As technology continues to evolve, better enabling the next generation of miners, artificial intelligence (AI) has seated itself top of mind for leaders throughout the mining, metals and materials (MMM) sector. In fact, nearly 80% of leaders throughout all enterprise trades anticipate that AI will significantly reshape their businesses within the next three years.[1] Deloitte reports that early uses of generative AI (GenAI) are aimed at enhancing efficiency and productivity, cutting costs and boosting product quality.
While the AI revolution is undoubtedly promising, it also poses questions about ethics, security and trust. As miners gear up to embrace AI, establishing clear guidelines and safeguards is essential.
Exploring AI
By entering an internet search or prompting a chatbot, users learn that AI is “the simulation of human intelligence in machines, enabling equipment to perform tasks that typically require human cognition, such as learning, problem-solving and decision-making.”[2]
Not long ago, machine learning (ML) was also a popular buzzword in the industry. Recently, however, companies that previously highlighted their ML, internet of things and digital twin technologies have begun using the term AI as a broad moniker for human-like machine intelligence. While this shift in terminology is technically accurate, it can also be somewhat misleading.
ML’s focus is on training machines to learn from data and improve in performance over time without being explicitly programmed. Engineers are familiar with the long cycles of training these models, requiring data preparation to feed the models and numerous iterations for satisfactory deployment.
When applied correctly, ML has proven highly beneficial for mining organizations, especially in reliability and equipment failure prevention use cases. It has also shown value in specific applications, such as metal recovery.
GenAI is the next iteration of operational ML, which enables self-machine learning based on patterns in existing data samples. This is the key factor transforming the abstract concept of augmented engineering into reality. By suggesting solutions, answering questions and explaining problem-solving methods, GenAI is bridging the divide between human expertise and advanced data analytics.
With nearly 30% of the North American industrial workforce set to retire in the next five to ten years, process manufacturers will need to leverage AI to fill in the human expertise shortage, using insights drawn from operational data. AI enhances humans’ ability to make sense of constantly evolving global markets and to address unseen future challenges.
Impact on the mining and metals industry
Global Market Insights estimated that GenAI will grow at a compound annual growth rate of 30.3% over the next 8 years throughout industry.[3] GenAI’s primary potential lies in reducing operational costs by 20% while increasing workforce productivity by more than 70%. The same study projects that over 40% of these cost reductions will be attributed to raw material and energy consumption optimization, 40% to material movement and logistical enhancements and 20% to improvements in finished product quality.
These statistics highlight key areas where industrial operations can gain substantial benefits in the coming years. Additionally, AI is set to play a pivotal role in transforming the MMM industry as we know it today.
According to the Deloitte survey, most industrial organizations are more focused on tactical benefits, including those related to productivity (56%) and variable cost reductions (35%).[4] Only 29% of companies are applying AI for strategic benefits, such as research and development.
Over the past decade, the MMM sector has undergone an advanced analytics revolution, and the evolution is continuing, with new applications and approaches emerging regularly. Some operational challenges are inherently more complex than others, leading to varying levels of data analysis complexity required to address them. Combining AI capabilities with these developing analytics trends is further enhancing operations throughout the industry, as exemplified by the following two case studies using different AI subsets.
Process optimization at a grinding mill
At one of Swedish miner Boliden AB’s grinding mills, the company identified and monitored a key performance indicator (KPI) related to energy consumption and torque in the grinding section to derive a 15% increase in long, sustained periods of high production.
Optimizing operations at a semi-autogenous grinding mill is critical because ore grinding is highly energy-intensive, consuming nearly half of the electricity used for all mineral processing at the mill. Additionally, ore grinding is responsible for producing adequately-sized particles, directly affecting mineral recovery efficiency in downstream flotation systems.
The ore grinding process is a major bottleneck in nearly every mineral processing plant, so enhancements applied to this step significantly impact final production outcomes. For example, increasing ore grinding production by 1% typically increases overall output by a nearly equivalent 1%.
Striving to identify ways of increasing sustained periods of high production, Boliden leveraged Seeq — an advanced analytics, AI and monitoring platform — enabling its engineers to obtain a more efficient production output to energy ratio.
The initial step required developing and running a simulated model of the grinding process, with which to compare current operations and potential changes in the process control system, to identify scenarios that could boost production. The platform tracked historical data on both absolute and relative power distribution to calculate available power, which served as the KPI for measuring the effort’s success.
Using this KPI and the assistance of the advanced analytics, AI and monitoring platform, the engineering team quantified the benefits of potential changes to the process control system (Figure 1).
This also empowered the organization to improve, refine and add nuance to the analysis, including setting up conditions to determine and correct load limitations for a more accurate evaluation (Figure 2).
The data revealed a 15% increase in long, sustained periods of high production.
Immortalizing knowledge
Many mill product suppliers are also investigating ways to integrate generative AI into their enterprise analytics solutions to benefit both existing and new users. Early adopters have reported that the use of large language models (LLMs), which underpin GenAI, accelerates adoption by at least eight times.
LLMs are expected to enable embedding industrial experts' knowledge into operational datastores, helping retain knowledge during staffing changes or deficiencies, such as through the loss of expert personnel to retirement. While human experts have traditionally handled the analytics required to detect data deviations and conducted root-cause analyses for abnormal operating conditions, these functions will be available using models.
These LLMs are trained in staff persona, question-response competence and tone, enabling GenAI to answer questions from personnel such as, “What is wrong with my continuous casting?” Advanced analytics platforms will then be able to pass along the conditions and properties from time periods of interest to the GenAI application to respond to a human-issued prompt.
For example, the Seeq AI Assistant, which is embedded in the Seeq platform, has access to models, training documentation and graphs that are woven into prompt responses. When asked about problem sources, the assistant will be able to quickly display signals that showcase inefficiencies or prove violations, and extrapolate the repercussions into daily production targets. The platform will also provide visibility to issues’ leading indicators, such as lower-than-normal feed flow or bad control valve position setpoints.
Other mill product manufacturers are adopting a more cautious approach by focusing on a few key signals and posing targeted questions. For instance, this method is expected to be used to optimize chemical usage during washing phases. In these cases, the advanced analytics platform’s AI assistant will respond with known relations between signals in the context of different operational modes, such as idle or running. After asking an initial question, SMEs will create follow-up queries to further refine the process, enhance asset efficiency, reduce energy consumption, minimize chemical use and prolong asset lifespan.
By following these methods, an LLM can be trained to optimize production based on mode of operation, chemical mix, product quality and more. Additionally, these enhancements become available for others using the same model every time it is tuned.
Implementation success
Focusing on tactical benefits is a powerful way to justify adopting and scaling AI applications, and mining organizations frequently find success when adopting a "think big but start small" strategy. This approach helps secure support for new technologies that drive innovation and change, while also establishing the necessary skills and organizational structure to back future strategic initiatives.
Implementing AI into operations requires assessing organizational readiness, particularly when proprietary or other intellectual property is at play. However, leading advanced analytics, AI and monitoring platforms segregate AI model data by organization so it cannot be used to train public or general models.
These platforms help miners achieve new levels of efficiency, accuracy and innovation. By incorporating the software tools into everyday workflows, teams can stay ahead in competitive markets.
