基于埃尔曼神经网络的高压涡轮叶片低周疲劳可靠性分析

Low-cycle Fatigue Reliability Prediction of High-pressure Turbine Blades Based on Elman Neural Network

涡轮叶片是航空发动机的关键部件,其主要失效模式为低周疲劳。为了提高涡轮叶片低周疲劳寿命的计算精度和效率,考虑材料参数和载荷的分散性,探索埃尔曼神经网络与蒙特卡罗抽样相结合的方法对涡轮叶片低周疲劳可靠性进行分析。首先利用有限元分析软件对涡轮叶片进行热-结构耦合分析,得到最大等效应力和应变数据;然后将数据代入到低周疲劳寿命理论模型中得到涡轮叶片的低周疲劳寿命;再将涡轮叶片有关随机参数作为输入变量,低周疲劳寿命作为输出变量,训练具有高精度的埃尔曼神经网络模型,实现对高压涡轮叶片的低周疲劳寿命预测和可靠性评估。与传统响应面法和蒙特卡罗方法进行对比,结果表明埃尔曼神经网络不但可以满足精度要求,而且计算效率得到显著提高。

Turbine blades are key components of aero-engines,and the main failure mode is low-cycle fatigue.Considering the dispersion of material parameters and loads,this paper explores the combination of the Elman neural network and Monte Carlo sampling to analyze the low-cycle fatigue reliability of a turbine blade.Firstly,the thermal-structural coupling analysis of the turbine blade is carried out by using finite element analysis software,and the maximum equivalent stress and strain data are obtained.Then the data are substituted into the fatigue life theoretical model to obtain the low-cycle fatigue life of the turbine blade.The random parameters and low-cycle fatigue life of the turbine blade are respectively taken as input variables and output variables to model a high-precision Elman neural network and then realize low-cycle fatigue life prediction and reliability evaluation of the high-pressure turbine blade.Compared with the traditional response surface method and Monte Carlo method,the results show that the Elman neural network can meet the accuracy requirement and significantly improve computational efficiency.