and beyond by
Plans for the future are also outlined.
However. it was hard to ensure that these programs would continue to work after delivery.
The solution was this: create one general-purpose program with:
This is why PHOENICS was created, and launched commercially,
in October 1981.
.... next
.... back or.... contents
Even the first PHOENICS possessed many of the features which are nowadays expected in a CFD code, being able to solve problems:
Moreover it already allowed users to add their own Fortran coding.
The latter facility is still much used, as witness presentations
at the
2002 International PHOENICS User Conference.
.... next
.... back or.... contents
Users were required to place FORTRAN statements in the SATELLITE input module, and then to compile and link.
The notion of a code series was however already present, as witness this picture from one of the earliest documents.
It was then supposed that:
This was made possible by enabling the computational grid to expand and contract in accordance with the motion of the piston.
This is now a common feature in CFD codes.
.... next
.... back or.... contents
This is the 'parabolic option' which permits accurate fine-grid solutions on small computers, when the main flow is uni-directional.
An animated plot from an early calculation is shown here.
It concerns the movement of smoke along a tunnel.
The grid had one million nodes; yet the computer was a 386 laptop.
This option is particularly useful for simulating boundary layers,
jets, wakes, duct flows, rocket exhausts, etc.
rocket-exhaust-plume
simulations, and is still in daily use.
.... next
.... back or.... contents
Here is shown one of the first examples of their use, for simulating the flow in a rotating centrifugal impeller.
The use of BFCs made it desirable that PHOENICS should possess its own graphical display package. This was created, and called PHOTON.
The latest (2002) version of PHOTON has been used to display the
grid just shown.
.... next
.... back or.... contents
This PHOENICS Input Language (abbreviated as PIL) could be used either:
Although PHOENICS has subsequently also developed menu-style input procedures, the ability to use PIL, interactively or not, has been steadily preserved by CHAM.
The reason is that menu-writers can never imagine all the inputs which a user may wish to make.
A menu is like a 'phrase-book' which assists those who are learning a language. It encapsulates what users have wished to say in the past.
More advanced users always wish to write their own never-before-spoken phrases. PIL therefore opens the way to the future.
.... next
.... back or.... contents
This library has since grown so large that finding desired items requires a 'search engine' which enables users to command:
"Find for me all entries which involve this, that and the other".
This is being shipped with PHOENICS-3.5 .
The current contents of the library can be seen
here.
.... next
.... back or.... contents
In the course of this work, the IPSA algorithm was invented; and it later found use in the simulation of:
CHAM's URSULA code, developed for the US Electric Power Research
Institute, became the industry-standard steam-generator simulator;
and PHOENICS later became the main steam-generator code used by the UK's
National Nuclear
Corporation.
.... next
.... back or.... contents
This can explain what to conventional turbulence models is incomprehensible, namely 'unmixing', as described here.
Other successes of the two-fluid model are simulating:
Perhaps because of inability to perform the calculations, competitors have not copied these developments ... yet.
Nevertheless they still stand as unanswered challenges.
.... next
.... back or.... contents
Here is shown an early example, an expanding and contracting lung.
The colour contours represent the oxygen content of the air in the lung.
Here is a somewhat later example, in which PLANT (see below) is used to cause a pipe wall to become crinkled, so inducing
a flow.
.... next
.... back or.... contents
One of the more successful is provided only by PHOENICS. It is CLDA, the Conservative Low-Dispersion Algorithm. When flow is oblique to the grid, the default "upwind-difference scheme" smears discontinuities, even with a fine (80 * 80) grid.
CLDA, on the other hand, induces no smearing, even for a coarser (40 * 40) grid
CLDA is another not-yet-copied device, probably because CHAM has not publicised it sufficiently. There are many such 'buried treasures' in PHOENICS.
A further development of CLDA, namely
"
X-cell has still not been 'officially' attached to PHOENICS
.... next
.... back or.... contents
For this to be possible, it was necessary to extend the simulation domain to the interior of the immersed solid objects, which, until that time, had been treated simply as 'blockages'.
Two pictures from 1990 are shown here (velocity vectors) and here (temperature contours).
This is an application sector in which CHAM's lead has
been followed by other vendors.
.... next
.... back or.... contents
To meet their needs, CHAM did two things, namely:
The enriched PIL remains in use today; but the menu system was
superseded in 1995 by a
more flexible and modern-looking one.
