You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/317581175

A comparison of graded PSD methods in slurry transport

Conference Paper · May 2017

CITATIONS READS

0 242

1 author:

Sape A. Miedema
Delft University of Technology
329 PUBLICATIONS   3,791 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The Delft Sand, Clay & Rock Cutting Model View project

Center for Dredging Engineering Study View project

All content following this page was uploaded by Sape A. Miedema on 14 June 2017.

The user has requested enhancement of the downloaded file.


A comparison of graded PSD methods in slurry transport

Sape A. Miedema
Delft University of Technology – Dredging Engineering

ABSTRACT

Most methods for determining the hydraulic gradient of slurry transport are based on a
single particle size. A few methods describe how to deal with a graded Particle Size
Distribution (PSD). The way these methods deal with graded PSD’s is very different. The
original Durand/Condolios method defines an equivalent particle diameter, based on a
weighted average of particle Froude numbers using the parallel resistor method. The
heterogeneous Wilson model uses a power for the dependency of the hydraulic gradient on
the line speed, where this power is smaller, the more graded the PSD. This power has a
maximum of 1.7 for single sized particle solids and a minimum of 0.25 in the Wilson
model. The Sellgren & Wilson 4 component model divides the PSD in 4 components based
on particle size boundaries. For each particle size based component, the corresponding
model is applied. These models are the homogeneous model for very small particles
influencing the viscosity, the reduced Equivalent Liquid Model for small particles, the
heterogeneous model for medium sized particles and the two-layer and three layer models
for large particles. The 4th model considered is the Delft Head Loss & Limit Deposit
Velocity Framework, dividing the PSD in fractions. The number of fractions is free to
choose, but normally 9 is sufficient. First the carrier liquid properties are adjusted based
on the fines fraction. Secondly the hydraulic gradient curve is determined for each fraction.
These hydraulic gradient curves are added, multiplied with the fraction. The paper
describes the 4 methods and gives pros and cons of each method. The 4 methods are
compared with each other.

1 NOMENCLATURE

Ax, A15, A85 Factor PSD -


A Correction factor pseudo homogeneous flow -
B Correction factor stratified flow -
Cvs Spatial volumetric concentration -
Cvt Delivered/transport volumetric concentration -
Cvb Volumetric bed concentration -
Cvr Relative volumetric concentration Cvr=Cvt/Cvb -
Cvs,r Remaining spatial volumetric concentration -
Cvs,x Volumetric concentration of fraction x -
Cx Durand & Condolios particle drag coefficient -
CD Particle drag coefficient -
d, dx, dy Particle diameter m
d50 Particle diameter 50% passing by weight m
d15, d85 Particle diameter 15%/85% passing by weight m
dlim Limiting particle diameter homogeneous fraction m
Dp Pipe diameter m
Erhg Relative excess hydraulic gradient (stratification ratio) -
f Correction factor determination v50 -
fy, fx Fraction passing -
Fr Froude number -
g Gravitational constant m/s2
il Hydraulic gradient liquid -
im Hydraulic gradient mixture -
if Hydraulic gradient adjusted carrier liquid -
K Durand & Condolios constant -
ΔL Length of pipeline m
M Power Wilson heterogeneous equation (0.25-1.7) -
n Porosity -
dp, pi Probability -
Δpl Liquid pressure kPa
Δpm Mixture pressure kPa
R Stratification ratio (Erhg) -
Rsd Relative submerged density -
Rsd,f Relative submerged density in adjusted carrier liquid -
Sf Relative density adjusted carrier liquid -
vls Line speed m/s
vls,ldv Limit deposit velocity m/s
v50 50% stratification ratio line speed for d50 m/s
v50,f v50 in adjusted carrier liquid m/s
vt Particle terminal settling velocity m/s
vsm Maximum limit of stationary deposit velocity m/s
vsm,f vsm in adjusted carrier liquid m/s
w50, w85 Particle associated velocity Wilson for d50 and d85 m/s
X Fraction -
Xf Fraction of PSD in homogeneous component -
Xph Fraction of PSD in pseudo homogeneous component -
Xh Fraction of PSD in heterogeneous component -
Xs Fraction of PSD in fully stratified component -
α15, α85 Ratio -
β Bed angle with vertical rad
λl Darcy Weisbach friction coefficient for liquid -
λf Darcy Weisbach friction coefficient adjusted for carrier liquid -
ρl, ρx Liquid density and Pseudo liquid density ton/m3
ρs Solids density ton/m3
ρm Mixture density ton/m3
σ Variance PSD -
νl, νw Kinematic viscosity liquid/water m2/s
νx Kinematic viscosity pseudo liquid m2/s
μl, μx Dynamic viscosity liquid/Dynamic viscosity pseudo liquid Pa·s
μsf Sliding friction factor -
Φ, ψ Durand & Condolios solids effect -
ξ, ξvsm Slip ratio and Slip ratio at vsm -
2 INTRODUCTION

