基于加权非局部平均算法的滚动轴承故障诊断

Fault diagnosis of rolling bearings based on weighted nonlocal means algorithm

非局部平均算法(Non-Local Means,NLM)拥有十分优异的去噪性能,被广泛应用于二维图像信号处理领域,并逐渐应用于一维轴承故障信号检测中。该方法能够利用信号中存在的冗余冲击成分,以包括局部结构的小窗口或邻域为单元,利用局部结构相似性进行加权运算,抑制随机噪声信号,使冲击特征得到增强。但对于强烈背景噪声干扰下的信号,诊断效果不够理想。提出一种基于非局部平均算法的权重包络谱诊断方法,该方法对信号各点进行加权运算,通过权值比对,使信号冲击分量的尖锐特性得到进一步增强。通过与EEMD方法对比,以及实验室轴承故障数据和工程案例分析,验证了该方法在检测轴承局部故障检测中的有效性和优越性。

The nonlocal mean （ NLM ） algoritym is widely applied in image processing at present and it effectively overcomes the limitations of neighborhood filters. NLM becomes very popular in fields of 2D image signal processing, and then is used in vibration signal processing for fault diagnosis of rolling bearings. NLM is an emerging method to tackle such problems with an ability to eliminate noise. Unfortunately, NLM is unable to trim down all noise in the presence of strong interferences. Aiming at such a dilemma, a novel fault diagnosis method for rolling element bearings was proposed based on weighted NLM de-noising. The impact components features were reflected throngh weight companion and the weighted operation. Then, envelope spectral analyses were performed with weights to allow easier detection of fault characteristic frequencies. Compared with EEMD, the effectiveness and superiority of the proposed method were verified with test data and a field case anallysis.