The era of artificial intelligence in aerospace is no longer a futuristic concept; it is actively reshaping production engineering workflows today. At a panel discussion at AIAA SciTech Forum 2026 in Orlando, Neil Ashton, Distinguished Engineer and Product Architect, NVIDIA; Thanos Margaritis, Commercial Director, Neural Concept; and Nicolò Vallana, Rotorcraft Technologies Specialist, Leonardo Helicopters, demonstrated how machine learning surrogates are delivering tangible speedups in critical design and manufacturing processes now.
Leonardo Helicopters showcased specific applications where AI surrogate models replaced traditional, slow physics simulations:
- Aeroacoustics. A geometric deep learning model predicts ground-level noise with a 540-fold speedup compared to standard computational fluid dynamics (CFD), allowing rapid compliance checks against noise regulations.
- Manufacturing. An AI model predicts porosity and solidification times for aluminum casting instantly, replacing 24-hour CFD runs and significantly reducing defect rates.
- Air Duct Design. A closed-loop system uses a CFD surrogate to instantly evaluate iterative design changes for cockpit fogging prevention before final high-fidelity validation.
The panelists agreed on key themes.
- Customization. There is no “one-size-fits-all” model; success depends on lightweight, problem-specific architectures tailored to unique physics.
- Data Scarcity. Unlike open-data fields, aerospace relies on proprietary, limited datasets (sometimes as few as 20 simulations), requiring techniques like sensitivity analysis to maximize value.
- Outlook. While design-phase AI is growing, operational use (like flight control) faces stricter regulatory hurdles, with EU certification guidelines expected in 2027. The future lies in hybrid models coupling surrogates for fluids, structures, and acoustics to enable rapid multi-physics simulations.
The panel concluded that AI acts not as a replacement for physics, but as a powerful accelerator that turns days of simulation into minutes.
