Improving the Reliability of Density and Grade Estimation from Borehole Geophysical Log Suites

Project Objectives

Coal density plays an important role in projecting and reconciling coal tonnage and quality variation. The best estimation of coal density is from direct measurement on core, but cored holes are sparse relative to chip holes due to the cost of drilling and subsequent laboratory analyses. All holes are geophysically logged, and if the geophysical data are accurately calibrated against core then they can be used to improve the sampling of spatial variability across a deposit. However, uncertainty about the precision and accuracy of the density estimation from geophysical logs still precludes its use for reserve analysis across the industry at large. A common comment is that “one accurate data point from laboratory analysis is better than hundreds of inaccurate ones”. The objective of this project was to improve the reliability of density and grade estimations from geophysical logs.

We tried to achieve this objective through examining quality control issues that might be addressed through simple procedures and their application. The approach of this project was to:

• Review the geophysical density logs in use within the coal industry;
• Analyse precision obtained by different geophysical logging measurements to estimate density; 
• Analyse the repeat measurements from a single calibration borehole to determine the precision of the measurement;
• Assess variability in correlations between geophysical logs, relative density and proximate analysis results for three mine site data sets of different rank (data from Clermont, Dawson and Norwich Park mines); 
• Examine causes of error and variation in geophysical logs and laboratory density measurements; and,
• Suggest guidelines to improve correlation and use of geophysics for density estimation.

Main Achievements and Deliveries

The key findings from this project are:

• The geophysical density logging has a solid physical foundation and provides a direct measurement of in-situ coal density. However, geophysical logging data have experimental errors that need to be reduced by implementing strict quality controls on borehole logging procedures. This can be done through establishing a logging instrument calibration facility at each mine site (either using selected calibration boreholes, or artificial testing blocks for known coal seam densities). This calibration should be carried out as part of routine logging calibrations before any new borehole is to be logged.
• Geophysical coal density logging is reliable to use as it can achieve an accuracy of about 2% and a precision (repeatability) of about 1.5% which has a combined error of about 3.5%, which is within the error range of 3% to 7.5% for the laboratory determination of relative density using the Standards Australia pycnometer bottle. Although this result is derived from only a single repeat log data set with the measurements from a single service company, it does provide a benchmark for what can be achieved from geophysical borehole density logging.
• Geophysical density logging has a limitation due to its intrinsic resolution, which is about 20cm for the coal density measurement from the short-spaced density log tool of Weatherford. This implies that the densities of coals less than 20 cm thick cannot be accurately measured by the density logging. The depth sampling interval for the logging does not affect the resolution of the density logging as demonstrated in this report.
• We proposed to reduce/eliminate the boundary effects by excluding the density portions the half-resolution (i.e. 10cm) distance away from the boundaries from the average density calculation. An automatic procedure has been created for such calculation. Due to the limitation of the density logging resolution, the average density of a coal seam from a geophysical density log will be affected by the coal seam boundary. For a coal seam of  1.4 g/cc surrounded by rocks of 2.8 g/cc, the thickness of the coal seam at least needs to be 5m to keep the boundary effect less than 1%. 
• A simple depth shift can improve the correlation between the geophysically logged density and lab measured relative density (RD), and so increase user confidence in the information from geophysics. The lab measured densities of the coal samples from the thin seam (<0.2m) or within the boundaries should not be directly compared with the average densities from the geophysical logs, as the geophysical densities are strongly influenced by the boundaries. Due to the boundary effect, a small depth-shift of a sample relative to the geophysical log will change the estimated density from geophysical log. The depth-shift should be corrected before calculation of the geophysical density for such sample. 
• Geophysical density logs correlate well with laboratory RD. The in-situ RD (RDis) have an average error of about 3% after excluding the thin-seam and near-boundary samples, which is consistent with comparable experimental errors of the laboratory measurement. 
• In-situ RD (RDis) can be estimated from geophysical density logs. Of the various density logs, the VECTAR (Vertical Enhancement by Combination and Transformation of Associated Responses) processed density log ADEN is the closest to the in-situ RD estimated using Preston-Sanders’ moisture compensation formula with the non-site specific in-situ moisture from Fletcher-Sanders’ ACARP Project C10041 report. This suggests that 1) the moisture correction is important to convert the laboratory RD to in-situ RD; 2) ADEN is a good measure of the in-situ coal density; and 3) the general in-situ moisture estimation from the ACARP Project C10041 is a valid approximation.
• Geophysical logs can be used to estimate in-situ density to similar accuracy as laboratory measured densities (RD & RDis) using the model-based regression, or multivariate Self Organising Maps (SOM) and Radial Basis Function (RBF) methods. These methods can be used as approaches for coal density calibrations. However, the SOM and RBF methods using the multiple geophysical logs do not improve the parameter estimation compared with the simple single-parameter model-based regression method for the data tested. The reasons for this are not clear yet but it does highlight the robustness of the conventional model-based regression methods.

During this project, three prototype testing software tools were developed:

• A command line program for extracting the average geophysical logging parameters for corresponding coal samples in the same depth range. This program also automatically matches the coal sample data with geophysical data to remove potential depth offsets.
• The RBF algorithm for coal quality parameter estimation from multiple geophysical logs. The algorithm is briefly described in the Appendix V.
• An automatic coal density estimation algorithm from geophysical density logs by using automatic blocking with automatic boundary adjustment based on the density logging resolution to reduce the boundary effects. 

Recommendations

Recommendations from this project include:

• As this project indicates that the geophysical density logs are valid in-situ coal density measurements, there is no reason why the coal density estimated from the geophysical density log cannot be used for resource and reserve estimation providing the geophysical logging quality is managed and the boundary effects have been removed. It would be useful to demonstrate the impact of coal resource and reserve estimations using better estimated spatial variations in coal density from geophysical density logs through a case study.
• The data used in this project are from the existing databases of Clermont, Dawson and Norwich Park mines. These data were not collected specifically for this project. One of the main objectives in this project is to examine if the geophysical density logs are a true measure of the in-situ coal density. This requires a reference data set to be used as a benchmark. However, such a data set was not available for this project. For a more strict comparison and calibration, it is recommended that a reference data set with known in-situ coal density and corresponding geophysical logs is collected for further investigation.
• From our analysis and observations, the geophysical density logs are reproducible and a reliable measure of the in-situ coal density. To achieve such results, it is recommended that strict quality control procedures are implemented by the mine sites to ensure all the geophysical data are of high quality and acquired in a consistent manner. Mine site calibration facilities and quality control implementations in software are recommended.