You are on page 1of 21

This article was downloaded by: [University of Waterloo]

On: 24 March 2015, At: 09:49


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Particulate Science and Technology: An International


Journal
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/upst20

PREDICTION OF VERTICAL LIQUID SOLID PIPE FLOW


USING MEASURED CONCENTRATION DISTRIBUTION
a b
A. S. BARTOSIK & C. A. SHOOK
a
Mechanical Engineering Department , Kielce University of Technology , 25–314, Kielce,
Poland
b
Chemical Engineering Department , University of Saskatchewan , Saskatoon,
Saskatchewan, S7N OWO, Canada
Published online: 24 Feb 2007.

To cite this article: A. S. BARTOSIK & C. A. SHOOK (1995) PREDICTION OF VERTICAL LIQUID SOLID PIPE FLOW USING
MEASURED CONCENTRATION DISTRIBUTION, Particulate Science and Technology: An International Journal, 13:2, 85-104, DOI:
10.1080/02726359508906672

To link to this article: http://dx.doi.org/10.1080/02726359508906672

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
PREDICTION OF VERTICAL LICIUID~SOLID
PIPE FLOW USING MEASURED CONCENTRATION
DISTRIBUTION

A. S. BARTOSIK
Mechanical Engineering Department
Kielce University of Technology
25-314 Kielce, Poland

C. A. SHOOK
Chemical Engineering Department
University of Saskatchewan
Saskatoon, Saskatchewan
S7N OWO, Canada
Downloaded by [University of Waterloo] at 09:49 24 March 2015

ABSTRACT

The performance of a standard (k-L) turbulence model for predicting slurry pipeline pressure
gradients is examined using experimental measurements of velocity and solids concentration
distribution for upward vertical flows. Assuming the solids concentration rises abmptly from
a value of zero at the pipe wall over a distance corresponding to one half a particle diameter,
satisfactory predictions were obtained for sand particles of median diameter up to about 800
pm in a 25.8 mm vertical pipeline. No damping of fluid turbulence or additional strwes due
to particle-particle interactions were included in the computational model.

Velocity distributions are found to be flatter than those of turbulent single-phase flows,
because of the variation of mixture properties in the vicinity of the pipe wall.

Key words: Fully developed slurry Oow. Flow in vertical pipes. Solid concentration
distribution. Modelling of turbulence.

I. INTRODUCTION

Liquid-solid turbulent pipe flows occur widely in the chemical and mineral industries.
Considering the complexity of these Oows, particularly with high solid concentrations, it is
desirable to search for better understanding of the phenomena and to create models which
can be used in engineering practice.

Steady single-phase turbulent flows have been investigated extensively and information
gained from experiments has contributed to formulation of reliable turbulence models suited
to prediction of the behaviow of such flows with reasonable accuracy, 'Launder and
Spalding, 'Launder et al. and 'Patel et al. The diiculties increase when we consider
two-phase flow and development of new turbulence models to predict two-phase flow
behaviour is limited by access to experimental data.

Particulate Science and Technology, 13:85-104. 1995 85


Copyright 1995 Taylor & Francis
0272-6351195 510.00 + .OO
86 A. S. Bartosik and C. A. Shook

Measurements at high concentrations of the dispersed phase are avoided because of the risk
of damage or contamination of intmsive hot-film or hot wire probes. Also, experiments are
more complicated in the vicinity of a solid waU. With optical methods, attenuation of the
beams by panicles occurs. As a result of these difliculties at solid concentrations above 14%
by volume. most of the measurements in the Literature concern gas-liquid or gas-solid flow
fields. Notable measurements of turbulent gas-solid pipe flow include these of 'Lee and
Durst. '.'Tsuji et al., 'Modarress et al. and 'Wang et al. Measurements of turbulent
liquid-solid pipe flow are scattered in the literature but one must note the experimental data
of 'Zisselmar and Mokms and '@Noun el al.

The aforementioned experiments have s h o w that the presence of a solid or liquid dispened
phase modifies the turbulent structure of the flow field. Thus, use of a standard turbulence
model to predict gas-solid or liquidsolid flow is not always appropriate. In recent years one
can find suggestioos in the literature of several turbulence models for two-phase flows. Most
Downloaded by [University of Waterloo] at 09:49 24 March 2015

of them are based on standard turbulence models, denoted as zero, one, and two equation
models, "Launder and Spalding, "Spalding. In the papers of '3.'4Hetsroni we find a wide
review of literature on the interaction between solid particles and the carrier fluid including a
review of turbulence models and matching of experimental data. The new turbulence
models, including these of "Danon et al., '6Melville and Bray, "Genchev and Karpuzov,
'%lghobashi and Abou-Arab, 'qPourahmadi and Humphrey, Wostafa and Elghobashi
I,
Chen and Wood, and more recently, =Lee and Chung =Bernard and Harlow, Ulssa, "Rizk
and Elghobashi are useful for gas-solid flows. Most of them have been verified for a narrow
range of parameters and for fully developed flow. Other two-phase one equation models of
turbulence have been suggested by =Roc0 and Balakrishnan and "Abou-Arab and Roc0 for
both liquid-solid and gas-solid pipe flows.

