Multiscale Flow Solver for Unstructured Networks
Robert Epp · Repository for Publications and Research Data (ETH Zurich) · 2016
The simulation of multiphase flow problems in highly heterogeneous porous media is computationally demanding due to the different length and time scales involved.By using the concepts of pore network modelling and multiscale flow solvers, an approximate solution is found which represents the most important macroscopic flow structures.In this work, the multiscale restriction-smoothed basis (MsRSB) method for the solution of the pressure distribution in porous media is extended for unstructured pore networks.A new method that creates fully connected support regions without isolated pore bodies is introduced.The implementation is purely algebraic and the geometric locations of the pores are not required for the computation.Consequently, the multiscale method can be used as a black-box solver for many different Poisson-type problems.It was observed that the method produces results of good quality for numerous test cases.However, unphysical coarse systems with negative transmissibilities can occur for certain scenarios, especially if large transmissibility contrasts on the fine-scale are involved, e.g.due to channels.This can result in approximate pressure solutions that violate the maximum principle and, if an iterative multiscale formulation is used, the solver may converge slowly or even diverge.The performance is substantially improved when either the restriction or the prolongation operator are adapted to the underlying transmissibility field.