A new artificial intelligence system reduces the preparation time for complex space maneuvers by a factor of 100—from months to seconds. This study focuses on calculating the jet impact of engines: a phenomenon that can damage the aircraft’s own structures during the operation of propulsion systems.
New calculation system
When a rocket engine ignites in a vacuum, the gas expands instantaneously in all directions—this places severe thermal and mechanical stresses on the spacecraft’s structure. To predict such effects, engineers traditionally ran millions of simulations, a process that took months.
The new system differs from conventional computing accelerators. It is based on NVIDIA PhysicsNeMo, a framework for developing AI models that combine the laws of physics with simulation data. It embeds the laws of physics directly into the model’s architecture.
As a result, AI doesn’t just process data; it understands natural patterns and operates within them. “We are transforming the role of simulation—not just speeding it up, but expanding the range of problems that engineers can confidently solve,” said Vera Yang, president and co-founder of Flexcompute.
Practical implications
A key advantage of the system is its built-in accuracy assessment. For each prediction, the model calculates a confidence level, which helps engineers understand the limits of the answer’s reliability.
This is critically important in real-world conditions, as during in-orbit docking or the execution of space robotics maneuvers, an error of just a few centimeters can jeopardize the mission.
Faster calculations open up broader engineering possibilities. More accurate modeling allows for the design of structures with a smaller safety margin and, consequently, a lower mass. At the same time, a more detailed calculation of jet interaction makes it possible to save fuel and extend mission duration.
According to Fahad Khan, Director of AI at Northrop Grumman, the partnership with Flexcompute and NVIDIA enables the company to accelerate project delivery at the pace customers require.
According to interestingengineering.com
