Miners are buying or developing AI tech to assist in everything from drillcore assays to autonomous rail haulage. Experts suggest that worker data-science training and empowerment might be the most beneficial investment of all.
By Russell A. Carter, Contributing Editor
Miners are trained to be familiar with tools. For anything from a jackleg drill to a production spreadsheet app, the principles are similar: Set and control the tool to do what you want it to do. When the job is done, put the tool aside and move on.
But suppose a tool requires a relationship — a constant back and forth dialogue about how best to approach a job and measure the value of the results? And suppose, in the end, that tool turns out to be smarter than the user? When it comes to Artificial Intelligence (AI), a tool that’s increasingly important for business success, these aren’t just remote possibilities.
They’re almost a given, if you take into account the rate of increase in industry interest and implementation of AI and the speed and scope of AI development. Mining is adopting AI technologies at an unprecedented rate, its suppliers are expanding or realigning their product and services portfolios to support that trend, and various social and technological trends are funneling the industry’s future format toward a mostly automated, robot-assisted scenario in which AI’s predictive algorithms control where to drill for, how to process and where to ship mineral products.
For the foreseeable future, however, AI won’t be a set-and-forget tool. Cliff Justice, U.S. leader for enterprise innovation at KPMG, offered this prediction on a recent episode of the MIT Technology Review Business Lab podcast: “The ecosystem that got you to a level in more of an analog-centric world is going to be very different in a more AI-centric world.
“That AI-centric world is going to accelerate everything digital has to offer. What I mean by digital are the new ways of working — the digital business models, the new ways of developing and evolving commerce, the ways we interact and exchange ideas with customers and with colleagues and coworkers. All of these are becoming much more digital-centric, and then artificial intelligence becomes one of the mechanisms that evolves and progresses the way we work and the way we interact. And it becomes a little more like a relationship with technology, as opposed to a tool that we program because AI is something that evolves and learns and develops the more it gets exposed to humans,” he explained.
Most social scientists will say that an important key to improving relationships is good communication between two parties. However, there are signs that future human relationships with AI may be skewed towards technology having the upper hand: Neural networks, for instance, are data-hungry, powerful AI tools that can be applied to problems that would simply take too long and require too much effort for humans to efficiently solve. But, as a recent article on the Scientific American* website pointed out, neural networks aren’t always forthcoming about how they produce results. Like a stubborn child, they might present a perfectly good answer to a problem but won’t tell you how they reached that solution.
According to author Robin Blades, “In the past 10 years, machine learning has become an extremely popular tool for classifying big data and making predictions. Explaining the logical basis for its decisions can be very difficult, however. Neural networks are built from interconnected nodes, modeled after the neurons of the brain, with a structure that changes as information flows through it. While this adaptive model is able to solve complex problems, it is also often impossible for humans to decode the logic involved.
“This lack of transparency has been nicknamed ‘the black box problem’ because no one can see inside the network to explain its ‘thought’ process. Not only does this opacity undermine trust in the results — it also limits how much neural networks can contribute to humans’ scientific understanding of the world.”
AI also poses the possibility of disrupting the traditional pattern of workforce roles. David Degerfeldt, program manager-artificial intelligence in mining at Boliden, touched upon this during a webinar held earlier this year titled AI: A Necessary Enabler for Sustainable Mining Production, sponsored by Swedish Mining Innovation, a joint partnership among Swedish government innovation agency Vinnova; Formas, a research council; and the Swedish Energy Agency.
Degerfelt outlined Boliden’s recent progress in AI initiatives in terms of an AI Maturity Timeline chart developed by the Global Mining Guidance Group (GMG) for its 2020 report Foundations of AI – A Framework for Ai in Mining. He estimated that his company is currently at Level 2, with a goal of eventually reaching Level 6, which is beyond the scope of the GMG chart and represents a fully autonomous operation.
He wondered, at that point, who the “heroes” of mining would be. “We would have a fully automated mine, with very few people on site. There would be no shifts. Machines controlled by AI don’t require breaks, they work 24/7.
“Of course, the heroes in today’s mine are the operators,” he noted. “They do all the heavy work and without them we couldn’t do anything. But in an automated mine, who would be the heroes? Consider the IT group, for example. It’s usually regarded as a support organization and not involved in production. In an automated mine, however, they might be considered the heroes that keep the operation running, while the operators would just step in when things go wrong on the ground. Would operators then be considered a support function?”
