The optimisation of digital twins for mine safety

It’s been ten years since the world’s first fully autonomous trucks appeared at Rio Tinto’s Yandicoogina and Nammuldi mines in the Pilbara. Since then, Australian mines have become global leaders in adopting automation and digitisation, with clear benefits for health and safety.

This technology has helped lift many workers out of the most hazardous parts of mine sites, yet fatalities and life-changing accidents still happen. In 2024, the number of deaths recorded was 39% higher than the 5-year industry average, according to Safe Work Australia data. Though mining still has a lower average fatality rate than the agriculture, transport, postal, warehousing and forestry and fishing industries, there is still room for improvement.

The mining industry is under pressure to improve its health and safety, and robotics and automation are emerging as a key solution. However, they come with new challenges, including the increased operational complexity these technologies can create.

“What we're seeing in the Australian mining industry is a turning point,” explains Cisco Sara, senior product marketing manager at software provider Accruent, which works with mining companies across Australia. “As operational complexity grows due to autonomous trucks, pressure for environmental change and falling resource quality, requiring companies to mine deeper to extract more from their sites, or to operate in more remote areas, there are increased risks which make operations harder.”

Miners, including BHP and Rio Tinto, are integrating digital twin technology across their operations to help manage some of this complexity. A 2024 report by Bentley Systems found that nearly 90% of surveyed mining organisations are either using, implementing or piloting digital twins, with health and safety cited as the biggest driver for this adoption.

Digital twins are dynamic virtual replicas of physical assets, processes and entire mine sites. A digital twin model is constantly updated from Internet of Things sensors on physical infrastructure, as well as from automated and digital systems. The technology enables remote monitoring, simulation of hazardous scenarios and predictive maintenance of safety-critical equipment, among other functions.

Emerging as an industry leader in this technology is BHP. The company is deploying digital twins across its operations, from pit to port, primarily to understand and manage complex, interconnected systems, rather than as a standalone safety tool, says Iván López, vice-president for value engineering at BHP.

“Modern mining operations are highly interconnected, with decisions in one part of the value chain affecting safety, productivity, cost, and environmental outcomes elsewhere. Digital twins provide a way to see and test those interactions explicitly, rather than relying on isolated assumptions,” he says.

“Safety, productivity, and sustainability benefits flow from that improved understanding. In many cases, safer outcomes are achieved because teams can avoid decisions that look optimal locally but create risk elsewhere in the system.”

A clear example of this, according to López, is the use of value chain digital twins at sites like BHP Mitsubishi Alliance, which produces metallurgical coal at its Copper South Australia province in Queensland. Another example is Escondida in Chile, where scenarios are tested end-to-end before changes are implemented in the field.

“By simulating different operating conditions, constraints and recovery actions, teams can identify safer operating envelopes, avoid high-risk short-term fixes and reduce the likelihood of reactive work under pressure, which is often where safety risk increases,” he adds.

The technology is supporting better planning and coordination, which directly reduces congestion, conflicting activities, and unplanned interventions, explains López.

“While these benefits are not always measured as a single ‘safety metric’, they materially reduce exposure by improving predictability and decision quality across the system,” he adds.

Meanwhile, at Copper South Australia, a digital twin is used to assess water, energy and material flows to support more stable operations of constrained systems, reducing the likelihood of “environmental exceedances or emergency responses,” says López.

In terms of the technology itself, BHP is using a combination of off-the-shelf platforms and bespoke models tailored to each asset’s operating reality and complexity.

However, most eye-catchingly, BHP is combining digital twins with generative AI to make the technology more user friendly. This has lowered the barrier between people and complex models, says López, though he notes it’s not yet the sole or standard interface for all interactions but currently implemented selectively where it adds “clarity and speed”.

“Traditionally, interacting with a digital twin required specialist skills. Generative AI allows users to ask natural-language questions, explore scenarios without needing to understand the underlying model structure and translate technical outputs into clear, decision-focused insights,” he says. “This is particularly powerful for broadening access beyond technical specialists and supporting leaders and frontline teams to engage with the insights.”

Rio Tinto has also adopted digital twins at its Gudai-Darri mine, its first greenfield site in the Pilbara. The company has said it’s using the technology to monitor and respond to data collected from the site’s processing plant. The same model is also used for an interactive 3D environment for virtual reality training, enabling team members to visually navigate the asset and plan their work using a to-scale 3D model.

Elsewhere, American software, hardware and services technology company Trimble is working with Australian mining firms using digital twin technology to help them maintain geotechnical integrity as they dig deeper to access more resources.

The company embeds sensors in mine slopes and the general mining environment to assess how the structure changes over time. The sensors, connected to software, automatically collect information in real time – rather than teams onsite taking measurements – and provide  geotechnical teams with analysis so they can make decisions around potential risks throughout the mines’ lifetime, and act accordingly.

“There's always a balance between how steep and deep the miner can go versus when it becomes too unsafe,” says director for monitoring, mining, and tunnelling at Trimble, Riley Smith.

She explains that Trimble’s technology creates a “real time digital component to the surface and the subsurface which provides obvious efficiency and health and safety benefits,” but also enables teams to be more focused on making decisions rather than collecting data.

The same principles can be applied to monitor open pit mine slopes to prevent failures, as well as to tailings dams infrastructure. The sensors will monitor the characteristics of the structure, soil, movement, temperature as well as wind and water pressure and volumes to understand how they’re behaving. High profile major incidents, such as the Brumadinho dam disaster in Brazil, which killed 270 people and saw regulations tighten globally, have largely driven the adoption of the technology for these purposes.

AI is also supporting decision-making on some applications, says Smith. For example, Trimble Mind Insights, a cloud application, leverages AI trained on a variety of geospatial data sets to assess the geometric characteristics of a slope, looking for potential areas that might be more likely to fail.

“It doesn’t predict a failure but classifies areas in the slope that might be more likely to fail because of their characteristics,” says Smith, who doesn’t believe AI should make these decisions alone. “This helps teams prioritise where they need to focus their efforts, such as increasing the range of information collected on those areas.”

Sara, whose company provides Electronic Document Management Systems which can support the adoption of digital twins on mine sites, believes the technology is compelling because as digitisation increases it enables operators to make use of data that would be unrealistic for humans to do alone.

“If you can get to the point where you’ve got all this data in the system and you can just ask it to give you the information needed to make decisions to keep your team safe and the site operating effectively – that’s powerful,” he says. “Missing data can cost lives; there was an incident a couple of years ago where two guys drove into a ditch because the information they had said the road was safe, but it wasn't.”

Thanks to readily available and more affordable satellite communications, adoption is also getting easier. However, Sara says he likes to remind those excited about the technology that they need “to make sure they’ve got their house in order first”.

“Make sure you do document management and other prep first – then you can start plugging sensor data in and start understanding and running simulations in a digital environment,” he says.

López echoes this point, saying adoption at BHP has been incremental.

“The biggest learning is that digital twins are not just a technology deployment; they require high-quality data foundations, strong operational engagement, clear decision use cases and integration into planning and governance routines,” he says.

“One of the key lessons has been that value only materialises when the digital twin becomes part of ‘how work gets done’, not an analytical exercise running in parallel.”

He adds that BHP plan to scale digital twin capability “where it delivers clear value”, but every site will not deploy the same solution in the same way or at the same time.

“Differences in asset maturity, system complexity, data readiness, and risk profile require the company to take a ‘fit-for-purpose approach’ as some sites benefit from full end-to-end value chain twins, while others focus on specific subsystems,” he concludes.