基于CYCBD与HRE的滚动轴承故障智能诊断方法研究.pdf

图书分类号图书分类号 TH133.3TH133.32 2 密级密级 非非密密 UDC 注注 1 621 621 硕硕 士士 学学 位位 论论 文文 基于基于 CYCBD 与与 HRE 的的滚动滚动轴承故障轴承故障智能智能诊断方法研究诊断方法研究 周杰周杰 指导教师（姓名、职称）指导教师（姓名、职称） 杜文华杜文华 教授教授 申请学位级别申请学位级别 工学硕士工学硕士 专业名称专业名称 机械工程机械工程 论文提交日期论文提交日期 2020 年年 5 月月 8 日日 论文答辩日期论文答辩日期 2020 年年 6 月月 6 日日 学位授予日期学位授予日期 年年 月月 日日 论文评阅人论文评阅人 赵韡赵韡、曾志强、李海虹、曾志强、李海虹 答辩委员会主席答辩委员会主席 靳宝全靳宝全 2020 年 6 月 8 日 注 1注明国际十进分类法 UDC的分类 万方数据 万方数据 中北大学学位论文 基于 CYCBD 与 HRE 的滚动轴承故障智能诊断方法研究 摘要 滚动轴承作为旋转机械的易损部件，对其进行故障诊断与状态识别具有重要意义。 在实际工况中，特别是在矿用环境下，采集的轴承振动信号通常包含大量噪声，而且故 障信号在传输路径上与其他振动信号耦合致使故障诊断难度加大。另外，滚动轴承振动 信号的非线性特征以及同一故障信号振幅的波动使得故障特征难以准确提取， 故障识别 准确率低。针对以上问题，本文以矿用滚动轴承为研究对象，对轴承故障信号降噪、特 征提取、状态识别方法展开研究。具体研究内容如下 （1）对最大循环平稳性盲反褶积（CYCBD）理论进行了研究。针对不当的滤波器 长度与循环频率影响 CYCBD 效果的问题，本文提出了基于自适应 CYCBD 的信号降噪 方法，并结合 Teager 能量算子解调法提取故障特征频率从而人为识别故障。首先，通过 仿真信号分析了 CYCBD 提取脉冲效果随滤波器长度与循环频率变化的规律；其次，针 对循环频率的确定问题，本文提出了基于形态包络自相关函数的循环频率估计方法；然 后，针对滤波器长度的确定问题，提出了综合考虑滤波效果与效率的性能效率比指数， 进一步结合等步长搜索策略提出了滤波器长度的自适应原则；最后，通过仿真与实验验 证了所提出方法的合理性。 （2）针对轴承故障信号的非线性特征以及同一故障信号振幅波动致使故障特征难 以准确提取的问题，本文提出了基于层次极差熵的故障特征提取方法。熵作为一种非线 性分析方法，能够表征不同故障的复杂程度。但常用的熵如样本熵等都没有考虑幅值波 动带来的熵值变化致使故障错误识别。极差熵虽然考虑了幅值波动的因素，但无法从多 个尺度分析信号。本文结合层次分析的优点，提出了层次极差熵指标。实测信号分析结 果表明，相比于只考虑低频成分的多尺度极差熵以及未考虑振幅波动的层次模糊熵，层 次极差熵提取的故障特征更加有效。 （3） 针对故障识别准确率低的问题， 本文提出了基于自组织模糊逻辑分类器 （SOF） 万方数据 中北大学学位论文 的故障识别方法。SOF 分类器具有高精度、高效率的优点，且在新增样本类别后可在原 有基础上进行训练，无需重新开始训练，但建立其结构时涉及到粒度级别与样本距离度 量方法的确定。本文通过实测信号的分析确定了粒度级别与样本距离度量方法，并结合 LDA 降维提出了轴承故障智能识别方法。实验结果表明1）特征冗余会使分类精度降 低， 通过 LDA 降维能够提高故障识别准确率。 2） 相比于未考虑振幅波动的层次模糊熵， 层次极差熵得到的分类精度更高。3）噪声会降低故障识别率，通过 CYCBD 降噪能够 提高故障识别准确率。 关键字矿用轴承，最大循环平稳性盲反褶积，层次极差熵，故障诊断，智能识别 万方数据 中北大学学位论文 Research on Intelligent Fault Diagnosis of Rolling Bearing Based on CYCBD and HRE Abstract As a vulnerable part of rotating machinery, rolling bearing fault diagnosis and state identification have important practical significance. In the actual working condition, the collected bearing vibration signal usually contains a lot of noise, and the fault signal is coupled with other vibration signals in the transmission path, which makes the fault diagnosis more difficult. Besides, the nonlinear characteristics of the bearing vibration signal and the fluctuation of the amplitude of the same fault signal make it difficult to extract the fault features accurately, and the accuracy of fault identification is low. Aiming at the above problems, this paper takes rolling bearing as the research object, and studies the s of bearing fault signal noise reduction, feature extraction and state recognition. The specific research contents are as follows 1 Study the theory of maximum cyclic stationary blind deconvolution CYCBD. Aiming at the problem that improper filter length and cycle frequency affect the effect of CYCBD, this paper proposes a signal denoising based on adaptive CYCBD. Combined with the Teager energy demodulation , a fault diagnosis for artificially identifying faults is proposed. Firstly, the law of the pulse extraction effect of CYCBD changing with the filter length and cycle frequency is analyzed by simulation signal. Secondly, in order to determine the cycle frequency, this paper proposes a of cycle frequency estimation based on morphological envelope autocorrelation function. Finally, in order to determine the length of the filter, the perance efficiency ratio index considering the filter effect and efficiency is proposed. Furthermore, the adaptive principle of the filter length is proposed by combining the equal step search strategy. Finally, the rationality of the proposed is verified by 万方数据 中北大学学位论文 simulation and experiment. 2 Aiming at the problem that the non-linear features of bearing fault signals and the fluctuation of the amplitude of the same fault signal make it difficult to extract the fault features accurately, this paper proposes a fault feature extraction based on hierarchical range entropy. As a nonlinear analysis , entropy can characterize the complexity of different faults. However, the commonly used entropy, such as fuzzy entropy, does not take into account the change in the entropy caused by the amplitude fluctuations, which causes the fault to be wrongly identified. Although the range entropy considers the amplitude fluctuation, it cannot analyze the signal from multiple scales. Based on the advantages of AHP, this paper puts forward the index of the entropy of the range of the hierarchy. The analysis results of measured signals show that the fault feature extracted by hierarchical range entropy is more effective than that of multi-scale range entropy which only considers low frequency components and hierarchical fuzzy entropy which does not consider amplitude fluctuation. 