基于 GMTCN 模型的轴承剩余寿命预测方法

taptap下载安装安卓学报 ›› 2025, Vol. 43 ›› Issue (3): 15-23.

基于 GMTCN 模型的轴承剩余寿命预测方法

taptap下载安装安卓 a.电子信息与自动化学院；b.中欧航空工程师学院，天津 300300

收稿日期:2023-12-21 修回日期:2024-05-06 出版日期:2025-07-12 发布日期:2025-07-12
作者简介:郭润夏（1981— ），男，山东诸城人，教授，博士，研究方向为机载设备故障诊断与健康管理、飞行器控制
基金资助:
国家自然科学基金项目（62173331）；天津市科技计划项目（23JCYBJC00060）；航空科学基金项目（2019ZD067007）；天津市优秀青年学者人才培养专项（TJTZJH-QNBJRC-2-19）；天津市教委科研计划项目（2023KJ222）

Prediction method for bearing remaining life based on GMTCN model

a. College of Electronic Information and Automation; b. Sino-European Institute of Aviation Engineering, CAUC, Tianjin 300300, China

Received:2023-12-21 Revised:2024-05-06 Online:2025-07-12 Published:2025-07-12

摘要/Abstract

摘要：

针对现有的轴承剩余寿命预测方法在处理多传感器数据时难以有效提取退化特征的问题，本文提出了一

种基于全局注意力的多尺度时间卷积网络（GMTCN， global attention and multi-scale time convolutional network）模型的轴承剩余寿命预测方法。首先，采用 GMTCN 模型对轴承的多传感器信号进行处理，借助两种

不同策略的时间卷积网络提取轴承在不同尺度下的退化特征；其次，利用全局注意力机制权衡来自不同传

感器和时间步长的数据在轴承剩余寿命预测中的贡献，并将提取的多尺度特征进行融合；最后，对轴承进行

剩余寿命预测。为对该方法进行性能评估，使用 PHM2012 轴承数据集和加速疲劳实验平台采集的退化数

据得到的轴承数据集进行剩余寿命预测，得到的均方根误差（RMSE，root mean square error）和平均绝对

误差（MAE，mean absolute error）均低于其他方法，评分函数（SCORE）的平均值在一定程度上有所提高，

证明了方法的有效性。

关键词:

轴承, 剩余使用寿命, 时间卷积网络, 特征提取, 注意力机制

Abstract:

Aiming at the problem that the existing prediction methods for bearing remaining life are difficult to effectively

extract degradation features when dealing with multi-sensor data, a prediction method for bearing remaining life

based on global attention and multi-scale time convolutional network (GMTCN) was proposed. Firstly, the GMTCN

model was used to process the multi-sensor signals of the bearing, and the degradation features of the bearing at

different scales were extracted with the help of two different strategies of temporal convolutional networks.

Secondly, the global attention mechanism was used to balance the contribution of data from different sensors and

time steps in the bearing remaining life prediction, and the extracted multi-scale features were fused. Finally, the

remaining life of the bearing is predicted. To assess the performance of this method, remaining life prediction were

conducted using the PHM2012 bearing dataset and a bearing dataset obtained from degradation data collected on

an accelerated fatigue testing platform. The root mean square error (RMSE) and mean absolute error (MAE) values

obtained were lower than other methods, while the average value of SCORE was increased to a certain extent,

proving the effectiveness of the method.

Key words:

bearings, remaining useful life, temporal convolutional network, feature extraction, attention mechanism

中图分类号:

TH133.3

郭润夏 a , 姜宇扬 a , 黄超 b.

基于 GMTCN 模型的轴承剩余寿命预测方法

[J]. taptap下载安装安卓学报, 2025, 43(3): 15-23.

GUO Runxiaa , JIANG Yuyanga , HUANG Chaob.

Prediction method for bearing remaining life based on GMTCN model

[J]. Journal of Civil Aviation University of China, 2025, 43(3): 15-23.

