基于改进奇异谱分解的形态学解调方法及其在滚动轴承故障诊断中的应用
第53卷第7期2017年4月
机械工程学报
JOURNAL OF MECHANICAL ENGINEERING
Vol.53 No.7
Apr. 2017
DOI：10.3901/JME.2017.07.104
基于改进奇异谱分解的形态学解调方法及其 在滚动轴承故障诊断中的应用*
鄢小安 贾民平
(东南大学机械工程学院南京211189)
摘要：针对强背景噪声及干扰源信号影响下滚动轴承故障特征难以检测的问题，提出一种基于改进奇异谱分解的形态学解调方法用于轴承故障诊断。首先，为了克服奇异谱分析按经验性选取嵌入维数长度的缺陷，采用一种新的自适应信号处理方法——奇异谱分解(Singular spectrum decomposition, SSD)进行振动信号分析，该方法通过构建一个轨迹矩阵与自适应选择嵌入维数长度，将非平稳信号从高频至低频依次划分为若干个单分量信号。针对奇异谱分解在分量序列重构过程中两端数据会偏离实际数据值进而引起端点效应现象的问题，提出运用特征波形匹配延拓法对奇异谱分解进行改进，提高其对振动信号的分解质量，获得一系列更接近实际曲线的单分量序列。为准确提取单分量中蕴含的有用故障特征信息，提出一种基于特征能量比自适应确定结构元素最佳尺度的自互补顶帽变换对单分量信号进行形态学解调。最后，分析解调结果的频谱特征并提取突出频率成分，实现轴承故障类型的准确判别。仿真和实测信号分析验证了方法的有效性。
关键词：奇异谱分解；端点效应；形态学解调；滚动轴承；故障诊断
中图分类号：TH17
Morphological Demodulation Method Based on Improved Singular Spectrum Decomposition and Its Application in Rolling Bearing Fault Diagnosis
YAN Xiaoan JIA Minping
(School of Mechanical Engineering, Southeast University, Nanjing 211189)
Abstract：Aiming at the difficulty of fault feature extraction of rolling bearing under strong background noise and interference sources, a morphological demodulation method based on improved singular spectrum decomposition (SSD) is proposed to detect bearing fault. Firstly, in order to avoid the empirical selection of embedding dimension length in singular spectrum analysis, a novel self-adaptive signal processing method named SSD is used to analyze the vibration signal. In this method, track matrix first is built and embedding dimension length is adaptively selected, and then several single components can be obtained by decomposing the non-stationary signal. However, boundary effect will appear in component reconstruction process of SSD. Aiming at this problem, a waveform matching extension method is used to improve the SSD, and several single components respectively located in high frequency to low frequency can be obtained, which are closer to the actual components curve. Besides, in order to extract the useful fault characteristic information of single components, a self-complementary top-hat transform is proposed to analyze the single components, which its optimal structure elements scale can be obtained by feature energy ratio. Finally, the fault types of bearing can be accurately identified by the spectrum analysis of morphological demodulation results. Simulation and practical signal analysis prove the validity of the proposed method.
Key words：singular spectrum decomposition；end effect；morphological demodulation；rolling bearing；fault diagnosis
0 前言
滚动轴承是机械设备中的核心组件，在长期高速运转、交变载荷等恶劣工况下，极易产生局部损伤并演化成晚期故障，影响着整个机械传递系统的
*国家自然科学基金(51675098)和高等学校博士学科点专项科研基金(20130092110003) 资助项目。20160607收到初稿，20170115收到修改稿工作性能。因此，针对轴承早期损伤阶段进行有效检测备受关注。而在实际工程中，轴承振动信号往往表现为周期非平稳特性，其受到部件间多传递路径耦合作用的影响，振动信号中掺杂着强背景噪声和干扰源信号，致使轴承的故障特征提取极其复杂和困难。因此，如何有效恢复强背景噪声及干扰源信号下的轴承故障特征信息值得深入研究。
奇异谱分析(Singular spectrum analysis, SSA)是一种基于主成分分析的非参数化谱估计方法，包括
万方数据