Computational aspects of the multiscale discontinuous Galerkin method for convection-diffusion-reaction problems (Q2228076)

In this paper, the matrix structure of the multiscale discontinuous Galerkin method (MDG) [\textit{M.-Y. Kim} and \textit{M. F. Wheeler}, Comput. Math. Appl. 68, No. 12, Part B, 2251--2261 (2014; Zbl 1361.35044)] is investigated. The MDG solution is obtained by composition of the discontinuous Galerkin (DG) and the inter-scale operator, whereas the matrix product of the DG matrix and the inter-scale matrix correspond to the local problem on the element. The ILU preconditioned GMRES is applied to effectively solve the resulting global system. Numerical examples for convection dominated problems show that the MDG method numerically is advantageous in adaptive \(p\) refinements and CPU time of the MDG solution is significantly reduced. In all numerical experiments MDG solution exhibits similar behavior of DG with much less degrees of freedom.