Force Schemes in Simulations of Granular Materials J. Shäfer, S. Dippel, D. Wolf To cite this version: J. Shäfer, S. Dippel, D. Wolf. Force Schemes in Simulations of Granular Materials. Journal de Physique I, EDP Sciences, 1996, 6 (1), pp.5-20. <10.1051/jp1:1996129>. <jpa-00247176> HAL Id: jpa-00247176 https://hal.archives-ouvertes.fr/jpa-00247176 Submitted on 1 Jan 1996 HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published 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. J. Phys. I France Force (1996) 6 Schemes Schifer(*), J. Simulations in S. JANUARY1996, 5-20 Dippel and 3 July1995, Forschungszentrum received final in 5 Materials Wolf D. E. H6chstleistungsrechenzentrum, (Received Granular of PAGE form and Jiilich, accepted D-52425 12 Jfilich, Germany 1995) October of granular flow, one widely used technique is classical In simulations Abstract. computer Molecular soft-sphere Dynamics, where the equations of motion of the particles are numerically integrated. This requires specification of the forces acting between grains. In this paper, we systematically study the properties of the force laws most commonly used and compare them with experiments on the impact of spheres. We point out possible problems and give criteria recent Finally, two generic problems of soft-sphere for the right choice of simulations parameters. are discussed. Computer modeling and PACS. 07.05Tp PACS. 46.30Pa Friction, PACS. 83.70Fn Granular 1. Introduction Flows of hardness, mechanical contacts, and tribology. solids. granular importance, simulation. adherence, wear, materials ubiquitous are properties (among which sticking [3], density waves [4, 5] in their flowing to Computer have turned simulations granular flow, especially valuable as difficulties far, and experimental are out are "silo and be to there industry and nature. Despite their technological from size segregation [1,2], sudden transitions is no a music" [6]) far are powerful tool to generally accepted from well understood. investigate the physics of theory of granular flow so A very simulation popular scheme is an adaptation of the classical Molecular Dynamics technique. It consists of integrating Newton's equations of motion for a system of "soft" grains starting from a given initial configuration. This requires giving an explicit expression for the forces that act In principle, between grains. should provide such the of touching bodies mechanics but problem contact two expressions, considerable. general conditions is very complicated (see, e.g., [7] and references therein). )(any more strongly simplified force schemes have thus been suggested and employed in simulations or (e.g., [8-20] ), often without a thorough discussion of their properties. three-fold: of The of the firstly, to give a brief theoretical present account paper is aim considerations results and experimental concerning the free impact of spheres; secondly, to critically compare the properties of existing force schemes on these grounds, and to give hints their and thirdly, to discuss problems that can in soft correct on use; some generic occur sphere simulations independently of the force laws used. ~ie perform example simulations to under less illustrate the properties of algorithm [21]. Although (*) Permanent Germanyj e-mail: © Les (ditions address: FB the force it is in 10, laws Theoretical j.schaefer%kfa-juebch.de de Physique 1996 with principle a constant-timestep desirable Physics, to Gerhard work with Mercator fifth order predictor-corrector units, they non-dimensional University, D-47048 Duisburg, JOURNAL PHYSIQUE DE I N°1 ~ ~ v~- v~ w~ Grain I Ri ~ R ~ r~ 2 ~ 2 2 Fig. Definition 1. quantities of the used description for impact. of the always practical; for example, non-linear force laws do not discussed later. Therefore, we decided to use SI units, and cellulose mimic a specific granular material, namely the acetate are will to not define a be tuned all by Drake [22-24] Thus, we are able and Foerster al. [25], of simulation et radius R 3 = as such as parameters spheres used and mm unique timescale, mass m experiments 1.48 x 10~~ kg. data [25]. in = directly to experimental compare our of ourselves non-cohesive, spherical grains which are restricted We limit dry, to the to case (two translational, one rotational), as would be the case m a twothree