You are on page 1of 14

285

Journal of Non-Newtonian Fluid Mechanics, 4 (1979) 285-298


0 Elsevier Scientific Publishing Company, Amsterdam - Printed in The Netherlands

Review

SCALE-UP PROBLEMS ARISING WITH NON-NEWTONIAN FLUIDS *

GIANNI ASTARITA
Istituto di Principi di Zngegneria Chimica, Universitd di Napoli, Naples (Italy)
(Received May 30, 1978)

Summary

Dimensional analysis, similarity theory and scale-up theory, as applicable


to non-Newtonian fluids, are critically reviewed. It is concluded that, in most
cases, a non-trivial scaling theory is impossible. A suggestion is made concern-
ing the possibility of a non-trivial scaling theory based on the time-tempera-
ture superposition principle.

1. Background

For very many years, engineers have considered the problem of predicting
the behaviour of a large-scale artefact (hereafter called a prototype) from ob-
servations made on a small-scale replica of it (hereafter called a model). Such
predictions require a scale-up theory; Vitruvius [ 1 ] considered the problem
and reached the conclusion that a scale-up theory is impossible. Leonardo
earnestly tried to construct an experimentally based scale-up theory for the
mechanics of solid materials [2] and, in more general terms, affirmed his
belief in the possibility of a scale-up theory: ‘Vitruvius says that small models
are of no avail for ascertaining the behaviour of large ones; and I here propose
to prove that this conclusion is a false one’ [3].
The intuitive idea that some similarity between model and prototype (at
least geometrical similarity) is needed for a scale-up theory to be useful was,
of course, implicit even in Leonardo’s work, but the concept of similarity
was first enunciated explicitly by Newton [4], although in qualitative terms.
A quantitative formulation was given for the first time in 1850 by Stokes [5].
Before Stokes’ work, whatever scale-up rules were used by engineers on an

* Lecture presented at the EUCHEM Conference, Santander, Spain, July 3-7, 1978.
286

intuitive basis, the fundamental rule was that p r o t o t y p e and model had to be
constituted by the same material; in m o d e m terminology, only homologous
models were considered. Stokes' work concerned the mechanics of Newtonian
fluids (the constitutive equation c o m m o n l y known as Newton's law should
in fact be called Stokes' law [6] ), and allowed similarity with different fluids
at the model and p r o t o t y p e levels. Non-homologous models are, in fact, pos-
sible only if a specific constitutive equation is assumed to hold.
The theory of similarity for Newtonian fluids was exploited in the second
half of the nineteenth century, mainly by Reynolds [7,8] and Froude [9].
The theory was limited to Newtonian flow, and was extracted directly from
the equations, with no grounds in dimensional analysis. Today, similarity
theory is usually illustrated on the basis of dimensional analysis, the first con-
crete example of which is possibly the work of Lord Rayleigh in 1899 [10].
The fundamental lay-out of the theory of dimensional analysis was developed
in 1914--15 by Buckingham [11] and Lord Rayleigh [12]. The b o o k of
Bridgman published in 1922 [13] is still possibly the most lucid illustration
of the subject. The algebraic structure of dimensional analysis has been dis-
cussed more recently by Langhaar [14] ; a rigorous and elegant p r o o f of the
fundamental pi theorem has been given b y Brand [15].
Although the very first paper on dimensional analysis [10] was concerned
with the constitutive equation for a gas, the classical theory has essentially
been developed by considering directly the solution of pragmatical problems,
paying no attention to the distinction between the fundamental laws of
physics and the constitutive equations. On closer inspection, it turns o u t that
the classical theory of dimensional analysis (and the scale-up theory based on
it) refers to phenomena taking place in materials described b y exceedingly
simple constitutive equations. The theological behaviour of fluids is regarded
as being identified entirely by only one parameter, namely the viscosity; the
heat conduction by only the conductivity; and so on. The success of the
classical theory is, in fact, due to the rather fortunate fact that a very large
number of pragmatically important phenomena can be analyzed in terms of
exceedingly simple constitutive equations.
The last statement m a y also be considered from the converse point of view.
The following quote from Truesdell and Noll [16] is illuminating: 'Only for
the simplest of (constitutive) theories are there any non-trivial scaling laws ...
The strongest experimental evidence favouring the classical theories comes,
not from the so-called fundamental experiments or the imaginary operational
definitions, b u t from the millions of successful (if rough or even crude) uses
of the scaling laws based u p o n these theories, scaling laws which in fact come
close to characterizing the classical theories'.
The cavalier attitude towards constitutive equations, typical of the classical
theory of dimensional analysis, cannot of course be maintained when materials
characterized by complex constitutive equations -- such as, typically, non-
Newtonian fluids -- need to be considered. Dimensional analysis of non-New-
tonian flow phenomena has been discussed in the literature [17], [18] partic-
287

ularly with regard to viscoelastic fluids. Certain degeneracies of special theories


of shear-thinning fluids have also been considered, though not in great depth
[ 191. The present work is an attempt to place the whole problem of scale-up
laws for non-Newtonian fluids in its proper perspective. Although the paper,is
meant mainly as a review, it is believed that some of the conclusions have a
degree of originality; in particular, the argument about the time-temperature
superposition principle seems not to have appeared in print before.

