基于局部均值分解的旋转机械故障诊断技术研究

发布时间：2019-05-24 12:02

【摘要】：旋转机械在现代化机械设备占很大的比重,为其进行状态监测和故障诊断已经成为重要的研究课题。在故障诊断中,最关键的问题是提取故障特征信息和故障类型识别部分。局部均值分解(Local Mean Decomposition,LMD)时频分析方法在分析机械振动信号时拥有很多优越之处,被广泛应用到旋转机械故障特征提取当中。然而,局部均值分解仍存在一些不足之处有待改进。本文重点研究了LMD时频分析方法的不足之处及改进办法,并研究了故障类型的模式识别方法和故障诊断系统的开发应用。首先,针对LMD存在的端点效应问题,分析其产生原因,并提出一种改进的方法——最大相似系数法,通过仿真和实验研究的对比分析,验证方法的有效性。其次,针对进行旋转机械故障特征提取时存在的微弱高频信号难以提取的问题,以及LMD分解结果存在的虚假频率问题,提出基于微分局部均值分解(Differential Local Mean Decomposition,DLMD)的故障诊断方法。采用仿真研究,验证该方法的可行性和有效性。并通过实际工程中的复合故障信号进行研究分析,验证该方法在实际应用中的可行性。然后,针对旋转机械故障类型的模式识别方面,将LMD方法与样本熵、模糊聚类结合,提出基于局部均值分解、样本熵和模糊聚类的旋转机械故障诊断方法。该方法首先对旋转机械振动信号进行LMD分解,分解得到的乘积函数(Product Function,PF)求取样本熵,以此作为特征向量来建立模糊矩阵,进行模糊聚类分析和模式识别,实现故障的分类和诊断。最后,结合MATLAB和Lab VIEW开发旋转机械故障诊断平台,应用Lab VIEW图形化编程语言的优势和MATLAB强大的数据处理能力,进行机械故障诊断界面的设计和故障数据的处理。
[Abstract]:Rotating machinery accounts for a large proportion of modern machinery and equipment, so it has become an important research topic to carry out condition monitoring and fault diagnosis for it. In fault diagnosis, the key problem is to extract fault feature information and fault type identification. Local mean decomposition (Local Mean Decomposition,LMD) time-frequency analysis method has many advantages in the analysis of mechanical vibration signals, and is widely used in fault feature extraction of rotating machinery. However, there are still some shortcomings in local mean decomposition that need to be improved. In this paper, the shortcomings and improvement of LMD time-frequency analysis method are studied, and the pattern recognition method of fault type and the development and application of fault diagnosis system are studied. Firstly, aiming at the problem of endpoint effect in LMD, the causes are analyzed, and an improved method, the maximum similarity coefficient method, is proposed to verify the effectiveness of the method through the comparative analysis of simulation and experimental research. Secondly, in order to solve the problem that it is difficult to extract weak high frequency signals in rotating machinery fault feature extraction, and the false frequency problem of LMD decomposition results, a differential local mean decomposition (Differential Local Mean Decomposition, is proposed. DLMD) fault diagnosis method. The feasibility and effectiveness of the method are verified by simulation. Through the research and analysis of the compound fault signal in practical engineering, the feasibility of this method in practical application is verified. Then, aiming at the pattern recognition of rotating machinery fault types, the LMD method is combined with sample entropy and fuzzy clustering, and a rotating machinery fault diagnosis method based on local mean decomposition, sample entropy and fuzzy clustering is proposed. In this method, the vibration signal of rotating machinery is decomposed by LMD, and the sample entropy is obtained by decomposing the product function (Product Function,PF), which is used as the eigenvector to establish the fuzzy matrix, and the fuzzy clustering analysis and pattern recognition are carried out. Realize the classification and diagnosis of faults. Finally, combined with MATLAB and Lab VIEW, the fault diagnosis platform of rotating machinery is developed, and the design of mechanical fault diagnosis interface and the processing of fault data are carried out by using the advantages of Lab VIEW graphical programming language and the powerful data processing ability of MATLAB.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TH165.3

【参考文献】