Minerals Engineering, Vol. 6, No. 4, pp. 3 9 7 - 4 0 4 , 1993 0892-6875/93 $6.00+0.

00
Printed in Great Britain © 1993 Pergamon Press Ltd

AN EXAMINATION OF PULP VISCOSITY IN TUBES AT HIGHER SHEAR RATES

W. WHITEN, P. STEFFENS and J. HITCHINS

JK Mineral Research Centre, Isles Road, Indooroopilly,

Qld. 4068, Australia
(Received 3 September 1992; accepted 29 October 1992)

ABSTRACT

Deficiencies in the description of pulp properties is a major factor limiting

improvements in modelling of the flow within many mineral processing
separation devices such as jigs and cyclones. Two factors severely hamper the
development of accurate descriptions of pulp rheology - the large number of
variables which influence pulp behaviour and the difficulty of making accurate
measurements of pulp viscosity. Many factors make measurements of pulp
viscosity difficult, including wall effects, turbulence, settling and the physical
size of particles, particularly at high shear rates. This paper examines the
measurement of pulp viscosity using capillary viscometers, and in particular the
high shear rate regions for which turbulent flow (or the transition to turbulent
flow) occurs. The results show a marked difference in the behaviour of the
pulps in this regime compared to Newtonian fluids. This is significant for
hydrodynamic simulations of pulps at high shear rates.

Keywords
Pulp viscosity; turbulence; tube viscometer; slurry viscometer

1: INTRODUCTION

The viscosities of ore and water mixtures commonly known as pulps or slurries are
important in many mineral processing units. These viscosities often have a large effect on
the process performance and power consumption. This applies particularly to units that
separate using particle settling rates, and they are even more important when the separation
forces are created by a vortex flow as in hydrocyclones and heavy media cyclones.

The properties of such pulps are not well understood, partly because of the range of
variables that can affect pulp behaviour and partly because of the difficulty of making
accurate measurements of pulp properties. For instance, particle density, particle size
distribution and particle shape characteristics all affect the pulp behaviour, while wall
effects, settling and the physical size of particles make the measurements of pulp properties
difficult. This paper looks at the measurement of pulp properties using capillary or tube
viscometers, particularly in the region of transition from viscous to turbulent flow.

2: VISCOSITY OF SLURRIES

In general, there is good agreement amongst the different theoretical studies, that the
viscosity of extremely dilute suspensions, #s, can be found using the relationship (for
instance see [1]):

~ , = ~o(l+k~ ~) (I)
397
398 W. WHITEN et al.

where ~bis the volume fraction solids, and k 1 is accepted to be 2.5 for no slip at the particle
surface (spherical particles), and for purely laminar flow.

For more concentrated suspensions the viscosities of pulps or slurries are affected by
percent solids, and by particle size, shape and size distribution ([2] pp. 411-412). Percent
solids or concentration is the major factor in slurry viscosity. For small concentrations,
increases in percent solids will increase the viscosity at the rate given by equation (1),
however as the concentration is increased, the viscosity is found to increase progressively
more rapidly with percent solids. The existence of a velocity gradient causes particle
rotation which increases with increasing particle size. Increased particle interaction,
including that caused by rotation, will result in an increase in viscosity. Particle shape is
important as minimum viscosity occurs when the slurry consists of spherical particles.
Increases in the particles asymmetry, at the same percent solids, will increase the viscosity.
Particle size distribution also affects the viscosity due to the quantity of smaller particles
filling the gaps between larger particles. This changes the effective concentration and hence
changes the viscosity.

The equation

/~s = /~o e2"5(#+5¢3) (2)

has been found to closely fit the viscosity data reported by Thomas, and as it also agrees
with our data, was used to represent the theoretical viscosity values in this analysis. For
small values of ~, it is equivalent to equation (1), but it predicts much higher viscosities at
higher values of ~b as is generally found.

