Please wait a minute...
浙江大学学报(工学版)
浙江大学学报(工学版)  2022, Vol. 56 Issue (7): 1447-1456    DOI: 10.3785/j.issn.1008-973X.2022.07.020
电气工程、机械工程     
基于多尺度特征与注意力机制的轴承寿命预测
莫仁鹏(),司小胜*(),李天梅,朱旭
火箭军工程大学 导弹工程学院，陕西 西安 710025
Bearing life prediction based on multi-scale features and attention mechanism
Ren-peng MO(),Xiao-sheng SI*(),Tian-mei LI,Xu ZHU
College of Missile Engineering, Rocket Force University of Engineering, Xi’an 710025, China
 全文: PDF(1709 KB)   HTML
摘要：

针对以往剩余使用寿命（RUL）预测方法对轴承退化信息挖掘不充分、忽视不同特征贡献度差异，影响预测准确性的问题，提出基于多尺度特征与注意力机制的轴承RUL预测方法. 在多个尺度下计算轴承原始振动信号的若干时域和频域特征，作为输入特征集. 将多尺度特征集输入到网络中，以注意力模块为不同特征自适应地分配最佳权重，以卷积神经网络（CNN）模块进行深层特征提取与多尺度特征融合，通过前馈神经网络（FNN）模块映射得到RUL预测值. 通过公开的轴承数据集进行实验验证，与其他RUL预测方法相比，所提方法的预测性能更优越.
关键词： 剩余使用寿命预测;  多尺度特征;  注意力机制;  卷积神经网络(CNN);  轴承    
Abstract:

A bearing RUL prediction method based on multi-scale features and attention mechanism was proposed aiming at the problem that the previous remaining useful life (RUL) prediction methods were insufficient in mining bearing degradation information and ignored the difference in the contribution of different features, which affected the prediction accuracy. Several time-domain and frequency-domain features of the original bearing vibration signal at multiple scales were calculated as the input feature set. The multi-scale feature set was input into the network, and the attention module was used to adaptively assign the best weights to different features. Then the convolutional neural network (CNN) module was used for deep feature extraction and multi-scale feature fusion. The RUL prediction value was obtained through the feedforward neural network (FNN) module mapping. The proposed method was applied to the public bearing datasets for comparative studies. Results showed the superior prediction performance of the proposed method.
Key words: remaining useful life prediction    multi-scale feature    attention mechanism    convolutional neural network (CNN)    bearing
收稿日期: 2021-07-06 出版日期: 2022-07-26
CLC:  TH 133  
基金资助: 国家自然科学基金资助项目（61773386，62073336）；国家自然科学基金优秀青年资助项目（61922089）
通讯作者: 司小胜     E-mail: renpengmo@163.com;sxs09@mails.tsinghua.edu.cn
作者简介: 莫仁鹏（1997—），男，硕士生，从事深度学习与剩余寿命预测的研究. orcid.org/0000-0002-4332-625X. E-mail： renpengmo@163.com
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章　　
莫仁鹏
司小胜
李天梅
朱旭

引用本文:

莫仁鹏,司小胜,李天梅,朱旭. 基于多尺度特征与注意力机制的轴承寿命预测[J]. 浙江大学学报(工学版), 2022, 56(7): 1447-1456.