Unrealistic expectations could also cause investment in AI to waver. As Javier Pigazo Merino, FLSmidth’s technical product line manager, group digital, warned in a recent online post, prospective AI customers need to be aware of the “Hype Cycle.”
“Especially with emerging technologies and trends in the industrial landscape, we hear bold promises from marketing materials or sales presentations — sometimes inherited from other sectors where maturity levels and/or conditions are far from similar,” Merino said. “This can make it very difficult for a non-technical audience to discern hype from what is technically viable and commercially profitable for their specific business needs.
“This overinflation of expectations, combined with low resistance to failure, leads to huge doses of frustration and early dropping of the investment, even before the learnings are incorporated into a new iteration or before a good productivity level is reached,” he explained.
Looking Downstream
The industry’s implementation of AI and other new, disruptive technologies may also have unanticipated downstream impact on ESG strategies and supply chain arrangements, according to a 2021 report from the International Institute for Sustainable Development, an independent think tank. The study, titled New Tech, New Deal: Mining Policy Options in the Face of New Technology, is the result of a project conducted as a partnership of the IISD, the Intergovernmental Forum on Mining, Minerals, Metals and Sustainable Development (IGF), the Columbia Center on Sustainable Investment (CCSI), and Mining Shared Value/ Engineers Without Borders Canada.
The authors pointed out that the advent of new technologies could pose significant implications for mining supply chains and is likely to transform the procurement function at the site and corporate levels. Three aspects of procurement will particularly be affected. In their words:
“Strategic sourcing — i.e., what mining companies buy, from where, and through the most effective market analysis — is becoming more predictive. This means mining companies are able to better plan and manage their procurement needs, select their suppliers, and secure the most competitive prices. Centralized decisions are expected to increasingly be made at corporate level, to then be redirected at the country or site level. Unless mandatory local procurement policies require them to do otherwise, this may reduce local sourcing, as local suppliers may not be able to compete with new global suppliers identified by AI and thus far unknown. Smaller suppliers may not have sophisticated enough structures to participate in digital platforms to serve mining operations.
“Second, with digitized sourcing platforms and payment systems, transactional procurement is being automated, allowing companies to centralize, analyze and structure their procurement orders more efficiently. Integration of different systems helps reduce risks and costs. Local suppliers may not have the same level of digitization and secured payment systems, which disadvantages them.
“Third, leaner systems are redefining mining companies’ relationships with their suppliers. Collaborative platforms and suppliers that are able to provide tailor-made solutions or proactive innovative technologies will benefit more from digitized procurement functions than traditional and small local suppliers.”
Those changes, the report’s authors contend, will have profound implications for local procurement strategies, from both mining companies (as buyers) and local suppliers, who will have to adjust their business models accordingly.
Sorting Out Strategies
Despite the potential for encountering these speed bumps, various surveys point to a high rate of AI interest and adoption throughout the mining industry — which is somewhat surprising given the industry’s reputation for technological risk aversion. For instance, earlier this year, Aspen Tech commissioned an independent survey of 200 North American and European IT and Operations decision-makers from across the industrial sector, including companies in construction and engineering, chemicals, energy, oil and gas, metals and mining and others. The results showed that nearly all respondents recognized the importance of Industrial AI to their organizations — 99% could name at least one business driver for adopting an Industrial AI strategy — but few actually had one. Overall, the survey indicated that:
• 79% said they either have an Industrial AI project live right now or are piloting one. Only 1% of respondents said they have either no initiatives for Industrial AI in place, or no plans to develop one.
• Although Industrial AI depends on ingesting quality data, much data is often left undiscovered due to a number of structural challenges. Nearly all (98%) of the IT and Operations decision makers polled said their organization was experiencing at least one of these industrial data challenges, inhibiting their data quality and management practices.
• In the same vein, on average, the survey respondents estimated that they only had visibility over about 66% of their organization’s industrial data. That means industrial organizations, on average, have little to no visibility over one-third of the industrial data they’ve collected.
• More than eight in 10 said Industrial AI has played a significant or major role in their organization’s broader digital transformation strategy in the last three years.
