SPONSOR:
Sandia National Laboratories
PROJECT PERIOD:
04/2019 – 09/2021
ABSTRACT:
This project uses a recently introduced deep learning technique, Deep Convolutional Generative Adversarial Neural Networks (DCGANs), to develop digital rock structures from image datasets. These digital images can then be used to perform numerical simulations and 3D printing work. The proposed DCGANs method will allow fast and reliable reconstruction of the spatial patterns of rock pore structures with channel connectivity and multi-scale variability. Rock physics analysis with digital datasets will lead to a better understanding of poromechanical and flow responses of geomaterials under different conditions.
Project Publication:
Kim, S.E., Y. Seo, J. Hwang, H. Yoon, and J. Lee. 2020. Connectivity-informed Drainage Network Generation using Deep Convolution Generative Adversarial Networks. arXiv:2006.13304
PRINCIPAL INVESTIGATOR
