Professional Documents
Culture Documents
1987 - Paper - Peden - Settling Velocity of Variously Shaped Particles in Drilling and Fracturing Fluids
1987 - Paper - Peden - Settling Velocity of Variously Shaped Particles in Drilling and Fracturing Fluids
Summary. The settling velocities of a variety of shaped particles to simulate drilled cuttings were measured in both Newtonian
and non-Newtonian fluids. The results showed that the particle drag coefficient is a function of the particle Reynolds ~umber and,
in the case of power-law-model fluids, of the flow behavior index. A new generalized model has been developed for predicting the
settling velocities of particles of various shapes in both Newtonian and power-law fluids over a range of flow regimes.
Theory of Drag-Coefficient Correlation For power-law fluids based on dimensional analysis, Cv is expect-
Assuming that the particles are separated sufficiently during settling ed to be a function of both the modified particle Reynolds number,
so that they do not collide or interact with each other, the force NRem• and the flow behavior index, n:
causing a particle to settle may be expressed as
Cv =J(NRem• n), · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (6)
Fg = 10- 6 VP(pP -p1 )g . ............................. (1) where
*Permanently employed at Research lnst. of Petroleum E&D, Beijing. (10 -3 ds)nvp (2-n) (103 Pj)
. .................. (7)
Copyright 1987 Society of Petroleum Engineers lc
SPE Drilling Engineering, December 1987 337
TABLE 1-PHYSICAL PROPERTIES AND DIMENSIONS
OF THE PARTICLES USED IN THE EXPERIMENT
Density Dimension
Materials (g/cm 3-) Shape (mm) Sphericity
Bauxite proppant 3.71 Sphere 0.3* 1.00
Intermediate Proppant 3.22 Sphere 0.4 to 0.75* 1.00
Sands 2.63 Well-rounded 1.2 to 2.2* 1.00
Glass beads 2.63 Sphere 4 to 6 1.00
Marbles 2.47 Sphere 15.5 to 25.1 1.00
Steel balls 7.75 Sphere 3.17 to 6.34 1.00
6.4 X 2.1 0.75
Aluminum 2.70 Disk 9.7x3.1 0.75
5.9x4.1 x2.1 0.73
Aluminum 2.70 Rectangle 10 X 8 X 2.1 0.62
6.4 X 2.2 0.76
. Plastic
. .
*These values are based on screen analysis.
2.28 Disk 9.6x3.2 0.76
For Newtonian fluids, the flow behavior index is, by definition, Examination of Eqs. 5 through 12 shows that a generalized form
unity and Eq. 7 reduces to Eq. 5. of the drag-coefficient correlation for both Newtonian and power-
In the laminar flow regime, where the inertia effect may be ne- law fluids, which is valid for all flow regimes, may be written as
glected, Stokes obtained the drag-coefficient correlation for spher-
ical particles within Newtonian fluids by theoretical analysis 16 as a
Cv=-- . ..................................... (13)
NRe~
24
Cv=--· (NRen<0.1) ............................ (8)
NRen For simplicity, and without significantly rectucing the accuracy, the
correlation described was used in this study. Dimensional analysis
For power-law fluids, the drag coefficient may be expressed as 3 shows that the coefficients a and e depend on only the fluid flow
regime for Newtonian fluids, and for power-law fluids depend also
on the flow behavior index, n.
24
Cv=--X (NRem <0.1), ......................... (9)
NRem Experimental Apparatus and Procedures
The experiments were conducted with two columns. One was a clear
where X is function of n and its lower and upper bounds have been perspex column (88-mm [3llz-in.] ID and 200-cm [6.56-ft] length)
evaluated by Wasserman and Slattery. 17 and the other a glass cylinder (60-mm [2.36-in.] ID and 45-cm
In the transitional and turbulent flow regimes, some empirical [1.48-ft] length). To simulate the drilled cuttings and to provide
correlations have been developed for Newtonian fluids. For spher- the drag.:coefficient values over a wide range of particle Reynolds
ical particles, Bird et al. 16 gave . numbers, solid particles were used that varied in maximum dimen-
sional size from 0.3 to 25 mm [0.01 to 1 in.], with densities rang-
18.5 ing from 2.5 to 7.8 g/cm 3 , and in a variety of shapes-e.g.,
Cv=-- (2<NRen <500) ...................... (10) sphere, disk, and rectangular plate. The properties and dimensions
6
NReR· of the particles used in the test are described in Table 1.
