Fault diagnosis method of planetary gearbox based on JS-VME-DBN and MS-UMAP
-
摘要:
为了解决行星齿轮箱振动信号存在噪声干扰和特征提取困难的问题,提出一种基于水母搜索优化变分模态提取(JS-VME)、深度置信网络(DBN)和监督型马氏距离的均匀流形逼近与投影算法(MS-UMAP)的行星齿轮箱故障诊断方法。采集行星齿轮箱的振动信号,利用JS-VME对其进行预处理,获得相关性较强的期望IMF(intrinsic mode function)分量;然后将该IMF分量应用DBN提取特征向量,构建高维故障特征集;采用MS-UMAP进行维数约减,获得低维、敏感的故障特征;将低维故障特征集应用水母搜索优化核极限学习机(JS-KELM)判别故障类型。行星齿轮箱故障诊断实验结果表明:与UMAP、t-SNE、Isomap、LPP、W-Isomap、LLE、LTSA和MDS等方法相比,MS-UMAP算法对JS-VME-DBN的特征提取结果有着最佳的降维效果,所提方法对行星齿轮箱的裂纹、磨损和缺齿等故障的识别率达到了100%,具有一定的有效性。
-
关键词:
- 故障诊断 /
- 行星齿轮箱 /
- 变分模态提取(VME) /
- 深度置信网络(DBN) /
- 均匀流行逼近与投影算法(UMAP) /
- 核极限学习机(KELM)
Abstract:In order to solve the problem of the noise interference and the difficulty in feature extraction in the vibration signal of planetary gearbox, a fault diagnosis method for planetary gearboxes based on jel-lyfish search optimization variational mode extraction (JS-VME), deep belief network (DBN) and supervised Mahalanobis distance uniform manifold approximation and projection algorithms (MS-UMAP) was proposed. The vibration signals of the planetary gearbox were collected, and JS-VME was used to preprocess them to obtain expected IMF (intrinsic mode function)component with strong correlation. Then, DBN was applied to the IMF component to extract feature vectors, and the high-dimensional fault feature set was built. MS-UMAP was used for dimensionality reduc-tion to obtain low-dimensional and sensitive fault features. The low-dimensional fault feature set was applied to the jellyfish search optimization kernel extreme learning machine (JS-KELM) to de-termine fault types. The experiment results of planetary gearbox fault diagnosis showed that com-pared with UMAP, t-SNE, Isomap, LPP, W-Isomap, LLE, LTSA and MDS, the MS-UMAP algorithm had the best dimensionality reduction effect on the feature extraction results of JS-VME-DBN. The fault recognition rate of the proposed method reached 100% with a certain validity in planetary gearbox, such as the cracks, wear and missing teeth.
-
表 1 6种数据集描述
Table 1. Descriptionof six data sets
数据集 总样本数 各样本数 训练样本 测试样本 sonar_all_data 208 97, 111 20, 22 77, 89 wine 178 59, 71, 48 12, 14, 10 47, 57, 38 iris 150 50, 50, 50 10, 10, 10 40, 40, 40 diabetes 768 500, 268 100, 54 400, 214 tic_tac_toc 958 626, 332 125, 67 501, 265 splice 1000 483, 517 97, 103 386, 414 表 2 分类器参数设置
Table 2. Classifier parameter settings
分类器 参数设置 PSO-KELM 局部搜索能力为2,全局搜索能力为2,
粒子群数为10,终止迭代为100GWO-KELM 灰狼种群规模数为10,终止迭代为100 ABC-KELM 蜂群规模数为10,初始蜜源为10,
控制参数为100,最大循环次数为100AO-KELM 金雕种群规模数为10,终止迭代为100 BA-KELM 蝙蝠种群规模设置为10,响度为0.5,
脉冲重复频率为0.01,最高频率为0.5,
最低频率为0.2,终止迭代次数为100JS-KELM 水母种群规模数为10,终止迭代为100 表 3 行星齿轮箱基本参数
Table 3. Basic parameters of planetary gearbox
