特征选择与t-SNE结合的滚动轴承故障诊断

殷秀丽; 谢丽蓉; 杨欢; 段智峰

doi:10.13433/j.cnki.1003-8728.20220156

特征选择与t-SNE结合的滚动轴承故障诊断

doi: 10.13433/j.cnki.1003-8728.20220156

殷秀丽^1,,
谢丽蓉^1, ,,
杨欢^{1, 2},
段智峰¹

1.
新疆大学可再生能源发电与并网技术教育部工程研究中心，乌鲁木齐　830047
2.
清华大学电力系统及发电设备控制和仿真国家重点实验室，北京　100084

基金项目: 国家自然科学基金项目（62163034）、新疆维吾尔自治区髙校科研计划自然科学重点项目（XJDU2020I004）及新疆维吾尔自治区区域协同创新专项（科技援疆计划）(2018E02072）

详细信息

作者简介:
殷秀丽（1996−），硕士研究生，研究方向为旋转机械故障诊断，xiuliyin@163.com

通讯作者:
谢丽蓉，教授，博士生导师，xielirong@xju.edu.cn

中图分类号: TG156
计量
- 文章访问数: 682
- HTML全文浏览量: 39
- PDF下载量: 30
- 被引次数: 0
出版历程
- 收稿日期: 2021-08-10
- 刊出日期: 2023-11-30

Rolling Bearing Fault Diagnosis Combined Feature Selectionwith t-distributed Stochastic Neighbor Embedding

YIN Xiuli^1
,,
XIE Lirong^{1
, ,},
YANG Huan^{1, 2},
DUAN Zhifeng¹

1.
Engineering Research Center for Renewable Energy Power Generation and Grid Technology of Ministry of Education, Xinjiang University, Urumqi 830047, China
2.
State Key Laboratory of Power System and Generation Equipment, Tsinghua University, Beijing 100084, China

摘要

摘要: 为准确识别滚动轴承当前故障状态，提出一种集合经验模态分解（EEMD）、特征选择与t-分布邻域嵌入（t-SNE）的诊断方法。采用EEMD分解故障信号获得若干本征模态函数（IMF），采用峭度准则筛选有效IMF分量并重构；求出重构信号的高维时、频域特征矩阵并对其归一化，采用t-SNE算法获得对故障状态更敏感的低维特征矩阵；将特征矩阵输入粒子群优化的最小二乘支持向量机（LSSVM）中，实现轴承的故障识别与诊断。采用实验分析并对比几种典型的降维法，证明了t-SNE的优越性，所提方法可以实现故障状态的100%识别，验证了该方法的有效性。
- 故障诊断 /
- 集合经验模态分解 /
- 特征选择 /
- t-分布邻域嵌入
Abstract: To identify the fault state of rolling bearing accurately, a diagnosis method is proposed which is based on ensemble empirical mode decomposition (EEMD) and t-distribution stochastic neighborhood embedding (t-SNE). Firstly, EEMD is used to decompose vibration fault signal to obtain several intrinsic mode functions (IMF), and the kurtosis criterion is used to select effective IMF components and reconstruct them. Secondly, the high dimensional time-frequency characteristic matrix of the reconstructed signal is obtained and normalized, and a low dimensional characteristic matrix which is more sensitive to fault states is obtained by t-SNE algorithm. Finally, characteristic matrix is input into the least squares support vector machine (LSSVM) classifier which is optimized by particle swarm optimization algorithm to realize the fault state recognition and diagnosis of bearing. In the case analysis, the bearing state data of Western Reserve University and MFPT is used, results show that the advantages of t-SNE by comparing several typical dimension reduction methods and the proposed method can realize 100% fault recognition.
- fault diagnosis /
- ensemble empirical mode decomposition (EEMD) /
- time-frequency domain characteristics /
- t-distributed stochastic neighbor embedding (t-SNE)

HTML全文

图 1 基于EEMD和t-SNE算法的滚动轴承故障诊断总体思路图

Figure 1. General idea diagram of rolling bearing fault diagnosis based on EEMD and t-SNE algorithm

