Neural predictor of the end point in a converter

The paper presents a system based on neural networks which is capable of predicting the so—called End Point of a converter by exploiting the estimates of the Oxygen content and of the temperature in order to predict the final Carbon content. This paper presents an alternative approach to the final Carbon content prediction, which is based on neural networks. The system has been trained and tested with real industrial data and its performance overcomes those provided by the analytical model, due to the capability of neural networks of inferring highly non linear relationships from experimental data.