Liquid phase fluidity study of iron ore fines based on improved CondInst

The liquid phase flowability of iron ore powder affects the quality of sintered ore. To better explore the liquid phase flow law of iron ore powder, this paper first adopted the improved CondInst (Conditional Convolutions for Instance Segmentation) to segment the image and achieved a segmentation accuracy of 96.61%. The segmentation accuracies on ResNet50 as well as ResNet101 were improved by 0.11% and 0.32%, respectively, relative to the original model. The image's height, area and wetting angle were used as the characteristic indexes of iron ore powder melting. The fitting curve was established by combining temperature and time to characterise the whole process of iron ore powder melting. Second, Factsage was used to simulate the liquid phase generation of iron ore powder, and a CatBoost regression model based on the liquid phase generation was constructed, and the maximum error between the predicted value and the real value was 3.74%. Finally, based on the equivalent mobility characteristic number, the liquid phase mobility performance of iron ore powder and the mechanism of alkalinity's influence on it were comprehensively analysed.