Rio Tinto has developed and deployed an advanced AI tool named ReconAI to enhance its defect elimination processes within its extensive rail operations, particularly targeting its autonomous locomotive fleet. This innovation is part of the miner’s broader strategy to maximise the use of data for improving asset reliability and availability, an imperative for its automated heavy-haul rail network operating in remote regions.
ReconAI automates crucial aspects of the defect elimination workflow—specifically classifying faults detected in asset health data and measuring the effectiveness of corrective actions. Traditionally, such classification required highly skilled engineers to manually catalogue faults among hundreds of systems and components on each locomotive. By taking on three of the six defect elimination steps, ReconAI significantly reduces the repetitive administrative burden on engineers, freeing them to focus on fieldwork such as delivering projects, collaborating with operations, and responding to breakdowns.
The tool integrates data drawn from multiple sources, including telemetry logs, alarms, and freeform textual maintenance records stored in systems like SAP and HOSDI. These records often contain detailed narratives entered by operational and workshop staff about fault diagnoses, troubleshooting, and repairs. Leveraging this wealth of qualitative and quantitative data, ReconAI provides engineers with a clear, granular view of each fault, identifying the precise equipment or subsystem and failure mode responsible.
Technically, ReconAI operates on a technology stack utilising cloud-native AWS services. Data from Rio Tinto’s rail and enterprise systems is ingested into Amazon S3 for analysis via Amazon Glue and Athena. The AI backend runs on Amazon EC2, employing retrieval augmented generation workflows powered by Anthropic Claude AI models via Amazon Bedrock. This approach enables not only high classification accuracy—improved from an initial 80% to 96% in production—but also transparency, as users can trace the AI’s reasoning and justifications throughout the classification process. This transparency has been critical to building trust among engineers, known for their scepticism of AI in complex, jargon-heavy environments.
Rio Tinto first applied ReconAI to its fleet, which numbers in the hundreds, that autonomously haul iron ore over a 1,500km network in Western Australia, from mine pits to ports. Minimising downtime is especially critical given the remote and autonomous nature of these operations; fault recovery can be time-consuming and costly when staff need to attend to breakdowns in the field. Following its successful deployment for locomotives, Rio Tinto is expanding ReconAI to cover other rail assets such as ore cars, signaling equipment, and port conveyors.
This AI-driven maintenance innovation aligns with Rio Tinto’s broader digital transformation journey. The company was an early pioneer in autonomous rail technology, having launched AutoHaul™ in 2018—the world’s largest autonomous heavy-haul rail network. The integration of AI tools like ReconAI complements other initiatives in predictive maintenance and smart operations, as Rio Tinto continues embedding AI and machine learning technologies across its mining and logistics assets to boost efficiency, safety, and operational predictability.
Furthermore, Rio Tinto has been involved in pilot programs and studies aimed at enhancing the safety and capability of its autonomous trains, such as testing AI-based hazard detection systems from vendors like Rail Vision and 4Tel. These efforts indicate a comprehensive approach, combining advanced AI for both asset health management and real-time operational safety.
Rio Tinto’s approach demonstrates not only significant technical innovation but also organisational learning. The deployment of ReconAI has improved data quality by providing feedback to maintenance supervisors on better data capture practices, enhancing the overall effectiveness of foundational operational processes. On the cost front, while current efforts focus on improving AI accuracy, the company is also working on optimising cost-efficiency through techniques like batch inferencing and prompt caching.
In summary, ReconAI exemplifies Rio Tinto’s commitment to leveraging cutting-edge AI in transforming traditional industrial maintenance. By automating fault classification and defect elimination, the miner not only improves asset uptime but empowers its engineers to focus on impactful fieldwork, setting new standards for autonomous rail operations within the mining sector.
Source: Noah Wire Services