2. General considerations

Dimensional considerations applied directly to constitutive equations,


rather than to the search for solutions of boundary-value problems, were orig-
inated in the early 1950s by Truesdell [20,21]. If attention is limited to
purely mechanical constitutive equations, that is, to the relationship between
kinematics of motion and stress, all quantities appearing in the constitutive
equations have dimensions expressed in terms of the three fundamental mech-
anical units, mass, length and time. Hence, classical dimensional analysis im-
plies that any constitutive equation can be made dimensionally invariant (or
‘complete’ in the terminology of Bridgman [ 131) by introducing at most three
dimensional parameters, with dimensions of either mass, length or time, or of
any triplet of independent combinations thereof.
Truesdell and No11 [ 161, starting from a different viewpoint, show that, in
the most general possible case, three dimensional parameters (with dimensions
of stress, time and length) are indeed required. At this level of generality, no
non-trivial result can be expected from dimensional analysis.
If, however, the principle of local action is assumed to hold at the first-
order level (that is, in the terminology of Truesdell and Noll, if attention is
restricted to simple materials), only two parameters are required, having
dimensions of stress and time. A simple material is defined as having no char-
acteristic Length. This is, of course, a result in the mathematical theory, but a
simple-minded physical interpretation may be offered as follows. At the first-
order level of the principle of local action, the distance over which two-point
interactions are significant has shrunk to zero; in this sense, rather than saying
that a simple materials has no characteristic length, one would say that its char-
acteristic length is zero. Indeed, by setting the characteristic length equal to
zero, the constitutive equation of a material of any grade reduces to that of
a simple material.
Of course, although only two dimensional parameters are needed, one may
need one, or more (or, in fact, infinitely many) dimensionless parameters to
characterize the constitutive equation of a simple material. This may make
non-homologous models almost impossible, since it may be hard to find two
materials with equal values of all the relevant dimensionless parameters. (In-
deed, this may not be as hopeless as it appears at first sight: for example, the
time-temperature superposition .principle may suggest that the same material
at two different temperatures exhibits the same values of the dimensionless
288

parameters, though n o t the same values of the dimensional ones.}


Of course, all dimensionless parameters appearing in the constitutive equa-
tion will necessarily be the same at the model and prototype levels if the same
material is used. Yet, in this case also, the two dimensional parameters will
have the same value, and therefore only very simple phenomena, where only
one dimensionless group is relevant, can be scaled. The classical example of
phenomena where both the Reynolds and Froude numbers need to be scaled,
which requires non-homologous models, shows directly the limitations imposed
by the restriction of using the same material at both levels.
If a specific constitutive equation is assumed to hold for a class of materials
which contains a finite number of dimensionless parameters in addition to the
two dimensional ones, a scaling theory along classical lines may be constructed.
When compared with the classical scaling theory relative to Newtonian fluids,
the existence of one more dimensional parameter makes the requirements for
similarity more difficult, but n o t impossible, to fulfill. Indeed, for viscoelastic
fluids, several dimensionless groups containing a characteristic time have been
proposed in the literature [ 18--25], (and the references quoted therein), and
the dimensional analysis of viscoelastic flow phenomena has been discussed
from the viewpoint of scaling theory [13--18]. This kind of approach is
reviewed in the fourth section of this paper.
Of course, materials may exist for which only one dimensional parameter
emerges from the constitutive equation. Classical Newtonian fluid theory is
the best known example (for a Newtonian fluid, in addition to the viscosity
which emerges from the constitutive equation, another dimensional param-
eter, that is, the density, enters the equation of m o t i o n but not the constitu-
tive equation; the latter part of this statement is, of course, true for all mate-
rials). The theory of elasticity is also a classical example. (It is interesting to
observe t h a t the two earliest formulations of similarity quoted before refer,
albeit indirectly, to elasticity [2] and Newtonian fluid mechanics [4].)
There is, however, a subtle difference in the reason why only one dimen-
sional parameter is required in the two theories, and this is worth pointing
out. In the case of elasticity, the constitutive assumption is that the stress is
determined uniquely by the strain. Since strain is dimensionless, a n y constitu-
tive equation of elasticity, no matter how non-linear it may be, only requires
one dimensional parameter having dimensions of stress, i.e. an elastic modulus.
In the simple linear case, only one dimensionless parameter (the Poisson ratio)
is required in addition to the elastic modulus. Since different materials with
equal Poisson ratios are easy to find, a scaling theory of the classical type is
obtained. In the non-linear case, more and more dimensionless parameters may
appear with increasing complexity of behaviour, but there is still only one
dimensional parameter, so t h a t a scaling theory of the classical type (with
possibly the restriction of using homologous models) is again obtained.
In the case of Newtonian fluids, the constitutive assumption is that the
stress depends only on the rate of strain, and in fact linearly [6]. Since the
rate of strain is not dimensionless, the fact that only one dimensional param-
289

