MMG Dugald River AI boosts zinc recovery & profits

MMG Dugald River has marked one year of operating an artificial intelligence control system at its zinc processing plant, reporting gains in zinc recovery and profitability.

Developed with engineering and professional services firm Hatch, the technology uses a digital twin of the mine's flotation process to adjust plant settings in real time. MMG said the investment paid for itself in less than three months after commissioning, making the project an early commercial test of AI-led process control in mining.

Dugald River, in north-west Queensland, has used the system to fine-tune an established processing operation rather than replace human oversight. The software uses Hatch's AI-Assisted Process Control, combining process models, machine learning and advanced process control to respond to changes in incoming ore feed.

According to MMG and Hatch, the system operates as a closed-loop control tool for the processing plant. It continuously adjusts operations to improve zinc recovery from each tonne of ore, while operators remain involved in plant decision-making.

Tim Akroyd, General Manager at MMG Dugald River, said the site adopted the technology after making gains through conventional process improvements.

"We've been running a successful zinc processing operation with clearly defined goals to deliver continuous improvement, and our team has delivered strong outcomes over many years," Akroyd said.

"As an operation, we are focused on identifying and implementing the right technology that can continue to drive safety and performance improvements. This is effectively like an always-on assistant that continuously adjusts how the plant runs to get the most zinc out of every tonne of ore processed."

Operational shift

The project centres on the flotation circuit, a key stage in zinc processing where minerals are separated from waste material. By modelling the process digitally and updating that model with live operating data, the system can make ongoing adjustments that would otherwise rely on manual intervention and operator judgment.

Hatch said the Dugald River installation is the first of its kind and offers evidence that digital twin technology can be used in live minerals processing at commercial scale. It described the project as a step towards broader autonomous control across concentrators, hydrometallurgical plants and smelters.

Alec Malcolm, Senior Metallurgist at MMG Dugald River, led the implementation at the site.

"The results have been positive, exceeding our previous performance benchmarks. I'm also encouraged by the flexibility of the system and how it can be tailored to different process goals," Malcolm said.

That emphasis on flexibility reflects a practical question around industrial AI projects: whether a control system can handle changing ore characteristics, plant constraints and shifting production objectives without becoming difficult for site teams to trust or manage.

Hatch said its model differs from black-box AI systems because it is designed to show operational relationships, report its own errors, encode operator and metallurgical knowledge, and validate decisions against process heuristics. The aim is to make the software easier for on-site personnel to understand and accept.

Industry test case

AI adoption in mining has often moved more slowly in processing plants than in areas such as maintenance, haulage or exploration, partly because control changes can affect recovery, stability and product quality within minutes. Dugald River's experience will therefore be watched as a test of whether more mines are prepared to hand a greater share of routine plant control to software.

Warwick Smith, Global Practise Lead for Digital Twins at Hatch, said MMG's role in deploying the technology had wider significance for the sector.

"MMG represents true forward-thinking for industry, and their successful adoption is giving confidence to others who want to explore how technology step-changes can greatly enhance their operations," Smith said.

"MMG has the right team on the ground, who trusted the process and understood the company's long-term view. We invested considerable time to explain how the technology would support their decision-making in real time and partnered closely with them to help them take the leap.

"By applying an incremental and metallurgically focused approach to digital twin technology in a complex process, from initial feasibility through to staged implementation, the site team was able to closely guide the development process and de-risk the overall implementation.

"Through steady confidence-building with site metallurgical and operational teams over time, Dugald River has achieved a level of autonomous, AI-driven plant control rarely seen in minerals processing."

The companies said the project drew on decades of metallurgical knowledge as well as machine learning techniques. That blend points to a broader pattern in industrial AI, where commercial value often depends less on generic algorithms than on how closely software reflects the specific behaviour of a plant and the operating experience of the people who run it.

For MMG, the immediate test is whether the system continues to deliver more consistent recovery and margin improvements over time. For Hatch, the site provides a working reference for using digital twins not just to advise operators, but to make and execute control decisions in production environments.

The one-year operating milestone comes with an economic claim likely to draw the most attention across the sector: analysis at Dugald River showed the investment paid for itself in under three months.