Most slurry transport models in the literature are based on a single particle size. The
experiments on which these models are based were carried out with either a narrowly-
graded Particle Size Distribution (PSD) or a single particle size. Most empirical models
are also based on the delivered volumetric concentration, since that is what was measured.
The more fundamental Two layer models (2LM) and Three layer models (3LM) are based
on a single particle size and a spatial volumetric concentration. Delivered concentration
curves are achieved by interpolation of delivered concentrations derived from slip ratios.
For graded or broadly graded PSD’s, the problem is, that the interaction between the
different fractions is not known. So, one can assume there is no interaction, all fractions
behave independently, or one can assume a certain interaction between the fractions. The
models considered here have some interaction. The models of Durand & Condolios (1952)
and Wilson et al. (1992) are both based on an adjustment of the equation for the
heterogeneous flow regime, but in different ways. The Sellgren & Wilson (2007) 4
component model (4CM) divides the PSD in 4 components based on particle size
boundaries. The Delft Head Loss & Limit Deposit Velocity Framework (DHLLDV) of
Miedema (June 2016) includes all flow regimes depending on the mean particle size of a
fraction.

3 THE ADJUSTED LIQUID PROPERTIES (4CM & DHLLDV)

The boundary between the homogeneous and the pseudo homogeneous flow regimes is
named the limiting particle diameter. The limiting particle diameter is determined, based
on a Stokes number of 0.03 for the DHLLDV Framework and 0.04 mm for the 4CM. The
value of 0.03 is found based on many experiments from literature. Since the Stokes number
depends on the line speed, here the Limit Deposit Velocity (LDV) is used as an estimate
of the operational line speed. The LDV is approximated by, giving for the limiting particle
diameter:

s  v ls,ldv  dlim
2
Stk  9  l   l  Dp
v ls,ldv =7.5  Dp0.4  Stk= d lim = (1)
9  l   l  Dp s  7.5  Dp0.4

The fraction of the particles (for example sand) in suspension, resulting in a homogeneous
pseudo fluid is named X. So, this is the fraction of particles smaller than dlim. This gives
for the density of the homogeneous pseudo fluid:

X  Cvs  Rsd
 x  l  l  if X  1  x  l  1  Cvs  Rsd 
1  Cvs  Cvs  X (2)

So the concentration of the homogeneous pseudo fluid is not Cvs,x=X·Cvs, but:

X  Cvs
Cvs,x 
1  Cvs  Cvs  X (3)

This is because part of the total volume is occupied by the particles that are not in
suspension, so the percentage of carrier liquid is reduced. The remaining spatial
concentration of solids to be used to determine the individual hydraulic gradients curves
of the fractions is now:
C vs,r   1  X   C vs (4)

The dynamic viscosity can now be determined according to Thomas (1965):


 x  l  1  2.5  Cvs,x  10.05  C2vs,x  0.00273  e
16.6Cvs,x
 (5)

The kinematic viscosity of the homogeneous pseudo fluid and the relative submerged
density are now:

x s   x
x  and R sd,x  (6)
x x

With the new homogeneous pseudo liquid density, kinematic viscosity, relative submerged
density and volumetric concentration the hydraulic gradient can be determined for each
fraction of the adjusted PSD in both the 4-component model and the DHLLDV
Framework. However, one can also combine this with the Durand & Condolios (1952) and
Wilson et al. (1992) models, although not mentioned by the authors. In this paper this is
not applied. In general, a higher pseudo liquid density and viscosity will increase the water-
based hydraulic gradient of homogeneous flow according to Darcy Weisbach. However,
this will decrease the water based hydraulic gradient in fully stratified flow (the sliding bed
regime) and in the heterogeneous flow regime. In both cases, due to the reduced relative
submerged density of the particles and in the heterogeneous flow regime also because of
the reduced terminal settling velocity due to the higher viscosity.

4 MODELS

A brief description is given here of the 4 models/methods considered. For a detailed


description of all models one can consult the appropriate chapters in Miedema (June 2016).

4.1 Durand & Condolios

In normal sands or other solids, there is not only one particle diameter, but a particle size
distribution (PSD) must be considered. The Froude number for a PSD can be determined
by integrating the Froude number as a function of the probability according to:

vt 1 1 1
Frp    
gd
  i  p i
1 n
Cx gd (7)
 vt
dp
i 1
Cx
0

It is also possible to split the particle size distribution into n fraction and determine the
weighted averaged particle Froude number. Gibert (1960) published a graph with values
for the particle Froude number that matches the findings of Durand & Condolios (1952).

