Fault diagnosis of rolling bearing based on visual spectrum signal feature extraction
-
摘要: 针对滚动轴承传统的故障特征提取方法易受外界的噪声干扰且包含大量的冗余信息的问题,提出一种可视图图谱信号特征提取的故障诊断方法。首先将振动信号转换成可视图图谱信号,计算每个可视图图谱信号的邻接矩阵与拉普拉斯矩阵,得到各种图谱指标,并采用双样本Z值方法筛选出合适故障特征作为故障特征向量,最后通过支持向量机分类算法得到轴承故障诊断分类结果。通过试验分析表明,与传统的故障特征提取方法比较,针对滚动轴承不同类型的故障诊断,采用基于可视图图谱信号的故障特征提取方法准确率提高了8.34%;为进一步证明该方法,针对滚动轴承外圈不同程度的故障诊断,该方法准确率提高了16.67%,更加表明基于可视图谱信号特征提取方法的优越性。Abstract: Aiming at the problem that the traditional fault feature extraction method of rolling bearing is easy to be disturbed by external noises and contains a lot of redundant information, a fault diagnosis method of feature extraction of visible spectrum signal was proposed. Firstly, the vibration signal is converted into viewable map signal, the adjacency matrix and Laplace matrix of each visual map signal are calculated, and various map indexes are obtained. The appropriate fault features are selected as the fault feature vector by using the double sample Z-value method. Finally, support vector machine (SVM) classification algorithm for bearing fault diagnosis classification results. The experimental analysis shows that, compared with the traditional fault feature extraction method, for different types of fault diagnosis of rolling bearing, the accuracy of fault feature extraction method based on visual map signal is improved by 8.34%; In order to further prove this method, for different degrees of fault diagnosis of the outer ring of rolling bearing, the accuracy of this method is improved by 16.67%, which shows the superiority of the signal feature extraction method based on visual spectrum .
-
Key words:
- feature extraction /
- rolling bearing /
- fault diagnosis /
- support vector machine /
- visible graph signals
-
表 1 可视图图谱指标
指标名称 指标表达式 图能量指标 $ {{{S}}_1} = \displaystyle\sum\limits_{i = 1}^n {|{\lambda _{{i}}}|} $ Estrada指标 ${ {{S} }_2} = \displaystyle\sum\limits_{i = 1}^n { { {{e} }^{ {\lambda _{{i} } } } } }$ Quasi-Wiener指标 ${ {{S} }_3} = {{N} } \cdot \displaystyle \sum\limits_{i = 1}^{ {{N} } - 1} {\dfrac{1}{ { {\mu _{{i} } } } } }$ Second Mohar指标 ${ {{S} }_4} = \dfrac{4}{ {N \cdot {\lambda _{N - 1} } } }$ 图拉普拉斯能量指标 ${ {{S} }_5} = \displaystyle \sum\limits_{i = 1}^n {\left|{\mu _{{i} } } - \frac{ {2{{m} } } }{n}\right|}$ 类拉普拉斯能量不变量指标 ${ {{S} }_6} = \displaystyle\sum\limits_{i = 1}^n {|\sqrt { {\mu _i} } |}$ L-Estrada指标 ${ {{S} }_7} = \displaystyle \sum\limits_{i = 1}^n { { {{e} }^{ {\mu _{{i} } } } } }$ 特征值的绝对偏差指标 ${ {{S} }_8} = \displaystyle\sum\limits_{i = 1}^N {\left| { {\lambda _i} - \overline \lambda } \right|}$ 特征值的平均绝对偏差指标 ${ {{S} }_9} = \dfrac{ { \displaystyle\sum\limits_{i = 1}^N {\left| { {\lambda _i} - \overline \lambda } \right|} } }{N}$ 表 2 基于不同方法特征提取结果