下载: 全尺寸图片幻灯片

图 2 轴承状态原始信号图

Figure 2. Original signal diagram of bearing conditions

下载: 全尺寸图片幻灯片

图 3 EEMD算法分解滚动体故障信号图

Figure 3. Decomposition of rolling element fault signals with the EEMD algorithm

下载: 全尺寸图片幻灯片

图 4 4种信号IMF分量峭度值

Figure 4. The kurtosis of the four signals′ IMF components

下载: 全尺寸图片幻灯片

图 5 去噪后4种轴承信号

Figure 5. Four kinds of bearing signal after denoising

下载: 全尺寸图片幻灯片

图 6 9种典型降维算法结果对比图

Figure 6. Comparison of results of nine typical dimension reduction algorithms

下载: 全尺寸图片幻灯片

图 7 MFPT的3种轴承状态信号图

Figure 7. Three kinds of MFPT′s bearing state signals

下载: 全尺寸图片幻灯片

图 8 EEMD分解外圈故障信号图

Figure 8. Decomposing the outer ring fault signals with the EEMD algorithm

下载: 全尺寸图片幻灯片

图 9 3种信号IMF分量峭度值

Figure 9. IMF component kurtosis values of three types ofsignals

下载: 全尺寸图片幻灯片

图 10 MFPT的3种信号去噪信号图

Figure 10. MFPT′s three kinds of denoised signals

下载: 全尺寸图片幻灯片

图 11 t-SNE和SNE降维结果对比

Figure 11. Comparison of dimensionality reduction results of t-SNE and SNE

下载: 全尺寸图片幻灯片

表 1 特征参数分类

Table 1. Characteristic parameter classification

类别		特征参数
时域特征参数	有量纲	最大值、最小值、峰峰值、平均值、均方根、方差、标准差、峰值因子
时域特征参数	无量纲	偏斜度、峭度、脉冲因子、裕度因子
频域特征参数		均方频率、均方根频率、频率方差、频率标准差

下载: 导出CSV

表 2 分类准确率及时间对比

Table 2. Classification accuracy and time comparison

分类器	测试数据分类准确率/%	分类时间/s
PSO-LSSVM	100	0.1305
SVM	85	2.2453
CNN	97.45	15.8864
LSTM	95.55	23.4521
RF	95	0.3580

下载: 导出CSV

表 3 MFPT轴承数据情况

Table 3. Conditions of MFPT′s bearing data

类别	负载/lbf	输入轴转速/Hz	采样速率/sps	持续时间/s
基线状态	270	25	97656	6
内圈故障	250	25	48828	3
外圈故障	250	25	48828	3

下载: 导出CSV

表 4 分类准确率及时间对比

Table 4. Classification accuracy and time comparison

分类器	测试数据分类准确率/%	分类时间/s
PSO-LSSVM	100	0.1142
SVM	83.33	1.9827
CNN	98.24	11.2759
LSTM	94.85	12.5783
RF	93.33	0.2790

下载: 导出CSV

参考文献(20)

