Identification of working conditions and prediction of FeO content in sintering process of iron ore fines

Abstract

The iron oxide (FeO) content had a significant impact on both the metallurgical properties of sintered ores and the economic indicators of the sintering process. Precisely predicting FeO content possessed substantial potential for enhancing the quality of sintered ore and optimizing the sintering process. A multi-model integrated prediction framework for FeO content during the iron ore sintering process was presented. By applying the affinity propagation clustering algorithm, different working conditions were efficiently classified and the support vector machine algorithm was utilized to identify these conditions. Comparison of several models under different working conditions was carried out. The regression prediction model characterized by high precision and robust stability was selected. The model was integrated into the comprehensive multi-model framework. The precision, reliability and credibility of the model were validated through actual production data, yielding an impressive accuracy of 94.57% and a minimal absolute error of 0.13 in FeO content prediction. The real-time prediction of FeO content provided excellent guidance for on-site sinter production.