As part of the AMIRA P843 project Keeney and Walters (2008) describe how the principal components analysis (PCA) method can be used to identify classes of samples with similar metallurgical characteristics. However PCA does not take account of the location of each sample in 3D space. This paper describes a study to analyse the spatial distributions of the classes and determine appropriate modelling methods to control the estimation of the comminution parameters and mill performance indices into a 3D block model. This includes but is not restricted to geostatistical techniques that are used for grade modelling._x000D_
The starting point for the study is the class definitions and the regression models for estimating the comminution parameters and performance indices. Each sample is assigned a class and the spatial distribution of each class has been analysed using different techniques including the calculation of variograms. This establishes whether the classes can be used as domain control for estimation into the block model - a procedure similar to using geological rock type domains for controlling the estimation of grades for a resource model._x000D_
However, standard geostatistical estimation methods such as ordinary kriging assume that the variable being estimated is additive or linear. This assumption is valid for variables such as grade but this is not necessarily the case for the comminution parameters or the performance indices. Therefore the estimation procedure must either use variables that are definitely additive or the effect of potential non-additivity must be tested and accounted for where possible.The results of the study show how performance indices such as recovery, throughput and specific power can be estimated into the block model._x000D_
This model is then combined with the geological model to provide a multi-parametric model that is used to optimise the mine planning and scheduling activities which can lead to significant improvements in NPV.