It is also appropriate to mention that a quite distinct approach to multi-phase fluid dynamics
has been followed by other workers, as for example 'Trowe and Sharma, and V r o w e and
Cboi, "Lee and Wiesler and "Kallio and Reeks who developed a particle-trajectory model.

In the absence of reliable turbulence data in the immediate vicinity of the wall over a wide
range of solids concentrations it is difficult to suggest a new turbulence model for liquid-solid
flow. However, it is possible to suggest modiications of standard turbulence models on the
basis of comparison between calculations and measurements of global parameters which are
particularly relevant in engineering practice.

When the particle diameter is sufficientlyh e , the particles can follow the motion of the iluid
so that turbulent diffusion produces a uniform concentration distribution within the pipe. In
this case, it is reasonable to assume the physical properties of the mixture (density and
viscosity) are constant across the pipe section. In this case, use of a single-phase turbulence
model to predict liquid-solid flow should be acceptable, at least to moderate concentrations,
"Bartosik and Sobocmski. These fine-particle slurries wiU be of the non-senling type.

However, there is not sufficient information in the literature concerning the Limiting
concentration and particle diameters for which the standard turbulence model is acceptable.
Of course there are two main reasons why a standard turbulence model will fail. Fim, the
concentration of the particles will not be constant. The measurements of "Nasr-El-Din et al.
and %mner et al. indicate the influence of panicle diameter on diiribution of
concentration, particular in the vicinity of the waU. Secondiy, because a standard model
neglects the two-phase nature of the flow and because a length scale must contain a d c i e n t
Concentration Distribution to Predict Pipe Flow 87

number of panicles to permit acceptable local delinitions of density and velocity, "Lumley,
ultimate failure at large panicle diameters must be expected.

This paper deals with fuUy developed turbulent venical pipe flow of liquid-solid mixtures
with moderate solids concentrations (C, = 10-40% by volume). The panicles used in these
studies were all closely sized. The numerical predictions were carried out for the water and
h e , medium, and coarse sands with median panicle diameters d=175 pm, d=470 gm, and
d=780 pm, respectively. These experimental conditions cover the vast majority of industrial
slurries of the "senling" type. A Newtonian fluid model is appropriate for these slurries,
"Shook and Roco. The results of numerical predictions are compared with experimental data
of -Sumner et al. in which the panicle density was 2650 kglm'. At the experimental
conditions the characteristic response time parameter (defined by equation (I)) for the fine
sand panicles is about 4.5 ms while for the medium and coarse sand it is about 32 ms and 89
Downloaded by [University of Waterloo] at 09:49 24 March 2015

ms, respectively.

Panicle-panicle and particle-fluid interactions depend on particle size, relative velocity and
the difference in densities. The measuremencs of "Yiannesltis and Whitelaw, conducted for
fully developed turbulent pipe flow for water-solid mixtures with panicles of mean diameter
270 pm and a few microns, indicated flow that was fully developed, with no significant
differences in the liquid and solid phase velocities. These authors cited earlier observations of
"Cox and Mason which also suggested that for similar particle diameters and for fully
developed flow there was no substantial slip between the panicle and the liquid.
The slip velocity at very low concentration can be estimated &om the terminal falling velocity
of a single particle and as the slurry concentration increases, the slip decreases. In the
experiments of =Sumner et al., the terminal velocity of the coarse sand particles was about
0.12 mls. Considering the high concentrations and mean flow velocities used in the
experiments of "Sumner et al., we assume that there is no imponant diierence between
panicle and nuid velocities. In this case, it seems reasonable to model a two-phase flow as a
singlephase flow which has a gradient of viscosity and density across the pipe.

The main purpose of the present paper is to examine the hihence of the actual concentration
distribution in prediction of global parameters of mixture flow in verticle pipelines. In so
doing we also seek to establish the validity of a standard turbulence model using particles of
moderate diameter.

1. M A T H E M A T I C A L M O D E L

We consider fully developed steady turbulent flow of an incompressible Newtonian


mixture in a circular vertical pipe. As further simplifications to the equations we a m m e that
slip between solid and liquid velocities is unimportant, as was mentioned earlier, and the flow
is axially symmeuical without a mean circumferential velocity component.

The mathematical model is based on Reynolds' equations in which turbulent stresses are
calculated by means of the Boussinesq hypothesis. The problem of closure is solved
employing 'PHassid and Poreh turbulence model. This turbulence model has been vedied for
88 A. S. Bartosik and C. A. Shook

a wide range of parameters both for steady liquid flow, 'OGibson et al., liquid-solid, "Banosik
and Shook and unsteady flow, 42Banosiket al.

The final form of the governing equations is presented in cylindrical coordinates with the
x-axis lying on the symmetry axis of the pipe and "r" being the radial diiance. Taking into
account the aforementioned assumptions, the final form of Reynolds' equation for x-direction
is,

where the total pressure gradient in the main flow direction we treat as a known quantity.
The momentum equation for the "r" direction indicates that pressure has a constant value
Downloaded by [University of Waterloo] at 09:49 24 March 2015

across the pipe.

