Abstract: To obtain compressive residual stress in milling aviation engine blade,the Radial Basis Function (RBF) neural network was introduced to predict the residual stress for high-temperature alloy GH4169.The lateral inclination angle,cutting speed and feed were taken as design factors to develop the ball-end milling experiments,and X-ray method was used to test residual stress.The RBF neural network was trained by using the experimental samples,and the result showed that RBF neural network had the best performance on prediction accuracy compared with Back Propagation (BP) neural network and conventional multiple linear regression method.The interaction effect of process parameters on residual stress was researched,and the result revealed that there was strongly nonlinear relationship between residual stress and process parameters.The proposed RBF model of residuals stress could be used for further process optimization of milling aviation engine blade.

Key words: aviation engine, blade, residual stress, radial basis function neural network, ball-end mill high-temperature alloy GH4169

摘要： 为了获得航空叶片精密铣削中的残余压应力,引入径向基函数神经网络对GH4169叶片铣削残余应力进行预测。以刀轴侧倾角、铣削速度和每齿进给量为设计因子开展球头刀铣削实验,采用X射线法测试残余应力;采用实验样本对径向基函数神经网络进行训练,获得残余应力的径向基函数模型,并与传统的多元线性回归模型与BP神经网络进行对比,结果表明径向基函数模型的预测精度最高。分析工艺参数对铣削残余应力的交互影响规律,结果表明铣削残余应力与3个工艺因子之间存在高度的非线性映射关联。所提出的径向基函数模型有望为航空叶片铣削工艺优化奠定理论基础。

关键词: 航空发动机, 叶片, 残余应力, 径向基函数神经网络, 球头刀高温合金GH4169