Design And Implementation Of Performance Prediction System Of Medium And Heavy Plate Products Based On SPSS

With the growing of the information level in the steel industry,a huge amount of engineer data,such as production data,technology data,marketing data and equipment data is accumulated during the long term production operation.These data provide strong data support for the enterprises to develop the fifth level system(Data Mining System),which makes it possible to further explore their hidden knowledge.The performance prediction model established base on Data Mining System(DMS)can predict the products quality,and it also has positive significance for the optimization of the production process and the development of new products.In this paper,we take the Bayuquan branch of the Anshan iron and Steel Group Company as the implementation unit,to predict the performance of the product.Firstly,this paper introduces the performance parameters of the selected plate which includes the tensile strength,yield strength and elongation.Secondly,the technological process of Bayuquan plate products is introduced and the analysis points that the process parameters of the products performance include the processing geometry size,chemical composition of raw materials,heating process parameters,rolling engineering parameters and performance test parameters.DMS collects the corresponding geometric size data,chemical composition data,heating engineering data,rolling engineering data,performance test parameter and mechanical performance indicator.Then these data are handled to create a new data source for the following data mining.Based on the data source of medium and heavy plate products,this paper uses the SPSS Modeler platform to select four kinds of DMS models,such as artificial neural network,linear regression analysis,classification and regression tree and generalized linear analysis.The analysis model with inputs of data are the processing geometry size,chemical composition of raw materials,heating process parameters,rolling engineering parameters and performance test parameters,and outputs of tensile strength,yield strength and elongation is established.Take 70%data of the prepared data source as training sample and 30%data as calibration sample to train the model and finally gett the optimal prediction model.According to the evaluation results,the artificial neural network model has the best performance compared with the other three models in the prediction accuracy.The error is less than 10%and the accuracy can reach more than 90%.Base on the best prediction model,the importance of the input parameter could be sorted to reflect the influence factor of the process parameters.Finally,this paper uses the trained artificial neural network model to establish the prediction platform of products performance.When product developers input the key process parameters and chemical composition value,the system will automatically output a predict performance value.The system can predict the performance of medium and heavy plate products,and has certain reference value to optimize the production process and develop new products.