eter is required hinges crucially on the assumed linearity. In general, a non-


lint ar dependence of stress on the rate of strain would produce two dimen-
sional parameters, so that, from the viewpoint of scaling theory, simple shear-
thinning makes matters as bad as non-linear viscoelasticity. There are, how-
ever, certain subtleties in this regard which will be discussed in the next sec-
tion.
The considerations developed so far refer to purely mechanical theories,
with no consideration of thermal effects. When non-isothermal phenomena
are considered, in principle, a complete thermomechanical constitutive theory
needs to be analyzed. In its most general formulation, such a theory could
not possibly produce non-trivial scaling law. It is interesting to observe that
specific theories produce, when analyzed dimensionally, quite interesting
results concerning the constitutive behaviour per se: in addition to Lord
Rayleigh’s pioneering research on the temperature dependence of viscosity
[lo], Truesdell’s analysis of the Stokesian fluid [30] and of the Maxwellian
fluid [ 311 are interesting examples in which a non-trivial dimensional analysis
of the constitutive equations is obtained for fluids which do not have a natural
time.
Apart from these general considerations, however, in practice the dimen-
sional analysis of non-isothermal flows of non-Newtonian fluids is carried out
by assuming (either implicitly or explicitly) that the differential energy balance
can be written in the same way as for ordinary Newtonian fluids [ 321. This
corresponds either to neglecting any elasticity of the fluid, or to assuming
that whatever elasticity there may be is entirely of entropic origin [ 331. The
latter requirement is fortunately not in conflict with the assumption that the
time-temperature superposition principle holds [ 341. In view of the earlier
comment made on the possibility of a scaling theory which makes crucial use
of the time-temperature superposition principle, the compatibility shown by
Sarti [ 341 is an important one. It is worth pointing out that there is at least
one analysis in the literature [ 351 which does not assume entropic elasticity;
only the dissipative part of the stress power is assumed to produce heating.
Correspondingly, an independent dimensionless group, called the viscoelastic
source number, emerges from the analysis of White and Metzner [35].
If the entropic elasticity assumption is accepted, the analysis of non-iso-
thermal flow problems for non-Newtonian fluids does not, in principle, imply
any additional complexity in comparison with its Newtonian counterpart
over and above those implied by the mechanical constitutive equation. In
practice, however, non-Newtonian fluids are often so highly viscous that the
contribution of the stress power to the .temperature field cannot be. neglected,
and viscous heating by itself may cause tem.perature rises large enough to
change significantly the value of the rheological parameters. From the view-
point of dimensional analysis, this means that the Brinkmann and Griffith
numbers need to be considered [36], in addition to the Prandtl number and
to whatever dimensionless groups the purely mechanical behaviour forces one
to consider. It is worth pointing outthat, in the case of non-Newtonian fluids,
290

these dimensio~ess groups need to be appropriately redefined. The assump-


t i o n o f entropic, elasticity makes t h e concept of the Brinkmann n u m b e r a
legitimate o n e , althot~gh~the stress power is n o t entirely dissipative. If the
fluid is n o n - N e w t o n i a n , o n e w o t d d j n principle need to consider as many dif-
ferent Griffith ~ t u n b e r s a s there are parameters, dimensional and dimension-
less, in the constitutive equation. This is never done, which is equivalent to
assuming that only one ,of~the parameters depends on temperature.
Apart f r o m ~;he diffictflties discussed above (and the conceptual difficulties
involved even i~ t h e very simplest case w i t h a scaliug theory including the
Griffith number ~[3~]) w h e n so. many~ d i ~ n s i o r t t e s s groups need to be con-
sidered in the requirements of similarity, i t is;in practice impossible to con-
struct a non-trivial ~aling theory. Therefoxe, the balance of this paper is only
concerned with scale.up problems related ~to isothermal flow of non-New-
tonian fluids.

