ABSTRACT The rolling process is one of the most important and widely used metal forming methods, applied to produce various products such as plates, rods, and profiles. In this research, due to the critical role of the main electromotor current in the hot rolling process and its dependence on several parameters, a multilayer perceptron artificial neural network (ANN) was utilized to predict this current. For model training, data on thickness reduction and electromotor current were collected directly from the rolling mill. Specifically, the ANN was trained using nine independent experimental datasets for the production of 6 mm plates and 11 datasets for 10 mm plates. After training, the model was employed to evaluate approximately 10,000 different thickness reduction regimes, ultimately identifying the optimal one that led to the minimum electromotor current. The predicted values from the ANN were compared with analytical results. In the middle passes of the rolling process, the maximum deviation observed was 10% for the 6 mm plate and 14% for the 10 mm plate, indicating a good level of agreement and confirming the reliability of the model. To quantitatively assess the accuracy of the ANN in predicting motor current, three statistical indices—Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the coefficient of determination ( R 2 )—were calculated. In both scenarios, R 2 values exceeded 0.99, while the 10 mm plate case showed lower MAE and RMSE values, suggesting better model precision and stability in thicker sheet production. Additionally, sensitivity analysis showed that the initial passes had the highest impact on electromotor current, especially in the 6 mm scenario. This highlights the necessity of precise control during the early stages of rolling thinner sheets. In contrast, 10 mm plate rolling exhibited a more uniform distribution of influence and greater overall process stability.
Afshari et al. (Sun,) studied this question.