基于CAE与LSTM的航空发动机剩余寿命预测  

Prediction of the remaining life of aerospace engine based on CAE and LSTM

作　　者：王旭 艾红[1] WANG Xu;AI Hong(School of Automation,Beijing Information Science&Technology University,Beijing 100192,China)

机构地区：[1]北京信息科技大学自动化学院,北京100192

出　　处：《北京信息科技大学学报（自然科学版）》2020年第4期57-62,共6页Journal of Beijing Information Science and Technology University

基　　金：北京市自然科学基金资助项目(4162025)。

摘　　要：通过深度学习方法构建航空发动机的健康状况评估模型,并在此模型基础上进行剩余寿命预测。基于卷积自编码器构建航空发动机的健康因子(health indicator,HI),以其HI值反映健康状况;通过长短期记忆网络(long short-term memory,LSTM)建立HI与剩余寿命的特征关系,实现剩余寿命预测。经比较验证,此间接方法的预测精度优于多层感知机、支持向量回归等浅层神经网络,以及卷积神经网络、多层LSTM等直接预测的深度学习方法。The deep learning method is used to construct the aerospace engine health assessment model,and the remaining life of aerospace engine is predicted based on this model.An aerospace engine health indicator(HI)is constructed based on convolutional autoencoder and the HI value reflects the health status;the characteristic relationship between the HI and the remaining life is established through LSTM(long short-term memory)to realize the prediction of remaining life.Through comparison and verification,the prediction accuracy of this indirect method is better than that of shallow neural networks such as multi-layer perception and support vector regression,and deep learning methods such as convolutional neural networks and multi-layer LSTM.

关 键 词：航空发动机 剩余寿命预测 健康因子 卷积自编码器 长短期记忆 

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]