How AI Blast Optimisation Turns a Drill Decision Into Mill Profits

A single software platform, running at one Western Australian iron ore operation, is estimated to unlock A$450 million in additional annual value. That figure comes from a 5.5% improvement in how much ore the processing plant can push through, and it was not produced by a demo or a sales deck.

It came from a live deployment, built on a digital twin trained on data from hundreds of millions of tonnes of ore. The approach behind it reframes where the hard work of breaking rock actually happens in a mine, and the industry has started to formally recognise it: PETRA Data Science’s MAXTA Drill&Blast is a 2026 finalist for the Drill and Blast Award run by Mining Magazine and GeoDrilling International.

This piece explains how a blasting decision made at the top of a mine turns into a dollar outcome at the bottom of a processing plant. It also explains what AI-driven blast optimisation means for how Australia’s mining sector is now spending its AI budget, and what has to be in place before a platform like this delivers on the numbers. You will finish knowing the mechanism, not just the marketing.

What winning the Drill and Blast Award actually signals for the industry

The Drill and Blast Award is not a small-vendor prize. Run by Mining Magazine and GeoDrilling International, it evaluates entrants on three things:

• Efficiency in mineral extraction
• Safety
• Environmental responsibility

That framing matters because it puts a data science company in direct competition with the industry’s largest operators.

For 2026, MAXTA Drill&Blast sits on the shortlist alongside Anglo American and OptiClino. The 2025 winner in the same category was Hexagon, recognised for its Drill Assist AI-powered drilling automation system. So the field MAXTA is being measured against spans global mining majors and established automation specialists, not niche software rivals.

That competitive context is the point. When a Queensland-headquartered data science firm is evaluated on equal footing with a company the size of Anglo American, it signals that AI-native businesses are now competing on the same technical ground as the industry’s incumbents. For operators and investors trying to read the sector’s innovation landscape, that is a genuine shift in who gets taken seriously.

PETRA Data Science was founded around 2015 and is headquartered in Toowong, Queensland. Its MAXTA libraries have been applied to copper-gold porphyry, iron ore, and epithermal gold operations worldwide.

Its corporate position changed this year. Mining software company Maptek completed a full acquisition of PETRA on 6 March 2026, lifting its stake to 100% from an initial 25% taken in 2019. PETRA now runs as a standalone, agile business inside the Maptek group.

Maptek’s assessment of PETRA Maptek described PETRA as “a pioneer in orebody learning and data fusion technology.”

Award recognition in a contested field gives you something vendor marketing cannot: an independent read on technical credibility. For the Australian market, it confirms that mine-to-mill AI has moved from experimental to formally evaluated.

How a digital twin connects a blast to a processing plant’s bottom line

Start with the rock. A blast does not just move material; it sets the size distribution of everything that will enter the processing plant afterwards. Get the fragmentation wrong and the plant inherits the problem.

That physical fact sits behind the whole economic argument. The mine-to-mill concept is built on one comparison: the energy used to break rock in a blast is far cheaper than the energy used to grind it in a mill. Shift more of the breakage upstream into the blast, and the mill receives feed that is easier and cheaper to process.

The mine-to-mill concept is one expression of whole-of-value-chain thinking, an operational philosophy that treats each stage of extraction, from drilling to dispatch, as a single interconnected system rather than a set of discrete cost centres to be optimised in isolation.

Finer fragmentation produces more fines that move quickly through grinding, effectively unloading the mill and improving grinding efficiency. Industry guidance holds that finer fragmentation can raise mill throughput by 20-30% or cut specific energy consumption by up to 30%.

Here is the part that changes the decision. Achieving finer fragmentation costs more in blasting, but the overall saving in mill operating cost can be 4-10 times the extra blasting spend.

That multiplier is the heart of it. Spending more on explosives is often the cheapest decision available to a mine, and the whole challenge is making that counterintuitive logic visible enough for an engineer and a CFO to act on together.

What the MAXTA platform actually does with that data

This is where the digital twin comes in. MAXTA Drill&Blast ingests geological, drill-and-blast, and processing datasets, then fuses them into a single analytical framework using proprietary ore-tracking. In the WA iron ore case, that twin was built on data from hundreds of millions of tonnes of ore.

The distinguishing capability is that data fusion. Standalone blast modelling tools stop at predicting fragmentation. MAXTA carries the chain forward, simulating a drill-and-blast configuration and immediately modelling its projected impact on plant outcomes: crusher blockages, oversize rock incidents, dig rates, crusher throughput, and grinding energy.

For a blast engineer, that means virtual experimentation. You can test configurations for each ore domain before committing a single blast design, then choose the setup that serves the plant, not just the blast face.

PETRA has retained a technology-agnostic approach inside the Maptek group, so the platform integrates with existing mine software environments rather than forcing a rip-and-replace. That matters for anyone weighing adoption cost.

What the numbers look like when this runs at scale

The clearest way to judge this is to look at what it has produced in the field, in enough detail to assess for yourself.

Start with the Western Australian iron ore operation, because the dollar figure is the largest. The deployment addressed existing problems with dig rates, crusher throughput, and processing bottlenecks, and the digital twin, trained on hundreds of millions of tonnes of ore, delivered a 5.5% annual improvement in processing plant throughput.

The headline outcome A 5.5% throughput uplift at the WA iron ore operation was estimated to be worth approximately A$450 million in additional annual value.

The Queensland coal case shows the same platform working at a different scale. There, simulations to find optimal blast configurations produced an estimated 10-20% improvement in excavator productivity and annual cost savings of between A$1.4 million and A$5.7 million. Broader AusIMM case studies using MAXTA-guided optimisation have reported 25-30% reductions in explosive costs through a switch to ANFO.

