Sei sulla pagina 1di 18

TECHNICAL PAPER

Title: Pressure Loss Equations for Laminar and Turbulent Non-Newtonian


Pipe Flow
Authors: R. A. Chilton, Bechtel Water Technology Limited
R. Stainsby, AEA Technology Limited
Date: May 1998
Publication/Venue: Journal of Hydraulic Engineering
Reprinted with permission
BECHTEL WATER TECHNOLOGY LIMITED
PRESSURE LOSS EQUATIONS FOR LAMINAR
AND TURBULENT NON-NEWTONIAN
PIPE FLOW
RA CHILTON
BECHTEL WATER TECHNOLOGY LIMITED, Chadwick House, Warrington Road, Risley,
Warrington WA3 6AE, UK
and
R STAINSBY
AEA TECHNOLOGY LIMITED, Risley, Warrington WA3 6AT, UK
ABSTRACT
The equations which define Newtonian pipe flow are well established and used routinely by engineers
and scientists throughout the world. The same cannot be said for non-Newtonian flows which have a
higher degree of complexity. This paper presents a coherent set of equations for the laminar and
turbulent flow of Herschel-Bulkley fluids. These equations are consistent with those used for
Newtonian fluids and previous work on the behaviour of generalised non-Newtonian fluids. A
numerical model for non-Newtonian flows is discussed and has been compared with experimental
measures from different sources. This model has been used to run a series of simulations to find the
coefficients required for a new turbulent friction factor correlation. Anew Reynolds number has been
defined which represents the conditions in turbulent flows more realistically than the existing
Metzner-Reed Reynolds number.
KEY WORDS
Pressure losses, non-Newtonian, Herschel-Bulkley, yield pseudoplastic, Bingham, power law,
laminar, turbulent, pipe flow
BECHTEL WATER TECHNOLOGY LIMITED
INTRODUCTION
The designer of pipe systems for the transport of Newtonian fluids has access to well established and
accurate methods for predicting pressure drops in both the laminar and turbulent flow regimes. These
methods are based upon well known analytical and experimental work by eminent researchers such
as Poiseuille, Reynolds, Prandtl, Nikuradse, Darcy, Weisbach, Colebrook, White and others. In
contrast, owing to the complexity and diversity of behaviour of non-Newtonian fluids, the
development of 'universal' equations for the prediction of pressure losses in non-Newtonian flows has
proceeded at a much slower pace. There are well established analytical methods for the laminar flow
of some non-Newtonian fluids. Whilst much work has been done on the development of prediction
methods for turbulent flows, the models so far produced cannot be claimed to be universal.
In any non-Newtonian turbulent flow model there are two sub-models, one which describes the
structure of the turbulence and the other which describes the rheology of the fluid. There are two main
difficulties in developing turbulence sub-models for non-Newtonian fluids. The first of these is
knowing whether the structure of the turbulence is modelled correctly and, the second, concerns the
interaction of turbulence with the rheological sub-model. Obtained from a laminar flow, the
rheological sub-model may not be accurate (or even applicable) at the higher strain rates encountered
in turbulent flows. There is, therefore, a problem in comparing the results of predictions with
experimental measurements - are the discrepancies due to inadequacies in the modelling of the
turbulence or in the modelling of the rheology, or perhaps in both? This uncertainty in the adequacy
of each of the sub-models results in models which, whilst appearing to perform well for particular
fluids over a particular range of strain rates, do not apply well to other fluids or over a different range
of strain rates.
The aim of the present work is to develop a consistent approach for predicting laminar and turbulent
pressure losses in both Newtonian and non-Newtonian fluids.
RHEOLOGY
In this paper the Herschel-Bulkley model (1926) is adopted to describe the rheology of the fluid;
t = ty + j(g)
n
where t is the shear stress, g is the shear strain rate, ty is the yield stress and, j and n are parameters
for a particular fluid. Whilst the Herschel-Bulkley model describes the behaviour of yield
pseudoplastic (or generalised Bingham plastic) fluids, Bingham plastic, power law (either
pseudoplastic or dilatant) and Newtonian fluids are all limiting cases.