Figure 1 shows these published values. If one uses the values of Gibert (1960), the whole
discussion about whether the CD or the Cx value should be used can be omitted. Analyzing
this figure however, shows that a very good approximation of the table values can be
achieved by using the particle Froude number to the power 20/9 instead of the power 1,
assuming that the terminal settling velocity vt is determined correctly for the solids
considered (Stokes, Budryck, Rittinger or Zanke, see Miedema (June 2016)).
Reciprocal Particle Froude Number √Cx Gibert & DHLLDV
5
Jufin-Lopatin
Jufin-Lopatin data
Durand, Condolios &
Gibert
H2 Sand, d50=0.20 mm,
√(Cx)=3.42
L3 Sand, d50=0.27 mm,
√(Cx,Gibert) or ψ*-2/3 (-)

√(Cx)=1.96
L4 Sand, d50=0.37 mm,
√(Cx)=1.34
L5 Sand, d50=0.58 mm,
√(Cx)=1.06
L6 Sand, d50=0.89 mm,
√(Cx)=0.88
L7 Sand, d50=1.33 mm,
√(Cx)=0.80
L8 Sand, d50=2.05 mm,
√(Cx)=0.72
A9 Gravel, d50=2.80
mm, √(Cx)=0.67
A10 Gravel, d50=4.20
mm, √(Cx)=0.62
Theory √(Cx)=1/Frp
0.5 DHLLDV & Gibert
0.1 1 10 √(Cx)=1/Frp^20/9
d (mm) Wilson NWL
© S.A.M.

Figure 1: Modified reciprocal particle Froude number,


determined experimentally for various sorts of sand and gravel by
Durand & Condolios (1952) and Gibert (1960).

Figure 1 shows the original data points, the theoretical reciprocal particle Froude numbers
using the Zanke (1977) equation for the terminal settling velocity of sand particles and the
curve using a power of 20/9. For large particles, there may be a small difference between
the original data and the theoretical curve applying the power of 20/9. The final equation
of Durand & Condolios (1952) and later Gibert (1960) for the pressure losses now
becomes:

pm  pl  1    Cvt  (8)

With the use of the PSD modified particle Froude number:

3/ 2
i i pm  pl  2
vls 
 m l   83    Cx  (9)
i l  Cvt pl  Cvt  g  Dp  Rsd 
 

Based on the current research and Figure 1 this can be written as:

3/ 2
 2
vls 1 
  83    20/9  (10)
 g  Dp  Rsd Frp 
 

4.2 Wilson et al.

Wilson et al. (1997) have defined a stratification ratio or relative solids effect, which
identifies which fraction of the particles is in suspension and which fraction is in the fixed
or moving bed, supported by granular contact. Wilson et al. (1997) give the following
general equation for the head losses in hydraulic transport, where μsf equals the friction
factor of a sliding bed, which was determined as μsf=0.44. For the 50% case at line speed
v50 this gives:

M
im  i l p m  p l  v 
Erhg  R    sf   50  (11)
Rsd  Cvt l  g  L  R sd  Cvt 2  v ls 

When the line speed vls equals the v50, the stratification ratio is 0.22 or half the sliding
friction coefficient μsf. This can be written in terms of pressures instead of hydraulic
gradient as:

M
sf v 
pm  pl   l  g  L   50   R sd  C vt (12)
2  vls 

Notice that here the solids effect does not depend on the carrier liquid pressure or hydraulic
gradient as it does in equation (8). This equation can be written in the more generic form,
matching the notations of the other theories:

 2 M 
sf  g  R sd  Dp M  1 
p m   p l   1    v 50      C vt  (13)
 l  v ls  
 

For the cross-sectional area-averaged line speed, where 50% of the particles are in granular
contact, v50, Wilson gives the following equation:

8  60  d 
v50  w 50   cosh  50 
and d50,4CM  0.0002  0.015  Dp (14)
l  Dp 
 

When the power M equals 1, equation (13) has the same form as the equation of Durand
& Condolios (1952), Gibert (1960), Fuhrboter (1961), Jufin Lopatin (1966) and Newitt et
al. (1955). The power M depends on the solids PSD and can be determined by:

 
1/ 2
M  0.25  13   2 (15)

The variance σ of the PSD can be determined by some ratio between the v50 and the v85:

 8  60  d  
 w 85   cosh  85  
 v 85   l  Dp  
  log   
  log   (16)
 v 50   8  60  d  
w
 50   cosh  50 
 Dp  
 l  

The particle associated velocity w, which is terminal settling velocity related, can be
determined by:

w  0.9  v t  2.7   R sd  g   l 
1/3
(17)
It seems this equation mixes the homogeneous and heterogeneous regimes. For very small
particles the second term gives a constant particle associated velocity, which matches
homogeneous behavior at operational line speeds. Since the homogeneous behavior does
not depend on the particle size, this gives a constant or asymptotic particle associated
velocity. The model of Wilson can be simplified with some fit functions, according to:

