某机翼的安全预测载荷模型建立

基于试飞阶段全V-N包线的实测飞行载荷，将改进遗传算法、线性回归与BP神经网络融合，给出了一种适用于全寿命周期的自适应安全预测载荷模型建立方法。将该方法应用于某飞机机翼的安全预测载荷模型建立，并对所建立的载荷模型进行了全V-N包线的验证。分析了样本空间与载荷模型精度的关系。结果表明：建立的弯矩预测载荷全包线最大误差为10.6%、平均误差为1.0%，剪力的最大误差为9.1%、平均误差为0.4%，比优化线性和分段线性的误差小，比神经网络的收敛性好。随着建模数据从全样本、1/2、1/3、…、1/10样本的变化，弯矩和剪力方程的全V-N包线的最大误差整体呈增大趋势，弯矩最大误差变化范围为10.6%~19.6%，最大剪力误差变化范围为9.1%~27.9%。

Construction of safety-predicting load model on certain wing

Using measured flight loads in full V-N envelope during flight tests, an adaptive safety-predicting load model for life cycles was built based on the combination of an improved genetic algorithm, a linear regression and BP neural network. The above adaptive method was used to build the safety-predicting load model of certain wing which was validated in full V-N envelope. Moreover, the effects of the sample on the model accuracy were analyzed. The results showed that the maximum and average errors of the predicted bending-moment in full V-N envelope are 10.6% and 1.0%, and those of the predicted shear in full V-N envelope are 9.1% and 0.4%, respectively. These errors are lower than those from optimized-linear and piece-wise-linear methods, and the convergence is better than that from the neural network. With the sample varying from full, 1/2, 1/3,…, 1/10, the maximum errors of the bending-moment and the shear change from 10.6% to 19.6% and from 9.1% to 27.9%, respectively.