It is useful to introduce the concept of an 'apparent viscosity'. From the definition of the viscosity
being the ratio of the shear stress to the strain rate in Newtonian flows, the same definition can be used
to define the apparent viscosity in non-Newtonian flows. For the Herschel Bulkley model, the
apparent viscosity is given by:
(1)
(2) j(g)
(n-1)
ma
g
t
g
ty
= = +
n = =
The Herschel-Bulkley model has been applied to a wide variety of fluids including sewage sludges,
kaolin slurries and mine tailings. However, it is known that the accuracy of the model deteriorates at
high strain rates which may or may not be significant depending upon the application. Kozicki and
Kuang (1993) confirmed theoretically what had often been observed experimentally, that all real
fluids, with or without a yield stress, have an upper limiting value of viscosity at the limit of infinite
strain rate. They also showed that at the limit of zero strain, fluids which do not exhibit a yield stress
have an apparent viscosity which is finite, and that fluids which exhibit a yield stress have an apparent
viscosity which is infinite. Because the upper limiting viscosity is finite, the shear stress/strain rate
relationship becomes linear as the strain rate increases or, in essence, the fluid becomes more
Newtonian. This return to Newtonian behaviour gives rise to the linear part of the stress/strain rate
curve being sometimes known as the "2nd Newtonian region". Therefore, the Herschel-Bulkley
model reproduces the correct behaviour only at low strain rates, because at high strain rates the
predicted viscosity tends to zero, and not to a finite value.
There are many other rheological models which have been developed for other types of non-
Newtonian fluids such as: Sisko (1958), Meter and Bird (1964) and Cross (1965). All of these are
limited in one way or another as to which fluids and shear strain rates they are applicable to, and their
ease of use.
Rabinowitsch (1929) and Mooney (1931) derived a pair of equations which, for any time-independent
fluid, relate the shear strain rate and the shear stress measured in a rheometer, to the mean pipe
velocity and pressure loss for laminar flow:
and
where gw is the shear strain rate at the wall, V is the mean pipe velocity, D is the pipe diameter, n' is
the apparent flow behaviour index, tw is the wall shear stress, DP is the pressure loss and L is the pipe
length. Their method is rigorous and applies to any time-independent purely viscous non-Newtonian
fluid. It does, however, require a full set of accurate rheometry data from which n' can be determined
graphically or by numerical differentiation. Alternatively, n' can be determined from the analytical
differentiation of a curve-fit to the rheological data. Whilst this approach appears attractive, Lazarus
and Slatter (1988) have shown that in practice, it is very difficult to obtain reliable values for n' from
actual rheometry data.
BECHTEL WATER TECHNOLOGY LIMITED
(3)
(4)
3n- 1
4n
-gW =
8V
D
d1n(tW)
d1n(8V/D)
d1n(DDP/4L)
d1n(8V/D)
LAMINAR FLOW
Laminar flow solutions for non-Newtonian fluids are derived in a manner analogous to that leading to
the Hagen-Poiseuille equation for Newtonian flows. The starting point is the rheological model of the
fluid. An expression for the velocity profile is obtained by re-arranging the rheological model and
integrating the strain rate over the pipe radius. Incorporation of the boundary conditions and a further
integration of the velocity profile yields the mean velocity as a function of the wall shear stress.
