Sie sind auf Seite 1von 9

See

discussions, stats, and author profiles for this publication at:


https://www.researchgate.net/publication/244155878

Drag Coefficient and Terminal Velocity of


Spherical and Non-Spherical Particles

Article in Powder Technology May 1989


DOI: 10.1016/0032-5910(89)80008-7

CITATIONS READS

750 2,011

1 author:

A. Haider
Texas Instruments Inc.
10 PUBLICATIONS 792 CITATIONS

SEE PROFILE

All content following this page was uploaded by A. Haider on 23 June 2016.

The user has requested enhancement of the downloaded file.


Powder Technology, 58 (1989) 63 - 70 63

Drag Coefficient and Terminal Velocity of Spherical and Nonspherical Particles

A. HAIDER and 0. LEVENSPIEL


Chemical Engineering Department, Oregon State University, Corvallis, OR 97331 (U.S.A.)
(Received September 19,1988)

SUMMARY results of various researchers, and using


nonlinear regression on 300 data points,
Explicit equations are developed for the proposed the following drag equation for
drag coefficient and for the terminal velocity Re < 3 X 105:
of falling spherical and nonspherical particles.
Cn = (2.25 Remoa31+ 0.36 Re0*06)3.45 (1)
The goodness of fit of these equations to the
reported experimental data is evaluated and Flemmer and Banks [4] proposed, for Re <
is compared with that of other recently 8.6 x 104,
proposed equations.
Accurate design charts for CD and ut are (2)
prepared and displayed for all particle spher-
icities. where
E = 0.261 Re*36g- 0.105 Re0*43
INTRODUCTION 0.124
-
1 + (log,, Re)2
There are well over 30 equations in the
literature relating the drag coefficient Cn to Turton and Levenspiel [ 51, using the equation
the Reynolds number Re of spheres falling form proposed by Clift and Gauvin [6] plus
at their terminal velocities. These correlations 408 previously reported experimental data
are of varying complexity, and contain as points, presented the following correlation
many as 18 arbitrary constants. Clift et al. for Re < 2.6 X 10:
[l] , Khan and Richardson [2], and Haider
[3] list many of these correlations. For non- CD = g (1 + 0.173 Reom6)
spherical particles, however, no generalized
expression for Cn vs. Re is available. This 0.413
paper develops and presents such a correlation.
Usually it is the terminal velocity, rather + 1 + 16300 Re-l*Og
than the drag coefficient, which is of ultimate After looking at the regression analysis for
interest. However, we have found only three spherical and nonspherical particles and the
equations in the literature which give ut results of Clift and Gauvin [6] and Turton
explicitly, again only for spheres, while none and Levenspiel [5], the following four-
are available for nonspherical particles. This parameter general drag correlation is proposed
paper also develops such a correlation for here :
particles of all sphericities.
C
CD= z(l+AReB)+ (4)
D
DRAG COEFFICIENT FOR, SPHERICAL l+Re
PARTICLES The values of A, B, C, and D in eqn. (4) were
found by minimizing the sum of squares
Consider the three most recently proposed error Q, which for n data points is defined as
drag equations, all of which happen to
contain five arbitrary constants. First, Khan
and Richardson [2] compiled experimental
Q= 5 tlog10 cD,
i=l
exp - log10 cD. cd2 (5)

0032-5910/89/$3.50 @ Elsevier Sequoia/Printed in The Netherlands


64

This was done with nonlinear regression soft- numbers in the last column of Table 1. Here
ware [7] which uses the Gauss-Newton the RMS deviation measures the average
method. The experimental data used in finding fractional displacement of the measured Cn-
the best values of the four parameters were values from the correlation line. Mathemat-
the same 408 data points compiled by Turton ically,
and Levenspiel [ 51. The final equation for l/2

predicting Cn of spheres is then found to be RMS= &


0n
24 0.4251
co= Re (1 + 0.1806 Re0*645g) +
6880.95 f: (loi?lOCD.exp - 1og10cD,cad2 12
1+ i=l
Re =
n
(6) [ I (7)
The goodness of fit of eqns. (l), (2), (3), Finally, the lowest curve in Fig. 1 shows the
and (6) to the data are compared by the fit of eqn. (6) to the experimental data.
TABLE 1
Comparison of fit to the data of the most recent drag correlations for spheres

Researchers Equation No. Re range RMS deviation in


calculated CD

Khan and Richardson [ 2 ] Eqn. (1) Re<3x105 0.041


Flemmer and Banks [ 41 Eqn. (2) Re < 8.6 x lo4 0.066
Turton and Levenspiel [ 5 ] Eqn. (3) Re < 2.6 x 10 0.025
This work Eqn. (6) Re < 2.6 x lo5 0.024

i-

isometric
I ch = 0.670

0.1
0.01 IO 100 IO3 IO4 IO5 IO6
dsph tf
Re= cI
Fig. 1. Reported drag coefficients for spherical particles (408 data points) and nonspherical particles (506 data
points).
65

