Original Article

A hybrid PCA-SOA-BP approach for predicting converter endpoint temperature in steelmaking

¹ State Key Lab of Advanced Metallurgy, University of Science & Technology Beijing, Beijing 100083, PR China
² Jianlong Xilin Iron & Steel Co. Ltd, Yichun, 153000, PR China
³ School of Metallurgical and Ecological Engineering, University of Science and Technology Beijing, Beijing 100083, PR China

^* e-mail: zhaolihua@metall.ustb.edu.cn

Received: 12 December 2023
Accepted: 30 September 2024

Abstract

Accurately controlling the temperature of the converter end is a crucial element of the steelmaking process. To enhance the accuracy of predicting the converter end temperature, we propose a hybrid model that utilizes principal component analysis (PCA) and snake optimization algorithm (SOA) in conjunction with a backpropagation algorithm (BP) neural network. The 16 parameters for smelting in converter steelmaking were reduced using principal component analysis to remove shared features. The nine principal components derived from this analysis were then used as input variables for an optimized BP neural network. An optimization algorithm was then employed to refine the initialized weights and thresholds of the BP neural network. The impact of neuron node quantity in the hidden layer on the BP neural network was examined. Results show that the ideal BP neural network is achieved with 19 neuron nodes in the hidden layer. Compared with ordinary BP neural network, PCA-BP neural network and SOA-BP neural network, the model proposed in this study can predict the end temperature of converter most accurately. In the temperature error range of ±10 °C and ±15 °C, the prediction accuracy of the model is 93% and 96%, respectively. Meanwhile, the model has been effectively applied in the industrial production of a steel plant in China. The results show that the prediction results of the model are in good agreement with the actual production data in the field. This accurate prediction can optimize the field operation process and realize the stable control of product quality.