degrees of freedom between two grains of radii R~, positions r~, dimensional velocities setup. A general contact ii sketched deformation of and angular velocities is The the 1, in Figure grains 2) I. w~ v~, overlap" parametrized by the "virtual (, is to results = j Two unit shear vectors and n s are (0, Ri max = used + R2 )r2 decompose to ri )). forces the and velocities normal into and components: n = s Thus, the normal relative velocity ~n ~~ " = ~~ ~~ )r2 ri = (n~, n~ (ny, -n~). " and ~n relative (V2 VI n (v2 vi s velocity shear + WI RI + ~~ are given by w2R2 with velocity component ~~ is equal to zero at the beginning of a contact, the impact normal, otherwise it is shearing or obliq~te. We begin our discussion in Section 2 or on normal impacts introducing the normal force Fn, and proceed to oblique impacts and the shear force F~ in If the is general, extremes are 3. colliding spheres undergo a perfectly inelastic and perfectly two of transformation heat) Section Impacts Normal 2. In shear head of [7] plastic kinetic energy deformation, into other deformation elastic. forms viscoelasticity which will be Possible of of the energy somewhere mechanisms which material, and for ultimately also elastic between the dissipation ii-e-, transform waves into excited FORCE N°1 by that energy one -~l/~l en the following, (final) quantities. According to Johnson [7], and little The elasticity of the Here SIMULATIONS IN but latter always present they make the noise carry so are Phenomenologically, them of dissipation. generally neglect source can as restitution impact is described by the coefficient of normal en, impact. the SCHEMES in the superscript the necessary lo, il. E = the refers pre-collisional (initial) to and f to post- collisional yield velocity causing plastic critical deformation is given by ~3 y5 107tj G3 ~),~>~ elf Ii) off (RiR2 /(Ri +R2) are the reduced mass and radius, (mim2) /(mi +m2) and R~n yield strength of the softer of the spheres, and E~n is related to Young's modulus E and v])/E2. If ~[ < ~~,~j~, ii VI) /Ei + ii Poisson's ratio v of both spheres through I/E~, deformation plastic during the impact, and all energy loss should be due to occur no may viscoelasticity. If ~[ > ~~,~,~, on the other hand, the energy loss due to plastic deformation dominate the energy loss due to viscoelasticity. have a critical yield Most materials must over velocity that is quite small (e.g. ~~,~,~ m 0.14 m/s for cellulose acetate, ~~,~,~ m 0.02 m/s for where m~a = = Y is the = steel). For the deformation, of plastic case increasing impact velocity. For the obtain coefficient close a to en) ii one, of normal c~ al. [28], Kuwabara and class of different materials compatible is hand, spheres which with show cases m/s < ~[ Modelling a 0.29 < and of sort some l.2 slight a of en narrow (Foerster ~yn is (, the is a force leads that dissipation. damping advantage that the calculation of ~[ on for et al. and the of en range e,i en [27], Bridges others spheres made of a large increasing ~[ which al. [29] for measured [26j Kono increasing ~[; for ~N.ith with impact of 0.87 m in the other acetate velocities used, velocity range a inelastic to The requires collisions simplest force with least at two properties desired the terms: the is repulsion damped force deformation the and m/s). Fn where Kuwabara Goldsmith decrease monotonic dependence rather was oscillator harmonic show systematic no both en decreases Experimental studies by [26] and Sondergaard et Kono all also that ~[~~~~ with to fall off like ~[~~~~ power law over several orders of magnitude of ~[. On by Drake and Shreve [22] and Foerster cellulose et al. [25] on a measurements in case restitution ~[~~~ et simple theory [7] predicts en of purely viscoelastic losses, a and km constant = is -kn( related to important quantities. en For instance, '~" =exp(- stiffness the grain. This model (also analytic solution (with initial of the its (2) 'fn(, referred conditions the to of as ((0) coefficient spring a linear = of 0 elongation whose spring-dashpot) has ~[) allows and ((0) = normal restitution tn), is (3) 2111~ff ~~~~~ ~ i~~i ~" denotes the duration of the collision. ~~~ ~2~ii~ ~ The fmax maximum < ~[t»/~, overlap during ~~~ a collision is is) PHYSIQUE DE JOURNAL I N°1 1.0 ~ ~ ~ ° < o °'~ o ~ o o n ~ ( ~ 8 o o 6 n- o o o ~ ~ ° T © ~ c o~~ ~ o n 0.8 -o ° ° i ~ ~ ° u ° 0.7 ° n ~ a 0.6 A o.5 ~~ jm/Sj ~~ Fig. Dependence 2. normal lo) forces. Kuwabara-Kono of coefficient (9), IA) force normal of spring-dashpot Linear restitution (2), In) Hertz force (10). Walton-Braun e~ impact the on law linear with velocity v[ for various damping (8), IO) Hertz- I). An accurate (reproducing the where the equality holds for elastic grains (en simulation analytic en with relative of the order 10~~) requires At m tn /100. time constant step a errors In principle, force (2) has no free since km and ~yn can be set adjusting en and parameters, exhibited experimental values by a given material in a velocity range tn to the corresponding = for relevant simulations. the = order In formulate to theory of elastic spheres: a behaviour its = refined more force stiffness with (6), en used 0.87 = in works numerous and tn I = x 10~~ s = for the Hertz case Drake's ~fi) properties and the to radii no ~~~ of = ~~~ ~[ the spheres, km 9.0 acetate longer independent of ~( [32] cellulose tn is time of (6) elastic the to collision 3.21 of the results force the -kn(~/~ For the tn been = connected 4/3/$E~n. = for (2), one can use the following repulsive than predicts Fn Here, in is a non-linear spheres through In 10~ Nm~~/~. Note that has it illustrated is kg/s). 2.06 [7, 31, 32], which contact advantages, of its Because [9-12, IS,16,30]. In Figure 2, (w km 7.32 x 10~ N/m, ~yn x (7) km This about In force Hertz for obtain to in an on no which a maximum fashion However, as m For low impact 10~~ x s, and we a -kn(~/~ leads force the have we set viscous a of the time simulation to impact velocity maximum tn /100. damping term At numerical during the = added was to the (8) ~yn(. = this choice [13,17,33]: studies contrary to increases, where the Hertz velocities, simulations, our some Taguchi [34] pointed out, impact velocity In The expected velocities Hertz-type force, Fn the collisions. to relative accuracy. tn m 1.88 dissipative ad hoc timescale intrinsic the numerical m/s, 100 order is depend must satisfactory ensure of there that means At step to collisions experimental results for elastic that evidence: contacts become ii should en be more elastic as ~(~~~~ [35]. regained, force c~ FORCE N°1 (8) produces a behaviour using ~yn of = that so Kuwabara and Kono material the and [26] be to In where the identical is m Figure 2, In zero. illustrate we covered by range this Foerster's of the elastic. -(n(~/~ approach Hertz §n(~/~(, theory Hertz in This and in is force (9) leads connected to a experimental 190 kg m~~/~ s~~. deformation radii of ii results: spheres of the normal restitution en) presented by was ~(~~~ c~ Walton and ki and k2, for the loading constants, spring the to coefficient with agreement = unloading part the original They derive extend [36] of = viscosity. increasing ~(, al. et instead [26, 35]. In Figure 2, we illustrate this using fin An approach guided by the picture of plastic different Braun [37]. They that there are assume and 9 velocity 0.87 in the Brilliantov from of bulk with decreases In the to coefficients two that en en viscoelastic Fn and approaches that restitution kg/s, 0.35 SIMULATIONS IN [25]. measurements assuming coefficient SCHEMES contact: kit ~ k2 If ~ j > 0 ( < 0 pleading) (Unloading) , (0) ~~~~ , to where the is circumstances, k2 a function made be As a (omitted the or remark, model, under abscissa fij. en = the given By making = behaviour want we this " function this the intersects In force F~'*~ achieved during loading, k2 ki + sfj-~, en (sv[(m~n/ki)~/~ + l)~~/~. of the impact velocity ~(, en maximum illustrate we final the unloading curve plastic deformation. ( where permanent of the decreasing a Figure 2, of value ki with comment to on 7.32 " the 10~ N/m and x of the use s = reduced 2 mass can In 10~ m~~. x in m~a the forces studies, it is understood as an additional prefactor for the some damping term [10, II,13,14,17,18, 30], leading to a coefficient of restitution decreases that with Other authors prefactor both the elastic the dissipative it and put term increasing as m~n. in [IS,16,34], such that the coefficient of restitution becomes independent of m,.,. If neither of the function of m~n. Unfortunately, two terms is given a prefactor m~~~ [8, 9,12], en is an increasing research dependence of knowledge there is