Derivation of a 'closed form' analytical solution depends upon the choice of rheological model or,
more precisely, upon whether the model equation can be integrated directly. The Herschel-Bulkley
model was chosen for this work because it can be integrated over the cross-section of the pipe to give
a closed form analytical solution for laminar flow. Chilton, Stainsby and Thompson (1996) present
such an integration which is a simpler version of that first presented by Govier and Aziz (1972) and
later, in more detail, by Lazarus and Slatter (1988). The resulting equation for the pressure loss is:
where X is given by:-
and
1
1 - aX - bX
2
- cX
3
BECHTEL WATER TECHNOLOGY LIMITED
3n + 1
4n
=
n n n
8V
D
1
1X
4j
D
DP
L
(5)
(5a)
(5b)
(5c)
(5d)
4Lty
DDP
X = =
ty
tW
1
(2n + 1)
a =
2n
(n + 1)(2n + 1)
b =
2n
2
(n + 1)(2n + 1)
c =
BECHTEL WATER TECHNOLOGY LIMITED
Given the mean velocity, Equation (5) can be solved for the pressure loss by a simple iterative
technique. Alternatively, Equation (5) can be re-arranged to give a direct solution for the mean
velocity for a given pressure loss. Because some of the simpler non-Newtonian models are special
cases of the Herschel-Bulkley model, their solutions can be obtained from simplifications of Equation
(5), viz:-
Bingham Plastic - Buckingham (1921) (n=1, j= coefficient of rigidity, h)
Power Law - Govier and Aziz (1972) (ty=0, j= power law viscosity, mp)
Newtonian - Hagen-Poiseuille equation (ty=0, n=1, j= viscosity, m)
By extending the work of Rabinowitsch and Mooney, Metzner and Reed (1955) derived a generalised
analytical solution for any time-independent fluid:
Metzner (1957) demonstrated the validity of this approach by deriving Equations (6) to (8) from
Equation (9). Equation (5) can be also be derived from Metzner and Reed's generalised solution if the
Herschel-Bulkley model is used to define K' and n'.
1
1 - 4X/3 + X
3
/3
=
8V
D
4h
D
DP
L
3n + 1
4n
=
8V
D
4m p
D
DP
L
n n
n
32mV
D
2
= =
8V
D
4m
D
DP
L
=
8V
D
4
D
DP
L
K
(6)
(7)
(8)
(9)
GENERALISED REYNOLDS NUMBERS
The Darcy-Weisbach equation is commonly used for relating the frictional pressure loss to the
velocity head in laminar and turbulent Newtonian pipe flow:
where f is defined as the Fanning friction factor with 4f being equal to l, the Darcy-Weisbach friction
factor according to the European convention, and r is the density.
For laminar Newtonian flow, Equations (8) and (10) can be equated giving the well known definition
for the friction factor:
Metzner and Reed applied this same reasoning to non-Newtonian fluids by equating their generalised
solution for pressure loss, Equation (9) with the Darcy-Weisbach equation, Equation (10) defining a
non-Newtonian friction factor as:
where ReMR is a generalised Reynolds number and is given by:
Equation (13) represents a Reynolds number which is applicable, for laminar flow, to any time-
independent fluid.
16
rVD/m
16
Re
BECHTEL WATER TECHNOLOGY LIMITED
=
=
4rf V
2
2D
DP
L
(10)
(11)
(12)
(13)
f =
f = 16/ReMR
rVD
ReMR =
K
8V
D
(n-1)
The authors have obtained the Metzner-Reed Reynolds number for Herschel-Bulkley fluids by
equating the analytical solution, Equation (5), with the Darcy-Weisbach equation, Equation (10)
giving:
Alternatively, this Reynolds number can be represented in a more traditional form by introducing the
concept of an effective viscosity:
where the effective viscosity m* is given by:
It can be seen when the Herschel-Bulkley model parameters are set to Newtonian values (ty=0, n=1
and j=m), then m* simply reduces to m and Equation (15) becomes the familiar Newtonian Reynolds
number.
Dodge and Metzner (1959) used the Metzner-Reed Reynolds number as the basis for friction factor
correlations for turbulent flow. In turbulent flow ReMR displays the correct qualitative behaviour, in
that it increases with increasing velocity and simplifies to the conventional form for Newtonian fluids.
However, because the effective viscosity m*, in the Reynolds number, is derived from a laminar
velocity profile, the value of m* which actually exists in a turbulent flow will be different. The
Metzner-Reed Reynolds number, therefore, is not physically realistic in turbulent flows except for
Newtonian fluids.
