ELMD熵特征融合与PSO-SVM在齿轮故障诊断中的应用
Application of ELMD Entropy Feature Fusion and PSO-SVM in Gear Fault Diagnosis

提出基于ELMD熵特征融合与PSO-SVM的齿轮故障诊断方法。该方法首先对原始信号进行总体局部均值分解（Ensemble local mean decomposition，ELMD），得到若干乘积函数（PF）；其次，对ELMD分解得到的前5个PF分量进行求取能量熵和近似熵，并利用KPCA对其进行特征融合；然后，选取部分融合特征作为训练样本，其余作为测试样本；最后，利用PSO优化的支持向量机对融合特征样本进行训练与测试。实验中，将单特征和融合特征分别进行SVM和PSO-SVM识别精度的对比。实验结果证明，所提方法可有效地应用在齿轮故障诊断中。
A gear fault diagnosis method based on ELMD entropy feature fusion and PSO-SVM is proposed in this paper. Firstly, the original signal is decomposed by ensemble local mean decomposition (ELMD), and several product functions (PF) are obtained. Secondly, the energy entropy and approximate entropy of the first five PF components obtained by ELMD decomposition were obtained and characterized by KPCA. Then, some of the fusion features are selected as training samples, the rest as test samples; finally, the PSO-optimized support vector machine is used to train and test the fusion feature samples. In the experiment, the singular and fusion features are compared with the recognition accuracy of SVM and PSO-SVM respectively. Experimental results show that the proposed method can be effectively applied in gear fault diagnosis