Data is the new oil: Rio Tinto builds new intelligent mine
An executive "innovation field trip" to America, marked the beginning of a new era for Rio Tinto. That happened merely a decade ago, and now as data turns the new oil, the company must follow its innovative approach and evolve.
A decade ago, one of the world’s largest metal and mining corporations, Rio Tinto has decided to harness new technologies to optimize the extraction of minerals from deep within the Earth while reducing environmental impacts and further improving safety. This change originated in an ‘innovation trip’ to America that Sam Walsh, a company executive who later became the CEO of Rio Tinto, and his team had conducted during 2005 [1]. On that trip, the team has deliberately avoided visiting mining companies and rather focused on paying visits to lead universities and companies outside of their industry. By that, the team gained insights about the benefits of using automation and big data analytics. The company’s management realized that although best practices have always been used to improve operational performance and employee safety, bringing in advanced modern technologies can achieve competitive advantage in a traditionally conservative. They decided to implement technologies such as autonomous transmission (for conveying cargo), advanced operation management systems and automated drills. They have formed a new growth and innovation group within the company, that aimed to rethink old processes. They named the program “mine of the future.”
In a couple of conferences during the past year, Sam Walsh and current head of innovation Stephen McIntosh, have provided a glimpse towards the next leap Rio Tinto is currently doing – using machine learning to keep improving. It appears that Rio Tinto iron ore business itself generates about 2.4 terabytes of data every minute, and the volume of seismic data crunched in oil exploration tends to add up to petabytes [2][3]. This ocean of data is worthless, as McIntosh called it, in its raw form. With that, the data could drive valuable insights that will help reducing maintenance and production costs from unplanned breakdowns. Using machine learning and advanced modeling Rio Tinto has been able to identify problems before they occur.
Rio Tinto is solving operational and control problems with machine learning as well [4]. For instance, in an interesting thought process, they have presented a challenge to five different groups of researches the company gathered from universities and startups.
The problem presented was the trucks’ carryback issue: Haul trucks dump their loads at processing plants and should leave empty. However, often some of the dirt sticks in the tray (McIntosh: “I am talking about tons at a time”). Carrybacks reduce the capacity of the vehicle for its next load. On some of the company’s sites, it costs Rio Tinto 5% of the haulage capacity.
The groups have used machine learning tools to provide possible solutions, and eventually came up with a new capability of predicting potential ‘carryback’ in a well-over 90% accuracy rate.
It seems that the implementation of machine learning methods at Rio Tinto is in its infancy, while the potential inherent in its various applications in the mining industry is enormous. At present, Rio Tinto uses machine learning mainly to optimize the company’s operations especially with its autonomous drills, loaders, and trains infrastructure. They do this while the mining industry as a whole has already begun to use machine learning for purposes the some might define as more sublime, such as locating high-quality natural resources through past data analysis and prediction of optimal locations. There are many applications for machine learning in the mining industry: Goldspot uses IPM’s Watson for gold and mineral exploration. TOMRA uses AI for sorting and filtering. [5]
hoI think that eventually, Rio Tinto will have to get a foothold in these areas as well. As it gets harder to produce natural resources, the efforts in the coming years should be focused their there. However, it might be not that easy to do. Current CEO, Jean-Sébastien Jacques warn in an interview that Australia has not enough data scientists to support the company’s aspirations in expanding its machine learning effort, and despite governmental assistance and reaching out to local universities, not much progress has been demonstrated [6]. He noted that reaching overseas might not be helpful as well, as big enterprises like Microsoft manage to better garner the interest of leading data scientists.
The coming years are a turning point. As Jean-Sébastien noted, in several years about two-thirds of his software developers should be data scientists, if the company wants to keep up the pace of the industry [7]. With the lack of personnel, it seems that there are merely two possible paths he can follow.
- He can form a Coopetition – competitors’ collaboration. It is reasonable to believe that the company shares concerns with its rivals, and as alone they cannot fight the headhunters of hi-tech giants, together they might find more success.
- He can outsource part of the data mining and machine learning development of the company, as Golspot did with Watson.
