Power draw estimation of ball mills using neural networks

Abstract

Estimating the power draw of a ball mill is of great importance from both operational and control standpoints. There are many factors that affect the power draw, and its prediction is particularly difficult in an operating plant. The important factors that affect power draw are the mill dimensions and operating parameters. Although mill dimensions remain the same for any given mill, the operational factors, such as ball load, mill filling and mill speed, may vary. This makes the modeling effort extremely cumbersome. It is for this reason that an artificial neural network is considered for the development of a black-box-type model to predict power draw. In this work, a total of 48 sets of plant data are used. Using these data, a neural network is built and trained. The trained network is used to predict power that is found to be in good agreement with operating mill power for many situations. In addition, the results of simulations are also compared to those obtained from a discrete-element method-based power-draw model.