Dynamic Regional Viscosity Prediction Model of Blast Furnace Slag Based on the Partial Least-Squares Regression

Abstract

Viscosity is considered to be a significant indicator of the metallurgical property of blast furnace (BF) slag. A model for viscosity prediction based on the partial least-squares regression of varietal quantity reference points is presented in this article. The present model proposes a dynamic regional algorithm for reference point selection. The study applied the partial least-squares regression to establish the dynamic regional viscosity prediction model on the basis of limited discrete points data. Then an actual prediction was carried out with a large amount of viscosity data of real and synthesized BF slags that was obtained from a certain steel plant in China. The results show that this advanced method turns out to be satisfactory in the viscosity prediction of BF slags with a low averaging error and mean value deviation.