3. Isothermal f l o w o f shear- ;thinning fluids

In the formal expansion of simple fluid theory, the appearance of normal


stresses(that is~ in loose terms~ of elasticity)is a second-order eHect, while
shear,thinning is a third-order e f f e c t [38]. Although this may be considered
as a purely theoretical result, the existence o f real :materials exhibiting large
normal stresses in shear flow, but a constant shear viscosity, has been demon-
s t r a t e d [ 39]. The converse experimental evidence (that is, fluids endowed
with appreciably shear-thinning viscosity which exhibit zero normal stresses),
if indeed available at all, is certainly not conclusive. The same conclusion is
reached experimentally when dilute polymer solutions are considered, and
theoretically f r o m an expansion of Oldroyd-type constitutive equations [40].
In spite of this, a large fraction of the engineering analyses of non-New-
tonian flow has been based o n the restriction to those flow problems where
only shear-thinning is important [ 41], possibly on the grounds that some
independent line o f reasoning allows one to disregard the elasticity of the
fluid.
The constitutive equation for a shear-thinning fluid is written in general
in the following form:
T ~- 2 7~(II D ) D , (1)
where • is the stress t e n s o r , D is the symmetric part of the velocity gradient,
the value of the 77(" ) function is the apparent viscosity, and lid is the second
invariant of tensor D. The dimensions of the apparent viscosity are those of
the ordinary Newtonian viscosity, while lid has dimensions of the square of
a frequency (for a sophisticated discussion of the dimensions of tensors, see
Dorgelo and Schooten [42] ). We here take ~ and D to have the dimensions
of their physical components.
In spite of the fact that, within the body of simple fluid theory, eqn. (1) is
so simple as to be degenerate from the viewpoint of dimensional analysis (and
291

hence of scaling theory), it is (almost) as hopeless as the general non-linear


viscoelastic case. In order to make the q(m)function dimensionally homogene-
ous, or complete, two dimensional parameters are in general required, say a
characteristic viscosity q. - which may for example be taken as the zero-shear
viscosity - and a characteristic time A, (we use the suffix r) to remind us that
A7, is to be extracted from an apparent viscosity curve (see the discussion in
[40]). One may thus write:
r = 2 Tjofi(A,211,)D, (2)
where now both the argument and the value of the ?j(-) function are dimen-
sionless. Thus a scaling theory would require the inclusion of both v. and A,
in the parameter listing, and that the dimensionless function fi(-) is the same
at the model and prototype levels. The complete simple fluid theory would
again require consideration of two dimensional parameters and equality of
the dimensionless constitutive functional at the two levels. Since it may be as
hard to find two different materials endowed with the same function !j(-) as
to find two different ones endowed with the same dimensionless functional,
a scaling theory based on eqn. (2).is not, in principle, any more fruitful than
a scaling theory based on general simple fluid theory.
This argument was already hinted at in the preceding section, but there is
a subtlety involved. Given a function of several variables f(xl, . ... x,), we call
it a ‘group’ if it has the following form:
f = KXT’ .. . X;“, (3)
where K, ol, . ... (Y, are constants. A wellknown theorem of dimensional anal-
ysis states that the dimensions of any secondary quantity in a linear system
of measurement is a group of the primary quantities [ 431. An implication of
this theorem has seldom been appreciated: i.e., that the number of dimen-
sional parameters needed in order to make a relationship complete collapses
to less than the maximum if the relationship is a group. In the particular case
of eqn. (l), the number of dirixensional parameters required collapses to one
if the ?j(‘) function is a group, i.e., for power-law fluids (for an early discussion
of this point, see Astarita [ 191; a hint is also made in Astarita and Sarti [ 441).
Before non-Newtonian flow came to the attention of engineers, the only
area where they were confronted with non-trivial constitutive equations was
in chemical kinetics, a field where non-linearity is the rule rather than the
(possibly important) exception. Scale-up theory and dimensional analysis of
chemical reactors has been developed [45,46], yet it appears that the dimen-
sional degeneracy of the power-law type of constitutive equation has never
been identified explicitly, in spite of the fact that the Dahmkahler numbers
can only be obtained with kinetic constitutive equations which are groups
[47]. Even a comparatively recent review seems to avoid the issue [48].
However, the issue is an important one. A specific and particularly simple
problem, namely pressure drop in flow through circular pipes, is particularly
illuminating. In the work of Metzner and Reed [49], the assumption that the
292

