基于EMD近似熵和LS-SVM的齿轮箱故障诊断研究

发布时间：2019-05-22 11:23

【摘要】：齿轮箱是机械设备中重要的传动部件，，对齿轮箱进行故障诊断研究有着非常现实的意义。本文将EMD(Empirical Mode Decomposition)近似熵和LSSVM(Least SquareSupport Vector Machine)相结合来实现对齿轮箱的故障诊断。 EMD方法对信号处理具有良好的局域化特性，同时针对非线性、非平稳的信号有着非常好的分解效果。近似熵在表征信号动力学特性方面能包含更多的信息，对提取信号的故障特征有着先天的优势。LSSVM是针对SVM(Support Vector Machine)作为分类算法中存在着运行时间过长和计算量过大的弊端做出的改进和变形，实验证明LSSVM在齿轮箱故障诊断中能准确而快速的实现故障识别。本文首先阐述了齿轮箱故障诊断的意义、目的及国内外研究现状，同时对目前的故障诊断技术进行了概述。其次介绍了齿轮箱振动机理和故障类型，接着重点研究了EMD方法在分解信号中存在着端点效应这样的弊端，提出了镜像延拓以及在信号序列上进行了加窗函数相结合的办法对EMD方法的改进。实验证明经过改进之后的EMD方法在信号分解上取得了非常好的效果。然后应用EMD和近似熵相结合的方法完成了对齿轮箱故障特征的提取，分别从理论上和具体实验中对SVM和LSSVM进行了对比，突出LSSVM在故障识别上的优势。最后利用改进后的EMD方法结合近似熵完成对故障特征的提取，利用LSSVM对提取的故障特征进行识别，然后通过对比其他几种不同的故障诊断方法，表明EMD近似熵和LSSVM能够提高齿轮箱故障诊断的准确率和效率。
[Abstract]:Gearbox is an important transmission component in mechanical equipment, so it is of great practical significance to study the fault diagnosis of gearbox. In this paper, EMD (Empirical Mode Decomposition) approximate entropy and LSSVM (Least SquareSupport Vector Machine) are combined to realize the fault diagnosis of gearbox. The EMD method has good localization characteristics for signal processing, and has a very good decomposition effect for nonlinear and non-stationary signals. Approximate entropy can contain more information in describing the dynamic characteristics of the signal. LSSVM has inherent advantages in extracting fault features of signals. LSSVM is an improvement and deformation for the disadvantages of SVM (Support Vector Machine) as a classification algorithm, which has too long running time and too much computation. The experimental results show that LSSVM can realize fault identification accurately and quickly in gearbox fault diagnosis. This paper first expounds the significance, purpose and research status of gearbox fault diagnosis at home and abroad, and summarizes the current fault diagnosis technology. Secondly, the vibration mechanism and fault type of gearbox are introduced, and then the end-point effect of EMD method in decomposing signal is studied. The mirror image extension and the combination of windowing function on the signal sequence are proposed to improve the EMD method. The experimental results show that the improved EMD method has achieved very good results in signal decomposition. Then the fault features of gearbox are extracted by using the method of EMD and approximate entropy. SVM and LSSVM are compared theoretically and in specific experiments to highlight the advantages of LSSVM in fault identification. Finally, the improved EMD method combined with approximate entropy is used to extract the fault features, LSSVM is used to identify the extracted fault features, and then several other fault diagnosis methods are compared. It is shown that EMD approximate entropy and LSSVM can improve the accuracy and efficiency of gearbox fault diagnosis.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TH165.3;TH132.41

【参考文献】