Abnormal Pattern Recognition Of Mixture Control Chart Based On Support Vector Machine

With the rapid development of manufacturing industry,quality control in complex production process is paid more and more attention.The goal of production process quality control is to produce high quality products at the lowest cost.In order to reasonably monitor the production process and improve the collection and analysis of manufacturing process data,all possible problems in the production process are dealt with.Statistical process control is a process control tool to help product quality balance and reduce failure cost.With the development of computer and computer technology,the traditional quality graph control method is not suitable for the era of big data,and machine learning has become the mainstream of quality graph control method.An intelligent network for pattern recognition of hybrid control charts is designed by machine learning algorithm.The proposed network is mainly composed of three parts : feature extraction,parameter optimization and classifier.In order to effectively obtain the features of each state,shape and statistical features are used for extraction respectively.Principal Component Analysis(PCA)is used for secondary extraction of features.Support Vector Machines(SVM)is used for feature extraction of data,and the control chart category is output.Reasonable selection of penalty parameters and kernel parameters can improve the performance of the model.In this study,the parameter selection of the model is optimized by genetic algorithm.To evaluate the performance of the proposed model,the effectiveness of the proposed method was evaluated through six basic control charts,four mixture control charts and three different experiments.Experiment 1 tested and compared the pattern recognition results of BP,MSVM,PCA＿MSVM and PCA＿MSVM＿GA,respectively.The experiment showed that the hybrid PCA＿MSVM＿GA method proposed based on shape and statistical characteristics had better performance than the other three models.Experiment 2 tests and compares the FFR,PD and SSF models.It is found that the FFR method with fusion features is more accurate than other models.The experiment based on PCA＿MSVM＿GA＿FFR can better realize the recognition of control charts.Experiment 3 compared the deep learning methods,and the results showed that although the process of deep learning method was faster,the accuracy was lower than that of the proposed method in this paper.To better realize the hybrid control chart pattern recognition problem,this paper first uses the shape and statistical characteristics of the control chart to extract,and then uses PCA to extract the features twice.In order to achieve better control chart recognition,the kernel parameters of SVM are optimized by GA,so that the model can achieve better performance.Finally,through experimental analysis,it is found that the FFR method combined with fusion characteristics can make the proposed model achieve higher recognition accuracy,effectively identify the control chart,and realize the monitoring of the production process.