Quantifying error and bias in sampling thin carboniferous reef types

No Thumbnail Available

Date

2016-04-05

Authors

Fourie, Andries

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

This study examines the history of underground face sampling since the earliest writers documented their findings regarding the method. An historical review of the errors associated with sampling is followed by a discussion of findings from a number of underground investigations. The thesis examines how errors introduced during and after sample extraction are responsible for error and bias. A Simulated Chip Sample Model was created and used to evaluate the nature of sampling errors and bias. Learning’s from this model include an understanding of the effects of sample extraction on variations in grade and precision. The effects of sample shape on error were considered and the aspect ratio of the sample was found to be a critical factor in minimising error It was found that if the homogeneity of the ore material decrease the error increase proportionally, despite the extraction error remaining constant. During extraction the material will be biased and the effect of using accumulation values will amplify the error and bias in samples; the error and bias in this case is secondary and has nothing to do with the extraction error. The historical ideas about error being a consequence of the nugget effect have been shown to be false; variability between samples in this case is a direct result of poor sample extraction rather than the occurrence of large gold grains in the ore. The thesis also identified sources of sampling bias and two distinct types, namely the soft-reef bias and waste-discard bias were identified. The perception that soft-reef bias is the main contributor to the deterioration of the MCF was found not to be the case. This finding is based on a comparison between chip sampling and perfect sampling, in the form of so-called “coffin samples”, that indicated there is no bias between these sample types. Other possibilities for sample error and bias were investigated and found to be related to human preference for selecting material that had to be discarded during or immediately after the sample had been taken. This was confirmed by a survey of 70 samplers with different experience levels and from different mines, who indicated the same preferences when selecting material they chose to discard from the sample material collected. The waste-discard bias is a better contender for introducing bias serious enough to affect the MCF because it occurs at each and every sample site, unlike the soft-reef bias. This type of bias was shown to mimic the soft-reef bias using the Simulated Chip Sample Model. Chip sampling has been the simplest and at the same time most misunderstood sampling method there has ever been, but it has stood the test of time and is shown to be without a meaningful replacement. Despite the appearance of poor extraction compared to other methods, chip sampling is an acceptable technique when one understands and eliminates biases during and after extraction.

Description

A dissertation submitted to the Faculty of Engineering and the Built Environment, University of the Witwatersrand, in fulfilment of the requirements for the degree of Masters in Engineering. 13st October 2015

Keywords

Citation

Collections

Endorsement

Review

Supplemented By

Referenced By