【作者】 鞠萍华；

【导师】 秦树人；

【作者基本信息】 重庆大学 ， 机械电子工程， 2010， 博士

【摘要】 旋转机械设备运行时,其振动信号一般是非常复杂的,各种振动因素综合起来,得到的机械系统的振动信号必然是非平稳非线性的多分量信号,不同的非平稳特性也预示着不同的机械故障形式。目前,时频分析技术在机械故障特征提取中取得了广泛的应用,但如何更准确地提取信号中的早期(微弱)故障特征仍然是该学科研究的热点和难点。在此背景下,深入研究了旋转机械及其早期故障的基本特点、多分辨分析技术、Hilbert-Huang变换、多分辨EMD方法、时频域平均技术、循环平稳理论以及基于局域均值分解的时频分析方法等信号分析处理技术,对旋转机械故障信号的消噪、解调、特征放大技术、时频表示等特征提取技术进行了深入的研究。首先对目前常用的时频分析工具进行了回顾和比较,对时频分析方法的发展历程及研究现状进行了简要介绍,对旋转机械的早期故障信号的特点和常用诊断方法进行了分析。旋转机械的振动信号由于具有循环平稳的特点,因此将其变换到平稳的循环域对信号进行平均处理能较好的提取特征信号而滤除干扰。频域平均不但能较好的分辨信号中的多个周期成分,实现设备的多故障检测,而且能使得信号的早期(或微弱)故障信号的能量被加强,特征更突出。提出了首先用多分辨分析的EMD方法获得信号的瞬时频率,再对信号进行频域平均处理的方法,并用齿轮传动故障检测实验装置进行实验,结果表明了该方法的有效性和可行性。时域平均对提取与回转频率直接相关的周期信号,是较为有效的信号分析和预处理方法,但当旋转机械运行不平稳时,该方法将失效。首先对循环平稳理论和时域平均方法进行了分析,为提高时域平均的诊断精度和适用范围,将齿轮箱输入轴瞬时速度的三次曲线拟合和信号重采样相结合,提出了非同步特征信号的时域平均提取方法。实验结果表明了该方法的有效性。目前,旋转机械故障振动信号的解调方法有多种,这些方法都只适用于单分量的AM-FM信号。局域均值分解是一种新出现的时频分析方法,比较EMD方法而言具有许多独特的优点。在介绍LMD方法的基础上,分析了直接计算瞬时频率的局限性,提出了将LMD方法与能量算子解调相结合多分量信号处理方法。由于LMD方法和EMD方法类似,亦是基于极值点来定义局域均值函数和局域包络函数,同样存在端点效应。在分析LMD的端点效应产生根源的基础上,提出了特征趋势正弦函数法实现数据延拓的端点效应处理方法。最后通过实验表明,基于LMD的能量算子解调法能有效地应用于旋转机械早期故障诊断中,所采用的端点效应处理方法简单有效。对虚拟仪器技术进行了研究。在上述理论研究成果的基础上,研制成功了齿轮箱早期故障诊断仪。该仪器兼具虚拟仪器和传统硬件化仪器的优点,并具有强大的信号分析能力,适合于科学实验和工程中的复杂信号分析。还通过大量的仿真实验和实际工程应用,对仪器功能的正确性和稳定性进行了验证。文章最后对本文工作进行了总结,并展望了下一步的研究方向。更多还原

【Abstract】 Vibration signals are usually very complex when the rotating mechanical equipment is running. After a variety of vibration factors are integrated, vibration signals of mechanical system that are gained are necessarily non-stationary non-linear multi-component signals, and different non-stationary characteristic also indicates different form of mechanical fault. At present, the time-frequency analysis technique is widely used in the mechanical fault feature extraction, but how to more accurately extract early (weak) fault features from signals is still a hot and difficult point of research of the discipline. Against this background,an intensive study is given to basic features of rotating machinery and early faults, multi-resolution analysis technique, Hilbert-Huang transform, multi-resolution EMD method, time and frequency domain average techniques, cyclostationarity theory, local mean decomposition-based time-frequency analysis method and other signal analysis and processing technologies, and to rotating machinery fault signal noise cancellation, demodulation, characteristic amplification technology, time-frequency representation and other feature extraction techniques.Conduct a review and compare of currently commonly-used time-frequency analysis tools, briefly introduce its development process and research status, and analyze characteristics and early diagnostic methods of rotating machinery early fault signals.Because of the cyclostationarity characteristic, vibration signals of rotating machinery are converted to a smooth cyclic domain to conduct an average treatment that can better extract characteristic signals to filter out interference. The time domain average is not only capable of better resolution of multiple periodic components to achieve multiple-fault detection in signals, and but is also capable of strengthening the energy of early (or weak) fault signals and making the characteristics more prominent. The EMD method of multi-resolution analysis, firstly bring forward firstly by me, is used to gain instantaneous frequency of signal; then conduct a frequency domain average treatment for signals, and use the gear drive fault detection experimental device to do experiments; results show that this method is effective and feasible.The time domain average is a more effective method of signal analysis and pretreatment method for extracting periodic signals directly related to the rotary frequency; but when the rotating machinery runs unsteadily, the method is out of work. In this article, I firstly analyze the cyclostationarity theory and the time domain average method, and in order to raise the time domain average diagnostic accuracy and extend its applicable scope, combine the fitting of cubic curve of instantaneous speed of gearbox input shaft with the re-sampling of signal to put forward time domain average extraction method of non-synchronous characteristic signals; experimental results show that the method is effective.Currently, there are multiple kinds of demodulation methods of rotating machinery fault vibration signals, but these methods are only applicable to single-component AM-FM signals. Compared with the EMD method, the local mean decomposition is a new emerging time-frequency analysis method, with many unique excellent features. In this article, I am based on the LMD method to analyze limitations of direct calculation of instantaneous frequency, and put forward a multi-component signal processing method that combines the LMD method with the energy operator demodulation. As the LMD method is similar to the EMD method, I am also based on the extreme value point to define the local mean function and the local envelop function, and similarly, the end effect existed. Based on analyzing the production source of LMD end effect, I bring forward an end effect treatment method that a data extension is achieved through the characteristic trend sine function method. At last, the experiment made by me shows that the LMD-based energy operator demodulation method can be effectively used in the early-fault diagnosis of rotating machinery, and that the end effect treatment method that is adopted is simple and effective.Virtual instrument technology are studied in this article. On the basis of the above-mentioned theoretical research results, a diagnosis instrument of early fault of gearbox is successfully developed. With excellent features of both virtual instrument and traditional hardware instrument as well as a powerful signal analysis capability, this instrument is applicable to scientific experiments and complex signal analysis in engineering. Accuracy and stability of instrument function are verified through a large number of simulation experiments and practical engineering applications.At the end of this article, I summarize the work done in this article, and give an outlook of next research direction.更多还原