[1]	王晓龙. 基于振动信号处理的滚动轴承故障诊断方法研究[D]. 北京: 华北电力大学, 2017. WANG X L. Research on fault diagnosis method of rolling bearing based on vibration signal processing[D]. Beijing: North China Electric Power University, 2017. (in Chinese)
[2]	HUANG N E, SHEN Z, LONG S R, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceedings of the Royal Society A:Mathematical, Physical and Engineering Sciences, 1998, 454(1971): 903-995. doi: 10.1098/rspa.1998.0193
[3]	HUANG N E, SHEN Z, LONG S R. A new view of nonlinear water waves: The Hilbert spectrum[J]. Annual Review of Fluid Mechanics, 1999, 31: 417-457. doi: 10.1146/annurev.fluid.31.1.417
[4]	WU Z H, HUANG N E. Ensemble empirical mode decomposition: a noise-assisted data analysis method[J]. Advances in Adaptive Data Analysis, 2009, 1(1): 1-41. doi: 10.1142/S1793536909000047
[5]	周建民, 王发令, 张臣臣, 等. 基于特征优选和GA-SVM的滚动轴承智能评估方法[J]. 振动与冲击, 2021, 40(4): 227-234. ZHOU J M, WANG F L, ZHANG C C, et al. An intelligent method for rolling bearing evaluation using feature optimization and GA-SVM[J]. Journal of Vibration and Shock, 2021, 40(4): 227-234. (in Chinese)
[6]	张淑清, 黄文静, 胡永涛, 等. 基于总体平均经验模式分解近似熵和混合PSO-BP算法的轴承故障诊断方法[J]. 中国机械工程, 2016, 27(22): 3048-3054. doi: 10.3969/j.issn.1004-132X.2016.22.012 ZHANG S Q, HUANG W J, HU Y T, et al. Bearing fault diagnosis method based on EEMD approximate entropy and hybrid PSO-BP algorithm[J]. China Mechanical Engineering, 2016, 27(22): 3048-3054. (in Chinese) doi: 10.3969/j.issn.1004-132X.2016.22.012
[7]	高彩霞, 吴彤, 付子义. 线性回归与EEMD的滚动轴承剩余寿命预测[J]. 机械科学与技术, 2019, 38(10): 1589-1597. GAO C X, WU T, FU Z Y. Remaining useful life prediction for rolling bearings based on linear regression and EEMD[J]. Mechanical Science and Technology for Aerospace Engineering, 2019, 38(10): 1589-1597. (in Chinese)
[8]	张琛, 赵荣珍, 邓林峰. 基于EEMD奇异值熵的滚动轴承故障诊断方法[J]. 振动、测试与诊断, 2019, 39(2): 353-358. ZHANG C, ZHAO R Z, DENG L F. Rolling bearing fault diagnosis method based on EEMD singular value entropy[J]. Journal of Vibration, Measurement & Diagnosis, 2019, 39(2): 353-358. (in Chinese)
[9]	胡超, 沈宝国, 谢中敏. 基于EMD-FastICA与DGA-ELM网络的轴承故障诊断方法[J]. 太阳能学报, 2021, 42(10): 208-219. HU C, SHEN B G, XIE Z M. Fault diagnosis method of bearing based on EMD-FastICA and DGA-ELM network[J]. Acta Energiae Solaris Sinica, 2021, 42(10): 208-219. (in Chinese)
[10]	陈俊洵, 程龙生, 胡绍林, 等. 基于EMD的改进马田系统的滚动轴承故障诊断[J]. 振动与冲击, 2017, 36(5): 151-156. CHEN J X, CHENG L S, HU S L, et al. Fault diagnosis of rolling bearings using modified Mahalanobis-Taguchi system based on EMD[J]. Journal of Vibration and Shock, 2017, 36(5): 151-156. (in Chinese)
[11]	许凡, 方彦军, 张荣. 基于EEMD模糊熵的PCA-GG滚动轴承聚类故障诊断[J]. 计算机集成制造系统, 2016, 22(11): 2631-2642. XU F, FANG Y J, ZHANG R. PCA-GG rolling bearing clustering fault diagnosis based on EEMD fuzzy entropy[J]. Computer Integrated Manufacturing Systems, 2016, 22(11): 2631-2642. (in Chinese)
[12]	王望望, 邓林峰, 赵荣珍, 等. 集成KPCA与t-SNE的滚动轴承故障特征提取方法[J]. 振动工程学报, 2021, 34(2): 431-440. WANG W W, DENG L F, ZHAO R Z, et al. Fault feature extraction of rolling bearing integrating KPCA and t-SNE[J]. Journal of Vibration Engineering, 2021, 34(2): 431-440. (in Chinese)
[13]	丁承君, 张良, 冯玉伯, 等. VMD和t-SNE相结合的滚动轴承故障诊断[J]. 机械科学与技术, 2020, 39(5): 758-764. DING C J, ZHANG L, FENG Y B, et al. Fault diagnosis method of rolling bearing combining VMD with t-SNE[J]. Mechanical Science and Technology for Aerospace Engineering, 2020, 39(5): 758-764. (in Chinese)
[14]	GAO Q, DUAN C, FAN H, et al. Rotating machine fault diagnosis using empirical mode decomposition[J]. Mechanical Systems and Signal Processing, 2008, 22(5): 1072-1081. doi: 10.1016/j.ymssp.2007.10.003
[15]	杨望灿, 张培林, 王怀光, 等. 基于EEMD的多尺度模糊熵的齿轮故障诊断[J]. 振动与冲击, 2015, 34(14): 163-167. YANG W C, ZHANG P L, WANG H G, et al. Gear fault diagnosis based on multiscale fuzzy entropy of EEMD[J]. Journal of Vibration and Shock, 2015, 34(14): 163-167. (in Chinese)
[16]	胡爱军, 马万里, 唐贵基. 基于集成经验模态分解和峭度准则的滚动轴承故障特征提取方法[J]. 中国电机工程学报, 2012, 32(11): 106-111. doi: 10.13334/j.0258-8013.pcsee.2012.11.006 HU A J, MA W L, TANG G J. Rolling bearing fault feature extraction method based on ensemble empirical mode decomposition and kurtosis criterion[J]. Proceedings of the CSEE, 2012, 32(11): 106-111. (in Chinese) doi: 10.13334/j.0258-8013.pcsee.2012.11.006
[17]	SAMUEL P D, PINES D J. A review of vibration-based techniques for helicopter transmission diagnostics[J]. Journal of Sound and Vibration, 2005, 282(1-2): 475-508. doi: 10.1016/j.jsv.2004.02.058
[18]	LEBOLD M, MCCLINTIC K, CAMPBELL R, et al. Review of vibration analysis methods for gearbox diagnostics and prognostics[C]//Proceedings of the 54^th Meeting of the Society for Machinery Failure Prevention Technology. Virginia Beach, 2000: 623-634
[19]	李海平, 赵建民, 宋文渊. 基于EMD-EDT的行星齿轮箱特征提取及状态识别方法研究[J]. 振动与冲击, 2016, 35(3): 48-54. doi: 10.13465/j.cnki.jvs.2016.03.008 LI H P, ZHAO J M, SONG W Y. Method of planetary gearbox feature extraction and condition recognition based on EMD and EDT[J]. Journal of Vibration and Shock, 2016, 35(3): 48-54. (in Chinese) doi: 10.13465/j.cnki.jvs.2016.03.008
[20]	VAN DER MAATEN L, HINTON G. Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008, 9(2605): 2579-2605.