Dynamic Modelling of SAG Mill Power Draw Using Neural Network Approach

Modelling and optimisation of the power draw of large SAG mills are important due to the large power draw which modern mills require (5~10 MW). The cost of grinding is the single biggest cost within the entire process of mineral processing. Traditionally, modelling of the mill power draw has been done using empirical models. Although these models are reliable, they cannot model mills and operating conditions, which are not within the model database boundaries. Also, due to their static nature, the impact of the changing conditions on the power draw, within the mill, cannot be deter-mined using such models. Despite advances in computing power, on-line modelling of large plant scale mills could be a time consuming task. On the other hand, using on-line parameters for model-ling, result in more reliable control of SAG mill without performing any more costly tests. In this paper, SAG mill power draw is modelled using neural network approach. Six inputs, namely: solid percent, two bearing pressures, noise, recycle and feed rate were used to model the power draw. The produced model shows an acceptable correlation coefficient, R2 =0.9623.