WEBINAR – 15 October 2026, 15:00 – 15:30 CEST
Using advanced neural networks to classify and map the connectivity of different types of pore architecture in porous media
Paul Glover, Professor of Petrophysics,
presenting the work of Wurood Alwan
University of Leeds in Leeds, United Kingdom
EXCITE User
Please note that the time is different from the usual time!

Join Paul Glover from the University of Leeds to discover how advanced AI and microscopy can reveal not just where pores are located, but also how they connect. This has important implications for understanding the subsurface and applications like CO₂ storage and geothermal energy.
In this webinar, Paul will explore how deep-learning approaches can be used to classify and map different types of pore architecture in porous media. This is based on the work done by Wurood Alwan using facilities within EXCITE Network. Focusing on high-resolution microscopy images of carbonate rocks, Paul will demonstrate how “attention-based U-Net architectures” can distinguish between connected pore pathways, isolated pores and mineral matrix.
In more detail
The ability to identify pore connectivity is crucial for understanding fluid transport and predicting the behaviour of porous materials. It is particularly relevant to applications such as CO₂ storage and geothermal energy, where the distinction between connected and isolated porosity can strongly influence flow, trapping and storage behaviour.
The speaker will introduce the multiscale Attention U-Net approach, discussing how neural networks can capture pore structures across different length scales and address the challenges posed by highly imbalanced classes and complex pore geometries. He will also show how connectivity-aware pore maps can provide valuable input for downstream digital-rock and petrophysical analyses.
