Académique Documents
Professionnel Documents
Culture Documents
Proceeding
Session: 23
Multiphase Flow/CFD
Internal Number: 195
Pipe Flow of Solid-Liquid Suspensions with a Broad Size Range
-2-
disperse from the regions of higher concentration, due to random motions. This behaviour is strongly controlled
by the properties of the particles and by the geometry of the pipe.
The settling-dispersion model was initially formulated by OBrien for the unidimensional situation [3] and, since
then, was used and improved by other authors [4,5,6]. In previous works we have extended the model to the twodimensional situation. Additionally, we have also shown elsewhere [7] that a better fit to the experimental data is
obtained when the model parameters (settling velocity and dispersion coefficient) are considered dependent on
the local conditions in the cross-section of the pipe, namely on local particle concentration. This model can be
mathematically translated by the convection-dispersion equation which, in steady state, can be written as:
) ( )
D c + v c = 0
(1)
where D is the dispersion tensor, c is the local particle concentration (v/v) and v is the velocity vector.
This equation can be solved, for an horizontal pipe, in cylindrical coordinates (r,,z). The boundary conditions
are described in more detail elsewhere [7].
In the present work we have reformulated the model to adapt it to the more realistic case where the solids exhibit
a range of diameters. Rewriting equation (1) for the particles of class i, diameter dpi, for a system with N size
classes, we have:
) ( )
Di c i + v i c i = 0
(2)
where D i is now the dispersion tensor for species i, v i is the velocity vector for the particles with a diameter dpi
and ci is the local volumetric concentration of those particles.
To obtain the distribution of the solids it is necessary to solve, simultaneously, the N equations for the different
particle size classes of the solid phase. The simplifications introduced, including the assumption of isotropic
dispersion, and the boundary conditions, were the same that were adopted for monodispersed systems [7]. Thus,
the simplified convection-dispersion equation, in cylindrical coordinates, for the generic species i, is:
D c
D c
vsi
c
c cos vsi
v cos ci
1 Di ci
+ i i +
+ i
+ vsi sin i + si
+ i i + 2
r r
r
r r r
r
r
(3)
2
2
ci
Di ci
Di 2 + 2
=0
r
r 2
ci sin
where Di is the dispersion coefficient for species i and vsi is the settling velocity for the same species.
The total volumetric concentration of the solids, for each nodal point of the cross-section of the pipe, is given by
the sum of the concentration of all the species:
N
c = ci
(4)
i =1
Once again the model parameters were considered dependent on local conditions. Additionally, since the settling
tendency of each particle will be influenced by the presence of the other species, the equation proposed by
Masliyah [8] has been used to take into account those interactions:
N
v si = (1 ci )v' oi (1 c ) i ck v' oi (1 c ) k
n 1
n 1
(5)
k =1
k i
In this equation n i is a function of the particle Reynolds number, v 'oi is the settling velocity of the particles with
diameter dpi , for infinite dilution.
The correlation selected for the calculation of the dispersion coefficient was the same that has been used for
monodispersed systems [7], which was adopted by analogy with the liquid-liquid diffusion process:
Di = D = D0 (c M c ) 2
n
(6)
-3-
In fact, as referred in previous publications, the empirical constant D0 does not depend on the type of particles
and, though n 2 may be a function of the particle characteristics this relation is not yet explicit. In equation (6) cM
is the packing concentration of the particles.
The system of non-linear algebraic equations was solved by the Brent algorithm [9], which is a software package
that uses a modification of the discrete Newton method.
In previous works [10] we have developed a pressure drop correlation that takes into account the solids
distribution in the pipe cross-section. That equation considers that the total head-loss is the result of three
contributions, as happens in a fluidised bed:
This equation has been fitted to a set of 60 experimental measurements leading to the following correlation:
P
= 5.825 10 3 E c + 382.447 E v + 5.539 10 2 E p
L susp
(7)
Once the solids distribution in the pipe is known, the pipe cross-section will be horizontally discretized, and
equation (7) will be applied to each slice of pipe, the total pressure drop being the sum of these contributions. In
this paper we will also present, for comparison, pressure drop values calculated using the Durand equation.
3 - S im u lat ed C o n ce nt ra t ion P ro f i le s
In this section we will present a selection of the simulated results produced. We will start by showing two
examples of solids distribution in the cross-section of a horizontal conveying pipe, that have been obtained
assuming the particulate phase to be monodispersed, that is, associating a single diameter to the solid phase. For
those examples we will show, in Fig. 1, both the 1D and 2D distributions, and also the calculated pressure drop
values using either the Durand equation or equation (7). The 1D distributions and the pressure drop values will
be compared with experimental data from the literature.