0.25 1
 d 
0.35
R 
0.45  l,actual    d 
v50  3.93   50    sd    and M   ln  85   (18)
 0.001   1.65    w,20   
    d50  

where the particle diameter d50 is in m and the resulting v50 in m/s. The third term on the
right had side is the relative viscosity, the actual liquid viscosity divided by the viscosity
of water at 20ºC. In normal (dredging) practice this term is about unity and can be
neglected. The factor 1.65 is based on sand in clear water. Later the simplified equation
for the v50 has been adjusted for particles with diameters from 0.2 mm to 0.5 mm with a
factor, Sellgren et al. (2016) and Miedema (June 2016):

dh  0.0002 dh  0.0001
f  or f  (19)
0.0005  0.0002 0.0004  0.0001

The second equation however gives a more reasonable fit for particles between 0.1 mm
and 0.4 mm. An even better simplification of the v50 is achieved with the following
equation (Figure 2 DHLLDV simplified, the dash-dot line):

0.25 0.092
 d 
0.45
R 
0.45  l,actual   Dp 
v50  3   50    sd       (20)
   1.65    w,20 
0.0005    0.1524 

The v50 versus the Particle diameter


8
Wilson Original

7
Wilson Simplified
6

Wilson Simplified
5 Modified 1
v50 m/s)

Wilson Simplified
4 Modified 2

3
DHLLDV
Simplified

2 DHLLDV
Framework
Simplified
1
DHLLDV
Framework
0
1.E-05 1.E-04 1.E-03 1.E-02
Particle diameter (m)
Dp=0.1524 m, Rsd=1.585, Cvs=0.300, μsf=0.416

Figure 2: The different v50 methods.


A much better approach to establishing v50 however, is the following equation, based on
the DHLLDV Framework:

1.96 0.092 0.092 0.27


 v   Dp  R   l,actual 
v50  3.4   t       sd    (21)
 gd   1.65    w,20 
   0.1524   

Figure 2 shows the different v50 methods. The thick black solid line shows the original v50
method of Wilson et al. (1992), equation (14). For very small particles and very large
particles this method overestimates the v50. This is because of homogeneous transport
under operational conditions for very small particles, and because of the occurrence of a
sliding bed for very large particles. For particles in the range of d=0.3 mm to d=0.8 mm
the estimated v50 values seem to be reasonable. The thin solid line shows the simplified
equation(18), which is close to the original in the range of d=0.1 mm to d=1 mm, although
the fit could be much better. The dash-dot green line shows the result of the DHLLDV
simplified equation (20), giving a much closer fit to the original Wilson et al. (1992)
method. Later the simplified equation (18) was adjusted with a factor to get rid of the
overestimation for small particles with equation (19).

Both possibilities of equation (19) are drawn in Figure 2. The first equation is the red
dashed line and the second equation the dash-dot red line. Still the equations include the
influence of the homogeneous regime for very small particles and the sliding bed regime
for very large particles. Using a factor to compensate for the homogeneous overestimation
is understandable, but not based on physics.

Based on the DHLLDV Framework the v50 can also be determined. Although this
Framework is more complicated, the line speed where the heterogeneous hydraulic
gradient matches the hydraulic gradient of a sliding bed with 50% of the sliding friction
coefficient is possible. This is the light brown dash-dot line in the graph, equation (21).
For medium sized solids this line matches both the original Wilson et al. (1992) method
and the simplified equations. It also matches the use of the factor (2 nd) to get rid of the
overestimation for very small particles. For very large particles this gives much smaller v50
values, since the formulation of the heterogeneous regime of the DHLLDV Framework is
not influenced by the sliding bed regime. Equation (21) is a simplification of the DHLLDV
Framework (the solid blue line), matching the full Framework very accurate. For the
Wilson et al. (1992) model for a solids PSD equation (21) will be used for the determination
of the v50, in order to have a fair comparison.

It should be mentioned that the original Wilson et al. (1992) method and the simplified
equation are not consistent regarding the viscosity. With an increasing viscosity, the
original method will give an increasing v50, while the simplified equation gives a
decreasing v50. This is caused by the second term in equation (17) for the particle associated
velocity, which increases with increasing viscosity, while the settling velocity will
decrease. The correct behaviour is a decreasing v50 with increasing viscosity, due to the
decrease in the settling velocity. In equation (21) the term containing the settling velocity
decreases much faster than the increase of the term with the viscosity, giving a decrease of
the v50 with increasing viscosity, like the simplified equations.

4.3 The modified 4 component model (4CM)

The Sellgren & Wilson (2007) 4 component model (4CM) divides the PSD in 4
components based on particle size boundaries. These components are:
1. Homogeneous flow (the fines d<0.04 mm).
2. Pseudo homogeneous flow (0.04 mm<d<0.2·νr mm).
3. Heterogeneous flow (0.2·νr mm<d<0.015·Dp).
4. Fully stratified flow (d>0.015·Dp).

