Feature Extraction and Recognition for Rolling Element Bearing Fault Utilizing Short-Time Fourier Transform and Non-negative Matrix Factorization

Due to the non-stationary characteristics of vibration signals acquired from rolling element bearing fault, thc time-frequency analysis is often applied to describe the local information of these unstable signals smartly. However, it is difficult to classitythe high dimensional feature matrix directly because of too large dimensions for many classifiers. This paper combines the concepts of time-frequency distribution（TFD） with non-negative matrix factorization（NMF）, and proposes a novel TFD matrix factorization method to enhance representation and identification of bearing fault. Throughout this method, the TFD of a vibration signal is firstly accomplished to describe the localized faults with short-time Fourier transform（STFT）. Then, the supervised NMF mapping is adopted to extract the fault features from TFD. Meanwhile, the fault samples can be clustered and recognized automatically by using the clustering property of NMF. The proposed method takes advantages of the NMF in the parts-based representation and the adaptive clustering. The localized fault features of interest can be extracted as well. To evaluate the performance of the proposed method, the 9 kinds of the bearing fault on a test bench is performed. The proposed method can effectively identify the fault severity and different fault types. Moreover, in comparison with the artificial neural network（ANN）, NMF yields 99.3% mean accuracy which is much superior to ANN. This research presents a simple and practical resolution for the fault diagnosis problem of rolling element bearing in high dimensional feature space.

Multi-view feature fusion for rolling bearing fault diagnosis using random forest and autoencoder

To improve the accuracy and robustness of rolling bearing fault diagnosis under complex conditions, a novel method based on multi-view feature fusion is proposed. Firstly, multi-view features from perspectives of the time domain, frequency domain and time-frequency domain are extracted through the Fourier transform, Hilbert transform and empirical mode decomposition (EMD).Then, the random forest model (RF) is applied to select features which are highly correlated with the bearing operating state. Subsequently, the selected features are fused via the autoencoder (AE) to further reduce the redundancy. Finally, the effectiveness of the fused features is evaluated by the support vector machine (SVM). The experimental results indicate that the proposed method based on the multi-view feature fusion can effectively reflect the difference in the state of the rolling bearing, and improve the accuracy of fault diagnosis.

Fractional Envelope Analysis for Rolling Element Bearing Weak Fault Feature Extraction

The bearing weak fault feature extraction is crucial to mechanical fault diagnosis and machine condition monitoring. Envelope analysis based on Hilbert transform has been widely used in bearing fault feature extraction. A generalization of the Hilbert transform, the fractional Hilbert transform is defined in the frequency domain, it is based upon the modification of spatial filter with a fractional parameter, and it can be used to construct a new kind of fractional analytic signal. By performing spectrum analysis on the fractional envelope signal, the fractional envelope spectrum can be obtained. When weak faults occur in a bearing, some of the characteristic frequencies will clearly appear in the fractional envelope spectrum. These characteristic frequencies can be used for bearing weak fault feature extraction. The effectiveness of the proposed method is verified through simulation signal and experiment data. © 2017 Chinese Association of Automation.

Predicting Reliability and Remaining Useful Life of Rolling Bearings Based on Optimized Neural Networks

In this study,an optimized long short-term memory(LSTM)network is proposed to predict the reliability and remaining useful life(RUL)of rolling bearings based on an improved whale-optimized algorithm(IWOA).The multi-domain features are extracted to construct the feature dataset because the single-domain features are difficult to characterize the performance degeneration of the rolling bearing.To provide covariates for reliability assessment,a kernel principal component analysis is used to reduce the dimensionality of the features.A Weibull distribution proportional hazard model(WPHM)is used for the reliability assessment of rolling bearing,and a beluga whale optimization(BWO)algorithm is combined with maximum likelihood estimation(MLE)to improve the estimation accuracy of the model parameters of the WPHM,which provides the data basis for predicting reliability.Considering the possible gradient explosion by training the rolling bearing lifetime data and the difficulties in selecting the key network parameters,an optimized LSTM network called the improved whale optimization algorithm-based long short-term memory(IWOA-LSTM)network is proposed.As IWOA better jumps out of the local optimization,the fitting and prediction accuracies of the network are correspondingly improved.The experimental results show that compared with the whale optimization algorithm-based long short-term memory(WOA-LSTM)network,the reliability prediction and RUL prediction accuracies of the rolling bearing are improved by the proposed IWOA-LSTM network.

