基于小波包变换与核主成分分析的铣削颤振识别

摘要

提出一种基于小波包变换（wavelet packets transform, WPT）与核主成分分析（kernel principal component analysis，KPCA）的颤振识别方法。铣削颤振会抑制或增强某些频段内的信号，利用四层小波包分解与重构，得到16个频段内的重构信号，获得各重构信号的面积，并进行归一化处理，完成铣削颤振特征向量的选择。继而通过对比基于主成分分析（principal component analysis，PCA）与核主成分分析的特征提取方法的特征提取效果，选取KPCA对特征向量进行降维处理，最后以降维后的数据作为最小二乘支持向量机分类器的输入对铣削状态进行识别。结果表明，在小样本的情况下仍能有效、准确地对铣削状态进行分类，分类准确率达95.0%。%A milling chatter identification method based on wavelet packet transform (WPT) and kernel principal component analysis (KPCA) is proposed. This method consists of four steps. In the first step, the measured milling force signals are decomposed and reconstructed by four-level WPT, so that the force signals can be allocated in a certain frequency band. In the second step, the reconstructed signal areas of different frequency bands are normalized and selected as a feature vector. In the third step, through the mutual comparison of the results of principal component analysis (PCA) method and KPCA method, the KPCA feature extraction method is selected for dimension reduction. Finally, the least squares support vector machine (LS-SVM) is designed for normal milling and chatter pattern classification. The experimental results prove that the method can identify the chatter accurately and effectively even in the case of small number of samples with an accuracy rate of 95.0%.