A new computational framework referred to as algebraic variational multiscale?
multigrid method is proposed, representing the initial step of our work on merging
the variational multiscale method with algebraic multigrid projection methods. This
new approach allows for a separation of resolved scales into various scale groups in a
purely algebraic way, that is, with no need to utilize further discretizations beyond
the basic one. By this means, it conveniently facilitates the application of modeling
terms either to all scale groups or only to selected scale groups, for instance, only
to the fine scales of the problem. Variational multigrid techniques form the basis
of our new approach. After presenting the cornerstones of the framework for an
abstract problem formulation, the close relationship of the variational multiscale
method to projection methods in general and to variational multigrid techniques
are discussed. The algebraic variational multiscale?multigrid method is then particularly
analyzed for convection?diffusion problems. The present initial study focuses
on exploiting the methodical aspects of the new framework by developing a fine?
scale discontinuity?capturing approach to diminish oscillations at sharp layers. Our
technique including a fine?scale discontinuity-capturing term is applied to numerical
examples of strongly convection?dominated convection?diffusion problems. The
results demonstrate that it enables the diminuation of local oscillations at parabolic
boundary and interior layers by about 60 - 80 percent without any notable smearing effect.
«
A new computational framework referred to as algebraic variational multiscale?
multigrid method is proposed, representing the initial step of our work on merging
the variational multiscale method with algebraic multigrid projection methods. This
new approach allows for a separation of resolved scales into various scale groups in a
purely algebraic way, that is, with no need to utilize further discretizations beyond
the basic one. By this means, it conveniently facilitates the application of...
»