Data Mining and Machine Learning to Predict the Sulphur Content in the Hot Metal of a Coke-Fired Blast Furnace

In order to combat climate change, the European Parliament adopted the European Climate Law, making the goal of climate neutrality by 2050 legally binding, in this sense, there has been increased interest in data mining and machine learning in the steel industry. Reducing energy consumption and neutralizing greenhouse gas emissions have become challenges for the steel industry. Nowadays, it is normal to use artificial intelligence to integrate Industry 4.0 technologies to improve and monitor production conditions in the steel industry. In the current scenario of the global economy, strict control of all stages of the production process is of utmost importance to increase productivity and reduce costs, decrease atmospheric emissions and reduce energy consumption. In steel production, the temperature of the hot metal is one of the most important parameters to evaluate, as a lack of control can negatively affect the final quality of the product and increase energy consumption. In this sense, data mining, machine learning and the use of artificial neural networks are competitive alternatives to contribute to solving the new challenges of the steel industry. The database used for the numerical simulation corresponds to 11 years of operation of a blast furnace. For this research, a Big Data with 301,125 pieces of information divided into 75 variables was used. The neural network input consists of 74 input variables and 1 output variable. The conclusion is that data mining and neural networks can be used in practice as a tool to predict and control impurities in the production of pig iron in a blast furnace, to reduce energy consumption, and to reduce the emission of gasses into the atmosphere.