Corrected according to the original reference:
Marschall, H. (2011).
Towards the numerical simulation of multi-scale two-phase flows.
PhD Thesis, TU München.
The syntax of this model has changed to permit transfers of species in
either direction. A list of transferring species is now given for each
phase, rather than identifying a single reacting phase. For example:
phaseTransfer
(
vapour_particles
{
type reactionDriven;
// TiO2 and TiO2_s are created by reactions in the vapour
// and are then transferred to the particles
species.vapour (TiO2 TiO2_s);
// H2O is created by reactions in the particles and is then
// transferred to the vapour
species.particles (H2O);
}
);
These models represent a phase nucleating directly out of a
multi-component mixture into isolated particles (i.e., homogeneous),
rather than onto an existing surface or impurity (heterogeneous).
The homogeneousCondensation model can represent the initial stages of a
gas condensing into a liquid. Example usage, in constant/fvModels:
homogeneousCondensation
{
type homogeneousCondensation;
libs ("libmultiphaseEulerFvModels.so");
// Phases between which the transfer occurs. The first phase is the
// gas, and the second is the condensed liquid.
phases (gas water);
// The specie that is condensing
specie H2O;
// Linearise the latent heat contribution into the energy equation?
energySemiImplicit no;
// Saturation curve for the specie in the gaseous phase
pSat ArdenBuck;
}
The homogeneousLiquidPhaseSeparation model can represent the initial
stages of a liquid solution precipitating out a solid or separating into
two immiscible liquid phases: Example usage, in constant/fvModels:
homogeneousLiquidPhaseSeparation
{
type homogeneousLiquidPhaseSeparation;
libs ("libmultiphaseEulerFvModels.so");
// Phases between which the transfer occurs. The first phase is the
// solution, and the second is the precipitate.
phases (liquid sugar);
// The specie that is condensing
specie C2H12O6;
// Linearise the latent heat contribution into the energy equation?
energySemiImplicit no;
// Solubility given in mass of solute per mass of solvent
solubility constant 0.9;
}
If population balance is being used, then both of these models require a
source term to be applied to the size-group equations. This is achieved
by means of a new nucleationSizeGroup field source. Example usage, in
0/fDefault.water:
sources
{
homogeneousCondensation
{
type nucleationSizeGroup;
libs ("libmultiphaseEulerFvModels.so");
}
}
Functions have been added to populationBalance to generate the
volumetric allocation coefficient etaV. This is needed if a volume or
mass source is to be conveniently distributed into the size groups. The
number-based allocation coefficient, eta, is still available and is
still used in all cases within populationBalance.
The allocation bounds handling has been removed. This, as it turns out,
was just an incomplete subset of having both number- and volume-based
allocation coefficients implemented.
Now that BasicThermo::Cp returns a volScalarField reference because Cp is cached
in BasicThermo code calling Cp can hold a reference rather than a copy for
efficiency.
The Clang compiler does not use std::move to transfer the result of the ternary
operator into the phase-fraction field resulting in it not being registered to
the database. To work around this limitation/bug the ternary operator is now
provided with tmp fields the result of which is passed with an IOobject to the
final field constructor to ensure it is registered and the IO options set
correctly.
An enumeration has been added to the arguments of the allocation
coefficient function, eta, to allow specification of how to allocate out
of bounds of the population balance size-groups. There are two options:
- "Clamp" will create an out-of-bounds allocation coefficient of exactly
one. This partitions unity across all size-space.
- "Extrapolate" will create an out-of-bounds allocation coefficient in
proportion to the ratio between the given size and the nearest
size-group size. This does not partition unity outside the range of
the size-groups.
The previous operation is equivalent to "Extrapolate".
It is not yet clear which method is preferable and under what
circumstances. More testing is required. The enumeration has been
created to facilitate this testing.
The specie molecular mass (Wi) and formation enthalpy (hfi) methods now
return dimensioned scalars. This permits their direct inclusion into
dimensioned field expressions. Non-dimensioned methods have been
retained with a "Value" suffix (i.e., WiValue and hfiValue).
The nearWallDist MeshObject is now deleted on mesh-change rather than updated
which is more efficient for cases with multiple mesh changes, e.g. motion,
stitching and mapping by avoiding unnecessary updates.
As a consequence of this change the wallDist::y() volScalarField reference
should not be cached across mesh changes so the turbulence models now obtain the
y field as required from the new momentumTransportModel::y() function, the
original near-wall distance function is now named momentumTransportModel::yb()
to clarify that it is the wall distance of the boundary cells.