.... next
.... back or.... contents
Only two of the less-conventional ones will be mentioned namely:
What EXPERT can achieve is shown by comparison of two 'monitor' plots for a square-cavity flow first without and then with EXPERT active.
Evidently EXPERT obtained convergence, and stopped the run, after
186 sweeps; without it, convergence had still not been
achieved in 10000 sweeps.
.... next
.... back or.... contents
The vectors in the next picture show the right-to-left velocity of the coolant, and the left-to-right displacements of the elements of the block caused by thermal expansion.
What makes this achievement possible is that the velocities and the displacements obey equations which are very similar, as is explained at length in a recent lecture.
This is another unique feature of PHOENICS which competitors have
not yet copied. But its advantages are so great that it surely
will be.
.... next
.... back or.... contents
For the first, CHAM developed "Parallel PHOENICS", using the technique of "domain-decomposition"; for the second, CHAM distributed an early version of PHOENICS free of charge, and with rights to copy.
The costs of hardware and software have both dropped dramatically since the developments, and emphasis has shifted towards clusters of linked PCs.
CHAM is still active in Parallel-PHOENICS development; but the
shareware release has not been updated.
.... next
.... back or.... contents
The distances are needed for many turbulence models; and, without LTLS, they can be expensive to calculate for complex geometries.
LTLS valid for any kind of grid as is shown by the following:
pictures of:
the grid,
the distance from the wall, and
the distance between the walls.
.... next
.... back or.... contents
A comparison between LVEL and other low-Reynolds-number turbulence models is shown in an ASME paper by Agonafer et al.
The LTLS and LVEL innovations may, by now, have been copied by competitors.
.... next
.... back or.... contents
Later, CHAM took over the whole development, so that the VR Front End
became the Environment/editor/viewer package that is used today,
seen here.
.... next
.... back or.... contents
Conventional turbulence models, such as k-epsilon, cannot predict PDFs; and, until MFM was invented, the only way of doing so was Pope's Monte-Carlo-based PDF-transport (PDFT) method, which is too expensive for routine use.
MFM is less expensive and more flexible than PDFT; and it offers a practical means of predicting the way in which turbulence affects chemical reactions.
Extensive documentation is provided with the PHOENICS package.
Other codes use the 'presumed-PDF' method, i.e. guesswork; but why
guess, if one calculate?
.... next
.... back or.... contents
The project showed that this was practicable, but, in those days, slow.
MICA was followed by a second EEC-supported project, ADELFI, which aimed to require the user to possess only browser software; and this led to the creation of the Simuserve operation, which provides remote computing as a commercial service.
The technology changes so rapidly that this service has not yet
settled into routine operation; but it will surely do so soon.
.... next
.... back or.... contents
This model embodies both the mathematics and the spirit of the LTLS wall-distance model; and like it, it gives exactly correct prediction in simple geometries and plausible predictions in complex ones.
Here are shown, for a two-dimensional box containing some radiating rods, the gas-temperature distribution and the (different) radiation-temperature distribution.
The calculated radiation fluxes to the wall are in good
agreement with exact solutions of the equations, obtained by more
expensive methods.
.... next
.... back or.... contents
PARSOL allows the flow around faceted objects to be rather
smoothly represented, as is shown by the following
'walking-man' picture,
.... next
.... back or.... contents
An example is provided by the three-part-airfoil simulation.
Since the multi-block grid is set up with a few mouse-clicks, and
the results are as accurate as those produced by difficult-to-create
BFC grids, PARSOL with FGEM has much to commend it.
.... next
.... back or.... contents
Its name was PLANT; and has proved to be an extremely powerful means of extending the simulation capabilities of PHOENICS.
Here for example is a picture of the flow induced by a paddle in a closed vessel.
What the user had to do can be seen by inspecting a fragment of the relevant Q1 file. The syntax is rather simple; but of course it has still to be learned.
PLANT then writes the corresponding Fortran, compiles, relinks,
and then performs the computation.
.... next
.... back or.... contents
Its first major use was for the creation of the SAFIR, blast-furnace model, which is characterised by having four phases flowing in the same space, namely:
MUSES accomplishes this by covering the same space twice, once for phases 1 and 2, and then for phases 3 and 4.
All relationships for transfer of heat, mass and momentum between
the phases, and for chemical reactions were introduced by way of PLANT.
.... next
.... back or.... contents
MUSES can be further illustrated by reference to a Moscow Conference paper about fuel-cell stacks.