Ren-peng MO,Xiao-sheng SI,Tian-mei LI,Xu ZHU. Bearing life prediction based on multi-scale features and attention mechanism. Journal of ZheJiang University (Engineering Science), 2022, 56(7): 1447-1456.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2022.07.020    或    https://www.zjujournals.com/eng/CN/Y2022/V56/I7/1447

图 1  剩余寿命预测的步骤
图 2  多尺度特征的提取
图 3  时间窗示意图
图 4  整体网络结构
图 5  注意力机制
工况 nr/(r·min?1) F/N
1 1 800 4 000
2 1 650 4 200
3 1 500 5 000
表 1  实验数据工况详情
图 6  轴承全寿命的振动加速度幅值
网络模块 网络层 超参数设置
注意力模块 全连接层 U = K
CNN模块 卷积层1 n = 32，k = 6，s = 1
CNN模块 卷积层2 n = 64，k = 6，s = 1
CNN模块 卷积层3 n = 64，k = 6，s = 1
CNN模块 卷积层4 n = 64，k = 6，s = 1
FNN模块 隐藏层1 U = 100
FNN模块 dropout层 r = 0.2
FNN模块 隐藏层2 U = 20
FNN模块 输出层 U = 1
表 2  网络超参数
图 7  不同时间窗尺寸下的RMSE和MAE
评价指标 RMSE MAE
尺度1 0.034 0 0.029 5
尺度1+注意力 0.031 7 0.027 3
尺度2 0.032 2 0.028 5
尺度2+注意力 0.030 3 0.026 4
尺度4 0.027 4 0.023 4
尺度4+注意力 0.026 8 0.021 9
多尺度 0.024 4 0.019 9
多尺度+注意力 0.021 2 0.016 7
表 3  消融实验的评价指标
图 8  消融实验的评价指标对比
图 9  不同单尺度模型的预测效果
多尺度组合 RMSE MAE ttr/s tte/s
1-2-4 0.021 2 0.016 7 4.64 2.73
1-2-4-8 0.022 6 0.019 3 7.15 3.99
1-2-4-8-16 0.021 1 0.017 4 11.20 5.20
表 4  不同多尺度组合的预测效果
网络 RMSE MAE ttr/s tte/s
DNN 0.065 6 0.057 7 1.32 0.91
CNN 0.030 7 0.026 0 5.05 2.47
MSCNN 0.023 1 0.018 6 52.00 8.69
BiLSTM 0.058 4 0.045 1 4.49 11.45
CNN-LSTM 0.056 0 0.046 2 51.00 44.48
本文方法 0.021 2 0.016 7 4.64 2.73
表 5  各方法的评价指标
图 10  各方法的预测结果对比
网络 轴承2-3 轴承3-1
RMSE MAE RMSE MAE
DNN 0.360 7 0.330 0 0.348 9 0.322 6
CNN 0.118 4 0.103 8 0.341 7 0.334 7
BiLSTM 0.165 8 0.135 9 0.229 4 0.210 1
本文方法 0.044 1 0.034 6 0.075 8 0.061 8
原训练集 0.448 6 0.426 8 0.438 9 0.420 9
表 6  不同工况轴承的预测性能
图 11  不同工况下的轴承RUL预测曲线
网络 轴承1-1 轴承2-2 轴承3-2
RMSE MAE RMSE MAE RMSE MAE
DNN 0.205 2 0.176 9 0.249 1 0.206 1 0.303 2 0.246 7
CNN 0.153 4 0.125 3 0.113 4 0.088 7 0.228 7 0.203 9
BiLSTM 0.144 8 0.123 5 0.102 6 0.088 7 0.207 9 0.164 9
本文方法 0.098 6 0.083 6 0.054 8 0.045 3 0.104 4 0.077 6
表 7  不同方法在XJTU-SY轴承数据集上的预测效果
图 12  测试集轴承的RUL预测曲线（XJTU-SY）
图 13  轴承1-3的RUL预测不确定性度量（FEMTO-st）
1 乔美英, 汤夏夏, 闫书豪, 等 基于改进稀疏滤波与深度网络融合的轴承故障诊断[J]. 浙江大学学报: 工学版, 2020, 54 (12): 2301- 2309
QIAO Mei-ying, TANG Xia-xia, YAN Shu-hao, et al Bearing fault diagnosis based on improved sparse filtering and deep network fusion[J]. Journal of Zhejiang University: Engineering Science, 2020, 54 (12): 2301- 2309
2 程卫东, 赵德尊 用于滚动轴承转频估计的 EMD 软阈值降噪算法[J]. 浙江大学学报: 工学版, 2016, 50 (3): 428- 435
CHENG Wei-dong, ZHAO De-zun EMD soft-thresholding denoising algorithm for rolling element bearing rotational frequency estimation[J]. Journal of Zhejiang University: Engineering Science, 2016, 50 (3): 428- 435
3 PECHT M. Prognostics and health management of electronics [M]. Hoboken: Wiley, 2008.
4 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[EB/OL]. [2021-06-30]. https://doi.org/10.16383/j.aas.c201068.
