Research On Aerodynamic Performance Of Helicopter Blades With Trailing Edge Flaps Based On Machine Learning

In order to improve the flight performance of helicopters,trailing edge flap devices can be added to the rotor wing pattern.Wind tunnel experiments are expensive and computational fluid dynamics(CFD)methods consume a lot of computational resources.It is not conducive to effective and rapid parametric evaluation of helicopter airfoil optimization design.With the rapid development of artificial intelligence theory and data science.Machine learning can skip the complex physical modeling process and obtain prediction results quickly and accurately by algorithms.Therefore,this paper combines machine learning algorithms and highly accurate CFD numerical simulations to build a machine learning model for helicopter airfoil lift-to-drag ratio prediction under different combinations of trailing edge flap aerodynamic parameters,and investigates the effects of different aerodynamic parameters on the aerodynamic performance of rotor blade airfoils through the SHapley Additive ex Planations(SHAP)method based on machine learning,which provides a reference for the aerodynamic design of rotor blade airfoils.In this paper,144 numerical simulations with different combinations of trailing edge flap parameters are performed sequentially by Ansys-Fluent software to construct the sample data set required for the training of machine learning algorithms.Then,the sample data set is put into three algorithms:support vector machine,random forest and e Xtreme Gradient Boosting(XGBoost)for model training and optimization of parameters to establish a lift-to-drag ratio prediction model of the airfoil with trailing edge flaps.In the training set the root mean square error(RMSE)is 0.625,0.7138 and0.423,and the coefficient of determination(R~2)is 0.99,0.9889 and 0.9988 for the support vector machine,random forest and XGBoost training results,respectively;in the test set,the RMSE is 0.895,0.9252 and 0.6719,and the R~2 is 0.9713,0.9601 and0.9889,respectively.All the three machine learning algorithms showed high prediction accuracy for wing lift resistance with trailing edge flaps in both the training and test sets.The XGBoost algorithm showed the best model prediction accuracy and no significant overfitting in both the training and test sets,and had the best model generalization capability.The SHAP method based on the XGBoost model was used to investigate the model contribution of the three parameters,and the results showed that the angle of attack was the most important parameter affecting the aerodynamic performance of the airfoil with trailing edge flaps,followed by the flap angle,and the flap length had the least influence on the aerodynamic performance.In addition,the effect of three aerodynamic parameters on the lift-to-drag ratio of the NACA 2412 airfoil was further analyzed and elucidated.The results showed that the NACA 2412 airfoil with trailing edge flaps had a significantly higher lift increase in the low angle of attack range of 0°to 5°compared to the airfoil without trailing edge flaps.However,when the angle of attack is greater than 6°,neither increasing the flap length nor the flap angle has an enhanced effect on the lift performance of the airfoil,and as the angle of attack increases,it has a weakening effect on the lift performance of the airfoil.The analysis of different flap angles shows that the flap angle is positively correlated with the lift-to-drag ratio of the airfoil in the range of 0°to 5°,but the rapid increase of drag coefficient leads to the decrease of lift-to-drag ratio when the angle of attack is larger than 6°.The analysis of different flap lengths shows that excessive flap length increases the lift of NACA 2412at the cost of increased drag,which in turn leads to a worse lift-to-drag curve.Therefore,increasing the flap length alone has limited effect in improving the lift-to-drag ratio of the helicopter airfoil.It is further concluded that the combined effect of trailing edge flaps on the airfoil is best when the flap length of the NACA 2412 airfoil is 10%of the chord length.Compared with the NACA 2412 airfoil without trailing edge flaps,the trailing edge flap airfoil with flap length L_P=10%C and flap angle?=10°improves the lift-to-drag ratio by 87.82%,64.94%,49.20%,33.30%,22.74%and 14.75%at the angles of attack?=0°,1°,2°,3°,4°and 5°,respectively,indicating that the NACA2412 airfoil with trailing edge flaps has significantly improved the lift-to-drag ratio in the low angle of attack range from 0°to 5°,which can effectively improve the lift performance of the helicopter airfoil.The machine learning model constructed in this paper provides an efficient and economical approach for the optimal design of helicopter airfoils,which can be used as a partial alternative to wind tunnel tests and numerical simulations,and has strong practical application significance.