A number of fluids were selected as test fluids to represent three
and Allen 18 obtained principal rheological types encountered in the problem previously
discussed. The first was Oi168, which represents a Newtonian fluid.
30 Carboxymethyl cellulose (CMC) and xanthan gum biopolymer (XC)
Cv - - - (1 <NRen < 10 3 ) . . . . . . . . . . . . . . . . . . . . . (11) solutions were used to provide power-law fluids, and hydroxyethyl-
NReR.625 cellulose (HEC) solution was used to represent a viscoelastic fluid 7
and to allow examination of the elastic effect. Through pilot tests,
Beyond NRen = 10 3 , application of Newton's law yields the concentrations of CMC, XC, and HEC were selected so that
proper distributions of n and K values could be obtained. The rheo-
logical properties were measured with two viscometers-a Fann
Cv=0.44 (NRen > 10 3 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12) 35A to provide the data under high shear rate (170 to 1,021
Concentration
(g/L) K
Pp 'Y*
CMC** CMCt XC
---
HEC _!_g_ (g/cm 3 ) n (Pa·sn) (seconds - 1 )
CMC 35 3 19.5 1.015 0.798 0.194 5 to 1,021
CMC/XC 6 1 21 1.002 0.627 0.1696 5 to 1,021
CMC/XC 10 1 18.5 1.003 0.597 0.4579 5 to 1,021
CMC/XC 8 3 21.3 1.003 0.449 0.7688 5 to 1,021
HEC 1.42 23 1.000 0.716 0.0374 5 to 1,021
HEC 4.28 22 1.00 0.59 0.5106 5 to 1,021
HEC 7.13 19.5 1.00 0.418 3.992 5 to 1,021
Shell Tellus Oil 68 20 0.88 1.000 0.1653
*These are shear-rate ranges within which n and K were determined.
• • Low-viscosity CMC.
tHigh-viscosity CMC.
0
cQ)
g
Qj
0
(..) 2
10
Ol
Stokes' Law ro
n=1 0
Stokes' Law
n=1
10
10
Particle Reynolds Number N Rem w-1
Particle Reynolds Number NRem
Fig. 1-Comparison between the theoretical correlation and
the correlation obtained with the experiment for Newtonian Fig. 2-Drag-coefficic:mt correlations obtained for CMC and
fluid in the laminar flow regime. CMC/XC fluids in the laminar flow regime.
seconds - 1 ) and a HAAKE-MVIIP VT181™ to provide the low- Fig. 1 shows that the experimental correlation for oil is basical-
shear-rate data (5 to 159 seconds - 1). The properties of the fluids ly consistent with the standard Newtonian fluid correlation-i.e.,
are listed in Table 2. Stokes' law. The experimental correlations for CMC and CMC/XC
A video camera and a recorder were assembled, and a computer fluids are shown in Fig. 2. It can be seen that these correlation lines
clock with an accuracy of 1 %/sec was incorporated so that the time lie close to that which defines the behavior of a Newtonian fluid
and the corresponding position of a settling particle could be record- where n = 1. However, the lines have different slopes that depend
ed simultaneously to allow its velocity to be calculated. When the on the value of the flow behavior index, n. From Table 3 it can
settling velocity was below about 10 mm/s [0.4 in./sec], a stop watch be seen that, as n increases, a and e decrease. Thus for power-law
was used. Proppant and sands were dropped into the column in a fluids, the drag coefficient depends not only on the value of N Rem
cluster and the particles settling midway were caught to measure but also on the value of n, as expected from dimensional analysis.
the settling velocity. Other particles were dropped individually. In As n approaches unity, however, these correlations failed to ap-
addition, the orientation of disks and rectangular plates during proach that for a Newtonian fluid.
settling was visually observed and recorded. For the same particle Fig. 3 presents the experimental correlations for HEC fluids. It
in the same fluid, about four to eight data points were obtained to can be seen that if HEC fluids were also characterized as power-
ensure repeatability and accuracy. Test results showed that the var- law-type fluids, the drag coefficient would be a function of both
iation in settling-velocity measurements for sands was within ± 10% NRem and nand would have a response similar to the CMC and
and for other particles within ± 5 % . CMCIXC fluids. Thus the elastic properties of HEC fluids are not
To allow a particle to reach its terminal settling velocity, the par- influential in this low-NRem region.
