A comparative study of black-box models for cement quality prediction using input-output measurements of a closed circuit grinding

This paper presents the methodology of design of three different modeling techniques for predicting cement quality using input-output measurements of the closed circuit grinding in a cement plant. The modeling approaches used are: statistical, artificial neural networks (ANN), and adaptive neuro-fuzzy inference systems (ANFIS). The data set for generating the predictive models are obtained from a database of the operation of the cement plant, UCEM-Guapan. An OPC (OLE for process control) network configuration in the SCADA system allows online validations of the proposed models in order to select the best approach for real-time prediction of cement quality.