DRAG COEFFICIENT FOR NONSPHERICAL


.
experience higher drag as they become less
PARTICLES spherical.
Here we correlate the reported experimental The data were fitted using the same equa-
drag data for nonspherical particles by a direct tion form as used for spheres, namely eqn. (4).
extension of the four-parameter equation Once again, the best values of the four param-
form of eqn. (4), but first we need a measure eters were evaluated and are summarized in
of particle shape and of particle size. Table 2.
Wadell [8] introduced the concept of Figure 1 shows the experimental data and
particle sphericity r$to account for particle the best fitted drag curves as predicted by
shape, thus eqn. (4) with parameter estimates given in
Table 2. Column 4 of Table 3 shows the good-
ness of fit of this equation to the data by
listing its RMS deviation at the various
where s is the surface of a sphere having the $-values.
same volume as the particle and S is the actual
surface area of the particle. TABLE 2
For close to isometrically shaped particles, Best values of parameters to be used in eqn. (4) for
predicting C, for particles of various sphericities
those with no one very much longer or very
much shorter dimension, the sphericity is B C D
@ A
a useful measure, most likely the best single
parameter for describing the shape for falling 1.000 0.1806 0.6459 0.4251 6880.95
particles. 0.906 0.2155 0.6028 0.8203 1080.835
As for particle size, a whole host of 0.846 0.2559 0.5876 1.2191 1154.13
0.806 0.2734 0.5510 1.406 762.39
measures have been proposed. We have chosen
0.670 0.4531 0.4484 1.945 101.178
to use the equivalent spherical diameter d,, 0.230 2.5 0.21 15 30
which is the diameter of a sphere having the 0.123 4.2 0.16 28 19
same volume as the particle. All correlations 0.043 7 0.13 67 7
presented hereafter are based on this measure 0.026 11 0.12 110 5
of particle size. For isotropically shaped
particles,
COMPREHENSIVE DRAG EQUATION FOR
d sph= Leen (9) FALLING PARTICLES

Experimental drag data for eight different Table 2 shows that the four parameters of
particle sphericities were compiled. For eqn. (4) are functions of 4, and once this
$J> 0.670, the data were adapted from functionality is established, it would make it
Pettyjohn and Christiansen [9], who studied convenient to interpolate for Cn for sphericities
the following isometric nonspherical shapes: other than the ones listed in Table 2. Keeping
this in mind, values of the four parameters
Cube octahedrons ($J= 0.906) 136 data points were plotted against 4 and a reasonable order
Octahedrons (@J= 0.846) 80 data points polynomial was fitted through the data using
Cubes (4 = 0.806) 136 data points least-squares fit. The result is
Tetrahedrons ($I= 0.670) 67 data points
A = exp(2.3288 - 6.45816 + 2.4486 rp2)
For lower sphericities, $ < 0.670, the data
used were for thin free-falling disks extracted (lOa)
from Schmiedel [lo], Squires and Squires B = 0.0964 + 0.5565 $I (lob)
[ 111, and Willmarth et al. [12]. In this case,
Cn, Re and r$were recalculated based upon C = exp(4.905 - 13.89444 + 18.4222@2
d,, using the information provided about the - 10.2599 f#JJ) (1Oc)
particles and the fluids used in the experiment.
The data were plotted for various @-values D = exp(1.4681 + 12.25844 - 20.7322 @2
and were observed to follow drag curves + 15.8855 e3) (1Od)
similar to that for spherical particles except
that the drag curves lie higher and higher as 4 Substituting the above relations into eqn. (4)
drops from unity. In other words, particles yields
TABLE 3
Deviation of predicted CD from experiment for different sphericities

Shapes Number of 4 RMS deviation of RMS deviation of RMS deviation of


data points eqn. (4) with eqn. (11) eqn. (12)
parameter values
given in Table 2

Spheres 408 1 .ooo 0.024 0.031 0.058

Isometric 136 0.906 0.017 0.024 0.040


solids 80 0.846 0.018 0.022 0.038
136 0.806 0.024 0.030 0.035
67 0.670 0.024 0.034 0.044

Disks, 10 0.230 0.049 0.067 0.160


Non-isometric 17 0.123 0.043 0.077 0.147
solids 30 0.043 0.107 0.095 0.112
30 0.026 0.107 0.154 0.217

Co = g [ 1 + exp(2.3288 - 6.4581$1+ 2.4486 @?) Ee(*0964+0-5565@)]

Re X exp(4.905 - 13.8944$ + 18.4222 42 - 10.2599 I#J~)


+ (11)
Re + exp(1.4681+ 12.25844 - 20.7322 d2 + 15.8855 G3)