3 For the problem of low accuracy of fault identification, this paper proposes a fault identification based on self-organizing fuzzy logic classifier SOF. The SOF classifier has the advantages of high accuracy and high efficiency, and can be trained on the original basis after the new sample category is added, without restarting training, but the establishment of its structure involves the determination of the granularity level and the sample distance measurement . In this paper, the measurement of filter level and sample distance is determined through the analysis of the measured signal, and the intelligent identification of bearing fault is proposed in combination with the LDA dimensionality reduction . The experimental results show that 1 Feature redundancy will reduce the classification accuracy, and LDA dimensionality reduction can improve the accuracy of fault recognition. 2 Compared with the hierarchical fuzzy entropy without considering the amplitude fluctuation, the classification accuracy obtained by the hierarchical range entropy is higher. 3 Noise will reduce the fault recognition rate, and CYCBD noise reduction can improve the fault recognition accuracy rate. 万方数据 中北大学学位论文 Keywords Rolling bearing, CYCBD, HRE, Fault diagnosis, Intelligent identificatio 万方数据 中北大学学位论文 I 目目 录录 1 绪论绪论 1.1 研究背景及意义研究背景及意义 ......................................................................................................... 1 1.2 滚动轴承故障诊断概述滚动轴承故障诊断概述 ............................................................................................. 2 1.3 滚动轴承故障诊断技术国内外研究现状滚动轴承故障诊断技术国内外研究现状 .................................................................. 3 1.3.1 基于反褶积的滚动轴基于反褶积的滚动轴承振动信号处理方法研究现状承振动信号处理方法研究现状 ................................... 4 1.3.2 滚动轴承故障特征提取方法研究现状滚动轴承故障特征提取方法研究现状 ........................................................... 5 1.3.3 滚动轴承故障分类识别方法研究现状滚动轴承故障分类识别方法研究现状 .......................................................... 8 1.4 问题提出问题提出 ...................................................................................................................... 8 1.5 本文的研究内容及结构本文的研究内容及结构 .............................................................................................. 9 1.5.1 论文研究内容论文研究内容 .................................................................................................. 9 1.5.2 论文组织结构论文组织结构 ................................................................................................ 10 2 滚动轴承故障机理分析及相关理论滚动轴承故障机理分析及相关理论 2.1 引言引言 ............................................................................................................................ 12 2.2 滚动轴承的基本结构滚动轴承的基本结构 ............................................................................................... 12 2.3 滚动轴承的故障类型与振动机理滚动轴承的故障类型与振动机理 ........................................................................... 13 2.3.1 滚动轴承的故障类型滚动轴承的故障类型 .................................................................................... 13 2.3.2 滚动轴承的振动机理与特征频率滚动轴承的振动机理与特征频率 ................................................................ 14 2.4 盲反褶积理论盲反褶积理论 ........................................................................................................... 15 2.4.1 盲反褶积基础盲反褶积基础 ................................................................................................. 15 2.4.2 CYCBD 理论理论 .................................................................................................. 16 2.5 熵理论熵理论 ........................................................................................................................ 18 2.5.1 近似熵与样本熵近似熵与样本熵 ............................................................................................. 18 2.5.2 极差熵极差熵 ............................................................................................................. 20 万方数据 中北大学学位论文 II 2.6 SOF 分类器分类器 ................................................................................................................ 21 2.6.1 离线训练环节离线训练环节.................................................................................................. 21 2.6.2 在线训练环节在线训练环节.................................................................................................. 23 2.6.3 测试环节测试环节 .......................................................................................................... 24 2.7 本章小结本章小结 ..................................................................................................................... 25 3 基于自适应基于自适应 CYCBD 的振动信号降噪方法的振动信号降噪方法 3.1 引言引言 ............................................................................................................................. 26 3.2 CYCBD 的参数对其效果的影响分析的参数对其效果的影响分析 ..................................................................... 26 3.3 循环频率的估计方法循环频率的估计方法 ............................................................................................... 33 3.3.1 形态包络自相关形态包络自相关函数函数 .................................................................................... 