According to FLSmidth’s Javier Merino, the benefits of AI are clear — particularly in processing applications — once certain misconceptions are overcome. For example, he pointed out that “APC [Advanced Process Control] systems are very often seen as one of the main drivers needed to reach the dream of autonomous operations. In this context, it’s commonly heard in the media that AI is replacing APC systems. But this wrongly assumes that AI is already a synonym for fully-autonomous operations. This kind of misrepresentation does not help, as such fully autonomous continuous-process plants are still not that close to reality.
“However, there are many examples where new technologies and workflows can heavily enhance the level of information that is gathered and analyzed, transforming it into much better actionable insights, to take decisions faster than ever,” he said. “This is what we call ‘intelligence augmentation’ and can clearly assist and elevate the performance of either existing APC systems or human-based control.”
There are three main areas where APC can and will benefit from AI, according to Merino:
Cognitive augmentation – The ability to gather, analyze and combine various data streams in real time can bring relatively quick benefits from operational and safety perspectives. One example would be building new virtual sensors to replace unreliable or unavailable signals, particularly when the instrumentation is placed in risky areas or is often out of service.
Smart controllers – In certain contexts, controllers, such as linear and non-linear MPCs, can be enhanced and complemented by virtual models of machinery or processes, known as digital twins. If the digital twins are done well, they can be used to find the controller’s optimum parameters, which leads to more stable processes, achieves higher production and quality levels, or decreases the amount of energy or water used.
Dynamic adaptiveness – Many industrial processes are by nature nonlinear and time-varying. This means that actions that were optimal to achieve specific goals yesterday (or even an hour ago) may be suboptimal or even inefficient now. The ability of AI technologies to continuously adapt to changing conditions to find the optimal operating parameters and targets is one of the key areas in which AI can improve the ability of APC systems to optimize cement and mining processes.
Scoping Industry Interest
Although mine operators may be technologically conservative by nature, cost-cutting and profitability are powerful magnets for attracting operational investment. Potential benefits lie in the areas of exploration, mobile and fixed asset optimization, worker safety and environmental compliance, among others. Now that we’ve seen some of the pros and cons of AI and sampled its rate of uptake, it’s fair to ask “where is the mining industry now in terms of AI integration?” A fair answer would be “all over the place.”
That’s not a flippant observation; it just reflects the widening scope of AI application capabilities and possibilities open to the industry. To provide a glimpse of how wide-ranging the industry’s interest in AI applications has become, here’s a quick rundown on just a few of the latest developments.
Minerva Intelligence, a knowledge engineering firm based in Vancouver, British Columbia, recently released the results of an evaluation of Freeport Resources’s Star Mountains property in Papua New Guinea. Minerva used DRIVER, its Al software, to perform an evaluation of multi-element drilling data. In addition, Minerva reinterpreted existing geophysical information on the project and completed traditional K-Means Cluster analysis on the multi-element data.
Gord Friesen, CEO of Freeport Resources, a Canadian junior company, said his company found Minerva’s analysis and the DRIVER system to be “very useful” for understanding the project. “DRIVER validated our geologists’ interpretation of the deposit zonation and gave indication of mineral potential beyond known resources on our properties and confirmed it in a fraction of the time. The synthesis of independent methodologies was a valuable contribution to the project and gave us confidence about the results.”
In September, the Silicon Valley investor-backed AI startup Kobold Metals announced it would work with BHP to discover battery minerals such as copper and nickel in various districts around the globe. Kobold said its data platform, TerraShed, aggregates and structures vast collections of scientific data and makes it rapidly available for analysis. Then Machine Prospector, a suite of algorithms, interrogates that data with a range of techniques — from ensemble machine learning, to full-physics joint inversions, to computer vision — to predict the composition of the subsurface in a statistically valid manner.
Kobold’s principal investors include Breakthrough Energy Ventures, a climate technology fund backed by Bill Gates and Jeff Bezos, among others; along with Andreessen Horowitz, a major Silicon Valley venture capital fund; and Equinor, the Norwegian state oil company.
