Sie sind auf Seite 1von 8

228 tCheologica Aeta, Band 7, He/t 3 (1968)

welche keine M6glichkeit zur B e r e c h n u n g ist, die Art der Berechnung so zu wghlen, dag die
des n e g a t i v e n W e r t e s des Relaxations- physikalischen Voraussetzungen fiir die unbekannte
Funktion in dem eigentlichen Bereehnungsverfahren
s p e k t r u m s ausschMtet. Die g e w o n n e n e n Er- numerisch ausgedriiekt werden.
gebnisse sind in der Tab. 1 u n d 2 angefiihrt
u n d graphisch in der Abb. 1 dargestellt. Summary
A general numerical solution of the integral equa-
Zusammen[a~'sung tion of first kind was discussed in connection with the
restriction arising from the physical meaning of relaxa-
Mittels des oben besehriebenen l~echnungsverfahrens tion spectrum. The Roessler-Twyman approximation
stabilisiert die gewonnene Funktion des Relaxations- method was modified in order to remove the appearance
spektrums ihre Grundform fiir die positiven Werte des in the course of iteration process of negative values of
Spektrmns ebenso sehnell wie die Funktionen, welche relaxation spectrum which have no physical meaning.
mittels der ursprfinglichen Methode Roesler-Twyman(2) The iteration process proposed becomes stabilised for
gewonnen werden. positive values of relaxation spectrum only. An example
Dureh die erw~hnte Modifikation wird die Kon- is given to illustrate the difference between the two
vergenz der ,Reproduktionsf~higkeit" des Relaxa- methods.
tionsspektrums etwas verz6gert, weil der Durehschnitts-
wert des Parameters ,,p" in dem Intervall (c, d) herab- Literatur
gesetzt wird. 1) Hlavd&]c, B. und V. Kotrba, Rheol. Acta 6, 288
Aus der graphischen Darstellung des Iterations- (1967).
verlaufs und aus der Tab. 2 geht hervor, dab man auch 2) Roesler, F. C. und W. A. Twyman, Proe. Phys.
mit der oben beschriebenen Modifikation der Berech- Soc. B 68, 97 (1957).
hung eine sehr gute ,,Reproduzierbarkeit" der Aus- 3) Schwarzl, F. und A. J. Staverman, Appl. Sci.
gangsfunktion des dynamischen Moduls erzielen kann, Res. &, 129 (1953).
unter Wahrung der Unterscheidbarkeit der diskreten 4) Phillips, D. L., J. Assn. Computing Machinery 9,
Relaxationszeiten. Der wesentliehe Vorteil der Modi- 84 (1962).
fikation der numerischen Methode Roesler-Twyman (2) 5) Twomey, S., J. Assn. Computing Machinery 10,
besteht darin, dages zu keiner Verzerrung der physika- 97 (1963).
lischen Bedeutung der Funktion des Relaxations-
spektrums in den hSheren Graden der Iteration kommt. Anschrift des Verfassers :
Aus den Ergebnissen der Arbeit von Twomey (5) und Dr. B. Hlavd~ek,
Phillips (4), sowie aus den Ergebnissen der vorliegenden ttochschule fiir chemischeTechnologic,
Arbeit kann man ersehen, d a g e s ffr numerisehe Instltug fiir technlsehe Physik, Teehnieks 5,
L5sungen der Integralgleiehung der ersten Art wichtig Prag 6 (CSSR)

From the Schools o/Chemical Engineering, University o/Brad]ord, Yorkshire (England)

The Correlation of Non'Newtonian Turbulent Pipe-Flow Data *)

B y J. H a r r i s
With 7 figures in 19 details and 1 table
(Received October 5, 1967)

