关键词: 飞机设计, 机翼, 多学科设计优化, 响应面, Kriging模型, 神经网络, 高亚声速运输机

Abstract: Multidisciplinary aerodynamic/structural design optimization is carried out for high subsonic transport wing using approximation technique. The framework of Multidisciplinary Design Optimization (MDO) based on approximation is presented and analyzed. The aerodynamic performance of wing-body combination in transonic flow is calculated with full-potential equation in conjunction with viscous correction method. Structural analysis is performed using finite element method to obtain stress and strain characteristics. The span, taper ratio, sweep angle and linear twist angle are chosen as design variables that define the aerodynamic configuration of the wing, and another four representing thicknesses of spars and skin are selected as the design variables for structural discipline. Uniform Design method is used to provide sample points, and approximation models for aerodynamic and structural discipline are constructed using quadratic response surface method (RSM), Kriging model (KM) and neutral networks (NN), respectively. The accuracy of each set of approximations is compared through numerical error analysis. The objective is to investigate whether KM and NN can construct more accurate global approximations than RSM in a real aerospace engineering application and finally choose that of best accuracy to be used in the present wing design optimization problem. It is found that KM and RSM have comparative high accuracies and both are more accurate than NN. Multi-objective optimization for the wing is performed based on RSM, with lift-to-drag ratio and weight as targets, and with lift, reference area, deform and equivalent stress as constraints. The optimum wing is proven to have better integrated performance and that the presented method is applicable in engineering for multidisciplinary aerodynamic/structural design optimization of high subsonic transport wing.

Key words: aircraft design, wing, multidisciplinary design optimization, response surface method, Kriging model, neural networks, high subsonic transport