基于改进多元多尺度加权排列熵的齿轮箱故障诊断

Gearbox Fault Diagnosis Based on Improved Multivariate Multiscale Weighted Permutation Entropy

齿轮箱振动存在多个传递路径,而典型齿轮箱故障诊断方法一般使用单个路径的单通道振动信号,易造成其它通道信息的遗漏。为充分利用不同路径振动信号故障信息,增强故障特征的质量,引入了多元多尺度加权排列熵,对其粗粒化方式进行了完善,提出了改进多元多尺度加权排列熵(IMMWPE),实现齿轮箱多通道振动信号的故障特征提取。基于此,提出了一种结合IMMWPE、成对邻近特征和粒子群优化支持向量机的齿轮箱故障诊断方法。通过齿轮箱多通道数据分析,将其与多元多尺度样本熵、多元多尺度排列熵和多元多尺度模糊熵等方法进行对比,结果证明该方法能够准确识别齿轮箱的各类故障,而且优于对比方法。

There are multiple transmission paths for gearbox vibration,and typical gearbox fault diagnosis methods generally use single channel vibration signal of a single path,which is easy to cause the omission of other channel information.In order to make full use of the fault information of vibration signals in different paths and enhance the quality of fault features,the multivariate multi-scale weighted permutation entropy is introduced,and its coarsening method is improved.An improved multivariate multi-scale weighted permutation entropy(IMMWPE)is proposed to achieve fault feature extraction of multi-channel vibration signals in the gearbox.Based on this,a gearbox fault diagnosis method combining IMMWPE,paire-wise features proximity and particle swarm optimization support vector machine is proposed.Through multi-channel data analysis of gearbox,the method is compared with the methods of multivariate multi-scale sample entropy,multivariate multi-scale permutation entropy and multivariate multi-scale fuzzy entropy.The results show that the method can accurately identify various faults of gearbox.