基于神经网络的双层壁叶片异型孔结构强度设计

Strength Design of Special⁃Shaped Holes in Double⁃Walled Turbine Blade Based on Neural Network

随着航空发动机领域的飞速发展,热端涡轮叶片服役环境越发恶劣.为了提高叶片的承温能力,行业内提出了双层壁叶片结构,其原理为“外壁承温,内壁承载”.本文针对双层壁叶片气膜孔结构强度问题,采用参数化建模构建叶型及外壁孔结构,通过仿真分析确定孔型设计参数对圆柱型气膜孔、圆锥型扩张孔和簸箕型扩张孔的孔周结构强度的影响,最后根据参数输入及对应响应构建神经网络模型,通过遗传算法优化径向基函数(Radial Basis Function, RBF)神经网络的权值与阈值来提高代理模型的准确性.寻优获得三种不同结构的气膜孔最优孔型设计方案,相较于圆柱型孔,圆锥型扩张孔孔周应力降低了25.12%,簸箕型扩张孔孔周应力降低了22.54%.

With the rapid development of aeroengine field,the service environment of hot end turbine blade is becoming increasing severe.In order to improve the heat bearing capacity of blade,the double-walled blade structure is proposed in the industry,which is based on the principle of"outer wall heat bearing,inner wall load bearing".In this paper,aiming at ensuring the structural strength of double-walled blade film hole,a parametric modeling is used to construct the blade shape and the outer wall hole structure,and the influences of hole design parameters on the perimeter structure strength of the cylindrical film hole,the conical expansion hole and the dustpan expansion hole are determined through simulation analysis.Then,according to the parameter input and the corresponding response,a Radial Basis Function(RBF)neural network model is constructed and further optimized through adjusting the weights and thresholds using the genetic algorithm to improve the accuracy of the surrogate model.Finally,the optimal hole designs for three film holes are obtained.Compared with the original cylindrical hole,the perimeter stress is reduced by 25.12%and 22.54%,respectively for the conical expansion hole and the dustpan expansion hole.