This simple model creates a heat transfer coefficient in proportion with
the corresponding drag model's momentum transfer coefficient. A
user-defined Prandtl number and a harmonic average of the phases'
specific heats are used to specify the constant of proportionality.
This model has no physical basis. It exists primarily for testing
purposes. It has the advantage of being applicable to any interface,
including those representing segregated configurations.
Example usage:
heatTransfer
{
gas_segregatedWith_liquid
{
type Prandtl;
Pr 0.7;
}
}
The patch-specific mapper interfaces, fvPatchFieldMapper and
pointPatchFieldMapper, have been removed as they did not do anything.
Patch mapping constructors and functions now take a basic fieldMapper
reference.
An fvPatchFieldMapper.H header has been provided to aid backwards
compatability so that existing custom boundary conditions continue to
compile.
Currently in compressibleVoF vDot contains only the compressibility dilatation
effect whereas in multiphaseEuler the effect of sources are also included but
this will be refactored shortly so that the handling of mass sources and
compressibility is consistent between VoF and Euler-Euler solvers.
The previously hard-coded 1e-4 division stabilisation used when linearising vDot
for bounded semi-implicit solution of the phase-fractions is now an optional
user-input with keyword vDotResidualAlpha, e.g. in multiphaseEuler:
solvers
{
"alpha.*"
{
nAlphaCorr 1;
nAlphaSubCycles 2;
vDotResidualAlpha 1e-6;
}
.
.
.
IsothermalSolidPhaseModel does not update energy and density from pressure
whereas IsothermalPhaseModel does to allow compressible fluid phases to change
volume due to pressure changes.
for thermophysical transport within stationary solid phases. This provides a
consistent interface to heat transport within solids for single and now
multiphase solvers so that for example the wallHeatFlux functionObject can now
be used with multiphaseEuler, see tutorials/multiphaseEuler/boilingBed.
Also this development supports anisotropic thermal conductivity within the
stationary solid regions which was not possible previously.
The tutorials/multiphaseEuler/bed and tutorials/multiphaseEuler/boilingBed
tutorial cases have been updated for phaseSolidThermophysicalTransportModel by
changing the thermo type in physicalProperties.solid to heSolidThermo. This
change will need to be made to all multiphaseEuler cases involving stationary
phases.
The interface for fvModels has been modified to improve its application
to "proxy" equations. That is, equations that are not straightforward
statements of conservation laws in OpenFOAM's usual convention.
A standard conservation law typically takes the following form:
fvMatrix<scalar> psiEqn
(
fvm::ddt(alpha, rho, psi)
+ <fluxes>
==
<sources>
);
A proxy equation, on the other hand, may be a derivation or
rearrangement of a law like this, and may be linearised in terms of a
different variable.
The pressure equation is the most common example of a proxy equation. It
represents a statement of the conservation of volume or mass, but it is
a rearrangement of the original continuity equation, and it has been
linearised in terms of a different variable; the pressure. Another
example is that in the pre-predictor of a VoF solver the
phase-continuity equation is constructed, but it is linearised in terms
of volume fraction rather than density.
In these situations, fvModels sources are now applied by calling:
fvModels().sourceProxy(<conserved-fields ...>, <equation-field>)
Where <conserved-fields ...> are (alpha, rho, psi), (rho, psi), just
(psi), or are omitted entirely (for volume continuity), and the
<equation-field> is the field associated with the proxy equation. This
produces a source term identical in value to the following call:
fvModels().source(<conserved-fields ...>)
It is only the linearisation in terms of <equation-field> that differs
between these two calls.
This change permits much greater flexibility in the handling of mass and
volume sources than the previous name-based system did. All the relevant
fields are available, dimensions can be used in the logic to determine
what sources are being constructed, and sources relating to a given
conservation law all share the same function.
This commit adds the functionality for injection-type sources in the
compressibleVoF solver. A following commit will add a volume source
model for use in incompressible solvers.
This change makes multiphaseEuler more consistent with other modules and
makes its sub-libraries less inter-dependent. Some left-over references
to multiphaseEulerFoam have also been removed.
requiring all parts of the moving phase solution algorithm to loop and operate
on the moving phases only making the code easier to understand and maintain.
The central coefficient part of the virtual-mass phase acceleration matrix is
now included in the phase velocity transport central coefficient + drag matrix
so that the all the phase contributions to each phase momentum equation are
handled implicitly and consistently without lagging contribution from the other
phases in either the pressure equation or phase momentum correctors.