ticle's settling velocity was measured after a certain distance of its
free fall. This distance was at least 60 em [2ft] for polymer fluids
in the perspex column and 15 em [0.5 ft] for oil in the glass cylin- TABLE 3-DRAG-COEFFICIENT CORRELATIONS FOR
der. The theoretical analysis and the test results showed that these SPHERICAL PARTICLES IN THE LAMINAR FLOW REGIME
free-fall distances were adequate.
a Average
Co=-- Data Error
Experimental Results and Analysis Fluid n NRe~ NRem Points (%)
Our experiments on particle settling velocity, ·vP, were performed . 32.1
in cylinders, and the experimental data must be corrected for the Oil 1.00 Co= N o.888
0.003 to 3 11 14.87
wall effect before they can be analyzed to determine the drag coeffi- Rem
cient. In this study, the following two simple empirical equations 18
for the correction factor, Fw, were selected to account for the wall 32.75
CMC 0.8 Co= N 0.006 to 4 9 12.60
effect. For laminar flow, o.824
Rem
10
10
Fig. 3-Drag-coefficient correlations obtained for HEC fluids Fig. 4-Drag•coefficient correlations obtained for CMC and
in the laminar flow regime. CMC/XC fluids in the transitional flow regime.
The experimental correlations in Table 3 were also compared with Fig. 4 shows the experimental correlations for CMC and
Eq. 9 and with the lower and upper bounds on X obtained by Was- CMC/XC fluids, and Allen's 18 and Bird's 16 correlations for New-
serman and Slattery. 17 The averages of the lower and upper tonian fluids. It can be seen that the drag coefficient is also a func-
bourtds range from 1 to 1.27 when n=0.7 to 1, which is basically tion of both NRem and n in this region, and as n increases, both
a and e decrease. However, some variation is shown in the drag
consistent with our experimental correlations. No effort was made
coefficient for power-law fluids compared with that fm Newtoni-
to fit the experimental data to Eq. 9, however, because inertia ef- an fluids. When NR~m is less than about 10, the experimental cor-
fect 'liad been neglected in its derivation, thus setting the exponent relation lines fall in the region between two Newtonian correlation
of NRem to Unity and restricting its validity to cases where lines, showing that the drag coefficient for power-law fluids is still
NRem < 0.1. Our experimental correlations were obtained in the close in value to that for Newtonian fluids. When NRem > 10, how-
range of 10- 3 <NRem<5, so the exponents of NRem are lower ever, all the exp~rimental correlation lines fall below the Newtoni-
than unity as the result of some inertia effect. an correlation lines, showing some "drag-reduction" effect. As
The experimental correlations for spherical particles in the range the value of n decreases, this drag-reduction effect becomes more
of 1 <NRem < 150 for CMCand CMC/XC fluids and in the range pronounced. A similar phenomenon has long been recognized for
of 1 <NRem <270 for HEC fluids are shown in Table 4. The aver- the friction factor in pipe flow of polymer fluids. 3 The experimen-
age deviations of the data from the corresponding correlations are tal correlationsJor HEC fluids are shown in Fig. 5 and some elas-
all less than ±9.2%. tic effect can be noticed by comparing Figs. 4 and 5.
Average
Data Correlation Error
Fluid n Co VS. NRem NRem Points Coefficient (%)
23.83
CMC 0.8 Co= N o.731 3 to 100 7 0.999 2.87
Rem
27.07
CMC/XC 0.6 to 0.63 Co= o.803 2 to 130 13 0.997 6.84
NRem
32.18
CMC/XC 0.45 Co= o.847 1 to 150 7 0.999 5.46
NRem
16.78
HEC 0.72 Co= o.624 2 to 270 7 0.997 6.47
NRem
:ja
25.41
HEC 0.6 Co= o.no 2 to 230 4 0.997 9.18
NRem
18.5
Bird's correlation: Co=--o.-6 for 1 <NRen <500
NRen
30
Allen's correlation: Co= N o.625 for 1 <NRen < 1,000
Ren
& 10
Allen's Correlation
c:
Q)
n=1
'(j
;;::
~ Bird's Correlation
() n=1
Ol
<ll
0
10 ~em <1
When NRem > 350, the limited experimental data showed that the and
drag coefficient became basically a constant and the value of this
constant depended on the flow behavior index, n. F 8 =2.6-1.61/; (NRem>l) .......................... (20)
Drag-Coefficient Correlations for Nonspherical farticles. To Combining with Eq. 13, the qrag coefficient of a nonspherical par-
simulate drilleq cuttings and cavings, disks and rectangular plates ticle becomes
were also used in the experiments. The results showed that the par-
ticle's orientation during settling was independent of its ori~nta F8 a
tion at release in all flow regimes. As shown in Fig. 6, when NRem Cv'=F8 Cv=--, ............................. (21)
was greater than about 10, these particles always settled flatwise NReme
but then would turn on edge when NRem was less than about 1.
