Model development for prediction of autogenous mill power consumption in the Sangan iron ore processing plant

Abstract

The variables including ore hardness based on the SAG power index (SPI), the particle size of mill product (P80), trunnion pressure of the mill free head (p), and the working time period of the mill liner (H) were considered as variables for the development of an adequate model for the prediction of autogenous (AG) mill power consumption in the Sangan iron ore processing plant. The one-parameter models (SPI as a variable) showed no adequate precision for the prediction of the Sangan AG mill power consumption. Two-parameter models (SPI and P80 as variables), proposed by Starkey and Dobby, showed no adequate precision for the Sangan AG mill power consumption. Nonetheless, by exerting an adjustment factor in the model (0.604513 which was obtained by what-if analysis using the Solver Add-Ins program), the model precision increased significantly (an error of 7.11%). Finally, a four-parameter model in which the Sangan AG mill power consumption is predicated as a function of SPI, P80, p, and H was developed. Hence, initially, the relationship between the mill power consumption and each of the variables was obtained and then the four-parameter model was developed by summation of these four equations and applying a similar coefficient of 0.25 for all of them. This model was modified by finding the best coefficients by what-if analysis using the solver Add-Ins program by minimizing the ARE error function. The error function for the training and testing data sets was determined to be 2.93% and 2.39%, respectively.