Mining's Artificial Intelligence

The world is facing huge technological change, and the mining industry could be amongst the major beneficiaries of a fusion of technologies that has become known as the fourth industrial revolution.

Go to the profile of Chris Hinde
Oct 09, 2019
0
0

The world stands on the brink of huge technological change, and as a capital-intensive, data-driven industry, mining could be amongst the major beneficiaries. This imminent fusion of technologies has become known as the 'fourth industrial revolution' (4IR). It follows the first industrial revolution of the 18th and 19th centuries (characterised by steam power and the rise of the iron and steel industries), the second revolution of 1870-1914 (which saw the harnessing of electricity for mass production) and the third, digital, revolution of the past 40 years.

At the centre of the creation of new integrated processes is artificial intelligence. AI is the intelligence demonstrated by machines that simulate the cognitive function of humans. This includes the acquisition of information and rules (learning), using rules to reach conclusions (reasoning) and self-correction.

4IR, and AI specifically, will lead to myriad changes in a 'reimagined' mining industry over the next 30 years. There are widespread applications, but the two most important groupings for the mining industry are perhaps automation and machine learning.

Automation: Particularly useful for high-volume, repeatable, tasks that humans normally perform. AI automation is different from conventional automation in that it can adapt to changing circumstances.

Machine learning: The science of getting a computer to act without programming, subsets of which include language processing and robotics. These algorithms can be 'supervised' (data sets are labelled so that patterns can be detected), 'unsupervised' (data sets are only sorted according to differences) or have 'reinforcement learning' (data sets aren't labelled or sorted but the AI system is given feedback).

Mining Automation

It has been argued that there are six levels of automation:

0 — No automation

1 — Initial mechanisation

2 — Line of site activity

3 — Remote activity

4 — Machine functions automated, and

5 — Intelligent machines.

The increased use of automated machines is already making mines safer, and the working environment more comfortable. AI 'supervision' of these machines will make them even more efficient, and also constantly survey their health and performance. AI will help prevent accidents by predicting equipment failures, ground movements and other potential hazards by analysing data patterns.

As a parallel benefit, this automation should also help increase the likelihood of more women being employed. Automation is of concern to unions, however, because of the threat, whether real or perceived, of lower levels of employment.

A good example of the improvements available to the mining industry is in autonomous vehicles and drills. Rio Tinto has been using autonomous haul trucks since 2008, and these trucks have been so effective that they have reduced fuel use by 13% and are much safer to operate. The trucks can also operate 24/7, without the need to stop for shift changes or bathroom breaks.

Rio Tinto has also been using autonomous loaders and blast-hole drill systems for several years. The drilling system lets one remote operator control multiple drilling rigs, and the company claims that its autonomous drills improve productivity by roughly 10%. Rio Tinto is currently introducing the world's first fully autonomous long-haul rail system.

Mining methods themselves have not changed much since the publication of De Rey Metallica, in which Georgius Agricola catalogued mining in 1556. An important change in the 21st century, however, will be the migration of 'single point' solutions to multiple solutions. A report by EY last year predicted that overall asset effectiveness, currently under 50%, will be 65% within 10 years. The main issue will be improving productivity by optimising the 'mine to mill' flow, and then extending it to 'mine to market'.

Mines of the future will be different, with significant automation leading to more complex operations (with up to 20 working places), different extraction and blending methods, and different equipment.

For example, as part of making the pit-to-port operations as intelligent as possible, Rio Tinto is creating an intelligent mine that should deliver its first ore by 2021. There are more than 100 innovations the company is evaluating, but one initiative, called digital twinning, first created by NASA, is now being adopted by many in the industrial sector.

By creating a virtual model that is fed real-time data from the field, scenarios can be quickly tested, and operations and production can then be optimised. This ability to test decisions before they are implemented leads to better outcomes, and to cost savings.

Machine Learning

Computers can be trained to identify faint patterns and trends. This has led to exciting developments in the mining industry, and technology companies are leveraging AI to reduce capital risk while increasing efficiencies and success rates for exploration, mine production, mineral processing and investment.

