Quality Prediction of Hot Rolled Products and Optimization of Continuous Casting Process Parameters Based on Big Data Mining

The continuous casting-rolling process is widely used due to the lower energy consumption and its compact process. However, the defect in continuous casting slab severely limits the development of the continuous casting-rolling technology. In this paper, based on big data mining technology, the quality prediction model for hot rolled coils and the corresponding optimization method for continuous casting parameters were proposed. Firstly, the GA (Genetic algorithm)-BP (Backpropagation) neural network prediction model with high accuracy was constructed according to the characteristics of actual production data. Then, the effect of continuous casting parameters on the probability of defects occurrence was investigated with the established model. The results show that the defects occurrence probability decreases firstly and then increases with the casting speed, as well as the temperature of molten steel, which are consistent with metallurgical theory. Meanwhile, the optimum casting speed and mold level are 1.3 m/min and 8200 mm, respectively. When the flow rate of argon blowing for stopper and nozzle are restricted to 8.5 and 8 L/min, the defects occurrence probability will be lower. Furthermore, the optimum critical values of temperature difference of mold cooling water and inlet temperature are 8 °C and 35 °C, respectively. This paper can provide the guidance for narrow range control of continuous casting parameters and contribute to the production of high-quality steel.