Gmdh Method Identification And Control Of Blast Furnace Ironmaking Process

Gang Zhou,Xiangguan Liu,Li Zhou,Shibo He

2010-01-01

Abstract:One of the most challenging jobs in Blast Furnace (BF) ironmaking process is the temperature control, due to the fact that there are hundreds of variables interacting with each other which have impacts on the temperature evolution of BF, and there does not exist an efficient method to characterize such interactions. All these make most existing approaches failed to work for temperature prediction in practice. In this paper, we propose to introduce a mathematical model by using GMDH method. Our proposed model is based on the theory of self-organizing data mining, and has capability of analyzing the relationships between the variables. With such explicit relationships we establish our model to predict and control the silicon content ([Si]) in hot metal, which is widely recognized as the most efficient indictor of temperature evolution. Through an experimental example, we show that our proposed model has a good performance of predicting [Si], with 86.9% percentage of target hitting in applications. Moreover, we discuss the relationships between variables obtained in the model, and interpret their implications in actual applications. All our observations lead to the conclusion that our proposed model based GMDH method can work as a powerful tool to facilitate the automatic control in BF ironmaking process.