In exploration, for example, data can be analysed from multiple sources and a machine-learning algorithm used to identify areas where minerals are likely to be found. Many miners are now utilising AI exploration techniques. Goldcorp, for example, is in an AI joint venture with IBM Watson to review geological information to identify drilling locations for gold in Canada.

There are numerous applications at the operating level. For example, it is possible for fragmentation analysis to more accurately measure rock breakage. This method of data collection can provide valuable feedback to the engineers, increase productivity and even optimise teeth changeouts by monitoring teeth wear. In Brazil, Vale SA is using AI in several areas; at the Salobo copper mine in Para, for example, there was a 30% increase in the lifespan of haul truck tyres in one year, and this same technique is now being applied at other mines. The company also uses AI to predict rail fractures, which is helping to reduce the occurrence of fractures by up to 85%.

In processing plants, mineral and ore-sorting equipment (including laser sorting and product-recognition scanners) is being utilised to separate ore from waste. The importance of AI was highlighted by the CEO of Freeport-McMoRan Inc., Richard Adkerson, at the company's Q1 conference call in April. Mr Adkerson said, "we are making progress in productivity and cost-control management by using advanced analytics at our Bagdad mine". This project is being seen as a test case for Freeport by "using data capabilities to measure things, and respond to them very quickly". Applying AI "in a business like ours is not something that people intuitively think about".

Mr Adkerson added that Freeport has brought in McKinsey, and other outside experts, to work with them. "We are really encouraged by what's happening in Bagdad now, and the next stop is going to be Morenci. But across the board, we're going to be using this big data analytics to help make our business better, reduce our costs and improve productivity."

AI is also increasing important with regard to improving health and safety. For example, at the Escondida copper mine in Chile, BHP has tested 'smart' hard hats that analyse the brain waves of drivers to measure fatigue. This has been integrated into over 150 trucks to boost productivity and increase safety.

Beacon on AI

AI is featured prominently at forthcoming Mining Beacon conferences, including the International Mining and Resources Conference (IMARC) in Melbourne, October 29-31, and Mining Technology London (MT London) on November 25-27.  

IMARC will include a session on AI hosted by McKinsey, and delegates will hear, no doubt, that the new technology is no 'magic bullet' and its implementation requires a buy-in from senior management. AI also needs to be discussed with staff in advance (especially in regard to any loss of jobs), requires the creation of multi-disciplinary teams and the tools need to be fully tested before being deployed.

The legal aspects of implementing AI also need to be considered. Who is liable if something goes wrong: is it the software provider, the mining company or a third-party consultant? There are also intellectual property considerations, such as whether the company has the necessary rights to the training data, and who will own any new IP that may be created.

These issues, and many others, will also be addressed in November at MT London. At this three-day meeting, Edson Antonio, Global AI Manager at Vale, will give a keynote address entitled 'Ensuring that your AI doesn't just deliver short-term quick wins, but delivers real step-change long-term benefit'.

Also featured at the event will be a 'fireside chat' between John Welborn, the CEO of Resolute Mining, and Tal Zarum, Head of Programmes & Supply Chain at Sandvik Automation Group, on how, and why, Resolute Mining invested in autonomous equipment. Michelle Ash, the Chairman of Global Mining Guidelines Group will also discuss how mining companies can turn data into dollars. MT London will include a presentation by Denise Callahan, Director of Analytics and Strategic Planning at The Doe Run Company on how junior producers can achieve significant ROI with predictive data analytics.

Go to the profile of Chris Hinde

Chris Hinde

Chief Commentator, Mining Beacon

Previously editorial director of Mining Journal, and more recently head of S&P Global Market Intelligence's metals and mining team, Chris is now Mining Beacon's editor-in-chief and lead commentator. He posts two blogs every week, one on Monday reviewing market conditions over the prior week, and a second on Thursday looking at issues on the global mining scene. There is also a quarterly blog on business opportunities in the sector.

No comments yet.