What way would be better for the company future? What should the CEO do?
(798 words)
[1] Brookes, J. (2018). The Innovation Culture that spurred Rio Tinto’s Mine of the Future – Which-50. [online] Which-50. Available at: https://which-50.com/innovation-culture-spurred-rio-tintos-mine-future/ [Accessed 12 Nov. 2018].
[2] McLean, A. (2018). Rio Tinto preparing for the Mine of the Future with automation | ZDNet. [online] ZDNet. Available at: https://www.zdnet.com/article/rio-tinto-preparing-for-the-mine-of-the-future-with-automation/ [Accessed 13 Nov. 2018].
[3] Winterford, B., Caruana, A., Winterford, B. and Caruana, A. (2018). Rio Tinto’s big data play delivers promised ore. [online] iTnews. Available at: https://www.itnews.com.au/news/rio-tintos-big-data-play-delivers-promised-ore-392359 [Accessed 13 Nov. 2018].
[4] Riotinto.com. (2018). [online] Available at: https://www.riotinto.com/documents/180514_Speech_Stephen_McIntosh_Breaking_from_tradition_The_Mine_fo_the_Future.pdf [Accessed 13 Nov. 2018].
[5] Walker, J. (2018). AI in Mining – Mineral Exploration, Autonomous Drills, and More. [online] TechEmergence. Available at: https://www.techemergence.com/ai-in-mining-mineral-exploration-autonomous-drills/ [Accessed 13 Nov. 2018].
[6] Crozier, R. (2018). Rio Tinto faces big shortfall of data scientists. [online] iTnews. Available at: https://www.itnews.com.au/news/rio-tinto-faces-massive-shortfall-of-data-scientists-487404 [Accessed 13 Nov. 2018].
[7] Financial Review. (2018). [online] Available at: https://www.afr.com/business/jeansebastien-jacques-is-redesigning-rio-tinto-for-the-new-world-order-20180701-h123db [Accessed 13 Nov. 2018].
A man after my heart! Great write up Gavriel. This space is very interesting and a lot of the major mining companies are following Rio’s lead. Barrick Gold Corp and BHP to name a few. With the automation of equipment in open pit settings the ability to interact with the type of data you’re describing and even intervene is significantly higher than it has traditionally been. However, the point around needing expertise to get these types of systems up and running is well founded. The reality is even with the data coming from existing system (mine dispatch software – similar to what Uber dispatching) that the horsepower and expertise does not exist on the company level to make use of it.
A good place to start is programs like Tableau. They’re intuitive and thereby quick to learn. Mining Engineers with a little training can be brought up to speed quickly and used to begin tapping in to the wealth of data available today. Somewhere to start at least!
Interesting piece Gavriel. You point out some clear benefits of machine learning for Rio Tinto. The shortage of talent makes me wonder for a company like Rio Tinto, would they be better off outsourcing the AI work? Just as many other companies has decided to stick to what they are good at, is this an instance where Rio just doesn’t have the core competency and will they always be struggling to keep up? Also, is there some unfortunate trade offs to be made? Maybe only the most promising projects get precious AI resources?
Thanks for writing this Gavriel! The competitive dynamics of whether or not to cooperate with other companies regarding machine learning is very interesting to me. I am concerned about cooperating with partners, either with IBM or with other mining companies, unless Rio Tinto believes that it cannot develop a comparative advantage in driving down operating costs or identifying potential new deposits. Currently, if Rio Tinto is the best at deploying data to improve operating costs, I believe that they should hold that information closely so that they could be the low cost provider and generate higher profits. If Rio Tinto begins to share this information, or even the scientists deploying the software behind this information, with comparable companies, they will be unable to use data to beat the competition. In this line of thinking, I would even be concerned about using a company like IBM to help them: it’s likely that they’d be able to use your data to provide competitors with better recommendations than if you hadn’t provided that information in the first place.
In summary, I think Rio Tinto needs to decide whether or not data is a comparative advantage for them (or needs to be such an advantage), regardless of data scientist availability issues, before they decide whether or not to cooperate with other companies.