Underground » Geology
The aim of this project was to develop techniques and tools to maximize the amount of geological, analytical and geomechanical information geologists can derive from their downhole geophysical logs.
The study analyses the data from 21 drillholes from Anglo Coal's Moranbah North project. It uses multivariate statistical techniques to determine if lithology, rock strength as estimated by the geologist, fracture frequency and durability (slaking potential) can be determined from the four downhole geophysical logs: natural gamma radiation, gamma-gamma (density), sonic and neutron-neutron.
Chapter 1 of this report reviews the state of the art which includes a summary of discussions with senior industry personnel and a detailed literature review describing over one hundred books and papers dealing with: logging tool principles, log quality control, log interpretation for petroleum, coal and geotechnical applications.
Chapter 2 describes the data and its format as received from Anglo Coal. Detailed descriptions of the many geophysical variables in the data set and the basis for selecting those used in this study are given in Appendix C. The chapter subsequently lists the basic statistics for the chosen geological and geophysical variables and the means of each of the geophysical variables for each geological category.
Chapter 3 introduces the two multivariate techniques which have been used in this study. These two techniques, namely Principal Component Analysis and Discriminant Analysis are explained in a qualitative manner using examples from the study data set. The mathematics behind them is described in detail in Appendix E.
Chapter 4 provides guidelines for auditing geological and geophysical data in coal exploration programmes. It introduces a graphical technique known as "box and whiskers" plots for analysing the geophysical data to find outliers such as values taken above the water level in a hole and detect changes of scale occurring between holes. "Box and whiskers" plots are also used to indicate holes where a particular rock type has an atypical response on one of the geophysical variables. Discriminant Analysis is subsequently utilised to produce plots indicating holes where a particular rock type has an atypical response over a number of geophysical variables. The calibration techniques used by the geophysical company are outlined. These are followed by lists of the tests that should be performed on-site immediately after data collection to check the quality of the geophysical data. Lastly the chapter provides graphical and mathematical techniques for checking the calibration of the geophysical data off-site before it is used for analysis purposes.
Chapter 5 uses Principal Components Analysis to display the lithology data as a function of the geophysical data. These diagrams show that the lithologies: coal, carbonaceous mudstone, mudstone/tuff, siltstone and sandstone can be reasonably distinguished based on their geophysical responses. Principal Components and Discriminant Analysis are used to display the rock strength estimated by the geologist and fracture frequency as functions of the geophysical data. These diagrams show that there is a relationship for both rock strength and fracture frequency with the geophysics and that this relationship is stronger for rock strength than fracture frequency. In both cases improved results were obtained after the data were divided into pseudo rock type categories as the relationships with the geophysics vary with rock type. For both rock strength and fracture frequency, however, there is a considerable overlap between the categories when interpreted from the geophysics. This overlap is less of a problem for the extreme categories (weak and very strong or very low and very high fracture frequency) than for the intermediate categories. Discriminant Analysis is also used to display the durability (slaking potential) as a function of the geophysical data. An excellent relationship is found between the durability and the geophysics but this needs to be treated with caution as the durability analysis was based on a very limited data set of only 34 samples.
Chapter 6 summarises the findings of the study, discusses future improvements to the data auditing. Lastly as the results of the study do not show a clear discrimination between various rock strength and fracture frequency categories but do provide very useful category probabilities for each sample, the chapter suggests that considerable advantage could be made of this information by incorporating it into conditional simulation models.