The flows in a 10-cell stack are illustrated here.
They all pass through the same space; and they interact with each other by heat and mass transfer.
Muses enables PHOENICS to simulate the processes by covering the same volume four times, each with a different part of the total grid.
Here is shown a set of contours computed for a single plane in all four parts,
Large amounts of Fortran coding were needed, but they were all created by PLANT. Here is an example of a simple user statement which PLANT turns into Fortran.
The National Research Council of Canada intends the following
further steps.
.... next
.... back or.... contents
This ShapeMaker package differed from the pre-existing VRGEOM package in having its own immediate-display capability,
ShapeMaker can be run from
within the VR-environment as is seen here.
.... next
.... back or.... contents
Its main merit is that, for many problems, it produces much more rapid convergence than does the built-in SIMPLEST solver, as may be seen from the following comparison.
MIGAL can not yet handle two-phase flows or multi-block grids; and it has not yet been extended to Parallel PHOENICS.
For other circumstances, however, it offers great reduction in
computing time.
.... next
.... back or.... contents
In-Form allows users to place in the input file lines like this:
(initial of temp is XG*XG + YG*YG)
or
(source of u1 at patch is 1.e-1 * (v1 + 4.0))
or
(property enul is 1.0 / log (tem1))
wherein what follows the is can be an expression of (almost)
unlimited complexity, qualified by a great variety of conditions.
.... next
.... back or.... contents
All that the user has to learn therefore are the simple rules for
expressing his wishes in the
(variable such-and-such is formula) mode.
All this is explained briefly in an Introductory Lecture on In-Form; and, with full details in the PHOENICS Encyclopaedia article.
So far as is known to the author, nothing having the power of In-Form,
or of PLANT for that matter, is possessed by competing codes. PHOENICS
appears to be two stages ahead.
.... next
.... back or.... contents
Some animated examples can be seen here.
Profiting from its experience with PARSOL, CHAM has adopted the policy of not modifying the grid, but causing the body to move through it, transmitting momentum to the fluid as it does so.
This method appears to be unique to PHOENICS at present; for other CFD codes require the grid to be distorted as the body moves. MOFOR does however allow for motion, with arbitrary acceleration, of the whole grid, with interesting effects.
Full information about the current state of MOFOR is contained in
a special lecture,
.... next
.... back or.... contents
Recently it has been enabled also to operate in 'inverse' mode, i.e. to predict the most probable inputs for prescribed outcomes.
COSP (originally Constant-Optimising Software Package) is the name given to the feature; for it was first used for determining what constants in empirically-based formulae for tar retention, best fitted cigarette-smoking data.
The complete story is told in a Moscow Conference lecture by JZ Wu,
This was not the only reference to inverse-problem solving at the Moscow Conference. An alternative approach was described by Norberto Fueyo, of The University of Zaragoza.
COSP and Fueyo's algorithm are not the same, the one being deterministic, the other stochastic, as shown by a final slide of his presentation.
Which is the better approach remains to be seen. In the mean
time, PHOENICS users have a choice.
.... next
.... back or.... contents
This is AC3D, which is described generally here.
and in relation to PHOENICS here.
It seems likely to prove a popular addition.
AC3D can be run from within the VR-environment.
.... next
.... back or.... contents
First back: it has been noted that not every innovation has been copied. Why is this?
Possible answers include: the innovation is
A historical note:
Almost all the 1960-70 Imperial
College turbulence-model research used the parabolic
computer program, GENMIX.
However, the merits are still to be incontrovertibly proved.
CHAM therefore intends to continue promoting its not-yet-fashionable
policies, including those relating to PARSOL, IMMERSOL, In-Form
and MOFOR.
.... next
.... back or.... contents
The major objectives of immediate work are:
However, the need is less for in-PHOENICS development (although there is still something to be done) and more for exemplification and publication.
For this we must look primarily to our users.
When the above-mentioned MOFOR developments have been made, it
will be time to compute the stresses within the moving object also
.... next
.... back or.... contents
Since, however, funding agencies also follow fashion, the funding will be easier to obtain when PHOENICS users have begun to exploit what has been provided so far, and to publish their results.
CHAM will be happy to collaborate with any such users, not least
by publishing their results on its website.,
.... next
.... back or.... contents
In this it expects to be assisted, as it always has been, by the communications which it receives from its users.
CHAM will continue to try to react promptly and helpfully to all
suggestions, criticisms and requests for help.
.... back or.... contents
