2.793

2018影响因子

(CJCR)

  • 中文核心
  • EI
  • 中国科技核心
  • Scopus
  • CSCD
  • 英国科学文摘

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

一种基于深度迁移学习的滚动轴承早期故障在线检测方法

毛文涛 田思雨 窦智 张迪 丁玲

毛文涛, 田思雨, 窦智, 张迪, 丁玲. 一种基于深度迁移学习的滚动轴承早期故障在线检测方法. 自动化学报, 2020, 45(x): 1−13 doi: 10.16383/j.aas.c190593
引用本文: 毛文涛, 田思雨, 窦智, 张迪, 丁玲. 一种基于深度迁移学习的滚动轴承早期故障在线检测方法. 自动化学报, 2020, 45(x): 1−13 doi: 10.16383/j.aas.c190593
Mao Wen-Tao, Tian Si-Yu, Dou Zhi, Zhang Di, Ding Ling. A new deep transfer learning-based online detection method of rolling bearing early fault. Acta Automatica Sinica, 2020, 45(x): 1−13 doi: 10.16383/j.aas.c190593
Citation: Mao Wen-Tao, Tian Si-Yu, Dou Zhi, Zhang Di, Ding Ling. A new deep transfer learning-based online detection method of rolling bearing early fault. Acta Automatica Sinica, 2020, 45(x): 1−13 doi: 10.16383/j.aas.c190593

一种基于深度迁移学习的滚动轴承早期故障在线检测方法

doi: 10.16383/j.aas.c190593
基金项目: 本文受国家重点研发计划重点专项项目(2018YFB1701400)和国家自然科学基金(U1704158)资助
详细信息
    作者简介:

    毛文涛:河南师范大学计算机与信息工程学院副教授. 主要研究方向为机器学习, 时间序列预测.E-mail: maowt@htu.edu.cn

    田思雨:河南师范大学计算机与信息工程学院研究生. 主要研究方向为深度学习与故障检测.E-mail: tianxiaosi77@163.com

    窦智:河南师范大学计算机与信息工程学院副教授. 主要研究方向为图像处理, 可重构计算, 机器学习.E-mail: 2015160@htu.edu.cn

    张迪:河南师范大学计算机与信息工程学院研究生. 主要研究方向为深度学习与异常检测.E-mail: vencent8692@gmail.com

    丁玲:河南师范大学计算机与信息工程学院研究生. 主要研究方向为深度学习与故障检测.E-mail: ddll1029@163.com

A New Deep Transfer Learning-based Online Detection Method of Rolling Bearing Early Fault

Funds: Supported by the National Key R&D Program of China (No. 2018YFB1701400) and the National Natural Science Foundation of China (No. U1704158)
  • 摘要: 近年来, 深度学习技术已在滚动轴承故障检测和诊断领域取得了成功应用, 但面对不停机情况下的早期故障在线检测问题, 仍存在着早期故障特征表示不充分、误报警率高等不足. 为解决上述问题, 本文从时序异常检测的角度出发, 提出了一种基于深度迁移学习的早期故障在线检测方法. 首先, 提出一种面向多域迁移的深度自编码网络, 通过构建具有改进的最大均值差异正则项和Laplace正则项的损失函数, 在自适应提取不同域数据的公共特征表示同时, 提高正常状态和早期故障状态之间特征的差异性; 基于该特征表示, 提出一种基于时序异常模式的在线检测模型, 利用离线轴承正常状态的排列熵值构建报警阈值, 实现在线数据中异常序列的快速匹配, 同时提高在线检测结果的可靠性. 在XJTU-SY数据集上的实验结果表明, 与现有代表性早期故障检测方法相比, 本文方法具有更好的检测实时性和更低的误报警数.
  • 图  1  多域迁移深度自编码网络结构图

    Fig.  1  Architecture of deep auto-encoder network with multi-domain transferring

    图  2  本文方法所提取的公共特征及其对应的排列熵值

    Fig.  2  The extracted common features and their corresponding permutation values

    图  3  图2的局部放大图

    Fig.  3  Partial enlarged view of Fig. 2

    图  4  所提方法流程图

    Fig.  4  Flow chart of the proposed method

    图  5  实验平台[23]

