Robert Bridson

Christopher Batty, Robert Bridson

We present a simple and efficient variational finite difference method for simulating time-dependent Stokes flow in the presence of irregular free surfaces and moving solid boundaries. The method uses an embedded boundary approach on staggered Cartesian grids, avoiding the need for expensive remeshing operations, and can be applied to flows in both two and three dimensions. It uses fully implicit backwards Euler integration to provide stability and supports spatially varying density and viscosity, while requiring the solution of just a single sparse, symmetric positive-definite linear system per time step. By expressing the problem in a variational form, challenging irregular domains are supported implicitly through the use of natural boundary conditions. In practice, the discretization requires only centred finite difference stencils and per-cell volume fractions, and is straightforward to implement. The variational form further permits generalizations to coupling other mechanics, all the while reducing to a sparse symmetric positive definite matrix. We demonstrate consistent first order convergence of velocity in L1 and Linf norms on a range of analytical test cases in two dimensions. Furthermore, we apply our method as part of a simple Navier-Stokes solver to illustrate that it can reproduce the characteristic jet buckling phenomenon of Newtonian liquids at moderate viscosities, in both two and three dimensions.

Essex Edwards, Robert Bridson

We present an algebraic method for constructing a highly effective coarse grid correction to accelerate domain decomposition. The coarse problem is constructed from the original matrix and a small set of input vectors that span a low-degree polynomial space, but no further knowledge of meshes or continuous functionals is used. We construct a coarse basis by partitioning the problem into subdomains and using the restriction of each input vector to each subdomain as its own basis function. This basis resembles a Discontinuous Galerkin basis on subdomain-sized elements. Constructing the coarse problem by Galerkin projection, we prove a high-order convergent error bound for the coarse solutions. Used in a two-level symmetric multiplicative overlapping Schwarz preconditioner, the resulting conjugate gradient solver shows optimal scaling. Convergence requires a constant number of iterations, independent of fine problem size, on a range of scalar and vector-valued second-order and fourth-order PDEs.