摘要
为简化多支持向量机识别模型的计算复杂度、提高动态过程质量异常模式的识别精度,提出一种基于多主元特征与支持向量机相结合的动态过程异常监控模型。利用主元分析方法对动态数据进行特征提取,将所提取的不同主元特征作为支持向量机分类器的输入对模型进行训练。将识别效率高的主元特征对应的转换矩阵与多支持向量机相结合,构建了基于多主元特征的多支持向量机识别模型,对质量异常模式进行识别。仿真实验表明,所提基于多主元分析支持向量机识别模型的识别精度比传统基于主元特征或其他特征提取方法的识别模型有显著提高,且训练所需时间大大减少。
To improve the recognition accuracy and reduce the computation complexity in dynamic process,a novel quality abnormal monitoring model based on Multiple Principal Component analysis integrated Support Vector Machine（MPC-SVM）was proposed.The different PC vectors were extracted from original data matrix by PCA technique,and the extracted principal components were taken as the input of SVM classifiers to train the model.After integrating transform matrix of PCA with MSVM,MPC-SVM recognition model for quality abnormal patterns in dynamic process was established.Compared with recognition model based on other traditional methods,the simulations illustrated that the proposed MPCA-SVM model had a good performance than other models in the fields of train time and recognition accuracy.
作者
刘玉敏
张帅
LIU Yumin , ZHANG Shuai(School of Business, Zhengzhou University, Zhengzhou 450001, Chin)
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2018年第3期703-710,共8页
Computer Integrated Manufacturing Systems
基金
国家自然科学基金资助项目(71672182
71272207
61271146
71711540309
U1604262)~~
关键词
动态过程
质量异常模式
多主元特征
支持向量机
dynamic process
quality abnormal monitoring
multiple principal component
support vector machine
