What is it about?
In this paper, a novel Physics-Guided Neural Network (PGNN) has been developed which can predict the cyclorotor academic performance such as thrust and power. This model combines with machine learning and physics based aerodynamic model where physics guided information was incorporated into a Neural Network architecture. This PGNN model can accurately predict the cyclorotor performance. This study compares this approach with conventional Neural Network and physics based aerodynamic model to show its superior performance.
Featured Image
Photo by Hyundai Motor Group on Unsplash
Why is it important?
We develop a physics-guided neural network (PGNN) that predicts cycloidal rotor performance by combining experimental data with physics-based aerodynamic insights. This is important for accelerating the design and optimization of emerging eVTOL propulsion systems, where accurate yet fast predictions are critical. Two significant contributions are that: a) the PGNN consistently outperforms both conventional neural networks and physics-based models across diverse configurations, and b) incorporating physics-guided features (analytical thrust and power) improves generalization, making the model robust even with limited experimental data.
Perspectives
Writing this article was an exciting journey as it brought together my passion for rotorcraft aerodynamics and my growing interest in machine learning. This work challenged me to think beyond traditional modeling and explore how physics and data can complement each other in powerful ways. I hope this article encourages others working in aerospace design to explore hybrid approaches like PGNN, especially for applications where both computational cost and accuracy matter. Most of all, I hope it sparks broader conversations about how we can make complex aerodynamic systems more accessible, efficient, and intelligent through thoughtful integration of domain knowledge and modern computational tools.
Nabia Fardin
Oklahoma State University Stillwater
Read the Original
This page is a summary of: Physics Guided Neural Networks Model for Predicting Cycloidal Rotor Performance, January 2025, American Institute of Aeronautics and Astronautics (AIAA),
DOI: 10.2514/6.2025-1447.
You can read the full text:
Contributors
The following have contributed to this page