Earlier this year, KoBold announced a partnership with iCRAG (Irish Research Center for Applied Geoscience) researchers to explore for critical minerals. iCRAG researchers will use the microscopy, geochemistry and spectrometry facilities in the iCRAG Labs at Trinity College Dublin to investigate minerals containing cobalt, as well as other elements often found proximity to cobalt, such as nickel and copper. In particular, the research project will focus on examining mineral samples from the Kisanfu cobalt-copper deposit in the Democratic Republic of Congo to better understand how cobalt deposits form. The analysis will form the basis of machine learning techniques developed by KoBold to gain insight into the exploration of critical raw materials in similar deposits around the globe.
Australian drill-tech company IMDEX recently paid about $20 million to acquire MinePortal software from California-based Data Cloud International. IMDEX describes MinePortal as a new-generation native cloud application that enables geological data modelling and real-time 3D visualization. Among its features are:
• Capacity to process high volumes of data in a cloud environment, while applying geostatistical and machine learning algorithms to identify orebody trends.
• Integration with IMDEXHUB-IQ to deliver a connected real-time orebody knowledge ecosystem.
• Ability to process IMDEX BLASTDOG data and other data sets, including MWD data and other IMDEX sensor data.
IMDEX said BLASTDOG is a semiautonomously deployed system for logging material properties and blast hole characteristics at high spatial density across the bench and mine and is commodity agnostic. It has been developed in collaboration with Universal Field Robots and tested at mines in Queensland, Western Australia, Chile and Nevada. At an industry event held in November, IMDEX said BLASTDOG will advance from engineering development to commercial prototype by the end of the year. Long-term, the company expects its acquisition of MinePortal to enhance the value of BLASTDOG for clients by translating sensor data into 3D visualization models.
Managing Mobile Assets
Last year, Caterpillar introduced MineStar Edge to augment its MineStar Solutions suite of technologies and to align with the way many mining operations manage their businesses. Edge, according to Cat, creates an operational ecosystem for mining companies. Rather than having data in individual silos, Edge brings visibility to the entire mining operation and enables managers to see how activities early in the process impact those further down the value chain. Edge enables supervisors to access real-time information from a computer or tablet from any place with an internet connection.
MineStar Edge also leverages cloud computing and AI so it can grow as it collects data, identifies patterns and learns to make decisions. These capabilities, said Cat, enable mine managers to focus on improving operations rather than collecting and interpreting data. Edge also automates data collection, which ensures accuracy, unburdens personnel and enables managers to trust the information they receive.
Cat claims that because Edge is delivered as a cloud-based, subscription managed application, it reduces costs of deployment, service and training. Caterpillar handles all office-based deployment, support, updates and upgrades. Customers select an offering by role, function or task — paying only for those functionalities they need.
Cognecto, a Bangalore, India-based company that has developed an AI-based platform that provides real-time analytics solutions and managed services related to heavy equipment, announced the launch of Machine Link, which it said is custom-built for the demanding environments of mining, material handling, construction, and logistics. According to the company, Machine Link will enable fleet owners to retrofit new and aged equipment with relevant sensors. It comes with the inbuilt capabilities of an advanced IoT device and easy to integrate tire pressure sensors, fuel sensors, load sensors, RFID sensors, and open channels to tap into the various electrical and mechanical sensors.
Cognecto said the modular design of the edge device helps customize the sensor inputs as per customer requirements and delivers ROI within few months. The base model enables connecting all the assets at the site to ensure visibility to complete material/process flow. It is pre-configured to connect with the Fleet Management Platform and is intended to help Cognecto remotely manage and upgrade the installed device without complex local support.
In June, Rio Tinto’s Iron Ore AutoHaul Rail announced a feasibility study of AI vision systems for hazard detection in its Pilbara, Western Australia, heavy haul railway network. AI Systems Ltd. is participating in the trial, now
under way.
The feasibility study involves the trial of AI Systems’ AI solution installed on a locomotive to validate its capability to detect and classify objects in the rail corridor ahead of the train.
Train control specialist company 4Tel Ltd. formed AI Systems as a special purpose company based in Newcastle, New South Wales, aimed at commercialization, management and further development of AI intelligence vision systems and intellectual property previously developed by 4Tel.
AI Systems’ technology is focused on object detection software, using a patented sensor array. The detection system can be fitted to both moving vehicles and to stationary locations such as level-crossings, stations and large intersections.