故障类型 识别错误数/正确率 图谱指标 时域指标 正常轴承 1/93.3% 5/66.7% 滚动体故障 1/93.3% 2/86.7% 外圈故障 0/100% 0/100% 内圈故障 0/100% 0/100% 表 3 基于不同方法特征提取结果
故障类型 识别错误数/准确率 图谱指标 时域指标 正常轴承 3/85.0% 9/40.0% 0.177 8 mm故障 0/100% 4/73.3% 0.355 6 mm故障 0/100% 0/100% 0.533 4 mm故障 0/100% 0/100% -
[1] 侯泽林. 旋转机械故障诊断的研究现状及发展前景[J]. 锻压装备与制造技术, 2021, 56(5): 33-37. [2] 宋宏智, 李力, 杨兴宽, 等. 高速机车轴承故障诊断与剩余寿命预测的发展及展望[J]. 轴承, 2020(3): 61-67. [3] Wang W G, Yuan F C, Liu Z S. Sparsity discriminant preserving projection for machinery fault diagnosis[J]. Measurement, 2020, 173: 108488. [4] 焦卫东, 丁祥满, 严天宇, 等. 基于频域特征波形模式匹配的故障诊断方法研究[J]. 华东交通大学学报, 2021(2): 73-81. [5] Mo Z L, Wang J Y, Zhang H, et al. Weighted cyclic harmonic-to-noise ratio for rolling element bearing fault diagnosis[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(2): 432-442. doi: 10.1109/TIM.2019.2903615 [6] 李恒, 张氢, 秦仙蓉, 等. 基于短时傅里叶变换和卷积神经网络的轴承故障诊断方法[J]. 振动与冲击, 2018, 37(19): 124-131. doi: 10.13465/j.cnki.jvs.2018.19.020 [7] 杨汉键. 基于图信号处理的滚动轴承故障特征提取方法研究[D]. 长沙: 湖南大学, 2016. [8] Khelaifi F, He H J. Perceptual image hashing based on structural fractal features of image coding and ring partition[J]. Multimedia Tools and Applications, 2020, 79: 19025-19044. [9] 王保云, 李沛. 分析大数据: 非规则结构与图信号[J]. 南京邮电大学学报:自然科学版, 2020, 40(5): 112-116. [10] Zhang R, Ashuri B, Deng Y. A novel method for forecasting time series based on fuzzy logic and visibility graph[J]. Advances in Data Analysis and Classification, 2017, 11(4): 759-783. doi: 10.1007/s11634-017-0300-3 [11] Xu P H, Zhang R, Deng R. A novel visibility graph transformation of time series into weighted networks[J]. Chaos, Solitons and Fractals:the Interdisciplinary Journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena, 2018, 117: 201-208. [12] Kaniska S, Soumya C, Rohit B. Cross-subject motor imagery tasks eeg signal classification employing multiplex weighted visibility graph and deep feature extraction[J]. IEEE Sensors Letters, 2020, 4(1): 1-4. [13] Supriya S, Siuly S, Wang H, et al. Weighted visibility graph with complex network features in the detection of epilepsy[J]. IEEE Access, 2016, 4: 6554-6566. doi: 10.1109/ACCESS.2016.2612242 [14] 高艺源, 于德介, 王好将, 等. 基于图谱指标的滚动轴承故障特征提取方法[J]. 航空动力学报, 2018, 33(8): 2033-2040. doi: 10.13224/j.cnki.jasp.2018.08.027 [15] 陈芒, 于德介, 高艺源. 基于可视图图谱幅值熵的滚动轴承故障诊断方法[J]. 振动与冲击, 2021, 40(4): 23-29. [16] 丁闯, 冯辅周, 张兵志, 等. 改进多尺度符号动力学信息熵及其在行星变速箱特征提取中的应用[J]. 振动与冲击, 2020, 39(13): 97-102,147. [17] 张淑清, 李威, 张立国, 等. 基于多元经验模态分解互近似熵及GG聚类的轴承故障诊断[J]. 中国机械工程, 2016, 27(24): 3362-3367. doi: 10.3969/j.issn.1004-132X.2016.24.017