no systematic experimental the to mass our on en from by work dating cited Goldsmith which indicates slight decrease 1864, except [27], some a of en with result This increasing m~~~ for spheres of equal size. suggests that the first of the possibilities described desirable above might be most suited. It is definitely most that more experimental work be done in order to settle However, when the granular flow this point. without large mass differences, putting m~~~ into the forces merely consists of particles amounts stiffness and for damping to redefining the constant. here for 3. Oblique We now of the force simplicity). Impacts to impacts or shear force Coulomb laws and ~ld that means ~1~ are the for the Contacts where ~( # 0, such that there for short. In general, the turn force, by the F~ Here Frictional and F~ ill) In of friction, ~1~Fn for static = ~1dFn for dynamic case of of static static a non-vanishing shear force is tangential connected to component the normal namely < coefficients is and friction (~~ friction = III 0), (~~ # 0). dynamic friction, respectively. The "<" compensates exactly the (unknown) friction, F~ (12) sign in external JOURNAL 10 PHYSIQUE DE N°1 I If Fj~~ > ~1sFn, one force Fz' applied to the contact, so that ~~ maintained. the 0 is enters dynamic friction regime and equation (12) applies. Normally, ~ls > ~ld and both are about 0.5. of two colliding bodies, a local version of equations ill and (12) In the contact convex area along the contact. For elastically similar bodies, the relates the normal and the shear stresses = shear the stresses results of the of do influence not theory Hertz contact can where area, the normal still be the normal distribution stress for the used stresses the over contact normal stresses. Then, small because the are in area the strains [7], so that regions small, one outer are general expect the condition for dynamic friction to be fulfilled, whereas in the central and friction This strains regions, where large normal static stresses present, are may occur. leads to the development of an ann~tl~ts of microslip surrounding an inner region of sticking friction Because the laws are strongly nonlinear, the size and form of the contact area. in annulus of microslip depend on the loading-unloading history of the contact, making the the deformation and friction forces, in a given prediction of tangential situation, complicated [38]. discussed friction forces Mindlin and Deresiewicz the tangential between elastic two [39] have loading-unloading and assuming the Hertz spheres for the case of several distinct histories theory to hold. Maw et al. [40, 41] and Walton [42] performed such an analysis for the case of the oblique impact of spheres. An interesting result is that due to the ability of the sticking "tangential" kinetic energy, there may be a reversion of tangential and store restore contact to for discs [41] and fact experimentally confirmed velocity ~~ under certain circumstances, a in must spheres [25, 2 7]. following, we implemented varying obliqueness. We directly compare the results to the experimental results of [25]. The linear spring-dashpot force (2) was used as normal force in all cases, because it is simple, robust and, most importantely, absolute makes the results of oblique impacts only dependent on the obliqueness, not on the value of the impact velocity ~[. We adopt the values en 7.32 x 10° N/m, 0.87 (~ km simulations otherwise. 2.06 kg Is) and ~1= 0.25 for all except where explicitly stated ~yn for the obliqueness of There the the angle impact, impact measures are various e-gillustrate To and them the carried properties of the out force shear discussed laws binary of free simulations test in the with impacts = = = ~ ~~~~] = ( ~~ . ~ arCCOS , or the In all We dimensionless are impacts going dimensionless tangential velocity, initial shown in the following, the quantities two to measure final tangential velocity as a ~f as a function of #' and the measured in the coefficient of tangential coefficient center-of-mass as function ~~s of total system particle spins initial f of the that #' tan ~. namely the = restitution fi a function ~l/~i = impact so obliqueness, / ~n restitution e~ zero, are ~f/~> = of siniJ. Plotting #~ ~ers~ts #', the SCHEMES FORCE N°1 SIMULATIONS IN 11 la) 16) 3.0 3.0 2. 2.0 /~ ° 9' o/f~ ° O$~ ~ O 1' O ~.