From Pit to Process Plant
The Weir Group is acquiring Motion Metrics, a Canadian developer of innovative AI and 3D technology used in mines worldwide. As part of the agreement, Motion Metrics’ Vancouver headquarters will become Weir’s global center for excellence in AI and machine vision technology.
Motion Metrics produces smart, rugged cameras that monitor and provide current data on equipment performance, faults, payloads and rock fragmentation. This data is then analyzed using embedded and cloud-based machine learning to provide real-time feedback to the mining operation. This enables immediate identification of potential issues that could impact safety and cause expensive unplanned downtime. This includes boulder and foreign body detection to dislodged ground-engaging tools that can critically damage crushing equipment if undetected. According to the company, it also provides information that can be used to optimize asset efficiency, supporting better decision making as miners seek to increase productivity while reducing energy consumption, particularly in areas such as comminution.
These technologies were initially developed for GET applications but have recently been extended into a suite of products and solutions that can be applied from drill and blast to primary processing. Motion Metrics technology is currently used on more than 80 mine sites.
Weir said Motion Metrics will join its ESCO division, reflecting the early adoption of its technology in ground engaging tools. Motion Metrics’ AI and machine vision capabilities are expected to be leveraged across the whole mining value chain served by the Weir Group.
Moving forward, Weir said ESCO’s focus is on accelerating growth through geographic expansion and the continued extension of its front-of-shovel offering. The Motion Metrics acquisition, according to Weir, is fully aligned to this strategy, including the early realization of the digitally instrumented smart bucket concept, which provides information on bucket and GET health alongside payload and ore fragmentation analysis.
Global chemical company BASF and IntelliSense.io, an industrial AI company, announced a partnership called the “BASF Intelligent Mine Powered by IntelliSense.io” that delivers AI solutions embedded with BASF’s mineral processing and chemical knowledge.
The partners said BASF Intelligent Mine Powered by IntelliSense.io is an open, real-time, decision-making platform that can be configured for individual sites, typically within three months. Each plant process, such as grinding, thickening, flotation and pumping, is supported by an Optimization as a Service (OaaS) application, which predicts and simulates future performance, generating process-specific recommendations for insights and optimization. As multiple OaaS applications link together, customers can generate efficiency gains throughout the entire mine-to-market value chain.
Remote operations access allows for 24/7 visibility of mine operational and financial performance, with BASF process experts available to provide real-time support. Additionally, the in-built simulation tool can be used to test alternative operating conditions, train staff and run non-intrusive “what-if” scenarios.
The AI solutions are based on a hybrid cloud architecture, enabling both on-site and cloud deployments.
Meanwhile, in response to studies that suggest industrial companies typically are able to use only about 20% of the data they generate, ABB has developed a solution called Ability Genix Industrial Analytics and AI Suite — a scalable advanced analytics platform offering pre-built applications and services. According to ABB, it collects, contextualizes and converts operational, engineering and information technology data into actionable insights, and the use of AI produces meaningful insights for prediction and optimization that can help improve business performance. Customers can subscribe to a variety of analytics on demand, as business needs dictate, speeding up the traditional process of requesting and scheduling support from suppliers. Genix supports a variety of deployments including cloud, hybrid and on-premise.
ABB additionally bolstered its digital offerings with the recent launch of the Ability Genix Asset Performance Management (APM) Suite for condition monitoring, predictive maintenance and comprehensive asset performance insights for the process industries and others. Genix APM is built on the Genix Industrial Analytics and AI Suite.
“The ABB Ability Genix Suite brings unique value by unlocking the combined power of diverse data, domain knowledge, technology and AI,” said Rajesh Ramachandran, chief digital officer for ABB Industrial Automation. “We have designed this modular and flexible suite so that customers at different stages in their digitalization journey can adopt ABB Ability Genix to accelerate business outcomes while protecting existing investments.”
A key component of Genix is the Ability Edgenius Operations Data Manager that connects, collects and analyzes operational technology data at the point of production. Edgenius uses data generated by operational technology such as DCS and devices to produce analytics that improve production processes and asset utilization. It can be deployed on its own, or integrated with Genix.
