Reviewed by Lexie CornerFeb 18 2025
Researchers from KTH Royal Institute of Technology and the Barcelona Supercomputing Center tested an artificial intelligence (AI) system designed to enhance the efficiency of experimental technologies for airflow regulation on wing surfaces.
The AI control system zeroes in on one particularly dangerous aerodynamic phenomenon known as flow detachment, or turbulent separation bubbles. Image Credit: David Callahan
AI in aircraft systems could help mitigate mid-air stalling and altitude loss. In a recent study, an international research team evaluated a machine-learning approach for turbulence management.
The findings indicate that these techniques are more effective when integrated with deep reinforcement learning (DRL), which enables the system to adapt to airflow dynamics based on prior data.
Ricardo Vinuesa, a researcher in fluid dynamics and machine learning at KTH, stated that the AI control system is designed to address flow detachment, also known as turbulent separation bubbles, a critical aerodynamic issue.
How Planes Stall
Flow detachment presents a significant aerodynamic risk. For an aircraft to maintain lift, airflow must remain slower beneath the wings and faster above them. The air passing over the wing surface must adhere to the wing shape, a condition known as "attachment."
According to Ricardo Vinuesa, when airflow over the wing surface no longer conforms to the wing shape and instead separates, it creates a turbulent recirculating region, or "separation bubble," which can lead to stalled airflow and reduced lift.
This usually occurs when the wing is at high angle of attack, or when the air slows down due to increasing pressure. When this happens, lift decreases, and drag increases, which can lead to a stall and make the aircraft harder to control.
Ricardo Vinuesa, Fluid Dynamics and Machine Learning Researcher, KTH Royal Institute of Technology
The Difference AI Makes
The researchers reported a 9 % reduction in the size of turbulence separation bubbles.
The study examined the effectiveness of AI in controlling synthetic jets—experimental devices that pulse air through a small aperture in the wing surface. While still in the experimental phase, these technologies are considered a potential complement to physical airflow management features such as vortex generators, which help maintain aerodynamic stability.
Previously, it was assumed that synthetic jet pulses would operate at fixed intervals. However, the study found that periodic activation reduced turbulence separation bubbles by only 6.8 %.
This study highlights how important AI is for scientific innovation. It offers exciting implications for aerodynamics, energy efficiency, and next-generation computational fluid dynamics.
Ricardo Vinuesa, Fluid Dynamics and Machine Learning Researcher, KTH Royal Institute of Technology
Journal Reference:
Font, B., et al. (2025) Deep reinforcement learning for active flow control in a turbulent separation bubble. Nature Communications. doi.org/10.1038/s41467-025-56408-6