Effect of ice geometry to airfoil performance using neural networks prediction

The ice accretion geometric characteristics have complex effects on airfoil aerodynamic coefficients. By using levenberg-marguardt (LM) learn algorithm of back-propagation (BP) network, several neural networks were established to get correlations between typical glaze ice geometry (ice horn leading-edge radius, ice height and ice horn position on airfoil surface) and airfoil aerodynamic coefficients. The neural network was also generated for predicting the maximum lift coefficient in typical ice accretion geometric parameters. In addition, the neural network that used to get the relationship of airfoil hinge moment coefficient and ice geometric was obtained to predict the angle of attack stall. The simulation results of neural networks have high coherence with experimental data and can predict coefficients at non-experimental conditions. The horn location on airfoil surface has the most severe effects on airfoil coefficients. The hinge moment coefficient breaks before the stall of angle of attack and can be used for predicting stall more safely.