基于MED-EMD和切片双谱的齿轮箱故障诊断研究

发布时间：2018-03-09 14:42

本文选题：齿轮箱　切入点：EMD　出处：《太原理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：在旋转机械中,齿轮箱是传递动力的主要部件,齿轮箱中的旋转部件如轴、轴承、齿轮等往往在复杂的变载荷环境下运行,这些部件因为疲劳损伤而产生微弱故障,而故障信号在强背景噪声下往往不易被察觉。如果任由微弱故障发展为严重故障,会造成严重的后果。因此选用恰当的降噪方法对早期故障特征提取意义重大。目前对于机械信号的处理多是忽略机械设备发生故障时所产生的非高斯非平稳振动信号,而传统的故障诊断方法又无法抑制高斯噪声对故障信号的影响,往往无法在故障初期提取到具有非平稳非高斯特性的微弱故障特征。针对齿轮箱的早期故障信号具有非线性、非平稳性、非高斯性且易受强背景噪声干扰的特点,本文提出了基于最小熵反褶积(MED)、经验模态分解(EMD)和切片双谱相结合的方法来提取微弱故障特征。主要研究内容如下:1)介绍了MED、EMD、互信息和切片双谱的各自的原理以及之间的联系和衔接,在此基础上运用MED-EMD切片双谱的方法分析了的仿真信号,对仿真信号经MED-EMD分解后的前三阶IMF分量进行了切片双谱分析后发现,在中低频段出现了载波频率1)1=70Hz,明显突出了调制频率1)2=200Hz及其边频簇、1)2的二倍频及其边频簇和载波频率1)9)=300Hz的边频簇。通过比较得到MED-EMD和切片双谱相结合的降噪效果在三种方法中是最优的,这验证了MED-EMD切片双谱方法应用到旋转设备故障诊断上是可行的。2)通过MED-EMD将原始信号降噪分解为多个本征模态函数(IMF),MED作为EMD的前置滤波器能够弥补强背景噪声下EMD分解的不足,选取和原始信号相关性强的IMF分量为有效分量并对其进行切片双谱分析,提取点蚀故障特征。切片双谱分析能够抑制高斯噪声对IMF分量的干扰,同时与传统的基于EMD分解的功率谱方法进行了结果的对比,验证了该方法的实用性和优越性。3)设计了具有冲击调制特点的仿真信号,通过对仿真信号的处理来完善算法;搭建齿轮箱试验台,采集并用文中方法分析了包含齿轮点蚀信息的振动信号,验证了基于MED-EMD和切片双谱的方法在应用到旋转设备故障诊断时的实用性;测试采集到的某型号风力发电机齿轮箱实际运行的故障信号,提取微弱故障特征并做出相应故障诊断,与开机检查的结果相匹配,得到了齿轮箱故障为中间轴靠电机侧的轴承内圈出现点蚀的正确结论,并与传统方法相对比,MED-EMD切片双谱的降噪明显提高了信噪比,表明该方法对微弱故障提取的有效性。
[Abstract]:In rotating machinery, gearbox is the main component of transmission power. The rotating parts such as shaft, bearing, gear and so on often run under complicated variable load environment, these parts are weak fault due to fatigue damage. However, the fault signal is often difficult to detect under strong background noise. If a weak fault is allowed to develop into a serious fault, Therefore, it is very important to select proper noise reduction methods for early fault feature extraction. At present, most of the mechanical signal processing is to ignore the non-stationary vibration signal of Gao Si produced when the mechanical equipment is in trouble. However, the traditional fault diagnosis method can not restrain the influence of Gao Si noise on the fault signal, and it is often unable to extract the weak fault characteristics with non-stationary non-#china_person1# characteristics at the early stage of the fault. The early fault signal of the gearbox is nonlinear. Characteristic of non-stationary, non-#china_person0# and easily disturbed by strong background noise, In this paper, a method based on minimum entropy deconvolution (MED), empirical mode decomposition (EMD) and slice bispectrum is proposed to extract weak fault features. The main research contents are as follows: (1) the principles of MED EMD, mutual information and slice bispectrum are introduced. And the connection and convergence between them, On this basis, the simulation signal is analyzed by using MED-EMD slice bispectrum method. The first three IMF components of the simulation signal decomposed by MED-EMD are analyzed by slice bispectrum analysis, and it is found that the first three IMF components of the simulation signal are decomposed by MED-EMD. In the middle and low frequency band, the carrier frequency 1T 1n 70 Hz appears, which highlights the double frequency of modulation frequency 1kW 200Hz and its edge frequency cluster, and the side frequency cluster and the edge frequency cluster of the carrier frequency 1t 9 / 300Hz. By comparison, the noise reduction effect of the combination of MED-EMD and slice bispectral spectrum is obtained. The three methods are optimal. This proves that it is feasible to apply MED-EMD slice bispectral method to the fault diagnosis of rotating equipment. 2) the original signal is decomposed into multiple intrinsic mode functions by MED-EMD. As a prefilter of EMD, it can make up for the deficiency of EMD decomposition under strong background noise. The IMF component with strong correlation with the original signal is selected as the effective component and analyzed by slice bispectrum to extract the pitting fault features. The slice bispectrum analysis can suppress the interference of Gao Si noise to the IMF component. At the same time, compared with the traditional power spectrum method based on EMD decomposition, the practicability and superiority of the method are verified. 3) the simulation signal with the characteristic of impulse modulation is designed, and the algorithm is improved by processing the simulation signal. The vibration signals containing pitting corrosion information of gears are collected and analyzed with the method of gearbox test bed. The practicability of the method based on MED-EMD and slice bispectrum is verified when it is applied to the fault diagnosis of rotating equipment. The fault signals collected from the gearbox of a certain type wind turbine are tested, the weak fault features are extracted and the corresponding fault diagnosis is made, which matches the results of the boot check. The correct conclusion that the inner ring of the bearing on the motor side of the middle shaft of the gearbox fault appears pitting corrosion is obtained. Compared with the traditional method, the noise reduction of the MED-EMD slice bispectrum obviously improves the signal-to-noise ratio (SNR), which shows that the method is effective for weak fault extraction.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH132.41

【参考文献】