Using Rabinowitsch and Mooney's first equation, Equation (3), and after some manipulation of
Equation (5), it can be shown that the effective viscosity m* is related to the apparent viscosity at the
wall mw by:
BECHTEL WATER TECHNOLOGY LIMITED
1
1 - aX - bX
2
- cX
3
rVD
1
1 - X
ReMR =
j
3n + 1
4n
8V
D
(n-1) n n
1
1 - aX - bX
2
- cX
3
1
1 - X
m* = j
3n + 1
4n
8V
D
(n-1) n n
rVD
ReMR =
m*
(14)
(15)
(16)
3n + 1
4n
m* =mW (17)
BECHTEL WATER TECHNOLOGY LIMITED
In turbulent flow, if the wall shear stress is known, mW can be derived from the Herschel-Bulkley
model:
A new Reynolds number is proposed which is equivalent to the Metzner-Reed Reynolds number in
laminar flow and physically more realistic in turbulent flow. The general form of this new Reynolds
number is:
or specifically for Herschel-Bulkley fluids:
Because of the equivalence of the new Reynolds number with the Metzner-Reed Reynolds number in
laminar flow, both are determined by Equation (13). In turbulent flow of Herschel-Bulkley fluids the
new Reynolds number is determined using Equation (20) with the apparent viscosity at the wall being
determined using Equation (18).
TURBULENT FLOW
Dodge and Metzner (1959) developed a semi-theoretical equation for the smooth wall fully developed
turbulent friction factor for time-independent, purely viscous non-Newtonian fluids. This was based
upon Prandtl's mixing length theory with suitable values for the empirical constants being obtained
from experimental data. Tomita (1959) and Clapp (1961) used a similar method to develop equations
specifically for Bingham plastic and power law fluids respectively. Torrance (1963) extended this
work to be applicable to yield pseudoplastics and to account for pipe roughness. As with Dodge and
Metzner's equation, the equations of Tomita, Clapp and Torrance all include empirical constants and
apply to fully developed turbulent flows.
Wilson and Thomas (1985) and Thomas and Wilson (1987) used a mixing length approach for
Herschel-Bulkley fluids. Their model included corrections to change the thickness of the viscous sub-
layer according to the rheology and a term to account for the central, un-yielded, core which
theoretically exists in fluids which possess yield stresses.
j
1-X
mW =
( )
= tW
tW
gW
n-1
n
( )
1
n
(18)
(19)
(20)
1
1 - aX - bX
2
- cX
3
rVD
Re =
3n + 1
4n
mW
rVD
Re =
3n + 1
4n
mW
BECHTEL WATER TECHNOLOGY LIMITED
Their work has theoretical justification and the empirical constants used in the approach take the same
values as those used for Newtonian flows. Within the limitations of the Herschel-Bulkley model, the
method has been shown to give fairly good agreement with experimental data and is the best available
analytical model for highly turbulent flows.
The preceding analyses all assumed that there is a discrete zone over the majority of the pipe cross-
section where the turbulent eddy viscosity dominates the molecular viscosity and, hence, the flow is
fully developed turbulent flow. Although this approximation is sufficiently accurate for Newtonian
and dilatant fluids at typical Reynolds numbers it is clearly inaccurate for fluids which exhibit a yield
stress and/or power law behaviour where the molecular viscosity increases towards the centre of the
pipe and the turbulent viscosity decreases. Fluids with yield stresses have a central core where the
molecular viscosity is theoretically infinite and the turbulence is completely suppressed. The
assumption normally made in Newtonian flows that the viscous sub-layer occupies a negligible
fraction of the pipe radius cannot generally be true in turbulent non-Newtonian flows, particularly at
the low Reynolds numbers found in practice.
NUMERICALANALYSIS
An alternative approach to the analytical and semi-analytical methods is to use numerical techniques
to solve the basic equations which describe the physics of the flow, the fluid rheology and turbulence.
The rheology can either be defined from tabular rheometer data or a suitable rheological model.
