Abstract: Aircraft engine maintenance is changing from original timing maintenance to condition-based maintenance, which is based on accurate estimate of the damage to the engine parts or the whole engine. Traditional damage prediction based on finite element theory only accurately predicts damage of the parts under nominal conditions, but it is difficult to guarantee the prediction accuracy of the component damage under other non-nominal conditions caused due to the change of external environment, noise. And its complex analysis process is not conducive to real-time operation of airborne. The paper takes thermo-mechanical damage of turbofan engine guide vane for example, establishes the neural network prediction model between the engine operating condition and vane damage, and then improves the prediction accuracy of the model using Monte Carlo simulation. The simulation results show that the error of predicted TMF damage is bellow 0.4%. And the model can be used to real-time prediction of airborne.

Key words: damage prediction, turbine blade, Monte Carlo simulation, aircraft engine

摘要： 目前航空发动机维修正从原来的定时维修向视情维修转换，而视情维修的基础则是准确预测发动机部件或整机的损伤。传统基于有限元理论的损伤预测仅能对标称条件下部件损伤进行精确预测，在因外部环境改变、噪声等因素引起的非标称条件下，则难以保证部件损伤预测精度，同时其分析过程复杂、工作量大，不利于机载实时运行。以某型涡扇发动机涡轮导向叶片的热机械疲劳损伤为例，建立发动机运行条件和叶片损伤之间的神经网络预测模型，并利用Monte Carlo仿真提高模型的预测精度。仿真结果显示，根据下一循环的飞行条件，叶片损伤预测结果相对误差在0.4%以下，且该模型可以应用于机载实时预测。

关键词: 损伤预测, 涡轮导向叶片, Monte Carlo仿真, 航空发动机