基于相空间重构与平稳子空间分析的滚动轴承故障诊断

刘尚坤，唐贵基，庞彬

(华北电力大学机械工程系，河北保定071003)

摘要：针对如何分离淹没在与转子转速相关的背景信号及其它噪声中的滚动轴承故障特征信号问题，提出将相空间重构与平稳子空间分析相结合的滚动轴承故障诊断方法。用相空间重构实现对滚动轴承故障振动信号升维；用平稳子空间对高维信号中平稳、非平稳源信号进行区分，并对峭度值最大的非平稳源信号进行最小熵解卷积降噪；对降噪信号进行包络谱分析提取轴承故障特征频率。仿真信号、故障诊断实例表明，诊断效果优于基于EMD的包络解调方法。

关键词：相空间重构；平稳子空间分析；最小熵解卷积；滚动轴承；故障诊断

中图分类号:TH133.3；TH165+.3文献标志码：A

Fault diagnosis for rolling bearings based on phase space reconstruction and stationary subspace analysis

LIUShang-kun,TANGGui-ji,PANGBin(Mechanical Engineering Department, North China Electric Power University, Baoding 071003, China)

Abstract:Rolling bearing’s fault feature signals are non-stationary, transient and often submerged in background signals associated with the components of rotating speed and other noises. How to separate the fault feature signals from the rolling bearing’s blind sources is an important issue. A method combining the phase space reconstruction technique with the stationary subspace analysis (SSA) was proposed. The fault vibration signal’s dimension was increased by the phase space technique and the non-stationary source components in the high-dimensional signals were distinguished from the stationary components by using the SSA. Then, the selected non-stationary component with the maximum kurtosis value was de-noised by the minimum entropy deconvolution (MED). Finally, the de-noised non-stationary component was analyzed by the envelope spectrum to extract the fault characteristic frequency. The simulation and experiment results indicate that the proposed method can extract the fault frequency better than the envelope demodulation method based on empirical mode decomposition (EMD).

Key words:phase space reconstruction; stationary subspace analysis; minimum entropy deconvolution; rolling bearing; fault diagnosis

滚动轴承作为旋转机械系统核心元件之一，对旋转机械设备的正常运行至关重要。由于承受载荷复杂且工作环境特殊，滚动轴承亦为脆弱部件之一，约30%的机械故障由滚动轴承引起。因此对滚动轴承故障特征诊断，尤其早期微弱故障诊断意义重大[1]。

相空间重构技术是为非线性问题处理方法[2]，适用于机械冲击故障的典型非线性特点，已证明重构高维相空间与系统的动力学特征相同，能在高维空间反映各故障状态下系统的动力学特征更清晰[3]。平稳子空间分析法(Stationary Subspace Analysis，SSA)可从高维信号中分离出平稳源、非平稳源信号[4-5]。较独立分量分析等传统盲源分离方法，该方法对原信号各分量的独立性并无要求。……