采用规范化局部保持投影的轴承故障诊断方法
|
摘要
针对轴承故障检测中特征融合导致的维度高、相关性强等问题,提出一种采用规范化局部保持投影算法(LPP)的轴承故障诊断(En-LPP)方法。首先,采用熵规范化的方法将相似度矩阵结合到传统LPP算法的优化函数中,与投影向量一并求解,得到一种规范化LPP降维算法;然后对原始轴承振动信号进行小波变换和经验模式分解得到10条信号分量,每个分量通过计算平均值、均方根等,提取12维统计特征,经归一化后生成特征向量;然后将特征向量输入到规范化LPP降维算法中进行迭代共同求解,得到满足终止条件的相似度矩阵和投影向量;最后利用降维后的特征集训练极限学习机模型确定轴承最终工作状态以实现故障检测。实验结果表明:与传统LPP方法以及其他降维方法相比,所提出的En-LPP方法对于轴承故障诊断的性能更好;在小波变换72维特征集合以及经验模式分解48维特征集合下的分类精度平均提升了7%以上;在4种不同分类器组合下的分类精度平均提升了17%以上;较好的降维特征区分能力使得En-LPP方法的故障诊断性能在不同条件组合下均具有很好的鲁棒性。
A novel diagnostic approach for bearing faults using regularized locally maintaining projections(En-LPP) is proposed to solve the problems of high dimension and strong relevance caused by feature fusion in bearing fault detection domains. Firstly, the entropy normalization method is used to combine a similarity matrix into the optimization function of the traditional locality preserving projections(LPP) algorithm, and to solve it along with the projection vector to obtain a regularized LPP algorithm with reduced dimension. Then, the original bearing vibration signal is decomposed into ten signal components through wavelet transform and empirical mode decomposition, and 12-dimensional statistical features for each component are extracted. Final high-dimensional feature vectors are generated by combining and normalizing these statistical features of all components. These feature vectors are then inputted into the proposed algorithm, and iteration is performed to obtain the similarity matrix and the projection vector that satisfy termination conditions. Finally, the reduced dimension feature set is used to train an extreme learning machine model for determining the final working state of a bearing to achieve fault detection. Experimental results and comparisons with the traditional LPP algorithm and other dimension reduction algorithms show that the proposed En-LPP algorithm has better performance for bearing fault diagnosis. The classification accuracy is improved by more than 7% on average under the 72-dimensional feature set of the wavelet transform and 48-dimensional feature set of empirical mode decomposition, and the averaged classification accuracy of the regularized LPP algorithm with combinations of 4 classifiers is about 17% higher than those of other dimension reduction algorithms. The proposed approach has better ability to distinguish dimension-reduced features, which makes its fault diagnosis performance robust under different combinations of conditions.
A novel diagnostic approach for bearing faults using regularized locally maintaining projections(En-LPP) is proposed to solve the problems of high dimension and strong relevance caused by feature fusion in bearing fault detection domains. Firstly, the entropy normalization method is used to combine a similarity matrix into the optimization function of the traditional locality preserving projections(LPP) algorithm, and to solve it along with the projection vector to obtain a regularized LPP algorithm with reduced dimension. Then, the original bearing vibration signal is decomposed into ten signal components through wavelet transform and empirical mode decomposition, and 12-dimensional statistical features for each component are extracted. Final high-dimensional feature vectors are generated by combining and normalizing these statistical features of all components. These feature vectors are then inputted into the proposed algorithm, and iteration is performed to obtain the similarity matrix and the projection vector that satisfy termination conditions. Finally, the reduced dimension feature set is used to train an extreme learning machine model for determining the final working state of a bearing to achieve fault detection. Experimental results and comparisons with the traditional LPP algorithm and other dimension reduction algorithms show that the proposed En-LPP algorithm has better performance for bearing fault diagnosis. The classification accuracy is improved by more than 7% on average under the 72-dimensional feature set of the wavelet transform and 48-dimensional feature set of empirical mode decomposition, and the averaged classification accuracy of the regularized LPP algorithm with combinations of 4 classifiers is about 17% higher than those of other dimension reduction algorithms. The proposed approach has better ability to distinguish dimension-reduced features, which makes its fault diagnosis performance robust under different combinations of conditions.