2.1: Viscometers

To measure the viscosity of a fluid or slurry there are two main categories of viscometer.
The first relies upon the creation of a velocity gradient in a thin layer of liquid and the use
of Newtons law of viscosity r = # d u / d y where r is the shear stress, d u / d y is the velocity
gradient and p the viscosity. This is conveniently done in a rotational or bobbin viscometer
which usually consists of a bobbin rotating at a known speed with respect to a stationary
outer concentric cylinder. The viscosity can then be found from/~=kM/fl where M is the
torque at the bobbin, fl is the angular velocity of the bobbin, and k is an instrument
constant dependent on the dimensions of the bobbin and the outer cylinder ([3] p.365).

The second type of viscometer is the capillary or tube viscometer. This requires the use of
the Hagen-Poiseuille equation for laminar flow in tubes ([3] p.193):

V = zxP DZ/(32 #L) (3)

where V is the average fluid velocity, D is the internal tube diameter, AP is the pressure
drop over a distance L, and/~ is the viscosity which is assumed independent of shear rate
i.e. a Newtonian fluid.

Several problems occur for both types of viscometer when used to measure the viscosity of
a slurry:
(i) They are both subject to Vand wall effects [4] which results in a layer of pure liquid
at the wall of the cylinder or tube, thereby altering the measured viscosity.
(ii) T h e y are both prone to turbulence. The frictional losses from these small scale
motions affect the macro losses being measured in a way which is difficult to
quantify.
(iii) The bobbin viscometer is prone to other secondary motions, known as Taylor
vortices, at high rotational speeds ([5] pp. 51-52) which have a similar effect to
turbulence.
(iv) The bobbin viscometer also exhibits a hysteresis effect where the viscosity at a
certain rotation speed varies depending on whether the speed was reached from a
faster or slower rotation velocity. This may be caused by the rheopectic and
 Pulp viscosity 399

thixotropic behaviour of suspensions, where the apparent viscosity of the suspension

increases and decreases respectively, with time ([2] p. 400). It has also been explained
by differences in the Taylor vortices [5].
(v) As particles in a slurry settle, it is often difficult to determine the solids content of
the portion of the slurry being measured.

Slurry viscometers generally need to be of a more robust construction with provision to

maintain the slurry uniformly in suspension and with larger physical apertures to allow free
passage of solid particles. As these units do not satisfy the requirements for direct
calibration, they are usually calibrated using fluids of known viscosities.

A tube viscometer was chosen for this work because of its simple construction and the ease
with which a range of physical dimensions could be tested.

2.2: Pipe Flow

For steady incompressible flow in a pipe, the Darcy-Weisbach equation ([3] p. 235):

hf = f LV2/(2gD) (4)

is generally adopted. With hf the head loss in the pipe of length L and internal diameter D,
and V the average velocity, f is a dimensionless friction factor required to correct the
equation for losses ([3] pp. 235-245). The friction factor depends upon velocity, diameter,
density p, viscosity/~, and certain characteristics of the wall roughness which are equal to
zero for smooth pipes. Thus f must be dependent on the dimensionless Reynold's number

Re = pDV/# (5)

Using the Hagen-Poiseuille equation substituted into equation (4), and making the
substitution AP = pghf, the simple equation

f = 64/Re (6)

is derived. This is the straight line representing laminar flow in pipes shown in Figure 1
which gives the results of experiments done by Nikuradse on the relative roughness of
pipes. Also shown in this figure is a line from the empirical relationship developed by
Blasius:

f-- 0.316/Re °'25 (7)

which describes turbulent flow, and is valid up to about Re -- 100 000.

The Darcy-Weisbach equation was corrected for minor head losses using KV2/(2g) where
K is a coefficient associated with the minor head losses and is determined by experiment.
Solving this for f, the equation used for the calculation of the friction factors became:

f _- [(W+L)/L][2gD/V2I-K(D/L) (8)
where W is the fluid head above the tube inlet.

