Aerodynamic parameter identification method based on physics-informed radial basis function-deep neural networks.

Area of Science:

• Aerospace Engineering
• Computational Fluid Dynamics
• Machine Learning

Background:

• Accurate estimation of aerodynamic parameters is crucial for flight dynamics and control.
• Real-world aerodynamic parameters often deviate from nominal values due to various factors, necessitating robust identification methods.

Purpose of the Study:

• To develop and validate a novel aerodynamic parameter identification method for precise estimation of perturbations.
• To enhance the fitting capability and accuracy of aerodynamic parameter identification using advanced neural network architectures.

Main Methods:

• Proposing a physics-informed radial basis function-deep neural network (PIRBF-DNN) for aerodynamic parameter identification.
• Utilizing an integration-based loss function within the PIRBF-DNN to precisely estimate parameter perturbations.
• Employing a radial basis function-deep neural network (RBF-DNN) structure to improve the network's fitting capabilities.

Main Results:

• The PIRBF-DNN method demonstrated precise estimation of aerodynamic parameter perturbations in simulation.
• Validation across different scenarios confirmed the effectiveness of the proposed identification technique.
• Comparative analysis showed superior performance against existing physics-informed neural network (PINN) based methods.

Conclusions:

• The PIRBF-DNN offers a powerful and accurate approach for identifying aerodynamic parameter perturbations.
• This method enhances the reliability of aerodynamic models by accounting for real-world parameter variations.
• The study highlights the potential of integrating physics-informed neural networks with RBF-DNNs for complex system identification.