Learning the matrix-fractures transfer rate using a convolutional neural network

N. Andrianov, H. M. Nick

Research output: Contribution to conferencePaper at conferencepeer-review

Abstract

One of the key elements in constructing of representative dual porosity/dual permeability models is to provide the mass transfer rate between the matrix and the fractures. Whereas it is possible to compute numerically this transfer rate for specific geometries, it is challenging to estimate the transfer function without running the CPU intensive computations. In this work, we demonstrate that a convolutional neural network can approximate a transfer function using the encoded fracture geometry and the precomputed fine-scale simulation results.

Original languageEnglish
DOIs
Publication statusPublished - 2020
Event1st EAGE Digitalization Conference and Exhibition - Vienna, Austria
Duration: 30 Nov 20203 Dec 2020

Conference

Conference1st EAGE Digitalization Conference and Exhibition
Country/TerritoryAustria
CityVienna
Period30/11/203/12/20

Programme Area

  • Programme Area 3: Energy Resources

Fingerprint

Dive into the research topics of 'Learning the matrix-fractures transfer rate using a convolutional neural network'. Together they form a unique fingerprint.

Cite this