一种时移多尺度排列熵与ELM相结合的滚动轴承故障诊断方法

A Rolling Bearing Fault Diagnosis Method of Time-shifted Multi-scale Permutation Entropy Combining with ELM

多尺度排列熵(Multi-scale permutation entropy,MPE)随着尺度因子的增加得到的粗粒化序列长度越来越短,造成时间序列信息的严重损失。为此,提出了时移多尺度排列熵(Time-shifted multi-scale permutation entropy,TSMPE)。首先,采用仿真信号分别对TSMPE与MPE做仿真对比分析,结果表明,TSMPE对原始振动信号的长度依赖性较小,得到的熵值更加稳定。进一步地,提出了一种基于TSMPE与极限学习机的滚动轴承故障检测与诊断方法,将其应用于两组实际滚动轴承测试数据对滚动轴承故障类型和程度进行识别,结果表明:所提出故障诊断方法不仅能够准确地诊断滚动轴承的故障类型和程度,而且识别率高于基于MPE与ELM的故障诊断方法。

When using the Multi-scale Permutation Entropy(MPE),with the increase of the scale factor,the obtained coarse-grained time series becomes shorter and shorter,which results in serious loss of time series information.For this purpose,a new Time-shifted Multi-scale Permutation Entropy(TSMPE)is put forward in this study.Firstly,the comparison of TSMPE and MPE is carried out with simulated signal.The results show that TSMPE has less dependence on signal length and the obtained entropy value is more stable.Furthermore,based on TSMPE and extreme learning machine,a fault detection and diagnosis method for rolling bearing is proposed and applied to two testing data of actual rolling bearings to identify the fault types and degrees.The results show that the proposed fault diagnosis method can not only accurately diagnose the fault types and degrees of rolling bearings,but also the recognition rates are higher than the fault diagnosis methods based on MPE and ELM.