Content area
Full text
Journal of Wuhan University of Technology-Mater. Sci. Ed. Feb.2012 187
DOI 10.1007/s11595-012-0433-3
Prediction of Free Lime Content in Cement Clinker Based on RBF Neural Network
YUAN Jingling1, ZHONG Luo1, DU Hongfu1, TAO Haizheng2*
(1.School of Computer Science and Technology, Wuhan University of Technology, Wuhan 430070, China; 2.State Key Laboratory of Silicate Materials for Architecture (Wuhan University of Technology),Wuhan 430070, China)
Abstract: Considering the fact that free calcium oxide content is an important parameter to evaluate the quality of cement clinker, it is very signicant to predict the change of free calcium oxide content through adjusting the parameters of processing technique. In fact, the making process of cement clinker is very complex. Therefore, it is very difcult to describe this relationship using the conventional mathematical methods. Using several models, i e, linear regression model, nonlinear regression model, Back Propagation neural network model, and Radial Basis Function (RBF) neural network model, we investigated the possibility to predict the free calcium oxide content according to selected parameters of the production process. The results indicate that RBF neural network model can predict the free lime content with the highest precision (1.3%) among all the models.
Key words: RBF neural network; cement clinker; free lime content
1 Introduction
Considering the fact that the manufacturing process of cement clinker is very complex, the association between the quality of cement clinker and the parameters of the making process is highly nonlinear. Therefore, using a conventional mathematical model, it is very difcult to describe this complex association. One model is to transform the highly nonlinear relationship into a linear one through various conversion methods. However, the accuracy of prediction through this method is very low, which is not so significant to predict the quality of cement clinker according to the parameters of manufacturing process. Another model is based on the neural network, which has the advantage of describing the nonlinear law with rather high precision[1-7].
Free lime content is a key index to evaluate the quality of cement clinker. Therefore, it is very signicant if the free lime content is precisely predicted
according to the parameters of manufacturing process. Utilizing Back Propagation (BP) neural network model, the reference[7] pointed out that the free lime content can be predicted with a high precision...