Research on data-driven continuous casting billet quality prediction modeling approach

The unreasonable use of process control procedures and composition ratios in the continuous casting production process will have an extremely bad impact, resulting in low steel qualification rate, smelting process energy consumption and other serious problems. In response to the above problems, a continuous casting billet quality prediction model is established from the causes of continuous casting billet quality defects, incorporating data-driven. The process parameters corresponding to the two continuous casting billet qualities, AQ0E1 and DV0A1, were first preprocessed using data cleaning and dimensionality reduction methods. A data-driven approach was adopted to establish a continuous casting billet quality prediction model using shuffling frog leaping algorithm (SFLA), genetic algorithm (GA) and BP neural network to form a hybrid evolutionary neural network (HENN) continuous casting billet quality classification prediction model. The prediction accuracy of AQ0E1 and DV0A1 can reach 88.33% and 90.38%, respectively. The model in this paper has higher accuracy and good stability than traditional methods, the continuous casting billet can be accurately predicted, it is convenient to quickly trace back to the cause of defects in continuous casting billet, and it has a certain guiding significance for the development of intelligent continuous casting industrial production.