Deep Residual Joint Transfer Strategy for Cross-Condition Fault Diagnosis of Rolling Bearings

Rolling bearings are key components of the drivetrain in wind turbines,and their health is critical to wind turbine operation.In practical diagnosis tasks,the vibration signal is usually interspersed with many disturbing components,and the variation of operating conditions leads to unbalanced data distribution among different conditions.Although intelligent diagnosis methods based on deep learning have been intensively studied,it is still challenging to diagnose rolling bearing faults with small amounts of samples.To address the above issue,we introduce the deep residual joint transfer strategy method for the cross-condition fault diagnosis of rolling bearings.One-dimensional vibration signals are pre-processed by overlapping feature extraction techniques to fully extract fault characteristics.The deep residual network is trained in training tasks with sufficient samples,for fault pattern classification.Subsequently,three transfer strategies are used to explore the generalizability and adaptability of the pre-trained models to the data distribution in target tasks.Among them,the feature transferability between different tasks is explored by model transfer,and it is validated that minimizing data differences of tasks through a dual-stream adaptation structure helps to enhance generalization of the models to the target tasks.In the experiments of rolling bearing faults with unbalanced data conditions,localized faults of motor bearings and planet bearings are successfully identified,and good fault classification results are achieved,which provide guidance for the cross-condition fault diagnosis of rolling bearings with small amounts of training data.

Application of Xgboost Feature Extraction in Fault Diagnosis of Rolling Bearing

Due to the difficulty that excessive feature dimension in fault diagnosis of rolling bearing will lead to the decrease of classification accuracy,a fault diagnosis method based on Xgboost algorithm feature extraction is proposed.When the Xgboost algorithm classifies features,it generates an order of importance of the input features.The time domain features were extracted from the vibration signal of the rolling bearing,the time-frequency features were formed by the singular value of the modal components that were decomposed by the variational mode decomposition.Firstly,the extracted time domain and time-frequency domain features were input into the support vector machine respectively to observe the fault diagnosis accuracy.Then,Xgboost algorithm was used to rank the importance of features and got the accuracy of fault diagnosis.Finally,important features were extracted and the extracted features were input into the support vector machine to observe the fault diagnosis accuracy.The result shows that the fault diagnosis accuracy of rolling bearing is improved after important feature extraction in time domain and time-frequency domain by Xgboost.

Application of Improved Deep Auto-Encoder Network in Rolling Bearing Fault Diagnosis

Since the effectiveness of extracting fault features is not high under traditional bearing fault diagnosis method, a bearing fault diagnosis method based on Deep Auto-encoder Network (DAEN) optimized by Cloud Adaptive Particle Swarm Optimization (CAPSO) was proposed. On the basis of analyzing CAPSO and DAEN, the CAPSO-DAEN fault diagnosis model is built. The model uses the randomness and stability of CAPSO algorithm to optimize the connection weight of DAEN, to reduce the constraints on the weights and extract fault features adaptively. Finally, efficient and accurate fault diagnosis can be implemented with the Softmax classifier. The results of test show that the proposed method has higher diagnostic accuracy and more stable diagnosis results than those based on the DAEN, Support Vector Machine (SVM) and the Back Propagation algorithm (BP) under appropriate parameters.