33 3.3.2 仿真分析仿真分析 ......................................................................................................... 33 3.4 CYCBD 滤波器长度的自适应原则滤波器长度的自适应原则 ......................................................................... 34 3.5 仿真分析仿真分析 .................................................................................................................... 37 3.5.1 自适应自适应 CYCBD 结果结果 .................................................................................... 38 3.5.2 对比分析对比分析 ......................................................................................................... 41 3.6 实验验证实验验证 .................................................................................................................... 43 3.6.1 实验台实验台 .............................................................................................................. 43 3.6.2 自适应自适应 CYCBD 的结果的结果 ................................................................................ 44 3.6.3 对比分析对比分析 ......................................................................................................... 46 3.7 本章小结本章小结 ..................................................................................................................... 48 4 基于层次极差熵的轴承故障特征提取方法基于层次极差熵的轴承故障特征提取方法 4.1 引言引言 ............................................................................................................................. 49 4.2 极差熵性能对比分析极差熵性能对比分析 ............................................................................................... 50 4.2.1.信号长度的影响信号长度的影响 ............................................................................................. 50 4.2.2 信号幅值的影响信号幅值的影响 ............................................................................................. 51 4.3 层次极差熵层次极差熵................................................................................................................. 51 万方数据 中北大学学位论文 III 4.3.1 层次分析层次分析 ........................................................................................................ 52 4.3.2 层次极差熵层次极差熵 ..................................................................................................... 54 4.3.3 参数选择参数选择 ......................................................................................................... 55 4.4 仿真对比仿真对比 .................................................................................................................... 56 4.5 实例验证实例验证 .................................................................................................................... 58 4.5.1 实例一实例一 ............................................................................................................. 58 4.5.1 实例二实例二 ............................................................................................................. 61 4.6 本章小结本章小结 .................................................................................................................... 62 5 基于基于 SOF 分类器的故障识别及实例验证分类器的故障识别及实例验证 5.1 引言引言 ............................................................................................................................ 63 5.2 SOF 分类器性能分析分类器性能分析 ............................................................................................... 63 5.2.1 SOF 分类器性能对比分类器性能对比 .................................................................................... 63 5.2.2 SOF 分类器参数确定分类器参数确定 .................................................................................... 64 5.3 基于层次极差熵与基于层次极差熵与 SOF 分类器的故障识别流程分类器的故障识别流程 ................................................. 66 5.4 案例分析案例分析 .................................................................................................................... 68 5.4.1 案例一案例一 ............................................................................................................ 68 5.4.2 案例二案例二 ............................................................................................................. 73 5.5 本章小结本章小结 .................................................................................................................... 77 6 总结与展望总结与展望 6.1 总结总结 ............................................................................................................................ 78 6.2 论文创新点论文创新点 ................................................................................................................ 79 6.3 展望展望 ........................................................................................................................... 79 参考文献参考文献 攻读硕士期间发表的论文及所取得的研究成果攻读硕士期间发表的论文及所取得的研究成果 致谢致谢 万方数据 中北大学学位论文 1 1 绪论 1.1 研究背景及意义 在科学技术飞速发展的今天，现代机械设备正朝着高智能、高精度的方向发展，其 性能受到了广泛的关注。 机械设备功能越多， 其规模也是越大， 各个部件之间相互联系、 相互影响， 甚至是不同的设备之间联系越发紧密， 形成一个统一的整体。 在这种情况下， 如果一个部件或零件发生了故障，其自身运行状况会急剧恶化，甚至可能导致部件或零 件发生故障，进而造成停工、产品质量不能得到保障，使企业效益大大降低，严重的情 况下还会导致重大伤亡事故[1]。因此，对机械设备进行故障分析和诊断，从而使机械设 备中存在的缺陷被发现并解决，这对企业的效益是极其重要的。一方面及时发现故障会 避免企业工人出现不必要的伤亡及带来的经济损失， 另一方面这会避免