Mathematical Modelling to Predict Fuel Consumption in a Blast Furnace Using Artificial Neural Networks

Abstract

The main form of processing of reduction of iron ore is by the blast furnace and although the principles of pig iron production are the same as those of a century ago, technology and the understanding of how the blast furnace works have evolved a lot from that time to the present day, ally with technological advances that allowed important evolutions in the monitoring of the process and procedural changes. Considering the importance of the pig iron production stage, for the production chain, there is a need for studies and the search for tools that can optimize or help control the process of obtaining it and consequently the production costs. Thus, considering the importance and complexity of operation of the blast furnace, combined with the potential reached by neural networks in process optimization, the purpose of this paper is to develop a source code based on artificial neural networks in the format of a committee machine to perform the monitoring the operation of the blast furnace and the prediction of results related to fuel consumption. The data set used in this research comes from the operation of blast furnace 1 of the company ArcelorMittal and the operational data correspond to 150 days of operation. Concluding, the high values of correlation mathematical show the good statistical performance of ANN and it shows that the mathematical model is an effective predictor and the results obtained in conjunction with the cross-validation of the data demonstrate the ANN's ability to generalize the acquired knowledge.