Research On Optimization Of Steel Rolling Control Based On Data Analysis

Steel hot rolling consists of heating furnace-rough rolling-finishing rolling-laminar cooling-crimping.The production process has complex features such as strong coupling,nonlinearity,distributed,and difficult parameter measurement.At present,the production and manufacturing in all walks of life have increasingly higher requirements for the quality of steel products.How to further improve the quality of steel products is the top priority of current steel production research.In order to achieve high-quality and high-volume steel rolling,it is imperative to study the optimization of the steel rolling process operation.The rolling force mathematical model is one of the main models in the finishing process of hot continuous rolling.The precise control of the rolling force model plays a vital role in improving the quality of steel products and accurately controlling the shape and thickness of steel products.At present,only the traditional rolling force model is far from being able to meet the precision requirements for the production of high-quality,multi-steel products.Taking the 2250 mm hot rolling production line of B Steel Plant as the research object,this paper combined industrial big data,expert knowledge,intelligent model and sensitivity analysis method to optimize the rolling process in order to solve the problem that the rolling force model of the production line has a large error between the calculated value and the measured value,that is,the rolling force model has a low calculation accuracy.The main research contents are as follows:(1)Problem description and program research of rolling process.Through reading a lot of literature and learning theoretical knowledge,familiar with the development status,technology and mechanism of the hot strip rolling process,and describe in detail the hot strip rolling process,control methods and control characteristics.The 2250 mm hot rolling production line of Steel B is reviewed.Aiming at the problem that there is a large error between the calculated value of rolling force mechanism model and the actual measured value,an intelligent control optimization scheme combining industrial big data,expert knowledge,intelligent model and sensitivity analysis algorithm is proposed.(2)Data analysis and preprocessing.Deeply excavate the influencing factors of the original system rolling force model,and determine the parameter variables that affect the rolling force model according to the rolling force mechanism model combined with expert experience.Since the production data has the characteristics of large amount of data,large detection noise interference,and large human influence factors,the pre-processing of data is an indispensable link before modeling.Taking selected variables as the data preprocessing objects,data cleaning,data filtering,data dimensionality reduction and data normalization are performed on them respectively,and the industrial field data is processed into high-quality modeling data.(3)Establish the rolling force decision model.Using the production data collected in the actual process of the 2250 mm hot rolling mill of Steel B,establish a support vector machine(SVM)decision model based on Linear kernel function(Linear),Polynomial kernel function(Poly),and Gaussian kernel function(RBF),and the three decision-making models are analyzed and compared through three evaluation indicators: coefficient of determination(R2),mean absolute error(MAE)and mean square error(MSE).(4)Optimization of decision-making model.Based on the decision model based on RBF?SVM with the highest model accuracy,the kernel function parameter gamma and penalty coefficient C of the RBF?SVM model are optimized through the differential evolution algorithm to improve the accuracy of the decision model.(5)Control optimization of rolling force model based on sensitivity analysis.Through learning the theoretical knowledge of the rolling system and deep analysis of the original system rolling force model,it is known that the main factor affecting the calculation of the rolling force is the deformation resistance,and the deformation resistance calculation is mainly controlled by the parameters in the table FSU0905 in the system C-TOLL interface.The sensitivity analysis method is used to analyze the influence of each parameter on the deformation resistance and find the key parameters.Establish evaluation rules based on expert knowledge,compare and evaluate model decision values with calculated values,adjust the calculated value of the system deformation resistance model by modifying key parameters,thereby changing the calculated value of the original system rolling force mechanism model.That is,the control parameters of the roll gap between the L2 level and the L1 level control finishing stand are changed,the finishing rolling process is optimized,and the product quality is improved.This subject takes the 2250 mm hot rolling mill in steel B as the research object,Taking the improvement of the quality of steel products as the control goal,and aiming at the problems in the rolling process,an intelligent control optimization scheme that combines industrial big data,intelligent models and expert knowledge is proposed.According to the actual industrial data of the steel mill,a rolling force prediction model was established.The experimental results show that the support vector machine rolling force decision model(DE?SVM)optimized by the differential evolution algorithm has high accuracy and meets the requirements of the site.Based on expert experience,the model decision value is compared and judged with the model calculation value.The sensitivity analysis algorithm is used to find out the key control parameters and accurately adjust the rolling process to achieve the goal of improving product quality.This subject provides a new idea for the optimization of rolling process control and has broad application prospects.