Application of machine learning algorithms for prediction of sinter machine productivity

Sinter machine productivity is key techno-economic parameter of an integrated steel plant. It depends upon the composition of different constituents like iron ore fines, flux and coke breeze which are agglomerated to produce sinter for blast furnaces. It is difficult to assess the interdependence of these constituents and their effect on sinter productivity through physical experimentation. In this paper, machine learning and data analytics approach have been applied to predict the sinter machine productivity. Industrial data of sinter machine productivity from an integrated steel plant have been collected. Linear regression and artificial neural network (ANN) models were developed to predict sinter machine productivity with the composition of constituent materials of the agglomerate as model inputs. The ANN model, developed in the present work, agrees well with measured sinter machine productivity. Sensitivity analysis identified that, percentage of MgO in flux and CaO in sinter have a highly detrimental effect whereas total Fe content in iron ore fines and percentage of SiO 2 in sinter have the most favorable impact on sinter machine productivity. •Sinter Plant Productivity can be efficiently predicted by neural network models.•Sensitivity analysis identifies important parameters affecting output.•Percentage of MgO in flux has the most adverse effect on sinter plant productivity.•Percentage of SiO2 in sinter is highly beneficial for sinter plant productivity.