The implementation of the optional non-uniform transformations in coupled
patches was based on transform property lists which could be either length 0 for
no transformation, 1 for uniform transformation or n-faces for non-uniform
transformation. This complexity was maintenance nightmare but kept to support
the hack in the original film implementation to partially work around the
conservation error. Now that film has been re-implemented in fully mass
conservative form this unphysical non-uniform transformation support is no
longer needed and the coupled patch transformations have been completely
refactored to be simpler and more rational with single values for the
transformation properties and boolians to indicate which transformations are
needed.
All of the film transport equations are now formulated with respect to the film
volume fraction in the region cell layer rather than the film thickness which
ensures mass conservation of the film even as it flows over curved surfaces and
around corners. (In the previous formulation the conservation error could be as
large as 15% for a film flowing around a corner.)
The film Courant number is now formulated in terms of the film cell volumetric
flux which avoids the stabilised division by the film thickness and provides a
more reliable estimate for time-step evaluation. As a consequence the film
solution is substantially more robust even though the time-step is now
significantly higher. For film flow dominated problem the simulations now runs
10-30x faster.
The inconsistent extended PISO controls have been replaced by the standard
PIMPLE control system used in all other flow solvers, providing consistent
input, a flexible structure and easier maintenance.
The momentum corrector has been re-formulated to be consistent with the momentum
predictor so the optional PIMPLE outer-corrector loop converges which it did not
previously.
nonuniformTransformCyclic patches and corresponding fields are no longer needed
and have been removed which paves the way for a future rationalisation of the
handling of cyclic transformations in OpenFOAM to improve robustness, usability
and maintainability.
Film sources have been simplified to avoid the need for fictitious boundary
conditions, in particular mappedFixedPushedInternalValueFvPatchField which has
been removed.
Film variables previously appended with an "f" for "film" rather than "face"
have been renamed without the unnecessary and confusing "f" as they are
localised to the film region and hence already directly associated with it.
All film tutorials have been updated to test and demonstrate the developments
and improvements listed above.
Henry G. Weller
CFD Direct Ltd.
and copy assignment operator for classes with a copy constructor
This is often described as the rule of 3 (or rule of 5 in C++11 if move
constructors and assignment operators are also defined) and makes good sense in
ensuring consistency. For classes in which the default bitwise copy constructor
or assignment operator are appropriate these are now specified explicitly using
the "= default" keyword if the other is explicitly defined fulfilling the rule
of 3 without the need to define the body of the function.
Currently these deleted function declarations are still in the private section
of the class declarations but will be moved by hand to the public section over
time as this is too complex to automate reliably.
Replaced all uses of complex Xfer class with C++11 "move" constructors and
assignment operators. Removed the now redundant Xfer class.
This substantial changes improves consistency between OpenFOAM and the C++11 STL
containers and algorithms, reduces memory allocation and copy overhead when
returning containers from functions and simplifies maintenance of the core
libraries significantly.
The base dynamicFvMesh now reads and stores the dynamicMeshDict and motion
solver receive it as a constructor argument.
Also rationalised the motionSolver diffusivity classes in which storing the
faceDiffusivity field provided no advantage; now it is created and returned on
demand.
The sub-loops of the solution control are now named more consistently,
with ambiguously named methods such as finalIter replaced with ones
like finalPimpleIter, so that it is clear which loop they represent.
In addition, the final logic has been improved so that it restores state
after a sub-iteration, and so that sub-iterations can be used on their
own without an outer iteration in effect. Previously, if the
non-orthogonal loop were used outside of a pimple/piso iteration, the
final iteration would not execute with final settings.
Registration occurs when the temporary field is transferred to a non-temporary
field via a constructor or if explicitly transferred to the database via the
regIOobject "store" methods.
for
db/functionObjects/timeControl/timeControl.H: timeControls
primitives/bools/Switch/Switch.H: class switchType
primitives/strings/fileName/fileName.H: fileType
primitives/strings/wordRe/wordRe.H: compOption
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.
This method waits until all the threads have completed IO operations and
then clears any cached information about the files on disk. This
replaces the deactivation of threading by means of zeroing the buffer
size when writing and reading of a file happen in sequence. It also
allows paraFoam to update the list of available times.
Patch contributed by Mattijs Janssens
Resolves bug report https://bugs.openfoam.org/view.php?id=2962
This is faster than the library functionality that it replaces, as it
allows the compiler to do inlining. It also does not utilise any static
state so generators do not interfere with each other. It is also faster
than the the array lookup in cachedRandom. The cachedRandom class
therefore offers no advantage over Random and has been removed.
Tree bound boxes are expanded asymmetrically to reduce the liklihood of
octree faces aliging with mesh faces and edges. The asymmetry is now
generated using hard-coded irrational numbers, rather than using a
random generator.
The asymmetry was effectively already hard coded. The random numbers are
only pseudo random, so the same numbers were being applied to the bound
boxes every time. This change simply removes the overhead of creating
the generator, and also gets rid of some duplicated code.
