Rio Tinto Employs AI to Document 30-Year-Old Production System

Quick Read

• Rio Tinto leverages AI to document its three-decade-old Metpro system.
• Metpro oversees the entire aluminium product lifecycle.
• The AI model utilizes Amazon Bedrock and SageMaker AI Jumpstart.
• Engineers can now rapidly grasp system dependencies.
• AI assists in preserving and activating historical system knowledge.

Utilising AI for System Documentation

Rio Tinto has harnessed artificial intelligence to thoroughly document the intricacies inherent in its longstanding manufacturing execution system, Metpro. The AI-driven domain assistant acts as a repository for the accumulated knowledge and complex interdependencies of this essential system.

The Role of Metpro

Metpro, a vital component of Rio Tinto’s aluminium operations, administers the lifecycle of aluminium products, from initial tapping to final dispatch. It functions as a core interface, connecting process control systems, data platforms, and operational frameworks.

Challenges of Fragmented Documentation

Throughout the years, the technical documentation of Metpro became fragmented, dispersed across numerous files. This disconnection, combined with the tightly-knit structure of the system, created considerable challenges in comprehending dependencies, heightening the risk for engineers during updates or alterations.

AI-Powered Solution

Preserving Embedded Knowledge

Rather than replacing the established Metpro system, Rio Tinto aimed to preserve and operationalize the vast knowledge it harboured. By developing a domain-relevant training dataset, they merged Metpro’s code with Amazon Bedrock Knowledge Bases and AgentCore, creating a unified understanding of the system.

Utilising Amazon SageMaker AI Jumpstart

Rio Tinto implemented Amazon SageMaker AI Jumpstart, using Llama 3.1 8B as the foundational inference model. The AI was methodically trained to comprehend the system’s operations, allowing engineers to quickly access essential system insights.

Real-Time Adaptation

The AI model, enhanced with an agent layer, sustains an up-to-date comprehension of the system, adjusting to new information without the need for frequent retraining. This guarantees the AI remains a trustworthy source of knowledge, aiding informed decision-making.

Revolutionising Engineering Processes

With the AI domain assistant, engineers can swiftly grasp system dependencies, enabling prompt and informed modifications. The time conserved from reverse-engineering intricate logic can be redirected toward innovation and enhancements.

Establishing a Future-Ready Foundation

This AI initiative has established a strong groundwork for future advancements, capturing and applying knowledge in a way that supports gradual, low-risk modernization while safeguarding the integrity of Rio Tinto’s crucial systems.

Summary

Rio Tinto’s strategic implementation of AI has transformed the documentation and comprehension of its Metpro system, ensuring ongoing operational efficiency and paving the path for future technological progress.

Q&A

Q: What is the primary function of the AI domain assistant?

A:

The AI domain assistant is designed to document and safeguard the knowledge, dependencies, and decision-making logic of Rio Tinto’s Metpro system.

Q: In what way does Metpro enhance Rio Tinto’s operations?

A:

Metpro oversees the complete aluminium product lifecycle from start to finish, serving as a central hub for process control and operational data.

Q: Why was AI selected instead of substituting Metpro?

A:

AI was chosen to maintain and operationalize the vast embedded knowledge within Metpro without the hazards linked to system replacement.

Q: How does the AI model keep itself updated?

A:

An agent layer within the AI model ensures it stays connected to real-time data, adapting without the necessity for regular retraining.

Q: What technologies were employed in the AI solution?

A:

The solution utilized Amazon Bedrock Knowledge Bases, AgentCore, and Amazon SageMaker AI Jumpstart featuring Llama 3.1 8B.

Q: What advantages does the AI solution offer engineers?

A:

Engineers can swiftly understand system dependencies, conserving time on reverse-engineering and enabling them to concentrate on innovation.