fluids considered actually o b e y e d the power-law constitutive equation was


carefully avoided; the parameters ~' (local consistency) and n' (local apparent
power-law index) were used in the correlation. Since 'local' is to be under-
stood as 'at the local value of the shear rate', this is equivalent to the imposi-
tion that the shear rate be the same at the model and p r o t o t y p e levels, a
requirement that would also emerge simply by considering the additional
dimensional parameter A~. After all, A~ is the natural yardstick for making
the shear rate dimensionless. In the debate following Metzner and Reed's
paper, Bird [ 50] pointed o u t the stronger results obtainable from dimensional
analysis b y assuming that the fluids considered do o b e y the power-law con-
stitutive equation, b u t failed to notice that this is related to the dimensional
degeneracy of that particular constitutive equation. Metzner [ 51], t h o u g h
stressing again the strength of a model which did n o t assume a particular con-
stitutive equation, did n o t point o u t that use of the parameters ~' and n'
makes the corresponding scaling theory (if n o t the dimensional analysis)one
step more awkward than that based on the power-law constitutive equation.
The more complete correlation of Dodge and Metzner [52], by requiring the
friction factor to be a function of b o t h the generalized Reynolds number and
the local apparent power-law index, makes the corresponding scaling theory a
trivial one (in the sense that only interpolation, b u t no extrapolation, is in
principle allowed for correlations between dimensionless groups), though this
point does n o t appear to have been realized. To be more explicit, the point is
as follows: if the fluid considered is indeed a power-law fluid, the friction
factor--Reynolds number curve for a given fluid is unique (according to Dodge
and Metzner [52] ) as it is for a Newtonian fluid. Hence the scaling theory is of
the same type, and allows actual scale-up by an arbitrary amount. If, on the
other hand, the fluid is n o t a power-law fluid, the relation is n o t unique, since
the values of ~' and n' to be used at a given value of the Reynolds number
depend on the pipe diameter -- and in principle a scaling theory allowing
actual scale-up outside of the range of experimental data is impossible.
The example discussed in some detail above points out the d i c h o t o m y one
is faced with when dealing with a scaling theory for shear-thinning fluids. On
the one hand, one may choose to regard fluids as n o t being described ade-
quately by the power-law constitutive equation, and therefore require that
any scaling law be applied 'at equal shear-rate range' [53]. This is equiv-
alent to introducing a second dimensional parameter (say A~), with all the
difficulties that this implies, which will be discussed in the next section, On
the other hand, one may choose to lose generality in favour of concreteness,
and assume that the power-law constitutive equation does hold at all shear
rates. Apart from the fact that this is known not to be true for any real mate-
rial, the resulting theory has many awkward aspects. Theoretically, similarity
solutions of boundary-value problems are not obtainable with the same ease
as with Newtonian fluids [ 54], although the dimensional analysis has the same
structure. Furthermore, creeping-flow asymptotes are of dubious value, since
when the Reynolds n u m b e r goes to zero, so also m a y the shear rate, and for
293

the power-law constitutive equation the viscosity goes to infinity. From the
experimental viewpoint, some results are very hard to explain from simple
dimensional analysis considerations (see, e.g., the problem of flow through
porous media [ 551). The explanation requires such a thorough understanding
of the fluid mechanics involved [ 551 that the scaling theory becomes a proce-
dure of dubious usefulness.

4. Isothermal flow of viscoelastic fluids

A scale-up theory for isothermal flow of viscoelastic fluids must necessarily


face the fact that two dimensional rheological parameters are required, say a
viscosity no and a natural time A. The latter needs to be considered if the
elasticity of the fluid is to play any role - if one assumes it does not, one
falls back to a shear-thinning theory. When both no and A are considered,
scaling theory implies (once geometrical similarity is fulfilled) that the follow-
ing two dimensionless groups be the same at the model and prototype levels:

Re = DVP/V~, (4)
We = RV/D, (5)
where Re is the Reynolds number, We is the Weissenberg number, and D is a
characteristic linear dimension. In addition, all the dimensionless rheological
parameters need to be equal.
The latter condition can be fulfilled rigorously only with homologous
models. With such models, a scaling theory based on both Re and We is trivial:
the scale D must be the same at the model and prototype levels. (This can
easily be seen by considering the dimensionless group Re/We = D2p/voA which,
in addition to the linear dimension D, contains only parameters of the fluid,
which are given once and for all in homologous modelling.) Any empirically
based correlation of data which uses the two groups in eqns. (4) and (5), or
any two groups which are a combination thereof (such as typically developed
in the analysis of drag reduction [ 561) only allows one to interpolate within
the size range covered by experiments, but does not allow extrapolation.
In the late 1960’s, Pawlowski [57,58] analyzed these problems and con-
cluded correctly that, no matter what the fluid’s rheological behaviour may
be, for homologous modelling only one more dimensionless group is required
than for Newtonian fluids. Though this may not seem to be much, it is in fact
enough to make a non-trivial scaling theory impossible.
There seem to be only three possible ways for circumventing the conclusion
reached above. These will be discussed in the following sections.

4.1 Creeping flow

The argument here is that, since the non-Newtonian fluids of interest are
generally very viscous, one may wish to restrict attention to creeping-flow
294

