Rosetta holds the key to linking analytical mineral data
Taking its name from the world-famous Rosetta Stone, which provided the clues to deciphering ancient Egyptian hieroglyphics, is a pointer to what a new technology does for mining companies. CSIRO’s Rosetta provides an illustrated guide to the characteristics of ore.
Unlike the Egyptian Rosetta Stone, which is now the most popular exhibit in the British Museum, the new Rosetta is not something that can easily be seen. It is effectively a data-driven software tool that involves pulling together analytical data gathered from measuring one or more samples and then using that knowledge to predict the character of rock to know where to mine.
The importance of studying the physical and chemical composition (e.g. mineralogy, chemistry and lithology) or “characterisation” of ore, lies in the way the acquired knowledge helps in understanding the quality of the rock during exploration targeting and how it will behave when mined and processed.
The machine learning process is how Rosetta gets its name, because on the Egyptian stone, three languages bearing the same message enabled a scholar who knew one of the languages (ancient Greek) to read the hieroglyphics and a third language on the stone, demotic script.
Deciphering multiple data sources to create a complete picture
Multiple layers of historical analytical data can be fed into the Rosetta tool. Hyperspectral information is one, but other examples of low cost data being fed into the learning process include x-ray fluorescence, and even core photography.
From a representative sample, Rosetta builds a characterisation image of the orebody. That knowledge becomes part of a process that progressively eliminates uncertainty through the addition of more representative samples – all leading to a point where a prediction about ore quality in exploration and ore behaviour in mining and processing can be made.
Trials of Rosetta in brownfields exploration at one of Australia’s biggest base metal mine sites have been successfully undertaken. It’s added a layer of confidence in the tool and its future use as a cost-saving process for existing mines with future potential in greenfields exploration.
Data integration is key
CSIRO’s resources data science research manager, Ryan Fraser, says the challenge that led to developing Rosetta, was finding a way to utilise the vast amounts of data gathered in mining and processing.
“Miners have many ways of analysing the rocks they’re proposing to mine, as well as the material they’re planning to process,” Mr Fraser says.
“Ultimately, the data integration solution at a mine is to use all of these sources to predict, for example, how much effort and cost it will take to dig the ore and put it through some kind of process.
“One way of answering the question of characterisation is to drill additional holes and assay the samples recovered.
“That would tell you a lot about the rocks in the mine, but it is an awfully expensive way of doing it.
“So, not all characterisation is equal in terms of the cost and the time it takes to get it.”
Rosetta’s advantage is that geoscience principles are “baked” into machine learning and analytical models to provide relevant predictions.
CSIRO geologist and resources data science lead, Jess Robertson, says a simple example of characterisation might be assessing rock chemistry, such as measuring how much gold or copper is present.
Looking deeper it can be important to know what other minerals are in the rock because some of them will go through the flotation stage of the treatment process, while other minerals do not.
“Hardness is another important measure, so we use characterisation as an umbrella term for all the different facets of a rock,” Dr Robertson says.
Both agree with a suggestion that the hunt for characterisation can be loosely equated with seeking “soul of the rock”, or its DNA to describe it another way.
Rosetta integrates mineral data from high and low cost analyses
Gaining the knowledge of what makes up a particular rock type is not cheap, which means bracketing the data being gathered into cheap-to-get data, and expensive-to-get data.
For example, one step involves using hyperspectral scanning, which is relatively low cost to use, although the machine itself is expensive.
When applied to drillcore, hyperspectral scanning provides a highly detailed picture of the mineralogy of a core sample that can be produced within minutes. An expensive alternative is sending rocks for assay and waiting weeks for the results.
Other tests become part of the overall Rosetta software tool, such as measuring how well a rock crushes and how easily a mineral can be liberated.
As more samples are processed, a library of information is developed, providing a picture of the mine and its orebody that can help determine inputs such as the energy required in processing.
Rosetta is a tool to aid rapid characterisation of ore, building orebody knowledge to drive efficient planning, improved control and decision-making.