EEMD分解的模糊熵t-SNE的齿轮故障诊断

Gear Fault Diagnosis Based on Fuzzy Entropy t-SNE of EEMD Decomposition

针对齿轮容易出现故障的问题,提出了一种通过求集合经验模态分解(EEMD)分量的模糊熵,然后再通过t分布的随机邻域嵌入算法(t-SNE)降维,最后将降维后的矩阵输入神经网络进行分类的齿轮故障诊断方法。利用该方法,对正常和具有三种不同裂纹程度的齿轮进行故障诊断和分类,并将分类结果与通过PCA降维的分类结果进行比较。结果表明:通过t-SNE降维的分类准确率达到了100%,说明提出的方法具有较好的故障诊断和分类效果。

Aiming at the problem that the gears are prone to faults,a gear fault diagnosis method is proposed,in which the fuzzy entropy of the set empirical mode decomposition(EEMD)component is calculated,and then the dimension is reduced by the random neighborhood embedding algorithm(t-SNE)of t distribution,and finally the reduced matrix is input into the neural network for classification.Through the fault diagnosis and classification of normal gears and gears with three different crack degrees,and compared with the classification through PCA dimensionality reduction.The results show that the classification accuracy of t-SNE dimensionality reduction reaches 100%,indicating that the proposed method has a better effect of fault diagnosis and classification.