LI Tian-mei, SI Xiao-sheng, LIU Xiang, et al. Data-model interactive remaining useful life prediction technologies for stochastic degrading devices with big data [EB/OL]. [2021-06-30]. https://doi.org/10. 16383/j.aas.c201068.
5 LU C J, MEEKER W O Using degradation measures to estimate a time-to-failure distribution[J]. Technometrics, 1993, 35 (2): 161- 174
doi: 10.1080/00401706.1993.10485038
6 LIAO G, YIN H, CHEN M, et al Remaining useful life prediction for multi-phase deteriorating process based on Wiener process[J]. Reliability Engineering and System Safety, 2021, 207: 107361
doi: 10.1016/j.ress.2020.107361
7 SI X S, LI T M, ZANG Q, et al Prognostics for linear stochastic degrading systems with survival measurements[J]. IEEE Transactions on Industrial Electronics, 2020, 67 (4): 3202- 3215
doi: 10.1109/TIE.2019.2908617
8 王泽洲, 陈云翔, 蔡忠义, 等 基于复合非齐次泊松过程的不完美维修设备剩余寿命预测[J]. 机械工程学报, 2020, 56 (22): 14- 23
WANG Ze-zhou, CHEN Yun-xiang, CAI Zhong-yi, et al Prediction of remaining life of imperfect maintenance equipment based on compound inhomogeneous Poisson process[J]. Journal of Mechanical Engineering, 2020, 56 (22): 14- 23
doi: 10.3901/JME.2020.22.014
9 KUNDU P, DARPE A K, KULKARNI M S Weibull accelerated failure time regression model for remaining useful life prediction of bearing working under multiple operating conditions[J]. Mechanical Systems and Signal Processing, 2019, 143: 106302
10 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
PEI Hong, HU Chang-hua, SI Xiao-sheng, et al Overview of equipment remaining life prediction methods based on machine learning[J]. Journal of Mechanical Engineering, 2019, 55 (8): 1- 13
doi: 10.3901/JME.2019.08.001
11 唐旭, 徐卫晓, 谭继文, 等. 基于LSTM的滚动轴承剩余使用寿命预测 [J]. 机械设计, 2019, 36(增1): 117-119.
TANG Xu, XU Wei-xiao, TAN Ji-wen, et al. Prediction of remaining service life of rolling bearing based on LSTM [J]. Journal of Machine Design, 2019, 36(supple. 1): 117-119.
12 王玉静, 李少鹏, 康守强, 等 结合CNN和LSTM的滚动轴承剩余使用寿命预测方法[J]. 振动. 测试与诊断, 2021, 41 (3): 439- 446
WANG Yu-jing, LI Shao-peng, KANG Shou-qiang, et al Combining CNN and LSTM to predict the remaining service life of rolling bearings[J]. Journal of Vibration, Measurement and Diagnosis, 2021, 41 (3): 439- 446
13 张钢, 田福庆, 佘博, 等 一种基于特定频段信息熵和RBM的健康因子构建方法[J]. 振动与冲击, 2020, 39 (6): 147- 153
ZHANG Gang, TIAN Fu-qing, SHE Bo, et al A health factor construction method based on information entropy and RBM in specific frequency bands[J]. Journal of Vibration and Shock, 2020, 39 (6): 147- 153
14 CHENG Y, PENG G, ZHU Z, et al A novel deep learning method based on attention mechanism for bearing remaining useful life prediction[J]. Applied Soft Computing, 2020, 86: 105919
doi: 10.1016/j.asoc.2019.105919
15 LI X, DING Q, SUN J Q Remaining useful life estimation in prognostics using deep convolution neural networks[J]. Reliability Engineering and System Safety, 2018, 172: 1- 11
doi: 10.1016/j.ress.2017.11.021
16 LUONG M T, PHAM H, MANNING C D. Effective approaches to attention-based neural machine translation [EB/OL]. [2021-06-30]. https://arxiv.org/abs/1508.04025.
