HAL Id: jpa-00246925
https://hal.archives-ouvertes.fr/jpa-00246925
Submitted on 1 Jan 1994
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Recurrent clogging and density waves in granular material flowing through a narrow pipe
Thorsten Pöschel
To cite this version:
Thorsten Pöschel. Recurrent clogging and density waves in granular material flowing through a narrow
pipe. Journal de Physique I, EDP Sciences, 1994, 4 (4), pp.499-506. �10.1051/jp1:1994155�. �jpa-
00246925�
Classification Physics Abstracts
05.60 47.25 46.10 02.60
Recurrent clogging and density
wavesin granular material flowing through
a narrowpipe
Thorsten Pôschel
HLRZ, KFA Jülich, Postfach 1913, D-52425 Jülich, Germany and
Humboldt-Universitàt zu Berlin, FB Physik, Institut für Theoretische Physik, Invalidenstrafle 110, D-10115 Berlin, Germany
(Received
29 November1993, received in final form l7December 1993, accepted 7January1994)
Abstract. We report on density waves m granular material, investigated both experimen- tally and numerically. When granular material faits through a long narrow pipe one observes
recurrent cloggmg. The kinetic energy of the
falling
partiales increases up to a characteristicthreshold corresponding to the onset of recurrent clogging and density waves of no definite wave- length. The distances between regions of high density depend strongly on the initial conditions.
They vary irregularly without any characteristic time and length scale. The partiale-flow was investigated using 2D Molecular Dynamics simulations. Experimental investigations lead to
equivalent results.
The
astonishing
elfects observed inmoving granular
material bave beensubjects
of interest for at least 200 years.Examples
of such eflects are trieangle
of reposeil,
2] sizesegregation [3-5], heap
formation onvibrating
media[6-9], density
waves [10] andgranular
fiow down in-dined surfaces
[11, 12].
Trieorigin
of these eflects is thatgranular
material can behave like a solid or like aliquid, depending
on itsdensity
[13].Particularly,
in recent timesdynarnic
aswell as static behaviour of
granular
mater1al bas beeninvestigated by
many authorsexperi- mentally
andtheoretically
with varioustechniques
eitheranalytical,
such asthermodynamic
[14] andhydrodynamic
[15]approaches,
or numericallike cellular automata [16], Monte-Carlosimulation [4] and molecular
dynamics
[3, 8,17].
Even random walkapproaches
bave beenproposed
[18].Trie atm of this Letter is to report trie eflect that
granular
mater1al like sand that fallsthrough
a vertical narrowpipe
orcapillary
with a diameter ofonly
fewpartiale-diameters changes
from ahomogeneous
to aninhomogeneous
flow when trie kinetic energyEk
of triefalling grains
reaches a characteristic threshold. In this state one observes recurrentdogging
of trie
partide
flow(stick slip motion)
anddensity
waves. We simulated this behaviourusing
moleculardynamics
and demonstrate that trie results agree with ourexperimenta1observations.
500 JOURNAL DE PHYSIQUE I N°4
The
experiments
were carried out in two vertical fixedglass-pipes
of diameters Dl = 2 mmand D2 " 4 mm and
length
L= 1.4 m. An upper funnel contained
enough granular
material toensure
time-independent
initial conditions.During
trie experiments veryfine-grained
sand oftypical partiale
diameterDg
= 0.18 mmbegins
to flowhomogeneously
with trievelocity
u = 0 at trie upper end of trie pipe. After atypical
distance ofapproximately
20 cm trie uniform flow becomes unstable and tutus into a fiow of recurrentclogging.
In thisregime density
waves can be observed.
Figure
1 shows triepipes
withtypical density
distributions. Triedistances between dark
high-density regions
varyirregularly.
There is no definitewavelength.
Trie distance between trie top of trie tubes and trie onset of trie stick
slip
motion as well as for trie distances between trieplugs depends sensitively
on triehumidity
of trie air and variessignificantly
from oneexperiment
to trie next one, 1-e-depend
on initial conditions. Similarexperiments
for trie fiow of sandthrough hourglasses
bave been described in [19]. As m trie case of our numerical simulationsthey
foundclogging,
trie distribution of trie distances between triehigh-density
regionsobeys
a power law.in fact we bave doue trie simulations first and trie
reported
experiments were intended to check trie accuracy of trie resultspredicted by
trie numerical simulations.According
to trieexperiments reported
above weinvestigated numerically
the flow ofsphencal- shaped grains
with rotationaldegrees
of freedomfalling
undergravity through
a pipe of diam- eter D(Fig. 2)
using 2D moleculardynamics
simulations. Trielength
of triepipe
consideredwas much
larger
trier its diameter L » D. Weinvestigated
systems ofpartiales
ofequal
radii r~ = R as well as ofpartiales
with a Gaussianprobability
distribution for trie radii with meanvalue R
(Fig. 3).