1. Introduction u p o n how well t h e d a t a for the shear stress -

There is usually little difficulty in cor- shear rate relation are established in t h e
relating n o n - n e w t o n i a n l a m i n a r pipe flow core of the pipe flow because the bulk of the
d a t a where, with some exceptions, the v o l u m e t r i c flow rate occurs in this region.
familiar engineering c h a r t of Zw against The prediction of the pipe velocity profile
8 u m / D will usually p r o d u c e a sufficiently near the axis usually calls for some quite
good correlation of d a t a for different pipe precise viscometric experiments.
diameters for a p a r t i c u l a r material and, The position in the t u r b u l e n t region is n o t
therefore, e n o u g h i n f o r m a t i o n for scale-up quite so simple however, here all fluids,
of d a t a if required. including the newtonian, p r o d u c e a family
B y an i n t e g r a t i o n process the "flow c u r v e " of curves in the t u r b u l e n t region, one for
g e n e r a t e d b y a plot of ~ against y o b t a i n e d each pipe diameter. There exists a v a r i e t y of
from r o t a t i o n a l viseometer tests can be used formulae for effecting the necessary cor-
to o b t a i n the a p p r o x i m a t e shape of the ~w relation in this region, b u t in recent engineer-
against 8 um/D g r a p h in the l a m i n a r region. ing t e x t s special p r o m i n e n c e has been given
The usefulness of this m e t h o d of estimating to the work of Metzner a n d associates, this
l a m i n a r pipe flow d a t a from l a b o r a t o r y work hinges on the so-called "generalized
r o t a t i o n a l viseometer experiments depends R e y n o l d s n u m b e r " which initially was deve-
H a r r i s , T h e Correlation o / N o n - N e w t o n i a n T u r b u l e n t P i p e _Flow D a t a 229

loped for the "power-law" fluid (6) and later B y substituting for Vwt in [2] using [3],
extended (2) to other materials which did
~ U2m-n ' jDn"
not exhibit a linear plot of In Vw against x~ - [4]
In (Sum/D)by using the equation of the 8n ' - I K '
tangent to the laminar curve at the current
value of (8urn/D). This work is sometimes For constant n' and K' the fluid is termed
classed as rigorous, its range of application "power-law".
is considered to include materials which are Eq. [4] is the generalized Reynolds number
both time-independent and homogeneous which has been widely quoted in the en-
(no wall-effect) in the laminar regime and gineering studies of non-newtonian flow.
time-independent, homogeneous, inelastic Since the only requirement in its derivation
materials in the turbulent regime. is that data from different pipe diameters
The alternative work b y Bowen (1) in shall correlate on the same curve of In Tw
correlating non-newtonian turbulent flow is against ln(8um/D) it is thought to be of
unusual in that no knowledge of the laminar universal significance for both laminar and
flow properties of the fluid is required. The turbulent flows with the above slight re-
correlation here is based on an empirical striction.
equation found to give good results in a I t m a y be noted that the relation
survey of the available literature, and un- Cf - - Twl 16
like Metzner's method it has been found 1 ~ Nit [5]
possible to also correlate fluids commonly y~um
accepted as being elastic as well as fluids
(such as solid-liquid suspensions) frequently is automatically obtained (1, 4) in the laminar
assumed to be inelastic. regime using experimental pipe flow data.
In some areas of work involving non- A plot of Cf against NR using data from a
newtonian fluids it is required to know the rotation viscometer experiment is more or
velocity profile in turbulent flow. Such in- less successful depending on how well the
formation is not easy to find directly due shear rate range of interest is covered, b u t
to the experimental and theoretical diffi- it cannot be expected to be as exact as the
culties involved. P a r t of this work is, there- pipe flow data even if factors such as
fore, devoted to estimating the profile shape secondary flow in the viscometer are absent.
starting from Bowen's formula.
The Dodge-Metzner Equations
Review of Metzner's Method Over more or less restricted ranges of flow
rate it is well established that a variety of
A quantity termed the "generalized Rey- non-newtonian fluids can be described suf-
nolds number" occupies a central role in the ficiently well in steady laminar shearing
work of Metzner and associates (2, 6). This motion b y the well-known "power-law",
quantity m a y be developed as follows (4)
T = K ~n. [6]
from the usual Reynolds number definition,
viz : The Dodge-Metzner formulae for the tur-
bulent velocity profile and the friction
NR -- q u~ D []]
factor - Reynolds number relation were
derived explicitly for the "power-law" fluid.
which m a y be written as, In the derivation the turbulent velocity
profile was assumed to have the form,
2v~ q u ~ D (8 u,~/D) [2]
Tw l u ~ / (9, Tw, K , n, y, roughness)

where Twz is the laminar wall shear stress and the velocity defect the form,
associated with the particular value of u 1 - - u = / (Vw, ~, a, y, n) .
For non-newtonian flow one m a y take the Using an anMogous argument to that
equation of the tangent to the laminar flow employed so effectively b y Millikan in the
curve of ha ~wz against In 8 - ~ - and obtain corresponding case of newtonian flow, similar
formulae were found for the variables,

