基于EEMD与MED的冲击信号自适应故障特征提取方法

发布时间：2017-12-27 14:27

本文关键词：基于EEMD与MED的冲击信号自适应故障特征提取方法　出处：《上海大学》2016年博士论文　论文类型：学位论文

【摘要】：振动信号是机械系统行为的最直观反映,它携带大量设备状态信息,可以及时地反映机械系统的动态变化过程。因此,振动信号被广泛地应用于状态监测与故障诊断。特征提取是故障诊断技术的关键环节,为了从复杂的非平稳、非线性信号中有效地提取冲击故障的特征信息,本文深入研究了EEMD与MED两种自适应故障特征提取方法,主要研究工作如下:(1)针对EEMD方法均值曲线构造过程中易出现过包络的问题,本文根据极值点存在的极值冗余或伪极值点现象,提出局部极值包络均值曲线构造方法。继而,将该方法替换EEMD原均值曲线构造方法,形成局部极值包络EEMD方法,通过与EEMD方法对比分析,发现局部极值包络EEMD方法在消除模态混淆方面可取得更好的实验效果。(2)针对如何从EEMD分解出的多个IMF中选取出反映冲击故障特征信息的问题,本文通过定义双值特征参数的“从属”和“不从属”逻辑关系,提出选取EEMD冲击故障特征信号的双值区间准则。根据峭度对冲击信号敏感的特点,将最大谱峭度频带区间作为双值特征参数,结合双值区间准则对冲击故障特征信号进行筛选。仿真实验及实测冲击信号分析表明,双值区间准则可方便地选取出EEMD冲击故障特征信号。(3)由于EEMD冲击故障特征信号在本质上仍然为冲击脉冲信号与传递函数卷积计算的结果,为了消除传递函数的干扰,对冲击脉冲信号进行自适应特征提取,本文将MED方法应用于EEMD冲击故障特征信号。重点分析了影响MED提取冲击脉冲信号效果的主要因素,提出EEMD-MED和形态学滤波的冲击脉冲信号增强方法,仿真实验和实例研究均表明该方法可有效地增强MED冲击脉冲信号的提取效果。(4)针对MED冲击脉冲信号在进行包络解调时会出现大量高次谐波成分的问题,本文利用冲击脉冲信号的卷积性质,将脉冲信号与高斯核函数进行卷积运算并对卷积结果进行傅里叶变换,提出冲击脉冲信号的解调思路。根据短时傅里叶变换的计算过程是将原信号与高斯核函数进行卷积,且其计算过程与MED脉冲信号的解调思路一致,本文提出增强MED短时傅里叶变换的脊线解调故障诊断方法。通过滚动轴承实例分析,并与EEMD包络解调方法及谱峭度包络解调方法进行对比,验证了该方法的有效性。
[Abstract]:The vibration signal is the most intuitionistic reflection of the mechanical system behavior. It carries a large number of equipment state information, which can reflect the dynamic change process of the mechanical system in time. Therefore, vibration signals are widely used in state monitoring and fault diagnosis. Feature extraction is the key technology of fault diagnosis, feature information for non-stationary and nonlinear signals effectively extract fault impact from the complex, this paper studies EEMD and MED two kinds of adaptive fault feature extraction method, the main research work is as follows: (1) easy to appear for process of the EEMD method in the mean curve structure the envelope of the problem, according to the extreme value redundancy limit exists or pseudo extremum phenomenon, put forward the local extremum of mean envelope curve construction method. Then, this method replaces the construction method of EEMD's original mean curve to form the local extreme envelope EEMD method. By comparing and analyzing with EEMD method, it is found that the local extreme value envelope EEMD method can get better experimental results in eliminating modal confusion. (2) for the problem of how to extract the characteristic information of impact fault from multiple IMF decomposed by EEMD, by defining the logical relationship between "subordinate" and "non subordinate" of two valued feature parameters, we propose a two valued interval rule that selects EEMD impulse fault characteristic signal. According to the sensitivity of kurtosis to the impact signal, the maximum spectral kurtosis band interval is used as a dual characteristic parameter, and the double fault interval criterion is applied to screen the characteristic signal of impact fault. The simulation experiment and the measured impact signal analysis show that the dual value interval criterion can easily select the EEMD impact fault characteristic signal. (3) the impact of the EEMD fault characteristic signal in essence is still the impulse signal and the transfer function of the convolution calculation results, in order to eliminate the interference of the transfer function, the impulse signal adaptive feature extraction, the MED method is applied to EEMD impact fault characteristic signal. The main factors that affect the effect of MED on extracting impulse signal are analyzed. A method of enhancing impulse pulse signal based on EEMD-MED and morphological filtering is proposed. Simulation experiments and case studies show that the method can effectively enhance the extraction effect of MED impulse signal. (4) the impact of the MED pulse signal will appear a large number of harmonics in the envelope demodulation problem, convolution of impulse signals using, convolution and Fourier transform of the convolution result pulse signal with Gauss kernel function, put forward ideas on pulse signal demodulation at. According to the computation process of short-time Fourier transform, the original signal is convoluted with Gauss kernel function, and its computation process is consistent with the demodulation idea of MED pulse signal. In this paper, a MED fault diagnosis method based on ridge transform demodulation is proposed. The validity of the method is verified by comparing the EEMD envelope demodulation method and the spectral kurtosis envelope demodulation method through the analysis of the rolling bearing case.
【学位授予单位】：上海大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN911.7

【参考文献】