Our leading experts are at the forefront of rock characterisation research. We are enhancing full core digital imaging and analysis past their physical limits with machine learning and efficient computational techniques. 

Surpassing hardware limitations for Micro-CT imaging 

  • Neural network noise removal allows high quality, transient, minute-span imaging 
  • Super resolution techniques boost resolution of micro-CT images and detail by an order of magnitude to the SEM scale 
  • Texture regeneration allows characterisation of sub-resolution features and mineral content quantification  

Automatic & unbiased mineral identification 

  • Automatic lithology classification from drill core images using convolutional neural networks for sample selection 
  • Determination of minerals in rocks using computer vision 
  • Reduced biased, automatic multi-mineral segmentation maximises physical accuracy of further analysis  

Fast & efficient flow simulation 

  • Machine learning accelerated flow modelling on massively parallel systems allows direct modelling of micro to centimetre scale samples 
  • Determination of permeability, relative permeability curves and capillary pressure curves on rock cores  

Recent research projects 

  • Boosting resolution and recovering texture of rock and medical images 
  • Physically accurate multi-mineral segmentation by neural networks 
  • Accelerated direct multiphase flow simulator 
  • Large image flow simulation by domain decomposition 
  • Predicting and accelerating flow simulation by deep learning successful applications 
  • Conventional reservoirs: sandstones and carbonates 
  • Unconventional reservoirs: tight sandstones and coal beds  

Facilities & infrastructure 

  • Nvidia Titan and Tesla GPU Clusters 

Our people

Contact us

For more information,  A/Professor Peyman Mostaghimi.  

E: peyman@unsw.edu.au