Updating Simulated Multi-Attribute Orebody Models Using High-Order Statistics Considering Production Data of Multiple Support Scales

Abstract Operating mining complexes constantly collect data from a wide variety of sources that directly or indirectly measure pertinent geological and geometallurgical attributes of mined material. These attributes are critical to processing performance and the quality of sellable products. Thus, integrating new information to rapidly update simulated orebody models of these attributes facilitates adaptive decision-making within the context of stochastic optimization for short-term production planning. The proposed framework updates multiple correlated attributes of simulated orebody block models given production data of two types. The first includes spatial measurements at point-support scale such as blasthole assays or drill penetration rates. The second type includes processing performance measurements such as mill throughput or metal recovery which depend on the properties of blended materials. The measurements are used as additional conditioning data to construct posterior probability distributions for attributes of nearby blocks. Within a block, values for multiple attributes are updated sequentially, with each attribute being used to condition the next. The high-order updating framework reproduces complex geological patterns and does not make any assumptions on the underlying distributions. A case study using a multi-element iron deposit illustrates the performance of the updating framework.