Signature Sponsor
Harnessing AI to Make Mining Safer

 


November 18, 2026 - Artificial intelligence is giving mine operators new ways to recognize hazards, monitor equipment and identify signs of worker fatigue. Its value depends on turning those observations into timely action by people who understand the mine and its risks.

The opportunities extend from underground coal workings to surface haul roads and preparation facilities. Some tools are commercially available, while others remain research projects or limited demonstrations. The distinction matters when evaluating what technology can deliver today.

A McKinsey analysis published September 23, and highlighted by International Mining on November 18, examines how AI could strengthen mining safety. It describes an unnamed company using a model across coal operations to bring together video, sensor readings and written reports, helping supervisors identify unsafe behavior and unusual conditions.

McKinsey also discusses using AI to review hazard reports and near-miss accounts for recurring problems. Generative systems could help prepare incident summaries, while digital simulations could support training. The analysis emphasizes human oversight and the organizational work needed to move beyond pilots. It does not identify the coal company or provide independently verified accident-reduction figures for that example. 

Recognizing Dangerous Situations

For coal miners, the potential applications address familiar hazards: confined work areas, poor visibility and people working close to moving equipment. NIOSH identifies these conditions as important contributors to machinery-related risk.

The agency’s proximity-detection research illustrates how technology can intervene. An intelligent system can restrict particular machine movements that would endanger a nearby worker. That is different from merely recording an incident or sounding an alarm after a dangerous movement has begun.

Proximity detection, remote control and AI are not interchangeable terms. A system may use programmed rules rather than a learning model. NIOSH also warns that software-controlled equipment can introduce failure modes different from those of conventional electromechanical machinery. Adding intelligence therefore makes functional-safety evaluation more important, not less. 

Coal-Mining Applications Already Being Developed

Huawei describes a named coal-industry example in its case study of Shandong Energy Group. According to the supplier, the company established an AI training center to develop applications involving extraction, underground development, transportation, ventilation and coal preparation.

The case study identifies Xinglongzhuang Coal Mine as a reference operation and describes plans to extend its model to additional mines. These are supplier-reported applications and expansion plans. They should not be read as independent confirmation that every planned site has adopted the technology or achieved a particular reduction in injuries.

This named example is separate from the unidentified coal operator discussed by McKinsey. The available evidence does not establish that they are the same company. 

Watching for Fatigue

Fatigue and distraction provide another practical use for AI. Caterpillar markets Driver Safety System technology that detects signs of inattention and alerts the operator. Its mining portfolio includes fatigue and distraction monitoring.

Caterpillar describes a combination of in-cab detection, immediate warnings and human review. Its DSS product documentation says AI identifies physiological signs of fatigue or distraction, with sound and seat vibration used to regain the operator’s attention. Trained personnel then review transmitted events and follow the customer’s intervention procedures.

The technology is intended to support a broader fatigue-management program. A warning does not resolve the underlying causes of tiredness, and equipment suitability must be checked for the particular machine and application. Commercial availability alone also does not establish a universal accident-reduction rate. 

Making AI Work Underground

The mine environment presents difficulties that an office-based AI system never encounters. Darkness, dust, blocked camera views, changing tunnel layouts and interrupted communications can all affect the information reaching a model.

A 2026 research paper, “Future Mining: Learning for Safety and Security,” by Missouri University of Science and Technology researchers examines these problems. It proposes an architecture combining different types of sensors, equipment-health monitoring, miner-location tools and protections against interference with AI systems.

The researchers explain that unreliable communications can leave systems working with outdated information. Poor-quality images and sensor readings can weaken hazard detection, while compromised data can distort a model’s decisions.

Their work is a research framework, not evidence of a fully deployed commercial safety system. It nevertheless identifies a central engineering challenge: a useful prediction must remain dependable under the conditions in which miners actually work. 

Building the Communications Foundation

Virginia Tech is working on part of that foundation. In an October 15 announcement, the university described its role in an Innovative Wireless Technologies-led project to develop and test advanced mining technologies.

Researchers plan to study underground wireless performance and develop 5G and 5G-Advanced communications. The Virginia Tech Quarry and Center for Autonomous Mining Systems will support testing of remote-controlled and autonomous equipment, sensors and training.

The university also plans systems to help machinery recognize people, equipment and terrain hazards, along with digital models that could identify maintenance needs. This is development and testing work; the announcement does not establish a finished AI safety installation at a particular coal mine. 

Keeping People Responsible

AI can create risks of its own. NIOSH’s workplace guidance highlights the difficulty of understanding how complex models reach decisions and the need to evaluate their safety before and after deployment.

The agency discusses collaboration between developers and safety professionals, workforce training, independent assessment and clear accountability. It also emphasizes workers’ privacy and their ability to understand how personal information is being used.

For a mine considering camera-based monitoring, those issues have practical consequences. Workers need to know what is collected, who reviews it and what happens when a system flags an event. Supervisors need enough training to question an unreliable recommendation rather than accept it automatically. 

Generative AI requires particular care. The National Institute of Standards and Technology warns that these systems can confidently produce incorrect information. An automatically drafted incident report or safety explanation therefore needs checking against original records and qualified personnel’s observations. 

For coal operators, the practical test is whether a tool helps prevent a defined hazard, works reliably at the site and supports a clear response. AI can extend what a safety team sees and how quickly it recognizes a problem. Its contribution becomes meaningful when that information helps miners and supervisors act before someone is hurt.