关键词: 控制图, 模式识别, 特征融合, 降维, 核主成分分析, 支持向量机

Abstract: In order to improve the intelligence of quality monitoring in machining processes, the paper proposed a control chart classification method based on fusion feature reduction and KPCA-SVM, on the basis of quality fluctuation which was described by control chart. Firstly, the Monte Carlo method was applied to generate the control chart data sets, statistical features and shape features were extracted to fuse with original features, then kernel principal component analysis was applied to reduce dimensionality of high dimensional fusion feature sets. Finally, genetic algorithm was used to optimize parameters of SVM. Recognition accuracy were compared through the simulation experiments with the applications of dimensionality reduction and different classification models, the results demonstrate that the higher recognition accuracy may be achieved by using the proposed method.

Key words: control chart, pattern recognition, feature fusion, dimension reduction, kernel principal component analysis（KPCA）, support vector machine（SVM）