The original 4 component model (4CM) (Sellgren & Wilson (2007)) assumes that each
coarser fraction is moving in a carrier liquid containing all finer fractions. So, the coarsest
stratified fraction is moving in a carrier liquid containing the homogeneous, pseudo
homogeneous and heterogeneous fractions, increasing the carrier liquid density and thus
decreasing the relative submerged density and the settling velocity. Only the homogeneous
fraction influences the viscosity of the carrier liquid. The original 4CM model up to
Sellgren et al. (2014) contains some errors regarding the determination of the different
carrier liquid densities, which have been corrected, after Miedema & Ramsdell (2015)
discovered this, in Sellgren et al. (2016) as also mentioned in Miedema (June 2016). There
is still an error in Sellgren et al. (2016) in equation 13 for the fully stratified flow. The
factor 0.55 should be to the power 0.25. In the original article, Sellgren et al. (2014) this
was correct. The original model for fully stratified flow is, using the sliding friction factor
of 0.44:

0.25 0.25
 0.55  vsm  v 
i m  i l  B' R sd  C vt     i l  B' R sd  C vt  2   sf   sm  (22)
 v ls   v ls 

Wilson et al. (1992) used a sliding friction factor of 0.44 in the derivation of the 2-layer
model. Based on his hydrostatic approach the factor 2 in equation (22) is valid for plug
flow if the spatial volumetric concentration equals the bed concentration. A more general
notation can be given, based on the spatial concentration (note the factor 2 is omitted):

2   sin       cos    
0.25
v 
im  i l  B  R sd  C vs  sf   sm  with: B  (23)
 v ls     sin     cos    
With the factor B according to the hydrostatic normal stress approach (with β the bed angle
to the vertical, so no bed β=0, 50% bed β=π/2 and 100% bed β=π). The factor B is included
because with β=π the term describes plug flow, B=2. For β<π the factor B decreases with
β, β=π/2 gives B=1.3 and β=0 gives B=1. So, for low concentrations and small fully
stratified fractions a value of B=1 should be chosen, which matches the later choice of
Wilson for a B’=0.5 (originally he mentioned B’ is close to unity). Sellgren et al. (2014)
use B’=0.25 (B=0.5), while in the reprint of the article Sellgren et al. (2016) use B’=0.35
(B=0.7). The weight approach of Miedema & Ramsdell (2014) also uses B=1. To convert
the equation from spatial volumetric concentration to delivered volumetric concentration,
the slip ratio ξvsm at the line speed vsm should be known, giving:

0.25
Cvt v 
im  i l  B  Rsd   sf   sm  (24)
1   vsm  vls 

The effect of the decreasing slip ratio with increasing line speed is taken into account with
the term (vsm/vls)0.25. The slip ratio can be estimated by the following empirical equation,
based on the two-layer model with sheet flow (a thin layer of fast moving particles on top
of the bed) added (so 3LM), Miedema (June 2016):
Cvt B
Cvr  and   0.58  Cvr0.42 and  Cvr0.5
Cvb 1   vsm
   v 
  (25)
   0.83 sf  C  0.5  0.075 D
  
2
 0.025 Dp  Dp0.025  ls  C0.65 
  4 vr p
  vsm 
vr

   1  Cvr   e 

This way the slip ratio is incorporated in the model and B’ depends on this slip ratio. A
better approximation of the hydraulic gradient is, using the line speed dependent slip ratio
(B=1 weight approach, B=equation (23) hydrostatic approach):

Cvt
im  i l  B  Rsd   sf (26)
1 

The maximum hydraulic gradient of course is the hydraulic gradient for plug flow, so if
plug flow is reached, the hydrostatic Wilson approach gives as an upper limit:

im  i l  2  Rsd  Cvb  sf (27)

The weight approach (assuming the hydraulic gradient is caused by the submerged weight
times the sliding friction factor) of Miedema & Ramsdell (2014) gives for the plug flow
upper limit (this follows from substituting equation (25) in equation (26) at vls=0):

im  i l  Rsd  Cvb  sf (28)

In the 4CM model each coarser fraction floats in all finer fractions. So, the buoyancy of
coarser fractions increases with increasing finer fractions. This seems strange since a bed
is a bed, including all fractions in the bed. The small fractions in the bed do not make the
coarse fractions lighter. So, in the modified 4CM model (see Miedema (June 2016)), only
the homogeneous fraction is assumed to influence the density and the viscosity of the
carrier liquid (see the adjusted liquid properties). This gives for the relative density Sf of
the homogeneous mixture of particles with d<0. 040 mm:

f Xf  Cvs  Rsd
Sf   1
1  Cvs  1  Xf 
(29)
l

The hydraulic gradient of the homogeneous flow regime if is now:

f  f  v ls
2
  v2 
if   =Sf  f ls = f  i l  Sf  i l (30)
l 2  g  Dp 2  g  Dp l

The difference between the pseudo fluid and pure carrier liquid is the Darcy Weisbach
friction factor, the relative density Sf and the viscosity of the resulting homogeneous
pseudo fluid. For small homogeneous fractions, the Darcy Weisbach friction factor will
not differ much from the factor determined for the carrier liquid. So, the main difference
is the use of the relative density Sf>1 instead of 1 and the viscosity. The resulting hydraulic
gradient of the modified 4 component model is now:
 v  
M
 A  X  i  X   sf   50,f  
f   ph f h
2  v ls  
im   i f  f  C vt  R sd,f   (31)
l l   vsm,f 
0.25 
Xs
 B    sf    
 1   vsm 
  v ls  

