A Machine Learning Model for Quickly Predicting the Inner States of Ironmaking Blast Furnaces

Wenbo Wu,Shibo Kuang,Lulu Jiao,Aibing Yu

DOI: https://doi.org/10.1016/j.powtec.2024.120137

IF: 5.64

2024-01-01

Powder Technology

Abstract:The inner states of ironmaking blast furnaces (BFs) govern their overall performance and thus are crucial for efficient and reliable BF production. However, the current control methods cannot directly consider the inner states because of the difficulty of accessing them. This paper introduces a machine learning (ML) model designed to predict the inner states according to injection parameters promptly. The model employs a modified ensemble learning method using data from a well-established mechanistic model. Two key modifications are implemented. A preprocessing method addresses the low prediction accuracy caused by large gradient data. A stack-based structure improves robustness across various inner states. Comparative analysis shows the proposed model predicts inner states with higher accuracy than existing ML models. Furthermore, the model outputs consistent resolutions while maintaining identical change trends for some key variables. The developed model offers a promising approach for implementing real-time BF prediction.