Abstract

In the present work an implementation of the Back and Forth Error Compensation and Correction (BFECC) algorithm specially suited for running on General-Purpose Graphics Processing Units (GPGPUs) through Nvidia’s Compute Unified Device Architecture (CUDA) is analyzed in order to solve transient pure advection equations. The objective is to compare it to a previous explicit version used in a Navier-Stokes solver fully written in CUDA. It turns out that BFECC could be implemented with unconditional stable stability using Semi-Lagrangian time integration allowing larger time steps than Eulerian ones.

Abstract

In this work, we consider a relatively simple case of fluid monitoring in a subsurface gas reservoir. Seismic wave velocities for porous rocks vary depending on fluid saturation, and our objective is to quantitatively evaluate the impact of the water/gas contact depth on elastic wave propagation. To efficiently test different contact depth scenarios and assess their impact on wave propagation, we propose to locally modify a 2D geological model and run time-dependent elastic simulations. The input model is a triangulated surface conforming to geological structures and representing physical properties. The 2D meshed model is locally updated, meaning that only a given region is modified and that the other parts of the mesh remain identical. To create several models by modifying only the reservoir layer, we insert a water/gas contact defined by a level-set at several depths with MMG. During the insertion, specific care is taken to maintain the conformity of the output mesh. As compared to global remeshing, the local modification reduces the cost of recomputing physical properties over the updated mesh. We run the numerical simulations by using Hou10ni2D code, which is based on a Discontinuous Galerkin method. Our results on a gas reservoir show a consistent behavior: we observe a correlation between the depth difference and L2-norm, the larger the distance from the reference depth contact, the higher the L2-norm. This approach could therefore be integrated into an inversion loop to determine the position of the fluid contact and reduce uncertainties in the reservoir model.