Abstract
Geometallurgical models predicting throughput/comminution performance of the processing plant often rely on hardness models based on sparse information and typically do not interact with production schedules while ignoring the non-additive nature of hardness.
A new approach integrating a prediction model of ball mill throughput into short-term production scheduling is proposed to overcome non-additivity issues using sensor and production data collected in a mining complex. First, the comminution behaviour of the mineral reserves is simulated by building additive hardness proportions from penetration rates of blasthole drilling. Material tracking through the mining complex, including stockpiling, is performed next to obtain hardness proportions and other rock properties of blended materials processed in the comminution circuit. Finally, a multiple regression model is used to predict throughput rates based on blended rock properties, which is then integrated into a stochastic optimization model for short-term production scheduling. The result is a weekly production schedule that can achieve planned production better by matching the scheduled materials with the predicted performance of the ball mill.