° l' _-~ O-O '~~~~~~~~~~~ aaoo° OOOO° ~'~0.0 2.0 1-o 3.0 ~0.0 4.0 2.0 0 / ll'~ o o 2.0 / / o l.0 0 )/ %fi' ~ °~ O-o o-o 2.0 0 ~a%a° ~0.0 4.0 3.0 2.0 0 ~~~ ~ % l-o ° ~'° o, o o f~ ° ° °'° g -10 a 1-o o-O 3.0 2.0 0 4.0 o-O 0 2.0 i'~ 3. ~i~-~i' plots for denote (14) with d) linear k$/kn theirs we (16) spring off for read any in Figure 3. role for decisive On the the to zero. force that (13) The circles denote grains, 0 e~ (reversal #~ vs. #' directly to we are going compare our the other hand, the form of e as a function of impact obliqueness plays dissipation of granular temperature in the granular system. In Figure 4, acetate Fs this (2). resulting curve. non-frictional For perfectly (frictional losses) I -I < < e~ < or even elasticity). Foerster et al. [25] used plots spheres; cannot provide is discontinuous = -/1. reversal at ~~ the following. the Coulomb of results to plot e(sin ~) for the force laws presented in simplest shear force [10,30] just applies Note force = The Obviously, normal with The [25]. = #~ from cellulose their combined speres = be characterize to a tangential 4.0 dotted, dashed and dot-dashed lines friction law (13) with y friction law 0.15, 0.25, 0.35; b) viscous of la) and 16) according to (15) with ~fs combination 1, 3, 20 kg/sj with ks/kn tangential spring (20) with 1, 2/7, 1/5; e) variable model (22) with (o "10~~ (only dotted line). equal to one; in practice, e~ tangential velocity due to tangential would of tangential forces impact of acetate 1, 2/3, 1/3j f) stick-slip be can the 3.0 i" various for data respectively: a) Coulomb kg/s; c) ~fs =1, 3, 20 = ° ° o / O-o °~ ~'° o experimental 4.0 ~~~ 2.0 the 3.0 3.o 3.o Fig. ~ ~ / , ~0.0 / o o l' aaoa° ~~- 4.0 3.0 2.0 ~~ 3.0 (d) (c) 3.0 ~9- o _~~-~~~~ o ~~~~~~/ O-O o / + law of dynamic friction, thus jfnj signjiJs). of = j13) tangential velocity; When 0 0. ~~ - it can only slow ~~ down (rolling regime), numerically JOURNAL 12 PHYSIQUE DE I N°1 (b) (a) i-o i-o 0.9 0.9 ~~ 0.8 © ~"~ /~ ~< 0.8 fi / ~ ~ 0.7 0.7 ~~0.0 0.2 0.4 0.6 ~~0.0 1-O 0.8 0.2 0.4 (cl 0.8 ,j~ ~ 0.6 0.8 0 0.6 0.8 0 0.9 ~/ ~ 0.8 /~ ~,,,~ ~ 0.8 ~, ~' 0.7 0.7 ~'~0.0 0A 0.2 0.6 ~'~0.0 1-O 0.8 0.2 0A le) (f) i o i o 0.9 0.9 /~ ~h~,~~ 0.8 w 0 (d) 0.9 w 0.6 / 0.8 ~~~~ 0.7 0.7 0.6 0.6 O-o 0.2 0.6 0.4 Sin Fig. 4. Dependence corresponding parameters gets F~ jumping one has effect no amplitude the on of 0A Sin linestyles between final 0.2 O-O 0 coefficient the and 1-o 0.8 as positive velocity ~(, of total in Figure the e on the offset impact sin ~. Forces, 3. negative and as restitution 0 values time average end of the of F~ instead of F~ vanishes 0. = as it However, this should, and the #~ vs. ~i~ as obtained with contact. distinguishes a regime of ~~ where, during the impact ~~ was slowed down to zero (rolling) and a regime where v) was too large, so that a finite ~( resulted. Both regimes are characterized by constant slope, and the transition between them is governed by the value of /~: the higher ~1, the longer the rolling regime is sustained. The corresponding e(sin~) diagram is displayed in Figure 4a. For normal impacts (sin ~ 0), e contribution of the impacts one sees a growing en; for increasingly oblique force to the total shear dissipation, and grazing impacts approach e I, as physical intuition (13) of the shown is in = jumping Figure 3a goes for to ~1 at zero = the 0.15, 0.25,0.35. One = = suggests. Some authors [12, IS, 20, 33,35] use a viscous F~ where ~y~ celeration is shear a is a linear damping function constant of the = force of the form (14) -'f~~~, physical velocity, one without initial friction interpretation. obtains a Because constant e~ here > 0 for the all de#' FORCE N°1 (Fig. 3b) rolling case is never go smoothly to one e (14) is not governed by the normal Thus, in the limits of nearly normal IN reached. On the does tution SCHEMES the grazing for not SIMULATIONS hence does nearly grazing and hand, the (Fig. 4b). impacts Fn and force other 13 coefficient of total This is due to the vanish for grazing not force impacts, resti- fact that impacts. (14) yields unphysical results. discontinuity The (13) in be can Fs Here, ~y~ considered be may properties collisional for to with (14) [II,13,17,19,43-45], (IS) rnirlll~/s~sl,IJLF«I) Sigrll~s). = to be not by combining (13) avoided a technical should which parameter