 |
Publications |
|
|
|
|
|
|
Abstract
This report describes the results of rock recognition analysis utilising drill performance and operational data gathered on a Terex Reedrill SKSS-15 blast hole drill.
The analysis has concentrated on a row of 28 holes drilled 3m apart and to a depth of 12m. The holes intersected a number of rock types including shale, ore and banded iron formation (BIF). The drilling was conducted with the drill operating in percussion mode and with the use of a vibration absorbing shock sub.
The analysis was undertaken using a variety of machine learning techniques but a technique known as boosting performed best. Learning data was supplied from geophysical logging results obtained in each of the boreholes (natural gamma, magnetic susceptibility, density and caliper logs). The geophysical data were obtained at a down-hole spacing of 0.1 m, the same spacing used for the reduction of the drill sensor data. From the geophysical data it was possible to identify the shale, ore and BIF. An estimate of ore hardness was made on the basis of variation in the density values within the ore zone.
The analysis of the drill sensor data was successful in that the results closely matched the geophysical interpretation and the geological expectations. Provision of suitable independent training data is an issue and future trials will need to ensure that adequate training data is available.
|
| Download |
PDF full text (11.68M)
|
|
Copyright Notice & Disclaimer |
|