For trie fiow we assumedperiodic boundary
conditions. Trie inner surface of triepipe
is build of smallerspheres
with a diameter rs=
)
R of trie same materai as trie grains to simulate arougir
surface. Since our model mcludes elastic and inelastic interaction between tl~e grains eacl~ otl~er and between triegrains
and tl~e watt we simulated tl~e wattexactly
in tl~esame mariner as trie
grains
but tl~epartiales
tl~epipe
is build of were neither allowed to movenor to rotate.
Trie
grains
with mass m~ were accelerated due togravity (g
= -9.81m/s~,
theirdensity
~~~Î~3
~ ~~~ ~° ~~~~~~'
Fz = -mi g
Ii)
where Fz is tl~e force
acting
upon trie itl~partiale
in trie direction of trie axis of triepipe.
Starting
withgiven
initialpositions
of trieparticles
andrandomly
chosen small initial hnear andangular
velocities triedynamics
of tl~e system was determinedby mtegrating
Newtonsequation
of motionnumerically
for eacl~particle
for ail furtl~er time steps. For tl~eintegration
we used a sixtl~ order
predictor-corrector
method for triepartiale positions
and a fourtl~ order for tl~e calculation of tl~epartiales' angular
movement [20].In our model
partiales
are allowed to penetrate eacl~ otl~ersligl~tly.
Tl~e force acting betweentwo
partiales
and j is assumed to be zero m case tl~epartiales
do net toucl~ eacl~otl~er,
1-e- when trie distance between their centres islarger
then trie sum of triepartiale
radii r~ + r~.Otherwise trie force separates into normal and shear force
(Eq. (2)).
~ =
FN ii
+ FSIl
&~)ii
if(xi x~l
< ri + r~
" ~~j
0 otherwise
with
FN " kN (r~ + r~ (x~ x~ ()~.~ ~iN
m~~(k~ k~) (3)
I
j
i
'
'
' i f
'
j <
~ ~
j
(
502 JOURNAL DE PHYSIQUE I N°4
acceleration due to
gravity
- DL
Fig.
2. The pipe. The rough surface consists of spheres with diameter rs=
j
R of the samematerial
as the grains.
Fig.
3. Emargement of a small part of the pipe filled with grains of dilferent radii. Gravity actshorizontally from left to right.
and
FS =
mm(-'fs
me~ urei P' (FNÎ)
~~~where
meoE "
~~
(6)m~ +Î~~
The
erms
itl~ partiale.
equation
(3) includes tl~e Hertzian contact force wl~icl~ rises
witl~
tl~e powerI.à
of
trie
penetrationdepth r~ + r~ - (x~ - x~( of two partiales. It
of tl~e grains
[21].
This nsatz
for force
ggested in [22j
and
ligl~tly
dified in [3].
quation 4 takes
into
account, that trie aximummomentum two
partiales
are able totransmit
~vhile
isdetermmed by trie
Coulomb friction law iii.
p
is trie oftrie
partiales.
To
discuss
lengtl~ of tl~e pipe
was
L = 666 . R andils widtl~
D
=was
~is " 3 x 10~ s~~, trie riction oefficient ~iN " 3x 0~ tl~e material constant
kN = lob )
and
p
=
o-S- For trie integrationlime step as hosen lit
=
10~~ s. Forvalue trie alculation
is
umencallyxact, 1-e- trie results do
non
depend on lit.After filling trie
artiales
into
trie pipe they start oving airest.
While fallingkinetic
energy
dueto
trie constantacceleration.
To
avoid infinitely rising velocity oftrie
E kin E_roll 2.5e10
2e10
1.5e10)
5e09
les 1.5e5 2e5 time steps
Fig.
4. Evolution of the kinetic energy of the transversal partiale movement and the rotational energy. Since the rotational energy is very small trie curveis almost indistinguishable of the time-
axis. While flowing accelerated, the system becomes unstable ai 35.000 lime steps approximately and
transits into another flow regime where ii reaches a state of relative steady kinetic energy
(E
G3 6 x 10~for this
simulation).
partiales
withoutcolliding they
were initialized with a very small randomvelocity perpendicular
to the axis of trie
pipe.
At a certain energy trie flow does net get faster anymore but triegrains
are decelerated very
intensively.