substantially With increasing differ from those of have force value a high enough #' for vs. (13). #~ approaches the behaviour is shown in Figure 3c. ~y~, force (IS) (13). The same trend is observed for e(sin~) (Fig. 4c). Apparently, a small ~y~ changes mainly the properties of "moderately oblique" impacts, making them more elastic. Considering that the form of e(sin~) is decisive for the dissipation of granular temperature in extended granular system, it seems very important that ~y~ is given a high enough value. an A simpler possibility is to use (13) instead of (IS), avoiding a physical parameter of unclear force this of force interpretation. presented so far do not account for tangential elasticity. Therefore, none negative contrary to the experimental a e~ in any part of the #~ vs. #' diagram, data. There is a further disadvantage to them which is important in static or quasi-static systems: a pile made of particles which interact through the force laws (13)-(15) is not stable. Stability would require that finite shear forces act between particles also at ~~ 0 in order to withstand gravitational force components in shear direction of the contacts. Tangential elasticity was first introduced by Cundall and Strack [8] and used by many others [9, 14, 16, 33, 46] writing All the of them force schemes shows = F~ where k~ took some is place that tangential since the min()ks(), = stiffness time ( denotes and to, when the (lt) It is essentially individual half the values period of ratio of the tangential was (16) displacement in the tangential first established, e. direction /~ ~slt')dt" determines k~ /kn that stiffnesses. This be can the II?) results oblique an considering of understood impact that k~ and the not determines a [47], oscillation (fill -i/2 ~ t~ the contact = sign((), )~1Fn)) = ~ + m~ii mR~ II)) (18) oscillation of inertia). Thus just like km determines a half period of normal in II is the moment the phase of the tangential oscillation when the determined by at the moment contact ceases, is fi$ (this proportionality is valid strictly only for en I). For uniform the ratio t~ /tn c~ spheres, = 1 and value force = 2 IS mR~, (19) 2/7. The periods of tangential and normal oscillation equal when k~ /kn are ~( additionally depends on ~(. In Figure 3d, we present ~fi~ vs. #' as obtained 2/7, one obtains a regime of (16) for k~/kn 1, 2/7, 1IS. With k~/kn the = of negative = e~ for small to = intermediate #', caused by the restoring of tangential kinetic actual using constant energy JOURNAL 14 by tangential spring. the regime of sliding a higher #', For region The friction. PHYSIQUE DE the Coulomb where reversal N°1 I condition of (16), governs and tangential velocity initial enters one observed is from normal impacts to ~~ m 1.6 (~ m 58°), in good agreement with the experimental by Foerster et al. [25]. For k~/kn smaller or higher than 2/7, the extension of the subtleties arise. For k~/kn I, there is a slight rise in (e~ < 0) region is unchanged, but new behaviour has been predicted theoretically by lilaw ~~ before it falls below the abscissa; this observed in Foerster's data. For k~ /kn elastic spheres, but is not I IS. et al. [40] for perfectly match the experimental values quite well. The behaviour the simulation results porresponding much of a difference which of e(sin~) is shown in Figure 4d. Here, it does not to make seem extends results = = k~/kn of value chosen. is leads to very realistic impact behaviour used by Walton and Braun was force and patterned after Mindlin's results for normal constant varying tangential force [39], assuming that in each time step, the normal force changes only by a influence force that will not significantly the tangential force. It introduces small amount a dependent k~, such that for each time step Another model Their [37]. that scheme is Fs~fl+ks'l(~l'l' where refers prime a respective the to values previous the in ~ if v~ in initial denotes during impact k~. In by kj stiffness such that (20) they with to seem Brilliantov (thus, = the surfaces et al. [36] experimental formulated values force a tangential force) apparent yield local the if is stress a F~ and [J~j surface saw-tooth tangential to 10~~ used in (o "10~~ m means the = truncation integer initially equal is in the normal value instantaneous kn(I v) Ill v/2). The Poisson 2/3. Figure 