Rare earth elements industry and AI

By | January 4, 2018

Rare earth elements (REE) have grown in importance over the years due to their increased usage in producing many key technologies in today’s world. These elements, which are not actually that rare, can be mined through various methods. As the world increases its development of various technologies, the increase in demand for REE is expected to accelerate. Because of this, the REE industry now faces a problem of keeping up with future demand.

In this post I’ll explore how the increase in demand for REE presents business opportunities for companies developing AI applications that can improve the REE mining process.

Rare earth elements (REE) overview

REE are a collection of 17 elements widely used in the production of modern technology. Many have properties that make them ideal for producing high quality magnets. Others are useful in the production of fuel cells, mercury vapor lights, electronics, glass strengthening techniques, computer applications, hybrid cars, advanced smart weaponry, nuclear power plant and weapons, medical technology as well as a myriad of other applications.

Examples of REE include dysprosium (renewable energy hardware), neodymium (magnets), lanthanum (cameras/telescopes), terbium (fuel cells), europium (optoelectronics), yttrium (superconductors) and praseodymium (hard metals in aircraft engines).

Major players in the industry (mainly China)

Currently, China is the world leader in producing REE with a dominant market share. According to the 2016 US Geological Survey, China produced over 105,000 MT of rare earth metals, which accounted for a market share of ~85%. Other players in 2016 include Australia (14,000 MT), Russia (3,000 MT), India (1,700 MT) and Brazil (1,100 MT).

The U.S is notably absent from the above list. Although the U.S has some REE reserves, the only U.S based REE company (Molycorp) filed for bankruptcy in 2015. As such the U.S imports 100% of its REE, a figure that the government has cited as an area for concern. The government has responded to these concerns in a few manners. The most recent initiative was when the Trump administration signed an executive order on December 21, 2017 to develop a federal strategy that ensures “secure and reliable supplies” of these critical minerals.

Demand forecast (Adamas Intelligence report)

With a world that’s increasing its implementation of technology in our daily lives and also increasing its reliance on renewable energy to combat climate change, it’s logical to assume the demand for REE will increase. Government policies reinforce the push towards renewable energy and many forecasts expect 50% of all new global REE demand growth over the next 10 years to be directly or indirectly led by government-led actions.

Surprisingly, China is expected to become a net importer of REE by 2025 (keep in mind they are by far the largest producer). Neodymium domestic demand for permanent magnets alone is poised to exceed total global production of neodymium by 9,000 tones in Adamas’ base case scenario.

From 2016 through 2020 demand for neodymium, praseodymium, dysprosium, and lanthanum will grow relatively strongly, but, from 2020 through 2025 the rate of global demand growth for these rare earths will accelerate year-over-year, resulting in major annual demand increases by 2025 that can only be satisfied by the continuous and accelerated development of new mines. This presents many opportunities and a strong demand for AI solutions that improve the efficiency of different parts of the REE mining process.

AI use cases in traditional mining
AI use cases in traditional mining are numerous. 5 main areas identified include:

  1. Prospecting and exploration – Answers the question of “where to explore”
  2. Discovery and Advanced Exploration – Answers the question of “what’s in the ground”
  3. Development/Construction – Answers the question of “how to build the mine”
  4. Operation and Production – Answers the question of “how to mine, mill and process the discovered ores”
  5. Reclamation – Answers “how do we rehabilitate the land and protect the environment, people, and animals”

A more detailed table can be found below:

Most relevant use cases for REE mining

There are many potential AI use cases for REE mining. A few of the most important include using AI to improve mineral exploration, environmental impact assessment (see reclamation category above) and advanced extraction.

Mineral exploration is the key first step in identifying sources of REE. Historically, mineral exploration has been conducted by relying on the intuition of geologists and using only basic heuristics, models and data sets. The financial upside is large for companies who can more accurately identify mineral deposits and the REE that lie in them.

Assessing different environmental impacts of REE mining is most applicable not only to China, but also to any country entering the industry who wants to avoid the environmental mistakes China made. One example of the consequences of these mistakes include the production of toxic waste and acidic waste water. Extraction of REEs includes acid baths and hydro-metallurgical techniques, both of which can be harmful to the environment.

Advanced extraction is another area that can benefit from AI. As an example, one could look at Japan’s recent discovery of REE found on the Pacific seafloor off eastern Japan, in an area covering around 950 sq. km, about half the size of Tokyo. While a large area, extracting REE from the deep sea is difficult and costly, not unlike drilling for oil in the deep sea.