This improves the conditioning of the pressure equation and convergence rate of
bubbly-flow cases.
Population balance size-group fraction 'f<index>.<phase>' fields are now
read from an 'fDefault.<phase>' field if they are not provided
explicitly. This is the same process as is applied to species fractions
or fvDOM rays. The sum-of-fs field 'f.<phase>' is no longer required.
The value of a fraction field and its boundary conditions must now be
specified in the corresponding field file. Value entries are no longer
given in the size group dictionaries in the constant/phaseProperties
file, and an error message will be generated if a value entry is found.
The fraction fields are now numbered programatically, rather than being
named. So, the size-group dictionaries do not require a name any more.
All of the above is also true for any 'kappa<index>.<phase>' fields that
are constructed and solved for as part of a fractal shape model.
The following is an example of a specification of a population balance
with two phases in it:
populationBalances (bubbles);
air1
{
type pureIsothermalPhaseModel;
diameterModel velocityGroup;
velocityGroupCoeffs
{
populationBalance bubbles;
shapeModel spherical;
sizeGroups
(
{ dSph 1e-3; } // Size-group #0: Fraction field f0.air1
{ dSph 2e-3; } // ...
{ dSph 3e-3; }
{ dSph 4e-3; }
{ dSph 5e-3; }
);
}
residualAlpha 1e-6;
}
air2
{
type pureIsothermalPhaseModel;
diameterModel velocityGroup;
velocityGroupCoeffs
{
populationBalance bubbles;
shapeModel spherical;
sizeGroups
(
{ dSph 6e-3; } // Size-group #5: Fraction field f5.air2
{ dSph 7e-3; } // ...
{ dSph 8e-3; }
{ dSph 9e-3; }
{ dSph 10e-3; }
{ dSph 11e-3; }
{ dSph 12e-3; }
);
}
residualAlpha 1e-6;
}
Previously a fraction field was constructed automatically using the
boundary condition types from the sum-of-fs field, and the value of both
the internal and boundary field was then overridden by the value setting
provided for the size-group. This procedure doesn't generalise to
boundary conditions other than basic types that store no additional
data, like zeroGradient and fixedValue. More complex boundary conditions
such as inletOutlet and uniformFixedValue are incompatible with this
approach.
This is arguably less convenient than the previous specification, where
the sizes and fractions appeared together in a table-like list in the
sizeGroups entry. In the event that a substantial proportion of the
size-groups have a non-zero initial fraction, writing out all the field
files manually is extremely tedious. To mitigate this somewhat, a
packaged function has been added to initialise the fields given a file
containing a size distribution (see the pipeBend tutorial for an example
of its usage). This function has the same limitations as the previous
code in that it requires all boundary conditions to be default
constructable.
Ultimately, the "correct" fix for the issue of how to set the boundary
conditions conveniently is to create customised inlet-outlet boundary
conditions that determine their field's position within the population
balance and evaluate a distribution to determine the appropriate inlet
value. This work is pending funding.
Now with the addition of the optional dependenciesModified() function classes
which depend on other classes which are re-read from file when modified are also
automatically updated via their read() function called by
objectRegistry::readModifiedObjects.
This significantly simplifies the update of the solutionControls and modular
solvers when either the controlDict or fvSolution dictionaries are modified at
run-time.
The momentum equation central coefficient and drag matrix is formulated,
inverted and used to eliminate the drag terms from each of the phase momentum
equations which are combined for formulate a drag-implicit pressure equation.
This eliminates the lagged drag terms from the previous formulation which
significantly improves convergence for small particle and Euler-VoF high-drag
cases.
It would also be possible to refactor the virtual-mass terms and include the
central coefficients of the phase acceleration terms in the drag matrix before
inversion to further improve the implicitness of the phase momentum-pressure
coupling for bubbly flows. This work is pending funding.
When the flow is stationary (e.g., at the beginning of a run) the
rDeltaT calculation now requires a maxDeltaT setting in the PIMPLE
sub-section of the fvSolution dictionary. This prevents floating point
errors associated with rDeltaT approaching zero.
The phase velocity mean adjustment was introduced for consistency with phase
flux mean adjustment which is necessary to ensure the mean flux divergence is
preserved. However for systems with very high drag it has proved preferable to
not adjust the velocity of the phases to conserve momentum rather than ensure
consistency with the fluxes.