SELF-ALIGNING EVEN LOAD MECHANISM OF MULTI-ROW BEARINGS OF LARGE STRIP ROLLING MILL

The load distribution of multi-row bearings of large strip rolling mill is fully analyzed by 3D contact boundary element method （BEM）. It is found out that bearings are frequently worn out due to serious uneven load on the multi-row rollers. The constraint mechanism of the previous rolling system is found to be unreasonable by theoretical analysis on heavy machinery structure. A mechanism of self-aligning even load for workroll bearing of 2 050 mm hot rolling mill of Baoshan I＆S Co. is developed. This device is manufactured with particular regard to the structure of 2 050 mm hot rolling mill mentioned above. Hence, uneven load on multi-row bearings is greatly relieved and their lives are remarkably prolonged. Meanwhile, theoretical analysis and on-spot tests prove the rationality and validity of the device.

Applications of Wigner high-order spectra in feature extraction of acoustic emission signals

The characteristics of typical AE signals initiated by mechanical component damages are analyzed. Based on the extracting principle of acoustic emission(AE) signals from damaged components,the paper introduces Wigner high-order spectra to the field of feature extraction and fault diagnosis of AE signals. Some main performances of Wigner binary spectra,Wigner triple spectra and Wigner-Ville distribution (WVD) are discussed,including of time-frequency resolution,energy accumulation,reduction of crossing items and noise elimination. Wigner triple spectra is employed to the fault diagnosis of rolling bearings with AE techniques. The fault features reading from experimental data analysis are clear,accurate and intuitionistic. The validity and accuracy of Wigner high-order spectra methods proposed agree quite well with simulation results. Simulation and research results indicate that wigner high-order spectra is quite useful for condition monitoring and fault diagnosis in conjunction with AE technique,and has very important research and application values in feature extraction and faults diagnosis based on AE signals due to mechanical component damages.

基于改进麻雀搜索算法优化LSTM的滚动轴承故障诊断

为了对滚动轴承的工作状态及故障类别进行准确的诊断,本文采用长短时记忆(LSTM)神经网络作为分类器对滚动轴承数据集进行分类诊断。首先,从滚动轴承原始运行振动信号中提取时域和频域特征参数,组成具有高维特征参数的数据集;使用核主成分分析(KPCA)方法对高维特征集进行降维处理,选取重要性程度高的特征构成输入特征向量。然后,针对LSTM神经网络在滚动轴承故障诊断中存在的超参数难以确定的问题,提出一种基于自适应t分布策略的麻雀搜索算法优化LSTM神经网络的故障诊断方法(tSSA–LSTM)。最后,使用凯斯西储大学滚动轴承数据中心的数据进行故障诊断精度测试、泛化性能测试及噪声环境下故障诊断性能测试等多个仿真实验,并将本文提出的诊断模型与麻雀搜索算法优化长短时记忆神经网络(SSA–LSTM)、遗传算法优化长短时记忆神经网络(GA–LSTM)、粒子群算法优化长短时记忆神经网络(PSO–LSTM)及传统LSTM诊断模型进行对比。结果表明:tSSA可以更有效地对LSTM的隐含层神经元数量、周期次数、学习率等超参数进行合理优化;所提方法的平均诊断准确率达到98.86%,交叉验证平均诊断结果为98.57%;所提方法在噪声干扰下的故障诊断准确率也优于对比方法。因此,本文提出的tSSA–LSTM模型不仅可以更精准地诊断滚动轴承故障状态,而且具有更强的泛化能力及抗干扰能力,有效地提高了滚动轴承故障诊断的性能。

改进卷积胶囊网络的滚动轴承故障诊断方法

目前许多基于卷积网络的滚动轴承故障诊断方法受噪声信号以及负荷变化的影响,存在诊断效果不佳、泛化能力差的问题。针对此问题提出一种改进卷积胶囊网络的滚动轴承变工况故障诊断方法。该方法设计了多尺度非对称卷积模块,其中采用不同尺度的非对称卷积层对输入数据进行特征提取,在实现最大化提取数据中的特征信息的同时,还能够有效减少参数量;在该模块中引入通道注意力机制,能更好地提取有用的通道特征,提高该方法特征提取的能力;通过将网络中的全连接层改进为胶囊全连接层,使得胶囊在输出向量特征信息时,避免了特征信息在空间中的丢失。使用凯斯西储大学轴承数据集和东南大学变速箱数据集来验证所提方法的诊断性能,并与其他深度学习方法进行了比较。实验结果表明,与其他深度学习方法相比,具有较好的泛化性,效果更佳。

