针对发动机热端部件工作环境恶劣,存在寿命评估困难的问题,提出一种基于有限元和神经网络联合模型的高压涡轮叶片寿命评估方法,深入研究叶片的服役损伤规律,准确预测叶片剩余寿命对发动机运行的安全性具有重要意义。从机理分析角度出发,利用热流固耦合仿真技术分析并考虑疲劳蠕变的交互作用,采用基于交互损伤相互作用因子的时间-寿命分数法开展高压涡轮叶片寿命预测模型的研究工作。为提高寿命预测的快速性和经济性,在有限元分析的基础上,结合神经网络预测技术建立以等效运行时间为输出量的涡轮叶片寿命评估数学模型,利用有限元软件计算得出100组数据进行网络训练与初步验证。有限元技术与神经网络联合模型搭建完成后, 将实际机组运行数据输入网络模型, 计算出叶片寿命预测结果。预测结果与国外商用软件的剩余寿命评估指标进行对比,一级静叶和一级动叶的等效运行时间相对误差分别在1%和3%以内,预测模型能够利用有限元原理与神经网络二者的优势,从而为快速准确预测叶片剩余寿命提供一种新方法。

Addressing the issue of harsh working conditions and difficulty in assessing the service life of the high-temperature components of the engine, it is of great significance to conduct in-depth research on the damage law of the high-pressure turbine blades and estab-lish a reliable residual life prediction model to improve the safety of engine operation. From the perspective of mechanism analysis, the thermal-fluid-solid coupling simulation technology is used to analyze and consider the interaction of fatigue and creep, and the time-life fraction method based on the interaction damage interaction factor is adopted to carry out the research on the high-pressure turbine blade life prediction model. One hundred sets of data are calculated by finite element software for network training and pre-liminary verification. Based on finite element analysis, a mathematical model for evaluating the life of turbine blades with equivalent operating time as the output quantity is established by combining neural network prediction technology. The predicted results of the model are compared with the calculated results of foreign commercial software, and the relative errors of the equivalent operating time of the first-stage stator blade and the first-stage rotor blade are within 1% and 3%, respectively. The predictive model of this re-search can take advantage of the strengths of both finite element principles and neural networks, thus providing a new method for rapid and accurate prediction of residual blade life.

V232.4

中国航发产学研基金合作项目（HFZL2019CXY028）和国家科技重大专项（2019-I-0019-0018）