基于流形学习的滚动轴承早期微弱故障诊断研究

发布时间：2018-02-11 08:27

本文关键词： 滚动轴承经验模态分解流形学习局部切空间排列算法 K-means分类器　出处：《安徽工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：滚动轴承发生故障时,其振动信号往往呈现出强烈的非平稳以及非线性特征,因此,如何从这些非平稳、非线性信号中获取故障特征信息并进行故障类型的识别是滚动轴承故障诊断的关键。由于经验模态分解(Empirical Mode Decomposition,EMD)能够对振动信号进行自适应分解从而提供信号的时域、频域局部信息,因此,本文将其用于提取滚动轴承振动信号的故障特征,构造原始特征量。但是,这些原始特征量维数较高、信息冗余,不利于分类识别。而流形学习局部切空间排列算法(Local Tangent Space Alignment Algorithm,LTSA)作为一种非线性信号处理方法,具有良好的维数约简能力。为此,将LTSA与EMD相结合,提出一种基于经验模态分解与流形学习的滚动轴承故障诊断方法:首先,利用EMD对滚动轴承振动信号进行自适应分解,并计算分解所得的内禀模态分量(Intrinsic Mode Function,IMF)的协方差矩阵特征值,组成滚动轴承原始特征集;然后利用LTSA对原始特征集进一步的融合提取;将所得新特征输入到K-means分类器中进行故障识别。论文主要内容如下:(1)以调心球轴承为实验对象,利用SG双金属电刻机进行故障加工,从而模拟调心球轴承滚子、外圈及内圈的早期微弱点蚀故障,并利用BVT-5轴承振动测量仪采集了调心球轴承正常、滚子故障、外圈故障及内圈故障四种状态的加速度振动信号。(2)通过对调心球轴承四种状态下振动信号的实例研究,验证基于经验模态分解与流形学习的滚动轴承故障诊断方法的有效性。(3)通过对基于时频特征提取与流形学习、基于相空间重构与流形学习和基于经验模态分解与流形学习这三种方法对不同轴向加载情况下的调心球轴承故障诊断实验对比分析,结果表明基于经验模态分解与流形学习的滚动轴承故障诊断方法最优,尤其在轴承早期微弱故障诊断方面,优势更加明显。
[Abstract]:When rolling bearings fail, their vibration signals often show strong nonstationary and nonlinear characteristics. Obtaining fault characteristic information and identifying fault types in nonlinear signals is the key to fault diagnosis of rolling bearings. Empirical Mode decomposition (EMD) can adaptively decompose vibration signals to provide the time domain of signals. Because of the local information in frequency domain, this paper uses it to extract the fault feature of vibration signal of rolling bearing and construct the original characteristic quantity. However, the dimension of the original characteristic quantity is high, and the information is redundant. However, as a nonlinear signal processing method, Local Tangent Space Alignment algorithm LTSA (Local Tangent Space Alignment algorithm) has good dimensionality reduction ability. Therefore, LTSA is combined with EMD. A fault diagnosis method for rolling bearing based on empirical mode decomposition and manifold learning is proposed. Firstly, the vibration signal of rolling bearing is decomposed adaptively by EMD. The eigenvalues of intrinsic Mode functions are calculated to form the original feature set of rolling bearing, and then the original feature set is extracted by LTSA. The new features are input into K-means classifier for fault identification. The main contents of this paper are as follows: 1) taking the centering ball bearing as the experimental object, the SG bimetallic electrolithography machine is used to process the fault, thus simulating the centering ball bearing roller. Early weak pitting fault of outer ring and inner ring, and using BVT-5 bearing vibration measuring instrument to collect normal ball bearing, roller fault, Vibration signal of acceleration in four states of outer ring fault and inner ring fault. The effectiveness of the fault diagnosis method for rolling bearings based on empirical mode decomposition and manifold learning is verified. Based on the three methods of phase space reconstruction and manifold learning and empirical mode decomposition and manifold learning, the fault diagnosis experiments of spherical bearings under different axial loading conditions are compared and analyzed. The results show that the fault diagnosis method of rolling bearings based on empirical mode decomposition and manifold learning is optimal, especially in the early weak fault diagnosis of bearings.
【学位授予单位】：安徽工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.33

【相似文献】