3: EXPERIMENTAL RESULTS
3.1: Apparatus

To examine the capabilities of a tube viscometer in measuring the viscosity of slurries, a

capillary system was designed and tested as follows.

Four identical glass head tanks were used in this experiment, with each inlet supplied by
one central four-way-splitter head tank. This enabled a simultaneous sample to be taken
400 W. WHITEN et al.

from each glass tube suspended from the outlet of each of the four head tanks. The tubes
were approximately 2, 3, 4, and 5 mm internal diameter, and there were three lengths of
each diameter, approximately 573,927, and 1500 mm, in order to determine the effect, if
any, of tube length on the calibration curve of friction factor versus Reynold's number.

0.10 IIII
~IIII
II lillll
IIIIII
II lI.IlILllll
IIIIII
II iI lll
III
I
0.09
0.08 ~II I IIIIII I I IIIIIII I I III I
0.07 l~ll I II!III I IIIIIIII I IIII I
0.06: I [l H±D _ ±30
0.05 I
III \.'1-,.I L~ilII .o-2.4~,c&111.111...':4 [~1 I i,-,411 D 61.2
~1~ ~ 0.04 III ~,. I "t'x,¢1
.,
IIII/
--
a,.--'r"F~[~I'IT'-"V-I'-I'I
"t ' , I I D-9.Scm"
Fl'ftl
1
I~
I ~"120
0.03 _~.
D
I
252

0.02 II I\ D 504

1
D 1014

0.011 10 3 10 4 I0s 106

VDp
I~.=~
bt
Fig. 1 Nikuradse's sand-roughened-pipe tests, (Streeter and Wylie p. 236)

The overflow from each head tank was returned to the sump, which contained a heating coil
operated by a PID controller to maintain a constant 35°C temperature. For each experiment,
measurements of fluid head height (above the tube inlet) and flow rate were taken.

3.2: C a l i b r a t i o n

Calibration was done using known viscosity sugar solutions (from tables of viscosities as
measured at the National Bureau of Standards) to establish a curve of friction factor versus
Reynold's number. This curve was necessary to determine Reynold's numbers from the
measured friction factors of the slurries, and hence to provide the slurry viscosities.

With the apparatus as given, sugar solutions of 0, 5, 10, 20, 30, and 40 % by mass were used
for the calibration. Initially the 2, 3, 4, and 5 mm tubes were used, with a constant head of
45 mm above the inlet of each tube. These tests were repeated with the heads constant at
25 mm. Tests with the l m m tube replacing the 4mm tube, and with a head of 25 mm, were
then completed using 0, 20, 30, and 40 % solutions. These tests produced more points at
lower Reynold's numbers and in the laminar to turbulent transition region.

For each experiment the Reynold's number and friction factor were determined according
to equations (5) and (8) respectively. The value of K (minor losses constant) in equation (8)
was considered to be invariant and was determined to be equal to 1.41 which is consistent
with reported values for similar entrance conditions to our experimental conditions ([3]
pp.243-245). This value was selected to minimise the deviation of the data from a spline
curve incorporating both the Hagen-Poiseuille laminar relationship (equation (3)) for
Re<1200 and the Blasius turbulent relationship (equation (7)) for Re>3000.
 Pulp viscosity 401

Figure 2 presents the results of these calibration experiments. Comparison with Figure 1
shows they are reasonably consistent with previously reported results. In the laminar region
for Re<1200 the experimental results lie slightly above the Hagen-Poiseulle curve as do the
earlier results. In the fully turbulent region, 3000<Re<12000, the results follow Blasius'
smooth pipe relationship well. However our results depart slightly from those reported in
Figure 1 in the transition region. Our results exhibit much more scatter in this region and
the onset of transition occurs earlier (Re -~ 1200 as opposed to Re --- 2000). These deviations
can both be explained by variation in entrance conditions, as flow in this transition region
is very sensitive to conditions at entrance. The major variation is from the largest diameter
tube which recorded a lower friction factor value in the first part of the transition region