AI startups in mining / REE mining

I’ve included a list of startups that are currently developing AI applications for use in the mining industry. None are currently focused on REE mining specifically, but they all have the potential to tailor their AI applications to land REE mining companies as clients.

Earth AI

Earth AI is an AI-powered mineral exploration platform. It passes a variety of data from geological maps, satellite imagery, geophysical and geochemical datasets through a neural net and auto-generates suggested locations for likely mineral deposits. Each of Earth AI’s customers uploads their own data (soil samples etc.) which increase the accuracy of the system’s predictions. Over time, this effect compounds, leading to more and more accurate predictions for all customers.

Earth AI was founded by Roman Teslyuk, who has been described in this medium article by Air Tree Ventures as the perfect combination of hustler and hacker. To find clients he first cold called a list of mining exploration companies eventually landing ActiveX as a launch customer. Earth AI now has dozens of mining firms using its software. If they could land any of the large Chinese REE mining companies as a client, that could be huge.

Earth AI is based in Australia and has raised $500k in seed money to date. Air Tree Ventures is their main investor.

Goldspot Discoveries

Goldspot Discoveries is revolutionizing the mineral exploration business by utilizing machine learning to target on a regional and localized scale. They were a finalist for the 2017 Disrupt Mining Competition. Their Goldspot Algorithm significantly decreases risk, while increasing the efficiency and success rate of mineral exploration.

Goldspot’s algorithms have been used for the Jerritt Canyon Project, majority owned by Sprott Mining inc. Jerritt asked Goldspot to assess a significant amount of data in order to assist with continued exploration. Goldspot consolidated over 30 years of historical remote sensing, mining, and exploration data into one comprehensive and functional geological model. Goldspot Artificial Intelligence was then able to use this geological model to identify correlations in the data layers of existing and historically mined deposits.

Total funding for the company is unknown. Major investors include ThreeD Capital, an emerging merchant bank, who announced a 22% stake in Goldspot in September 2017.

Kore Geosystems

KORE was awarded $500,000 from 2017 Disrupt Mining Competition for its idea of installing measurement instruments onto drill rigs for real-time automated geological data acquisition. According to the company’s President, Vince Gerrie, the solution would fast-track exploration campaigns while reducing costs. Gerrie expects to have KORE’s technology up and running at a Goldcorp mining site by year end.

Total funding for KORE is unknown. If their technology is proven to work for Goldcorp, many opportunities for expansion will become available. In addition to traditional mining, KORE might be able to tailor their technology to REE mining and capture the increased market demand for AI solutions in the industry.

Ozius Spatial

Ozius Spatial’s Naxia is a machine-learning artificial intelligence solution for environmental monitoring. Naxia consumes existing field data and uses remote-sensing algorithms, supported by machine learning to enable large tracts of land to be analysed for environmental risk, consistently and efficiently without needing to be on-site. Data is fused with imagery from satellites, aircraft and drones to create new landscape intelligence. Naxia reduces the subjectivity and misclassification that can come from off the shelf spatial systems alone. Information generated by Naxia can also integrate directly into existing enterprise systems.

Larger companies investing in AI solutions for mining / REE mining

Rio Tinto

By far the industry leader in investing in AI technology. Jean-Sébastien Jacques, Rio’s chief executive, says it is ten years ahead of mining rivals in autonomous technology. So far the company has utilized autonomous trucks and drills to increase efficiency. The 76 autonomous vehicles in Rio’s 400-strong truck fleet in the Pilbara are an estimated 15% cheaper to run than the rest. The company is also upgrading the locomotives that haul ore hundreds of miles to ports. Once finished, these upgrades will allow the trains to drive themselves, and be loaded and unloaded automatically.

Yandex data factory

Yandex Data Factory is a subsidiary of Yandex, a Russian search engine that is among the largest in the world. They provide AI-based solutions that directly increase productivity, reduce costs, and improve energy efficiency. Gazprom Neft and Yandex Data Factory have agreed to start collaboration in developing big data analytics, applying machine learning and artificial intelligence to drilling and completion, as well as using these technologies to optimise other production processes. The agreement was signed at St. Petersburg International Economic Forum by Vadim Yakovlev, First Deputy CEO of Gazprom Neft, and Alexander Khaytin, Chief Operating Officer of Yandex Data Factory

TOMRA

TOMRA has developed separation technology for REE mining. It’s been proven to increase efficiency, precision and speed. Materials which pass through the sorting process are often heavy, dusty and abrasive, and thus demand a technological design which is incredibly robust in every respect. TOMRA Sorting Solutions combine material handling, recognition and pressurized air ejection technologies in a thoroughly optimized system which reliably separates valuable mineral ores from waste rock.