基于KNN-EMD算法的机车轴承故障诊断方法  

Locomotive Bearing Fault Diagnosis Method Based on KNN-EMD Algorithm

作　　者：王嘉浩 罗倩[1] 胡园园 WANG Jia-hao;LUO Qian;HU Yuan-yuan(College of Information and Communication Engineering,Beijing Information Science&Technology University,Beijing 100192,China)

机构地区：[1]北京信息科技大学信息与通信工程学院,北京100192

出　　处：《计算机仿真》2021年第11期129-132,172,共5页Computer Simulation

摘　　要：机车滚动轴承发生故障时信号是非平稳的,故障发生在不同部位时,其振动信号能量分布不同。传统KNN算法中,欧式距离主要针对空间中两样本点间的距离计算,对空间中不同位置分布间的相似性度量难以实现有效判断。针对上述问题,将一种相似度度量Earth Mover’s Distance (EMD)引入到KNN算法中,代替传统的欧式距离对故障部位进行分类。该上述方法使用‘db3’小波对轴承震动信号进行三层小波包分解,并计算第三层各结点能量作为该信号的能量分布特征。计算能量分布间的EMD,根据EMD大小对分布间的相似度进行判断。结合KNN中多数表决分类决策对故障部位进行定位。仿真结果表明,所提方法诊断准确率达到99.23%,相较于传统KNN诊断方法在诊断准确率上提升了0.77%。该方法能够准确有效地诊断滚动轴承故障,可以应用到工业生产中。When the rolling bearing of the locomotive fails, the signal is non-stationary, and the energy of each frequency band of the vibration signal changes accordingly. Aiming at the feature that the vibration signal energy distribution is different when the fault occurs in different parts, a similarity measure Earth Mover’s Distance(EMD) is introduced into the KNN algorithm to classify the fault parts instead of the traditional Euclidean distance. This method uses the ′db3′ wavelet to perform a three-layer wavelet packet decomposition of the bearing vibration signal and calculates the energy of each node in the third layer as the energy distribution characteristic of the signal. The obtained distribution is processed, the EMD between the energy distributions is calculated, and the similarity between the distributions is judged according to the EMD size. Combine the majority voting classification decision in KNN to locate the fault location. Simulation results show that the diagnostic accuracy of this method reaches 99.23%, which is 0.77% higher than the traditional KNN diagnostic method. This method can accurately and effectively diagnose rolling bearing faults and can be applied to industrial production.

关 键 词：轴承故障诊断 小波包分解 能量分布特征 最近邻 推土机距离 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]