Predicting Iron Ore Losses and Dilution Factors Using Conditional Simulations

The local cumulative distribution function (lcdf) of a selective mine unit (SMU) can be estimated by conditional simulations using the statistics derived from the sample points, generating multiple equally probable scenarios. Conditional simulations permit the evaluation of the lcdf at any support, taking into account the spatial variability, ie the variogram, and the two statistical moments, mean and variance, of the conditioning samples. The SMU ore losses and dilution depend on the mining method and mining equipment selectiveness that, in its turn, depends on the predicted block size. A new approach to predict ore loss and dilution was developed in this study. Monte Carlo simulation is used to randomly generate a grade value derived from the lcdf at the mining stage (Zs_mc(t)) in order to make a decision about the SMU destination, ie processing plant (ORE) or waste dump (WASTE). If Zs_mc(t) is lower or equal to the cut-off (Z_cut-off) the block is ORE and, conversely, if Zs_mc(t) is higher than the cut-off (Z_cut-off) the block is WASTE. Considering these binary destinations ORE/WASTE, at the mining time t, and combining each realisation of the simulated random function Zcs_i (X), there are four possibilities for each SMU simulated value: 1. the block was simulated as ore and is actually ore (OO), 2. the block was simulated as ore and is actually waste (OW),3. the block was simulated as waste and is actually ore (WO), and 4. the block was simulated as waste and is actually waste (WW).The dilution factor is given by possibility 3 (WO), while the ore loss factor is represented by possibility 2 (OW). These results comprise the classical classification and reconciliation plots used in the mining industry. The new procedure proposed in this paper is based on the use of all the equally probable simulations to define the ore loss and dilution factors. An application of this methodology was carried out at Vale's iron mine in Brazil. The grade of silica was simulated and 50 multiple equally probable 3D models were generated. The mean and the standard deviation of the ore tonnage (OO + WO) for the 50 realisations, was 233 Mt and 0.5 Mt, respectively. The mean grade increased for the diluted ore (OO + WO) as well as their respective coefficient of variation (CV). The CV increases from 62 per cent in the undiluted ore (OO), to 101 per cent in the ore, considering the dilution factor (OO + WO). The mean ore loss (OW) was 10 Mt and the waste tonnage dilution (WO) was 9 Mt.