Shape Prediction And Model Optimization For Hot Tandem Rolling Process Based On Mechanism Fusion Data

In this paper,a 1580 mm seven-stand hot tandem rolling production line is studied as the research object,and the data-driven methods and rolling mechanism fusion artificial intelligence algorithm are employed as the mathematical tools.Meanwhile,aiming at the CVC(Continuous variable crown)hot rolling mills,the machine learning methods guided by the rolling mechanism are studied to realize the multi-objective optimization control of strip crown-thickness of hot strip rolling.Besides,the machine learning(ML)and optimization algorithms are used for in-depth research.Three strip shape prediction models of KPLS-SVM(Kernel partial least squares-Support vector machine),PCA-CS-SVM(Principal component analysis-Cuckoo search-Support vector machine)and DP-M-SVR(Dimensionality processing-Multi-output support vector regression)are proposed ultimately,which have an important theoretical guiding significance and practical application value for realizing accurate control of strip shape and improving the quality of hot rolled products.The main study contents of this paper are as follows:(1)Based on a large number of hot strip process data collected on rolling site,the multivariate statistical methods and data processing techniques are employed to analysis and processing the experimental data,which can fully obtain the main information of experimental data.Meanwhile,the method of PCA is used to reduce the dimensionality of experimental data with variable coupling,abnormal data and a large amount of noise,which can effectively reduce the coupling relationship between variables,enrich data information and ensure data quality.(2)The kernel partial least squares(KPLS)method based on data-driven is used to deal with the strong coupling,multivariate and nonlinear relationship between process parameters and quality indicators effectively.Besides,a KPLS-SVM hot strip crown quality prediction model based on data-driven methods combined with machine learning is proposed,and the particle swarm optimization(PSO)algorithm is used to optimize the key parameters of SVM model,which can fully improves the prediction accuracy of strip crown for hot strip rolling.(3)In order to make up for the poor quality of traditional control methods in the process of hot strip roll and meet the ever-increasing requirements of strip crown accuracy,the model of SVM combined with PSO algorithm and principal component analysis combined with cuckoo search(PCA-CS)optimization strategy are proposed.Finally,the experimental results show that the PCA-CS-SVM model has the highest prediction accuracy and the fastest convergence speed,which has a theoretical guiding significance for improving the quality of hot-rolled products.(4)To address the difficulties of the multivariable,non-linear,strong coupling,time-varying and the conventional SVM is only an one-dimensional output problem in the hot strip rolling process,a model of machine learning prediction guided by rolling mechanism is established.Among them,on the basis of the rolling process data,the rolling mechanism data are calculated as extra input features to participate in the training process of the proposed model through the dimension processing.At the same time,genetic algorithm(GA)is employed to optimize the major parameters of multi-output support vector regression(M-SVR)model to further improve the prediction accuracy.The experimental results fully prove that the proposed model has stronger versatility.In addition,it has theoretical guiding significance and practical application value for realizing precise control of the shape quality and improving the quality of hot-rolled products.