Twl= K' (-~-) n'. [3] = A(n) log [NR -b B (n) [7]

230 Rheologica Acta, Band 7, Heir 3 (1968)

and diminishing values of n also lead to flatter

u+ = C (n) log y+ + D (n) [8] velocity profiles which should lead to a higher
where, rate of radial momentum transport and,
u + = u/u* (u* = l/@w/@)) [9] therefore, a higher friction factor. The two
trends are, therefore, inconsistent.
y+ = yn (u*)2 n ~/K . [10] c) Real non-newtonian materials invariably
Dodge (2) and later T h o m a s (8) evaluated have an out of phase stress-strMn rate
the parameters A and B in the above relationship (fundamental and harmonics) in
equations as shown in table 1. the frequency range of interest (3, 4, 5)
(0-10,000 c.p.s.). This may be due to e i t h e r
Table 1 viscoelasticity or/and thixotropy. Particle
inertia can also contribute to this effect.
A B Fig. 1 shows results obtained during
frequency response experiments on two
4.0 0.4
Dodge nO.Tg nl.~ solid-liquid systems which in an ordinary
steady shearing experiment would appear
4.0 0.4 well suited to "power-law" representation.
I t is clear t h a t these two slurries are highly
unsuited to the "power-law" in unsteady
To attempt to extend the treatment to motion.
include non "power-law" fluids the generaliz- Far from being rigorous as the D o d g e -
ed Reynolds number [4] was substituted into M e t z n e r formulae have been classed hitherto,
[7] by Dodge (2). is apparent t h a t any success obtained with
them is fortuitous.
3. C o m m e n t s on the Dodge-Metzner Formulae The introduction of the tangent form [3]
I t will be noticed t h a t a fundamental and the associated "generalized Reynolds
assumption in the D o d g e - M e t z n e r treatment number" [4] into [7] cannot eliminate the
is t h a t the "power-law" eq. [6] for steady difficulties because [3] is the equation of the
laminar shearing holds even under the equivalent "power-law" at a given value of
conditions prevailing in turbulent motion. (8 urn/D).
Objections to this assumption are summariz-
ed below. 4. B o w e n ' s Method
a) Assuming the shear rate in unsteady All prior techniques for correlating non-
flow can be expressed as the sum of a constant newtonian pipe-flow data have involved
term plus a time-dependent perturbation, creating a Reynolds number from some form
7 = 7. + ~(t) [ill of assumed steady shearing rheological
equation. B o w e n ' s method is unusuM in this
then from [6] respect in t h a t no assumption is made about
T = K (Yo + e(t)) n [12] the steady shearing rheologicM model, only
and the time-average stress is data from turbulent pipe flow experiments
are required.
1 f~(t)dt [133 From a survey of the turbulent pipe flow
data available up to that time, B o w e n found
t h a t data obtained for a particular fluid using
where T is the interval over which the different pipe diameters would correlate
averaging is carried out. using an empirical equation of the form,
For the same form of "power-law" to hold
in both the steady laminar shearing and the D ~vw = L u r e w. [15]
unsteady case requires, Work in this laboratory confirms the
excellence of the method.
Assuming the effectiveness of [151 in the
but this is only true for small e(t) turbulent region and recasting it in the form,
b) Examination of the friction factor - ~w 8w D ~ - w
Reynolds number chart given by Dodge and
M e t z n e r shows t h a t for a given Reynolds and assuming, for the moment, t h a t in the
number diminishing values of n give di- laminar region
minishing values of the friction factor. But -Cw~ = K ' (8 urn~D) n [3]
Harris, The Correlation o/Non-Newtonian Turbulent Pipe Flow Data 231

