Key results
Result highlights
- Processing time cut from hours to seconds
- Transition from 2D to 3D urban simulations
The story
An aerospace engineering research laboratory focused on fluid mechanics and urban airflow simulations for climate resilience.
Traditional computational fluid dynamics methods forced a trade-off between accuracy and speed, restricting models to 2D planes that could not capture complex turbulence. Scaling to 3D city environments required computational resources that made real-time analysis and rapid design iteration impossible.
The team developed a framework using NVIDIA PhysicsNeMo that combines generative AI with physics-informed machine learning. This system utilizes GPU acceleration to transition simulations from 2D limitations to fully interactive 3D environments capable of modeling entire metropolitan areas. The platform processes complex turbulence data on multi-node infrastructure to visualize wind patterns and pollutant hotspots in real time.
Quotes
“Aerodynamics and turbulence are unsolved problems of physics. The answer must be in the data, so we were inspired and intrigued by using AI methods to really interrogate this data, and that’s the type of solutions we have been obtaining to develop very efficient frameworks.”
“To achieve city-scale 3D modeling with actionable results for urban planning, we needed computational power far beyond traditional CPU-based approaches. We recognized that GPU acceleration was essential to make high-resolution 3D urban flow reconstruction practical, rather than purely theoretical.”
The company
University of Michigan
umich.eduPublic research university with comprehensive academic and graduate programs.
Scope & timeline
- 10x faster model training
Implementation partner
Finnish Center for Artificial Intelligence (FCAI)
fcai.fiProvided hands-on engineering assistance for NVIDIA frameworks to the VinuesaLab research team.