The remaining figures show the simulated distributions obtained supposing the particles to be polydispersed
(with a range of particle diameters), which have been calculated using equation (3). The tests presented show the
influence of the particle size range on the solids distribution in the pipe, namely on the extent of the segregation
that can develop in the cross-section of the pipe. The influence, on the calculated profiles, of the size of the class
ranges which have been considered during the calculations, will also be discussed. The distributions obtained by
assuming the particles to be polydispersed will be compared with the results from the monodispersed model.
Furthermore, a comparison with experimental measurements will also be attempted, though the experimental
data for polydispersed systems is rather difficult to obtain, and thus, quite scarce in the literature.
3.1- Monodispersed Systems
Table 1 summarises the operating conditions of the experimental tests that we have used as input for the
simulation studies [11]. The calculated profiles were obtained using the model for monodispersed systems.
TEST
1
2
TEST
1
2
-4-
The value of Do was adjusted for each test in order to obtain the best agreement between the simulated and the
experimental profiles. However, it was obvious a coherent variation of that parameter as a function of the
conveying velocity, larger conveying velocities implying larger values for Do.
In Fig. 1 the simulated profiles for the operating conditions of the tests in Table 1 are compared with measured
profiles. The comparison is based on the vertical variation of the chord average concentration, since most of the
experimental data available is presented in that manner. The deviation between the simulated and experimental
profiles is also shown in each graph (Dev.). Looking at those graphs it is obvious that, generally speaking, a
good agreement between simulated and experimental profiles is obtained. This adjustment is better for the
smaller particles and, other studies have shown that it is also better for the lower concentrations. For these tests,
the pressure drop calculated using the simulated profiles is also very near the experimental value, more close
than the values obtained with the Durand equation.
Making a more detailed analysis of the results in Fig. 1, we can conclude that the model is capable of translating
the influence of the operating conditions on the solids distribution in the cross-section of the pipe. For instance,
the model is adequate to describe the effect of the diameter of the particles on the solid phase distribution in the
cross-section of the horizontal pipe: larger settling velocities, i.e., larger particles, correspond to less uniform
distributions for the same conveying velocity. Fig. 1 shows also the 2D profiles that can be obtained with the
model, giving a full picture of the solids distribution in the cross-section of the pipe.
TEST 1
90
TES T 1
45
120
S im u l.
E xp e r.
Concentration (%Vol.)
40
Pressure D rop (P a /m ):
M e a s u re d = 1 8 2 2 .4 5
S im u la te d = 1 9 8 4 .2 0 5
D u ra n d E q .= 1 6 6 2 .9 3
D ev= 1 0 .4 %
35
30
60
3.00%
150
30
5.00%
25
10.00%
20
180
15
15.00%
10
20.00%
25.00%
330
210
0
0.0
0.2
0.4
0.6
0.8
35.00%
1.0
y/D
240
40.00%
300
270
TEST 2
90
TEST 2
60
120
S im u l.
E xper.
Concentration (%Vol.)
50
Pressure Drop (P a /m ):
M e s u re d = 2 07 4 .6 2
S im u la ted = 2 05 8 .53
D u ra n d E q .= 18 4 9.7 1
Dev= 1 7 .9%
40
60
3.00%
30
150
5.00%
30
10.00%
0
180
15.00%
20
20.00%
10
25.00%
330
210
0
0,0
0,2
0,4
0,6
0,8
1,0
y/D
240
35.00%
40.00%
45.00%
300
270
-5-
3 . 2 - Po lyd i sp e rs ed S y s t e m s
As referred previously, the Settling-Dispersion model has been adapted in order to incorporate the handling of a
solid phase with a wide particle size distribution. In this case, the solid phase size distribution is divided in a
certain number of fractions, an average diameter being attributed to each fraction (species diameter).
Fig. 2 presents the profiles (1D and 2D) obtained considering the particle size distribution to be described by
three size classes (ternary system). Tables 2 and 3, summarise the operating conditions and model parameters
used in the simulation of the multispecies systems (tests 3 and 4). The diameter of the pipe was 410-2 m. The
graphs in Fig. 2 refer both to the total solids distribution and to the distribution of the individual species. Tests 3
and 4 correspond to different particle size ranges, and it is apparent from the analysis of Fig. 2 that the larger the
size range (test 4) the more evident is the segregation that pushes the smaller particles towards the top of the pipe
(compare Figs. 2 (c) and (f) ). Altough the ratio Do/vo was kept the same in both tests, which means that the
conveying velocity has been increased for the wider distribution, with a larger average settling velocity,
segregation is more pronounced in the second case and the total solids distribution is less uniform, the smaller
particles showing a maximum in the central zone of the cross-section. The model calculates also the total solids
load in the pipe and the average concentration of each species, as shown in table 3.
