Prediction of rolling force deviation in the hot rolling process of strip steel based on LSTM-attention and process knowledge

In the hot strip rolling process, factors such as product diversity and frequent roll changes lead to the challenge of predicting rolling force under non-steady-state working conditions. This paper proposes a hybrid artificial neural network model based on Long Short-Term Memory (LSTM) and attention mechanisms to predict the horizontal force during hot strip rolling. First, a mathematical model for rolling force prediction is established, and the key parameters influencing the rolling force are thoroughly analyzed. Then, an attention mechanism is introduced to enhance the model’s ability to capture important information from the input sequence, improving prediction accuracy. By combining LSTM and attention mechanisms, a hybrid neural network model is constructed, effectively modeling the complex nonlinear relationships in the rolling process. Experimental results show that the proposed hybrid model significantly improves prediction accuracy compared to traditional methods. The mean of the predictions is close to the actual mean, and the variance of the predictions is smaller, indicating that the model is highly stable and reliable. Additionally, the model demonstrates strong generalization ability, making it well-suited to predicting rolling force under varying working conditions. This study not only enriches the theoretical framework for modeling the hot strip rolling process but also provides a scientific basis for quality control and energy reduction in actual production. In the future, we will further optimize the model structure and explore the application of more advanced neural network techniques to promote the intelligent development of hot strip rolling technology.