o o
o o
ID w o
~ +.+.4-

'~ 'o,'o " ~o ~'o

Frequency cps

70 ++ +

6o + ~ 60 o o
o o
50 Fig. 1. Frequency response
of(O) Bentonite, (+) China
clay eonc. 5 gins/25 ml (dis-
} ,~ tilled) water. Shear Amp.
o ,} e. o ; 1~ I} 0.214
Frequency cps Frequency cps

then from [7] that the resulting expression should be

2-n n ( ) l-nl~ independent of pipe diameter gives the well-
NR c}_nI~ - ~ um D 2L known 1/7th power-law velocity profile,
8n-1 K" ~-w D
~ um z [17]
u = ul (y/a) lp 9 [22]
I f relation [7] is independent of tube If the same similarity hypothesis is taken
diameter as it should be, then with B o w e n ' s expression [15] and the mean
velocity is eliminated then,
u = constant • (~-)l/WymD(z~w)m [23]
x--2_ n [18] Assuming this should be independent of
diameter (4) then,
which is not generally true. x
m -- [24]
5. The Velocity Profile
and hence the expression for the turbulent
In newtonian flow a useful approximation velocity profile is,
to the turbulent velocity profile is obtained u = u 1 (y/a) zlw . [25]
from a hypothesis of similarity of velocity
profiles together with the well known em- The dimensionless velocity profile for a
pirical Blasius relation, yielding respect- smooth tube m a y be expressed in functional
ively, form by,
u = u 1 (y/a) m [19] u+ = ](y+) [26]
or, what is equivalent where

u = constant x Urn ( y / a ) m [20]

and y+ is a dimensionless wall distance whose
and explicit form is yet to be determined.
Cf = 0.079 2V~ 1/4. [21] The mean velocity m a y be obtained from
[25] as,
Eliminating the mean velocity between 2Ul
[20] and [21] together with the assumption Um (m + 1) ( m n c 2) [27]

232 Rheologlca Acta, Band 7, Heft 3 (1968)

therefore, recast as,

(m -I- ]) (m + 2)
u : 2 Um (y/a) m [28] 2
urn+ - - (m -b 1) (m -I- 2) E(a+)m [38]
Now Um m a y be eliminated between [15] The friction factor CI is defined by,
and [28] yielding,
-- [39]
[29] CI 1/2 ~ U*m
u 2
Dividing [29] through by ]/(~w]~) gives
the dimensionless velocity profile, q~m+ = V~-1" [40]
Hence combining [40] and [38],
u+ = (m + 1) (m + 2) DmTw~ - 7 ~189 [30]
2 Lllw ~y~,alm 1 r
1/~-~f = (m + 1) (m + 2) E(a+)m. [41]
u+ = E (y+)~ [31]
Now defining a Reynolds number as,
Um D
Y "~'=' NI~ = v' [42]
y+ = [32]

and where v' is given by eq. [24]. This Reynolds

E = 2 m-1 0.0395 -4m (m + 1) (m + 2) [33] number is related to a+ by,
and the kinematic viscosity term is given by,

v' = 4.1 9 lO 5 - - [34]

Making use of [43] to eliminate a+ in [41]
~llm then,
Ct = F iv~-~ [44]
The dimensionless profile [31] is a more where
general form of the equivalent newtonian F =---2.0.03958m/(m + 1) [45]
expression, and
u+ = 8.6 y+11, [35]
i~ . [46]
to which it degenerates on substitution of m+l
the appropriate terms, Eq. [44] is the relation sought. Since no

1 1 further assumptions other than similarity
1~=0.0395~v1#; x=-~-; m:-~-
and also ~ = v'. to

6. The Friction Factor Reynolds Number-

The functional form for the friction factor -
I5 J
Reynolds number relation is,
C! = / ( N I ~ )
(assuming a smooth wall again).
For newtonian fluids the particular em-
pirical Blasius expression is given by [21].

