With the inclusion of boundary layer modelling in the gas, the
separation of wave perturbation from and mean flow became less useful,
and potentially prevents further extension to support similar boundary
layer modelling in the liquid.
The mean velocity entry, UMean, is now needed in the
constant/waveProperties file rather than in the waveVelocity boundary
condition.
In order to increase the flexibility of the wave library, the mean flow
handling has been removed from the waveSuperposition class. This makes
waveSuperposition work purely in terms of perturbations to a mean
background flow.
The input has also been split, with waves now defined as region-wide
settings in constant/waveProperties. The mean flow parameters are sill
defined by the boundary conditions.
The new format of the velocity boundary is much simpler. Only a mean
flow velocity is required.
In 0/U:
boundaryField
{
inlet
{
type waveVelocity;
UMean (2 0 0);
}
// etc ...
}
Other wave boundary conditions have not changed.
The constant/waveProperties file contains the wave model selections and
the settings to define the associated coordinate system and scaling
functions:
In constant/waveProperties:
origin (0 0 0);
direction (1 0 0);
waves
(
Airy
{
length 300;
amplitude 2.5;
phase 0;
angle 0;
}
);
scale table ((1200 1) (1800 0));
crossScale constant 1;
setWaves has been changed to use a system/setWavesDict file rather than
relying on command-line arguments. It also now requires a mean velocity
to be specified in order to prevent ambiguities associated with multiple
inlet patches. An example is shown below:
In system/setWavesDict:
alpha alpha.water;
U U;
liquid true;
UMean (1 0 0);
This is to make it clear that the value supplied is the scalar mean
velocity normal to the patch, and to distinguish it from other instances
of the keyword "UMean" which take a vector quantity.
The Scaled Function1 removes the need for classes to hold both a value
and a ramping function. If it is desired to ramp up a velocity up to
(10 0 0) over the space of 5 seconds, that can be achieved as follows:
velocity
{
type scale;
scale
{
type halfCosineRamp;
duration 5;
}
value (10 0 0);
}
Also, as a result of this change, the velocityRamping fvOption has
become a general acceleration source, based on a velocity Function1. It
has therefore been renamed accelerationSource.
Now for transient simulations "Final" solver settings are required for ALL
equations providing consistency between the solution of velocity, energy,
composition and radiation properties.
However "Final" relaxation factors are no longer required for fields or
equations and if not present the standard value for the variable will be
applied. Given that relaxation factors other than 1 are rarely required for
transient runs and hence the same for all iterations including the final one
this approach provide simpler input while still providing the flexibility to
specify a different value for the final iteration if required. For steady cases
it is usual to execute just 1 outer iteration per time-step for which the
standard relaxation factors are appropriate, and if more than one iteration is
executed it is common to use the same factors for both. In the unlikely event
of requiring different relaxation factors for the final iteration this is still
possible to specify via the now optional "Final" specification.
to avoid the need to evaluate departure functions and simplify evaluation of the
temperature. In general it makes more sense to use and e/Cv based
thermodynamics when solving for internal energy rather than h/Cp and have
convert between the energy forms.
All related tutorials and test cases have also been updated.
This provides more flexibility in specifying the allowed internal and boundary
extrema.
For driftFluxFoam and other settling problems it is beneficial to set the
boundaryExtremaCoeff to 1 to allow rapid accumulation of the partials on the
bottom wall (which was the previous default behaviour) but this is not suitable
for many Euler-Euler cases for which a uniform etrema coefficient is preferable,
either 0 or a small value.
Now by default boundaryExtremaCoeff is set to extremaCoeff which defaults to 0
which provides the behaviour before
OpenFOAM-dev commit cb2bc60fa5
and the driftFluxFoam tutorials have been updated adding
boundaryExtremaCoeff 1;
to the MULES controls in fvSolution so reproduce the previous behaviour.
The LBend was set to run for 2 s, but at about 1.95 s the packed region
builds up to the inlet and the simulation diverges. The end time has
been reduced to 1.9 s so that this does not occur.
snappyHexMesh now generates a face-zone for the AMI-s, and createBaffles
and mergeOrSplitPoints -split are used to create the patches. Before,
snappy generated AMI patches directly, which were then converted to
AMI-s with createPatch.
This way, the AMI-s match exactly at the start of the simulation. For
more complicated cases that may be derived from this tutorial, this
could be important.
With the -noFields option the mesh is subset but the fields are not changed.
This is useful when the field fields have been created to correspond to the mesh
after the mesh subset.
including third-body and pressure dependent derivatives, and derivative of the
temperature term. The complete Jacobian is more robust than the incomplete and
partially approximate form used previously and improves the efficiency of the
stiff ODE solvers which rely on the Jacobian.
Reaction rate evaluation moved from the chemistryModel to specie library to
simplfy support for alternative reaction rate expressions and associated
Jacobian terms.
Temperature clipping included in the Reaction class. This is inactive by default
but for most cases it is advised to provide temperature limits (high and
low). These are provided in the foamChemistryFile with the keywords Thigh and
Tlow. When using chemkinToFoam these values are set to the limits of the Janaf
thermodynamic data. With the new Jacobian this temperature clipping has proved
very beneficial for stability and for some cases essential.
Improvement of the TDAC MRU list better integrated in add and grow functions.
To get the most out of this significant development it is important to re-tune
the ODE integration tolerances, in particular the absTol in the odeCoeffs
sub-dictionary of the chemistryProperties dictionary:
odeCoeffs
{
solver seulex;
absTol 1e-12;
relTol 0.01;
}
Typically absTol can now be set to 1e-8 and relTol to 0.1 except for ignition
time problems, and with theses settings the integration is still robust but for
many cases a lot faster than previously.
Code development and integration undertaken by
Francesco Contino
Henry G. Weller, CFD Direct
The nonRandomTwoLiquid and Roult interface composition models have been
instantiated (and updated so that they compile), and a fuller set of
multi-component liquids and multi-component and reacting gases have been
used.
The selection name of the saturated and nonRandomTwoLiquid models have
also been changed to remove the capitalisation from the first letter, as
is consistent with other sub-models that are not proper nouns.
An additional layer has been added into the phase system hierarchy which
facilitates the application of phase transfer modelling. These are
models which exchange mass between phases without the thermal coupling
that would be required to represent phase change. They can be thought of
as representation changes; e.g., between two phases representing
different droplet sizes of the same physical fluid.
To facilitate this, the heat transfer phase systems have been modified
and renamed and now both support mass transfer. The two sided version
is only required for derivations which support phase change.
The following changes to case settings have been made:
- The simplest instantiated phase systems have been renamed to
basicTwoPhaseSystem and basicMultiphaseSystem. The
heatAndMomentumTransfer*System entries in constant/phaseProperties files
will need updating accordingly.
- A phaseTransfer sub-model entry will be required in the
constant/phaseProperties file. This can be an empty list.
- The massTransfer switch in thermal phase change cases has been renamed
phaseTransfer, so as not to be confused with the mass transfer models
used by interface composition cases.
This work was supported by Georg Skillas and Zhen Li, at Evonik
