旋转机械振动信号特性提取技术研究

发布时间：2018-11-28 19:35

【摘要】：旋转机械设备在工作时的振动参数信号，特别在设备存在故障的条件下，绝大多数都是非平稳信号，其信号特征中频率是随时间而变化的，如果只在时域或频域中分析是远远是不够的，需掌握非平稳信号在频率—时间尺度中幅值或能量分布情况。故障特征提取过程是旋转机械故障诊断中的最关键且重要的问题，考虑旋转机械系统非平稳信号特征，对比其他时频分析方法，本文对现今非平稳信号故障特征提取方法研究热点—基于希尔伯特黄变换时频分析方法进行了深入的研究，总结经验模式分解的算法存在端点效应和模态混叠的问题。针对旋转机械现场信号往往参杂大量随机噪声和脉冲干扰问题，对奇异值分解和形态滤波理论进行研究，结合奇异值分解可有效消除随机噪声和形态滤波能较好抑制脉冲干扰特点，提出了奇异形态滤波去噪方法，该方法可有效消除现场测量中参杂的随机噪声和脉冲干扰，避免经验分解的模态混叠现象。针对经验模式分解在实际使用中存在端点效应和模态混叠等问题，提出了适用于旋转机械非平稳信号的微弱故障特征提取方法—集总极值域均值分解算法，该方法可有效解决经验模式分解算法的局限性。最后，采用QPZZ-II旋转机械振动故障模拟实验平台，模拟滚动轴承故障形式，分别模拟滚动轴承的外圈损失故障、内圈损伤故障和滚动体损伤故障进行，，对本文提出的方法进行试验验证，试验结果表明，本文方法对强噪声非平稳振动信号能够有效提取出故障特征频率。
[Abstract]:The vibration parameter signals of rotating machinery and equipment, especially under the condition of equipment failure, are mostly non-stationary signals, and the frequency of the signals is changed with time. If it is not enough to analyze only in time domain or frequency domain, it is necessary to master the amplitude or energy distribution of non-stationary signal in frequency-time scale. Fault feature extraction is the most critical and important problem in rotating machinery fault diagnosis. Considering the non-stationary signal characteristics of rotating machinery system, compared with other time-frequency analysis methods, fault feature extraction process is the most important problem in rotating machinery fault diagnosis. In this paper, the current research focus of fault feature extraction for non-stationary signals, time-frequency analysis method based on Hilbert-Huang transform, is deeply studied, and the problem of endpoint effect and modal aliasing in empirical mode decomposition algorithm is summarized. The singular value decomposition (SVD) and morphological filtering theory are studied to solve the problem of random noise and pulse interference in rotating machinery field signals. Combined with singular value decomposition (SVD) can effectively eliminate random noise and morphological filter can better suppress the characteristics of pulse interference, a singular morphological filter denoising method is proposed, which can effectively eliminate random noise and pulse interference in field measurement. The mode aliasing phenomenon of empirical decomposition is avoided. In order to solve the problem of endpoint effect and modal aliasing in practical application of empirical mode decomposition, a new method for extracting weak fault feature of non-stationary signals of rotating machinery, called lumped Polar mean decomposition algorithm, is proposed. This method can effectively solve the limitation of empirical mode decomposition algorithm. Finally, QPZZ-II rotating machinery vibration simulation experiment platform is used to simulate the fault form of rolling bearing, which simulates the outer ring loss fault, inner ring damage fault and rolling body damage fault of rolling bearing, respectively. The experimental results show that the proposed method can extract the fault characteristic frequency effectively for the strong noise non-stationary vibration signal.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.7

【参考文献】