l°l--iiiiiii:
1
-------:;;:----iiiiiiii
giliniinilllllli
N I I I lI lI i l l l i l l l l
l l l l U l i k ~ I l i l l l i l i
i i

friction factor
n n n n l n n n ~ , - l i n n l m l l l
i l l l l l i / R l m m n i l l l i
nnlnuuih~,~lnniiml
0.1 miinnnuiil|guu:] i

i i i iIlii i "-~ ' i

0.011 . t n t t Ett I I l i iilll i i ,
10 100 1000 10000 100000
Reynold's number

Fig.2 Friction factor v's Reynolds number calibration curve

from sugar data (K= 1.41) with Blasius overlay

3.3: Slurry Experiment and Results

Slurries of 10, 20, and 30% (by volume) sand in water and 10, 20, 30 and 40% (by volume)
sand in a 30% (by mass) sugar solution were run through the same apparatus with the 2, 3,
4, and 5 mm tubes. The sand had a size range of approximately 125-250~m.The conditions
of the calibration experiment were replicated, with the exception of a stirrer included in the
sump to discourage settling and to keep the slurry well mixed. However, the % solids
measured at the discharge from the tubes was often noticeably different from the target
value. Samples from each tube and for each % solid were taken to give flow rates, densities,
% solids, and size distributions.

Friction factors were calculated from equation (8) and the calibration spline curve was used
to produce Reynold's numbers for each size tube and % solid. The viscosities were then
calculated from equation (5). For friction factors which did not yield a unique solution for
Reynold's number, ie. in the transition region, an uncertainty band was established. More
specifically, for friction factors between .03 and .07 the range of possible Reynold's number
was calculated. The lower limit of this range was the Hagen-Poiseuille relationship for
laminar flow and the upper limit of this range an extrapolated linear (on log-log axes)
approximation to the turbulent calibration curve, similar to the Blasius relationship. In
addition, where it was possible to identify the most likely section of the calibration curve,
this was also done.

Figures 3 and 4 show these results plotted in terms of slurry viscosity versus % solids for
the sand/water and sand/sugar solution slurries respectively. Also plotted on these curves
is the previously established empirical relationship (equation (2)). The fluid viscosity in this
equation was taken as the average of the calculated viscosities of the 0% solids experiments.
 402 W . WHITEN et al.

12

,/
10
/

Viscosity' (cp) 6 TTJ "•

J

0 5 10 15 20 25 30 35 40
% solids (by volume)

Fig.3 Viscosity v's % solids for sand in sugar. Triangles represent the measured data,
while the squares indicate viscosity ranges.

Viscosity (cp) 4 //
3
J
2 J

1 J
t'------
0
0 5 10 15 20 25 30 35 40 45
% solfds (by volume)

Fig.4 Viscosity v's % solids for sand in water. Triangles represent the measured data,
while the squares indicate viscosity ranges.

Clearly those points which are not in the transition region are in close agreement with this
empirical curve (std error = 0.40 cPas), although Figure 4 has only a few of these points as
the lower viscosity of the carrying fluid brought most of the test conditions into the
transition region. Further, the uncertainty bands for all those points within the transition
region pass across, or close to, this empirical curve. We therefore conclude that this
empirical representation of slurry viscosity is a good approximation over the range of our
data. The data for which there is a clear calibration point does not show evidence of non-
linear effects as the shear rate changes. There is no evidence in the data for an effect of
tube length.

We subsequently used the viscosity obtained from this relationship to recalculate the
Reynold's number for each experimental point. The resulting friction factor versus
 Pulp viscosity 403

Reynold's number plots are shown in Figure 5. Figure 6 further shows an expanded view
of a section of this curve in which each experimental point has been classified according to
its % solids rounded to the nearest multiple of 10.

friction factor 0.1 i L i i i ~l

i

I"-.~ ;,3, . . . .I. . . I I II

I
A L I

0.01
100 1000
"liirii 10000 100000

Reynolds number

• sand m Spline • s a n d in sugar

Fig.5 Friction factor v's Reynold's number for sand in water and in sugar,
using viscosities from polynomial with spline curve overlay.

