Design And Application Of Multivariate Data Fusion Analysis Software For Process Industry

Industrial data is an important asset of industrial enterprises.By effectively using factory production,management and operation data,the enterprises can better achieve refined management.However,the current information systems in process industry enterprises are of wide varieties and have many data types.Non-standards and high thresholds for data analysis make it impossible to truly realize the value of data mining.In this thesis,the main technologies are object modeling technology,data fusion processing technology,big data analysis and mining technology.Then,combined with the characteristics of process industry enterprise data and industry knowledge,it use object modeling technology to organize different types of data in the process industry,use the data fusion processing technology to clean and in-depth processing of the original data in order to provide data sources for upper-level industrial applications,use the packaged big data analysis technology to achieve low-threshold data analysis.By discussing the design and implementation process of multiple data fusion and analysis software,this thesis built a multi-data fusion analysis software for the process industry.The main content can be summarized as the following aspects:1)Study the data characteristics and data fusion requirements of process industry enterprises.2)Research object-oriented modeling technology,data fusion processing technology,big data analysis and mining technology,and based on these technologies,design the overall architecture of the process industry-oriented multi-data fusion analysis software.3)Detailed design of industrial data asset management components based on object modeling technology.Use unified metadata language to build multiple factory models.Use knowledge graph technology to establish business connections between industrial objects in order to provide effective support for upper-level data applications.Solve the problem of non-standard management of multiple data.4)Detailed design of industrial big data analysis components based on machine learning and data mining.Provide data mining and predictive analysis capabilities.And through packaging technology,make it easy for users to gain insights into data intuitively and help enterprise can realize data analysis on lower threshold.5)Use the multivariate data fusion analysis software in the pilot application.The prediction results of the smoke emission content in the pilot application of the three regression algorithms of LSTM(Long Short Term Memory)time series regression prediction algorithm,BP(Back Propagation)neural network algorithm and linear regression algorithm were compared.The application results showed that the designed multivariate data fusion analysis software can effectively integrate and analyze multiple enterprise data,and provide decision support for enterprise production and management.