A framework for upscaling and modelling fluid flow for discrete fractures using conditional generative adversarial networks

Scaling up highly heterogeneous aperture distributions of fractures into equivalent permeability tensors enables a substantial reduction in the computational cost simulating fluid flow fractured porous media by allowing employment coarser grids while keeping accuracy an explicit model. This work proposes adaptation and application conditional generative adversarial networks (CGAN) for upscaling single fractures. Three different types are used as input this work: layered media, Zinn & Harvey transformations self-affine fractals. As output, model predicts pressure inside fracture which is calculation tensor. Our results show that framework employing CGAN provides can capture accurately both angle anisotropy discrete fractures, with time when compared to traditional frameworks rely on numerical simulations.