3e and Figure 4e =1, 2/3, 1/3. The results look the direction. tangential ratio bit stiffness v vs. quite similar is better for k) /kn the -~1Fn certain of force of z, (o is a using (16), though 2/3. threshold. length of normal obtained With their transmission contacting model, they (o~ (o value occurs materials. most assuming that the tangential moment by microscopic asperities on the a the to mediated exceeds the evaluation related 1/3 for e(sin~) those = in to set inevitably that of Fn k) #~ and to and to 0 force of order is #~ show m derive where F] change for by using the theory [39], the initial match which The accounted elastic kj/kn kj /kn opposite stiffness. direction. its reverses is Mindlin's tangential initial the ~~ (~~~ l/3 F) ~1Fn + S whenever direction ~j i if ~~ in Here, k) step and ~~~ ks of F~ time ~i) ~~ ~0 120) (22) connected to material constants expressed in terms microscopic Force (22) is a parameters. function in ( with period (o. Clearly, (o should be much smaller than (~, the final displacement immediately before the contact ends. (~ is in a range from about 10~~ roughness, and (from nearly simulations. our m (changing (o ~1 is stiffness to nearly grazing) for the impact angles and normal properties of force (22) are shown m Figure 3f and Figure 4f for 10~~ m or10~~ m does not visibly alter the curve). One obtains normal The to of SCHEMES FORCE N°1 SIMULATIONS IN 15 3.0 %1' ~Q ° O/,~ ° '/ II / /~i ,</ l-Q ~9- ,/ p/O O/~' / ~ ,/ II __---~~' Q-Q O coop* ~° 1' / ' '/ -1.0 I-O O-O 2.0 3.0 4.0 if' combination of Hertz-Kuwabara-Kono Fig. 5. normal force (9) with linear tangential ~i~ vs. ~i' for elasticity (16) for v]~, 0.1 m Is (dotted line), 1.0 m/s (dashed line), 10.0 m Is (dot-dashed line). The circles denote the experimental values for the impact of spheres [25]. acetate = of reversion is smaller e seems sin ~ to with be (16) forces overestimated and with for (20). regard small but the hand, the other forces the to angles, impact On the (16) of total (20) especially and region of this extension coefficient restitution the in region o.7. m Finally, let from tangential velocity initial than (2) normal discuss us applied. are force indirectly changes the coefficient The through F~ collisional in behaviour when normal force different laws the tangential restitution on e~ and #~ vs. ~' depend f(Fn) and through tn, the normal oscillation half period: of = ~( ~[ = /~~ A~ (23) F~ dt. c~ force (2) by other changes in the behaviour of normal forces, some expected. We discuss those changes for the dynamic friction force (13) and the elastic tangential force (16). force 11 3) is not likely to be very The non-elastic sensitive to the normal force chosen, because either to rolling or slipping. Of course, choosing a non-linear normal force, it essentially leads velocity-dependent, and the same must be true for A~. In practice, however, the in becomes velocity dependence in (7) is so weak that for a large range of ~[, the changes in #~ vs. #~ are small to be represented here. too tangential force (16), the situation is different. Here, as pointed out before, For the elastic the impact results depend on the ratio of the half period of tangential oscillation t~ and the collision normal In order to get meaningful results, k~ must be set such that for time tn. particular impact velocity, t~/tn I. According to (18) and (19), this leads to the one c m replaces Therefore, if one oblique impacts must be = condition ~ k~ Figure 5, In combination put k~ the = 9.43 different show we with x the 2/7 ~ m~~~ (24) tnc #~ vs. #' for impacts obeying the Hertz-Kuwabara-Kono tangential force (16) with varying impact velocities ~]~, lo~ N/m impact = m velocities order to achieve affect and tn t~ /tn = therefore I for ~)~, the = l collision m Is. It is behaviour. force )v] clearly = On (9) v)). seen the in We how other JOURNAL 16 PHYSIQUE DE I N°1 .o o-g ~~~- ~,, '~, 0.8 w ~ / ', --,__ ~~-, i If ~,~j~_ ' _,/, ~'~'~ 0.7 0.6 O-O 0A 0.2 0.6 sin Fig. Dependence 6. forces and of the parameters coefficient Figure as of total k~ constant selected is smaller times have constants Since with than restitution coefficient the whereas of km, total restitution tangential dissipation, the normal force especially for small impact angles. curve, dependent for true by non-elastic Generic The and and