Research on roll profile electromagnetic control ability in optimal electromagnetic stick parameter

Roll profile electromagnetic control technology (RPECT) is a new strip flatness control technology. As the control element, electromagnetic sticks have a great effect on the control ability of RPECT. To improve control ability and extend service life, effective control ratio of electromagnetic stick is presented in this paper. The ratio is designed based on the structure character of electromagnetic stick, and can be used to evaluate the key parameter of electromagnetic stick. Based on the coupled FEM, the heat flux density of the roll inner hole and the temperature distribution of electromagnetic stick are analyzed for different effective control ratios; the average contact pressure between electromagnetic stick and electromagnetic control roll is studied to evaluate the change of force roll profile; the state of roll profile and the stress state of the roll are researched to analyze the comprehensive control ability. Through the verification on the roll profile electromagnetic control experimental platform, the reasonable selection range of effective control ratio, which can be used to expand the roll profile axial affected area, is from 0.5 to 0.583. In order to increase the roll crown, the selection of ηd needs to consider the current density and the optimal selection range of effective control ratio.

Modeling the stress state of steel strip at roller levelling machine under cyclic alternating deformations

The final stage in the production of hot rolled steel is leveling on roller levellers under cyclic alternating deformation. When laser is cutting a sheet it may bend due to the release of residual stresses that are unevenly distributed over the volume. The majority of roller leveller models for calculating the process under cyclic alternating deformation does not provide an adequate assessment and prediction of residual stresses in a steel sheet. On the basis of finite element analysis, formation of residual stresses owing to roller levelling of hot rolled strip is disclosed. The implementation of a model of the levelling process was performed in SIMULIA Abaqus. Models are verificated by comparing forces under the rollers. We have experimentally confirmed the convergence of the simulation results with the measurements of the strip flatness obtained after sheets plasma cutting. It was found that after levelling, tensile longitudinal residual stresses remain on the upper surface of the sheet, compressive ones remain on the lower surface, stresses are zero in the middle in thickness, and the stress values are opposite in sign in the remaining parts of the section. It was established that the same parameters of the levelling process of different strength categories lead to different deviations of stresses. An increase in yield strength of the strip leads to an increase in the deviation of residual stresses along the strip thickness. The proposed method of simulation of roller levelling process should be used to study the stress-strain state of hot-rolled steel and to design improved strip levelling setting modes with minimal residual stress deviations.

Self-supervised pre-training of CNNs for flatness defect classification in the steelworks industry

Classification of surface defects in the steelworks industry plays a significant role in guaranteeing the quality of the products. From an industrial point of view, a serious concern is represented by the hot-rolled products shape defects and particularly those concerning the strip flatness. Flatness defects are typically divided into four sub-classes depending on which part of the strip is affected and the corresponding shape. In the context of this research, the primary objective is evaluating the improvements of exploiting the self-supervised learning paradigm for defects classification, taking advantage of unlabelled, real, steel strip flatness maps. Different pre-training methods are compared, as well as architectures, taking advantage of well-established neural subnetworks, such as Residual and Inception modules. A systematic approach in evaluating the different performances guarantees a formal verification of the self-supervised pre-training paradigms evaluated hereafter. In particular, pre-training neural networks with the EgoMotion meta-algorithm shows classification improvements over the AutoEncoder technique, which in turn is better performing than a Glorot weight initialization.

Multi-Objective Optimization of Intermediate Roll Profile for a 6-High Cold Rolling Mill

The multi-objective optimization of the SmartCrown intermediate roll profile for a cold rolling mill was proposed in this paper in order to improve the strip flatness quality. A coupling model of roll profile and strip flatness was established, and the roll gap profile, roll gap crown adjustment range, rolls contact pressure, and strip flatness under different intermediate roll profile parameters were calculated based on the coupling model. The results showed that the roll gap crown adjustment range and rolls contact pressure difference increased with increasing roll profile parameters. The roll profile parameters were multi-optimized based on the non-dominated sorting genetic algorithm II (NSGA-II). The minimum rolls contact pressure difference and maximum roll gap crown adjustment range were taken as the objective function of multi-objective optimization. The optimal roll profile parameters were applied to a six-high five stand tandem cold rolling mills, which improved the flatness quality of the DP780 steel strip.