problems only, say to the limit R e ~ O. As long as the Reynolds number is


very small, at both the p r o t o t y p e and model levels, it does not matter whether
it has different values.
There is, however, a subtle difficulty involved here. As long as a scaling
theory includes all the relevant dimensionless groups, it does not matter which
particular set of independent groups has been chosen, since the scaling laws
are independent of that choice. For the case in hand, if R e is considered, it
does n o t matter which particular dimensionless group containing A is chosen.
If, however, one considers the asymptotic behaviour of a scaling theory
when one of the groups goes to zero, the choice of the others is n o t irrelevant.
(Suffice it to consider that one possible initial choice is one in which, what-
ever the groups, they are all multiplied by the one that is then allowed to go
to zero!) For the case in hand, different creeping-flow scaling theories can be
constructed according to which group containing A is chosen. For example,
the We group would require, for homologous modelling, that the model veloc-
ity is less than the p r o t o t y p e velocity. But if one chooses the group ApV2D7o,
which is known to be relevant in certain problems where propagation of dis-
continuities is possible [ 59], one would require the model and p r o t o t y p e
velocities to be equal.
Since We and ApV2/~o are equally legitimate choices from the viewpoint
of dimensional analysis, no scaling theory can be obtained without such a
thorough understanding of the fluid mechanics involved as is required to
decide which particular group containing A is pertinent for the problem at
hand. Rather subtle questions of asymptotic behaviour arise [60], b u t the
main issue is that, if one has that much understanding of the fluid mechanics
involved, a scaling theory becomes useless.

4.2 A s s u m p t i o n o f a specific constitutive equation

If a specific constitutive equation, containing few or even no dimensionless


rheological parameters, is assumed to describe the behaviour of the fluid at
both the model and p r o t o t y p e levels, non-homologous modelling becomes
possible. Of course, once the restriction to homologous modelling is relaxed,
a non-trivial scaling theory which includes both R e and We can easily be ob-
tained. The difficulties and pitfalls of this approach have been discussed [17].
In spite of the rather drastic assumptions required for this approach, it is
in a sense popular, since it gives the promise of obtaining at least qualitative
information on the behaviour of, say, a polymer melt in a large-scale process
from experiments made with, say, p o l y m e r solutions on a small scale. Apart
from the fact that this hope seems the only pragmatical justification for the
vast b o d y of academic research on the fluid mechanics of p o l y m e r solutions,
it has in fact been stated explicitly and strongly [61]. The issue is a contrc~
versial one, b u t it is fair to say that this approach has been far less successful
than had been hoped ten or fifteen years ago.
295

4.3 Time-temperature superposition principle

In spite of the large rheological literature which uses, or refers to, the time-
temperature superposition principle, I have been unable to find any hint of
the possibility of constructing a non-trivial scaling theory based on it. The
principle (which has been formulated rigorously only recently [34,62]) can
be interpreted to imply that the same fluid at two different temperatures has
the same values for all dimensionless rheological parameters, though different
values of the dimensional ones. If so, non-homologous modelling is possible,
with non-homologousness being obtained not by changing the fluid, but by
changing the temperature. Both the prototype and the model would need to
be isothermal, but the two temperatures would need to be different.
Consideration of the dimensionless group D2p/qOA, and of the fact that p
is a weak function of temperature, implies that, for non-homologous modelling
based on the time-temperature superposition principle, the product ?&,A
should be less in the model than in the prototype, i.e. the model would need
to be hotter.
This may not be the conclusion preferred by academic research groups,
but there is nothing which can be done about it.
Since, as far as I know, the arguments above are the first to appear in the
literature concerning a scaling theory based on the time-temperature super-
position principle, it is impossible to assess the potentialities of this approach.

5. Conclusions
Apart from a few cases where one is willing to make rather drastic assump-
tions on the appropriate constitutive equation, a non-trivial scaling theory can-
not be constructed for non-Newtonian fluids. Earnest and intelligent attempts
on the part of engineering scientists to build such a theory have, in general,
failed, and extrapolation to a scale substantially different from the one on
which experiments are made is impossible without a thorough understanding
of the fluid mechanics involved, which of course when available makes the
need for a scaling theory obsolete. The scaling theory based on the time-tem-
perature superposition principle has never been tested. Leonardo’s programme
[ 31, successful as it has been with ordinary materials described by very simple
constitutive equations, has failed to contradict, in the general case, Vitruvius’
mistrust [l] in small-scale modelling.
Acknowledgements
I am grateful to my son Mr. Tom Astarita for help in tracing the references
previous to 1845. This work was supported by C.N.R., Grant No. 76.01284.
References

The following list is by no means an exhaustive bibliography of the subject, and is


limited to works actually read while preparing this paper. In particular, many arguments
296

developed refer to a paper which is meant only as an example to which the argument
applies; many, and possibly better, alternative examples could be found in the literature.

