Application Research of Bearing Condition Identification using EEWD and SODN
-
摘要: 传统滚动轴承工况识别方法存在轴承振动信号人工特征提取困难的问题, 提出一种基于增强经验小波分解(Enhanced empirical wavelet decomposition, EEWD)和自组织深层网络(Self-organizing deep network, SODN)的工况识别方法。首先改进经验小波分解的频谱分割方式, 将滚动轴承振动信号自适应分解为若干本征模态分量; 然后利用综合评价指标筛选出最能反映信号工况特征的本征模态分量并重构信号; 最后构造自组织深层网络, 将重构后的滚动轴承振动信号输入SODN进行自动特征学习与工况识别。实验结果表明: EEWD结合SODN方法相比于其它深度学习方法在信号特征提取和工况识别准确率方面更具优势。Abstract: Considering that traditional methods for rolling bearing condition identification were difficulty in manual feature extraction of bearing vibration signals, a new method based on enhanced empirical wavelet decomposition (EEWD) with self-organizing deep network (SODN) was proposed. Firstly, the segmentation method of spectrum of empirical wavelet decomposition was enhanced, and the vibration signals of rolling bearings were adaptively decomposed into several intrinsic modal functions. The intrinsic modal functions which can best reflect the condition characteristics of the raw signals were selected by the comprehensive evaluation index and then reconstructed. Secondly, the self-organizing deep network was constructed. Finally, the reconstructed signals were fed into SODN for automatic feature learning and automatic condition identification. The experimental results indicate that the method based EEWD and SODN is superior than other deep learning methods in signals feature extraction and condition recognition accuracy.
-
表 1 7种滚动轴承工况
故障状态 代号 编码 转频/Hz 样本数量 正常 a 1 000 000 35.0 12 000 内圈轻微 b 0 100 000 37.5 12 000 内圈中度 c 0 010 000 40.0 12 000 外圈轻微 d 0 001 000 35.0 12 000 外圈中度 e 0 000 100 37.5 12 000 滚动体轻微 f 0 000 010 40.0 12 000 滚动体中度 g 0 000 001 37.5 12 000 表 2 不同方法的识别结果
方法 识别正确率/%±标准差 训练时间/s SODN 98.93 ± 0.11 112. 98 DAE 90.13±1.08 191.19 DBN 91.00±1.00 150.64 DDAE 91.97±1.32 110.47 DSAE 93.29 ± 0.41 142.23 DCAE 93.97 ± 0.52 175.42 DWAE 95.62 ± 0.37 159.17 表 3 第4组不同方法的精确率和召回率
工况 SODN DWAE DCAE P R P R P R a 95.19 93.43 90.91 97.12 84.51 77.82 b 95.37 92.24 91.09 80.35 85.34 80.79 c 96.12 95.37 90.97 98.68 84.99 78.58 d 96.01 94.15 90.91 80.89 83.51 80.89 e 95.97 94.09 88.13 86.17 88.98 99.12 f 94.19 93.49 91.13 80.67 82.43 70.87 g 94.63 91.26 89.13 96.24 82.43 76.24 表 4 第4组不同方法的F1值
工况 F1 SODN DWAE DCAE a 95.19 90.91 84.51 b 95.37 91.09 85.34 c 96.12 90.97 84.99 d 96.01 90.91 83.51 e 95.97 88.13 88.98 f 94.19 91.13 82.43 g 94.63 89.13 82.43 表 5 不同激活函数对识别准确率的影响
激活函数 识别率/% ReLU 95.14 LReLU 95.82 ELU 94.64 Gaussian wavelet 99. 08 Swish 95.17 Morlet wavelet 98.12 Mexican hat wavelet 98.09 -
[1] 孙国栋, 王俊豪, 徐昀, 等. CEEMD-WVD多尺度时频图像的滚动轴承故障诊断[J]. 机械科学与技术, 2020, 39(5): 688-694 https://www.cnki.com.cn/Article/CJFDTOTAL-JXKX202005007.htmSUN G D, WANG J H, XU Y, et al. Rolling bearing fault diagnosis based on CEEMD-WVD multi-scale time-frequency image[J]. Mechanical Science and Technology for Aerospace Engineering, 2020, 39(5): 688-694 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JXKX202005007.htm [2] SCHMIDHUBER J. Deep learning in neural networks: an overview[J]. Neural Networks, 2015, 61: 85-117 [3] SHAO H D, JIANG H K, LI X Q, et al. Intelligent fault diagnosis of rolling bearing using deep wavelet auto-encoder with extreme learning machine[J]. Knowledge-Based Systems, 2018, 140: 1-14 doi: 10.1016/j.knosys.2017.10.024 [4] SHAO H D, JIANG H K, ZHANG X, et al. Rolling bearing fault diagnosis using an optimization deep belief network[J]. Measurement Science and Technology, 2015, 26(11): 115002 doi: 10.1088/0957-0233/26/11/115002 [5] SHAO H D, JIANG H K, WANG F A, et al. An enhancement deep feature fusion method for rotating machinery fault diagnosis[J]. Knowledge-Based Systems, 2017, 119: 200-220 doi: 10.1016/j.knosys.2016.12.012 [6] SHAO H D, JIANG H K, ZHANG H Z, et al. Rolling bearing fault feature learning using improved convolutional deep belief network with compressed sensing[J]. Mechanical Systems and Signal Processing, 2018, 100: 743-765 doi: 10.1016/j.ymssp.2017.08.002 [7] 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 [8] 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 [9] YEH J R, SHIEH J S, HUANG N E. Complementary ensemble empirical mode decomposition: a novel noise enhanced data analysis method[J]. Advances in Adaptive Data Analysis, 2010, 2(2): 135-156 doi: 10.1142/S1793536910000422 [10] DRAGOMIRETSKIY K, ZOSSO D. Variational mode decomposition[J]. IEEE Transactions on Signal Processing, 2014, 62(3): 531-544. doi: 10.1109/TSP.2013.2288675 [11] GILLES J. Empirical wavelet transform[J]. IEEE Transac-tions on Signal Processing, 2013, 61(16): 3999-4010 doi: 10.1109/TSP.2013.2265222 [12] 朱艳萍, 包文杰, 涂晓彤, 等. 改进的经验小波变换在滚动轴承故障诊断中的应用[J]. 噪声与振动控制, 2018, 38(1): 199-203 doi: 10.3969/j.issn.1006-1355.2018.01.039ZHU Y P, BAO W J, TU X T, et al. Application of enhanced empirical wavelet transform to rolling bearings fault diagnosis[J]. Noise and Vibration Control, 2018, 38(1): 199-203 (in Chinese) doi: 10.3969/j.issn.1006-1355.2018.01.039 [13] 雷亚国, 韩天宇, 王彪, 等. XJTU-SY滚动轴承加速寿命试验数据集解读[J]. 机械工程学报, 2019, 55(16): 1-6 https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201916001.htmLEI Y G, HAN T Y, WANG B, et al. XJTU-SY rolling element bearing accelerated life test datasets: a tutorial[J]. Journal of Mechanical Engineering, 2019, 55(16): 1-6 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201916001.htm [14] WU J, TANG T, CHEN M, et al. A study on adaptation lightweight architecture based deep learning models for bearing fault diagnosis under varying working conditions[J]. Expert Systems with Applications, 2020, 160: 113710 doi: 10.1016/j.eswa.2020.113710 [15] DU G Y, XU Q, YANG X Y. Fault diagnosis of rotating machinery components using a deep kernel extreme learning machine under different working conditions[J]. Measurement Science and Technology, 2020, 31(11): 115901 doi: 10.1088/1361-6501/ab9037