Encyclopaedia Index
What CFD can and cannot do
Contents
- General considerations about the capabilities of CFD
- Why the subject is important
- What affects the degree of difficulty
- How to get value from CFD despite the difficulty
- What is easy
- What is difficult
- Transition from laminar to turbulent flow
- Chemical reaction in turbulent fluids
- Fluid-structure interactions
- Erosion
- Atomisation
- What is impossible
All who are interested in using CFD need to understand that, although all fluid-flow
phenomena are in principle amenable to simulation by CFD techniques , it is only
the minority of those which arise in practice which are easy to simulate; most are
rather difficult; and many will have to be regarded for many years as impossible.
The sources and natures of the difficulties need to be recognised; and it is useful to
distinguish those difficulties which can be surmounted by spending more money (for example
on more-powerful software or hardware) and those which derive from inadequacies of
scientific knowledge.
Most important of all is to recognise that, even in cases of great difficulty, much
value can be obtained from CFD simulations, if the problem statement is judiciously
simplified, and the demand for accuracy is reduced.
The features which affect the relative ease or difficulty are:
- Dimensionality in time. Problems can be:
- zero-dimensional, i.e. steady (not varying with time), or
- one-dimensional, i.e. transient (varying with time).
- Dimensionality in space. Problems can be:
- zero-dimensional, i.e. without variations with position, as in a well-stirred cooking
pot;
- one-dimensional, i.e. with significant variations in one direction only, as in a long
insulated pipe;
- two-dimensional, i.e. with significant variations in two directions, as when the pipe is
neither long nor insulated, so that both radial and longitudinal temperature gradients are
of importance;
- three-dimensional, i.e. with significant variations in every direction, as in a
living-room with a heating stove in one corner.
- Complexity of the geometry of the fluid-solid interfaces. Thus, simulation of the
flow of air in a room is:
- easiest if the room is empty,
- harder if it contains a large amount of furniture,
- harder still if human beings or other curved-surface items are present,
- almost impossible if the people move about .
- Complexity of the flow-influencing boundary conditions. Thus, simulation of the
flow of air in a room is:
- easiest if the temperatures of walls and solid objects are prescribed and uniform, and
if doors and windows are shut and leakproof;
- harder if the prescribed temperatures are not uniform and if in-flow rates of air are
prescribed at some apertures and pressures at others; and
- harder still (but not impossible) if neither temperatures nor flow rates nor pressures
are prescribed, but all have to be deduced from information about the external wind flows
and solar radiation.
- Complexity of the physical processes which have to be taken into account. Thus
the simulation of the flow phenomena in the room is:
- easiest if the temperatures are uniform, and the air is set in motion by a single fan;
- harder when the temperature is not uniform, so that buoyancy effects (i.e. free
convection) influence the air motion;
- even harder when the temperatures are sufficiently high for radiative heat transfer to
play a significant role;
- harder still when a fire is ignited in a waste-paper basket, so that heat and smoke are
generated at rates which must be computed from empirical formulae;
- well-nigh impossible when the conflagration has engulfed all the furniture and other
combustible material in the room.
- The number of fluid phases which participate. Thus the flow in the room is:
- single-phase when only air (possibly mixed with smoke) is present;
- two-phase when the fire-fighters arrive, and start pumping in water; and
- multi-phase when sparks start to fly.
In summary, the degree of difficulty of a flow-simulation problem is strongly dependent
on its magnitude, ie on the number of:
- dimensions,
- objects,
- boundary conditions,
- participating processes, and
- physical phases.
The greater the magnitude of the problem, the greater must the size of the computer,
and the time for which it must run; and computers possessing the necessary power and speed
for even moderately large problems are simply not available.
However, there is another equally insuperable source of difficulty: scientific
knowledge is also lacking or inadequate for many of the processes and materials to
which one might wish to apply CFD.
Thus the fire in the waste-paper basket cannot truly be simulatd by CFD, even if
simplified in respect of geometry, because the chemistry and physics of the combustion of
paper have not yet been reduced to quantitative scientific order.
Faced with the difficulties just enumerated, some would-be CFD users may decide to
proceed no farther, preferring either to obtain the information which they require by
making physical experiments, or to do without it.
Most, however, wisely decide to lower their expectations, and to learn what they can
from the simulation of a simpler situation which still, they may believe, contains the
essence of the more complex one.
Thus, they may reduce the size of the problem by:-
- reducing the number of time dimensions, confining attention to a steady-state flow
rather than simulating the transient phenomena which precedes it;
- reducing the number of space dimensions, presuming that variations of conditions in one
of the directions are much less important than those in the others;
- reducing the number of objects to be considered, for example omitting all those below a
certain size, or representing them by way of a
"distributed-resistance"coefficient;
- reducing the number of physical processes to be considered, for example neglectiing heat
transfer by radiation in comparison with that by conduction and convection;
- reducing the number of phases under consideration, for example neglecting the fact that
the steam and the water in a boiler have different velocities at each location, and
treating the steam-water mixture as a single fluid.
In addition to these quantitative reductions, the prudent user of CFD will also look
for qualitative ones, determining to represent some physical phenomena in an
approximate manner, even when science knows better.
For example, in a combustion simulation, he or she may choose to use the
"mixed-is-burned" presumption, thus dispensing with the necessity to simulate in
detail the many chemical reactions which are known to take place.
Such simplifications abound in CFD and are necessarily much employed. So important is
this that the selector of a CFD software package should ask not only what complex problems
it can solve exactly but how many and which affordable approximations it possesses.
As will be seen, PHOENICS possesses quite a large number of these, for example: LVEL, IMMERSOL and the multi-fluid model of turbulence .
The special features of transition
What PHOENICS can do now
What PHOENICS could do
The special features of turbulent chemical reaction
The special difficulty about predicting correctly the rates of chemical reaction which
take place in turbulent fluids is discussed in several places within the PHOENICS
documentation, for example:
What PHOENICS can do now
What PHOENICS could do
The special features of Fluid-structure interactions
What PHOENICS can do now
What PHOENICS could do
The special features of erosion
What PHOENICS can do now
What PHOENICS could do
The special features of Atomisation
What PHOENICS can do now
What PHOENICS could do