Original Article

Prediction of end-point phosphorus content in converter based on integrated transfer learning strategy

¹ School of Automation and Electrical Engineering, Inner Mongolia University of Science and Technology, Baotou 014000, Nei Mongol, PR China
² Infrastructure Department of Inner Mongolia University of Science and Technology, Baotou 014000, Nei Mongol, PR China

^* e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

Received: 28 April 2025
Accepted: 12 May 2026

Abstract

The prediction of phosphorus content in molten steel at the end of converter is of great significance to control the quality of steel and reduce the production cost. In order to achieve accurate prediction, aiming at the problem of small sample size and poor stability of single model, this paper proposed a prediction method of terminal phosphorus content combining Dung beetle optimization (DBO) algorithm and ensemble transfer learning strategy(EL-TL). Firstly, the improved Dung Beetle optimization algorithm (IDBO) was improved to optimize the hyperparameters of the Multi-layer Perceptron (MLP). Secondly, the transfer learning strategy was introduced and the multi-kernel maximum mean difference was improved to solve the problem of insufficient prediction accuracy of small samples. Finally, the ensemble learning strategy and adaptive weight improvement were introduced to further solve the problem of insufficient stability and easy overfitting of a single model. The ablation experiment verifies that the proposed method achieves 85% hit rate within the error range of plus or minus 0.002%, which can provide guidance for practical production.