Bulk sampling of materials is commonly performed using a sampling device that passes through the stream of material flowing from the ends of conveyor belts. The accuracy of the sampling is very difficult and expensive to establish experimentally. DEM modelling allows any bias to be simply and cheaply evaluated. The major advantage of DEM for this application is not its ability to predict the composition of the sample taken, but its ability to determine what the reference sample should be. The reference sample is the imaginary sample to which the real sample is compared to assess the bias and consists of all the particles that should have been sampled. The difference between these is that all particles were not sampled because of their interactions with the cutter and the resulting interactions with other particles. Previously, modelling of sampling was in 2D and needed large ensembles (Robinson and Cleary, 1999).

Figure 2 shows a sampling system consisting of a conveyor belt, head pulley, and sampler. The particles are initially shaded mid-grey. As they pass through a sampling plane near the head pulley, their trajectories are assessed and particles that should be sampled are re-shaded light grey and form the reference sample. At the end of the simulation, the actual sample is compared to the reference sample. In this case, several larger particles from the reference sample are seen to be deflected away from the sampler opening and escape leading to bias against the larger particles. This experiment...