摘要：This paper studies the quality pattern recognitionof industrial processbased on the statistical process control(SPC). An abnormal quality pattern recognition model based on multi-support vector machinewasproposed, which can be used to solve the problem of abnormal pattern recognition in the intelligent manufacturing processfor products.The combination of "one-to-one" and "one-to-many" support vector machine (SVM) classifiersis arranged according to the structure of directed acyclic graphsin the model. At the same time, a structural optimization method was proposed to reduce the cumulative error problem. The model uses the originalfeatures of the datastream of quality. For the support vector machine classifier with low recognition accuracy, the statistical features and shape features form the datastream of quality are integrated withthe original features. Relief algorithm is used to reducethe fusion featuresin order to reduce the consumption caused by increased features. The experimentalresults demonstrate thatthe model improves the accuracy of the recognitionof abnormal patterns, and its structure also has a good time advantage.
