自适应变分模态分解的齿轮箱故障诊断研究

强噪环境下,复合故障特征提取难度更大,VMD(Variational Mode Decomposition)被大量应用于齿轮箱故障诊断中;但是它属于参数型分解方法,K过大或过小都会导致过分解或欠分解现象,因此分解的层数需要自适应的确定。提出了一种多点峭度和VMD的复合故障特征提取方法。考虑到多点峭度可以提取多故障的冲击性周期的个数;周期性冲击个数决定VMD的分解层数K,通过VMD处理后,进一步通过FFT确定故障特征。所提出的自适应复合故障特征提取方法和EEMD(En?semble Empirical Mode Decomposition)对比分析,验证了它可以克服模态混叠的特征,通过对实测性信号处理进一步确定了此方法的有效性。最终确定了齿轮剥落和轴承滚珠等复合故障特征。

Fault Feature Extraction of Gearbox based on Adaptive VMD

In the noisy environment,the composite fault feature extraction is more difficult. The VMD is widely used in gearbox fault diagnosis,but it is a parametric decomposition method. If K is too large or too small,it will lead to over-decomposition or under-decomposition. The number of layers needs to be determined adaptively,a multi-point kurtosis-VMD(Variational Mode Decomposition)composite fault feature extraction method is proposed. Considering the multi-point kurtosis,the number of impact cycles of multiple faults can be extracted,the number of periodic impacts determines the number K of decomposition layers of the VMD,and after VMD processing,the fault features are further determined by FFT. The proposed adaptive composite fault feature extraction method and Ensemble Empirical Mode Decomposition(EEMD)comparison analysis verify that it can overcome the characteristics of modal aliasing. The effectiveness of this method is further determined by the measured signal processing. The composite fault characteristics such as gear spalling and bearing balls are finally determined.