No abstract

No abstract

To meet the “dead flat” transverse thickness profile requirement of electrical steel strip in 5‐stand 6‐high universal crown mill (UCM) tandem cold rolling mills, a random forest predictive model for electrical steel strip profile is established to reduce the dimension of the data and make predictions for the strip transverse thickness difference (TTD), based on the collected and processed real‐time data of the steel strip transverse thickness profile and multi‐methods of the industrial production mills. The control strategy of multi‐methods including positive and negative hydraulic work roll bending system (WRB), positive hydraulic intermediate roll bending system (IRB), and hydraulic intermediate roll shifting system (IRS) of different stands in 5‐stand 6‐high mills is proposed by comprehensively considering association rules mining which shows the optimal combination of ranges of key control parameters and analysis of 5‐stand 6‐high mills with the developed edge drop control work rolls for non‐shifting of work rolls (EDW‐N) with divided width groups technology on stand No. 1 and No. 2. The strategy is continuously and stably applied to 1420 mm 5‐stand 6‐high UCM tandem cold rolling mills and shows remarkable results. The rate of TTD less than or equal to 7 μm increases from 38.58% to 67.74%.

To meet the shape quality requirements of “Dead flat” rectangular section of electrical steel in the cold rolling process, the transverse thickness difference (TTD) prediction model of 6‐high tandem cold rolling mills (TCMs) based on genetic algorithm, particle swarm optimization, and support vector regression (GA‐PSO‐SVR) is proposed. The TTD prediction model uses 5000 coils of data obtained from a cold rolling line. The GA‐PSO‐SVR model is obtained by the GA‐PSO hybrid algorithm to search and obtain the optimal parameter of the SVR to improve the prediction model accuracy. The results reveal that using multiple evaluation indicators the GA‐PSO‐SVR model has preferably adaptability and higher accuracy. Meanwhile, the relative importance of input variables is calculated based on the GA‐PSO‐SVR model, which indicates that the shape control methods in stands 1–5 have the most important influence on the TTD. The TTD prediction model is continuously applied to 1420 mm 6‐high TCMs; the results show that the rate of the TTD less than 7 μm increased from 29.6% to 63.85%.