Planning a Strategy
Big-data information science companies are buying startups or competing tech developers on an almost weekly basis. One of the larger AI-related M&A transactions of interest to the mining industry is AVEVA’s recent $5 billion acquisition of OSIsoft, finalized earlier this year. OSIsoft, a California-based company founded in the early 1980s by Dr. Patrick Kennedy, developed the PI System, a widely used data management platform for industrial operations. AVEVA, based in the U.K. with a majority shareholder and strategic partner in Schneider Electric, specializes in engineering, design and operations software for heavy industries and has nine of the top 10 mining companies in the world as customers. The acquisition allowed AVEVA to add OSIsoft’s PI System platform to its existing software product lineup, which includes AVEVA System Platform, AVEVA Production Management, AVEVA APC, AVEVA Predictive Analytics, AVEVA Unified Engineering and others.
E&MJ recently had an opportunity to speak with Martin Provencher, AVEVA’s global head of mining, about some of the general challenges and choices mining companies may encounter when considering an AI strategy.
E&MJ: The use of AI technologies is quickly gaining momentum in the mining industry. For companies that are just formulating an AI strategy, or may have already tried AI but failed to achieve expected results, what are some recommended steps to help achieve success?
Provencher: There are a lot of challenges that need to be considered, but they mainly fall into three areas. First, a company needs to identify its goals for using AI — is it interested in cutting down-
time and increasing reliability of its equipment, for example, or mitigating operational risk, or in achieving autonomous operations?
Then, it should determine what level of analytics it needs to achieve its goals. Are predictive analytics required, and if so, how powerful do they need to be? Does a company’s AI strategy require the use of perceptive analytics, which enables users to analyze audio or visual data in addition to standard digital input? For example, AVEVA can provide a library of solutions that utilize both preset conditions and variable machine learning to trigger action for possible issues related to asset operations.
And finally, a company needs to take a hard look at what type of organizational structure will be needed to achieve its goals with AI. Is it prepared to assemble a system in house, or go with an existing system that can be tailored to its needs? Does it have the resources to train its staff to maintain an in-house system? I work exclusively with our mining customers and although they all have roughly similar long-term concerns — production efficiency, environmental compliance, things like that — I help them identify their principal objectives and then recommend the type and level of information technology required to reach them.
E&MJ: AI depends heavily on access to large volumes of data, some of which may need to be shared. Mining companies have traditionally been reluctant to share data with anyone. Do you see any signs that this hesitancy is diminishing?
Provencher: I do see it starting to change. We’re not completely there yet. Many companies have an operational data platform in place, and OEM vendors are trying to obtain access to that data. But it can still be a bit of a struggle as data continues to be siloed among various organizations, and OEMs increasingly are turning to connected machines to directly acquire the information they need. So, mining customers who buy that equipment are already aware that data will be going directly to the vendor. I have had many discussions with users of our PI System on how to make that arrangement work most effectively for them. It’s always best to have only one version of the truth, and part of that truth may be through access to external data that enables a customer to improve equipment efficiency. I think the industry will eventually reach a “sharing” perspective.
E&MJ: Is a corporate culture change often necessary for success in AI?
Provencher: Success in this area usually only comes with strong support from top management. It’s a transformative process based on a firm business decision, but I’ve seen many companies that assume once they’ve bought into an AI platform, everything will just naturally fall into place and their staff will know what to do with the data and produce the kind of information needed. But, companies first need to empower their people to not only understand and work with the AI platform, but to be able to add insights — based on their expertise — to the process. They need to become, in effect, digital engineers.
E&MJ: What IT infrastructure or other supporting technology usually needs to be in place to carry out an AI strategy?
Provencher: That brings us back to the one version of the truth principle that I mentioned earlier. Right now, mining customers have the challenge of handling data from many different sources — control systems, PLCs, etc. — and these may have trouble speaking to each other. At that point, a company faces the choice of buying or building an AI platform. I see companies that reach this decision point but may not have the expertise to choose the right path. If they choose incorrectly, the data they need may not be usable or even available.
Establishing a consolidated data structure is very important, and that’s what PI System is very good at doing. Many companies already have an ERP platform, an operational data platform, along with a planning and scheduling platform. Once the data structure to accommodate all of these is in place, it becomes much easier to leverage technology to achieve the desired results.