Wing aerodynamic robustness optimization based on neural network response surface

The robustness problem in the aerodynamic optimization of an aircraft wing is discussed in reference to the undulation of aircraft performance derived from uncertainty factors. Aerodynamic performance robustness constrained models are built which are subject to the uncertainty factors of velocity and twist angle. By dint of the BP (Back Propagation) neural network response surface based on the uniform design which is constructed through MATLAB, two schemes, whose difference lies in whether or not robustness is taken into account, are respectively obtained based on genetic algorithm. The results suggest that the maximum lift-drag ratios at cruising speeds for both schemes are higher than those of the initial scheme. Though the scheme with the consideration of robustness is 0.027 9 lower than that of the scheme without it, the variation of maximum lift-drag ratio of the former scheme is respectively 0.034 0 and 0.001 6 less than the latter within the range of thecruise Mach number and the twist angle. Other aerodynamic performances of the design which takes robustness into consideration are also much more stable than those which does not. Therefore the aerodynamic robustness optimization method in this article is shown to be useful and efficient.