From now on trie system moves withapproximately
constant kinetic energy.Figure
4 shows trie evolution of trie kinetic energy of trie transversalparticle
movement Ek and trie energy of trie
partiale
rotation Er.In trie case of trie low
velocity regime
triegranular
materai movesthrough
triepipe
with almosthomogeneous partiale density. Figure
5displays snapshots
of trie pipe each 1,000 time steps. Time increasesupwards
whilehorizontally
one sees trie evolution of triedensity
wave from trie left to trieright. Gravity
acts from left toright.
In trieearly
times triepartiale-density
is
approximately homogeneous
due to trie initial conditions.Dunng
trie evolution ii becomesmore and more
inhomogeneous
anddensity
waves are observed. As visible at later times trie flow is unstable: trieregions
ofhigh density
candiverge
as well as converge while trie averagevelocity
remainsapproximately
trie same as shown infigure
4.After trie transition into trie
recurrent-plug-flow regime,
where trievelocity profile
does not vary with trie distance from trie centre on trie pipe, trie system reaches adynamic
state ofnearly
constant kinetic energy due to trieequilibrium
between constant acceleration and energydissipation by
friction. Nevertheless one can observe small fluctuations of trie energy whichare
approximately
due to triegenesis
andvanishing
ofregions
ofhigher density
with lime, as visiblecomparing
trie evolution of trie energy(Fig. 4)
with triecorresponding snapshot
ai triesame lime
(Fig. 5).
This behaviour proves that trieinhomogeneous
flow is unstable and ii could be an indication ofcoexisting
metastable states but we dia trot check ibispossibility
yen.Our simulations show further that there are no
significant
diflerences between trie behaviour ofa system with
partiales
ofequal
andrandomly
distributed radii.Simulating
trie system withslightly
dilferent initial conditions we got very similar results for trie energy evolution and for trie structure of triedensity
wave but trie concretedensity
wave as it cari be observed as dark andbright remous
infigure
5 variessignificantly. Very
small dilferences in trie initial conditions lead to very dilferent macroscopic behaviour. In ibis sense we call trie flowchaotically.
There are essent1al dilferences between trie behaviour of a
liquid
and a fluidizedgranular
material: for trie case of an
incompressible
fluid one never findsdensity
waves. Nevertheless there are similanties too: The evolution of trie energy of trie system is similar to trie evolution504 JOURNAL DE PHYSIQUE I N°4
t=26
t=24
t=22
t=20
t=18
t = iô
t= 14
t=12
t = io
t = 8
t = 6
t = 4
t = 2
t = 0
Fig.
5. Snapshots of the pipe each 1,000 time-steps. The time is given in lo~ time steps. Thedensity fluctuations
are net equidistant and net stable. They con converge as well as diverge. Dark regions correspond to
higher
densities. The partiales of which the wall consists ofare non drawn.
of trie
velocity
of aliquid flowing through
a pipe [23] while trie pressure is increasedgently.
Ai low flowvelocity
anincompressible
viscous fluid forms a laminar flow(Poisseuille-flow)
which gels faster withrising
energy. If trievelocity
reaches a certain value itc trie Poiseuille-flowbecomes unstable and an
inhomogeneous
flowregime
becomes stable. For trie case of a fixed pipe geometry and given fluid parameters trie criticalvelocity
of trie flow itc is describedby
triecntical
Reynolds-number
[24] Rec=
fi
m 2, 300 where ~lk is trie kinematic
viscosity
of trie fluid. Provided there are critical fluctuationsk trie system tums into a non-laminarregime.
If trie pressure is increased verygently
and trie system is net disturbed in another way,however,
it is
possible
to maintain trie Poiseuille-flow at velocities far over Re~ m trie unstableregime.
Trier a small fluctuation causes trie system to turn to trie stable state, 1-e- it
suddenly
lowers trievelocity
of trie flow anddissipates
trie energy. As shown for trie fiow of agranular
materialthere exist dilferent fiow
regimes
too and trie transition between them is due to a sudden energydissipation.
Hence trie transition from low energy tohigh
energyregime
ofgranular
flow in a pipe is similar to trie transition of a laminar fiuid-fiow into a non-laminar one.Experimental
and recent numericalinvestigations
of trailic fiows [25] lead to similar resultsas those
presented
here. At agiven
averagecar-velocity
triehomogeneous
trailic flow becomes unstable and trafficjamming
occursprovided
tl~ere are randomperturbations
of tl~evelocity
of tl~e cars whicl~correspond
to tl~e collisions of triegrains
witl~ trie wall in our case.Concluding
we state thatgranular mater1alfalling through
a narrow pipe appears elfects far from trie behaviour of a laminar fluid. At a certain energy triegranular
flow transits from triehomogeneous
to aninhomogeneous flow-regime
whereclogging
anddensity
waves can be observed. Trie regions ofhigh partiale density
can converge as well asdiverge,
their distances varyirregularly.
