Basic Oxygen Furnace (BOF), TBM type (Thyssen – Blas – Metallurgie)is one of the heat units occurring in a steel production process. The refractory lining of BOF consists of several zones and is lined with MgO-C bricks. For the above mentioned zones refractories with different properties are selected due to the different factors influencing the corrosion process. Intense wear of refractories is observed mainly at the slag spout zone in accordance to the influence of thermochemical, thermomechanical factors (including the oxidizing atmosphere).
The aim of this paper is to find the regression formula with satisfactory forecast measure of fit, which will make it possible to predict the refractory material wear in the slag spout zone of BOF depending on the real wear measurement made during the BOF operation. Calculations were conducted with the use of regression trees with CART algorithm (Classification and Regression Trees), Multivariate Adaptive Regression Splines (MARS), Boosted Trees algorithm and Multilayer Neural Networks MLP type (Multilayer Perceptron).Selected metallurgical parameters registered during the BOF campaign are the independent variables discussed in refractory material wear models, whereas the wear rate of refractory materials calculated per one heat is set as a dependent variable

Keywords: Basic oxygen furnace, Refractories, Machine learning, MgO-C