Construction Of Data Warehouse For Converter Steelmaking Production And Development Of Endpoint Prediction Model

In recent years,the wide application of big data technology make it possible to process excess data efficiently.Combined with big data analysis technology,through the construction of production line quality control system based on big data technology,the relevant parameters are analyzed,and a more reliable and effective quality prediction and control model can be obtained,which can guide the production of products.Due to the complex physical-chemical reactions occurring in the process of converter steelmaking,it is difficult to solve relatively complex production problems only by using traditional mathematical models.In addition,due to steelmaking is a process of multiple factors influencing each other,the application scope and precision of the traditional model is greatly affected.Therefore,the more accurate and efficient endpoint prediction model is established,which has important theoretical significance and practical application value for field production and process optimization.The neural network technology has a good ability in self-learning and fault tolerance,a large number of high quality historical data is used for model training,and model parameters are optimized to improve the accuracy of the model.This paper is based on the production data of converter steelmaking process in one steel plant,the data warehouse of converter steelmaking is established after data preprocessing,the key parameters influencing endpoint carbon content and temperature in converter steelmaking are determined.Based on the improved radial basis function neural network and the improved generalized regression neural network,the endpoint prediction models of the converter are constructed,and the forecast accuracy are analyzed and evaluated.The main conclusions are as follows.The data warehouse of converter steelmaking process was established.There are 11077 groups of basic data in the data warehouse,and the number of sample data is finally determined as 2610 groups after data preprocessing,among which 2300 groups are training samples and 300 groups are test samples.The key parameters affecting the end-point carbon content and temperature of molten steel are the quality of hot iron,hot iron composition(carbon,silicon,manganese,phosphorus,sulfur),the temperature of hot iron,the scrap melting procedure,the quality of lime,the quality of light burning,the thickness of slag,smelting cycle,tapping time,oxygen consumption of furnace.The improved radial basis function(RBF)neural network and the improved generalized regression(GRNN)neural network are selected to establish the endpoint forecasting model of converter respectively.The improved RBF neural network optimizes the clustering center using the k-means++,which accelerates the convergence speed and improves the accuracy of the model.The improved GRNN neural network uses the fruit fly algorithm to optimize the smoothing factor,which improves the forecasting speed and accuracy.Based on the improved RBF neural network,when error scope of endpoint carbon content is±0.003%,the prediction hit probability is 92%,when error scope of endpoint temperature is±15℃,the prediction hit probability is 92%.Based on the fruit fly algorithm to optimize the GRNN neural network,when error scope of endpoint carbon content is ±0.003%,the prediction hit probability is 93%,when error scope of endpoint temperature is ±15℃,the prediction hit probability is 90.3%.The improved RBF neural network has a higher accuracy in predicting the end-point temperature of converter,and the improved GRNN neural network has a higher accuracy in predicting the end-point carbon content of converter.Python is used to establish the GUI of the endpoint prediction model in converter.The interface has the characteristics of freely selecting the prediction algorithm,simplifying the operation process,intuitive prediction results,timely displaying and commenting abnormal data,etc.The models for predicting the endpoint quality of converter based on big data are established,the stability of the converter steelmaking process is monitored and the quality of product is analyzed,it provides theoretical basis for optimizing and controlling process.