Stainsby, Chilton and Thompson (1994) describe how a one-dimensional finite difference model and
a commercially available general purpose computational fluid dynamics code (CFX) have been
successfully used for this purpose. The one-dimensional finite difference model solves the basic
partial differential equation for fully developed flow in a circular pipe, with an apparent molecular
viscosity being calculated (iteratively) from the Herschel-Bulkley model. Turbulent viscosity is
defined by the low Reynolds number k-e model of Launder and Sharma (1974) using their
(Newtonian) values for the constants. The standard high Reynolds number k-e model cannot be used
for highly non-Newtonian flows, with the exception of dilatant fluids, as it inherently assumes that the
molecular viscosity is negligible everywhere except for the viscous sub-layer next to the wall. The
extra terms in the low Reynolds number model allow for the ability of the molecular viscosity to
dissipate turbulence across the whole flow region including the central core where an un-yielded and,
therefore, non-turbulent plug may exist. Further details of this model can be found in Stainsby,
Chilton and Thompson (1994) or Chilton, Stainsby and Thompson (1996).
COMPARISON OF NUMERICALAND
EXPERIMENTAL RESULTS
Before applying the numerical model to non-Newtonian fluids, the technique has been verified against
the Hagen-Pouiseuille Equation for laminar water flow and against pressure losses calculated from the
Moody Diagram for turbulent flow. In both flow regimes the model showed excellent agreement with
errors of less than 1% for laminar flow and between 1.5% to 3% for turbulent flows.
BECHTEL WATER TECHNOLOGY LIMITED
For laminar non-Newtonian flow, the numerical model agrees very well with both the analytical
models and the experimental results.
Laminar and turbulent flow predictions for yield pseudoplastics have been compared with
experimental results from various sources, the fluid properties and pipe diameters of some cases are
presented in Table 1. It can be seen that the pipe sizes, fluids and flows cover a wide range of practical
applications.
Table 1
Parameters use for Numerical Flow Predictions
1 SS
2 SS 0.157 1011 0.727 0.069 0.664 41
(*)
0.157
m kg/m
3
N/m
2
Pa.s
n
1024 1.268 0.214 0.613 34
(*)
3 SS
4 SS 0.157 1016 1.273 0.189 0.594 48
(*)
0.157 1013 2.827 0.047 0.806 44
(*)
5 KS
KS 6 0.140 1061 1.040 0.014 0.803 KERM2408
(**)
0.140 1071 1.880 0.010 0.843 KERM0608
(**)
7 KS
8 KS 0.079 1071 1.880 0.010 0.843 KERS0608
(**)
9 KS 0.079 1061 1.040 0.014 0.803 KERS2408
(**)
10 KS 0.079 1105 4.180 0.035 0.719 KERS2607
(**)
0.140 1105 4.180 0.035 0.719 KERM2607
(**)
No FLUID D r
ty
j n IDENTIFIER
SS - Sewage Sludge ; KS - Kaolin Slurry
(*) - Ackers and Allen (1980) ; (**) - Slatter (1995)
BECHTEL WATER TECHNOLOGY LIMITED
It became apparent to the authors, when looking for suitable experimental data, that accurate
rheological measurements are essential if the flow predictions are going to be meaningful. A large
fraction of the published data is of little worth because of poor rheometry, incorrect fitting and
inappropriate use of rheological models or simply incomplete data sets. The rheological data used in
this analysis was obtained from good quality tube and concentric cylinder rheometers, and data
reduction techniques similar to those described by Lazarus and Slatter (1988) for fitting the Herschel-
Bulkley model were applied.
The comparison of the numerical and experimental results (Nos. 1 to 10) are presented in Figures 1
to 7 (please contact the authors). The agreement is very good, both within the flatter, laminar, flow
region and the steeper, turbulent, region. Pressure loss predictions are consistently within 15%. There
is a clear tendency for the numerical method to under-predict the pressure losses. This under-
prediction is believed to be attributable to the deterioration of the Herschel-Bulkley model in the 2nd
Newtonian region. Wilson and Thomas's model is also presented and shows good agreement with the
experimental results and with the numerical model. Wilson and Thomas's theory consistently predicts
higher pressure drop values than the numerical model at low velocities with the difference reducing
as the velocity increases. This behaviour is expected because Wilson and Thomas's model is based
upon the assumption of a fully developed turbulent flow.