The turbulent shear stress in equation ( 2 ) is calculated in accordance with Boussinesq


hypothesis, as follows:

The final form of the k-L model is as follows:


- kinetic energy of turbulence:

-length scale:

Turbulent viscosity is related to the local value of kinetic energy of turbulence and the
modified length scale.

d e r e the damping function is determined as,

and

The final form of the mathematical model consists of a set of two partial differential
equations, namely equations ( 2 ) and (4) and the complementary relations (3) and (5HS).
Concentration Distribution to Predict Pipe Flow 89

'%assid and Poreh suggested the following values of the k-L model: C,=0.22; C,=0.416;
A;0.012; o,=1.0.
Calculations were performed using a finite difference scheme with relaxation factors, taking
into account the boundary conditions (U=k=O.O at the wall and d U l d ~ 0 . 0in symmetry axis)
and the criterion of convergence described by equation (9) where a,"is the value of at the
j' grid node aRer the n' iteration cycle and , ' that for the (n-I)' iteration cycle.
Computations of U and k-equations have been performed until convergence, defined by
equation (9). was achieved.

Solving the set of differential equations by means of the TDMA method with iterations.
Downloaded by [University of Waterloo] at 09:49 24 March 2015

"Roache, the velocity distribution and turbulence parameters are obtained. Numerical
calculations were carried out for 80 nodal points of the differential grid, distributed on the
radius of the pipe. Most of them were located near the waU. The number of grid points was
set experimentally to provide nodally independent predictions.

3. PREDICTIONS AND RESULTS OF MEASUREMENTS


The predictions for slurries of the three sands were carried out for fully developed flow in a
pipe of diameter D=25.8 mm for comparison with the measurements of "Sumner el al.
The assumption of constant density and viscosity across the pipe section is probably
acceptable for sufficiently small panicle diameters but these experimental data confirm that
for moderate particle diameters the dinribution of concentration (particularly in the "inner"
region near the wall) is not uniform, "Nasr-El-Din et al. and "Sumner et al. Other
measurements con6rming this observation have also been reponed, -Newin. Thus, the
mean velocity distribution in the inner region is determined by the wall shear stress, local
density, local viscosity and the distance 6 o m the wall. In the inner region of the boundary
layer, the energy exchange process are in a state of near equilibrium and as a result t l ~ elocal
production and the local dissipation of energy compensate. For this reason the inner region
of the flow is governed wentially by the local conditions, "Cebeci and Smith, 'Mrnze.
Therefore, use of a constant density and viscosity could produce sipticant error in
predictions in cases when the particle diameter is not sufficiently small.
The results of concentration distribution measurements for these slurries indicate that
concentration has a minimum value in the immediate vicinity of the wall. This phenomenon is
pronounced at higher concentration andlor larger panicles, as shown in Fig. I.
To interpret the experimental pressure drop data we introduce the concept that the minimum
distance from the wall at which we can expect the concentration to reach the measured value
is one half the particle diameter. We also assume that the concentration between the wall and
the first experimental point varies in a linear way from a value of zero at the wall. However,
it must be mentioned that the results of numerical predictions for the slurries considered here
did not show any essential differences between linear and parabolic earapolation to the wall
from the first measured concentration point. The appropriate position at the wall for the
concentration measurement of "Sumner et al. cannot be specified exactly but it must lie
between one half the particle diameter and 0.5 mm, since the domain of the concentration
sensor was found to be I mm. For particles of the size considered here, the effect of this
uncertainty on the concentration distribution is minor. The results of interpolation between
the rest of the experimental points are shown as solid lines in Fig. I.
A. S. Badosik and C. A. Shook
Downloaded by [University of Waterloo] at 09:49 24 March 2015

FIGURE 1 . Concentration distribution. a . h e sand. b.medium sand. c.coarse sand. Squares:


data ofMSumneret al. at C,=O. I ; Triangles: data of YSumneret al. at C,=0.3; Circles: data of
"Sumner et al. C,=0.4. Solid lies: smoothed relationships used in computations.
Concentration Distribution to Predict Pipe Flow 91

The local density was calculated from solids concentration using equation ( 10). For panicles
of the size considered here, direct measurements of slurry viscosity as a function of solids
concentration is impossible because of settling so that correlations reflecting the results of
experiments conducted by other researchers must be considered.

For spheres with narrow size d i b u t i o n s , "Thomas proposed equation (I I). Equation (I I)
provides, in effect, a lower limit to the slurry viscosity. Using sand panicles settling slowly in
a high viscosity mineral oil, %aniel's sand slurry viscosity measurements could be correlated
by equation (12) for C<0.35.
Downloaded by [University of Waterloo] at 09:49 24 March 2015

