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

Abstract

This paper investigates the perturbations estimation between the real and nominal aerodynamic parameters. To address this issue, this study proposes an aerodynamic parameter identification method based on the physics-informed radial basis function-deep neural network (PIRBF-DNN). PIRBF-DNN utilizes an integration-based loss function to achieve precise estimation of aerodynamic parameters perturbations and adopts a radial basis function-deep neural network (RBF-DNN) structure to enhance fitting capability of the network. The proposed identification method is validated through simulation in different scenarios and comparison with other aerodynamic parameters identification methods based on physics-informed neural networks (PINNs).