引文
[1]剡昌锋,朱涛.基于马田系统的滚动轴承初始故障检测和状态监测[J].振动与冲击,2017,36(12):155-188.YAN Changfeng,ZHU Tao.Incipient fault detection and condition monitoring of rolling bearings by using the Mahalanobis-Taguchi system[J].Journal of Vibration and Shock,2017,36(12):155-188.
[2]沈微,陶新民.基于同步挤压小波变换的振动信号自适应降噪方法[J].振动与冲击,2018,37(14):239-247.SHEN Wei,TAO Xinmin.Self-adaptive de-noising algorithm for vibration signals based on synchrosqueezed wavelet transforms[J].Journal of Vibration and Shock,2018,37(14):239-247.
[3]陈娜,杨绍普.基于匹配追踪预处理的铁道车辆滚动轴承故障诊断方法[J].振动与冲击,2017,36(21):132-137.CHEN Na,YANG Shaopu.Railway vehicle rolling bearing fault diagnosis method based on the matching pursuit pretreatment[J].Journal of Vibration and Shock,2017,36(21):132-137.
[4]王林,蔡改改.基于改进MP的稀疏表示快速算法及其滚动轴承故障特征提取应用[J].振动与冲击,2017,36(3):176-182.WANG Lin,CAI Gaigai.Fast sparse representation algorithm based on improved MP and its applications in fault feature extraction of rolling bearings[J].Journal of Vibration and Shock,2017,36(3):176-182.
[5]武英杰,辛红伟.基于VMD滤波和极值点包络阶次的滚动轴承故障诊断[J].振动与冲击,2018,37(14):102-107.WU Yingjie,XIN Hongwei.Rolling bearing fault diagnosis based on the variational mode decomposition filtering and extreme point envelope order[J].Journal of Vibration and Shock,2018,37(14):102-107.
[6]SWAGATAM D,SHOUNAK D.Handling data irregularities in classification:foundations,trends,and future challenges[J].Pattern Recognition,2018,81:674-697.
[7]SHAO H D,JIANG H K.Rolling bearing fault feature learning using improved convolutional deep belief network with compressed sensing[J].Mechanical Systems and Signal Processing,2018,100:743-765.
[8]SHAO H D,JIANG H K.A novel method for intelligent fault diagnosis of rolling bearings using ensemble deep auto-encoders[J].Mechanical Systems and Signal Processing,2018,102:278-297.
[9]WANG S B,IVAN S.Nonconvex sparse regularization and convex optimization for bearing fault diagnosis[J].IEEE Transactions on Industrial Electronics,2018,65(9):7332-7343.
[10]柯燕亮,王华庆.基于Teager峰值能量的低转速轴承故障特征提取方法[J].振动与冲击,2017,36(11):124-133.KE Yanliang,WANG Huaqing.Fault feature extraction of low speed roller bearings based on Teager peak energy[J].Journal of Vibration and Shock,2017,36(11):124-133.
[11]唐贵基,庞彬.基于时时能量阶比谱的变转速工况滚动轴承微弱故障诊断研究[J].振动工程学报,2017,30(5):856-864.TANG Guiji,PANG Bin.Research on fault diagnosis of rolling bearing’s weak fault under variable speed conditions based on time-time energy order spectrum[J].Journal of Vibration Engineering,2017,30(5):856-864.
[12]孔佑炳,郭瑜.基于EEMD的混合陶瓷球轴承故障双冲击特征提取[J].振动与冲击,2016,35(1):17-22.KONG Ruobing,GUO Yu.Double-impulse feature extraction of faulty hybrid ceramic ball bearings based on EEMD[J].Journal of Vibration and Shock,2016,35(1):17-22.
[13]谷朝健,马增强.多通道相关-自适应轴承故障诊断方法[J].航空动力学报,2018,33(7):1750-1756.GU Chaojian,MA Zengqiang.Adaptive bearing fault diagnosis method of multi-channel correlation[J].Journal of Aerospace Power,2018,33(7):1750-1756.
[14]赵孝礼,赵荣珍.全局与局部判别信息融合的转子故障数据集降维方法研究[J].自动化学报,2017,43(4):560-567.ZHAO Xiaoli,ZHAO Rongzhen.A method of dimension reduction of rotor faults data set based on fusion of global and local discriminant information[J].Acta Automatica Sinica,2017,43(4):560-567.