Daniel's sand particles were somewhat more angular than those used by "Sumner et al., so
that equation (12) provides an upper limit for the viscosity concentration function. This
function was used at concentrations beyond 35% by volume, however.
Numerical predictions of pressure gradient versus bulk velocity using the k-L turbulence
model and the actual concentration distribution are compared with experimental data for fine
sand, medium sand and coarse sand slurries in Fig. 2, Fig.3 and Fig. 4, respectively. In the
aforementioned figures are also shown results of predictions for a constant density and
viscosity across the section (dashed lines) where both bulk coefficients p and p are established
from equations (lo), and (1 I) andlor (12).
With the assumption of uniform density and viscosity across the pipe, the discrepancy
between predictions and experimental data is high for the h e , medium and coarse sand
sluries (dashed lines in Fig. 2, Fig. 3 and Fig. 4) and the discrepancy has a clear tendency to
increase with increasing concentration of solid particles.
AAer introducing the actual concentration distribution, the numerical predictions for the h e
sand flow are very close to the measurements when equation (I I) is used, as shown in Fig.2,
as a solid l i e . However, it must be mentioned that differences between equations (I I) and
(12) are slight for low concentrations. For this reason there are only small differences
between predictions using viscosity functions (1 I) or (12). This results from the fact that in
the turbulent boundary layer the laminar viscosity coefficient is most important in the viscous
sublayer and pan of the buffer layer. In this region the concentration is always lower than the
mean value, "Nan-El-% et al. and "Sumner et al.
Predictions for the medium sand and coarse sand slurries (using equation (12)) are much
better than in the case of constant density and viscosity (solid line in Fig.3 and Fig 4).
However, calculations with the actual concenwation dishibution indicate a systematic error.
The maximum average relative error in bulk velocity in the range of solids concentration
C,=O. 1-0.4 is about 8% for the medium sand sluny (at C, =0.4), while for the coarse sand
slurry it is about 7% (at C,=0.3). The relative errors are highest at the lowest velocities,
possibly because the precision of the experimental pressure drops becomes the limiting factor.
Downloaded by [University of Waterloo] at 09:49 24 March 2015
Downloaded by [University of Waterloo] at 09:49 24 March 2015
A. S. Badosik and C. A. Shook
Downloaded by [University of Waterloo] at 09:49 24 March 2015

BULK VELOCITY, m/s

FIGURE 4. Pressure gradient versus bulk velocity for coarse sand slurry, C,=O. 1; 0.3; 0.4.
Squares: data of "Sumner et al. Dashed lines: predictions for constant solids concentration
distribution; Solid lines: predictions for actual solids concentration distribution [C=f(r)].
Concentration Distribution to Predict Pipe Flow 95

It is also useful to compare the predicted and measured velocity profiles for fine, medium and
coarse sand slurries with measurements of % m o e r et al. However, it is appropriate to
reiterate that the distance from the wall is somewhat uncertain for the measurements closest
to the pipe wall. As noted earlier, the appropriate distance is between one half particle
diameter and 0.5 mm. For this reason, in Fig. 5, Fig. 6 and Fig. 7 we locate the first
measured point at 0.25 mm from the wall and therefore do not interpret quantitatively the
predictions of velocity at the wall.

The predicted velocity profiles for h e sand slurries are somewhat more curved than the
experimental data. This is shown more clearly for higher bulk velocities and higher
concentration (see Fig. 5b and 5c at U,=6.4 mls). Nevertheless, the maximum relative
difference between measured and predicted local velocity in the range of solids concentration
C,=O. 1-0.4 is less than 5%. Furthermore, the predicted velocity profile for fine sand is flaner
Downloaded by [University of Waterloo] at 09:49 24 March 2015

using the actual concentration distribution than for constant physical properties across the
pipe. This fact was noted by %mner et al. for all their experiments and is discussed later in
this paper.

In contrast with the results obtained with the fine sand, predictions of velocity profiles for
coarse sand slurry differ qualitatively e o m the measurements at C,=0.3 and C,=0.4 (Fig. 7).
The experimental data suggest that radially segregated flow could exist, with a maximum
velocity located near the wall and uniform core region. Alternatively, particle friction with
the measuring device may have been higher in the centre of the pipe. The qualitative
difference between predicted and measured velocity prolles limits application of the
mathematical model to estimates of the relationship -dPldx=f(U,). On the basis of the
experimental data we therefore conclude that coarse sand slurry flow is more complex than
that slurries of fine and moderate particles. Additional stresses due to pamcle-particle
interactions could play a significant role in the core region and stresses of this type were
detected in Couene flow by 'PBagnold.

The predicted velocity prolles for fine, medium, and coarse sands slurries are all flatter using
the actual concentration distribution than for constant density and viscosity across the pipe.
As an example, in Fig. 8 we show predicted dimensionless velocity profiles for fine sand
slurry, for water, and the experimental data of MSumner et al. at C,=0.4. It is evident that
the predicted s h y and water curves are not parallel and the highest differences are
observable in the inner region, excluding R*=O-12.

It is appropriate to mention that we also performed numerical predictions with the gravity
component in equation (2) evaluated as p,g. If the bulk density for the d o l e cross section
(p,) were used, the effect on predicted -dP/dx=f(U,), is small compared to predictions for
original form of equation (2). For medium and coarse sand the maximum dierence in bulk
velocity is less than 5%, while for fine sand the difference is close to zero. This discrepancy
is directly dependent on particle diameter. However, for the medium and coarse sands slurries
there is an observable difference in velocity profile. As an example, in Fig. 6c we show
numerical predictions using the bulk density (dashed lines) in the last component of
momentum equation (2).
A. S. Bartosik and C. A. Shook
Downloaded by [University of Waterloo] at 09:49 24 March 2015

