【作者】 邹岩崑；

【导师】 马孝江；

【作者基本信息】 大连理工大学 ， 机械电子工程， 2004， 博士

【摘要】 近年来,设备监测诊断中非平稳信号的分析一直是一个复杂而有意义的研究课题,局域波分析方法在几年的发展过程中,逐渐成为分析非平稳信号的有效方法之一。本文在总结前人研究成果基础上,结合工程实际需要,对局域波分析方法作了进一步的研究和发展,并以此为基础,在机械设备的诊断方法上进行了探索性的研究。 针对设备监测诊断中存在的非平稳问题,引入了局域波分析方法。该方法从信号瞬时频率的角度出发,将非平稳时变信号分解成为有限个局域波分量,每一个分量描述了时变信号中不同频率和尺度范围的固有振动模式,瞬时频率可以在每一个分量中随处定义。经过Hilbert变换得到局域波时频谱能够同时提供时域和频域的信息。 针对局域波分析方法目前的研究状况,分析了该方法中存在的问题:分解方法中的边界问题、分量的瞬时频率估计以及如何合理选择采样频率等问题。提出了基于包络均值法的改进算法;分析并验证了利用相位差分法进行瞬时频率估计的有效性和精确性;给出了进行局域波分析,信号采样频率的选择标准。 在此基础上,研究了局域波分析的频率多分辨特性及分解的尺度滤波特性,给出了不同时频谱图的频率分辨率的计算式和滤波器的表达式。通过对加噪信号的有效去噪检验了尺度滤波的可行性。提出了基于局域波时频谱的边界谱分析,以及频带能量分析方法,并运用这些方法对旋转机械和往复机械的不同故障进行了有效的诊断。 以一维局域波分解方法为基础,研究了二维局域波分解方法,对边界处的极值判断方法作了改进,有效的解决了插值曲面的边界摆动。制定了合理的筛选停止准则和分解停止准则。通过对图像去噪,检验了二维分解对于提取图像局域细节信息的有效性。同时,提出了基于二维局域波分解的图像诊断方法,通过对时频灰度图像的二维分解提取表征故障信息的图像细节部分,有效地实现了故障的特征提取。 以上研究工作在一定程度上丰富和完善了局域波分析方法,诊断应用表明本文中提出的方法能够有效的识别故障,解决实际问题。更多还原

【Abstract】 In recent years, analysis of the non-stationary signal in equipment fault diagnosis, is always a complex and significant work. With the hard research, the local wave method is gradually turned into one of the effective method to solve non-stationary problems. In this paper, based on the former achievements linking to practical project, the local wave method is deeply developed and expanded. And the fault diagnosis method on mechanical equipment is researched on the basis of former work.Local wave method is imported especially for solving non-stationary problems of the equipment diagnosis. The method, which can decompose time varying signal into several local wave functions, is based on the instantaneous frequency of signals. Every function describes different frequency and scale of the inherence oscillation mode. Instantaneous frequency can be defined everywhere in every function. By Hilbert transform, local wave time-frequency spectrum can give enough information in both time domain and frequency domain.Aimed at the research status of local wave method, the problems in this method is analysed, such as the end effect, estimate of instantaneous frequency and how to choice the sampling frequency. The way is put forward for improvement on arithmetic based on envelope mean. The validity and accuracy of phase difference method is analysed and validated. The criterion of how to choice the sampling frequency is given in local wave method.Based on the former work, the characteristic of multi-distinguish frequency analysis of local wave method and time-scale filtering is researched. The expressions of the frequency distinguish ratio in the time-frequency spectrum is given. The formula of filter can also be found in this paper. The feasibility is proved by noise canceling of signals with noise. An approach, which can be called boundary spectrum and energy of frequency belt, is proposed based on the local wave spectrum. With these methods we can find the fault in rolling machine and reciprocating machine.Two-dimensional local wave decomposition is investigated on the basis of one dimensional decomposition method. A new way to obtain the local maxima and minima at boundary is brought forward. The boundary swing problem is effectively solved. The criterions for the sifting process and decompose process to stop is also given. An examination on image de-noising is carried out. It proves that detail information of image can be obtainedby two-dimensional local wave decomposition which. Image diagnosis method based on decomposition is put forward. Details of time-frequency grayer degree image with fault information can be picked up by two-dimensional decomposition. During this process, we may find the fault character.The research work mentioned above enhances the local wave method. It also proved that the methods which are introduced in the paper can effectively identify the fault and solve practical problems.更多还原