作者: Shriram Srinivasan , Jeffrey D. Hyman , Daniel O'Malley , Satish Karra , Hari S. Viswanathan
DOI: 10.1016/BS.AGPH.2020.08.001
关键词: Machine learning 、 Network connectivity 、 Graph theory 、 Graph (abstract data type) 、 Artificial intelligence 、 Computer science
摘要: Abstract This is primarily an account of the role machine learning, including played by graph theory, in development reduced-order models (ROM) flow and transport through fractured media that are modeled with DFN approach. We describe construction a DFN, necessary governing equations for various representations it. The ROMs then traced divided into different unifying themes. set out, general terms, learning constructing place three approaches have been developed this context abstract perspective, explain fundamental ideas each show differ identity elements being classified, rule assigning labels to elements. By choosing as paths, rather than fractures, truly physics-informed method results preserves network connectivity reduced networks.