Research Of Strip Shape Prediction Based On Rolling Mechanism And Industrial Data

With the improvement of the quality requirements of cold-rolled strip in the downstream industries such as high-grade automobile sheets and household appliances,the cross-sectional shape and straightness of strip have become the most important quality index of cold-rolled strip products.The flatness of cold-rolled strip is affected by multi-factors of nonlinear,strong coupling and time-variant,such as mill type,process parameters and rolling state,so the mechanism of flatness control is complex.The establishment of a high precision mathematical model of flatness prediction is of great practical significance.In this paper,an approach by three-dimensional(3D)elastic-plastic finite element method(FEM)is used to study the influence of the process parameters on the rolled flatness in the rolling process,and the off-line prediction model based on the rolling mechanism and on-site production data is developed.The main research contents are follows:(1)Two 3D elastic-plastic finite element models of the traditional UCM mill and the new HYPER UCM mill are respectively established according to the size parameters and the physical performance parameters of the mill,as well as the strip mechanical property parameters obtained from the rolling production field.The setting of various parameters,the selection of material model,the application of boundary conditions and loads during the modeling process are given.The accuracies of the finite element models are verified by the comparison between the theoretical calculation values and the actual measured values of rolling force.The comparison results show that the maximum relative error of the rolling force calculated by the model is less than 5%.(2)The simulation experiments of strip rolling process with different work roll bending forces,intermediate roll bending forces and intermediate roll shifting values are carried out by using the controlling variable method.Based on the analysis of the simulation results of rolled flatness,the effects of different flatness control actuators of traditional UCM mill and new HYPER UCM mill on the cross-sectional shape,crown,edge drop and flatness of rolled strip are studied,and the flatness control capacities of different actuators of two kinds of mill are investigated.(3)The actual flatness control effect of a 1450 mm tandem cold strip rolling production line is deeply studied,and the field flatness related data are collected.On the premise of data preprocessing,combined with the multi-layer perceptron network and the extreme learning machine network in the machine learning algorithm,the off-line prediction model of prone area of flatness defects in cold-rolled strip based on field data-driven is established.The prediction accuracy and generalization ability of the two network models are analyzed and compared.The results show that the prediction effect of the model based on ELM algorithm is better than that based on the multi-layer perceptron network,and the off-line prediction effect of flatness in cold-rolled strip with error less than 10%is realized.