FIGURE 5. Velocity profiles for fine sand slurry. Squares, triangles and circles: data of
"Sumner el al. a.C,=O.l; U,=2.74; 4.45; 6.30 mls. b.q=0.3; U,,=2.64; 4.58; 6.40 mls.
c.C,=0.4; U,=2.69; 4.46;6.44 mls. Solid Lines: predictions for actual solids concentration
distribution [C=f(r)].
Concentration Distribution to Predict Pipe Flow
Downloaded by [University of Waterloo] at 09:49 24 March 2015

FIGURE 6. Velocity profile for medium sand slurry. Squares, triangles and circles: data of
YSumner et al. a.C,=O. I; U,=2.63; 3.58; 5.29mIs. b.C,=0.3; Ub=2.80; 3.71; 5.29mls.
c.C,=0.4; U,=2.80; 3.66; 5.33 mls. Solid limes: predictions for actual solids concentration
distribution [C=Rr)]. Dashed lines: prediction for constant solids conceotration distribution.
A. S. Bartosik and C. A. Shook
Downloaded by [University of Waterloo] at 09:49 24 March 2015

FIGURE 7. Velocity profile for coarse sand slurry. Squares, mangles and circles: data of
Y
Sumner et al. a.C,=O. I ; U,=2.78; 4.43; 6.24 mls. b.C,=0.3; U,=3.10; 5.09; 5.89 mls.
c.C,=0.4; U,=3.14; 5.25 m/s. Solid lines: predictions for actual solids concentration
distribution [C=qr)].
Concenrntion Distribution to Predict P i ~ Flow
e
Downloaded by [University of Waterloo] at 09:49 24 March 2015

FIGURE 8. Dimensionless velocity profile for 6ne sand slurry at Cb=0.4. Squares: data of
Y
Sumner et al. Triangles: predictions for slurry. Circles: predictions for water.

In d e w of the discrepancy between predictions and experiments and the inherent difficulty of
obtaining precise measurements of friction for venical flows we have also examined the
experimental data of *Newin et al. They presented measurements for a coarse sand slurry at
solid concentrations of 15% and 25% (by volume) in a one inch diameter pipe. The mean
particle diameter and density were d,,=710 pm and p,=2600 kglm'. As mentioned
previously, if the mean density were used in the gravity component of equation (2), the
effect on predicted -dP/dx=f(U,) is small compared to predictions for the original form of
equation (2). Because of the small difference, we can use the present results to reconsider
data in the literature which used the mean density to correct observed pressure gradient.
&4
Newin et al. reponed that headlosses for coarse panicles suspensions were essentially the
same as water. Within the limits of experimental error this also seems to be true of the data
of ?hmner et al. Because we do not have the results of concentration distribution
measurements for Newin's coarse sand slurry we have assumed that the distribution of
concentration was the same as for the coarse sand in "Sumner et al. experiments.

In Fig.9 we show experimental frictional pressure gradients of UNewin et al. (estimated as


five points, shown as squares) and numerical predictions using the half panicle diameter
concept (solid lines). We also show predictions for a constant concentration distribution
(dashed lines) where the mkture viscosity was calculated by equation (12). AAer
approximating the concentration distribution with the half particle diameter concept, the
calculations show good agreement with measurements. However, it is appropriate to
emphasize that experimental date of -Newin et al. cover a smaller range of bulk velocity
than the experiments of YSumner et al. The small discrepancy between Newin's and
100 A. S. Bartosik and C. A. Shook

Sumner's results could be attributed to panicle shape differences and to precision of the
measurements at the lowest velocities.
Downloaded by [University of Waterloo] at 09:49 24 March 2015

BULK VELOCITY, rn/s


FIGURE 9. Pressure gradient versus bulk velocity for coarse sand slurry at C,=O. 15 and
C,=0.25. Squares: data of "Newin et al. Dashed lies: predictions for constant solids
concentration distribution; Solid lines: predictions for actual solids concentration distribution
[c=Wl.

The half panicle diameter concept and the turbulence model thus combine to predict
frictional pressure gradients which decrease as the pamcle diameter increases, in vertical flow.
The experiments of uNewin et al. were broadly in agreement with this conclusion. Because
of possible additional stresses and turbulence damping implied by the experiments of
Y
Sumner et al., it seems that the model and the half particle diameter concept are i o n
useful for coarse panicles in terms of pressure gradient predictions.

Horizontal sluny flows have been studied much more extensively than vertical ones and, in
conditions where the flow is close to axisymmetric (i.e. line particles, smaU density
Concentrarion Distribution to Predict Pipe Flow 101

dserences between solid and liquid phases and/or very high flow velocity) frictional pressure
gradients are considered to be proportional to the density of the mixture. This simplification
is sometimes called pseudohomogeneous flow approximation. It evidently conflicts with both
model predictions and experimental observations for coarse particle slurries in vemcal flow.

4. DISCUSSION OF RESULTS

The numerical computations of fully developed turbulent mixture flow in vertical pipes for
slurries of fine and moderate particles diameter indicate that it is possible to extend the
application of standard turbulence model to predict global parameters if one takes into
account the actual concentration distribution. This concept is also useful for slurries of
coarse particles in terms of -dP/dx=qU,), although there may be a limiting pamcle diameter
for which the half panicle diameter concept and a standard turbulence model gives
Downloaded by [University of Waterloo] at 09:49 24 March 2015

reasonable predictions.

