Article from resourceful: Issue 12
RESEARCH PERSPECTIVE
It's hard to make the most of an economic resource when you only know part of the picture.
For instance, you might be making decisions without knowing the answer to critical questions on mineral relationships, such as whether or not the gold is refractory, what minerals host key elements and what mineral phases will impact on processing steps.
Now, advances in mineral analysis and characterisation technology are deepening our knowledge and understanding so that it's possible to paint a more complete picture of a resource to feed into decision making across the value chain.
We can process big datasets – with a wealth of information – faster and more cost-effectively than ever before.
It means that the chemical makeup or characteristics of a resource can now be analysed and understood at a fundamental, even atomic, level.
Only with data from every sample scale (from kilometre to atomic), can we fully understand the true value of a resource and how it will behave during processing to maximise extraction and recovery.
Understanding the bigger picture looking at the micro-scale
We know that because when we review large-scale data – such as regional geochemistry – with understanding from micro-scale data, we often find that the "bigger picture" model doesn't hold up. This calls for a reinterpretation of large-scale datasets.
Micro-characterisation is also valuable for validating an exploration tool or confirming process understanding for increasing confidence in the application of a method.
Bias can be introduced in sampling and data collection in a multitude of ways. For example, drill core logging on an exploration camp or minesite may be undertaken by numerous geologists over time, each of whom may have slightly different interpretations of the logging codes.
Using a data driven approach to create objective logging methods improves correlation and supports more informed sub-sampling. Objectively analysing characterisation data at every scale underpins the collection of representative samples.
Data integration
Machine learning provides a tool for integration of micro-scale data back into the larger models. Such approaches allow micro-scale data to be used to support the prediction of chemistry or other ore characteristics from bulk datasets, allowing for earlier and more informed decision making. For example, in helping to determine the optimal process pathway to take to maximise recovery and increase productivity.
With the development of a new drill core laboratory, as part of CSIRO's Advanced Resource Characterisation Facility expansion, we envisage creating a new workflow where data is collected on whole drill cores, and the subsequent categorising (or domaining) of that data, drives an informed and representative approach to progressive sub-sampling. This creates a framework for integrating the data from micro-scale analysis back into mine models.
In the same way that geochemical data has become routine in minesite data collection; the use of micro-characterisation data is poised to become standard practice and fundamental to decision making at any mine operation. As a result, value will be created for industry at every stage – from exploration to processing.