A method for unsteady aerodynamic modeling of biaxial coupled oscillation based on CV-SMO-SVR

This paper presents a precise unsteady aerodynamic modeling method of biaxial coupling oscillation which overcomes the drawbacks of the nonlinearity, coupling and hysteresis of the aerodynamic during post-stall maneuver. In order to establish a large number of experimental data model rapidly, a method of unsteady aerodynamic modeling on the basis of Sequential Minimal Optimization - Support Vector Regression (SMO-SVR) is proposed. The input variables, output variables and kernel functions of the SVR model for unsteady aerodynamic modeling are determined relying on the analysis of the wind tunnel test data. To improve the modeling accuracy, Cross Validation (CV) is successfully applied to adjust the parameters of the proposed SMO algorithm. The accurate unsteady aerodynamic model can be obtained from the random training data and the random testing data. The unsteady aerodynamic modeling under the pitch-roll and the yaw-roll oscillation is completed. Comparing with the Back Propagation Neural Networks (BPNN) method, the method proposed in this paper has characteristics of high accuracy and strong versatility.