Multi phase solver

If you are creating a two fluid simulation you will have to specify some parameters of the multiphase solver. The best tested options are the SinglePhase and the BlendedAlgebraicVOF solvers. You can also add your own multi phase model, see Writing a custom multiphase model.

Single phase

The default value of the multi phase solver setup is to use a single phase method:

multiphase_solver:
    type: SinglePhase

Since this is the default, the multiphase_solver section can be omitted from the input file when running single phase simulations.

Algebraic Volume of Fluid method

This is an algebraic VOF method where the interface is captured as the 0.5 level set of a colour function which should be strictly within the range from 0 to 1 inclusive.

When using the multi phase VOF solver by specifying type: BlendedAlgebraicVOF the following parameters can be specified:

function_space_colour

CG for continuous Galerkin, DG for discontinuous Galerkin (default)

polynomial_degree_colour

The degree of the approximating polynomials, default 0 (piecewise constant)

num_subcycles

Number of times the VOF method is run per time step. Running VOF with a smaller time step can be beneficial to decrease the Courant number. The VOF calculations are typically much faster than the Navier-Stokes solver. Default value 1, using about 5 sub cycles can to give good results in many cases without much impact on running time.

force_static

Do not move the free surface. Can be useful for testing in some cases.

project_uconv_dgt0

Project the velocity that will convect the colour function into a DGT0 field, piecewise constant on each facet. This is consistent with how the finite volume method handles advection, it can be more stable (lower local Courant numbers) and it is mass conserving in DG0 space. Default on.

In addition you will have to specify a convection scheme for the VOF colour function in order to keep the free surface sharp. For specifying the convection scheme, see Convection.

Other methods

There are some other options that exist in the code base, but these are not well tested and very likely need some work before they can run. They are

  • VariableDensity (for mixtures without sharp interfaces)

    This model was actually used a bit some time ago, so it may work with a few modifications. It can be used with higher order descriptions of the density, and with an appropriate slope limiter it could potentially deal with sharp interface problems as well as diffusive mixes. There is a convergence test in cases/convergece-variable-density-disk that is not tested by default currently, but has been used to show that the multi phase model setup can produce the correct order of convergence for velocity, pressure and density. If you want to resurrect this method then starting with getting the convergence case running is probably the best option. It probably works with few or no code modifications.

  • HeightFunction

    Implemented for 2D simuations. A single valued height function separates two fluid domains. Uses the height function to compute a VOF field which reuses most of the BlendedAlgebraicVOF machinery. Not used in a long while, so the code is probably somewhat bitrotted (probably no parallel support etc).

  • HeightFunctionALE

    Implemented for 2D simuations. The mesh is moved according to the vertical fluid velocity at the free surface after each time step. Not used in a long while, probably somewhat bitrotted. Would need stablisation/smoothing to combat sawtooth instabilities if it was ever to be used for something.

    Not used in a long while, so the code is certainly bitrotted and the ALE code in Ocellaris in general is not tested at all at the moment, so expect problems. There is also probably no support for running in parallel with MPI etc.

  • Lagrangian

    A purely Lagrangian multiphase model. The mesh is moved according to the calculated fluid velocity after each time step. This will obviously distort the mesh in allmost all calculations.

    This was implemented as a stepping stone to ALE, and to test hydrostatic pressure calculations where the correct answer is zero velocity everywhere for all time and ALE should not be necessary.

    To initialise the multi phase field the colour function must be specified in the input file (as initial condition for “cp”). The colour function is unity when rho=rho0 and nu=nu0 and zero when rho=rho1 and nu=nu1