Table 2: Suspension characteristics/Ternary systems: Tests 3 and 4
TEST 3
TEST 4
SPECIES
1
2
3
1
2
500
300
250
750
500
Diameter-dpi (m)
Terminal Settling Velocity - voi (m/s)
0.0402
0.0190
0.0142
0.0678
0.0402
Particle Reynolds Number- Repi
7.13
2.03
1.26
18.05
7.13
0.0245
0.0407
Average Terminal Settling Vel. - vo (m/s)
342
3
250
0.0142
1.26
500
Test
3
4
1
5.94
6.12
2
9.31
8.45
The solids distributions obtained assuming a polydispersed particulate phase have also been compared with the
profiles calculated supposing a monodispersed solid, with a particle diameter equal to the average diameter of
the multispecies system. Table 4 summarises the conditions used in those studies. The size range of the particles
in test 5 is the same as in test 3 (see table 3), for test 6 the size range is equal to the one used in test 4. Moreover,
table 4 presents data for binary and ternary particulate phases, since the results that will be presented in Fig. 3
were obtained applying the polydispersed model both to a two and three species system. Fig. 3 compares the one
and two dimensional distributions obtained for those two situations with the results for the monodispersed
system. Looking at this figure, the differences between the results obtained with the polydispersed and
monodispersed models are obvious, even though the maximum size range in the polydispersed system is only 3.0
(dpmax/dpmin) in test 6. These differences increase when the size range increases (compare Figs. 3 (a) and (c)),
being more pronounced in the regions of higher concentration. For the particle size ranges tested, going from a
binary to a ternary distribution does not alter the overall solids distribution significantly.
Table 4: Operating conditions and suspension characteristics: Tests 5 and 6
TEST 5
TEST 6
Monod
Bin.
Tern.
Monod
Bin.
Average Concentration (%vol.)
21.09
21.16
21.11
20.00
20.09
Particle Reynolds Number
2.89
7.30
0.0226
0.0272
0.0245
0.0407
0.0410
Average Terminal Settling Vel. - vo (m/s)
Average Particles Diameter (m)
Do (m2/s) 104
Do / vo (m) 102
345
4.81
2.04
350
5.55
2.04
340
5.00
2.04
505
8.30
2.04
502
8.36
2.04
Tern.
20.13
0.0407
507
8.31
2.04
-6-
TEST 4
TEST 3
90
90
4.00%
120
150
60
5.50%
6.00%
4.50% 2.00%
7.00%
5.00%
8.00%
5.50%
4.00%
9.00%
120
Species1
Species2
Species3
4.50%
Species1
Species2
Species3
60
4.00%
30
4.70%
4.80%
5.00%
4.90%
2.00%
150
30
3.00%
7.00%
8.00%
4.00%
0
180
180
6.00%
7.00%
8.00%
10.00%
10.00%
210
330
330
210
13.00%
13.00%
240
15.00%
240
300
16.00%
18.00%
300
270
270
90
90
120
120
60
60
10.00%
12.00%
30
150
30
150
15.00%
15.00%
19.00%
20.00%
0
180
180
25.00%
26.00%
330
210
330
210
30.00%
30.00%
32.00%
240
240
300
35.00%
300
270
270
Species1
Species2
Species3
Total
30
Species1
Species2
Species3
Total
35
Concentration (%Vol.)
Concentration (%Vol.)
35
25
20
15
10
5
30
25
20
15
10
5
0
0,0
0,2
0,4
0,6
y/D
0,8
1,0
0
0,0
0,2
0,4
0,6
0,8
1,0
y/D
-7-
TEST 5
TEST 5
90
45
40
Concentration (%Vol.)
120
Tern.
Bin.
Monod.
Tern.
Bin.
Monod.
60
12.00%
8.00%
12.00%
150
35
10.00%
15.00%
12.00%15.00%
30
15.00%
30
19.00%
20.00%
180
25
25.00%
20
15
210
10
5
0,0
0,2
0,4
0,6
0,8
25.00%
26.00%
30.00%
35.00%30.00%
30.00%
40.00% 32.00%
32.00%
240
330
300
270
1,0
y/D
50
TEST 6
90
Tern.
Bin.
Monod.
45
Concentration (%Vol.)