基于OFMD和FSC的滚动轴承复合故障诊断

针对滚动轴承的复合故障诊断问题,深入研究了一种基于优化特征模态分解和快速谱相关的复合故障诊断方法。首先,通过理论分析,提出脉冲能量因子指标来实现特征模态分解的参数选择以及最优分量的选取;然后,基于快速谱相关原理设计谱相关相对强度曲线和改进快速谱相关图,用于确定不同故障调制后对应的最优载波,对最优载波进行包络处理,从而分离轴承的复合故障特征,最终实现复合故障的准确性诊断。通过模拟故障试验和工程案例分析结果表明,该文所提方法相比于经验模态分解能够有效滤除噪声干扰,具有良好的鲁棒性,同时,避免了解卷积方法设定参数的缺陷,且与Autogram方法相比,能够有效分离复合故障特征,避免复合故障特征成分耦合。

基于振动信号的滚动轴承复合故障诊断研究综述

滚动轴承是旋转机械的关键部件。工作原理与工作环境决定了其具有易损、易耗特点。对其进行故障识别与诊断是保证设备运行安全可靠的必要手段。在工程应用中,轴承复合故障发生率高于单一故障,且特征识别较为困难。文章面向基于振动信号的滚动轴承复合故障诊断领域,按照传统诊断、智能诊断分类,从算法历程、基本原理、应用效果、算法优缺点等角度,对各种诊断方法进行了论述和分析,对轴承复合故障诊断方法的研究趋势进行展望。

新小波阈值法与VMD相结合的滚动轴承特征提取

针对滚动轴承故障信号弱以及难提取等问题,提出了一种新小波阈值方法与VMD相结合的轴承故障信号特征提取方法。首先,利用一种改进的指数小波阈值函数来优化传统小波降噪方法,克服其存在间断点和恒定偏差等问题;然后,结合VMD提取滚动轴承的有效故障特征;最后,以6205-RS号轴承内圈故障数据作为原始信号进行实验验证。实验结果表明,该方法能够有效提高降噪信号的信噪比,降低均方根误差,保证滚动轴承微弱故障信号特征提取的完整性和有效性。

基于共享近邻加权局部线性嵌入的轴承故障诊断

针对传统局部线性嵌入算法在挖掘局部流形结构时未充分考虑样本邻居分布信息,且在降维过程中默认样本具有相同的重要性导致提取鉴别特征不明显的问题,提出基于共享近邻的加权局部线性嵌入(weighted local linear embedding based on shared neighbors,SN-WLLE)算法,并用于滚动轴承故障诊断.该算法首先使用余弦距离划分样本邻域;其次计算样本邻域对相似度用以评估样本共享近邻信息,并结合样本的6种邻居分布修正局部结构挖掘,提高多共享近邻的k近邻重构准确性;接着从多流形的角度评估样本点与近邻点间的稀疏分布一致性,以获得样本的重要性指标,并在低维空间保持该信息,进而提取准确的鉴别特征;最后结合KNN分类器构建出完备的轴承故障诊断模型.采用凯斯西储大学轴承数据集和实验室测试平台轴承数据集,从可视化评估、定量聚类评估、故障识别精度评估及鲁棒性评估等方面进行分析.结果表明:SN-WLLE算法的F值保持在108以上水准,平均故障识别精度最低可达0.9734,不仅具有较好的类内紧致性与类间可分性,还对近邻参数k具有低敏感性.

