Comparing Power and Performance Trends from Dem and Jk Modelling

This paper compares Discrete Element Model predictions for power and variations in impact and abrasion (low angle impact) breakage with power and grinding rate predictions from the more "traditional" JKMRC models. The DEM predictions are based on measured mill load sizings from a large industrial SAG mill at two operating conditions. The load sizings can also be predicted with reasonable accuracy by the JK SAG model from feed sizing, ore characterisation and mill geometry and operating conditions. Opportunities for a "hybrid" model are discussed. This model might use JK ore characterisation and traditional models to predict average performance mill load. A discrete element model could then use this load as a base case to predict more detailed variations such as those resulting from changes to lifter geometry. This paper reports on progress and problems encountered to date. BACKGROUND A comparison of charge flow and velocity patterns in a scale model autogenous mill demonstrated that three dimensional discrete element modelling could provide an accurate description of the flow patterns over a wide range of mill loads, mill speeds and liner conditions. This comparison is reported in Cleary, Morrison and Morrell (2001). It also IV-285 suggested that 2D DEM could not provide an adequate description. The accurate motion description encouraged the authors to compare a well characterised full scale mill with a full scale DEM simulation. This paper reports on progress and problems encountered to date. COMPARISON OF DEM AND JK MODELS DEM considers the motion of each particle as a consequence of the interactive forces with the other particles and the mill liner. The standard approach to collision modelling is to use a spring and dashpot model for normal and shear particle interactions. The CSIRO approach is described in Cleary (1998, 2001). The first step for a DEM model is a description of all the particles in the mill or at least in a slice of it. The enormous increase in the power of desktop computers allows simulations with several hundred thousand particles to be completed in a “reasonable” time – days to months – depending on how long the model takes to reach steady state. A 0.5m slice of an 11m (36 foot) diameter mill typically contains about 200,000 particles larger than 25mm. Extending the model to 12mm would typically double the number of particles to be considered. Hence, computational time is still an issue for large industrial mills and a 1.8m diameter by 0.6m pilot mill is worth considering as a modelling target. Cleary (2001) has shown that truncating the mill load size distribution for a ball mill at between 1 and 10 mm has minimal effect on flow patterns within the mill. For the cases considered in this paper, the charge has been truncated at 25 mm to keep the number of particles in the calculation to less than 200,000. The DEM model estimates the forces acting on each particle during every collision. The forces are integrated over distance to estimate the energy absorbed in each collision. If we can estimate the breakage that would result from the energy expended during each collision, then mill production can be estimated. The JKMRC has been characterising single particle breakage (Narayanan and Whiten, 1983, Leung et.al. 1987 and Banini 2001) for more than 20 years. These characterisation techniques relate the degree of breakage to input energy levels. These characterisation techniques have been licensed world-wide and are in common use for AG/SAG mill optimisation and design. JK breakage characterisation considers both impact and abrasion breakage which might reasonably be considered to be related to normal and shear forces respectively.