Research on Thickness Error Optimization Method of Rolling System Based on Improved Sparrow Search Algorithm–Bidirectional Long Short-Term Memory Network–Attention

Abstract

With the development of technology, the working processes of rolling equipment have become more and more complex, and the traditional rolling model encounters difficulties in meeting current demands for accuracy. To reduce the thickness error of the rolling system, we propose a high-precision rolling force prediction method based on SSA–Bilstm–Attention, which reduces the thickness error of the rolling system by predicting the high-precision rolling force. Firstly, a mechanical model is established, and the parameters involved are analyzed to extract suitable parameters as inputs to the network to reduce the feature loss of the network inputs. Secondly, an improved sparrow search algorithm is used to search for the hyperparameters of the network to obtain better training results. Finally, the attention mechanism is introduced to increase the network’s training accuracy. A stochastic small-batch gradient descent method is used to improve the training speed of the network. In addition, this paper establishes a web-based host computer, which provides an effective data source for the experimental analysis. The experimental results show that the optimized model has a mean square error of 1.22%, which is more accurate than other models, and has good generalization ability. The experiments confirm the method’s effectiveness in improving the thickness accuracy of the rolling system and provide a new optimization scheme for the industry.

Keywords

• rolling model
• improved sparrow algorithm
• Bilstm
• attention mechanism
• rolling force prediction