[15]吴广宁,袁海满,高波,等.基于特征评估与核主元分析的电力变压器故障诊断[J].高电压技术,2017,43(8):2533-2540.WU Guangning,YUAN Haiman,GAO Bo,et al.Fault diagnosis of power transformer based on feature evaluation and kernel principal[J].High Voltage Engineering,2017,43(8):2533-2540.
[16]张绍辉,李巍华.基于特征空间降噪的局部保持投影算法及其在轴承故障分类中的应用[J].机械工程学报,2014,50(3):92-99.ZHANG Shaohui,LI Weihua.Locality preserving projections based on feature space denoising and its application in bearing fault classification[J].Journal of Mechanical Engineering,2014,50(3):92-99.
[17]王广斌,杜谋军.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.WANG Guangbin,DU Moujun.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.
[18]丁晓喜,何清波.基于WPD和LPP的设备故障诊断方法研究[J].振动与冲击,2014,33(3):89-93.DING Xiaoxi,HE Qingbo.Machine fault diagnosis based on WPD and LPP[J].Journal of Vibration and Shock,2014,33(3):89-93.
[19]杨望灿,张培林.基于邻域自适应局部保持投影的轴承故障诊断模型[J].振动与冲击,2014,33(1):39-44.YANG Wangchan,ZHANG Peilin.Bearing fault diagnosis based on neighborhood adaptive locality preserving projections[J].Journal of Journal of Vibration and Shock,2014,33(1):39-44.
[20]HE Xiaofei,YAN Shuicheng,HU Yuxiao,et al.Face recognition using Laplacian faces[J].IEEE Trans on Pattern Analysis and Machine Intelligence,2005,27(3):328-340.
[21]FADI D,AMMAR A.Enhanced and parameterless locality preserving projections for face recognition[J].Neurocomputing,2013,99(1):448-457.
[22]ZHENG Heng.A new construction method of neighbor graph for locality preserving projections[J].Journal of Information and Computational Science,2013,10(5):1357-1365.
[23]Case Western Reserve University.Bearing data center website:bearing data center seeded fault test data[DB/OL].(2016-04-06)[2018-07-15].http:∥csegroups.case.edu/bearingdatacenter/pages/downloaddata-file.
[24]GYEONGYONG H,PAUL G.Robust kernel discriminant analysis using fuzzy memberships[J].Pattern Recognition,2011,44:716-723.
[2]沈微,陶新民.基于同步挤压小波变换的振动信号自适应降噪方法[J].振动与冲击,2018,37(14):239-247.SHEN Wei,TAO Xinmin.Self-adaptive de-noising algorithm for vibration signals based on synchrosqueezed wavelet transforms[J].Journal of Vibration and Shock,2018,37(14):239-247.
[3]陈娜,杨绍普.基于匹配追踪预处理的铁道车辆滚动轴承故障诊断方法[J].振动与冲击,2017,36(21):132-137.CHEN Na,YANG Shaopu.Railway vehicle rolling bearing fault diagnosis method based on the matching pursuit pretreatment[J].Journal of Vibration and Shock,2017,36(21):132-137.
[4]王林,蔡改改.基于改进MP的稀疏表示快速算法及其滚动轴承故障特征提取应用[J].振动与冲击,2017,36(3):176-182.WANG Lin,CAI Gaigai.Fast sparse representation algorithm based on improved MP and its applications in fault feature extraction of rolling bearings[J].Journal of Vibration and Shock,2017,36(3):176-182.
[5]武英杰,辛红伟.基于VMD滤波和极值点包络阶次的滚动轴承故障诊断[J].振动与冲击,2018,37(14):102-107.WU Yingjie,XIN Hongwei.Rolling bearing fault diagnosis based on the variational mode decomposition filtering and extreme point envelope order[J].Journal of Vibration and Shock,2018,37(14):102-107.
[6]SWAGATAM D,SHOUNAK D.Handling data irregularities in classification:foundations,trends,and future challenges[J].Pattern Recognition,2018,81:674-697.
[7]SHAO H D,JIANG H K.Rolling bearing fault feature learning using improved convolutional deep belief network with compressed sensing[J].Mechanical Systems and Signal Processing,2018,100:743-765.
[8]SHAO H D,JIANG H K.A novel method for intelligent fault diagnosis of rolling bearings using ensemble deep auto-encoders[J].Mechanical Systems and Signal Processing,2018,102:278-297.
