Boundary Classifi cation for Automated Geological Modelling

摘要

Chemical species distributions in rock sequences can be modelled using Gaussian processes (GPs) tornpredict the weight percentage between known data points. This method can be improved by dividingrnthe modelled area into regions of different mineralogy. In the context of many iron ore deposits,rnmarker shales give the initial guidance of where these regions may lie. These shales give rise torndistinctive peaks in the natural gamma downhole logs, which are conventionally identifi ed by hand.rnA GP method has been developed to automate this task.rnOnce the appropriate boundary has been located on the basis of shale occurrences, chemical assaysrnfrom exploration drill holes are used to fi nd the exact boundaries of interest. These boundariesrndivide the drill hole stratigraphy into regions of different mineralogy, each of which display theirrnown distinct correlations between the main elements and oxides (Fe, SiO_2 and Al_2O_3). Iron ore showsrna negative correlation between Fe and Al_2O_3, but in banded iron formation (BIF) there is a positiverncorrelation between these species. Similarly, SiO_2 and Al2O3 have a positive correlation in the shalesrnand ore but a negative correlation in the BIF. Correlations obtained within ore-, BIF- and shaledominatedrnregions are therefore better than those obtained using the entire log and can be used tornimprove the results obtained when modelling.