Research on Zinc Layer Thickness Prediction Based on LSTM Neural Network

Hot-dip galvanizing is a widely used steel anti-rust method. By immersing clean steel products in molten zinc solution, a uniform zinc layer is attached to the surface of steel products, which can slow down the corrosion of steel products to a certain extent. In the products produced by the hot-dip galvanizing process, the thickness, uniformity and firmness of the zinc layer are important technical performance indicators to measure product quality. In this paper, using the advantages of LSTM neural network in processing long sequence information, a zinc layer thickness prediction model based on LSTM neural network is proposed, and the zinc layer thickness prediction model is established based on the data of hot-dip galvanizing production line. The experimental results show that the mean square error of the test set prediction results is 2.790, the average absolute error is 1.359g/m2, and the average absolute percentage error is 1.824%, which achieves effective prediction of zinc layer thickness and proves the feasibility of applying LSTM neural network to zinc layer thickness prediction Sex. Compared with the multiple linear regression model, the MSE, MAE, and MAPE of the LSTM neural network model are better than the multiple linear regression model, and have higher prediction accuracy.