Gardet, Caroline; Le Ravalec, Mickaele et Gloaguen, Erwan (2016). Pattern-based conditional simulation with a raster path: a few techniques to make it more efficient. Stochastic Environmental Research and Risk Assessment , vol. 30 , nº 2. pp. 429-446. DOI: 10.1007/s00477-015-1207-1.
Ce document n'est pas hébergé sur EspaceINRS.Résumé
Mass transport is known to depend on heterogeneity in geological formations. This entails geological bodies with complex geometries. The major interest of multiple-point simulation is its ability to reproduce such geological features through the use of a training image. The idea behind the training image is to describe a geological concept with the expected geological architecture. Its structural content is then used to infer multiple-point statistics. This yields a database with a variety of possible patterns or events. In this paper, we present a hybrid algorithm combining geostatistical multiplepoint and texture synthesis techniques for simulating geological reservoir models constrained to hard data. The proposed algorithm is a two steps process, involving first analysis with the building of an organized database from the training image content, and second synthesis with the simulation of a realization. Various tests are performed to investigate the potential of the algorithm in terms of computation time and ability to properly reproduce the shapes and connectivity features of the objects represented in the training image. We also propose a few improvements to make the algorithm more efficient. Last, six examples are presented based upon different kinds of training images depicting large-scale channelized and fractured media as well as fine-scale porous media.
Type de document: | Article |
---|---|
Mots-clés libres: | geostatistics; multiple-point simulation; training image; texture synthesis; pattern classification |
Centre: | Centre Eau Terre Environnement |
Date de dépôt: | 20 déc. 2016 16:05 |
Dernière modification: | 20 déc. 2016 16:05 |
URI: | https://espace.inrs.ca/id/eprint/3975 |
Gestion Actions (Identification requise)
Modifier la notice |