AI, machine learning to deliver ‘wave of discoveries’

The past 20 years have seen remarkable advances in the mining industry, particularly in mineral exploration technologies with vast volumes of data generated from geologic, geophysical, geochemical, satellite and other surveying techniques. However, the abundance of data has not necessarily translated into the discovery of new deposits, according to Colin Barnett, co-founder of BW Mining, a Boulder, Colorado-based data mining and mineral exploration company.

“One of the problems we’re facing in exploration is the huge increase in the amounts of data we have to look at,” said Barnett, in his presentation at PDAC 2020, during a session on managing and exploring big data through artificial intelligence and machine learning. “And although it’s high-quality data, the sheer volume is becoming almost overwhelming for human interpreters, and so we need help in getting to the bottom of it.”

By integrating hundreds or even thousands of interdependent layers of data, with each layer making its own statistically determined contribution, machine learning offers a solution to the problem of tackling the massive amounts of data generated, and a powerful new tool in the search for mineral deposits.

But, in an interview with The Northern Miner, he cautioned that to fully exploit the potential of machine learning in mineral exploration, “prospectors will still need to devote considerable time and effort to the preparation of data before machine learning techniques can add value for companies.”

To illustrate his point, Barnett demonstrated how he and his partner at BW Mining, Peter Williams, are using machine learning to analyze data from geological, geochemical and geophysical surveys of the Yukon in northwestern Canada to locate new deposits.

The Yukon became famous for the Klondike gold rush during the late 1890s, which petered out after a few years as prospectors moved onto Alaska. Today the area is experiencing a renewed interest in what has become known as the Tintina Gold Belt, with significant lode deposits being found over the past two decades and, according to Barnett, “more waiting to be discovered.”

“We used the Yukon bedrock geology map published by the Yukon Geological Survey, which is very detailed and shows over 200 different geological formations,” explained Barnett. “However, you can’t simply put 200 formations into a machine learning process. First, the data requires special treatment.”

By representing each of the formations with a separate grid and by continuing the grids upward, they were able to see overlaps between formations, allowing them to consolidate the data by grouping the formations by rock type and age, and thereby reducing the data set down to around 50 discrete and different formations. They then used the same process to represent structural data provided by the map.

Image of a typical target. Red areas are highly favourable, while purple areas are shown as unfavourable for gold. Credit: BW Mining.

“The structural data is important because it represents the pathways that the mineralization generally took to reach the surface,” explained Barnett. “We then used geophysical maps of the area provided by Natural Resources Canada, which contain enormous amounts of information that can be extracted and subjected to the same statistical treatment.”

Applying the same approach to geochemical, gravity, topographical and satellite data, they were able to generate detailed data sets covering over 300,000-400,000 sq. km of the study area.

“The most critical layer of data for our machine learning process is the known deposits because this is used to train our artificial neural network against all the other layers to identify deposit formations,” said Barnett.

Artificial neural networks operate much like the human brain. They can recognize patterns in the different layers of data and cluster or classify them into groups according to similarities in the input data. They are then capable of discriminating between zones of high and low mineral potential.

After scouring through geologic publications, company websites and NI 43-101 technical reports, Barnett and Williams were able to develop accurate mineral footprints for more than 30 deposits using their model, which, according to Barnett, reportedly contain over 46 million oz. of gold.

They then used an artificial neural network to establish the statistical favourability of a location containing an economically viable deposit across the entire region of interest. This approach is essentially an inversion process that uses exploration data relating to a given location as inputs to the network, which then produces the corresponding favourability as the output.

“This requires very sophisticated software to analyze and interpret the data, so you can’t just use off-the-shelf software,” explained Barnett. “We first started analyzing the data on a parallel-processor in the basement of the University of Sussex [in England] back in 1992, where my partner was a professor. But it would take five days to get an answer — by which time we’d forgotten what the question was.”

However, with improvements to computer software and hardware, they are now able to generate an answer in a matter of hours using a common laptop.

Barnett’s and Williams’ use of artificial intelligence and machine learning has led to a highly-focused target map that assigns numerical probabilities of making an economic discovery anywhere in the region of interest. And it can be used to systematically rank and rate targets and plan cost-effective follow-up programs that take into account the expected return on investment for any given target.

Although Barnett believes there is currently a lack of understanding of artificial intelligence and machine learning in the industry, he is convinced that as “these techniques become more widely used and available, machine learning and artificial intelligence will lead to a wave of discoveries. And within ten years, they will be commonly used tools in the mineral exploration industry.”