基于全映射复合多尺度散布熵的滚动轴承故障诊断

为有效提取滚动轴承振动数据中的非平稳故障特征,将复合多尺度散布熵(CMDE)中的不同映射方式进行集成,形成了一种新的测量轴承振动信号复杂度和自相似度的方法,即全映射复合多尺度散布熵(FCMDE)。在此基础上,提出了基于FCMDE和k近邻(KNN)的滚动轴承故障诊断方法,利用FCMDE计算轴承振动信号的熵值并提取轴承的故障特征,将高维故障特征输入KNN分类器中进行滚动轴承的故障识别,采用西储大学和江南大学轴承数据集的验证结果表明,FCMDE方法能够有效识别滚动轴承的故障类型,准确率分别达到了100%和95.83%。

增强组合差分乘积形态学滤波的轴承故障特征提取方法

针对滚动轴承故障信号的非线性、非平稳、强噪声特性导致的常规时频域特征提取方法受限问题,提出一种增强组合差分乘积形态学滤波的轴承故障特征提取方法。在分析数学形态学4种基本运算的正、负冲击脉冲提取特性的基础上,运用级联、差分、乘积构造的一种新的组合差分乘积算子(combination difference multiply operator,CDMO)具备了同时提取正、负冲击脉冲的能力,并发挥梯度乘积运算对脉冲提取更敏感的优势,实现故障信息的充分提取。引入故障特征频率比指标优化CDMO结构元素参数,修正待处理信号的几何特征,提取与结构元素相匹配的信号特征信息。在CDMO滤波的基础上,借助三阶累积量切片谱技术能够抑制高斯噪声、突出二次耦合分量的优势,准确提取故障特征频率及其倍频,增强轴承故障特征提取能力并抑制噪声干扰。依托2种不同来源的工程实际信号并与经典故障特征提取方法对比分析,验证了所提方法的有效性。

基于深度卷积测量网络的滚动轴承压缩域故障特征提取方法

压缩感知可有效降低机械状态监测信号的数据存储和传输压力,而现有压缩感知方法在故障诊断的应用中存在压缩效率低下、信号重构过程缓慢等问题。本文利用自编码网络与压缩感知的对应关系,提出了一种基于深度卷积测量网络的滚动轴承压缩域故障特征提取方法。针对无噪声的故障信号样本难以获取的问题,提出一种利用故障机理构建数据集的方法,利用该仿真数据集训练得到的模型适用于不同工况下的实测轴承信号。构造网络层数由所需要的信号压缩率确定、隐含层与原信号的频率呈对应关系的深度卷积去噪自编码网络。截取训练完备的编码子网络(即深度卷积测量网络)代替传统的观测矩阵对滚动轴承振动信号进行压缩测量,实现压缩域的故障特征提取。仿真分析验证了所提数据集构造方法及压缩域特征提取方法的有效性。滚动轴承实验信号分析进一步验证了采用所提方法训练得到的深度卷积测量网络具有很好的泛化性,且能够在压缩率远低于传统压缩感知方法的情况下有效地提取轴承故障特征成分并进行故障诊断。

基于注意力机制和深度残差网络的滚动轴承故障诊断

针对现有的滚动轴承诊断模型特征提取能力不足、诊断准确率不高的问题,提出一种注意力机制与一维深度残差网络相结合的故障诊断方法。该方法首先通过引入残差结构来防止深度网络性能退化,然后结合注意力机制来提高网络的特征提取能力,最后使用原始的滚动轴承振动信号训练故障特征分类器。针对变工况故障诊断,本文采用小样本迁移学习框架。在两个开源实验平台上的结果表明,该方法能够有效地提高滚动轴承故障诊断的准确率,为实际应用提供一定的理论参考。

CEEMDAN和盲源分离在轴承复合故障诊断中的应用

滚动轴承的复合故障信号中往往含有多个特征信息及背景噪声,为更高效实现故障信息的提取,提出一种基于具有自适应白噪声的完备集成经验模态分解(CEEMDAN)和盲源分离的滚动轴承复合故障特征提取方法。对实验所获取的故障数据进行CEEMDAN分解,得出一组固有模态函数(IMF),利用加权峭度因子选取其中有效IMF重构信号,再将重构的信号进行BSS分离。对分离出的信号做解调包络分析,从其解调谱中提取故障信号的特征频率。结果证明了此方法可以有效地分离轴承的内外圈故障,使故障特征更易被提取。