基于深度学习的导航装备轴承剩余使用寿命预测  

A Navigational Equipment Bearing Remaining Useful Life Prediction Based on Deep Learning

作　　者：党慧莹 李海林[1] 吴北苹 余佳宇 庄银传 DANG Huiying;LI Hailin;WU Beiping;YU Jiayu;ZHUANG Yinchuan(Information and Navigation School,Air Force Engineering University,Xi’an 710077,China;Unit 31401,Tonghua 134000,Jilin,China;Air Force Communication Noncommissioned Officer School,Dalian 116000,Liaoning,China;Unit 95486,Chengdu 610041,China;Unit 93127,Beijing 100834,China)

机构地区：[1]空军工程大学信息与导航学院,西安710077 [2]31401部队,吉林通化134000 [3]空军通信士官学校,辽宁大连116000 [4]95486部队,成都610041 [5]93127部队,北京100834

出　　处：《空军工程大学学报》2025年第2期81-88,共8页Journal of Air Force Engineering University

摘　　要：作为导航装备的重要部件,轴承影响着导航装备的定位精度和保障能力。在装备剩余使用寿命(RUL)预测中,传统的机器学习算法在处理复杂非线性传感信号问题上存在局限性,为此提出了一种基于注意力机制(AM)和双向长短期记忆网络(Bi-LSTM)的轴承RUL预测框架(Bi-LSTM-A)。该框架在前端加入一维卷积神经网络(CNN)从原始传感器信号中提取局部特征,然后利用双向长短期记忆网络与注意力机制相结合的方式对信号进行分析预测,最后经过网络末端的全连接层输出预测结果。通过与同类算法的对比实验表明,该方法能够准确地预测装备剩余使用寿命,具有较好的预测效率和预测精度。As a crucial component of navigation equipment,bearings affect the positioning accuracy and safeguarding capability of the navigation equipment.In predicting the remaining useful life(RUL)of equipment,traditional machine learning algorithms are limited to dealing with the problems of complex nonlinear characteristic signals.For the above-mentioned reasons,a new prediction framework for RUL of bearing based on attention mechanism(AM)and bidirectional long short-term memory(Bi-LSTM)is proposed(Bi-LSTM-A).First,a one-dimensional convolution neural network(CNN)is added to the front of the structure to extract local features from the original signal sequence,and then,the signals are analyzed and predicted by combining bidirectional long short-term memory network with attention mechanism.finally,the predicted results are output through the fully connected layers at the end of the network.in comparison with the similar algorithms,the results show that the proposed method can accurately predict the equipment remaining useful life,and is good in predicting efficiency and accuracy.

关 键 词：轴承 深度学习 长短期记忆网络 注意力机制 剩余使用寿命 

分 类 号：TH17[机械工程—机械制造及自动化]