Research on the Prediction Model of Zinc Coating Thickness of Continuous Hot‐Dip Galvanized Strip Steel Based on Gaussian Process Regression

In continuous hot‐dip galvanizing production of steel strips, the zinc coating thickness detector is located ≈150 m away from the air knife control point, resulting in a significant time lag. This leads to insufficient control accuracy of the zinc coating thickness along the entire length of the strip and causes wastage of zinc resources. Leveraging the knowledge of zinc coating formation mechanisms, this study collects substantial industrial production data, processes the data, extracts relevant features, and constructs a dataset. Based on this dataset, a Gaussian Process Regression (GPR) model is developed to predict zinc coating thickness, and its performance is compared with Random Forest and Backpropagation Neural Network models. Under unified feature selection and hyperparameter optimization conditions, the GPR model achieves an R 2 value of 0.9913 and demonstrates significantly superior smoothness and physical consistency compared to the other models. Industrial applications show that the GPR model can rapidly reflect changes in zinc coating thickness in response to variations in air knife parameters. The prediction results based on this model can enhance the real‐time performance and accuracy of zinc coating thickness control.