1 Vitruvius, De Architectura, ca. 35 B.C.


2 Leonardo Da Vinci, Ms. A, f. 48 v; see also, for deviations related to buckling, Ms. A,
f. 3 v, ca. 1500.
3 Leonardo Da Vinci, Ms. F, back cover. Translation as reported in: R.E. Johnstone and
M.W. Thring, Pilot Plants, Models and Scale-up Methods in Chemical Engineering,
McGraw-Hill, New York, 1957.
4 I. Newton, Philosophiae Naturalis Principia Mathematica, Lib. II, Prop. XXXII, 1687.
5 G.G. Stokes, On the effect of the internal friction of fluids on the motion of pendu-
lums, Trans. Cambr. Phil. Soc., IX, part 2, (1852), Read on Dec. 9, 1850.
6 G.G. Stokes, On the theories on the internal friction of fluids in motion, and of the
equilibrium and motion of elastic solids, Trans. Cambr. Phil. Soc., 8 (1845) 287.
7 O. Reynolds, An experimental investigation of the circumstances which determine
whether the motion of water shall be direct or sinuous, and of the law of resistance in
parallel channels, Trans. R. Soc., 174 (1883) 935.
8 O. Reynolds, On the dynamic theory of incompressible viscous fluids and the deter-
mination of the criterion, Phil. Trans. R. Soc., (1895).
9 W. Froude, The Papers of, The Institution of Naval Architects, London, 1955.
10 J.W. Strutt (Lord Rayleigh), On the viscosity of argon as affected by temperature,
Proc. R. Soc., 66 (1899) 68.
11 E. Buckingham, On physically similar systems: illustrations of the use of dimensional
equations, Phys. Rev., 4 (1914) 347.
12 J.W. Strutt (Lord Rayleigh), The principle of similitude, Nature, 95 (1915) 6 6 , 5 9 1 ,
614.
13 P.W. Bridgman, Dimensional Analysis, Yale University Press, New Haven, 1922.
14 H. Langhaar, A summary of dimensional analysis, J. Franklin Inst., 242 (1946) 459.
15 L. Brand, The pi theorem of dimensional analysis, J. Ratl. Mech. Anal., 1 (1952) 125.
16 C. Truesdell and W. Noll, The non-linear field theories of mechanics, Enc. of Physics,
Vol. III/3, Springer Verlag, Berlin, 1965, pp. 64--65.
17 G. Astarita, Dimensional analysis of viscoelastic flow phenomena, Chem. Engng. Sci.,
29 (1974) 1973.
18 G. Astarita and G. Marrucci, Principles of non-Newtonian Fluid Mechanics, McGraw-
Hill, Maidenhead, 1974, Chap. 7.
19 G. Astarita, Variational principles and entropy production in creeping flow of non-
Newtonian fluids, J. Non-Newtonian Fluid Mech., 2 (1977) 343.
20 C. Truesdell, The Mechanical foundations of elasticity and fluid dynamics, J. Ratl.
Mech. Anal., 1 (1952) 125.
21 C. Truesdell, A program of physical research in classical mechanics, Z. Angew. Math.
Phys., 3 (1952) 79.
22 A.B. Metzner, J.L. White and M.M. Denn, Behavior of viscoelastic materials in short-
time processes, Chem. Eng. Proc., 62(12) (1966) 81.
23 J.L. White and A.B. Metzner, Constitutive equations for viscoelastic fluids with applica-
tion to rapid external flows, Am. Inst. Chem. Engrs. J., 11 (1965) 324.
24 A.B. Metzner, J.L. White and M.M. Denn, Constitutive equations for viscoelastic fluids
for short deformation periods and for rapidly changing flows: significance of the
deborah number, Am. Inst. Chem. Engrs. J., 12 (1966) 863.
25 G. Astarita, G. Marrucci and L. Nicodemo, Gruppi adimensionali caratteristici della
fluodinamica di liquidi viscoelastici, Q. Ing. Chim. Ital., 3 (1967) 97.
26 G. Marrucci and G. Astarita, Significance of the Deborah number in steady flows,
Meccanica, 2 (1967) 141.
27 G: Astarita, Two dimensionless grouPs relevant in the analysis of steady flows of
viscoelastic materials, Ind. Eng. Chem. Fund., 6 (1967) 257.
28 G. Astarita, Spherical gas bubble motion through Maxwell liquids, Ind. Eng. Chem.
Fund., 5 (1966) 548.
29 R.R. Huilgol, On the concept of the Deborah number, Trans. Soc. Rheol., 19 (1975)
297.
297