参考文献 18

[1]	黄庆卿, 胡欣堪, 韩延, 等. 多源域子域自适应的滚动轴承剩余寿命预测方法[J]. 电子测量与仪器学报, 2022, 36(10):100-107.
[2]	裴洪, 胡昌华, 司小胜, 等. 基于机器学习的设备剩余寿命预测方法综述[J]. 机械工程学报, 2019, 55(8):1-13.
[3]	AL AZZAWI D, MONCAYO H, PERHINSCHI M G, et al. Comparison of immunity -based schemes for aircraft failure detection and identification [J]. Engineering Applications of Artificial Intelligence,2016, 52: 181-193.
[4]	CUI L L, WANG X, WANG H Q, et al． Research on remaining useful life prediction of rolling element bearings based on time -varying Kalman filter[J]. IEEE Transactions on Instrumentation and Measurement, 2020,69(6): 2858-2867.
[5]	LI N P, LEI Y G, LIN J, et al． An improved exponential model for predicting remaining useful life of rolling element bearings[J]. IEEE Transactions on Industrial Electronics, 2015, 62(12): 7762-7773.
[6]	LIN Y H, LI G H． A Bayesian deep learning framework for RUL prediction incorporating uncertainty quantification and calibration[J]． IEEE Transactions on Industrial Informatics, 2022, 18(10): 7274-7284．
[7]	李天梅, 司小胜, 刘翔, 等. 大数据下数模联动的随机退化设备剩余寿命预测技术[J]. 自动化学报, 2022, 48(9): 2119-2141.
[8]	SATEESH BABU G, ZH AO P L, LI X L． Deep convolutional neural network based regression approach for estimation of remaining useful life[C]//Database Systems for Advanced Applications: 21st International Conference, DASFAA 2016, April 16 -19, 2016 , Dallas, TX, USA.Cham: Springer International Publishing, 2016: 214-228.
[9]	CHOW T W S, FANG Y. A recurrent neural-network-based real-time learning control strategy applying to nonlinear systems with unknown dynamics[J]. IEEE Transactions on Industrial Electronics, 1998, 45(1):
	151-161.
[10]	WU Y T, YUAN M, DONG S P, et al. Remaining useful life estimation of engineered systems using vanilla LSTM neural networks[J]. Neurocomputing, 2018, 275: 167-179.
[11]	GUO L, LI N P, JIA F, et al. A recurrent neural network based health indicator for remaining useful life prediction of bearings [J].Neurocomputing, 2017, 240: 98-109.
[12]	BAI S J, KOLTER J Z, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling [EB/OL].(2018-03-04)[2024-03-15]. https://arxiv.org/pdf/1803.01271.pdf.
[13]	HE K, SU Z Q, TIAN X Q, et al． RUL prediction of wind turbine gearbox bearings based on self -calibration temporal convolutional network [J].IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-12.
[14]	WANG Y W, DENG L, ZHENG L Y, et al. Temporal convolutional network with soft thresholding and attention mechanism for machineryprognostics[J]. Journal of Manufacturing Systems, 2021, 60: 512-526.
[15]	梁浩鹏, 曹洁, 赵小强. 基于并行双向时间卷积网络和双向长短期记忆网络的轴承剩余使用寿命预测方法[J/OL]. 控制与决策:1-9[2024-03-15]. https://doi.org/10.13195/j.kzyjc.2023.0152.
[16]	卢瑾, 张永平. 基于注意力机制的滚动轴承剩余使用寿命预测方法[J]. 机电工程, 2023, 40(4):516-521.
[17]	WANG H Q, ZHANG X S, GUO X D, et al. Remaining useful life prediction of bearings based on multiple-feature fusion health indicator and weighted temporal convolution network[J]. Measurement Science and Technology, 2022, 33(10): 104003.
[18]	NECTOUX P, GOURIVEAU R, MEDJAHER K, et al. PRONOSTIA: an experimental platform for bearings accelerated degradation tests[C]//IEEE International Conference on Prognostics and Health Management,
	2012, Denver, CO, USA. PHM′12. IEEE Catalog Number: CPF12PHMCDR, 2012: 1-8.