FeO Content Classification of Sinter Based on Semi-Supervised Deep Learning

Sinter is the main raw material for blast furnace ironmaking. The FeO content in sinter is used as an important indicator for evaluating the quality of sinter, especially the strength and reducibility of sinter. In view of the existing sinter FeO content detection methods that have strong hysteresis, reliance on manual experience, and possible cognitive limitations, this paper uses Dense Convolutional Network (DenseNet), taking the section image of sintering machine tail as input to classify FeO content. For the problem of insufficient labeled data, the semi-supervised learning method FixMatch is used to train the model to achieve accurate classification of FeO content. The experimental results indicate that the FeO content classification results are within an acceptable range and can be fed back in real time at the sintering production site. The use of the above methods also provides references for sintering production and related research.