Any change of properties in the viscous sublayer and/or buffer layer affects ~ r b u l e n t
properties in the rest of the sluny flow. The half panicle diameter concept. introduced to
interpret the measurements of concentration, results in changes of turbulent properties in the
whole boundary layer, although density and viscosity change only in the immediate vicinity
of the wall. Thus, if the diameter of the particle increases, the turbulence propemes will also
change, compared to sluny flow with a constant concentration distribution. The final result
of this changes is a decrease in the turbulence level.

"Gore and Crowe have suggested a critical parameter (the ratio of pamcle diameter to
turbulent length scale associated with the fluid phase) which is closely connected to the level
of turbulence intensity. The critical value of this ratio is equal to 0. I and defines a particle size
above which an increase and below which a decrease of turbulence intensity in the canier
phase will appear. In case of gas-solid flow this criterion appears to be useful. However, for
liquidsolid flows we can find examples which are not in coincidence uith this criterion. If
the length scale is estimated as 0. I R, as it oflen is for turbulent flows, the ratios in the present
work are 0.14 ( h e sand), 0.36 (medium sand) and 0.60 (coarse sand). The measurements in
medium and coarse sand slurries suggest that there is significant suppression of turbulence
although the ratio of particle diameter to turbulent length scale is significantly greater than
0.1.

The computational model has s h o w that the flaner velocity distributions measured by
"Sumner et al. are consistent with the concentration variations they observed. The model
suggests that there are two reasons for these flaner distributions: the variation of fluid
properties at the wall and the variation of the gravity component in the momentum equation.
If additional stresses or increased damping of turbulence kinetic energy occur these e5ec1s
must tend to compensate for each other in vertical flows.

Although no simple criterion has been obtained for defining the extent of the region of rapidly
varying viscosity and density in terms of particle diameter, it seems probable that the region at
the pipe wall requires special treatment in any model of pipe flow. The pseudohomogeneous
flow approximation would thus appear to be restricted to comparatively fine pamcles (finer
than 175 pm) old low concentrations. This is particularly important for development of
horizontal flow models, such as the two layer model of "Wilson, as modified for example by
12
Gillies et al.
A. S. Bartosik a n d C. A. Shook

NOMENCLATURE

3 --- turbulence model constant


concentration o f solid particles, volume fraction
turbulence model constants (i=D, p)
Ci
d - panicle diameter, m
d,, - mass median particle diameter in a mixture, m
D - pipe diameter, m
fp - damping function in turbulence model
g -acceleration due to gravity, mls
k - kinetic energy o f turbulence, mls
1. - length scale in turbulence model, m
I' -
static pressure, N/m2
r - distance from symmetry axis, m
Downloaded by [University of Waterloo] at 09:49 24 March 2015

R -pipe radius, m
R' - dimensionless position, yUT/v
Re - Reynolds number
I* - characteristic time-response parameter, s
-
u',v' fluctuating components of velocity in directions x and r, respectively, mls
U -velocity component in axial directioo, mls
U' - dimensionless velocity, U N ,
U, -
friction velocity, ( I J ~ ) ' ~mls
,
x -axial coordinate, m
y -distance from the pipe wall, m
& - rate o f dissipation of turbulent energy, m'/s3
)I - coefficient o f dynamic viscosity o f slurry, Ns/ma
p - density o f slurry, kglm'
4 - effective Prandtl-Schmidt number for k
@ -general dependent variable (@=U,k)

Subscripts and Superscripts


b -bulk
j - nodal point
-
L liquid phase
n - number o f iteration
S -solid phase
t -turbulent
- time averaged value
REFERENCES
I. Launder, B.E. and Spalding, D.B., 1972, The Turbulmce Models and their Aapplication to the
Prediction of Internal Flows, Hear amfFluidFlow.wl. 2. pp. 43-54.
2. Launder. B.E..Morse, A . Rodi, W., Spalding, D.B., 1972. The Prediction of Rree Shear Flaws
- A Comparison of the Performance of Six Turbulence Models. Proc. NASA Conj: on Free
Turbulent Shear Flons, Langley, MI. 1, pp. 361-426.
3. Patel. V.C., Rodi, W.. Scheuerer, G., 1985, Turbulence Models for Near Wall and Low
Reynolds Number Flows: A review, A M J.,wl. 23. pp. 1308-13 19.
4. Lee. S L and Durn. F., 1982. On the Motion of Particles in Turbulent D u a Flows, Inr. J.
Mulrrphase Flow. MI. 8, pp. 125-146.
Concentration Distribution to Predict Pipe Flow 103

