A Sparsified Multiscale Preconditioner for Fractured Porous Media

Summary In this work, we propose an algorithm to reduce the sparsity of the coarse-scale operator of multiscale methods for fractured porous media, enabling faster and efficient computations. Our approach draws on stencil collapsing techniques used in multigrid methods. A similar strategy was applied to address the lack of monotonicity in multiscale methods. In this work, we mainly focus on applying the sparsified multiscale operator as a preconditioner to a GMRES solver. Consequently, we evaluate its impact on overall convergence performance, as well as the time required to construct and solve the multiscale coarse-scale system.