The factor A<1 is included because often the excess hydraulic gradient is smaller than the
Equivalent Liquid Model (ELM) would give, because of near wall lift. A value of A=0.5-
0.6 is found to be reasonable. For the factor B a value of 1 is applied. Sellgren et al. (2016)
give a smaller value of about 0.7, which is difficult to compare, since they do not use the
slip ratio. The power M in the 4CM model is assumed to be 1. The v50,f and vsm,f values
are determined based on the adjusted carrier liquid. Physically this is a three-layer model,
with the homogeneous fraction forming an adjusted carrier liquid, the fully stratified
fraction forming a sliding bed, the heterogeneous fraction on top of the sliding bed and the
pseudo homogeneous fraction at the top of the pipe. For small concentrations, however,
there is not much difference between this model and original 4CM model, but for large
concentrations there may be a difference depending on the PSD.

4.4 The DHLLDV Framework

The DHLLDV Framework is extensively described in Miedema (June 2016) and will not
be described in detail here. The Framework combines the 5 main flow regimes; the
stationary bed regime, the sliding bed regime, the heterogeneous flow regime or sliding
flow regime and the homogeneous flow regime, for single sized particle solids (for
example sands and gravels) and constant spatial volumetric concentrations. The result is a
hydraulic gradient curve where all flow regimes may be present depending on pipe and
particle diameter, concentration and line speed. Based on a Limit Deposit Velocity model
of Miedema (June 2016) the slip ratio curve is constructed and based on the slip ratio curve,
the constant delivered concentration curve is determined. For graded sands or gravels, the
PSD is divided into several fractions. First the liquid properties are adjusted as described
in this paper. Secondly the PSD is adjusted, so that it does not contain the fines anymore.
For each resulting fraction the hydraulic gradient curve is determined based on the
spatial/delivered concentration of the whole PSD to take hindered settling into account in
an appropriate way. The resulting hydraulic gradient curves are multiplied with the
corresponding fraction and pseudo liquid to liquid density ratio and summed. The result is
a hydraulic gradient curve for the whole PSD.

5 EXAMPLE OF GRADED SOLIDS (SAND)

As an example of the comparison of the 4 methods a pipe diameter of Dp=0.1524 m and a


d50=0.5 mm are chosen, because the models for solids having a single particle size give
very similar hydraulic gradients under operational conditions (line speeds). The different
v50 equations also give about the same v50. This way only the grading of the solids PSD
may be the reason for the differences.

Figure 3 shows the hydraulic gradient curves of 9 fractions and the resulting hydraulic
gradient curve for the DHLLDV Framework. The PSD given in Figure 4, is a wide PSD
to emphasize the effect of grading. This figure also shows the PSD corrected for the fines
of the DHLLDV Framework and the heterogeneous PSD as used in the 4CM model, as
well as the PSD for the delivered solids. Figure 3 shows that for small line speeds the
resulting hydraulic gradient (the thick dashed line) is smaller than the corresponding
hydraulic gradient of the single sized particle sand (the thick solid line). For larger line
speeds (above about 2.6 m/s) however, the resulting hydraulic gradient is larger.

Hydraulic gradient im, il vs. Line speed vls, Graded


0.40
Liquid il curve
0.36 Equivalent Liquid
Model
Homogeneous
0.32
Hydraulic gradient im , il (m water/m)

Sliding Bed
0.28 Cvs=c
d05=0.033 mm
0.24 d15=0.100 mm

0.20 d25=0.180 mm
d35=0.282 mm
0.16
d50=0.500 mm
0.12
d65=0.888 mm
0.08 d75=1.386 mm

0.04
d85=2.500 mm
d95=7.682 mm
0.00
0 1 2 3 4 5 6 7 8 9 10 Limit Deposit
Line speed vls (m/sec) Velocity
Graded
© S.A.M. Dp=0.1524 m, Rsd=1.585, Cvt=0.300, μsf=0.416

Figure 3: The hydraulic gradient curves of the fractions and the resulting hydraulic
gradient curve, according to the DHLLDV Framework.

Cumulative Grain Size Distribution


100

90 PSD Original
Spatial
80

70
PSD Remaining
60
DHLLDV
Framework
% Passing

50

40 PSD 4 CM
Heterogeneous
Fraction
30

20
PSD Delivered
10

0
1.0E-06 1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01 1.0E+00
Particle Diameter (m)
© S.A.M.

