Development Of The Crown Prediction And Intelligent Prediction System For Hot Strip Rolling Based On Convolution Neural Network

Abnormality of rolling parameters in the hot strip rolling process will lead to surface quality problems such as edge drop and sickle bending on the surface of strip steel products.Based on this problem and the need of society for high-quality strip steel products,a mechanism model collaborative convolution neural network is combined with big data analysis and target optimization algorithm to establish a mechanism collaborative convolution neural network intelligent plate crown prediction system,Realized accurate prediction of strip steel crown under different rolling conditions and developed the plate crown early warning function.The specific research content of this paper is as follows:Firstly,the hot strip rolling site production data were screened and preprocessed,and the deep feedforward neural network model and the convolutional neural network model were built according to the deep neural network theory.The prediction accuracy of the deep feedforward network is 88.54% and that of the convolutional neural network is91.6%.The obtained convolutional neural network model has better prediction performance and higher accuracy,and it is chosen as the base model of the forecasting system.Secondly,a mechanistic model of roll temperature field and wear was established based on rolling theory,and the accuracy was verified by comparing with the field measured data.A neural network classification model was established to classify the production data into four working conditions according to the rolling parameters for classification training.The combined mechanism model,convolutional neural network,and big data analysis established a mechanism collaborative neural network plate convexity forecasting system,which shortened the prediction time by 83% compared with the traditional mathematical model,and the error value of plate convexity was between±10μm,with a prediction accuracy of 95.9%.The early warning mechanism of plate convexity was also developed.Finally,the intelligent plate convexity forecasting software was developed based on the mechanistic cooperative neural network forecasting system to achieve accurate prediction of work roll temperature,wear and strip plate convexity.Based on the developed plate convexity warning mechanism combined with the actual rolling production case of 1750 hot rolling line,the function of timely warning when the rolling process or parameters are abnormal is verified,which reduces the production failure rate of strip steel products.This research is conducive to improving the precise control of the crown of hot strip steel,and has theoretical and practical significance for the intelligent production of hot strip steel and improving the surface quality of the strip steel.