In the range of 1 <NRem < 10, unstable settling occurred and the where CD is the drag coefficient of a sphere whose volume is. equal
particle tended to sett1e in a zigzag path. These results are in good to that of the particle concerned.
agreement with those obtained by Walker and Mayes. 11 By use of Eqs. 19 through 21 to correlate the shape effect of the
The drag-force analysis showed that a particle tended to settle experimental data from disks and rectangular plates, the average
in the orientation that offered the greatest possible resistance in all deviation qf these data from the corresponding drag-coefficient
flow regimes. This can be better understood by analyzing Eqs. 2 correlation~ for spherical particles is ± 10.2%. Fig. 7 shows the
through 4. When NRem > 10, pressure drag dominates the total drag results in the CMC/XC fluids when n=0.6 to 0.63 in the transi-
force and the particle settles flatwise, thus offering a larger area tional flow regime! and Fig. 8 shows the results in the HEC fluid
AN and corresponding pressure drag force, and accordingly a larg- when n=0.42 in the laminar flow regime.
er total drag force. However, when NRem < 1, viscous drag
dominates the total drag force ap.d the particle settles edgewise, Parti~le Settling-Velocity Calculation
offering a larger area AP, which results in both a larger viscous From our analysis, a generalized model has been developed that
drag and corresponding total drag force. In the range ·of allows the calculation of the settling velocity of particles of vari-
1 <NRem < 10, neither pressure drag nor viscous drag dominates; ous shapes in both Newtonic!.fl and power:..law fluids. The model
thus, unstable settling results. takes the form of
According to previous investigations, 18 the drag coefficient for
nonspherical particles in Newtonian fluids is a functio11 of both the
particle Reynolds number and the particle's sphericity, 1/;, which
is defined as 1/;=A; IA 8 •
Because the drilled cuttings can be characterized either as disks
or rectangular plates in addition to spheres, the above expression
can be simplified as follows. For disks, where d 8 is the diameter of a sphere or the volume-equivalent di-
ameter of a nonspherical particle and Fs is the shape factor deter-
mined from Eqs. 17 through 20. For Newtonian fluids and spherical
2
2.621 xFhb h p~rticles, n=1, Ic=Jl, and F 8 =1. When the coefficients a and e
1/;= ' ............................... (17) have valt!es of 24 and 1, respectively, pq. 22 reduces to the Stokes'-
1+2XFhb law model, and when they have values of 0.44 and 0, the equation
is that defined by Newton's law.
where Fhb =hs /dd. For rectangular plates, For power-law fluids, on the basis of our experimental data, the
following empirical equations m(j.y be used to approximate a and e.
2 In the laminar flow regime, ·
2.418 xFhb h
' .......................... : .... (18)
a=39.8-9n .................................... (23a)
anq
where Fhb=h 8 1beq and beq=Jib.
On the basis of our experimental data, the following empirical
e=l.2-0.47n (NRem<5 and n>0.45), ........ : .... (23b)
equations were developeq to correlate the effect of the particle's
shape on the drag coefficient, Fs:
and in the transitional flow regime,
E
Q)
"(j
~0
u
Ol 0 Disk Ol Correlation tor Spheres
l1l l1l
0 o Rectangle 0 HEC Solution n =0.42
10 ~------------------~------------------~
2
10
Fig. 7---Drag-coefficient correlation for spherical and non- Fig. a-Drag-coefficient correlation for spherical and non-
spherical particles after correction for shape effect in the tran- spherical particles after correction for shape effect in the lami-
sitional flow regime. nar flow regime.