The Predicting Of Cold Rolled Products’s Quality Properties Based On Data Mining Technology

Cold-rolled products with a high added value are the steel companies leading product,particularly like the automotive steel production expand rapidly with the market demand.But in the car plate developing process are usually deficiencies in the experimental conditions,large resource-consuming on experiments,long development cycle,unstable on new product quality control and product performance often fluctuate so on.It is difficult to meet the large demand of cold-rolled products market.Therefore,iron and steel enterprises need to build a simulated environment using for new products development,to reduce the number of experimental production times,test、direct and optimize the design,shorten product development cycles,bring down development costs and improve the product quality.The article discusses how to use data mining technology to developing various types of data existing in the enterprise information systems,converting it into knowledge and process it for cold-rolled products quality performance prediction.It demonstrates the feasibility and operability of cold-rolled products quality predict the performance theoretical based on data mining techniques in iron and steel enterprises.This paper discusse several aspects such as follows:from the significance of cold-rolled products mechanical properties prediction parameters using correlation analysis from to determine the performance of the key factors;from data preparation,data cleaning,data integration to build data marts;from the choice of data mining algorithms for modeling,training,to the model assessment,forecasting.Dissertation focuses a deep study on the process of establishing of cold-rolled products quality performance prediction model(including neural networks,decision trees,multiple linear regression model),by using the S+ miner and SPSS statistical analysis software to build decision tree,BP nerve network cold-rolled products performance prediction model,obtain the model’s input parameters by correlation analysis,the cold-rolled products important mechanical performance parameters of the yield strength,tensile strength,elongation as output,by training model,described production of complex functional relationship between process parameters,chemical composition and mechanical properties finally.Comparing the real sample data of mechanical properties with the results of performance prediction models of cold-rolled products,proved the effectiveness of cold-rolled products quality performance prediction based on BP neural network model.