Prediction of Iron Ore Sinter Properties Using Statistical Technique

Due to the drastic change in iron ore qualities, maintaining consistency in sinter property has become a challenge for steel manufacturing industries, resulting the irregularities and disturbances in the blast furnace iron making. The present work aimed to develop a prediction model for physical, mechanical and high temperature properties of iron ore sinter well in advance from the plant process parameters to support smooth functioning of the furnace. Correlation matrix, multiple linear regression and artificial neural network (ANN) analysis were performed for the development of the model. This predictive system for three important sinter properties [Mean particle size (MPS), tumbler index (TI) and reduction degradation index (RDI)] was established based on back propagation (BP) neural network, which was trained and tested by actual plant data. The prediction accuracy of MPS, TI and RDI using that ANN model is 79, 91 and 76 %, respectively. A user-friendly graphical user interface has been developed using C language based on the model. The application results showed that the prediction system had high accuracy rate, stability and reliability.