and manufacturing.
The central innovation of this work lies in how it layers different levels of computational precision to balance accuracy and efficiency in its predictions while quantifying uncertainty.
Near the surface of a high-speed vehicle — where heat, pressure and material loss are most intense — Gao’s research team uses high-fidelity solvers, which are specialized computer programs designed to calculate complex physical behaviors with great accuracy. These solvers model physics at a fidelity closer to “first principles,” meaning they are based directly on the underlying laws of physics rather than approximations.
For example, individual molecules may be represented to capture critical details like how shockwaves interact with the surface and cause issues such as ablation, when material is worn away or vaporized due to extreme heat and pressure. However, the computational cost of simulating these kinds of conditions is extraordinary. Overall, less than an inch cubed of the complete flow field can be solved with the high-fidelity solver.
Elsewhere, lower-fidelity solvers can be used to simulate the broader flow with less computational cost. In the low-fidelity solvers, approximate models are used to represent the true physics with reduced accuracy in regions where extreme precision is less critical or the risks are lower. Gao’s innovation layers the high-fidelity and low-fidelity solvers so that its models can efficiently deliver accurate predictions.
You don’t need to model every inch of space with the same level of detail.
“You don’t need to model every inch of space with the same level of detail,” Gao explained. “You need to know where the fine details matter, and how those details interact with and influence the larger system. That’s where the predictive power comes from.”
What makes Gao’s approach especially powerful is how these different layers of modeling communicate. The high-fidelity and low-fidelity solvers exchange information continuously, allowing the simulation to adapt in real time and update predictions across the entire system. Classical numerical algorithms are augmented by artificial intelligence to minimize errors in the process. This dynamic coupling ensures that fine-scale physical effects are accurately represented while still maintaining a practical, scalable simulation framework.
It’s a method Gao describes as “adaptive mesh, model and algorithm refinement,” with the high- and low-fidelity solvers coordinated by an error-learning framework designed to create predictions both efficient and trustworthy. Within that framework, Gao and her team use state-of-the-art practices for high-performance computing, uncertainty quantification, artificial intelligence and data management, all of which support the simulation on DOE exascale computer resources that are roughly one million times more powerful than a smartphone. The result is a multiscale simulation engine capable of handling vast complexity while preserving computational efficiency.
UVA’s Investment in Hypersonics Made a Difference
The SAGEST Predictive Simulation Center reflects UVA Engineering’s deep and growing strength in hypersonics. Over the past several years, the school has made strategic hires and programmatic investments to build one of the nation’s leading hypersonics teams. The recent hiring of Professor Gao substantially strengthened the team's capability for computational research.
“This award is a testament to Dr. Gao’s vision and leadership,” said Richard W. Kent, chair of the Department of Mechanical and Aerospace Engineering and Frederick Tracy Morse Professor. “It also reflects the strength of the entire ecosystem we’ve built around hypersonics: a collaborative, interdisciplinary environment that’s equipped to lead in predictive science, high-performance computing and hypersonics all at once.”
The Center also draws on UVA’s growing strength in artificial intelligence, including tools that help simulations learn from data, adapt across scales and produce faster, more confident predictions.
A Model for Simulation and Education
Headquartered at UVA Engineering, the SAGEST Center is also a training ground for interdisciplinary research and collaboration. Gao will lead a broad multi-university team that includes faculty and students from UVA Engineering and UVA’s School of Data Science as well as collaborators from the University of Utah and The Ohio State University who bring specialized expertise in communication and data movement for exascale computing. Additional team members are from the University of Minnesota and the University of Iowa. Together with researchers from DOE/NNSA laboratories, faculty from multiple institutions are building a shared language and framework for approaching complex systems across departments, skillsets and academic cultures.
Students working within SAGEST will gain experience across fields, combining deep technical knowledge with systems-level thinking. Gao emphasizes that this kind of cross-training is essential, not just for future research, but for solving real-world problems that don’t fit neatly into disciplinary boxes.
Her philosophy also extends beyond academia. Gao fosters a research culture where mentorship, collaboration and personal growth matter just as much as technical achievement.
“I want my students to look back in five or six years and say, ‘That was one of the best times of my life,’” she said. “They should walk away not just with a degree, but with a sense of purpose, clarity and connection to the larger research community.”
Predictive Tools With Broad Impact
While the Center’s immediate focus includes aerospace use cases, its simulation platform can be applied to any field where physical experiments fall short. This includes complex energy systems, advanced manufacturing, materials development and biomedical modeling. For the NNSA, this research helps advance science-based modeling and simulation that supports nuclear safety and security, including efforts in nonproliferation, disarmament and maintaining the safety and reliability of the nation’s defenses.
Ultimately, Gao sees the Center as advancing the tools engineers can use to design with confidence from the outset, particularly in cases where direct testing is limited or not feasible.
“If we can give people simulation tools that are reliable, accessible and provide a measure of trustworthiness,” she said, “we can speed up discovery, reduce costly trial-and-error, and support safer, more effective systems in every part of society.”