基于多尺度基本熵和参数优化KELM的电机轴承故障诊断

摘要

Aiming at solving the problems of the low computational efficiency in signals feature extraction u-sing mutilscale sample entropy (MSE) and mutilscale permutation entropy (MPE) methods, a novel diag-nosis method for motor bearings based on mutilsclae base-scale entropy and parameters optimum KELM was presented. Firstly, the feature information of roller bearing vibration signals was extracted using MBSE. Meanwhile,the computational efficiency of MBSE was compared with MSE and MPE. Finally,the differ-ent conditions of the rolling bearings were judged using the kernel extreme learning machine(KELM) clas-sifier. And the key influencing parameters of KELM were optimized by artificial fish swarm algorithm(AF-SA). The experimental results showed that the proposed method could effectively identify the operating con-ditions of the rolling bearings.%针对信号特征提取中多尺度样本熵(MSE)与多尺度排列熵(MPE)算法计算效率差的问题,提出一种基于多尺度基本熵(MBSE)和参数优化核极限学习机(KELM)的电机轴承诊断新方法.该方法先通过MBSE来提取所拾取滚动轴承振动信号的特征信息,同时对比分析了多尺度基本熵、多尺度样本熵与多尺度排列熵的计算效率.最后利用KELM分类器对滚动轴承的不同状态进行判定,并通过人工鱼群算法(AFSA)对KELM的关键影响参数进行寻优.实验结果表明所述方法能够对滚动轴承的运行状态进行有效识别.