U-M Faculty Receive Award to build AI-driven Scientific Computing Resource

A 2023 Defense University Research Instrumentation Program (DURIP) has awarded Michigan Aerospace’s Venkat Raman and Alex Gorodetsky and ME’s Aaron Towne a $2M award for computing infrastructure.

The highly competitive annual DURIP award is administered jointly by the Air Force Office of Scientific Research, the Army Research Office, and the Office of Naval Research through a merit competition. Proposals are awarded based on their foundational science and engineering research relevant to national defense. For the FY 2023 competition, the Service research offices received 522 proposals requesting $171 million in funding. 

In FY23, 147 university researchers were awarded projects totaling $59 million with the goal of enabling universities to perform state-of-the-art research that augments current and develops new capabilities.

The awarded U-M project, “Graphics Processing Units based Computing System for Emerging Hypersonics Solvers,” will bring a powerful computer resource to campus, and enable novel methods to accelerate simulations of complex multi-physics systems in a way that adds a critical level of detail often unavailable in physical experimentation. The cluster machine will use graphics processing units or GPUs, which can carry out large numbers of calculations using a fraction of the power needed by more traditional computers. In particular, GPUs have been central to the vast growth in machine learning and AI technologies.

This project turns machine learning inwards, using such methods for understanding how computer algorithms can be optimized on a given architecture. “We are designing new tools using machine learning models that will enable supercomputers to teach themselves how to solve physics problems more efficiently. These computing resources will enable us to develop and deploy new capabilities for tackling large scale problems of critical societal importance,” Raman states.

The team will initially focus on hypersonics, a field that deals with flight at speeds far exceeding current aircraft capabilities. “This on-campus computing resource will help understand the complex physics found inside high-speed engines, probing molecular level interactions that enable efficient and reliable flight”, said Raman. This award is part of continued investment from DoD in his research program, specifically aimed at enabling hypersonic flight capabilities.

Looking forward, this on-campus resource has broad applications, including the field of computing-driven certification of complex systems. For instance, a 2030 roadmap for aircraft systems emphasizes the use of virtual testing using detailed simulation tools to accelerate the process of certifying new aircrafts and engines. Many such engineering problems span multiple physics domains, requiring different mathematical representations, algorithms, and implementation paradigms.

“We are particularly interested in applications that involve large-scale particle simulations, for example simulations of plasma systems and non-equilibrium flows. Our primary research will target two areas: accelerated Monte Carlo simulation methods and more efficient in-situ lossy compression for fast analytics and data transfer,” comments Gorodetsky.  “The simulations enabled by this resource open exciting new opportunities for data-driven modeling and control of high-speed jets and other turbulent flows”, said Towne.   

Raman and Gorodetsky represent an important focus on simulation technology applications in aviation and aerospace at U-M Aerospace.

A faculty member at U-M since 2014, Raman focuses on the development of computational models for turbulent reacting flows with application to aircraft and scramjet engines, stationary power generation, and synthesis of novel materials. His research group uses high-performance supercomputers and detailed numerical simulations to study the performance of combustion devices. He runs the Advanced Propulsion Concepts Lab at U-M. Raman received the NSF CAREER award in 2008 and was recently named as associate editor of Combustion & Flame Journal.

On faculty since 2018, Gorodetsky’s research interests include developing applied mathematics and computational science to enhance autonomous decision making under uncertainty. He is especially interested in controlling systems, like autonomous aircraft, that must act in complex environments that are often represented by expensive computational simulations. He is the recipient of the John von Neumann Postdoctoral Research Fellowship in Computational Science in 2016 and the Air Force Office of Scientific Research Young Investigator Program in 2018.

Towne has been on the faculty in Mechanical Engineering since 2018. His research develops reduced-complexity models that can be used to understand, predict, and control turbulent flows, using both physics-based and data-driven methods.  He is a recipient of the Air Force Office of Scientific Research Young Investigator Program Award in 2020 and the National Science Foundation CAREER Award in 2022.