Research On Abnormal Quality Pattern Recognition Based On CEP And MSVM

Manufacturing the Internet of Things is a new manufacturing model and service model that deeply integrates the Internet of Things technology and the manufacturing industry.It can improve the competitiveness and influence of manufacturing companies in China.But This model still has many challenges in the development process.For example,in the aspect of network transmission,constrained resources and heterogeneous network convergence seriously affect the reliability and real-time of data transmission.In terms of data processing,due to the characteristics of massive data sources,high spatial-temporal correlation,and timeliness,the limitations of limited storage and computation are insufficient to support the complete processing of data.The real-time monitoring of processing quality,as a reliable service for manufacturing the Internet of Things,requires processing more quality data per unit of time,thereby increasing the speed of recognition of abnormal quality patterns.The domestic and international researches on quality control charts has begun to focus on using artificial intelligence algorithms for active identification.And through the extraction of control chart features and the screening of key features,the recognition accuracy is improved.However,it still needs to consider the probability of occurrence of different control chart patterns to adjust the classification structure,thereby further increasing the recognition speed.In addition,the use of complex event processing techniques to obtain quality data from manufacturing data streams will further increase the overall recognition speed of online recognition of abnormal quality patterns.Therefore,this dissertation combines multi-support vector machines and complex event processing to improve the online recognition efficiency of abnormal quality patterns in manufacturing the Internet of Things.The main research work has the following two aspects:(1)This paper analyzes the research status and requirements of intelligent quality control,and proposes a recognition model of quality abnormal patterns for production process based on optimized feature selection and dynamic combination of multi-support support vector machines.The model firstly adopts the relief and random forest algorithm for feature selection,and takes the extracted feature subset as the input of the multi-support vector machine classifier based on the dynamic combination of the probability of abnormal patterns.Then,the particle swarm optimization algorithm is used to optimize the hyperparameter.Finally,the optimized classifier is used to recognize the abnormal quality patterns,so as to take into account both the recognition accuracy and the recognition speed.(2)For the sharing of multiple complex event queries in data streams,this paper proposes a multi-pattern complex event detection method based on double-array trie-tree,builds a multi-pattern matching automaton model to reduce redundant detection and computation,and fully compresses the storage space using the double-array trie-tree method,thus improve the efficiency of the complex event processing.Simulation experiments show that the quality data acquisition scheme proposed in this paper reduces the space consumption while taking into account the detection efficiency;the recognition model of quality abnormal patterns has better recognition accuracy than traditional recognition methods,and its identification structure is simpler and the dynamic adjustment structure based on the probability of occurrence helps to increase the recognition speed in practical applications.