Particle.news
Download on the App Store

IBM and Dallara Team on Physics-Based AI to Speed Aerodynamic Design

The partners plan to validate results with wind‑tunnel and track data to expand models to more driving scenarios.

Overview

  • A collaboration announced Thursday between IBM and Dallara applies physics-based AI to aerodynamic design and begins exploring quantum methods.
  • In an LMP2-style rear diffuser test, the AI evaluated multiple designs in about 10 seconds and matched the optimal choice found by CFD that took hours.
  • The model learned from Dallara’s validated high‑fidelity simulation data, with real‑world checks to follow using wind‑tunnel and on‑track measurements.
  • Scaled to hundreds of geometry options, the speedup could cut days of screening to minutes, letting engineers compare more designs earlier in development.
  • The work builds on IBM’s Gauge‑Invariant Spectral Transformers research detailed in an April 20 arXiv preprint and presented April 26 at ICLR, and Dallara says gains could carry over to passenger cars and aerospace.