5. Tsuji, Y., Morikawa, Y.. Shiomi. H.. 1984. LDV Measurements of an Air-Solid Two-Phase
Flow in a Venical Pipe. J. FluidMeetr., vol. 139, pp. 417-434.
6. Tsuji. Y.. Morikawa. Y.. Tanaka, T., Karimine. K.. Nishida, S., 1988. Measurements of an
Axisymmaric Jet Laden u l t h Coarse Particles, Inl. J. Muhiphare Flow, vol. 14, pp. 565-574.
7. Modarress, D., Wuerer. J.. Elghobadu. S., 1984, An Experimental Study of a Turbulent Round
Two-Phase Jet, Chem. Eng. Commun., wl. 28, pp. 341-354.
8. Wang, S.K., Lee. S.J.. Jones, J.O.C. and Lahey, J.R.T.. 1987. 3-D Turbulence Stmaure and
Phase Distribution Measurements in Bubbly Two-Phase Flows, k r J. Mulriphase Flow. MI.13.
pp. 327-343.
9. Zisselmar, R. and Molerus. 0.. 1979, InvesIigation of Solid-Liquid Plpe Flow u l t h Regard to
Turbulent Modification. Chem. Engng J . , vol. 18, pp. 233-239
101. Noun, J.M., Whitelaw, J.H., Yianneskis, M., 1987, Panicle Motion and Turbulence in Dens
Two-Phase Flows, lnr. J. Muillphase Flow, vol. 13, pp. 729-739.
II Launder. B.E. and Spalding. D.B.. 1972, MarhemahcalModels o/Turbulence, Academic Press.
London-New Yo&.
Downloaded by [University of Waterloo] at 09:49 24 March 2015

12 Spalding, D.B.. 1983. Turbulence Models - A Ledure Course. Lmperial College, Mech Engng.
Dept., CFD18214 Rept.
13 Hetsroni, G., 1988. Panicles-Turbulence Interaaion. 3rd Int. Symposium on Liquid-Solid
Flows, pp. 1-12. Ed. M.C. Roco. ASME.
14 Hasroni. G., 1989. Panicles-Turbulence Interaaion, Int. J. Multiphase Flo~v.wl. 15, pp.
735-746.
15 .
Danon. H., Wolfrhtein. M.. Henron~.G 1977, Numerical Calculations ofTwo PhaseTurbulent
Rowid Jet. Inr. J. Mulliphare Flow, wl. 3, pp. 223-234.
16. Melville, W.K. and Bray, K.N.C., 1979. A Model ofthe Two Phase Turbulent Ja. h ~ rJ. . Heor
Mms Transfer, MI. 22, pp. 647456.
17. Genchev, ZH. and Karpuzov. D.S., 1980, E f f m of the Motion of Dust Panicles on Turbulence
Transport Equations, J. FluidMech., wl. 101, pp. 833-842.
IS. Elghobashi, S.E. and Abou-Arab. T.W., 1983, A Two Equation Turbulence Model for
Two-Phase Flows, Physics of Fluidr. MI.26. pp. 93 1-938.
19. Paurahmadi, F. and Humphrey, A.C., 1983, Modelling Solid-Fluid Turbulent Flows ulth
Application to Predicting Erosive Wear. PCH Phpico-Chemical Hydro+namics, vol. 4, pp.
191-219
~.
~

20. Mostafa. A.A. and Elghobashi, S.E., 1985, A Two-Equation Turbulence Model for Jets Flows
Laden ulth Vaporizing Dropla, Inr. J. Multiphase Flow. wl. II,pp. 515-533.
21. Chm. C.P. and Wood. P.E.. 1986, A Turbulence Closure Modelling of Dilute Gas-Panicle
Axisymmaric Jb, AIChEJ.. vol. 32, pp.163-166.
22. Lee. K B , and Chung M.K.. 1987, Refinement ofthe fixing Length Model for Prediaion of
Gas-Panicle Flow in a Pipe. Inr. J. Mulriphase Flow, vol. 13, pp.275-282.
23. Besnard, D.C. and Harlow. F.H., 1988, Turbulence in Multiphase Flow, 611.J. Multiphare
Flow, vol. 14, pp. 679499.
24. Issa. RI.. 1988. Prediaion of Turbulent Stratified Two-Phase Flow in Inclined Pipes and
Channels, /,a. J. Mulriphare Flow, vol. 14, pp. 141-154.
25. Rizk. M.A. and Elghobashi, S.E.. 1989, A Two-Equation Turbulence Model for Dispersed
Dilute Confined Two-Phase Flows, Inr J Mulriphase Flow. wl. 15, pp. 119-133.
26. Roco, M.C. and Balakrishnan. N., 1985, Multi-Dimensional Flow Analysis of Solid-Liquid
Mixtures, J. Rheologv. MI.29, pp. 431-456.
27. Abou-Arab, T.W. and Roco. M.C., 1988, Solid Phase Contribution in the Two-Phase
Turbulence Kinaic Energy Equation, Proc. 3rd Jnt. Symp. an Liquid-Solid Flows. ASME, pp.
13-28. ed. M.C. Roco.
28. Crowe, C T , and Sharma, M.P.. 1978, A Novel Physico-Computational Model for Quasi One
Dimensional Gas-Particle Flows, Tram. ASME, J. Fluids Engng., vol. 100, pp. 343-349.
29. Crowe, C.T. and Choi. H.N.. 1979. Two Dimensional Numerical Model for Steam Water Flow
in a Sudden Contraaion. in Two-Phare Momentum Hear &Mars - . ed. F. Durst. G.
Tramfer.
Tsiklauri, N. Afgan. Hemisphere. Washington.
104 A. S. Bartosik and C. A. Shook