0.1

"" Spllne

00%

• -10%
Friction factor
• -20%

• -30%

o -40%

0.01
1000 10000

Reynold's number

Fig.6 Friction factor v's Reynold's number using viscosities from polynomial
for both sand in water and in sugar, grouped by measured % solids.

These curves reveal a different trend through the transition region of the slurries compared
to the fluids. The slurries have less of a marked transition and tend to move more slowly
from laminar flow to turbulent flow as % solids increases. As shown in Figure 6, the slurry
curves are all much smoother than the 0% curve through the transition region and do not
exhibit the "bump" associated with transition of Newtonian fluids. In the transition region
for the fluids (1000<Re<3000), the introduction of solids appears to increase the friction
factor for a given Reynold's number, and therefore the frictional resistance to flow.
However, for larger Reynold's numbers (Re>3000), for which the fluids exhibit fully
turbulent flow, the slurry curves exhibit a lower friction factor for a given Reynold's
number, tending more towards laminar behaviour. A plausible explanation for this
occurrence is that the inclusion of particles in the flow dampen the small scale motions
associated with turbulence.
404 W . W H I T E N et al.

4:CONCLUSIONS

The viscosities of slurries is a difficult area in which to undertake experimental work. The
design of test equipment that gives clear results independent of other possible effects is not
easy. There are also a large number of variables related to the particles, such as shape, size,
size distribution, and surface properties, that potentially affect the measurements.

The use of a tube viscometer for the measurement of slurry viscosity must be undertaken
within a range of constraints: the tube diameter is limited below by the need to avoid
blockage and above by turbulence and vortices at the inlet. It is not possible to reduce the
effective gravitational force by inclining the tube as settling and possibly blockages occur.
Within the range determined by the physical requirements we have obtained good results in
the laminar region. This requires small particles and low diameter tubes. A more viscous
fluid, if appropriate, would be useful in extending the technique to larger particles.

For the particular slurries tested, we found no evidence in the laminar range of non-linear
viscous properties, ie the measured viscosity appeared independent of the shear rate. In the
transition region the situation is more complex. We consider that these results are best
interpreted as a change in turbulence within the slurry: an increasing concentration of
particles progressively reduces the amount of turbulence, resulting in a progressively
smoothed and delayed transition to turbulent conditions. The alternative of a viscosity
versus percent solids relation that includes large inconsistencies is not considered likely.

This is important for the measurement of slurry viscosity in the transition/turbulent region
as the calibration is usually against fluids that produce similar viscous forces at the same
shear rate. Similarly care is needed in the hydrodynamic simulation of slurries in this
regime.

There is a clear need for more experimental work in a range of devices and on slurries of
different compositions to further determine slurry viscous properties.

REFERENCES

. Thomas, D.G., Transport Characteristics of Suspensions: VIII. A Note on the

Viscosity of Newtonian Suspensions of Uniform Spherical Particles. Journal of
Colloid Science 20, 267-277 (1965).
. Jinescu, V.V., The Rheology of Suspensions International Chemical Engineering 14,
No. 3, 397-420. (July 1974).
3. Streeter, V.L. & Wylie, E.B., Fluid Mechanics. First SI Metric Edition. 562. McGraw-
Hill London, (1981).
4. Goldsmith, H.L. & Mason, S.G., The Microrheology of Dispersions. in Rheology-
Theory and Applications 4, Eirich, F.R. (ed.), 85-233 (1967).
5. Castro, O., An Investigation of Pulp Rheology Effects and Their Application to the
Dimensionless Type Hydrocyclone Models. Masters Thesis, Julius Kruttshnitt Mineral
Research Centre, University of Queensland. (1990)