基于局部均值分解的齿轮故障诊断方法

发布时间：2018-06-17 09:09

本文选题：齿轮故障诊断 + 局部均值分解　；参考：《湖南大学》2012年硕士论文

【摘要】：齿轮是机械设备中一种常见的通用零部件，它通常负责连接和传递动力。由于工作环境恶劣等因素，齿轮极易发生故障而直接影响到相关机械设备的运行状态，因此，对齿轮进行故障监测和诊断具有重要意义。齿轮故障诊断的核心和关键就是提取齿轮故障特征，对齿轮的故障位置和损坏程度作出判断，而齿轮振动信号与齿轮工作状态密切相关，其中往往包含大量的故障信息。本文正是针对齿轮故障振动信号的非平稳性及其多为多分量的调制信号之和等特性，将局部均值分解（Local mean decomposition，简称LMD）引入齿轮故障诊断，进一步将LMD方法与能量算子解调、循环频率解调、谱峭度方法相结合应用于齿轮故障诊断，并在LMD的基础上提出了局部特征尺度分解（Local characteristic-scale decomposition，简称LCD）方法。本文的主要研究内容如下： 1、针对齿轮故障振动信号大多数为若干的调幅调频信号之和这一特点，将LMD方法应用于齿轮故障诊断中。LMD是一种新的自适应时频分析方法，它能将复杂的非平稳多分量调幅调频信号从高频到低频分解为有限个单分量的调幅调频信号之和，因此非常适合齿轮故障信号分析。通过仿真信号对比分析和实际的齿轮故障信号分析可知，LMD方法具有更好的自适应性和时频聚集性，能够更精确地获得信号的瞬时频率和瞬时幅值，能够得到更加清晰和完整的时频分布。 2、针对齿轮故障信号的调制特性，且从调制信息中通常能提取出故障特征这一特点，提出了基于LMD的能量算子解调和循环频率解调齿轮故障诊断方法。该方法先利用LMD将信号分解为一系列的乘积函数（Product function，简称PF），然后分别利用能量算子解调和循环频率解调获得相关PF分量的幅值调制信息和相位调制信息，从而提取出故障特征进行故障诊断。将该方法运用于齿轮振动信号故障诊断中，并和LMD直接法进行对比分析，证明了该方法的优越性。 3、针对齿轮故障振动信号在解调分析之前一般要通过滤波来确定包含故障信息的最佳频段，而滤波参数又无法准确确定，只能依靠历史数据和人工经验这一问题，，提出了基于LMD的谱峭度齿轮故障诊断方法。该方法在LMD时频分析的基础上获得信号峭度图，根据最大峭度原则在峭度图上选取最佳滤波频段从而获得最佳滤波参数对原始信号进行滤波，再对滤波后的信号进行包络解调分析提取出故障特征。实验结果证明，该方法具有有效性。 4、针对LMD计算量大，运算速度慢，不适于在线监测的特点，采用基于极值点的局部特征尺度参数，定义了另一种瞬时频率具有物理意义的单分量信号——内禀尺度分量（Intrinsic scale component，简称ISC），在此基础上提出了LCD方法。LCD方法也是一种自适应的分解方法，能够将多分量的信号分解为单分量信号之和，通过分别与EMD和LMD方法对比证明其具有不会产生包络误差，且运算速度较快等优点。同时，本文还将LCD方法成功地运用于齿轮箱故障诊断。
[Abstract]:Gear is a common common component in mechanical equipment. It is usually responsible for connecting and transmitting power. Because of bad working environment and other factors, the gear is easily malfunction and affects the operating state of the related machinery directly. Therefore, it is of great significance to monitor and diagnose the gear fault.
The core and key of gear fault diagnosis is to extract the feature of gear fault and judge the fault position and damage degree of the gear, and the gear vibration signal is closely related to the working state of the gear, which often contains a large number of fault information. This paper is aimed at the non stationarity of the vibration signal of the gear fault and its multi component modulation. The Local mean decomposition (LMD) is introduced into the gear fault diagnosis, and the LMD method is combined with the energy operator demodulation, the cyclic frequency demodulation, and the spectral kurtosis method is applied to the gear fault diagnosis, and the local feature scale decomposition (Local characteristic-sc) is put forward on the basis of LMD. Ale decomposition, referred to as LCD). The main contents of this paper are as follows:
1, in view of the characteristic that most of the gear fault vibration signals are the sum of amplitude modulation and frequency modulation signals, the LMD method is applied to the gear fault diagnosis..LMD is a new adaptive time-frequency analysis method. It can decompose the complex nonstationary multicomponent FM signal from high frequency to low frequency into a limited single component FM signal. It is very suitable for the analysis of the gear fault signal. Through the analysis of the simulation signal contrast analysis and the actual gear fault signal analysis, it is known that the LMD method has better adaptability and time frequency aggregation. It can get the instantaneous frequency and instantaneous amplitude of the signal more accurately, and can get a clearer and complete time frequency distribution.
2, in view of the modulation characteristics of the gear fault signal and the feature that usually can extract the fault feature from the modulation information, a method of gear fault diagnosis based on LMD's Energy Operator Demodulation and cyclic frequency demodulation is proposed. The method first uses LMD to decompose the signal into a series of product functions (Product function, for short, PF), and then profit separately. The amplitude modulation information and phase modulation information of the related PF components are obtained by using the Energy Operator Demodulation and the cyclic frequency demodulation. The fault features are extracted and the fault diagnosis is extracted. The method is applied to the fault diagnosis of the gear vibration signal and is compared with the LMD direct method, and the superiority of the method is proved.
3, before the demodulation analysis of the gear fault vibration signal, the best frequency band which contains the fault information is usually determined by filtering, and the filter parameters can not be accurately determined. It can only rely on the problem of historical data and artificial experience. The method of diagnosis of the spectral kurtosis of gear barrier based on LMD is put forward. This method is based on the basis of LMD time-frequency analysis. On the basis of the maximum kurtosis principle, the best filter band is selected on the kurtosis map to obtain the best filter parameters to filter the original signal, and then the signal is extracted and analyzed by the envelope demodulation analysis. The experimental results show that the method is effective.
4, in view of the characteristics of LMD with large computation, slow computing speed and unsuitable for on-line monitoring, a single component signal, an intrinsic scale component (Intrinsic scale component, simply called ISC), is defined by using local feature scale parameters based on extreme points, and based on this, a LCD method.LCD method is also proposed. An adaptive decomposition method can be used to decompose the multicomponent signals into the sum of the single component signals. By comparing with the EMD and LMD methods, it is proved that it has the advantages of no envelope error and faster operation speed. At the same time, this paper also successfully applied the LCD method to the gear box fault diagnosis.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【参考文献】