局部特征尺度分解方法及其在机械故障诊断中的应用研究

发布时间：2018-03-14 04:33

本文选题：局部特征尺度分解　切入点：自适应　出处：《湖南大学》2014年博士论文　论文类型：学位论文

【摘要】：机械设备状态监测和故障诊断对于保证机械设备的健康运行、早期故障的预警以及故障发生的正确定位与诊断都有重要的理论和实际意义。机械设备振动信号大都是非线性和非平稳信号，因此，机械设备故障诊断的关键是如何从非线性和非平稳信号中提取故障特征并进行模式识别。时频分析方法由于能够同时提供振动信号时域和频域的局部信息而在机械故障诊断中得到了广泛应用。近年来，小波变换、经验模态分解(Empirical Mode Decomposition, EMD)、局部均值分解(Local Mean Decomposition, LMD)等时频分析方法由于特别适合处理机械振动信号而被国内外相关学者应用到机械故障诊断领域，并取得了许多非常可喜的研究成果，但这些时频分析方法都有各自不同的局限性。局部特征尺度分解(Local Characteristic-scale Decomposition, LCD)是一种新的非平稳信号自适应分析方法，该方法在定义瞬时频率具有物理意义的单分量信号——内禀尺度分量(Intrinsic Scale Component, ISC)基础上，自适应地将一个复杂信号分解为若干个ISC分量之和，从而得到原始信号完整的时频分布。与EMD、LMD等方法相比，LCD在端点效应的抑制、计算速度和分解效果等方面具有一定的优越性。论文在国家自然科学基金项目（编号：51075131）的资助下，，对局部特征尺度分解方法进行了深入的研究，对其理论进行了完善，在此基础上，将局部特征尺度分解方法及其理论应用于旋转机械故障诊断。论文主要研究工作和创新性成果有： 1.对LCD方法的理论进行了研究，解决了均值曲线定义存在的不足、模态混叠的抑制等问题。 (1)将LCD方法与EMD进行了对比分析，仿真和机械故障振动信号的分析结果表明了LCD方法的优越性； (2)针对LCD中均值曲线中直线连接极值会与数据交叉的问题，提出了基于分段多项式的改进LCD方法，并将其应用于仿真和转子碰摩故障振动信号分析，结果表明了ILCD方法的有效性； (3)针对基于筛分的自适应信号分解方法中由于均值曲线不同而导致分解结果差异的问题，提出了一种新的非平稳信号的自适应分解方法——广义局部特征尺度分解（Generalized LCD, GLCD），GLCD通过从不同均值曲线筛分的结果中选择最优分量，再对剩余信号重复筛分过程，从而保证了最终的分解结果也是最优的。分别采用仿真和机械故障振动信号将其与EMD、LCD方法进行了对比，结果表明GLCD方法在正交性、分解能力等方面有一定的优越性，从而能够得到更好的分解结果。 (4)针对LCD分解过程中可能出现的模态混叠问题，分别提出了部分集成和完备总体平均局部特征尺度分解等方法，对仿真和机械故障振动信号的分析结果表明，所提出的方法能够有效地抑制LCD的模态混叠现象。 2.对ISC分量的瞬时频率估计方法进行了研究，提出了两种新的瞬时频率估计方法和多分量信号解调方法。 (1)针对希尔伯特变换、能量算子解调和标准希尔伯特变换等常用的瞬时估计方法存在的不足，提出了一种新的瞬时频率估计方法——经验包络法，仿真信号分析结果表明了其优越性。同时，针对机械故障振动信号的调制特性，提出了基于LCD的经验包络解调方法，并将其应用于滚动轴承的故障诊断，结果表明了所提出方法的有效性； (2)针对标准希尔伯特变换和直接正交法存在的问题，提出了归一化正交法。并针对多分量信号的解调问题，提出了基于GLCD和归一化正交的时频分析方法，仿真和实验信号的分析结果表明了所提出方法的优越性； 3.对LCD方法在机械故障诊断中的应用进行了研究，与其它数学方法相结合，提出了多种基于LCD的机械故障诊断方法，实验数据分析结果表明了LCD方法可以有效地应用于机械故障诊断。 (1)在对多尺度模糊熵进行改进的基础上，提出了基于LCD和模糊熵的振动信号自适应多尺度复杂性分析方法；在多尺度排列熵的基础上，提出了基于LCD和排列熵的振动信号自适应多尺度随机性检测方法，并将它们应用于机械故障振动信号特征的提取； (2)将基于变量预测模型的模式识别(Variable Predictive Model based ClassDiscriminate, VPMCD)方法的应用扩展到机械故障诊断领域，VPMCD方法基于特征量之间的内在关系建立预测模型，通过对特征量进行预测，从而实现模式的分类。在VPMCD的基础上，结合LCD，提出了相应的旋转机械智能故障诊断方法。
[Abstract]:Mechanical equipment condition monitoring and fault diagnosis to ensure the healthy operation of mechanical equipment, has important theoretical and practical significance to correctly locate the fault early warning and fault early diagnosis. And the vibration signals of mechanical equipment are nonlinear and non-stationary signal, so the key to fault diagnosis of mechanical equipment is from nonlinear and non-stationary signal in fault feature extraction and pattern recognition. The time-frequency analysis method for local information can also provide the vibration signal in time domain and frequency domain is widely used in mechanical fault diagnosis.
In recent years, wavelet transform, empirical mode decomposition (Empirical Mode, Decomposition, EMD), the local mean decomposition (Local Mean Decomposition, LMD) and other methods as particularly suitable for processing of mechanical vibration signals by domestic and foreign scholars applied to mechanical fault diagnosis field frequency analysis, and made a lot of very gratifying results, but these the time-frequency analysis method has its own limitations. The local characteristic scale decomposition (Local Characteristic-scale Decomposition, LCD) is a new adaptive non-stationary signal analysis method, the single component of intrinsic scale components the signal has a physical meaning in the definition of instantaneous frequency (Intrinsic Scale Component, ISC) based on adaptive to be a complex signal is decomposed into several ISC components, and thus, the time-frequency distribution of original signal integrity is obtained. Compared with EMD, LMD, L Inhibition of CD in the end effect, has certain advantages of calculating speed and decomposition effect. Based on the project of National Natural Science Foundation (No. 51075131) under the support of local characteristic scale decomposition method is studied, the theory was improved based on the local characteristic scale decomposition the theory and method applied in fault diagnosis of rotating machinery.
The main research work and innovative achievements of the paper are as follows:
