摘要
涡扇发动机的整个生命周期在恒定的退化模式中,对其进行健康管理和剩余使用寿命预测具有重要意义。在发动机设备发生退化恒定故障中,为保证长期有效可靠性维护需要,该文基于机器学习与深度学习对涡扇发动机进行寿命预测,采用梯度提升决策树与随机森林模型对涡扇发动机特征进行重要性排序并建立Stacking集成学习模型,同时采用双向长短时记忆网络(BiLSTM)模型进行涡扇发动机寿命预测。结果表明,使用特征重叠后的Stacking算法模型表现优异,均方根误差(RMSE)较低,拟合优度约0.96,在涡扇发动机寿命预测方面表现良好。
The entire life cycle of a turbofan engine is in a constant degradation mode,and it is important to manage its health and predict its remaining service life.In order to ensure the longterm effective reliability maintenance of engine equipment in the event of degradation and constant fault,this study predicts the life of turbofan engines based on machine learning and deep learning,uses gradient boosting decision trees and random forest models to rank the importance of turbofan engine features and establishes a Stacking ensemble learning model,and uses bidirectional long shortterm memory network(BiLSTM)model to predict turbofan engine life.The results show that the Stacking algorithm model using feature overlap has excellent performance,low root mean square error(RMSE),and good fit of about 0.96,which performs well in turbofan engine life prediction.
作者
周先烨
陈里里
王彪
白怀伟
ZHOU Xianye;CHEN Lili;WANG Biao;BAI Huaiwei(School of Electro Mechanical and Vehicle Engineering,Chongqing Jiaotong University,Chongqing 400074,China)
出处
《自动化与仪表》
2023年第8期51-55,共5页
Automation & Instrumentation
基金
中国博士后科学基金面上资助项目(2020M683256)
重庆市技术创新与应用发展专项重点项目(cstc2020jscxgksbX0010)
交通工程应用机器人重庆市工程实验室2020年度开放课题(CELTEARKFKT202003)
重庆市研究生教育教学改革研究项目(重点)(yjg182027)。
关键词
寿命预测
集成学习
特征阈值
特征重叠
随机搜索
lifetime prediction
integrated learning
features threshold
features overlap
random search
