RESEARCH NEWS
CFAL: Physics-Informed Machine Learning for Fluid Dynamics and Aerodynamics Dr. Michael Kinzel The Computational Fluids and Aerodynamics Laboratory (CFAL) applies machine learning to advance modeling and prediction in complex fluid systems. A primary research focus is the use of Physics-Informed Neural Networks (PINNs) to model atmospheric boundary layers, where sparse measurements and high uncertainty traditionally limit predictive accuracy. By embedding governing physical laws directly into neural network training, PINNs maintain physical consistency while improving predictive capability. CFAL integrates these approaches with classical potential flow methods and modern computational fluid dynamics (CFD), creating hybrid modeling frameworks that enhance both efficiency and fidelity. These tools support a range of aerospace applications, including atmospheric energy harvesting for long- endurance uncrewed aerial systems. A second research thrust involves developing surrogate aerodynamic load models using Gaussian process regression and neural networks. These models replace traditional lookup tables with function- based representations derived from high-fidelity CFD datasets. While they require an initial investment in CFD computation, they significantly improve flexibility, efficiency and uncertainty quantification in engineering analysis. CFAL is expanding this work through collaboration with the University of Southampton to reduce CFD dependence and enable more cost-effective aerodynamic modeling workflows.
Reinforcement Learning and Immersive Aerospace Education Drs. Hao Peng, David Canales Garcia and Morad Nazari Recent contributions include reinforcement learning-based spacecraft attitude estimation and collaborative research on extended reality and AI-enhanced interactive learning environments for aerospace education. These efforts explore how intelligent systems can improve both spacecraft autonomy and engineering education.
6 | AERONEWS 2025 - 2026
Made with FlippingBook - professional solution for displaying marketing and sales documents online