Modeling of Aircraft Nonlinear Unsteady Aerodynamics at High Angle Attack Based on Random Forest

A new modeling technique for aircraft nonlinear unsteady aerodynamics and aerodynamic moments at high angle attack was proposed. Traditional modeling methods based on the physical mechanism analysis and experimental observation possess some limitations leading to lower model accuracy. Aiming at the problem, a random forest method was proposed for modeling. Combining with the principle of random forest, the input features of the random forest model were determined according to the characteristics of large amplitude pitch oscillations in the wind tunnel. Through error analysis experiment, some key parameter values of the random forest model were determined, including the number of the decision trees, the number of variables randomly sampled as candidates at each split, and so on. Finally, experiments were carried out with the data obtained in the low speed tunnel with F-18 shrinkage ratio model. The results show that, the prediction of random forest model is closer to the wind tunnel data compared to polynomial model, which verifies random forest can be more effective for modeling the nonlinear unsteady aerodynamics and aerodynamic moments at high angle of attack.