Resource estimation in a heavy mineral sand deposit using below detection limit data.
No Thumbnail Available
Date
2012-08-17
Authors
Boekhoud, Karina
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The objective of this research report is to investigate the variation of resource estimation at a heavy mineral sand deposit using below detection limit data
Best practise throughout the mining industry suggest regular audits to ensure the appropriate application of modelling techniques, reliability of the grade model and other factors contributing to the grade model such as analytical values (Dominy, Annels and Noppe, 2002). The heavy mineral sands industry is market driven. Correct resource and reserve estimation is essential when considering market negotiations.
Current estimation methodology at Hillendale mine uses the laboratory reported ‘zero’ data. The ‘zero’ data used for estimation is not considered to be best practise. External audits suggest the replacement of ‘zero’ data to ‘null’ data to lower kriging variance. The current exploration strategy at Hillendale Mine comprises of WAC drilling, sample preparation and analyses of the sample with a Carpco® lift machine which uses magnetic susceptibility. Carpco® separated fractions are further analysed after compositing with grain counting and XRF analyses. This process is in general similar to other heavy mineral sand operations in South Africa.
Two datasets are used to estimate a geological resource. The ‘zero’ dataset is the original borehole data as analysed by the laboratory. In the ‘zero’ dataset, all the below detection limit data is assigned a ‘0’ and the value is added as a number to the average calculation. With the ‘null’ dataset, the below detection limit data are left blank and are no value is added to the
average calculation. All other parameters and methodologies are kept the same for comparative purposes. The effect of the two datasets is compared to one another through basic statistical analyses, geostatistical analysis and visually through trend analysis on 50 metre slices through the mine. The geological estimates are compared to actual mine plant data through reconciliation to evaluate the influence of the ‘zero’ data compared to the ‘null’ data in a producing heavy mineral sand mine.
The result of the study will show that the global estimation and methodology for both datasets are similar. Cross validation checks using the ‘null’ dataset with the magnetic fractions are not possible. Both datasets follow similar trends of over- and underestimation. In general the resources’ are overestimated for the mineral deposit compared to the actual plant data. The overestimation of the ‘zero’ dataset is less than the ‘null’ dataset therefore a better estimation of the expected value.
It is recommended that future work is done in terms of investigating the appropriateness of the datasets, the effect of cut-off values and the appropriateness of the estimation method. In addition the effect of mining, in terms of mining to plan, as well as actual plant recovery should be investigated.