1. the theory of LCD method is studied, which solves the problem of the deficiency of the definition of the mean curve and the suppression of the modal aliasing.
(1) the LCD method and the EMD are compared and analyzed. The simulation and the analysis results of the mechanical fault vibration signal show the superiority of the LCD method.
(2) in view of the problem that the line connection extremum in the mean value curve of LCD will be intersecting data, an improved LCD method based on piecewise polynomial is proposed. It is applied to simulation and rotor rub impact fault vibration signal analysis, and the results show the effectiveness of ILCD method.
(3) based on adaptive signal decomposition method in screening due to the mean curve due to different decomposition results of different problems, put forward a new kind of non-stationary signal adaptive decomposition method -- generalized local characteristic scale decomposition (Generalized LCD, GLCD, GLCD) by selecting the optimal component from different mean curve screening results. And then the rest of the signal repeat screening process, so as to ensure the final result of the decomposition is optimal. By simulation and mechanical fault vibration signal with the EMD, compared with LCD method, the result showed that GLCD method in orthogonal, have certain superiority decomposition ability, which can get better decomposition results.
(4) for possible modal decomposition process of LCD mixing, respectively put forward some integrated and complete the overall average local characteristic scale decomposition method, the analysis results of the simulation and mechanical fault vibration signals show that the proposed modal method can effectively suppress LCD aliasing.
2. the method of instantaneous frequency estimation for ISC component is studied, and two new instantaneous frequency estimation methods and multi component signal demodulation methods are proposed.
(1) based on Hilbert transform, instantaneous energy operator demodulation and standard Hilbert transform popular estimation method and its shortcomings, puts forward a new method of instantaneous frequency estimation -- envelope method, signal analysis simulation results show its superiority. At the same time, modulation characteristics for mechanical fault vibration signal, puts forward experience envelope demodulation method based on LCD and its application in fault diagnosis of rolling bearing. The results show the effectiveness of the proposed method;
(2) according to the existing standard of Hilbert transform and direct orthogonal method, proposed the normalized orthogonal method. Aiming at the problem of multi - component signal demodulation, time-frequency analysis method GLCD and normalized orthogonal was put forward based on the analysis results of the simulation and experiment signals show that the proposed method is superior;
3., the application of LCD in mechanical fault diagnosis is studied. Combined with other mathematical methods, a variety of LCD based mechanical fault diagnosis methods are put forward. Experimental data analysis results show that LCD method can be applied to mechanical fault diagnosis effectively.
(1) improved based on multi-scale fuzzy entropy, adaptive method is proposed to analyze vibration signal LCD and fuzzy entropy based on multi scale complexity based on multiscale permutation entropy; on the proposed adaptive vibration signal LCD and multiscale permutation entropy method based on random detection and extraction, and their application on the characteristics of mechanical fault vibration signals;
(2) the pattern recognition based on variable prediction model (Variable Predictive Model based ClassDiscriminate, VPMCD) the application of the method is extended to the field of mechanical fault diagnosis, the VPMCD method based on the intrinsic relationship between the features to establish prediction model, through forecasting the characteristic quantities, so as to realize the pattern classification. On the basis of VPMCD, combined with LCD, put forward the corresponding intelligent fault diagnosis of rotating machinery.

【学位授予单位】：湖南大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TH165.3

【参考文献】