Research on the BOF steelmaking endpoint temperature prediction

Abstract

One aim of basic oxygen furnace (BOF) steelmaking endpoint control is the temperature control. For the majority of the china's small or medium BOF, sublance can not be used as a result of restrictions of production conditions, so, researching the BOF endpoint control without sublance has a significant application value. For the data's characteristics of nonlinearity and high noises in the field, a new algorithm combining the DBSCAN clustering algorithm with the RBF neural network algorithm was proposed. It was used to achieve effective treatment of the noises, furthermore, it will be applied to the prediction of endpoint temperature of BOF steelmaking. Simulation results show that the RBF neural network with DBSCAN has more advantages than the original RBF neural network in dealing with sample sets of high noises. The new RBF neural network has a higher hit rate, which improved the versatility and practicality of RBF neural network.