Numerical simulation of the effective mechanical properties of the core samples by GPU computing.

Autor: Yakovlev, Maxim, Nikitin, Leonid, Levin, Vladimir
Předmět:
Zdroj: AIP Conference Proceedings; 2023, Vol. 2899 Issue 1, p1-7, 7p
Abstrakt: In this article, a method and some results of the numerical simulation of the effective elastic properties of core samples are described. The method consists in carrying out a series of stress-strain calculations on a 3D digital model of a core sample. The digital model is a voxel structure obtained from core sample CT scan data. Based on the features of the core sample, a representative volume is determined, on which several static boundary value problems of elasticity with different boundary conditions are solved. A digital core sample model can contain up to one billion voxels and more. Therefore, the relaxation method and the finite difference method (FDM) with explicit time scheme, which do not require a large amount of RAM during the calculation, are used to solve static elasticity problems. In addition, the calculations are parallelized using the CUDA technology for high-performance GPU computing. The effective mechanical properties of the core sample are estimated by averaging over RVE generalized Hooke's law using the results of solving each static problem. The paper presents GPU calculation results for two core samples: sandstone and limestone. The calculations are carried out on full-size core samples. The comparison of results for small core fragments with CAE Fidesys is carried out for verification. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index