Parametric reduced-order model of unsteady aerodynamics based on incremental learning algorithm

The aerodynamic Reduced-Order Model (ROM) is a useful tool in the prediction of nonlinear unsteady aerodynamics with reasonable accuracy and low computational cost. The efficacy of this method has been validated by recent studies. However, the robustness of ROMs with respect to flight parameter variations should be further improved. To enhance the prediction capability of ROMs for varying flight parameters, this paper presents two parametric reduced-order models based on the Least Squares Support Vector Regression (LS-SVR) and the incremental learning algorithm. LS-SVR is a class of regression methods with good generalization ability, and the main contribution of the incremental learning algorithm based on LS-SVR is that it is not necessary to relearn the whole data set while the new sample sets are incremented. To illustrate the approach, we construct the parametric unsteady aerodynamic ROMs of the NACA64A010 airfoil model with two degrees of freedom. The Mach number and angle of attack are considered as the additional model inputs to train the relationship between aerodynamic inputs and the corresponding outputs. It is demonstrated that the model can accurately capture the dynamic characteristics of aerodynamic and aeroelastic systems for varying flight parameters. © 2021, Beihang University Aerospace Knowledge Press. All right reserved.