Data-driven Multivariate Analysis of Basic Oxygen Furnace Used in Steel Industry

This paper presents early results of the development steps taken in order to build a static model of a basic oxygen furnace for SSAB Raahe steel plant. The model aims for being used as a part of a plant-wide control system. Model development includes pre-processing and actual modelling steps. The modelling step uses a pre-selected set of input variables and non-linear multivariate regression models. The results show that the models for Mn, P and S changes are able to capture the main interactions even though the variance around the main trend is high. The prediction of Fe changes, however, is not successful. This research indicates that grouping or clustering of data or usage of other modelling techniques may improve the results greatly.