Equation (11) predicts Cn quite accurately TERMINAL VELOCITY OF FALLING SOLIDS


but it is rather tedious to use. As an approxi- To find the terminal velocity ut of particles
mation, a linear relation was developed for from any of the proposed Cn us. Re expres-
each of the parameters as a function of $.I. sions requires a tedious trial and error
When substituted into eqn. (4), this gives procedure since ut is present in both variables.
a much simpler expression, Thus, it would be useful to have an expression
which explicitly gives ut in terms of the
system variables.
Cn = 2 [I + [8.1716 exp(-4.0655 $)] So far, three explicit equations have been
reported in the literature to determine the
x Re0.0964 +0.5565 $)
1 terminal velocities of free-falling spheres,
73.69 Re exp(-5.0748 4) however none have been proposed for non-
(12) spherical particles. Here, we develop a simple,
+ Re + 5.378 exp(6.2122 4) explicit, and reasonably accurate equation
capable of predicting terminal velocities of
Table 3 also compares the goodness of fit free-falling spheres as well as nonspherical
of eqns. (11) and (12) to the experimental particles.
data. For isometric particks (r#~ > 0.67), the Before presenting anyones results, we
fit is quite good, butkpoorer for disks. Also, define two dimensionless quantities: a dimen-
for spheres we may prefer to use the simpler sionless terminal velocity u, and a dimension-
eqn. (6) with its RMS error of 2.4% rather less particle diameter d, as follows
than eqns. (11) or (12) with their RMS errors
4 Re Ii3
of 3.1% and 5.8%, respectively. u, = --
Figure 2 presents a design chart for the ( 3 cD 1
drag coefficient Cn using eqn. (6) for spheres

1
l/3
Pt2
and the comprehensive drag equation, eqn. =
4 (13)
(ll), for all other sphericities. W(P* -Pt)
67

0.1
0.1 1.0 IO 100 IO3 IO4 IO5 IO6

dsph Pf
Re= ~

Fig. 2. Design chart for drag coefficients of single free-falling particles.

and

This correlation comes from the general two-


parameter expression for u, :
(14)

Zigrang and Sylvester [13] presented an


explicit equation for particle settling velocities
in solid-liquid suspensions. In the limit as the In the present study, values of Kl and K2 for
volume fraction of solids in the system the five different sphericities (0.67 < 4 < 1.00)
approaches zero, their equation gives the were found by minimizing the sum of squares
terminal velocity of a single sphere settling error for u ,,,, defined similarly as in eqn. (5).
in a liquid as Experimental data analysed are the same 408
[(14.51 + 1.83d*32)i2- 3.81]2 points for spherical particles as previously
u* = (15) used in the analysis of drag coefficients, plus
d, all the data from Pettyjohn and Christiansen
Khan and Richardson [2], using nonlinear [ 91 for nonspherical isometric particles.
regression on their compiled experimental Table 4 gives the best-fit values of Kl and K2
data, presented the following explicit correla- along with RMS deviation for various spher-
tion for ut of spheres: icities, and Fig. 3 shows the resulting fit of
Ke = (2.33d,e.os - 1_53d,-0.06)s.s eqn. (18) to the data. For the sake of clarity,
(16) the $ = 0.846 line and data are omitted from
Recently, Turton and Clark [ 141, making this figure.
use of the asymptotic expressions for ut for One may notice from Table 4 that the
very low and very high Re, came up with values of K2 for different $Jstay close to 1.
68

TABLE 4 for K1 and the corresponding RMS deviation


Best values of K1 and K2 to be used in eqn. (18) for are shown in Table 5.
calculating the terminal velocity of particles of various One notices from this table that K1 increases
sphericities regularly as C#I
decreases. This may be approxi-
mated by the linear relationship
9 KI K2 RMS deviation
of u, K1 = 3.1131- 2.3252 $I (29)

1.000 0.7554 0.8243 0.0245 Substituting eqn. (20) into eqn. (19) yields
0.906 0.9999 0.9677 0.0212 the following simple general correlation for
0.846 1.1272 0.9697 0.0257
0.806 1.2024 1.0222 0.0327
0.670 1.5469 0.9464 0.0275
TABLE 5
Best values of K1 to be used in the simple eqn. (19)
Setting Kz = 1 in eqn. (18) essentially amounts for calculating us of particles of different sphericities
to adding low and high Re contributions in
parallel. On doing this, eqn. (18) reduces to RMS deviation
the simple one-parameter expression of u,

1 .ooo 0.8039 0.0403


u, = 18, 0) 0.906 1.0142 0.0218
&t2 0.846 1.1416 0.0261
0.806 1.1926 0.0329
Once again, nonlinear
regression was performed
0.670 1.5824 0.0288
on eqn. (19). The resulting best-fit values