17 郭宝震, 左万利, 王英 采用词向量注意力机制的双路卷积神经网络句子分类模型[J]. 浙江大学学报: 工学版, 2018, 52 (9): 1729- 1737
GUO Bao-zhen, ZUO Wan-li, WANG Ying Two-way convolutional neural network sentence classification model using word vector attention mechanism[J]. Journal of Zhejiang University: Engineering Science, 2018, 52 (9): 1729- 1737
18 BAHDANAU D, CHOROWSKI J, SERDYUK D, et al. End-to-end attention-based large vocabulary speech recognition [C]// International Conference on Acoustics, Speech and Signal Processing. Shanghai: IEEE, 2016: 4945-4949.
19 雍子叶, 郭继昌, 李重仪 融入注意力机制的弱监督水下图像增强算法[J]. 浙江大学学报: 工学版, 2021, 55 (3): 555- 562
YONG Zi-ye, GUO Ji-chang, LI Chong-yi Weakly supervised underwater image enhancement algorithm incorporating attention mechanism[J]. Journal of Zhejiang University: Engineering Science, 2021, 55 (3): 555- 562
20 BA J, MNIH V, KAVUKCUOGLU K. Multiple object recognition with visual attention [EB/OL]. [2021-06-30]. https://arxiv.org/abs/1412.7755.
21 SONG Y, GAO S, LI Y, et al Distributed Attention-based temporal convolutional network for remaining useful life prediction[J]. IEEE Internet of Things Journal, 2020, 8 (12): 9594- 9602
22 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. Piscataway: IEEE, 2012: 1-8.
23 SOUALHI A, MEDJAHER K, ZERHOUNI N Bearing health monitoring based on Hilbert–Huang transform, support vector machine, and regression[J]. IEEE Transactions on Instrumentation and Measurement, 2014, 64 (1): 52- 62
24 REN L, CUI J, SUN Y, et al Multi-bearing remaining useful life collaborative prediction: a deep learning approach[J]. Journal of Manufacturing Systems, 2017, 43: 248- 256
doi: 10.1016/j.jmsy.2017.02.013
25 张继冬, 邹益胜, 邓佳林, 等 基于全卷积层神经网络的轴承剩余寿命预测[J]. 中国机械工程, 2019, 30 (18): 2231- 2235
ZHANG Ji-dong, ZOU Yi-sheng, DENG Jia-lin, et al Bearing remaining life prediction based on fully convolutional neural network[J]. China Mechanical Engineering, 2019, 30 (18): 2231- 2235
26 孙鑫, 孙维堂 基于多尺度卷积神经网络的轴承剩余寿命预测[J]. 组合机床与自动化加工技术, 2020, (10): 168- 171
SUN Xin, SUN Wei-tang Prediction of bearing remaining life based on multi-scale convolutional neural network[J]. Modular Machine Tool and Automatic Manufacturing Technique, 2020, (10): 168- 171
27 韩林洁, 石春鹏, 张建超 基于BiLSTM的滚动轴承剩余使用寿命预测[J]. 制造业自动化, 2020, 42 (5): 47- 50
HAN Lin-jie, SHI Chun-peng, ZHANG Jian-chao Prediction of remaining service life of rolling bearing based on BiLSTM[J]. Manufacturing Automation, 2020, 42 (5): 47- 50
doi: 10.3969/j.issn.1009-0134.2020.05.011
28 曹正志, 叶春明 基于并联CNN-SE-Bi-LSTM的轴承剩余使用寿命预测[J]. 计算机应用研究, 2021, 38 (7): 2103- 2107
CAO Zheng-zhi, YE Chun-ming Bearing remaining service life prediction based on parallel CNN-SE-Bi-LSTM[J]. Application Research of Computers, 2021, 38 (7): 2103- 2107
29 WANG Biao, LEI Ya-guo, LI Nai-peng, et al A hybrid prognostics approach for estimating remaining useful life of rolling element bearings[J]. IEEE Transactions on Reliability, 2018, 69 (1): 401- 412
30 GAL Y, GHAHRAMANI Z. Dropout as a Bayesian approximation: representing model uncertainty in deep learning [C]// International Conference on Machine Learning. New York: [s. n.], 2016: 1050-1059.
