Comparison of temperature control time prediction models for wide and thick plates based on machine learning

• Main Authors: Zhang, Zhuolun, Li, Tieke, Wang, Bailin, Yuan, Shuaipeng
• Format: Conference Proceeding
• Language: English
• Subjects: Computational modeling; Feature extraction; feature selection; machine learning; Neural networks; prediction models; Predictive models; Robustness; Stability analysis; Temperature control; temperature control time
• ISSN: 2473-3547
• DOI: 10.1109/ISCID56505.2022.00014

Temperature control time is a key parameter in the hot rolling process of wide and thick plate. Accurate prediction of temperature control time can help to make efficient production scheduling. The features related to temperature control time are extracted from the historical rolling data, and the classified features are converted into numerical values by using the target variable coding method. All the sample data are normalized, and the Pearson correlation coefficient between all the features is calculated to remove redundant features and features with low correlation with the target variable. Using mean absolute error, root mean square error and coefficient of determination as evaluation indicators, the accuracy and robustness of five machine learning models in predicting temperature control time are compared and analyzed. The experimental results show that the BP neural network has the best performance.