2024 Capability Portfolio Digital | Page 155

Clean Energy Capability Portfolio | Energy Markets

Artificial Intelligence

for Rock and Ore
Characterisation

An innovative platform for using advanced data analytics methods , based on convolutional neural networks and generative adversarial networks . These methods enable the prediction of important properties of rock and ore , to improve efficiency and promote automation in minerals and energy resources engineering .

Competitive Advantage

• Ability to determine mineral contents of rock / ore at high-resolution using machine learning
• Ability to predict porosity , permeability , and relative permeability curves of reservoir rock
• Leading the automatic identification of :
• patterns and features in rock / ore images
• fractures in drill cores

Impact

• Automated analyses of drill cores for identifying features of interest
• Improved efficiency in the mining industry through the reliable determination of mineral contents
• High-fidelity reservoir models to optimise recovery

More Information

Professor Peyman Mostaghimi School of Mineral and Energy Resources Engineering T : + 61 2 9385 5122 E : peyman @ unsw . edu . au
Professor Ryan Armstrong School of Mineral and Energy Resources Engineering T : + 61 2 9385 5122 E : ryan . armstrong @ unsw . edu . au
Dr Ying Da Wang School of Mineral and Energy Resources Engineering T : + 61 2 9385 5122
E : yingda . wang @ unsw . edu . au

Successful Applications

• Tested on several reservoir rocks where petrophysical properties were predicted with high accuracy
• Tested on multi-mineral rocks with mineral contents identified

Capabilities and Facilities

• Computational facilities for training and testing
• High-resolution X-ray CT scanners

155