31 牟含笑, 郑建飞, 胡昌华, 等. 基于CDBN与BiLSTM的多元退化设备剩余寿命预测[EB/OL]. [2021-08-26]. http://kns.cnki.net/kcms/detail/11.1929.v. 20210510.1354.004.html.
MOU Han-xiao, ZHENG Jian-fei, HU Chang-hua, et al. Residual life prediction of multivariate degraded equipment based on CDBN and BiLSTM [EB/OL]. [2021-08-26]. http://kns.cnki.net/kcms/detail/11.1929.v.20210510.1354.004.html.
[1] 鞠晓臣,赵欣欣,钱胜胜. 基于自注意力机制的桥梁螺栓检测算法[J]. 浙江大学学报(工学版), 2022, 56(5): 901-908.
[2] 王友卫,童爽,凤丽洲,朱建明,李洋,陈福. 基于图卷积网络的归纳式微博谣言检测新方法[J]. 浙江大学学报(工学版), 2022, 56(5): 956-966.
[3] 张雪芹,李天任. 基于Cycle-GAN和改进DPN网络的乳腺癌病理图像分类[J]. 浙江大学学报(工学版), 2022, 56(4): 727-735.
[4] 许萌,王丹,李致远,陈远方. IncepA-EEGNet: 融合Inception网络和注意力机制的P300信号检测方法[J]. 浙江大学学报(工学版), 2022, 56(4): 745-753, 782.
[5] 柳长源,何先平,毕晓君. 融合注意力机制的高效率网络车型识别[J]. 浙江大学学报(工学版), 2022, 56(4): 775-782.
[6] 陈巧红,裴皓磊,孙麒. 基于视觉关系推理与上下文门控机制的图像描述[J]. 浙江大学学报(工学版), 2022, 56(3): 542-549.
[7] 农元君,王俊杰,陈红,孙文涵,耿慧,李书悦. 基于注意力机制和编码-解码架构的施工场景图像描述方法[J]. 浙江大学学报(工学版), 2022, 56(2): 236-244.
[8] 刘英莉,吴瑞刚,么长慧,沈韬. 铝硅合金实体关系抽取数据集的构建方法[J]. 浙江大学学报(工学版), 2022, 56(2): 245-253.
[9] 董红召,方浩杰,张楠. 旋转框定位的多尺度再生物品目标检测算法[J]. 浙江大学学报(工学版), 2022, 56(1): 16-25.
[10] 王鑫,陈巧红,孙麒,贾宇波. 基于关系推理与门控机制的视觉问答方法[J]. 浙江大学学报(工学版), 2022, 56(1): 36-46.
[11] 陈智超,焦海宁,杨杰,曾华福. 基于改进MobileNet v2的垃圾图像分类算法[J]. 浙江大学学报(工学版), 2021, 55(8): 1490-1499.
[12] 雍子叶,郭继昌,李重仪. 融入注意力机制的弱监督水下图像增强算法[J]. 浙江大学学报(工学版), 2021, 55(3): 555-562.
[13] 陈涵娟,达飞鹏,盖绍彦. 基于竞争注意力融合的深度三维点云分类网络[J]. 浙江大学学报(工学版), 2021, 55(12): 2342-2351.
[14] 陈岳林,田文靖,蔡晓东,郑淑婷. 基于密集连接网络和多维特征融合的文本匹配模型[J]. 浙江大学学报(工学版), 2021, 55(12): 2352-2358.
[15] 陈雪云,夏瑾,杜珂. 基于多线型特征增强网络的架空输电线检测[J]. 浙江大学学报(工学版), 2021, 55(12): 2382-2389.