Figure 4: The PSD’s of the solids (sand) considered.

This behavior is similar to the effect of a reduced power M in the Wilson heterogeneous
v50 method, with a v50 of about 3 m/s for all equations. Figure 5 and Figure 6 show a
comparison of the 4 different methods. In Figure 6 the relative excess hydraulic gradient
or stratification ratio as Wilson named it is shown, which is defined as:
im  i l im  i l
Erhg  or Erhg  (32)
Rsd  Cvt Rsd  Cvs

Hydraulic gradient i m, il vs. Line speed vls


0.30
Liquid il curve
0.28

0.26
Limit Deposit
0.24
Velocity
Hydraulic gradient im , il (m water/m)

0.22

0.20
4 Component
Model
0.18

0.16 Wilson
Heterogeneous
0.14

0.12 Durand &


0.10
Condolios
Graded
0.08 DHLLDV Graded
0.06 Sand Cvs=c.
0.04
DHLLDV Graded
0.02 Sand Cvt=c.
0.00
0 1 2 3 4 5 6 7 8
Line speed vls (m/sec)
© S.A.M. Dp=0.1524 m, d=0.500 mm, Rsd=1.585, Cv=0.300, μsf=0.416

Figure 5: The resulting hydraulic gradient curves.

Relative excess hydraulic gradient E rhg vs. Hydraulic gradient i l


Limit Deposit
Velocity
Relative excess hydraulic gradient Erhg (-)

1.000
4 Component
Model

Wilson
Heterogeneous

Durand &
0.100 Condolios
Graded
DHLLDV Graded
Sand Cvs=c.

DHLLDV Graded
Sand Cvt=c.
0.010
0.001 0.010 0.100 1.000
Hydraulic gradient il (-)
© S.A.M. Dp=0.1524 m, d=0.500 mm, Rsd=1.585, Cv=0.300, μsf=0.416

Figure 6: The resulting relative excess hydraulic gradient curves, also named the
stratification ratio.

6 CONCLUSIONS & DISCUSSION

The Durand & Condolios (1952) model always gives a lower hydraulic gradient curve for
graded solids, compared with single sized particle solids, which contradicts with the 3 other
methods. This is caused by the shape of the √Cx number graph in Figure 1. In the example
considered the value increases from 1.13 for single sized particle sand to 3.54 for graded
sand (the thin dash dot line). The effect of this is a reduction to 20% of the solids effect in
the hydraulic gradient for the graded solids. This does not seem to be reasonable and also
contradicts measurements of Clift et al. (1982) with broad graded crushed granite, which
matches the other 3 methods. The Durand & Condolios model contradicts with these 3
models and is rejected.

The Limit Deposit Velocity, the line speed above which there is no stationary or sliding
bed, is estimated by the DHLLDV Framework to be about 3.1 m/s. Under operational
conditions, if line speeds in the range of 3-5 m/s in the pipe diameter are considered, the
Wilson heterogeneous model (thick dash dot line), the 4CM model (thick short dashed
line) and the DHLLDV Framework (thick long dashed line) give very similar hydraulic
gradients and Erhg curves, of which the 4CM gives a slightly higher hydraulic gradient.
The DHLLDV Framework has 10.3% of the particles to adjust the carrier liquid properties,
which is a 3% concentration in the case considered (30% solids). About 16.3% of the
particles are in the sliding flow regime, which is an integral part of the Framework. The
original Wilson heterogeneous model gives a power M=0.587, the simplified model
M=0.621. For the 4CM model a power of M=1 is applied, according to Sellgren et al.
(2016). The 4CM has about 6% of the particles in the homogeneous component (1.8%
concentration), 21% in the pseudo homogeneous component, 57% in the heterogeneous
component and 16% in the fully stratified component. For both the DHLLDV Framework
and the 4CM, the carrier liquid properties were adjusted, but only slightly.

The differences between the 3 models occur outside the operational line speed range. The
solids effect of the Wilson heterogeneous model and the 4 CM go to infinity for very small
line speeds, due to the formulation of the heterogeneous component, see equation (12) and
equation (23) for the stratified component in the 4CM. The solids effect of the Wilson
heterogeneous model goes to zero at very high line speeds, see also equation (12). The
solids effect of the 4CM will increase at very high line speeds due to the effects of the
homogeneous and pseudo homogeneous fractions. The 4CM curve would be a bit lower if
the reduced relative submerged density was used as in the original model.

The main difference between the 4CM and the DHLLDV Framework is, the 4CM uses
fixed boundaries based on particle diameters to divide the PSD into 4 components. These
boundaries do not depend on the line speed. The DHLLDV Framework determines the
hydraulic gradient curves for each fraction separately. The result is a changing flow regime
division depending on the line speed and the particle diameter. Particles that may be in a
sliding bed at low line speeds, will behave heterogeneous at a higher line speed and
homogeneous at a very high line speed. Here the PSD is divided into 9 fractions, which
seems to be enough, but the number of fractions can be unlimited.