    Fig.  5  Experiment platform[23]

    图  6  采用多层深度自编码器提取的三种工况下轴承正常状态数据的特征分布

    Fig.  6  Feature distribution of bearing normal state data extracted by multi-layer deep auto-encoder under three working conditions

    图  7  采用本文方法提取的三种工况下轴承正常状态数据的特征分布

    Fig.  7  Feature distribution of bearing normal state data extracted by the proposed method under three working conditions

    图  8  本文方法训练损失变化趋势图

    Fig.  8  Changing trend of training loss of the proposed method

    图  9  三种工况下15个轴承正常状态数据的排列熵

    Fig.  9  Permutation entropy of 15 bearings in normal state under all three working conditions

    图  10  工况1下轴承5的特征走势及对应的排列熵值, 其中红线与蓝线分别为加入和未加入Laplace正则项的本文方法所对应效果

    Fig.  10  Feature trend and the corresponding permutation entropy value of bearing 5 under working condition 1, where the red and blue curves show the effect with or without Laplace regularizer, respectively

    图  11  工况2下轴承4的特征走势及对应的排列熵值, 其中红线与蓝线分别为加入和未加入Laplace正则项的本文方法所对应效果

    Fig.  11  Feature trend and the corresponding permutation entropy value of bearing 4 under working condition 2, where the red and blue curves show the effect with or without Laplace regularizer, respectively

    图  12  Bearing1-5轴承的检测结果

    Fig.  12  Detection results of bearing 1-5

    图  13  Bearing2-4轴承的检测结果

    Fig.  13  Detection results of bearing 2-4

    图  14  本文方法与LOF算法的检测结果对比图, 其中, 本文方法横坐标为序列编号(即样本号除以100), 标签值大于0表示样本识别为正常样本, 小于0表示识别为异常样本

    Fig.  14  Comparative detection results between the proposed method and LOF algorithm, while x-coordinate of the proposed method denotes sequence number (equal to the sample number divided by 100), the label value greater than 0 means the corresponding sample is recognized as normal sample, else as anomaly

    表  1  本文所用符号与对应描述

    Table  1  Symbols and corresponding descriptions used in this paper

    符号描述符号描述
    ${\cal D}{^{ s}}$源域${\cal D}{^{ t}}$目标域
    $W,b$自编码器的权重矩阵和偏置$f,g$激活函数
    ${s_c}$工况数量$m$监测数据的组数
    $X$数据矩阵$Y$标签矩阵
    ${{x}}$单个样本$y$单个样本的标签
    $n$样本数量$H$特征矩阵
    $p$数据分布$W'$$W$的转置
    下载: 导出CSV

    表  2  XJTU-SY数据集中三种工况描述

    Table  2  Description of three working conditions in XJTU-SY dataset

    运行条件径向力(kN)转速(rmp)轴承数据集
    工况1122100Bearing 1_1 Bearing 1_2 Bearing 1_3 Bearing 1_4 Bearing 1_5
    工况2112250Bearing 2_1 Bearing 2_2 Bearing 2_3 Bearing 2_4 Bearing 2_5
    工况3102400Bearing 3_1 Bearing 3_2 Bearing 3_3 Bearing 3_4 Bearing 3_5
    下载: 导出CSV

    表  3  七种方法对比检测结果

    Table  3  Comparative detection results of seven methods

    检测方法Bearing1_5检测结果误报警数Bearing2_4检测结果误报警数
    BEMD-AMMA[25]111000
    局部异常值检测(LOF)[26]1096359621
    iFOREST[27]12501096013
    SDFM[28]115009800
    DAFD[18]1210011000
    SRD[29]117069652
    本文方法109009600
    下载: 导出CSV
  • [1] Lei Y G, Li N P, Guo L, Li N B, Yan T, Lin J. Machinery health prognostics: A systematic review from data acquisition to RUL prediction. Mechanical Systems and Signal Processing, 2018, 104: 799−843
    [2] Liu R N, Yang B Y, Zhang X L, Wang S B, Chen X F. Time-frequency atoms-driven support vector machine method for bearings incipient fault diagnosis. Mechanical Systems and Signal Processing, 2016, 75: 345−370
    [3] 杨蕊, 李宏坤, 贺长波. 利用最优小波尺度循环谱的滚动轴承早期故障特征提取. 机械工程学报, 2017, 54(17): 209−217

