New Method for Mixed Abnormal Pattern Recognition Using Multi-Class Support Vector Machines

Control charts is an important tool of statistical quality control (SQC), and the recognition of mixed abnormal pattern exists on the control chart is one of difficulties of on-line intelligent process quality diagnosis. After limitations of control chart-recognizers used in practice were analyzed, a novel intelligent process quality diagnosis method is proposed with a special model of multi-class support vector machine (MSVM). In this model, the binary decision tree is firstly used in recognizing the controlled process sample with upper on-line recognition rapidity. Then five single feature abnormal models are recognized using multi-class SVM classifiers in one-versus-one (OVO) decomposition. Finally, to make full use of the unclassifiable regions existing in the traditional "Max-Wins" voting (MWV) strategy, the capability is realized to recognize mixed abnormal patterns. Numerical results are given to demonstrate that, the proposed method can achieve the better classification ability and resolve the mixed abnormal pattern recognition problem. So, it provides a candidate for the small-batch production process quality online diagnosis and control.