基于信息融合与VPMCD的滚动轴承智能诊断研究

发布时间：2018-10-16 18:02

【摘要】：滚动轴承作为机械系统中关键的支撑部件,其运行状态的好坏直接影响到整台设备的安全运行。因此,开展对滚动轴承的故障诊断很具现实意义。在滚动轴承运行状态监测中,为了确保状态信息采集的完备性、可靠性,通常在测点上布置多个传感器。采集到的多通道信号大多数呈现出非线性、非平稳特性。但是传统的时频方法很难实现多通道信号的同步联合分析,而多维经验模态分解可以有效解决这个问题,同时能够保证所分解的多通道信号的IMF分量按照频率尺度对齐,这就为多通道信息融合提供有利条件。本文将其分别和多元多尺度熵及全矢谱技术相结合应用到滚动轴承故障特征提取中,并将基于信息融合技术提取的特征利用基于变量预测模型识别方法(VPMCD)进行识别。本文的主要研究工作如下:1、提出了基于多维经验模态分解(MEMD)与多元多尺度熵(MMSE)的退化特征提取方法。首先,利用MEMD算法对滚动轴承不同退化状态的多通道信号进行同步自适应分解;然后,对多尺度IMF分量重构的信号进行多元多尺度熵分析。最后通过实例分析,验证了该方法能有效反映滚动轴承退化趋势。2、提出了一种基于噪声辅助多维经验模式分解(NA-MEMD)与全矢谱结合的滚动轴承故障诊断方法——全矢NA-MEMD。首先,利用NA-MEMD对同源双通道信号和噪声辅助信号构成的多通道信息自适应分解成一系列IMF分量;然后,根据相关系数从同源双通道中选取包含故障主要信息的IMF分量进行重构;最后,将重构信号进行全矢信息融合来提取故障特征。通过仿真信号和实验信号验证该方法的有效性。3、应用变量预测模型(VPMCD)对上述两种基于信息融合的故障特征提取方法所提取的特征进行分类,实现轴承故障程度的识别和故障类型分类;首先将基于MEMD与MMSE提取的多尺度熵作为特征值构建特征向量,将其输入VPMCD实现轴承退化程度的识别;然后提取全矢NA-MEMD包络谱中各种故障特征频率处的幅值作为特征值构建特征向量,将其输入VPMCD分类器中实现故障类型的分类。最后通过两种方法最终实现滚动轴承的定性和定量智能诊断。
[Abstract]:As a key supporting part in mechanical system, rolling bearing has a direct impact on the safe operation of the whole equipment. Therefore, the development of rolling bearing fault diagnosis is of practical significance. In order to ensure the completeness and reliability of the state information collection, multiple sensors are usually arranged on the measuring points in order to ensure the completeness and reliability of the rolling bearing operation state monitoring. Most of the collected multi-channel signals are nonlinear and non-stationary. However, the traditional time-frequency method is difficult to realize the simultaneous joint analysis of multi-channel signals, and multi-dimensional empirical mode decomposition can effectively solve this problem and ensure that the IMF components of the decomposed multi-channel signals are aligned according to the frequency scale. This provides favorable conditions for multi-channel information fusion. In this paper, it is combined with multivariate multi-scale entropy and full-vector spectrum technology to extract fault feature of rolling bearing, and the feature extracted based on information fusion technology is identified by variable prediction model recognition method (VPMCD). The main work of this paper is as follows: 1. A method of extracting degenerate features based on multi-dimensional empirical mode decomposition (MEMD) and multivariate multi-scale entropy (MMSE) is proposed. Firstly, the multi-channel signals in different degraded states of rolling bearings are decomposed synchronously and adaptively using MEMD algorithm, and then the multi-element and multi-scale entropy analysis of the signals reconstructed by multi-scale IMF components is carried out. Finally, through the analysis of examples, it is proved that the method can effectively reflect the degradation trend of rolling bearing. 2. A new fault diagnosis method of rolling bearing based on noise assisted multidimensional empirical mode decomposition (NA-MEMD) and full vector spectrum is proposed, which is called full-vector NA-MEMD.. Firstly, NA-MEMD is used to decompose the multi-channel information composed of the homologous two-channel signal and the noise auxiliary signal into a series of IMF components, and then, according to the correlation coefficient, the IMF component containing the main fault information is selected from the homologous dual-channel to reconstruct. Finally, the reconstructed signal is fused with the full vector information to extract the fault features. Simulation signals and experimental signals are used to verify the effectiveness of the method. 3. The features extracted by the two fault feature extraction methods based on information fusion are classified by variable prediction model (VPMCD). Firstly, the multi-scale entropy extracted from MEMD and MMSE is used as the eigenvalue to construct the feature vector, and the bearing degradation degree is recognized by input VPMCD. Then the amplitudes of all kinds of fault feature frequencies in the full-vector NA-MEMD envelope spectrum are extracted as the eigenvalues to construct the eigenvector, and the feature vectors are input into the VPMCD classifier to realize the classification of fault types. Finally, the qualitative and quantitative intelligent diagnosis of rolling bearing is realized by two methods.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.33

【相似文献】