    Yang Rui, Li Hong-Kun, He Chang-Bo. Early fault feature extraction of rolling bearing by using optimal wavelet scale cyclic spectrum. Journal of Mechanical Engineering, 2017, 54(17): 209−217
    [4] 孙鲜明, 刘欢, 赵新光. 基于瞬时包络尺度谱熵的滚动轴承早期故障奇异点识别及特征提取. 机械工程学报, 2017, 53(3): 73−80 doi: 10.3901/JME.2017.03.073

    Sun Xian-Ming, Liu Huan, Zhao Xin-Guang. Identification and feature extraction of early fault singularity of rolling bearing based on instantaneous envelope scale spectrum entropy. Journal of Mechanical Engineering, 2017, 53(3): 73−80 doi: 10.3901/JME.2017.03.073
    [5] Dhamande L S, Chaudhari M B. Compound gear-bearing fault feature extraction using statistical features based on time-frequency method. Measurement, 2018, 125: 63−77 doi: 10.1016/j.measurement.2018.04.059
    [6] Li F, Wang J X, Chyu M K, Tang B P. Weak fault diagnosis of rotating machinery based on feature reduction with Supervised Orthogonal Local Fisher Discriminant Analysis. Neurocomputing, 2015, 168: 505−519
    [7] Amar M, Gondal I, Wilson C. Vibration spectrum imaging: A novel bearing fault classification approach. IEEE Transactions on Industrial Electronics, 2014, 62(1): 494−502
    [8] Fernández-Francos D, Martínez-Rego D, Fontenla-Romero O, Alonso-Betanzos A. Automatic bearing fault diagnosis based on one-class v-SVM. Computers & Industrial Engineering, 2013, 64(1): 357−365
    [9] 张柯, 姜斌. 基于故障诊断观测器的输出反馈容错控制设计. 自动化学报, 2010, 36(2): 274−281 doi: 10.3724/SP.J.1004.2010.00274

    Zhang Ke, Jiang Bin. Fault diagnosis observer-based output feedback fault tolerant control design. Acta Automatica Sinica, 2010, 36(2): 274−281 doi: 10.3724/SP.J.1004.2010.00274
    [10] Jia F, Lei Y G, Guo L. A neural network constructed by deep learning technique and its application to intelligent fault diagnosis of machines. Neurocomputing, 2018, 272: 619−628 doi: 10.1016/j.neucom.2017.07.032
    [11] Shao H D, Jiang H K, Zhang X. Rolling bearing fault diagnosis using an optimization deep belief network. Measurement Science and Technology, 2015, 26(11): 115002
    [12] 庄福振, 罗平, 何清, 史忠植. 迁移学习研究进展. 软件学报, 2015, 26(1): 26−39

    Zhuang Fu-Zhen, Luo Ping, He Qing, Shi Zhong-Zhi. Research progress in transfer learning.Research progress in transfer learning. Journal of Software, 2015, 26(1): 26−39
    [13] 贺敏, 汤健, 郭旭琦, 阎高伟. 基于流形正则化域适应随机权神经网络的湿式球磨机负荷参数软测量. 自动化学报, 2019, 45(2): 398−406

    He Min, Tang Jian, Guo Xu-Qi, Yan Gao-Wei. Soft sensor for ball mill load using DAMRRWNN model. Acta Automatica Sinica, 2019, 45(2): 398−406
    [14] 张雪松, 庄严, 闫飞, 王伟. 基于迁移学习的类别级物体识别与检测研究与进展. 自动化学报, 2019, 45(7): 1224−1243

