Speaker
Description
Permeability determines how easily fluids move through porous materials, controlling flow in natural and engineered systems such as groundwater filtration, enhanced oil recovery, and CO2 sequestration. Traditional approaches to permeability calculations, based on direct experiments or numerical flow simulations, are accurate but computationally expensive.
In the first part of this presentation, we explore the utility of machine learning informed by topology and network descriptors applied to three-dimensional (3D) synthetic data to predict permeability efficiently while maintaining interpretability. Our approach combines geometric analysis, pore-network modeling, and topological data analysis (TDA) to build predictive models that are both data-driven and physically meaningful.
In the second part of the talk, we discuss the application of TDA to the experimental data (3D micro-CTs of porous rocks), focusing on understanding scalability and answering the following question: How large an experimental sample needs to be so that the computed measures are system-size-independent?
Acknowledgment: This work is supported by NSF Grants DMR-2410985 and DMS-2201627, and NJIT GHAIRI program.
| Country | USA |
|---|---|
| Acceptance of the Terms & Conditions | Click here to agree |
