多维多尺度齿轮故障特征提取与分类的研究

发布时间：2019-02-26 14:10

【摘要】：齿轮故障诊断中关键的两个部分就是信号特征的提取和模式识别。本课题综合分析了国内外对齿轮故障诊断技术的研究现状，在齿轮的振动机理及齿轮故障的振动信号特征的基础上，提出了一种多维多尺度特征提取算法和一种基于模糊C聚类的故障分类算法。具体的研究内容如下： (1)在总结了齿轮不同状态下的振动信号特征的基础上，采用仿真信号，模拟了齿轮正常、磨损和断齿的振动信号特征，更为直观地掌握了不同故障的振动信号特征，同时也为故障分类提供了仿真信号。 (2)提出了一种多维多尺度特征提取算法。待处理信号首先经过奇异值分解，将一维信号重构成多维信号，凸显出更多的特征信息；然后采用奇异值分解，分解成不同尺度的调幅——调频信号，取能量高的乘积函数求和重构；最后，，利用形态差值滤波器提取出特征信息，通过仿真实验和齿轮故障模拟实验验证了该方法的有效性。 (3)在特征提取的基础上，将模糊C聚类算法引入到故障分类中，从提取的特征信息中选择合适的特征量，通过仿真实验和齿轮故障模拟实验验证了模糊C聚类算法是一种有效的故障分类算法。本课题提出了一种有效的故障特征提取算法和故障分类算法，也验证其有效性，为齿轮故障诊断提供一种有效、准确的方法。
[Abstract]:Two key parts in gear fault diagnosis are signal feature extraction and pattern recognition. This paper comprehensively analyzes the research status of gear fault diagnosis technology at home and abroad, based on the vibration mechanism of gear and the characteristics of vibration signal of gear fault. A multi-dimensional and multi-scale feature extraction algorithm and a fault classification algorithm based on fuzzy C clustering are proposed. The specific research contents are as follows: (1) on the basis of summarizing the vibration signal characteristics of gear under different states, the vibration signal characteristics of gear normal, wear and broken teeth are simulated by using simulation signal. The vibration signal characteristics of different faults are grasped more intuitively, and the simulation signals are also provided for fault classification. (2) A multi-dimensional and multi-scale feature extraction algorithm is proposed. After singular value decomposition (SVD), the one-dimensional signal is reconstituted into multi-dimensional signal, which highlights more characteristic information. Then singular value decomposition (SVD) is used to decompose the amplitude modulation-FM signal of different scales, and the product function of high energy is taken to sum and reconstruct. Finally, the morphological difference filter is used to extract the feature information, and the effectiveness of the method is verified by simulation experiment and gear fault simulation experiment. (3) on the basis of feature extraction, fuzzy C clustering algorithm is introduced into fault classification, and the appropriate feature quantity is selected from the extracted feature information. The fuzzy C clustering algorithm is proved to be an effective fault classification algorithm by means of simulation experiments and gear fault simulation experiments. In this paper, an effective fault feature extraction algorithm and fault classification algorithm are proposed, and the validity of the algorithm is verified, which provides an effective and accurate method for gear fault diagnosis.
【学位授予单位】：武汉科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TH165.3

【参考文献】