    Zhang Xue-Song, Zhuang Yan, Yan Fei, Wang Wei. Status and development of transfer learning based category-level object recognition and detection. Acta Automatica Sinica, 2019, 45(7): 1224−1243
    [15] Shen S, Sadoughi M, Li M. Deep convolutional neural networks with ensemble learning and transfer learning for capacity estimation of lithium-ion batteries. Applied Energy, 2020, 260: 114296 doi: 10.1016/j.apenergy.2019.114296
    [16] Din I U, Islam N, Rodrigues J. A novel deep learning based framework for the detection and classification of breast cancer using transfer learning. Pattern Recognition Letters, 2019, 125(1): 1−6
    [17] 雷亚国, 杨彬, 杜兆钧, 吕娜. 大数据下机械装备故障的深度迁移诊断方法. 机械工程学报, 2018, 55(7): 1−8

    Lei Ya-Guo, Yang Bin, Du Zhao-Jun, Lv Na. Deep transfer diagnostic method for mechanical equipment failure under big data. Journal of Mechanical Engineering, 2018, 55(7): 1−8
    [18] Lu W N, Liang B, Cheng Y, Meng D S, Yang J, Zhang T. Deep model based domain adaptation for fault diagnosis. IEEE Transactions on Industrial Electronics, 2017, 64(3): 2296−2305 doi: 10.1109/TIE.2016.2627020
    [19] Wen L, Gao L, Li X. A new deep transfer learning based on sparse auto-encoder for fault diagnosis. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2019, 49(1): 136−144 doi: 10.1109/TSMC.2017.2754287
    [20] Yang B, Lei Y, Jia F, Xing S. An intelligent fault diagnosis approach based on transfer learning from laboratory bearings to locomotive bearings. Mechanical Systems and Signal Processing, 2019, 122: 692−706
    [21] Cai D, He X F, Hu Y X, Han J W. Learning a spatially smooth subspace for face recognition // 2007 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). IEEE Computer Society, Minneapolis, USA, June 2007, pp. 1−7.
    [22] 冯辅周, 饶国强, 司爱威. 基于排列熵和神经网络的滚动轴承异常检测与诊断. 噪声与振动控制, 2013, 33(3): 212−217

    Feng Fu-Zhou, Rao Guo-Qiang, Si Ai-Wei. Abnormality detection and diagnosis of rolling bearing based on permutation entropy and neural network. Noise and Vibration Control, 2013, 33(3): 212−217
    [23] Wang B, Lei Y G, Li N P, Li N B. A hybrid prognostics approach for estimating remaining useful life of rolling element bearings. IEEE Transactions on Reliability, 2020, 69(1): 401−412
    [24] Cho K, Courville A, Bengio Y. Describing multimedia content using attention-based encoder-decoder networks. IEEE Transactions on Multimedia, 2015, 17(11): 1875−1886 doi: 10.1109/TMM.2015.2477044
    [25] Li Y B, Xu M Q, Liang X H, Huang W H. Application of bandwidth EMD and adaptive multi-scale morphology analysis for incipient fault diagnosis of rolling bearings. IEEE Transactions on Industrial Electronics, 2017, 64(8): 6506−6517 doi: 10.1109/TIE.2017.2650873
    [26] Ma H H, Hu Y, Shi H B. Fault detection and identification based on the neighborhood standardized local outlier factor method. Industrial Engineering Chemistry Research, 2013, 52(6): 2389−2402
    [27] Domingues R, Filippone M, Michiardi P. A comparative evaluation of outlier detection algorithms: Experiments and analyses. Pattern Recognition, 2018, 74: 406−421
    [28] Mao W, Chen J, Liang X. A New Online Detection Approach for Rolling Bearing Incipient Fault via Self-Adaptive Deep Feature Matching. IEEE Transactions on Instrumentation and Measurement, 2020, 69(2): 443−456 doi: 10.1109/TIM.2019.2903699
    [29] 郭小萍, 刘诗洋, 李元. 基于稀疏残差距离的多工况过程故障检测方法研究. 自动化学报, 2019, 45(3): 617−625

    Guo Xiao-Ping, Liu Shi-Yang, Li Yuan. Fault detection of multi-mode processes employing sparse residual distance. Acta Automatica Sinica, 2019, 45(3): 617−625
  • 加载中
计量
  • 文章访问数:  81
  • HTML全文浏览量:  28
  • 被引次数: 0
出版历程
  • 收稿日期:  2019-08-18
  • 录用日期:  2020-04-27

目录

    /

    返回文章
    返回