Research And Development Of Performance Prediction System For New Products Of Medium-thick Plate

Under the condition of the domestic iron and steel industry is very competitive,enterprise for profit space,more pursuit of new products or high quality products brought competitive advantage,Ansteel as the old manufacturing base,pays more attention to the new products,high quality research and development.And the medium-heavy plate products are Ansteel's important steel products,including ship plate,military industry,nuclear power steel with steel and space steel,etc.,are the strategy of Ansteel products.Therefore,development of new products or high-quality plate products has great strategic significanceThe medium-heavy plate products are typical hot-rolling(or controlled rolling)state products that delivered directly to the end user to use,product performance is the main basis that measure the quality of the products.The main factors affecting the performance of the product is the product of chemical composition and production process,Therefore research and develop products performance prediction system,understand and master the influence of the product chemical composition and production process for product performance has very important significance.Research and development of new products Plate-Performance-Prediction-System can use the massive data in Ansteel's data warehouse,combined with today's mainstream big data technologies-data marts and data mining technology,to research product performance prediction model based on the technology of big data,and develops an interactive software systems,it can make technical person to achieve new product development in product performance prediction function in a simulated environment,and can greatly reduce the time and cost of new product development plate.This paper focuses on three aspects:data mart application technology,data mining(analysis)technology and development of interactive software system.First,use DataStage development tools to design,build data marts;Second,using Neural Networks and Extreme Learning Machine both model theory,with the aid of SPSS Modeler modeling tools and MATLAB tool to create,train performance prediction model.Finally,the use of JAVA language developing interactive software system.and comparing the traditional neural network modeling techniques,to research Extreme Learning Machine(ELM)technical advantages on prediction problems,ELM can better avoid the data sample shortage and local optimal solution problem,in order to better solve problem of part of the production process has too little amount of data in Ansteel's plate production data.Lay the foundation for development of new products under more production process.