基于LSTM网络参数优化的航空发动机寿命预测  

Aeroengine life prediction based on LSTM network parameter optimization

作　　者：宋慧 陶冠叶 曲大义 曲亚川 SONG Hui;TAO Guanye;QU Dayi;QU Yachuan(School of Mechanical and Automotive Engineering,Qingdao University of Technology,Qingdao 266525,China)

机构地区：[1]青岛理工大学机械与汽车工程学院,青岛266525

出　　处：《青岛理工大学学报》2023年第5期112-117,共6页Journal of Qingdao University of Technology

基　　金：国家自然科学基金资助项目(51678320);山东省重大科技研发计划项目(2019GGX101038)。

摘　　要：作为飞机正常运行的关键部件,航空发动机在保证飞机整体可靠性方面起着至关重要的作用。为了精确预测航空发动机剩余使用寿命(Remain Useful Life,RUL),提出了一种基于混沌博弈(Chaos Game Optimization,CGO)算法优化长短时记忆(Long Short-Term Memory,LSTM)网络的预测方法。首先,基于航空发动机历史运行数据构建CGO算法的适应度函数;其次,利用CGO算法对LSTM网络内的神经元个数、学习率、迭代次数等参数进行调整;最后,利用训练好的LSTM网络预测在役航空发动机的RUL。基于NASA提供的航空发动机退化仿真数据集,实验结果表明,该方法的预测精度优于单一LSTM网络预测方法。As the key component of the normal operation of aircraft,aeroengine plays a vital role in ensuring the overall reliability of the aircraft.In order to precisely predict the aeroengine Remain Useful Life(RUL),a prediction method based on Chaos Game Optimization(CGO)algorithm is proposed to optimize Long-Short Term Memory(LSTM)network.Firstly,the fitness function of CGO algorithm is constructed based on historical aeroengine operation data.Secondly,CGO algorithm is used to adjust the number of neurons,learning rate,iteration number and other parameters in LSTM network.Finally,the trained LSTM network is used to predict the RUL of in-service aeroengine.With reference to the aeroengine degradation simulation dataset from NASA,the experimental results of this study show that the proposed method outperforms the single LSTM network-based prediction method in prediction precision.

关 键 词：航空发动机 剩余使用寿命(RUL) 混沌博弈优化(CGO)算法 长短时记忆(LSTM)网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] V23[自动化与计算机技术—控制科学与工程]