Presently, the numerical model is only applicable to smooth pipe wall conditions. In most practical
situations, at relatively low Reynolds numbers, the non-Newtonian molecular viscosity effects will
dominate the roughness effects.
CRITICAL REYNOLDS NUMBERS
Having demonstrated that the numerical model is in reasonably close agreement with experimental
results, it is possible to use the method to perform a series of numerical experiments to find the
transition point between laminar and turbulent flow. The transition velocity is found from the cross-
over point between the laminar and turbulent predictions. This approach gives the minimum velocity
at which turbulent flow can exist - transition in real pipelines can occur at higher velocities if the
laminar flow is very stable. Agreement of the predicted critical Reynolds numbers with the
experimental results presented by Dodge and Metzner (1959) is good. Moreover, by using the new
definition of Reynolds number the critical value was consistently found to lie between 2500 and 5000,
generally increasing with increasing non-Newtonian behaviour, particularly with increasing yield
stress.
NON-NEWTONIAN FRICTION FACTORS
As is the case in Newtonian flows, it is convenient to present the predicted pressure losses in a non-
dimensional form. Figure 8 (please contact the authors) shows the variation of predicted friction
factor with the Metzner-Reed Reynolds number for power law fluids with n varying between 1 and
0.4. The results of Dodge and Metzner (1959) and Wilson and Thomas (1985) are shown for
comparison. The predicted friction factors show good agreement with Newtonian fluids but an
increasing tendency to under-predict as n decreases. The numerical predictions compare well with the
theory of Wilson and Thomas (1985) particularly as the Metzner-Reed Reynolds number increases.
BECHTEL WATER TECHNOLOGY LIMITED
When the predicted friction factors are plotted against the new Reynolds number, instead of the
Metzner-Reed Reynolds number, the variation due to non-Newtonian effects reduces. Figure 9
(please contact the authors) demonstrates this for power law fluids.
The predicted variation of friction factor with the new Reynolds number for Bingham plastic fluids
for different X values is presented in Figure 10 (please contact the authors). Comparison of Equation
(14) for the Metzner-Reed Reynolds number with Equations (18) and (20), which define the new
Reynolds number, shows that the two Reynolds numbers are identical for the case of Bingham plastic
fluids.
A NEW TURBULENT FLOW CORRELATION
About the same time that Prandtl and Von Karman derived semi-analytical expressions for friction
factors in turbulent flow, some researchers developed purely empirical correlations - generally of the
form:
f = A + B(Re)
C
Where A, B and C are constants. Table 2 (below) gives values of the constants and the upper limit of
Reynolds numbers to which the correlations are stated to hold.
Table 2
Coefficients for Newtonian Turbulent Friction Factor Correlations
(21)
SOURCE
Blasius
Lees
Hermann
Nikuradse
Numerical Model
Numerical Model
A
0.0000
0.0018
0.0013
0.0008
0.0020
0.0014
0.079
0.153
0.099
0.055
0.138
0.090
-0.250
-0.350
-0.300
-0.237
-0.350
-0.293
100,000
400,000
2,000,000
?
100,000
1,000,000
B C Re
The differences between the values of A, B and C are not only due to the inaccuracies in the different
experiments but are also due to the fact that all of these equations only approximate the true friction
factor variation. In essence, the value of the constants and accuracy of the correlation depend upon
the range of Reynolds numbers over which the correlation is fitted.
The last two rows in Table 2 were obtained by fitting the results of the numerical model applied to a
Newtonian flow, over different ranges of Reynolds number, to obtain A, B and C. Acorrelation based
on these values is accurate to +/- 2 % over the given ranges - an accuracy which is comparable to that
of Prandtl's equation for smooth pipes.