30. Lee, J. and Wiesler. M A . . 1987. Theory of Transverse Migration o f Panicles in a Turbulent
Two-Phase Suspension Flow Due to Turbulent Diffusion-I, Int. J. Multiphare Flow, MI. 13, pp.
99-11],
31. Kallio. G.A. and Reeks. M.W.. 1989, A Numerical Simulation o f Panicle Deposition in
Turbulent Boundary Layers. 611.J. Multiphace Flow. MI.IS, pp. 433.446.
32. Banosik, A.S., Sobocikki, R.A., 1988, Energaic Property of Turbulent Hydrotrampon of
Slurry with Optimal Concentration o f Fluxer, Proc. 6th Inr. Conj on Trnmporr and
Sedinrentnrion of SolidParticles. Wroclaw (Poland), No. 176, pp. 57-66.
33. Nasr-El-Din, H., Shook, C.A.. Colwell, J., 1987, The Lateral Variation ofSolids Concentration
in Horizontal Slurry Pipeline Flow. hr. J. Multiphme Flow, vol. 13. pp. 661-670.
34. Sunmer. R.J.. McKibben, M., Shook. C.A., 1990. Concentration and Velocity Distribution in
Turbulent V e n i a l Slurry Flow. J. SolidLiqtidFlow, vol. 2. No. 2, pp. 33-42.
35. Lumley, J.L., 1976, in Turbulence, Ed. P. Bradshaw. TwwPhdw and NowNewtonian Rows.
pp. 290-295. Berlin, Springer-Verlag.
36. Shook, C.A. and Row, M.C.. 1991. Slurry Flow: Principles nnd Pracrice,
Downloaded by [University of Waterloo] at 09:49 24 March 2015

Bunenvonh-Heinemam, Boston.
37. Yianneskis, M. and Whitelaw, J., 1983, Velocity Charaueristics of Pipe and Je2 Flows with
High Panicle Concentration, in ASME Symposium on Liquid-Solid Flows and Erosion Wear in
lndusrrinl Equipment, pp. 12- 15.
38. Cox, P. and Mason. S.. 1975. Flow o f Fluids in Pipes in the Presmce of Suspended Partides.
Rheology of Suspensions.
39. Hassid. S. and Poreh. M.. 1975. A Turbulent Enemv - Reduction. J.
-. Model for Flows with Drae.
Fluids Engng., vol. 97, pp. 234i241.
40. Gibson. M.M.. Soaldine. . .
-. D.B.. Zinser. W.. 1978. Boundaw Laver Calculations Usinz -
assi id-~oreh 6 n e ' ~ ~ u a t i o~n n e &~ o~d e l .Lerren i'n Heat and Mars Transfer. vol. 5. pp.
73-80.
41. Banosik, AS. and Shook C.A.. 1991, Prediction of Slurry Flow with Non-Uniform
Concentration Distribufion - Comparison of the Performance o f Four Turbulence Models. Proc.
711, lnl. Corg on Nurne~cal~ e r h a d sin Laminar and Turbulenr Flow, Stanford. CA, Pan I.
wl. 7. pp. 277-287.
42. Banosik. AS.. Sobocihrki, R.A. and Wanik, A.J., 1987. Prediction o f Fully Developed
Pulsating Flow with the Aid of Low-Reynolds Number Turbulence Models. Proc. 5th 1111.Conj
on NumekalMethods in Lnminnr and Turbulent Flow, Swansea, vol. 5, Pan 1, pp. 521-532.
43. Roache, P.J., 1982, Compurntionnl Nurd Dynamics, Hermosa Publ., Albuquerque.
44. Newin. D.M.. Richardson. J.F.. Gliddon, BJ.. 1961. Hydraulic Conwying of Solids in Vertical
Pipes. Trnns. Imtn. Chem. Engrs.. MI.39, pp. 93-100.
45. Cebeci. T, and Smith, A.M.O., 1974, Ana&sis ojTvrbulenr Boundary Layers. Aadmnic Press.
New York.
46. Hinze. J.O.. 1975, Turbulence, 2nd ed..Mffiraw-Hill.
47. Thon~as.D.G., 1965, Transpon Charaueristics of Suspension: V m . A Note on the Viswsity o f
Newonian Suspensions o f Uniform Spherical Panicles, J Colloid Sci., vol. 20, pp. 267-277.
48. Daniel. S.M., 1965, Flow of Suspensions in a Redangular Channel, Ph.D. 7hesis. University of
Saskatchewan.
49. Bagnold. R.A.. 1954. Experiments on a Granty-Free Dispmsion of Large Solids Spheres in a
N m o n i a n Fluid Under Shear. Proc. Roy. Soc.. Ser. A, ~ 1 . 2 2 5 pp.
, 4963.
50. Gore. R.A. and Crowe, C.T., 1989, Effecl o f Particle Size on Modulating Turbulent Imensity,
Inr. J. Multiphme Flow, MI.15, pp. 279-285.
51. Wilson. K C . . 1976, A Unified Physically Based Analysis o f Solid Liquid Pipeline Flows, Proc.
41h lnr. Cottj on Hydrnulic Trnnsporr of Solids in Pipes, Cramield, BHRA Fluid Engng., Paper
Al.
52. Gillies, R.G., Husband, W.H.W. and Small, M., 1985, A Study of Conditions Arising in
Horizontal Coarse Slurry Shon Dinance Pipdining Practice, Saskatchewan Rerearch Council,
R-833-Z-C-85 Rept.

You might also like