100

IO

d,= d sph

Fig. 3. Reported terminal velocities for both spherical and nonspherical particles.
69

predicting terminal velocities for isometric Table 6 shows the goodness of fit of eqn. (21)
particles, given information on the particles to the data for different sphericities as well as
and physical properties of the fluid, the goodness of fit of all the recommended

,I
(2.3348 - 1.7439 r$) - equations for spheres.
u* =
I_E+
k2
d 0.5
*
0.5 < 4 < 1
Finally, Fig. 4 displays the best-fit terminal
velocity curves for different particle spher-
icities.
(21)

TABLE 6
Goodness of fit of various explicit equations for u* to the reported data

Researchers 9 Equation No. RMS deviation


of eauation from data

Zigrang and Sylvester [ 13 ] 1.0 Eqn. (15) 0.041


Khan and Richardson [ 2 ] 1.0 Eqn. (16) 0.033
Turton and Clark [ 141 1.0 Eqn. (17) 0.024
This work, for spheres 1.0 Eqn. (21) 0.041
This work, for nonspherical particles 0.906 Eqn. (21) 0.022
0.846 Eqn. (21) 0.026
0.806 Eqn. (21) 0.035
0.670 Eqn. (21) 0.029

Fig. 4. Design chart for finding the terminal velocity of single free-falling particles.
70

FINAL RECOMMENDATIONS g acceleration due to gravity, = 9.81


m/s2
Spherical Particles: 408 data points, Re < KI, K2 fitted constants in eqn. (18).
2.6 X lo5 Re Reynolds number based on equivalent
- For finding Cn use eqn. (6): RMS devia- spherical diameter of particle,
tion = 2.4% = dsphutpf
- For finding ut use eqn. (17): RMS devia-
cc
tion = 2.4% RMS root-mean-square deviation of data
Nonspherical Particles: 419 isometric data points from correlation, on log-log
points, Re < 25 000; 87 disk data points, plot, see eqn. (7)
Re < 500 terminal velocity of particle in fluid,
ut
- For finding Cn with a cumbersome but
m/s
accurate expression use eqn. (11): RMS dimensionless particle velocity, see
u*
deviation = 3% for isometric particles. eqn. (13)
- For finding Co with a simpler expression
use eqn. (12): RMS deviation = 5% for Greek symbols
isometric particles. viscosity of fluid, kg/m. s)
P
-For finding ut for $ # 1 use eqn. (21):
Pi density of fluid, kg/m3
RMS deviation = 3% for isometric particles.
PS density of particle, kg/m3
Design Charts:
# particle sphericity, see eqn. (8)
- Figure 2, based on eqn. (6) for 4 = 1 and
eqn. (11) for $ + 1, relates Cn with Re.
-Figure 4 is for finding u, and conse- REFERENCES
quently ut. It is based on eqn. (18) with Ki
1 R. Clift, J. R. Grace and M. E. Weber, Bubbles,
and Kz values given in Table 4 for spheres Drops and Particles, Academic Press, New York,
and on eqn. (21) for nonspherical isometric 1978.
particles. A. R. Khan and J. F. Richardson, Chem. Eng.
Comm., 62 (1987) 135.
A. M. Haider, M.S. Pro_iect, Oregon State Univer-
ACKNOWLEDGEMENT sity (1987).
R. L. C. Flemmer and C. L. Banks, Powder
A. H. acknowledges the financial assistance Technol, 48 (1986) 217.
extended by the National Science Foundation R. Turton and 0. Levenspiel, Powder Technol.,
for this project under NSF Grant # CBT- 47 (1986) 83.
R. CIift and W. H. Gauvin, Proc. Chemeca 70, 1
8420034. (1970) 14.
Bio Medical Data Processing Statistical Software
LIST OF SYMBOLS Manual, Program BMDP-BR, University of
California Press, 1985.
A B, fitted constants in eqn. (4) 8 H. Wadell, J. Fmnklin Inst. (April 1934) 459.
9 E. S. Pettyjohn and E. B. Christiansen, Chem.
C, D Eng. Prog., 44 (1948) 157.
4 gd,a (P, - pi)
CD drag coefficient, = - - 10 V. J. Schmiedel, Physik. Zeit., 29 (1928) 593.
3 ut2 Pf 11 L. Squires and W. Squires Jr., Trans. Am. Inst.
d, screen size of particles, m Chem. Eng., 33 (1937) 1.

d, equivalent spherical diameter, or 12 W. W. Willmarth, N. E. Hawk and R. L. Harvey,


Physics of Fluids, 7 (1964) 197.
diameter of sphere which has same
13 D. J. Zigrang and N. D. Sylvester, AZChE J., 27
volume as particle, m (1981) 1043.
dimensionless particle diameter, see 14 R. Turton and N. N. Clark, Powder Technol, 53
eqn. (14) (1987) 127.

View publication stats