基于LMD样本熵与ELM的行星齿轮箱故障诊断

Planetary Gearbox Fault Diagnosis based on LMD Sample Entropy and ELM

为了解决行星齿轮箱故障特征提取困难的问题,考虑到行星齿轮箱振动信号的耦合、非线性的特点,提出基于局域均值分解(LMD)的样本熵和极限学习机(ELM)结合的行星齿轮箱故障诊断方法。首先,利用局域均值分解方法将振动信号自适应地分解为多个PF分量,结合相关系数选取包含主要故障信息的前4个PF分量。其次,应用样本熵方法进行计算,组成特征向量。最后,将特征向量输入极限学习机进行故障分类。在行星齿轮箱实验台上进行了实验,与基于概率神经网络(PNN)分类算法进行了对比,并与基于奇异值分解(SVD)构成的特征向量进行了对比,结果验证了该方法的有效性。

In order to solve the difficult problem of early fault feature extraction of planetary gearbox and consider that the planetary gearbox vibration signal is coupling and nonlinear,and the signal has multiple transmission paths,a planetary gearbox fault diagnosis method based on Local Mean Decomposition(LMD) and Sample Entropy and Extreme Learning Machine(ELM) is proposed.Firstly,the vibration signal is adaptively decomposed into a plurality of PF components by LMD,and the first four PF components including the main fault information are selected in combination with the correlation coefficient and the variance contribution rate.Secondly,the Sample Entropy of the signal is calculated to form a feature vector.Finally,the feature vector is input into ELM for fault classification.Experiments are carried out on the planetary gearbox test bench,compared with the probabilistic neural network classification algorithm,and compared with the feature vector based on Singular Value Decomposition(SVD).The results verify the effectiveness of the proposed method.