The heterogeneous line of Wilson in the Erhg graph pivots around the v50 when the power
M is changed. The concept of the v50 pivot point of Wilson, matched the results of the
DHLLDV Framework. At line speeds, higher than the v50 the solids effect increases while
at lower line speeds it decreases with wider PSD’s, although the line speed of this pivot
point is not exactly the same in both models, but it is close.

Under operational conditions, line speeds above the LDV, all 3 models, the Wilson
heterogeneous model, the 4CM and the DHLLDV Framework, can be used. Outside the
operational conditions, low and high line speeds, the DHLLDV Framework takes the
behavior of a possible sliding bed and homogeneous flow better into account.
7 REFERENCES

Clift, R., Wilson, K. C., Addie, G. R., & Carstens, M. R. (1982). A mechanistically based
method for scaling pipeline tests for settling slurries. Hydrotransport 8 (pp. 91-
101). Cranfield, UK.: BHRA Fluid Engineering.
Durand, R., & Condolios, E. (1952). Etude experimentale du refoulement des materieaux
en conduites en particulier des produits de dragage et des schlamms. Deuxiemes
Journees de l'Hydraulique., 27-55.
Durand, R., & Condolios, E. (1952). Etude experimentale du refoulement des materieaux
en conduites en particulier des produits de dragage et des schlamms. (Experimental
study of the discharge pipes materieaux especially products of dredging and
slurries). Deuxiemes Journees de l'Hydraulique., 27-55.
Fuhrboter, A. (1961). Über die Förderung von Sand-Wasser-Gemischen in Rohrleitungen.
(On the advances of sand -water mixtures in pipelines). Mitteilungen des Franzius-
Instituts, H. 19.
Gibert, R. (1960). Transport hydraulique et refoulement des mixtures en conduites.
Annales des Ponts et Chausees., 130(3), 307-74, 130(4), 437-94.
Jufin, A. P., & Lopatin, N. A. (1966). O projekte TUiN na gidrotransport zernistych
materialov po stalnym truboprovodam. Gidrotechniceskoe Strojitelstvo, 9., 49-52.
Miedema, S. A. (June 2016). Slurry Transport: Fundamentals, A Historical Overview &
The Delft Head Loss & Limit Deposit Velocity Framework. (1st Edition ed.). (R.
C. Ramsdell, Ed.) Miami, Florida, USA: Delft University of Technology.
Miedema, S. A., & Ramsdell, R. C. (2014). An Analysis of the Hydrostatic Approach of
Wilson for the Friction of a Sliding Bed. WEDA/TAMU (p. 21). Toronto, Canada:
WEDA.
Miedema, S. A., & Ramsdell, R. C. (2015, May). Pages from The Delft Head Loss & Limit
Deposit Velocity Framework: Wilson. Retrieved from ResearchGate:
https://www.researchgate.net/publication/277340666_Pages_from_The_Delft_H
ead_Loss_Limit_Deposit_Velocity_Framework_Wilson
Newitt, D. M., Richardson, J. F., Abbott, M., & Turtle, R. B. (1955). Hydraulic conveying
of solids in horizontal pipes. Transactions of the Institution of Chemical Engineers
Vo.l 33., 93-110.
Sellgren, A., & Wilson, K. (2007). Validation of a four-component pipeline friction-loss
model. Hydrotransport 17 (pp. 193-204). BHR Group.
Sellgren, A., Visintainer, R., Furlan, J., & Matousek, V. (2014). Pump and pipeline
performance when pumping slurries with different particle gradings.
Hydrotransport 19 (pp. 131-143). Denver, Colorado, USA.: BHR Group.
Sellgren, A., Visintainer, R., Furlan, J., & Matousek, V. (2016). Pump and pipeline
performance when pumping slurries with different particle gradings. The
Canadian Journal of Chemical Engineering, Vol. 94(6), 1025-1031.
Thomas, D. G. (1965). Transport characteristics of suspensions: VIII. A note on the
viscosity of Newtonian suspensions of uniform spherical particles. Journal Of
Colloidal Sciences, Vol. 20., 267-277.
Wilson, K. C., Addie, G. R., & Clift, R. (1992). Slurry Transport using Centrifugal Pumps.
New York: Elsevier Applied Sciences.
Wilson, K. C., Addie, G. R., Clift, R., & Sellgren, A. (1997). Slurry Transport using
Centrifugal Pumps. Glasgow, UK.: Chapman & Hall, Blackie Academic &
Professional.
Zanke, U. C. (1977). Berechnung der Sinkgeschwindigkeiten von Sedimenten. Hannover,
Germany: Mitteilungen Des Francius Instituts for Wasserbau, Heft 46, seite 243,
Technical University Hannover.

View publication stats

You might also like