Correlations similar to those presented in Table 2 can be fitted to Herschel-Bulkley fluids based upon
the new definition of Reynolds number. A series of numerical simulations have been performed to
find the coefficients and their relationship with ty and n over different ranges of Reynolds numbers
and rheological parameters. Only the simplest relationship was used to fit the data which resulted in
an expression which is based on Blasius's equation:
If the correlating function, involving X and n, is absorbed into a modification of the new Reynolds
Number, Re', as shown in Equation (23), then the friction factor correlation is identical to the Blasius
equation. It has been found that the more accurate Newtonian equations, such as Prandtl's equation,
Equation (24) or the Moody diagram, can also be used for Herschel-Bulkley fluids if Re' is substituted
for the Newtonian Reynolds number:
f
-0.5
= 4.0Log10(Re(f)
0.5
)-0.4
Including the extra terms in the definition of Reynolds number essentially makes turbulent non-
Newtonian and Newtonian pipe flow calculations 'universal'. To demonstrate this the friction factors
derived from the experimental results given in Table 1 are plotted against Re' on a Moody diagram,
Figure 11. In general, all of the experimental data appears to lie on the same curve, the accepted
Newtonian 'hydraulically smooth' line. The difference between the experimentally derived friction
factors and the hydraulically smooth line is generally within 15%. Except for experimental scatter,
there is a systematic deviation at high Reynolds numbers which is believed to be due to the limitations
of the Herschel-Bulkley model from which Re' and the subsequent correlation was derived.
Alternative rheological models which will perform better at higher Reynolds numbers are currently
being investigated.
BECHTEL WATER TECHNOLOGY LIMITED
(22)
(23)
f = 0.079 = 0.079(Re)
-0.25
-0.25
Re
n
2
(1-X)
4
BECHTEL WATER TECHNOLOGY LIMITED
CONCLUSIONS
A numerical model for the simulation of laminar and turbulent flow of Herschel-Bulkley fluids has
been developed. This model was found to agree closely with analytical solutions in laminar flow. In
turbulent flow the numerical model predicts pressure losses which are generally lower than the
experimental values with less than 15% error. Wilson and Thomas's semi-analytical approach agrees
well with the numerical model, with the agreement improving as the velocity increases.
The Metzner-Reed Reynolds number is derived for Herschel-Bulkley fluids and is shown to be
consistent with previous work and with Newtonian fluids in laminar flow.
A new Reynolds number is proposed which is equivalent to the Metzner-Reed Reynolds number in
laminar flow and is able to predict the critical velocity at which transition to turbulence occurs. More
importantly the new Reynolds number is demonstrated to be more physically representative of
transitional and turbulent flow conditions and gives almost 'universal' behaviour when plotted against
friction factor.
Numerical simulations have been performed for a wide range of Herschel-Bulkley parameters from
which a correlation for friction factor as a function of the new Reynolds number has been derived. It
has been shown that this correlation goes some of the way towards a 'universal' approach for
determining pressure losses in Newtonian and non-Newtonian pipe flows.
ACKNOWLEDGMENTS
The authors would like to thank Prof T Jones at the University of Plymouth for his work on rheometry
and Dr P Slatter at Cape Technikon for providing his experimental data.
BECHTEL WATER TECHNOLOGY LIMITED
APPENDIX I
REFERENCES
Ackers P. and Allen M.C. (1980)
"The laminar flow of sewage sludge through pipelines" The Public Health Engineer, Vol 8, p88-106.
Buckingham E. (1921)
"Plastic flow through capillary tubes" Proc ASTM, Vol 21, p1154
Chilton R.A., Stainsby R. and Thompson S.L. (1996)
"The design of sewage sludge pumping systems" J Hyd Res, Vol 34, p395-407 IAHR
Clapp R.M. (1961)
"Turbulent heat transfer in pseudo-plastic non-newtonian fluids" Proc Conf Int Dev in Heat Transfer,
Univ Colorado, Boulder, pt III, Sec A, p652-661
Cross M.M. (1965)
"Rheology of non-Newtonian fluids: a new flow equation for pseudoplastic systems" J Colloid Sci,
Vol 20, p417-437
Dodge D.W. and Metzner A.B. (1959)
"Turbulent flow of non-Newtonian systems" AIChE J, Vol 5, p189-204
Govier G.W. and Aziz K. (1972)
"The flow of complex mixtures" Von Nostrand-Reinhold, New York, USA
Herschel W.H. and Bulkley R. (1928)
"Measurement of consistency as applied to rubber-benzene solutions" Proc ASTM, Kolloid Z,
Vol 26, p621
Kozicki W. and Kuang P.Q. (1993)
"Prediction of lower/upper limiting viscosities" Can J Chem Eng, Vol 71, p329-331
Lazarus J.H. and Slatter P.T. (1988)
"A method for the rheological characterisation of tube viscometer data" J Pipelines, Vol 7, p165-176
Launder B.E. and Sharma B.I. (1974)
"Application of the energy dissipation model to the calculation of flow near a spinning disc" Lett Heat
and Mass Transfer, Vol 1, p131-138.