/9 /
On the basis of the dimensionless velocity
profile [31], which gave rise to the kinematic
viscosity expression [34], it is possible to
form an equivalent expression for non-
newtonian flow.
From [31] the m a x i m u m velocity is, ,'o.2 %. ,~' ;o
u~+ = E (a+) m [37] 0.3% r

and the mean velocity [27] m a y be similarly Fig. 2. Results f r o m Dodge's w o r k 0 1/,, d i a . + 1" dia.
Harris, The Correlation o/Non-Newtonian Turbulent Pipe Flow Data 233

of velocity profiles and the suitability of
Bowen's formula are used, it may in the cases
0.2 where Bowen's formula applies, provide an
indirect check on the validity of the velocity
profile similarity hypothesis and hence of
the velocity profile [25].
2- It is emphasised that the friction factor -
Reynolds number relation [44] cannot be any
more useful than the Bowen formula from
.008 which it was derived and is much less con-
venient to use in scaling up data. In addition
.006 it depends upon similar velocity profiles.

6. Experimental Results
.00~ 0 The work of several authors has been
analysed on the basis of eq. [44] with the
.o'o, .0o6 .;oe ,'~-2 ~.2 ;., object of indirectly observing the effective-
c,,;-H ness of the velocity profile similarity concept.
0.7~ CMC Results of these investigations are shown in
Fig. 3. Results from Dodge's work 0 I/,"dia. + 1" dia. figs. 2-7. The work of Dodge (2) and
[] 2" dia. Shorbaghi (7) show best agreement between

.007 .007 /
.oos .005

.003 I I I I I ,003 I I t I
9003 .005 .007 .003 .005 .007
Cf TH. CfTH.
13 % Attusol 15 % Attasol


,/ /

.008 .O08
~-g +


/ t



.006 .008
Cf TH. CfTH.
0,3% Carbopol 0.4%Curbopol
Fig. 4. Results from Dodge's work 9 1/,, dia. + 1" dia. [] 2" dia.
234 Rheologica Acta, Band 7, H e f t 3 ( 1 9 6 8 )

/ /I
~- oo6 .006 oJ+

,003 ~ * i i [ ~ .003 I I i I I i 1
,003 .006 102 903 .006 102
0.44% Am AIg. 0.75% Am AIg.

o~ .ooe

.003 I i i i I I I ,003 i i t i t I I
.003 ,006 102 .003 .006 152
Cf TH CfTH Fig. 5. l~esul~sfrom Thomas's work
0 68 % MajoI 1.4e~ Majol (~ 5 / 1 6 t / dia. + 1" dia. [] 2" dia

.006 o/ ~.006 o/ .006

0 J+++ + o/+ +
,004 .004 ' .004

.002 . , .002 , J i
.002 .o'o4 .0'06 ,b-2 ~o2 .o~4 .o~8 ,'~2 ~176 .oh4 .oh6 10-2
CfTH Cfl"H Cfl"H
O,53 % Cellosize 0.73 % Cellosize 0.7% Ceurlose

Fig. 6. Results from T h o m a s ' s work O 5/16" dia. + 1" dia. [ ] 2 "dia.,

the e x p e r i m e n t a l and theoretical values of 7. Conclusions

the friction factor. T h e full line on each of The main conclusions of this work are;
the curves expresses e q u a l i t y of friction (i) T h a t the Dodge-Metzner formulae for the
factors. friction f a c t o r - Reynolds n u m b e r relation
T h e work of Thomas (8) was less well a n d the t u r b u l e n t velocity profile which were
suited to this investigation because t h e r e was derived for the " p o w e r - l a w " fluid are invalid
usually little overlap between t h e t u r b u l e n t for real materials.
results of In ~w against In 8 um/D for different (ii) T h a t t h e Bowen f o r m u l a for correlating
pipe diameters. B u t a sufficient n u m b e r of t u r b u l e n t pipe flow d a t a gives excellent
e x p e r i m e n t a l points lie a b o u t t h e curve results and is m u c h more c o n v e n i e n t to
Cfexp. = Cstheor. which t a k e n t o g e t h e r with the e m p l o y t h a n the friction f a c t o r - R e y n o l d s
r e m a i n d e r of the curves suggest t h a t t h e n u m b e r formula.
similarity hypothesis e m p l o y e d m a y not be (iii) T h a t a useful a p p r o x i m a t i o n to t h e
too far from reality. real t u r b u l e n t velocity profile m a y possibly
Harris, The Correlation o/Non-Newtonian Turbulent Pipe Flow Data 235