30 C. Truesdell, A new definition of a fluid. I, The Stokesian fluid, J. Math. Pure Appl.,
29 (1950) 215.
31 C. Truesdell, A new definition of a fluid. II, The Maxwellian fluid, J. Math. Pure
Appl., 30 (1951) 111.
32 G. Astarita and R.A. Mashelkar, Heat and mass transfer in non-Newtonian fluids, The
Chem. Engr., 317 (1977) 100.
33 G. Astarita and G.C. Sarti, Thermodynamics of compressible materials with entropic
elasticity, in Theoretical Rheology, Hutton, Pearson and Walters (Eds.), Appl. Sci.
Publ., Barking, 1975.
34 G.C. Sarti, Thermodynamics of polymeric liquids: simple fluids with entropic elasticity
obeying the time-temperature superposition principle, Rheol. Acta, 15 (1977) 516.
35 J.L. White and A.B. Metzner, Thermodynamics and heat transport considerations for
viscoelastic fluids, Chem. Engng. Sci., 20 (1965) 1055.
36 J.R.A. Pearson, Heat transfer effects in flowing polymers, Prog. Heat Mass Tran.,
5 (1972) 73.
37 J.R. Serrin, Mathematical principles of classical fluid mechanics, Enc. of Physics,
III/l, Springer Verlag, Berlin, 1962, Sec. 66.
38 B.D. Coleman and W. Noll, Normal stresses in second-order viscoelasticity, Trans.
Sot. Rheol., 5 (1961) 41.
39 D.W. Boger, A highly elastic constant-viscosity fluid, J. Non-Newtonian Fluid Mech.,
3 (1977) 87.
40 G. Astarita and A.B. Metzner, Viscoelastic properties of very dilute solutions of
polymeric materials, Rend. Act. Lincei, VIII-46 (1966) 74.
41 A.H.P. Skelland, Non-Newtonian Flow and Heat Transfer, J. Wiley, New York, 1967.
42 H.B. Dorgelo and J.A. Schooten, On unities and dimensions, Proc. Kon. Net. Akad.
Wet., 48 (1946) 124.
43 G. Astarita, Dispense de1 Corso di Principii di Ingegneria Chimica, L’Arte Tipografica,
Napoli, 1962, pp. 43-4.
44 G. Astarita and G.C. Sarti, A class of mathematical models for sorption of swelling
solvents in glassy polymers, Polym. Engng. Sci., 18 (1978) 388.
45 K. Vulis, Issledovanie protsessov goreniya naturalnogs topliva, Knorre, Moscow, 1947.
46 R.C.L. Bosworth; Transport Properties in Applied Chemistry, J. Wiley, New York,
1956.
47 G. Dahmkohler, Der Chemie Ingenieur, Eucken and Jacob (Eds.), Vol. III, part 1,
Edward Publishers, Ann Arbor, 1937.
48 D.F. Boucher and G.E. Alves, Fluid and particle mechanics, in Chemical Engineers
Handbook 5th Edn., Perry and Chilton (Eds.), McGraw-Hill, New York, 1973.
49 A.B. Metzner and J.C. Reed, Flow of non-Newtonian fluids: correlation of the laminar,
transition and turbulent flow regions, Am. Inst. Chem. Engrs. J., 1 (1955) 434.
50 R.B. Bird, Correlation of friction factors in non-Newtonian flow, Am. Inst. Chem.
Engrs. J., 2 (1956) 428.
51 A.B. Metzner, Response (to [50]), Am. Inst. Chem. Engrs. J., 2 (1956) 10(s).
52 D.W. Dodge and A.B. Metzner, Turbulent flow of non-Newtonian systems, Am. Inst.
Chem. Engrs. J., 5 (1959) 189.
53 A.B. Metzner and R.E. Otto, Agitation of non-Newtonian fluids, Am. Inst. Chem.
Engrs. J., 3 (1957) 3.
54 S.Y. Lee and W.F. Ames, Similarity solutions for non-Newtonian fluids, Am. Inst.
Chem. Engrs. J., 12 (1966) 700.
55 R.E. Sheffield and A.B. Metzner, Flow of non-linear fluids through porous media,
Am. Inst. Chem. Engrs. J., 12 (1966) 700.
56 G. Astarita, G. Greco and L. Nicodemo, A phenomenological interpretation and
correlation of drag reduction, Am. Inst. Chem. Engrs. J., 15 (1969) 564.
57 J. Pawlowski, Relationship between process equations for processes in connection
298

with Newtonian and non-Newtonian substances, A m . Inst. Chem. Engrs. J., 15 (1969)
303.
58 J. Pawlowski, Zur Theorie der Ahnlichkeitsubertragung bei Transportvorgangen in
nicht-Newtonschen Stoffen, Rheol. Acta, 6 (1967) 54.
59 C.J.S. Petrie, S o m e Problems in unsteady flow for co-rotational rheological models,
Proc. VIIth Int. Congr. Rheology, Klason and Kubat (Eds.), Chalmers Univ. of Tech-
nology, Goteborg, 1976, p. 446.
60 J.R. Black, M.M. Denn and G.C. Hsiao, Creeping flow of a viscoelasticliquid through
contractions:~a numerical perturbation solution, in Theoretical Rheology, Hutton,
Pearson and Waiters (Eds.), Appl. Sci. Publ., Barking, 1975.
61 D.V. Boger and H.L. Williams, Predicting melt flow instabilityfrom a criterion based on
the behavior of polymer solutions, Polym. Eng. Sci., 12 (1972) 309.
62 M.J. Crochet, A non-isothermal theory of viscoelasticmaterials, in Theoretical
Rheology, Hutton, Pearson and Walters (Eds.), Appl. Sci. Publ., Barking, 1975.

You might also like