Digital transformation in the mining industry is not simply about collecting and analysing data, it is about the application of artificial intelligence to provide feedback loops to the sensors, and machines operating on mine sites. SUE KEAY writes

Smiling woman with red-blonde long hair

Dr Sue Keay, CSIRO's Data61 Research Director for Cyber-Physical Systems, believes the keys to embracing a brave new digital future is people

Working in Cyber-physical systems gives me a box seat to witness the transformation possible by combining the digital with the physical, sometimes referred to as the fourth industrial revolution.

In this new era, it is not sufficient for machines to be automatised, they must operate with true autonomy, relying on feedback from sensor data to modify their actions.

Digital transformation in the mining industry is not simply about collecting and analysing data, it is about the application of artificial intelligence to provide feedback loops to the sensors, and machines operating on mine sites.

It is an important step in the ambition to achieve a zero entry, fully automated, integrated mine that can be applied to any orebody type and at any scale.

The most fun part of my job is getting to see many of the new developments in embedded artificial intelligence, that is, AI applied to physical devices.

Embedded AI is manifested in the fields of robotics, computer vision, sensing systems and cybernetics.

While there is no shortage of new technologies being developed in these fields that can be applied to the mining industry, the barrier to adoption remains developing in a business case and system-level approach to technology implementation that would see them applied on a large scale rather than in a piecemeal fashion.

This is difficult to achieve unless incorporated in new mine designs, and why it is important to consider the opportunities that new embedded AI technologies may deliver in the future and plan accordingly.

Below are some of the trends that I see in the future based on the work being conducted within my research program at CSIRO.

The robotisation of commodity products

While nothing is more exciting than seeing a new robot being designed and built from scratch, there are a lot of advantages to turning something that already works well into a robot.

The robot product cycle can last decades before you achieve a robust and reliable mass commodity product (think robotic vacuum cleaners).

What if you could just convert your existing vacuum cleaner into a robot?

There is a reason that self-driving cars are based on existing car designs – because they have been extensively safety tested before being allowed on our roads.

Using modular robotic components, in the future we will be able to convert commodity products into robots – think self-driving car kits.

Robust and ubiquitous sensing systems

In developing sensing systems at continental-scale for monitoring the environment, increasingly, our sensors are becoming smaller, cheaper, more robust, reliable and able to operate with no or minimal energy requirements in communication-denied environments.

We are developing algorithms that require low computation power allowing edge computing and teaming this with direct-to-orbit satellite communication from our devices to enable remote IoT (Internet of Things).

Applied to mine sites, integrated sensing systems will enhance safety and real time decision-making, allowing tracking of all mine assets (including people) in real-time, information which could be used, for example, to ensure that blast zones are clear.

Computer vision systems increasingly trained on synthetic data

Computer vision is fast becoming an indispensable tool for obtaining situational awareness, predicting future events and enabling perception for autonomous machines.

Despite rapid advances in machine learning, computer vision systems remain limited by access to large datasets of labelled images, that are hard to obtain for all but tech giants and unreliable at identifying rare events due to the scarcity of training data.

Many high-risk activities are rare but may have catastrophic impact, so it is important that computer vision systems are able to correctly identify any signs of such an event.

It is now possible to create data simulators to generate contextually relevant data with quality labels in order to train computer vision algorithms and which can in turn be used to train robots for different activities.

The accuracy of synthetic data can be tested against real data, for example using motion capture systems such as Australia’s largest MOCAP system at CSIRO’s Data61 facility at the Queensland Centre for Advanced Technologies that is open for use by industry. 

Synthetic data has the potential to disrupt the tech giants whose power lies in access to proprietary datasets (data moats) and will allow start-ups to scale into many new areas.

Data61's Robotics Innovation Centre is a purpose-built facility enabling world-leading robotics research across industries of strategic importance to Australia and the world, such as manufacturing, agriculture, mining, biosecurity and biodiversity

Collective Intelligence

Australia’s Team CSIRO Data61 recently competed in the DARPA Subterranean Challenge in the US. The SubT Challenge requires teams to deploy robots underground with no access to communications (save what the robots can carry) where they are given one hour to explore and find artefacts, replicating a disaster response scenario.

This brings together several strands of our research over many years.

One of our most exciting outcomes has been the successful application of multiple robots working together as a team with a common navigation platform to share and correct one another’s maps and understanding of the world in real-time under challenging circumstances.

The IoT era means we will be increasingly dependent on multiple sensors (some carried by robots) and we will rely on these sensors to validate each other’s data.

This helps to ensure the accuracy of the sensor data itself, where we are increasingly using blockchain to record verified data and monitor the integrity of supply chains to spot substitutions or product loss.

As well as applications to supply chain management, the same collective intelligence can be applied to many other areas on a mine site such as predictive maintenance.

Self-repair and self-fabrication

The advent of 3D printing has enabled us to rapidly prototype-test-redesign repeat robot components to optimise their performance.

Pairing this with new developments in materials science and the increasing ease of multi-material 3D printing opens up the possibility, long dreamed, of self-maintaining and even self-creating robots.

This is an important consideration for projects such as the “moon village”, a European Space Agency backed lunar construction project where the aim is for building to be completed before a human ever sets foot in the village.

There are similar potential applications in many parts of Australia that rely on remote operations, and in the future, we will need a range of skilled people and insights from AI to continue to explore new and creative ways of solving problems using intelligent machines.

The Future

The range of new technologies being developed opens up the possibility of us solving more and more problems that were once considered unsolvable.

But technology is just a tool.

The key to embracing a digital future is people.

While it’s important to understand what tools are now available that work better than tools of the past, we require people with the vision and drive who will seek to make whole-of-system changes and put the hard work in to demonstrate the business case that makes the adoption and integration of many of these technologies, a wise choice for the mining sector.

Without that vision and drive, mining will not see the same level of productivity improvements from the application of digital technologies that we are starting to see in other sectors.

***

Sue Keay is the Research Director for Cyber‑Physical Systems at CSIRO’s Data61.

With more than 20 years experience in the research sector, Sue has worked with physical scientists, engineers, social science researchers and economists and loves the challenge of managing research across distributed sites.

Sue set-up the world’s first robotic vision research centre and led the development of Australia’s first robotics roadmap, outlining how robotics and automation will impact on every sector of the Australian economy.

She has a PhD in Earth Sciences from the Australian National University, an MBA with UQ Business School, is a graduate of the Australian Institute of Company Directors and serves on the Board of the CRC for Optimising Resource Extraction and Women in Robotics International.

Resourceful magazine

Resourceful brings you the latest in innovation, technology and research from across the minerals value chain — from mineral exploration to mining and metal production. Get it by email or in print.

Contact us

 
Your contact details
0 / 100
0 / 1900
You shouldn't be able to see this field. Please try again and leave the field blank.

For security reasons attachments are not accepted.