基于SVD与ELM的齿轮故障诊断技术研究

发布时间：2018-06-26 22:20

本文选题：齿轮箱 + 奇异值分解　；参考：《沈阳理工大学》2017年硕士论文

【摘要】：齿轮箱作为旋转机械中重要部件之一,其能否正常运行对整台甚至整套机械设备的正常工作和使用产生具有重大影响。因为齿轮箱特殊的工作环境,容易引发多个组件、零件故障,所以及时对齿轮箱振动数据进行检测,准确识别出齿轮箱运行过程中的故障并且及时更换故障零件,对提高旋转机械整体运行可靠性具有重大意义。本文叙述集合经验模态分解、奇异值分解分解及极限学习机齿轮箱故障诊断和识别领域的使用情况。通过对齿轮箱故障诊断和识别模型的研究,提出了EEM D-SVD故障特征提取和EEMD-ELM故障识别的方法。对振动数据做EEMD分解得到一系列固有模态分量,对其进行有效的筛选并且重构,对重构的信号构造Ha nkel矩阵,再通过SVD对矩阵做正交分解,利用奇异值差分谱来选择奇异值进行SVD重构,由此实现对故障特征的提取。然后提取与原信号相关较大的IMF能量值作为故障识别模型的输入向量,建立齿轮箱ELM故障分类模型。最后,通过齿轮箱实验验证EEMD-SVD方法可以高效、准确提取齿轮和轴承故障特征。对比ELM与SVM的识别结果,ELM比SVM有更快的运行速度、更高的分类精度。
[Abstract]:Gearbox is one of the important parts in rotating machinery. Its normal operation has great influence on the normal operation and use of the whole machine and even the whole set of mechanical equipment. Because of the special working environment of the gearbox, it is easy to cause many components and parts faults, so the vibration data of the gearbox can be detected in time, the faults in the gearbox running process can be accurately identified and the faulty parts can be replaced in time. It is of great significance to improve the overall operating reliability of rotating machinery. This paper describes the application of set empirical mode decomposition singular value decomposition and extreme learning machine gearbox fault diagnosis and identification. Based on the research of gearbox fault diagnosis and identification model, a method of EEM D-SVD fault feature extraction and EEMD-ELM fault identification is proposed. A series of inherent modal components are obtained by EEMD decomposition of vibration data, which are effectively screened and reconstructed. Ha nkel matrix is constructed for the reconstructed signal, and then orthogonal decomposition is made for the matrix. Singular value difference spectrum is used to select singular value for SVD reconstruction, and thus fault feature extraction is realized. Then the IMF energy which is related to the original signal is extracted as the input vector of the fault identification model and the ELM fault classification model of the gearbox is established. Finally, the EEMD-SVD method is proved to be efficient and accurate to extract gear and bearing fault features by gearbox experiments. Compared with the recognition results of ELM and SVM, ELM has faster running speed and higher classification accuracy than SVM.
【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH132.41;TH17

【参考文献】