级数 齿轮 齿数 级数 齿轮 齿数 1 太阳轮 20 2 太阳轮 28 行星轮 40 行星轮 36 内齿圈 100 内齿圈 100 表 4 JS-VME-DBN的特征提取效果
Table 4. Feature extraction effect of JS-VME-DBN
分类方法 不同分类器对测试样本故障识别精度/% 正常 裂纹 磨损 缺齿 平均 PSO-KELM 52.50 100 80.00 100 83.13 ABC-KELM 87.50 100 82.50 100 92.50 GWO-KELM 97.50 100 67.50 100 91.25 AO-KELM 77.50 100 100 100 94.38 BA-KELM 40.00 100 77.50 100 79.38 JS-KELM 100 100 90.00 100 97.50 表 5 9种算法降维后的故障识别精度和性能指标
Table 5. Fault identification accuracy and performance index after dimensionality reduction of nine algorithms
降维方法 多故障分类器对测试样本故障识别精度/% 降维性能指标 正常 裂纹 磨损 缺齿 平均 ${S_{\rm{b}}}$ ${S_{\rm{w}}}$ ${S_{\rm{b}}}$/${S_{\rm{w}}}$ t-SNE 100 100 90.00 82.50 93.13 428.53 43.06 9.95 Isomap 85.00 100 100 100 95.00 2.33$ \times $10−4 2.29$ \times $10−5 10.17 W-Isomap 97.50 100 100 90.00 96.88 1.16$ \times $10−4 4.67$ \times $10−6 24.84 LPP 40.00 100 72.50 100 78.13 1.32$ \times $10−4 7.11$ \times $10−5 1.86 LLE 100 60.00 100 100 90.00 1.28$ \times $10−2 2.11$ \times $10−3 6.07 LTSA 100 80.00 82.50 100 90.63 6.20$ \times $10−3 8.79$ \times $10−3 0.71 MDS 100 57.50 100 100 89.38 2.31$ \times $10−4 2.29$ \times $10−5 10.10 UMAP 52.50 100 100 100 88.13 59.92 36.14 1.66 MS-UMAP 100 100 100 100 100 82.52 1.40 58.94 -
[1] 曹蔚, 苟臻元, 韩昭, 等. 基于GWO-TVF-EMD方法的行星齿轮箱齿面剥落故障诊断[J]. 航空动力学报, 2021, 36(11): 2418-2429.CAO Wei, GOU Zhenyuan, HAN Zhao, et al. Fault diagnosis of tooth surface spalling of planetary gearbox based on GWO-TVF-EMD method[J]. Journal of Aerospace Power, 2021, 36(11): 2418-2429. (in Chinese) [2] 丁闯,张兵志,冯辅周,等. 局部均值分解和排列熵在行星齿轮箱故障诊断中的应用[J]. 振动与冲击,2017,36(17): 55-60.DING Chuang,ZHANG Bingzhi,FENG Fuzhou,et al. Application of local mean decomposition and permutation entropy in fault diagnosis of planetary gearboxes[J]. Journal of Vibration and Shock,2017,36(17): 55-60. (in Chinese) [3] 何田,林意洲,郜普刚,等. 局部均值分解在齿轮故障诊断中的应用研究[J]. 振动与冲击,2011,30(6): 196-201.HE Tian,LIN Yizhou,GAO Pugang,et al. Application of local mean decomposition in gear fault diagnosis[J]. Journal of Vibration and Shock,2011,30(6): 196-201. (in Chinese) [4] 邓博元,崔伟成,曾庆松,等. 基于信号经验模态分解与分集合并的齿轮故障诊断[J]. 计算机测量与控制,2021,29(9): 43-49.DENG Boyuan,CUI Weicheng,ZENG Qingsong,et al. Gear fault diagnosis based on empirical mode decomposition and signal diversity merging[J]. Computer Measurement & Control,2021,29(9): 43-49. (in Chinese) [5] 郑近德,姜战伟,代俊习,等. 基于VMD的自适应复合多尺度模糊熵及其在滚动轴承故障诊断中的应用[J]. 航空动力学报,2017,32(7): 1683-1689.ZHENG Jinde,JIANG Zhanwei,DAI Junxi,et al. VMD based adaptive composite multiscale fuzzy entropy and its application to fault diagnosis of rolling bearing[J]. Journal of Aerospace Power,2017,32(7): 1683-1689. (in Chinese) [6] 唐贵基,王晓龙. 变分模态分解方法及其在滚动轴承早期故障诊断中的应用[J]. 