Active filters :
Discard Filter

Pattern recognition in transformed data – the search for domain shapes in mine geology data


Want a discount? Become a member!

Author S Juras, S McKinley and E Uludag


More and more, today’s mining geologists are asked to identify and create 3D value domains for multidisciplinary usage: grade definition, geometallurgical performance and geotechnical characteristics, to name a few. Often, that domain shape is the result of an assumed or conjectured value as a surrogate to a set of conditions (eg kaolinite-rich, sericite alteration). Would it not be instead better to create the respective domain through identification of a consistent quantifiable geologically-based pattern as its basis?

Critical to the identification of patterns is the transformation of 3D point data into a model by a method that uses few, if any, preselected thresholds. Too often we complicate the reviews of the modelled data by prejudging the value range of interest. And when we do revert to 2D views (ie sections and plans), we tend to only examine numbers or line traces. These hinder and obscure identification of geological patterns particularly in deposits of complex geology. Instead, these data should be looked at with minimal prejudgment and interrogated in 2D space by one of the best and most eye-catching display types: filled contour plots.

Here we present two different processes that transform data with minimal preselection into formats that allow for pattern recognition by filled contour plots. These are: probability assisted constrained kriging (PACK) and multiple interval interpolant formation (Aranz Geo’s Leapfrog® software). The former creates a 3D model of probabilities on a threshold value; the latter creates numerous (≥10), regularly-binned 3D interval interpolants. The results of each method are examined as filled contours over numerous sections or plans. Any distinct patterns are quickly discerned thereby readily identifying the best value to base the creation of a domain. These processes therefore provide a way to marry recent advances in software tools with what is supposed a geologist’s core skill: pattern recognition.


Juras, S, McKinley, S and Uludag, E, 2017. Pattern recognition in transformed data – the search for domain shapes in mine geology data, in Proceedings Tenth International Mining Geology Conference 2017, pp 37–42 (The Australasian Institute of Mining and Metallurgy: Melbourne).