Meter D.M. and Bird R.B. (1964)
"Tube flow of non-Newtonian polymer solutions: Part 1 laminar flow and rheological models" AIChE
J, Vol 10, p878-881.
BECHTEL WATER TECHNOLOGY LIMITED
APPENDIX I (Contd)
REFERENCES
Metzner A.B. (1957)
"Non-Newtonian fluid flow - Relationships between recent pressure drop correlations" Ind Eng
Chem, Vol 49, p1429-1433
Metzner A.B. and Reed J.C. (1955)
"Flow of non-Newtonian fluids - correlation of the laminar, transition, and turbulent-flow regions"
AIChE J, Vol 1, no. 4, p434-440
Mooney M.J. (1931)
"Explicit formulas for slip and fluidity" J Rheol, Vol 2, p210-222
Rabinowitsch B. (1929)
"Ueber die viskositat und elastizitat von solen" Z Phisik Chem Ser A, Vol 145, p1-26
Sisko A.W. (1958)
"The flow of lubricating greases" Ind Eng Chem, Vol 50, p1789-1792
Slatter P.T. (1995)
Personal communication
Stainsby R, Chilton R.A. and Thompson S.L. (1994)
"Prediction of the head losses in non-Newtonian sludge using CFD" Proc 2nd CFDS Int User Conf,
Pittsburgh, USA, p259-272
Thomas A.D. and Wilson K.C. (1987)
"New analysis of non-Newtonian turbulent flow - Yield-Power-Law fluids" Can J Chem Eng, Vol 65,
p335-338
Tomita Y. (1959)
"A study of non-Newtonian flow in pipelines" Bull JSME, Vol 2, p469.
Torrance B. McK. (1963)
"Friction factors for turbulent non-Newtonian fluid flow in circular pipe"
J South African Mech Eng, Vol 13, p89-91.
Wilson K.C. and Thomas A.D. (1985)
"A new analysis of the turbulent flow of non-Newtonian fluids" Can J Chem Eng, Vol 63, p539-546.
BECHTEL WATER TECHNOLOGY LIMITED
APPENDIX II
NOTATION
The following symbols are used in this paper:
a = coefficient defined by eqn 5b;
A = constant in eqn 22;
b = coefficient defined by eqn 5c;
B = constant in eqn 22;
c = coefficient defined by eqn 5d;
C = constant in eqn 22;
D = pipe diameter;
f = Fanning friction factor;
j = Herschel-Bulkley parameter;
K' = apparent consistency coefficient;
L = pipe length;
n = Herschel-Bulkley index;
n' = apparent flow behaviour index;
DP = pressure loss;
Re = Reynolds number (including new definition);
ReMR = Metzner-Reed Reynolds number;
Re' = modified new Reynolds number;
V = mean pipe velocity;
X = yield and wall shear stress ratio;
g = shear strain rate;
gW = shear strain rate at the wall;
h = coefficient of rigidity;
l = Darcy-Weisbach friction factor = 4f;
m = viscosity;
ma = apparent viscosity;
mp = power law viscosity;
mw = apparent viscosity at the wall;
m* = effective viscosity for Herschel-Bulkley fluids;
r = density;
t t = shear stress;
t tw = wall shear stress;
t ty = yield stress;

Potrebbero piacerti anche