Development And Application Of Real-time Monitoring Platform Of Big Data For Polymer Extrusion Process

Real-time monitoring of the polymer extrusion processing production process has become an effective means of the product quality management system.The massive amounts of data will be generated in polymer extrusion process,forming big data of polymer extrusion process with the characteristics of large volume,different structure,and rapid growth.In the analysis and application of these big data,there are problems such as scattered data storage,low efficiency of single-machine serial calculation,poor data sharing,low information integration,and difficulty in fusion analysis.In view of the above problems,with the support of the national key R&D project,a real-time monitoring platform of big data for polymer extrusion process was developed by using big data technology in this article.The main tasks completed in the paper were as follows:(1)Solutions were proposed for the first four problems mentioned above.Among them,the unified storage of big data in the polymer extrusion process was realized through Hadoop Distributed File System(HDFS)and My SQL relational database;the big data technologies such as Kafka,flume and spark are used to realize unified distributed stream computing for big data in polymer extrusion process;the Model-View-View Model(MVVM)architecture was proposed to solve the problems of platform data sharing and low information integration.(2)The platform visualization system including real-time monitoring,statistical analysis,data center and batch processing center module was developed.This system integrated functions such as real-time monitoring and statistical analysis of key information in the polymer extrusion process,sharing of standard reference data,and scheduling of batch processing tasks.The technical system of the platform and the developed visualization system of the platform were deployed and tested.The test results showed that the big data real-time monitoring platform of the polymer extrusion process ran stably,the functions met expectations,the performance met the needs of use,and it had good practicability and reliability.(3)Through the application of the platform in the real-time monitoring of polymer melt density,the online monitoring function of the platform was reflected.In this application,aiming at the problem that it was difficult to integrate and analyze multi-source heterogeneous data in the polymer extrusion process,the multi-source heterogeneous big data fusion technology was applied to the development process of the melt density inline monitoring model.Based on the ultrasonic,near-infrared spectroscopy and Raman spectroscopy data in the platform,this paper proposed a feature-level multi-source heterogeneous big data fusion prediction method based on Depthwise Separable Convolutions Neural Networks(DSCNN).The soft sensing method based on single-source ultrasound had been compared and verified on different polymer processing systems.The results showed that the melt density soft sensing method based on single-source ultrasonic was suitable for systems sensitive to high-frequency ultrasonic signals and the method had good repeatability,but the accuracy needed to be improved.The fusion prediction method of melt density multi-source heterogeneity based on DSCNN was suitable for more polymer processing systems with high accuracy,Moreover,the method could integrate the characteristics of various polymer extrusion process data,and had universality and complementarity.