40
120
Tern.
Bin.
Monod.
60
10.00 %
150
35
10.00 %
10.00 %
30
15.00 %
15.00 %
15.00 %
18.00 % 20.00 %
30
25
180
20.00 %
20
25.00 %
15
210
25.00 % 25.00 %
30.00 %
35.00 %
10
5
0,0
0,2
0,4
0,6
0,8
240
1,0
330
30.00 %
30.00 %
35.00 %
35.00 %
300
270
y/D
TEST 7
90
50
120
Concentration (%Vol.)
Exper.
TotalBin.
Monod.
40
30
150
Species1
Species2
60
5.00%
5.50%
1.00%
6.00%
6.50%
7.00%
2.00%
30
180
20
10.00%
15.00%
210
10
240
0
0,0
0,2
0,4
0,6
0,8
1,0
330
20.00%
25.00%
30.00%
35.00%
300
270
y/D
-8-
TEST
Do 104
(m2/s)
9.8
4 - C o n c l u s io n s
It has been shown here that in a hydraulic conveying system , where a heterogeneous suspension is flowing, the
concentration profiles which build up in the cross-section of the pipe can be adequately described by the
Settling-Dispersion model, based on the balance between the tendency of the particles to settle and their
tendency to disperse from the regions of higher concentration.
Additionally, a new pressure drop correlation, which takes into account the distribution of the solids in the pipe,
was applied to the simulated profiles, having produced calculated pressure drop values very close to the
experimental ones.
However, a more realistic approach to the flow of a true suspension has to consider the particle size range of the
solid phase. The Settling-Dispersion model has been modified in order to cover that situation. The results
presented seem quite promising, though the experimental validation is still scarce. In fact, the existence of good
quality measurements supplying a reliable picture of the local conditions in the conveying pipe (solids and
velocity distributions) is considered of crucial importance.
To summarise, it can be said that by being able to predict, in an accurate way, the solids distribution in the
conveying pipe, it is possible to improve the design of such systems. The model presented here provides a
relatively simple and flexible tool that can be used to interpret and predict experimental data on hydraulic
transport systems.
Acknowledgements
This project was funded by PRAXIS XXI and by the Portuguese Foundation for Science and Technology (FCT).
One of the authors would also like to acknowledge the receipt of a scholarship from FCT.
REFERENCES
1- Durand, R.; Basic Relationship of the Transport of Solids in Pipes- Experimental Research, in Proc. Int.
Ass. For Hydraulic Research, Minneapolis, USA (1953).
2- Roco, M.C., Shook, C.A.; New Approach to Predict Concentration Distribution in Fine Particle Slurry
Flow; Physico Chem. Hydrodynamics, 8, 1, p. 43 (1987).
3- OBrien, M.P.; Review of the Theory of Turbulent Flow and its Relationship to Sediment Transportation,
Trans. Am. Geophysical Union, Section of Hydrology, p. 487 (1933).
4- Shook, C.A., Daniel, S.M., Scott, J.A., Holgate, J.P.; Flow of Suspensions in Pipelines, part. 2: Two
Mechanisms of Particle Suspension, The Can. J. Chem. Eng., 46, 4, p. 238 (1968).
5- Karabelas, A.J.; Vertical Distribution of Dilute Suspensions in Turbulent Pipe Flow, AIChE J., 23, 4, p.
426 (1977).
6- Wilson, K.C., Pugh, F.J.; Dispersive-Force Modelling of Turbulent Suspension in Heterogeneous Slurry
Flow, The Can. J. Chem. Eng., 66, p. 721 (1988).
7- Rasteiro, M.G., Figueiredo, M.M., Maia, M.C.,Scarlett, B.; Modelling of Solid/Liquid Flow in Pipes,
Powder Handling & Processing, 5, 3, p. 253 (1993).
8- Masliyah, J.H.; Hindered Settling in a Multi-Species Particle System, Chemical Eng. Sc., 34, p. 1166
(1979).
9- Mor, J.J., Cosnard, M.Y.; BRENTM, a Fortran Subroutine for the Numerical Solution of Systems of NonLinear Equations, ACM Trans. On Math. Software, 6, 2, p.241 (1980).
10- Rasteiro, M.G., Figueiredo, M.M., Franco, H.; Pressure Drop for Solid/Liquid Flow in Pipes, Particulate
Science and Technology, 11, p. 147 (1993
11- Gillies, R.G., Shook, C.A; Concentration Distribution of Sand slurries in Horizontal Pipe Flow,
Particulate Science and Technology, 12, p. 45 (1994).