Trie results of trieexperiment
shown in triebeginmng
agree with trie numencalsimulations.
Acknowledgments.
Tl~e author tl~anks J.
Gallas,
H.Herrmann,
G. Ristow and S. Sokolowski for fruitful andstimulating
discussions and H. Pul~l forexperimental
assistance.References
iii
Coulomb C. A., Acam. R. Sa. Mem. Math. Phys. par Dmers Savants 7(1773)
343.[2] Jaeger H. M., Lin C. and Nagel S., Phys. Reu. Lent. 62, 40
(1989);
Lee J. and Herrmann H. J., J. Phys. A 26
(1993)
373.[3] Half P. K. and Werner B. T., Powder Technol. 48
(1986)
239.[4] Devillard P., J. Phys. France 51
(1990)
369.[Si Williams J. C., Powder Techn. 15
(1976)
245;Rosato A., Strandburg K. J., Prinz F. and Swendsen R. H., Phys. Reu. Lent. 58
(1987)
1038;Ohtsuki T., Takemoto Y. and Hayashi A., prepnnt.
[6] Faraday M., Phil. Trans. R. Soc. Land. 52
(1831)
299.[7] Dinkelacker F., Hübler A. and Lüscher E., Biot. Cybern. 56
(1987)
51;Walker J., Soi. Am. 247
(1982)
167.[8] Gallas J. A. C., Herrmann H. J. and Sokolowski S., Phys. Reu. Lent 69
(1992)
1371; J. Phys.France 2
(1992)
1389;Sokolowski S. and Herrmann H. J. E~rophys. Lent. 18
(1992)
415;Taguchi Y. H., Phys. Reu. Lent. 69
(1992)
1367;Taguchi Y. H., report titcmt-1992-7
(1992);
Herrmann H. J., Physica A191 (1992 263.
[9] Evesque P. and Rajchenbach J., Phys. Reu. Lent. 62
(1989)
44;Rajchenbach J., E~rophys. Lent. 16
(1991)
149.[10] Baxter G. W., Behringer R. P., Fagert T. and Johnson G. A., Phys. Reu. Lent. 62
(1989)
2825.[Il]
Drake T. G., J. Geophys. Res. 95(1990)
8681; J. Fhid Mech. 225(1991)
121.[12] Savage S. B. and Hutter K., J. Fhid Mech. 199
(1989)
177;Pôschel T., J. Phys. France 3
(1993)
27.[13] Reynolds O., Philos. Mag. 20
(1885)
469.[14] Edwards S. F. and Oakeshott R. B. S., Physica Amsterdam157A
(1989)
1080;Edwards S. F., J. Star. Phys. 62
(1991)
889;506 JOURNAL DE PHYSIQUE I N°4
Herrmann H. J., J. Phys. France 3
(1993)
427;Mehta A. Ed., Granular Matter
(Springer,
New York,1993).
[15] Savage S. B., J. Fhid Mech. 92
(1979)
53.[16] Baxter G. W. and Behringer R. P., Physica D 51
(1991)
465; Phys. Reu. A 42(1990)
1017.[17] Thompson P. A. and Grest G. S., Phys. Reu. Lent. 67
(1991)
1751;Hong D. C. and McLennan J. A., preprint;
Ristow G. H., J. Phys. France 2
(1992)
649.[18] Caram H. and Hong D. C., Phys. Reu. Lent. 67
(1991)
828.[19] Schick K. L. and Verveen A. A., Nat~re 251
(1974)
599.[20] Allen M. P. and Tildesley D. J., Computer Simulations of Liquids
(Clarendon
Press, Oxford,1987);
Walton O. R., Braun R. L., J. Rheol. 30
(1986)
949.[21] Landau L. D. and Lifschitz E. M., Theory of Elasticity
(Pergamon
Press, London, 1959).[22] Cundall P. and Strack O. D. L., Géotechmpe 29
(1979)
47.[23] Landau L. D. and Lifshitz E. M., Fluid Mechanics,
(Pergamon
Press, London, 1959).[24] O. Reynolds, Phil. Trans.
(1883);
Froc.Roy.Soc.(1894).
[25] Herman R. and Prigogme I., Science 204
(1979);
Nagel K. and Schreckenberg M., J. Phys. France 2
(1992)
2221;Nagel K. and Herrmann H. J., Physica A 199
(1993)
254;Treiterer J., Technical Report PB 246 094, Ohio