BHP Group unlocks 1 mil mt/y of extra iron ore in Western Australia with AI

Artificial intelligence helped BHP Group Ltd. unlock an extra 1 million metric tons of iron ore in 2025, Digital Officer Mikko Tepponen said at the Global Resources Innovation Expo in Perth on May 6.

BHP, one of the major producers that makes Australia the world's largest iron ore exporter, had lost about 1,000 hours of crusher downtime, along with additional disruptions at its West Australian ports, over three years before an AI-driven solution implemented in 2025 fixed the problem, Tepponen told the expo.

"One of the persistent challenges in our iron ore operations here in Western Australia is the presence of oversized rocks and foreign objects entering the processing system," Tepponen said. "When this occurs, the impact is actually quite significant" and represented a "clear operational constraint."

"Our response was oversight and foreign object detection -- a computer vision solution that uses cameras and machine learning models to actually detect those anomalies in real time, integrated directly into the process control system," Tepponen added.

Computer vision is used at key points along conveyors at BHP's Western Australia Iron Ore operations to help its teams spot oversized rocks or foreign objects and then remove them before they create safety risks, damage equipment or cause unplanned stoppages.

"That one solution is actually delivering to us a little bit under 1 million [metric] tons per annum of uplift potential, which translates then to almost $50 million in value, a 20% reduction in the crusher downtime and up to a 60% reduction in related disruptions," the BHP officer said.

After the technology was implemented in 2025, there have since been no issues caused by oversized or foreign objects, Tepponen told Platts, part of S&P Global Energy, on the sidelines of the expo.

Execution over invention

However, the most important aspect of the technology's effectiveness was not so much the technology itself but how the solution was delivered.

It was "co-designed with the front line [staff] from the get-go, integrated into the existing systems, delivered with pace and agility and quickness and designed to scale from the outset," Tepponen said.

This was then applied from one site to multiple sites and then to other use cases using computer vision, "creating this foundation for future and additional computer vision use cases, not just a single solution."

In this way, BHP was "focusing on building capability, not just a pilot" for technology, which is a "similar pattern we can observe across the mining industry," such as autonomous haulage, which evolved from controlled pilots into large-scale operational capability, Tepponen said.

"What enabled that shift was not just the technology alone, but actually the system around that, the integration into operations, alignment to the operating model and the ability to scale across the network. The same shift from trial to system-level is happening in AI," Tepponen added.

However, a gap remains not only in mining but across other industries using AI "between what is being built and what is actually scaling" in the industry, the BHP officer said.

"That gap will not close with more pilots. It closes when we design for that scale from the start and connect those solutions into real, practical challenges faced by our operations," Tepponen said.

"The future of AI mining will not be defined by how much we experiment, but ... by how effectively we scale what works, because in the end, execution matters more than invention," Tepponen added.