振动工程学报,2016,29(4): 638-648.TANG Guiji,WANG Xiaolong. Variational mode decomposition method and its application on incipient fault diagnosis of rolling bearing[J]. Journal of Vibration Engineering,2016,29(4): 638-648. (in Chinese) [7] 向玲,鄢小安. 汽轮机转子故障诊断中LMD法和EMD法的性能对比研究[J]. 动力工程学报,2014,34(12): 945-951.XIANG Ling,YAN Xiaoan. Performance contrast between LMD and EMD in fault diagnosis of turbine rotors[J]. Journal of Chinese Society of Power Engineering,2014,34(12): 945-951. (in Chinese) [8] 陈鹏,赵小强. 基于优化VMD与改进阈值降噪的滚动轴承早期故障特征提取[J]. 振动与冲击,2021,40(13): 146-153.CHEN Peng,ZHAO Xiaoqiang. Early fault feature extraction of rolling bearing based on optimized VMD and improved threshold denoising[J]. Journal of Vibration and Shock,2021,40(13): 146-153. (in Chinese) [9] NAZARI M,SAKHAEI S M,NAZARI M,et al. Variational mode extraction: a new efficient method to derive respiratory signals from ECG[J]. IEEE Journal of Biomedical and Health Informatics,2018,22(4): 1059-1067. doi: 10.1109/JBHI.2017.2734074 [10] CHOU J S,TRUONG D N. A novel metaheuristic optimizer inspired by behavior of jellyfish in ocean[J]. Applied Mathematics and Computation,2021,389: 125535. doi: 10.1016/j.amc.2020.125535 [11] SHI Huaitao,GUO Jin,BAI Xiaotian,et al. Research on a nonlinear dynamic incipient fault detection method for rolling bearings[J]. Applied Sciences,2020,10(7): 2443. doi: 10.3390/app10072443 [12] HINTON G E,SALAKHUTDINOV R R. Reducing the dimensionality of data with neural networks[J]. Science,2006,313(5786): 504-507. doi: 10.1126/science.1127647 [13] 杨大炼,张帆宇,李仁杰,等. DBN参数对双转子不对中故障特征提取的影响及综合评估优选研究[J]. 振动与冲击,2021,40(12): 151-158.YANG Dalian,ZHANG Fanyu,LI Renjie,et al. A study on the influence of DBN’s parameters on dual-rotor misalignment fault feature extraction and its optimization based on comprehensive evaluation[J]. Journal of Vibration and Shock,2021,40(12): 151-158. (in Chinese) [14] YU Xiao, REN Xiaohong, WAN Hong, et al. Rolling bearing fault feature extraction and diagnosis method based on MODWPT and DBN[C]//2019 11th International Conference on Wireless Communications and Signal Processing. Piscataway, US: IEEE, 2019: 1-7. [15] XU Fan,FANG Yan,WANG Dong,et al. Combining DBN and FCM for fault diagnosis of roller element bearings without using data labels[J]. Shock and Vibration,2018,2018: 1-12. [16] 任朝晖,于天壮,丁东,等. 基于VMD-DBN的滚动轴承故障诊断方法[J]. 东北大学学报(自然科学版),2021,42(8): 1105-1110.REN Zhaohui,YU Tianzhuang,DING Dong,et al. Fault diagnosis method of rolling bearing based on VMD-DBN[J]. Journal of Northeastern University (Natural Science),2021,42(8): 1105-1110. (in Chinese) [17] 罗廷金. 基于流形学习的数据降维算法研究[D]. 长沙: 国防科学技术大学, 2013.LUO Tingjin. Research on dimension reduction algorithms based on manifold learning[D]. Changsha: National University of Defense Technology, 2013. (in Chinese) [18] 姜战伟,郑近德,潘海洋,等. 基于多尺度时不可逆与t-SNE流形学习的滚动轴承故障诊断[J]. 振动与冲击,2017,36(17): 61-68, 84.JIANG Zhanwei,ZHENG Jinde,PAN Haiyang,et al. Rolling bearing fault diagnosis method based on multiscale time irreversibility and t-SNE manifold learning[J]. Journal of Vibration and Shock,2017,36(17): 61-68, 84. (in Chinese) [19] 施莹,庄哲,林建辉. 基于卷积稀疏表示及等距映射的轴承故障诊断[J]. 振动、测试与诊断,2019,39(5): 1081-1088, 1138.SHI Ying,ZHUANG Zhe,LIN Jianhui. Fault diagnosis of bearing based on CSR-ISOMAP[J]. Journal of Vibration, Measurement & Diagnosis,2019,39(5): 1081-1088, 1138. (in Chinese) [20] 李建斌,武颖莹,李鹏宇,等. 基于局部线性嵌入和支持向量机回归的TBM施工参数预测[J]. 浙江大学学报(工学版),2021,55(8): 1426-1435.LI Jianbin,WU Yingying,LI Pengyu,et al. TBM tunneling parameters prediction based on locally linear embedding and support vector regression[J]. Journal of Zhejiang University (Engineering Science),2021,55(8): 1426-1435. (in Chinese) [21] 王广斌,杜谋军,韩清凯,等. 基于多尺度子带样本熵和LPP的轴承故障诊断方法[J]. 振动与冲击,2016,35(20): 71-76, 97.WANG Guangbin,DU Moujun,HAN Qingkai,et al. A bearing fault diagnosis method based on multi-scale sub-band sample entropy and LPP[J]. Journal of Vibration and Shock,2016,35(20): 71-76, 97. (in Chinese) [22] 常春,梅检民,赵慧敏,等. 基于局部切空间排列和最小二乘支持向量机的气缸压力识别[J]. 振动与冲击,2020,39(13): 16-21, 63.CHANG Chun,MEI Jianmin,ZHAO Huimin,et al. Recognition of cylinder pressure based on LTSA-LSSVM[J]. Journal of Vibration and Shock,2020,39(13): 16-21, 63. (in Chinese) [23] 张安安,杨林,何嘉辉,等. 基于MDS的电缆附件局部放电模式识别[J]. 电子科技大学学报,2019,48(2): 202-207.ZHANG An’an,YANG Lin,HE Jiahui,et al. Pattern recognition for partial discharge of cable accessories based on multidim ensional scaling[J]. Journal of University of Electronic Science and Technology of China,2019,48(2): 202-207. (in Chinese) [24] VLACHOS M, DOMENICONI C, GUNOPULOS D, et al. Non-linear dimensionality reduction techniques for classification and visualization[C]//Proceedings of the eighth ACM SIGKDD international conference on Knowledge discovery and data mining. New York: ACM, 2002: 645-651. [25] LELAND M,HEALY J,SAUL N,et al. UMAP: uniform manifold approximation and projection[J]. Journal of Open Source Software,2018,3(29): 861. doi: 10.21105/joss.00861 [26] HUANG Guangbin. An insight into extreme learning machines: random neurons, random features and kernels[J]. Cognitive Computation,2014,6(3): 376-390. doi: 10.1007/s12559-014-9255-2 [27] 吕启,窦勇,牛新,等. 基于DBN模型的遥感图像分类[J]. 计算机研究与发展,2014,51(9): 1911-1918.LÜ Qi,DOU Yong,NIU Xin,et al. Remote sensing image classification based on DBN model[J]. Journal of Computer Research and Development,2014,51(9): 1911-1918. (in Chinese)