Research For Quality Abnormal Pattern Recognition In Dynamic Process Based On Feature Fusion

With the increasing automation,continuous and complex production process,the abnormal quality pattern recognition and quality diagnosis of dynamic process has attracted many scholars' attention.The accuracy of pattern recognition of abnormal quality mainly depends on the pattern classification feature and classifier of the two key factors,component characteristics extracted from the original data quality model,not only can effectively reflect the nature and state of the mode of quality,enhance the distinction between different modes,can also reduce the redundancy and complexity of data to a great extent,has become one of the effective means to enhance the precision and quality of pattern recognition.But because of the complexity of data dynamic process,any kind of abnormal mode,only rely on the data characteristics of a single type of difficult to obtain a higher recognition accuracy,the urgent problem is how to make the information fusion screening and efficient use.Therefore,how to extract feature data with low dimension and strong detail information for dynamic data stream is the key to improve the efficiency of dynamic process pattern recognition.On the basis of collecting a large number of domestic and foreign research literature,this paper studies the dynamic process quality anomaly pattern recognition method based on feature fusion.Firstly,on the basis of the summary of dynamic process pattern recognition,feature extraction and quality diagnosis,this paper defines the dynamic process model.Then,a dynamic process quality anomaly pattern recognition method based on feature fusion is proposed,in addition,the particle swarm optimization algorithm is used to find the optimal parameters of SVM.Finally,the effectiveness of the proposed method is verified by three simulation experiments.The results show that:In this paper,the method of feature recognition based on feature fusion and multi support vector machine is more efficient than the traditional one.The particle swarm optimization algorithm is used to find the optimal parameter combination of the support vector machine,and the fusion reduction feature is used as the input vector of the multi support vector machine.The model is used to reduce the dimension of feature set,which can reduce the dimension of feature and eliminate redundant and irrelevant features.The research and innovation of this paper lies in:This paper presents a method ofdynamic process quality anomaly pattern recognition based on fusion features.The rough set method is applied to the feature combination and optimization of the quality anomaly pattern recognition,which can eliminate the small contribution to the classification or the redundant information,and then get the feature set which is easy to be classified.The support vector machine is chosen as the classifier of the quality anomaly pattern recognition,and the particle swarm optimization(PSO)algorithm is used to find the optimal parameter combination.This study has overcome the characteristics of a single type of data can only reflect the dynamic process of running condition and the loss of information from a defect information,and effectively compress the information and realize the real-time processing and diagnosis,for petroleum,chemical,provide real-time quality monitoring and fault diagnosis technology of tobacco industry automation.