基于机器学习的TC17合金超高周疲劳寿命预测
Prediction of very-high-cycle fatigue life of TC17 alloy based on machine learning
石炜 1钱泓江 2黄志勇 2赵伟 1郭建英1
作者信息
- 1. 中国航发四川燃气涡轮研究院,成都 610500
- 2. 四川大学 空天科学与工程学院,成都 610065
- 折叠
摘要
航空发动机压气机叶片材料TC17超高周疲劳试验周期长、成本高,且疲劳寿命数据分散度大,导致其疲劳寿命预测较难,预测模型精度不高.机器学习(ML)具有强大的数据处理能力,采用了蒙特卡罗模拟(MCS)对TC17超高周疲劳寿命数据进行了有效的扩展和增强,提出了一种具有动态记忆建模能力的机器学习网络模型,可检验MCS的有效性和提高超高周疲劳寿命数据预测的准确性.研究结果表明,应力比R=0.1工况下经MCS数据增强作用后的网络模型预测精度最大可提高约63.05%,预测结果均在5.0倍分散带以内.
Abstract
Based on the challenges of the fatigue life of TC17,a titanium alloy compressor blade material for aero-engines,which is difficult to predict due to the large dispersion of fatigue life and the limitations of high test cost and long test period.Machine learning(ML)has powerful data processing ability,this paper uses Monte Carlo simulation(MCS)to extend and enhance the fatigue life of TC17 at very-high-cycle fatigue life,and uses machine learning to verify the accuracy of fatigue life prediction.The prediction results for the stress ratioR=0.1 show that the prediction accuracy of the ML model after data enhancement has improved by 63.05%,and also those predictions are all within the scatter band of 5.0.
关键词
航空发动机/机器学习/TC17/超高周疲劳/寿命预测/蒙特卡罗/数据增强Key words
aero-engines/machine learning/TC17/very-high-cycle fatigue/life prediction/Monte Carlo/data enhancement引用本文复制引用
出版年
2024
NETL
NSTL
万方数据