[9]WANG S B,IVAN S.Nonconvex sparse regularization and convex optimization for bearing fault diagnosis[J].IEEE Transactions on Industrial Electronics,2018,65(9):7332-7343.
[10]柯燕亮,王华庆.基于Teager峰值能量的低转速轴承故障特征提取方法[J].振动与冲击,2017,36(11):124-133.KE Yanliang,WANG Huaqing.Fault feature extraction of low speed roller bearings based on Teager peak energy[J].Journal of Vibration and Shock,2017,36(11):124-133.
[11]唐贵基,庞彬.基于时时能量阶比谱的变转速工况滚动轴承微弱故障诊断研究[J].振动工程学报,2017,30(5):856-864.TANG Guiji,PANG Bin.Research on fault diagnosis of rolling bearing’s weak fault under variable speed conditions based on time-time energy order spectrum[J].Journal of Vibration Engineering,2017,30(5):856-864.
[12]孔佑炳,郭瑜.基于EEMD的混合陶瓷球轴承故障双冲击特征提取[J].振动与冲击,2016,35(1):17-22.KONG Ruobing,GUO Yu.Double-impulse feature extraction of faulty hybrid ceramic ball bearings based on EEMD[J].Journal of Vibration and Shock,2016,35(1):17-22.
[13]谷朝健,马增强.多通道相关-自适应轴承故障诊断方法[J].航空动力学报,2018,33(7):1750-1756.GU Chaojian,MA Zengqiang.Adaptive bearing fault diagnosis method of multi-channel correlation[J].Journal of Aerospace Power,2018,33(7):1750-1756.
[14]赵孝礼,赵荣珍.全局与局部判别信息融合的转子故障数据集降维方法研究[J].自动化学报,2017,43(4):560-567.ZHAO Xiaoli,ZHAO Rongzhen.A method of dimension reduction of rotor faults data set based on fusion of global and local discriminant information[J].Acta Automatica Sinica,2017,43(4):560-567.
[15]吴广宁,袁海满,高波,等.基于特征评估与核主元分析的电力变压器故障诊断[J].高电压技术,2017,43(8):2533-2540.WU Guangning,YUAN Haiman,GAO Bo,et al.Fault diagnosis of power transformer based on feature evaluation and kernel principal[J].High Voltage Engineering,2017,43(8):2533-2540.
[16]张绍辉,李巍华.基于特征空间降噪的局部保持投影算法及其在轴承故障分类中的应用[J].机械工程学报,2014,50(3):92-99.ZHANG Shaohui,LI Weihua.Locality preserving projections based on feature space denoising and its application in bearing fault classification[J].Journal of Mechanical Engineering,2014,50(3):92-99.
[17]王广斌,杜谋军.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.WANG Guangbin,DU Moujun.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.
[18]丁晓喜,何清波.基于WPD和LPP的设备故障诊断方法研究[J].振动与冲击,2014,33(3):89-93.DING Xiaoxi,HE Qingbo.Machine fault diagnosis based on WPD and LPP[J].Journal of Vibration and Shock,2014,33(3):89-93.
[19]杨望灿,张培林.基于邻域自适应局部保持投影的轴承故障诊断模型[J].振动与冲击,2014,33(1):39-44.YANG Wangchan,ZHANG Peilin.Bearing fault diagnosis based on neighborhood adaptive locality preserving projections[J].Journal of Journal of Vibration and Shock,2014,33(1):39-44.
[20]HE Xiaofei,YAN Shuicheng,HU Yuxiao,et al.Face recognition using Laplacian faces[J].IEEE Trans on Pattern Analysis and Machine Intelligence,2005,27(3):328-340.
[21]FADI D,AMMAR A.Enhanced and parameterless locality preserving projections for face recognition[J].Neurocomputing,2013,99(1):448-457.
[22]ZHENG Heng.A new construction method of neighbor graph for locality preserving projections[J].Journal of Information and Computational Science,2013,10(5):1357-1365.
[23]Case Western Reserve University.Bearing data center website:bearing data center seeded fault test data[DB/OL].(2016-04-06)[2018-07-15].http:∥csegroups.case.edu/bearingdatacenter/pages/downloaddata-file.
[24]GYEONGYONG H,PAUL G.Robust kernel discriminant analysis using fuzzy memberships[J].Pattern Recognition,2011,44:716-723.