Abstract

As global demand for critical minerals and sustainable practices intensifies, the mining industry faces the dual challenges of maximizing operational efficiency and reducing environmental impact. Traditional methods often struggle with fragmented data systems, lack of interoperability, and an inability to adapt quickly to changing conditions. Addressing these limitations, Airth Intelligence, built atop the robust Airth Core modules, offers the industry’s most advanced, AI-driven platform. This sophisticated solution integrates schema and graph ontology and Graph Neural Networks (GNNs) to provide actionable insights and prescriptive analytics, transforming mining operations into autonomous, efficient, and sustainable ecosystems. By optimizing resource use and minimizing environmental impact, Airth Intelligence aligns with ESG and sustainability goals, enabling mining operations to not only meet regulatory standards but actively contribute to sustainable resource management. 

Airth Core serves as the foundation of Airth Intelligence, providing a low-code, vendor-neutral architecture that integrates seamlessly across the entire mining value chain. Core’s ontology based system enables real-time data contextualization, validation and interoperability, unifying diverse data sources such as planning, operations, equipment health, and financial metrics. This dynamic ontology allows Airth Intelligence to continuously adapt to changing data and operational changes in the mine and deliver contextually aware, agile, data-driven decision-making that fosters both sustainability and profitability. Airth Intelligence has been enhanced by Airth Plato, a mining-specific Large Language Model (LLM) and Language Action Model (LAM) that has been trained on mining data and is trained on the mine site specific data for site enhancements. This grounding of the LLM using the ground truths of the mine and context of the mine ontology helps the Airth Intelligence system to guide users in interpreting data and prioritizing actions, making advanced insights generated by the system actionable. 

Airth Intelligence utilizes graph structures and databases to model complex, interrelated mining data and uses a combination of Convolutional Neural Networks (CNN) and Graph Neural Networks (GNN) to uncover insights that traditional methods may miss. GNNs enhance predictive accuracy by capturing the intricate relationships between equipment performance, resource allocation, and environmental conditions. With these capabilities, Airth Intelligence facilitates predictive maintenance and precise operational planning, significantly reducing unplanned downtimes, optimizing fleet usage, and supporting on-the-fly adjustments in equipment and resources. For example, a recent proof of concept study that is done at a major copper mining site achieved notable improvements in equipment uptime and fleet utilization, and ultimately can yield into multimillion-dollar savings while reducing emissions and resource waste.