基于自适应MOMEDA与VMD的滚动轴承早期故障特征提取

Feature extraction for rolling bearing incipient faults based on adaptive MOMEDA and VMD

轴承故障衍生早期,由于故障尺寸较小且易受环境噪声和信号衰减的影响,因此故障冲击信号往往非常微弱。变分模态分解(Variational Mode Decomposition,VMD)已经在轴承的故障特征提取中有一定的应用,但对于背景噪声较强时的滚动轴承的微弱故障提取效果并不理想。针对这一问题,将改进多点优化最小熵解卷积(Multipoint Optimal Minimum Entropy Deconvolution Adjusted,MOMEDA)与VMD相结合,研究了滤波器长度对MOMEDA效果的影响,提出基于进退法确定最优滤波器长度的自适应MOMEDA方法。利用自适应MOMEDA对信号降噪并避免传统MED迭代以及滤波后可能出现的虚假峰值。将自适应MOMEDA降噪后的信号使用VMD进行分解,然后依据谱峭度大小进行重构,对重构之后的信号进行故障特征提取,取得了较好的效果。最后通过实验验证了方法的可行性及有效性。

Due to bearing incipient fault size being smaller and susceptible to environmental noise and signal degradation,fault impact signals are often very weak.Variational mode decomposition(VMD)has a certain application in bearing fault feature extraction,but bearing weak fault extraction is not ideal under stronger background noise.Here,aiming at this problem,the multipoint optimal minimum entropy deconvolution adjusted(MOMEDA)and VMD were proposed to do bearing early fault diagnosis.Influences of filter length on MOMEDA effect were studied.An adaptive MOMEDA method based on the forward and backward one was proposed to determine the optimal filter length.The adaptive MOMEDA was used to de-noise signals and avoid false peal values after traditional MED iteration and filtering.The de-noised signals were decomposed with VMD to do signal reconstruction based on spectral kurtosis.Fault features were extracted from the reconstructed signals to gain the better effect.Finally,the feasibility and effectiveness of the proposed method were verified with tests.