Fatigue kills miners. That’s not dramatic language. It’s statistical fact. The Queensland Mines Inspectorate’s 2025 annual report identified fatigue as a contributing factor in 34% of serious incidents in underground mines. The numbers are similar in Western Australia and New South Wales.

The industry has been working on this problem for decades, from basic hours-of-work restrictions to wrist-worn actigraphy devices. But the latest generation of AI-powered fatigue monitoring systems represents a genuine step forward in detection capability. The question is whether the technology is ready for widespread underground deployment.

How the Technology Works

Modern AI fatigue monitoring systems use one or more of these approaches:

Camera-based facial analysis. Cameras mounted in the operator cabin track eye movements, blink frequency, blink duration, and head position. AI algorithms compare these metrics against the operator’s baseline to detect fatigue indicators. Companies like Seeing Machines and Caterpillar’s DSS (Driver Safety System) are the most deployed systems in Australian mines.

Steering and control input analysis. AI monitors the operator’s control inputs, looking for patterns associated with fatigue: micro-corrections, delayed reactions, inconsistent operating patterns. This approach doesn’t require cameras and works entirely from existing vehicle telemetry.

Wearable biometric monitoring. Devices worn on the wrist or embedded in PPE track heart rate variability, skin conductance, and movement patterns. AI processes this data to estimate alertness levels. SmartCap’s Life system, which embeds sensors in the operator’s hard hat headband, is the most well-known example in Australian mining.

Multimodal fusion. The most advanced systems combine multiple data sources. For example, Seeing Machines’ camera data combined with vehicle telemetry and wearable biometrics creates a more reliable fatigue picture than any single data source alone.

What’s Working

Based on conversations with safety managers at five underground operations across Queensland and Western Australia, here’s what the technology is doing well:

Early warning catches real events. Multiple sites reported instances where the AI system alerted an operator to fatigue before a near-miss occurred. At one gold mine in WA, the fatigue monitoring system averaged 3.2 genuine fatigue alerts per week across their underground fleet. Before the system was installed, they were relying entirely on self-reporting, which produced maybe one fatigue report per month.

Data patterns inform scheduling. The aggregate fatigue data is helping mines optimise shift patterns. One operation found that fatigue alerts spiked dramatically in the last two hours of 12-hour shifts (not surprising), but also that alerts were 40% higher on the third consecutive night shift compared to the first. That data influenced their roster design.

Operator acceptance is improving. Early fatigue monitoring systems faced significant pushback from operators who saw them as surveillance tools. The newer systems have better privacy features (data is processed locally and only aggregate trends go to management) and operators are seeing the safety benefit firsthand. “When the system tapped me on the shoulder at 4am and I realised I’d been microsleeping, I became a believer,” one loader operator at a NSW coal mine told me.

What’s Not Working Yet

False positive rates in underground environments. This is the biggest technical challenge. Underground lighting conditions, dust, and vibration create difficult conditions for camera-based systems. False positive rates that sit around 5-8% on the surface can jump to 15-20% underground. Operators quickly learn to ignore a system that cries wolf.

Compatibility with PPE. Safety glasses, dust masks, and face shields all interfere with camera-based facial analysis. SmartCap’s in-headband approach avoids this, but it only measures one biometric channel. The ideal multimodal approach requires multiple sensors that all need to coexist with existing PPE.

Connectivity challenges. Underground mines have notoriously patchy network coverage. Systems that need real-time cloud processing struggle. The better systems process locally on the vehicle, but this limits the sophistication of the AI models that can run.

Individual calibration time. AI fatigue systems need to learn each operator’s baseline. That calibration period, typically one to two weeks of normal operation, means the system is less accurate for new operators or those who change roles frequently.

The Regulatory Landscape

Australian mining regulators are watching this technology closely but haven’t mandated it yet. Queensland’s mine safety legislation requires operators to have a fatigue management plan, but doesn’t specify particular technologies. Western Australia’s approach is similar.

Industry bodies like the Minerals Council of Australia are encouraging voluntary adoption. The International Council on Mining and Metals (ICMM) included AI fatigue monitoring in its 2025 Innovation for Safety guidelines.

The practical reality is that insurers are driving adoption faster than regulators. Several mining insurance underwriters now offer premium reductions for operations that deploy approved fatigue monitoring systems. That financial incentive is proving more effective than regulatory requirements.

What Comes Next

The technology is improving fast. The two developments to watch:

On-device AI processing. New edge computing hardware means more sophisticated AI models can run on the vehicle without cloud connectivity. This directly addresses the underground connectivity challenge and will improve both accuracy and false positive rates.

Predictive fatigue modelling. Current systems detect fatigue as it’s happening. The next generation aims to predict fatigue hours before it occurs, based on the operator’s sleep data, shift history, and work intensity patterns. Several research programs at Australian universities, including the University of Queensland’s Sustainable Minerals Institute, are working on this.

Should You Deploy It Now?

If you’re running an underground operation with mobile equipment operators, yes. The technology isn’t perfect, but it’s significantly better than relying on self-reporting and hours-of-work restrictions alone. Choose a system rated for underground conditions, accept that you’ll need to manage false positive fatigue during the calibration period, and commit to using the aggregate data to improve your shift design.

The cost is meaningful but not prohibitive. Expect $3,000-5,000 per vehicle for hardware and installation, plus $50-100 per vehicle per month for software licensing and support. For an operation with 20 underground vehicles, that’s a total annual cost of around $75,000-$85,000. Compared to the cost of a single serious fatigue-related incident, both human and financial, it’s an investment that makes clear sense.

Fatigue monitoring won’t eliminate operator fatigue. Nothing will, as long as people work underground in shifts. But AI is making it possible to catch fatigue earlier, respond faster, and design better systems around the human limitations that have always been part of mining.

MinerMundo covers technology and innovation in the mining sector.