The Anglo Platinum Group, the world's leading primary producer of platinum metals, recently identified a pressing need to implement more efficient ways to analyse its geological assay data, and turned to Octoplus and the SAS JMP data visualisation tool.

Anglo Platinum is listed on the Johannesburg and London stock exchanges and drills between 600 and 700km of exploration boreholes per year, with samples taken at regular intervals across reef intersections. These samples are used to evaluate the potential of future mining at various sites around the Bushveld Complex. Each sample is analysed at one of a number of a laboratories and checked for Quality Assurance and Quality Control (QAQC).

"Each sample received is sent to one of our laboratories and analysed for elements such as platinum, palladium, rhodium, gold, etcetera. This is done twice in what we call a Twin Stream format,” says Markus van der Neut at Anglo Platinum. “Every month thousands of results are sent to the central database and then used for resource estimations and mine planning. It is of utmost importance that the quality of the data is assured for precision and accuracy prior to being used for any resource evaluations."

In the past, each laboratory made use of Microsoft Excel to help analyse the data it generated. However, modern governance controls require that the end-user must verify the acceptability of the data prior to its use. The data limitations, particularly with the Excel 2003 version, such as the ability to cater for only 65 000 records, an xy-scatter plot for only 32 000 results, the fact that all data had to be input via copying and pasting, the absence of the ability to create an audit trail, the slow and cumbersome processes as well as the large margin for human error were cause for management concern.

It was for these reasons that Anglo Platinum began the process of identifying a technology tool that could assist it with its data analysis, data representation, reporting, auditing and to a large extent automate a laborious process. It turned to Octoplus as a technology partner with the SAS JMP data visualisation tool.

"The process is not as simple as it sounds; we use a number of different data types for a QAQC investigation for example: the assay results of the submitted samples, the results of Blanks and Standards added by the relevant laboratory, the assay results of an external umpire laboratory as well as blind Blanks and Standards added to the sampling stream by the geologist,” adds Van der Neut. “These Blind standards are submitted by the geologist under a different name so as to conceal their true identity to the laboratory; this is done in order to determine how close to a known and fixed value the laboratory can report, which helps to ensure that no mistakes are made and that data is accurate while at the same time testing for contamination."

The first application of the JMP tool provided Anglo Platinum with a simple tool based on a number of primitive scripts, which covered a very basic set of calculations such as a single outlier method, precision and basic statistical parameters. While all the results could be recorded in a Journal file, it was not compulsory to do so in order to move to the next script - this meant it was possible to forget to journal, in which case one had to start over again.

This process was subsequently replaced by Octoplus with a customised set of scripts and tools, all grouped under an "Angloplat Button" on the toolbar. With JMP, every action by a user is now menu-based and involves four separate components: evaluation of assay data; evaluation of Blanks and Standards; a set of useful tools and lookup tables (which can be edited by the user); and an extraction utility which allows the user to extract data from SQL-based databases. The menu system also provides an audit trail with the evaluation facility recording every action that the user makes and storing these in a separate journal file.

"Apart from the standard features such as calculating the various basic statistical parameters, we can now select from three techniques instead of having one choice of outlier method,” adds Van der Neut. “Furthermore, we now have a comparative facility, where we can see which records have been identified as an outlier by the different techniques before making a final outlier choice. Only once this has been made are the outliers removed from the working file and stored in a rejected file – this is all automated and done in minutes as opposed to days."

A user-friendly interface allows the user to extract data directly from the Anglo database into JMP; the almost limitless number of records JMP can handle sets it apart from Excel. Often data from one particular laboratory over a number of years needs to be extracted for evaluation purposes in JMP; these datasets can easily reach several hundred thousand records.

JMP is a data visualisation and statistical discovery software that combines the use of graphical capabilities, the ability to manage virtually unlimited volumes of data and seamless integration with SAS Analytics tools. Traditionally used by scientists and engineers, JMP provides business users the ability to replace static charts with motion-enabled, interactive plots that uncover hidden trends and predict the future. Through graphical data querying, enhanced script editors and project collaboration tools, data discovery is made more comprehensive.

"With JMP, Anglo Platinum is now able to extract data more easily, we can identify outliers quicker, we can view our results via means of visual representation and can have accurate information back to our mining geologists quickly," Van der Neut adds.

"I cannot stress enough the usefulness of our current JMP evaluation system. Its strength lies in the fact that the user is still very much in control and has several evaluation methods to choose from, prior to making final choice, yet still following a standardised, auditable procedure. It is not just another 'press button and all is done' system, but a very powerful evaluation tool and one that has changed the way in which we are able to deliver results," he ends.

Source: SAS Institute, IT Web Enterprise