The credibility test is whether these outcomes line up with independently documented results. They do. A machine-learning blast optimisation solution deployed by Orica at Roy Hill reduced drill-and-blast costs by A$7.5-7.6 million over under two years and lifted average instantaneous dig rates from 7,380 tonnes per hour to 9,006 tonnes per hour.

The consistency between MAXTA’s reported figures and the independently documented Orica-Roy Hill results is the signal worth holding onto. These numbers are not vendor outliers; they reflect what structured, AI-driven blast optimisation reliably produces at scale in Australian iron ore.

MAXTA is now live at 18 sites globally, with five Tier 1 or Tier 2 operations added in the two years prior to September 2026. For anyone modelling a return, the WA and Queensland cases together give you a realistic range across commodity types and mine scales, rather than a single best-case headline.

Past performance does not guarantee future results. Financial projections are subject to market conditions and various risk factors.

Where Australian mining’s AI investment is heading, and what could slow it down

The market backdrop explains why this category is drawing so much capital. Australia commands roughly 74% of global AI-in-mining investment, ahead of China at 12% and the United States at 9%.

The local market was valued at US$316.1 million in 2025 and is projected to reach US$1,263.2 million by 2030, a compound annual growth rate of 21.9%. Globally, AI-in-mining spending is expected to grow from US$2.7 billion in 2024 to US$13.1 billion by 2029, with roughly 70% of mining companies already integrating AI-driven technologies.

Then comes the honest part. A mid-2025 Mining Magazine survey found that only 39% of respondents had fully implemented operational AI solutions. Investment intent is running well ahead of deployment reality.

The gap has specific, practical causes. The main implementation risks fall into three categories:

• Data quality and readiness: operational data is often not AI-ready
• The model-reality gap: models trained on historical data can underperform when ore characteristics or conditions shift
• Workforce trust: opaque, black-box recommendations meet resistance from operators who trust experience-based judgment

The data problem is the one to fixate on. Modern mines generate between 1.5 and 4.2 TB of operational data daily, yet the readiness figure is stark.

The number behind the adoption gap Less than 8% of the operational data generated at modern mines is considered AI-ready.

That 8% figure is the most important number here for any operator or investor. It means most mines considering a platform like MAXTA face a data infrastructure problem first, and the timeline and cost of solving it belong in the adoption decision from day one.

AI readiness in mining is not a software procurement problem; it is a data infrastructure problem, and operations that treat it as the former consistently discover the latter only after a deployment stalls.

So is the sector early or late? Neither. It is mid-transition, with real headroom remaining but genuine prerequisites that decide whether a deployment delivers its modelled value or stalls short of it.

What separates a transformative AI blast platform from a modelling tool

Given all of that, the practical question becomes how you tell a full mine-to-mill platform apart from ordinary blast modelling software. Three architectural characteristics separate them.

The first is end-to-end data integration across geological, blasting, and processing domains, rather than a model that stops at the blast face. The second is a digital twin calibrated on the site’s own operational tonnes, not generic industry parameters. The third is a feedback loop that updates the model as ore and conditions change.

The Maptek acquisition matters here because it extends that integration surface. Access to Vulcan and Evolution platform workflows creates a more complete data pathway, from geological modelling through to plant performance, with MAXTA sitting in the middle.

Scale supports the point. MAXTA has been deployed across 18 sites globally since 2016, spanning mine types and operational scales, and the five Tier 1 or Tier 2 operations added in the two years prior to September 2026 point to accelerating enterprise adoption. PETRA’s technology-agnostic approach has been retained through the acquisition.

The distinction to hold onto is this: a fragmentation predictor tells you what size rocks you will get, while a mine-to-mill platform tells you what that means for your processing cost per tonne. That gap is where most of the financial value either materialises or disappears.

Three questions to ask before committing to an AI blast platform

1. Does the platform model outcomes at the processing plant, or only at the blast face? If it stops at predicting fragmentation, it cannot show you the cost-per-tonne effect where most of the value sits.
2. Is the digital twin calibrated on your site’s operational data, or on generic industry parameters? A twin transplanted from another mine without site-specific tuning is where the model-reality gap opens up.
3. Can the vendor show a documented throughput or cost outcome from a comparable operation? A verified result from a similar commodity and scale is worth more than any modelled projection.

The case for AI-driven blast optimisation goes beyond efficiency

Pull the threads together and one picture emerges. The award recognition, the mine-to-mill mechanism, the WA and Queensland outcomes, and the market data all point the same way: AI-driven blast optimisation is moving from competitive advantage to operational standard in large-scale Australian mining.

That transition will be uneven. The data readiness gap and the model-reality risk mean Tier 1 and Tier 2 operations will lead, while smaller operators face a longer runway before the numbers work for them. The gap between the 70% of companies integrating AI and the 39% with fully implemented solutions is exactly where the next tier of returns will be captured.

Australia’s 74% share of global AI-in-mining capital, and a market growing at a 21.9% CAGR to 2030, suggests the country is positioned to lead that shift rather than follow it. Where that leaves you depends on your seat:

The intersection of AI integration and critical minerals strategy in Australia is reshaping how Tier 1 operators prioritise capital allocation, with productivity gains from AI deployment increasingly factored into project economics for new mineral developments rather than treated as post-commissioning optimisation.

• For operators: the question is no longer whether to adopt, but which data infrastructure to build first.
• For investors: the growth trajectory and Australia’s capital concentration make this a segment worth watching closely.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions.