【作者】 胡劲松；

【导师】 吴昭同； 严拱标； 杨世锡；

【作者基本信息】 浙江大学 ， 机械制造， 2003， 博士

【摘要】 基于经验模态分解(Empirical Mode Decomposition，EMD)的时频分析方法，被认为是近年来对以傅里叶变换为基础的线性和稳态谱分析的一个重大突破。本文结合“基于独立分量分析的旋转机械故障诊断新方法的研究”国家自然科学基金项目(50205025)和“基于CORBA分布计算策略构建远程监诊的新方法研究”省自然科学基金项目(501061)，以面向大型旋转机械故障诊断的时频分析为目标，研究了该方法的基本原理和算法：针对其端点效应等不足，提出了新的数据延拓技术；针对EMD算法的精度和速度上的不足，提出了新的改进算法；设计和开发了基于EMD时频分析的软件系统；以模拟转子实验台为基础构建了实验系统，并进行了实验研究。本文的研究成果将为旋转机械信号处理与故障诊断，尤其是为非线性和非稳态的故障分析与诊断给出了一条新的途径。全文的主要研究内容如下： 第一章论述了面向旋转机械故障诊断时频分析的研究意义；介绍了旋转机械故障诊断理论与技术概况；综述时频分析方法的发展及其在机械故障诊断中的应用及存在问题；介绍了基于EMD的时频分析方法及国内外研究现状，以及把该方法引入旋转机械故障诊断的重要性；最后，提出了本文的研究思路与研究内容，并给出了本文的总体框架和创新之处。 第二章介绍了基于EMD时频分析方法的一些基本概念；阐述了EMD方法和基于EMD的希尔伯特变换的基本原理和算法；在此基础上，用基于EMD的时频分析方法对仿真和实际信号进行了分析验证，结果表明把该方法引入旋转机械振动信号时频分析和故障诊断领域是有效的。 第三章介绍了EMD算法端点效应的机理；然后系统地研究了直接信号序列延拓技术、基于时间序列预测和基于神经网络预测的数据序列延拓技术的特点及性能；最后，对各种延拓技术进行了比较研究，分析了各种延拓技术的优缺点。提出了基于时间序列建模和基于BP、径向基神经网络的EMD数据延拓技术。 第四章从提高速度与精度两方面着手，对EMD算法进行了系统地研究。首先研究样条插值的端点条件对EMD算法的影响，结果表明第一个端点条件和第二个端点条件对EMD分解效果较好；然后研究基于低次插值的EMD算法和效果，结果表明该算法速度较快但精度不好；提出了基于高次(高于三次)样条插值的EMD算法并考察了效果，表明高次样条插值能提高EMD算法精度，但耗时增加；在论述简单平均的EMD算法的基础上，分析了基于自适应时变滤波的EMD算法和效果，结果表明该算法没有明显的优点；根据EMD算法的使用特点，提出了基于有效数据的EMD快速算法，并对该算法的效果进行了系统地研究，结果表明该算法效果很好。 第五章针对旋转机械信号进行EMD时频分析的要求，提出了提高整周期采样精度的预测算法和适合于单片机计算的近似公式；论述了模态裂解现象，并通过仿真信号和实际信号的研究，提出了使用滤波技术来减弱模态裂解现象的方法；给出了适合于旋转机械振动信号基于EMD时频分析的工作步骤；研究了基于EMD的旋转机械振动信号自适应滤波技术。 第六章把基于EMD的时频分析方法分别与短时傅里叶变换、Wigner分布分析、小波变换进行了比较研究，结果表明，基于EMD时频分析方法比上述方法有效；针对Wigner分布分析的交叉项干扰，提出了基于EMD的Wigner分布分析新方法，研究结果表明新方法能有效地消除交叉项干扰。 第七章以模拟转子实验台为基础构建了实验系统；用基于EMD的时频分析方法对转子冲临界过程进行了实验研究与分析；在论述几种典型故障特点的基础上用基于EMD的时频分析方法对这几种故障振动信号进行了实验研究与分析，获得了时频特征。浙江大学博士学位论文摘要 第八章分析、设计并实现了基于EMD时频分析的软件系统;结合Matlab和C++Builder的优点，介绍了C++ Bui lder和Matlab混合编程的实现方法;介绍了基于EMD时频分析软件系统的基本组成和功能。 第九章总结了全文的研究成果和创新之处;并对今后的工作提出了展望。更多还原

【Abstract】 Compared to the Fourier-based linear and stationary spectral analysis, the empirical mode decomposition (EMD) based time-frequency analysis method was considered a great breakthrough. Based on the "Research on the new method for rotating machine faults diagnosis based on independent component analysis"(National Nature Science Fund Project, No: 50205025) and "Study on the new method of CORBA-based distributed calculation tactics construction in remote condition monitoring and diagnosis"(Zhejiang Province Nature Science Fund Project, No: 501061), this paper aimed to the rotating machinery faults diagnosis oriented time-frequency analysis. The EMD-based basic theories and methods were studied; To weaken the EMD method’s end effect, some new data extention means were proposed; To improve the EMD method’s precision and efficiency, some new algorithms were proposed; The EMD based time-frequency analysis software was designed and developed; A experiment system at the base of rotor test bed was built to study the rotor faults diagnosis using the new time-frequency analysis method, then studed several typical faults. The study results of the dissertation provide a new way to the faults analysis and diagnosis of rotating machine, especially to non-linear and non-stationary’s. The details were studied as follows:Chapter one explained the importance of the rotating machinery faults diagnosis oriented time-frequency analysis study and briefly introduced the general situation of faults diagnosis of the rotating machine. The developing and the current situations of the time-frequecy theories were summarized. At last, the way of study,the main contents, general structure scheme and innovation points of this dissertation were presented.Chapter two introduced the basic ideas of EMD theories briefly, and then discussed the EMD-based time-frequency analysis method’s basic agrithom. the conclusion was deduced that the method is valid in the field of signals processing and fault diagnosis in rotating machinery by simulation and real signal analysis.Chapter three discussed the origin of EMD method’s end effect, then studied the direct, time-series based and neuron network based data extention technologys to weaken the end effect. At last, compared the three data extention technologys merits.The time-series based and neuron network based data extention methods were proposed.Chapter four studied the improvement of the EMD algorithm al efficiency and precision.Reserthed the effect of spline interpolation’s end condition to the EMD algorithm’s precision , then discussed the low-order and propoed high-order spline interpolation based EMD algorithm and their effects. Under the study of simple average EMD algorithm.deduced the self-tune time-variety based EMD algorithm and investigated the effect of the algorithm. At last, the valid data based fast EMD algorithm was proposed.Chaper five aimed to the requirement of the EMD based time-frequency analysis in the field of rotating machinery signal analysis and fault diagnosis.studied the method of rotatory speed prediction to improve the precision of complete period sampling of vibration signals in rotating machinery, then, discussed the phenomenon of mode fission in noise signal anlysis by EMD and provided the filter method to weaken the phenomenon. After then, found a process to analysis the signals using EMD-based method in the field of rotating machinery signal analysis and faultdiagnosis. At last, EMD-based filter techonolgy was researched.Chaper six compared the EMD-based time-frequency analysis method to the typical’s :short time fourier analysis, Wigner distribution and wavelet anlysis,the result showed that the EMD-based’s is better than the others.To weaken the crossing interfere in multi-components signal in Wigner distribution the EMD-based Wigner distribution was proposed and the study result showed that the method is valid.Chapter seven built a experiment system at the base of rotor test bed, discussed characteristics of several typical fault in rotating m更多还原