.008 .008 /
~ .006
ooo,. d"

/+J i

.004 .004 -

, i i I J i m

.004 .006 .008 10-2 .004 .006 .008

8,16 % Titanium Dioxide 0.I % CMC



.004 -b
/ n:r
o;8 7o-2 Fig. 7. Results from Shor-
Cf TH bagi's work O i/2" dia.
0.2% CMC + 1" dia. [] 2" dia.
be made using a diameter independence Angestellte Uber]egungen bezfiglich der Dodge-
hypothesis (which produces the 1/7th power- Metzner-Gleichung lassen erkennen, dab diese fiir reale
Flfissigkeiten nicht gilt. Uberdies gibt die empirische
law velocity profile in newtonian flow) and Formel yon Bowen (1), die keine Daten der laminaren
Bowen's correlating formula. Str6mung benStigt, um auf das turbulente Str6mungs-
verhalten zu schlieflen, ausgezeichnete Werte und ist
S ummary daher vorzuziehen.
Bei Anwendung einer Elypothese, wonach eine Un-
Several formulae have been proposed for treating abh~ngigkei~ vom Durchmesser vorhanden ist, nnd
non-newtonian pipe flow friction data, and with one der Bowensehen Formel, ist es mSglich, einen Ausdruek
exception they rely in some way on laminar data for fiir das Turbulenz-Gesehwindigkeits-Profil zu flnden.
predicting turbulent data. Perhaps the most widely Diese Beziehung wird indirekt dutch Vergleieh der
accepted and quoted of these formulae is that due to theoretischen und experimentellen Reibungsfaktoren
Dodge and Metzner (2) which is also stated to be iiberpriift. In einer Anzahl yon F~llen (aber nicht in
rigorous. alien) ergibt der Vergleieh giinstige gesultate.
Some consideration is given to the Dodge.Metzner
equation which seems to show that far from being
rigorous it is not valid for real fluids. Moreover the Re]erences
formula of Bowen (1), which is the exceptional case not 1) Bowen, Le. R., Chem. Engrg. J. June 12, 243
requiring laminar flow data, gives superior correlations (1961); June 20, 127 (1961); July 10, 147 (1961); J u l y
and is, therefore, to be preferred. 24, 143 (1961); August 7, 129 (1961); August 21, 119
Using a diameter independence hypothesis and (]961); September 4, 131 (196t).
Bowen's formula it is possible to find an expression 2) Dodge, D. W., Ph.D. Thesis, Univ. (Delaware
for the turbulent velocity profile. This is checked 1957).
indirectly by comparing the theoretical and experi- 3) Harris, J., Nature 207, 744 (1965).
mental friction factors. In a number of cases (but not 4) Harris, J., Nature 211, 579 (1966).
all) the comparison is favourable. 5) Harris, J., The Chemical Engineer 45, CE. 30
6) Metzner, A. B. and J. C. Reed, A . I . C h . E . J . 43~,
Zusammen[assung 1 (1955).
Mehrere Formeln sind fiir die Behandlung der l%ei- 7) Shorbagi, I. S., ~.Sc. Thesis, Univ. (St. Andrews
bung beiin DurehfluB nicht-Newtonseher Flfissigkeiten 1967).
durch Rohre vorgeschlagen worden. Mit einer Ausnahme 8) Thomas, G., Ph.D. Thesis, Univ. (Wales 1960).
werden dabei Daten aus der laminaren Str5mung dazu
verwendet, um Daten der turbulenten StrSmung vor- Author's address :
aussagen zu kSnnen. Die bekannteste dieser Formeln Dr. J. Harris
ist diejenige von Dodge und Metzner, yon der auch be- Schools of Chemical Engineering University of Bradford
hauptet wird, dab sic streng gilt. Bradford, Yorkshire (England)