feature describe First consider apart and all grains the whole that a away grain to catch wall impact on a a up wall) hand, limiting damped the contact the In smaller m the with large number considerably mutual [35], discussed of binary force tat other, £~ ~N l at oscillators the case same rather to case between coefficient to/tn lead to e(sin ~) as in the and e(sin ~) of the a velocityThis is influenced is Figure 5. to equations (4) as of I IN, the N /tn the to » wall, a until wait coefficient of restitution ~f N~~ I /tn, where when the I, the grains. restitution so " is total /~o is the coefficient restitution impact, en. a binary hitting the wall leads The with total the needed time for suddenly stopped (e.g. by hitting the wall leads to a for interact grains. to latter column's The column's grains distinct hit column the i=I of it the and < time, f ~ this measure i on front in grain than normal according that is Let and ~~ ~; depends e,~, limiting the that how parameters models ao~is. common each and ~ collisions than spring velocity-dependent. 6 shows and the behaviour forces also Figure forces both the normal both Simulations i=I et al. of on (2), force I. m c function becomes alike. same normal influence non-linear forces along the ~o the achieve column, Luding same decreases with increasing stiffness. We tn that connected duration of contacts. to the finite column of N grains of same and material, all a distance size a ~ As the duration finite are velocity from linear to e(sin~) soft-sphere all one-dimensional a having move by have effects two strong the using the direct a a Soft-Sphere shared collisions normal of en with With tangential elastic of sin ~ for results is has restitution, normal Problems common (7), now of velocity dependence the 4. so coefficient offset impact of those magnitude order of same of the on e Figure 3d (obtained with force (2) the tangential as long as spring with force (9), the numerical value of k~ is nearly that Note care. that be of the to 1-O 5. hand, the results of Figure 5 look very similar to and varying k~ /kn) and also fit the experimental 1000 0.8 0 wall restitution as a On all chain thus the is other grains into of coupled coefficient is close SCHEMES FORCE N°1 SIMULATIONS IN 17 ~ ~lel Fig. Sketch 7. illustrate to brake the failure effect. disperses. detachment Therefore, this regime was called the regime [35]. This example illustrates that it is very important to choose realistic collision times, realistic stiffnesses, especially when dense systems are considered. Note also that in the i-e"inelastic collapse" observed by McNamara and Young in ID granular gases [48] instantaneous of vital importance. contacts are The second effect is related to described by Schifer times in oblique impact and was contact Consider and Wolf [30]. other with relative velocity ~' two grains hitting each )v]~,) under of the duration impact angle ~ as sketched in Figure 7. We ask for the contact, t~~,,,. an 0), t~~,,~ Of course, for normal impacts (sin ~ tn, but for an oblique impact this is not necessarily so. One can distinguish two regimes: a hard-sphere regime, where the stiffness of the tn still holds, and a soft-sphere regime, where the stiffness grains is so high that t,~,,~ unhinderedly. In this case, t~~,,, each other ~/~~, is so low that the grains essentially cross to this in one the case, column = = = = = where in ~ Figure we 7. a of the Cartesian chord initial between coordinate with system and the final contact z-axis m the p$in.t as direction drawn of v]~~, ~[ ~[ on the initial velocity ~' in the two limiting dependence of A~~ hard-sphere limit, the collision follows approximately the law of reflection, such ~~. In the soft-sphere limit, one has the = the In A~~ length is the Introducing ask for now regimes. that 2Rcos~ = c~ ~~~ ~ mn< l/~(t) l/~ t~,i~. dt ~ during the impact, ll~, is independent of ~~ m this regime, while t~~,,~ c~ 1/~', 1/~~. The braking efficiency goes down with increasing ~~; therefore, this regime has been called the brake fail~tre regime. It is clear that in a given impact it depends on ~', for the between the regimes estimate transition km, and ~ which of the regimes applies. An ~fi ~/~'; solving for u~ gives the critical velocity for provided by equating tn m is The so mean that ~ A~~ force c~ = brake failure ~~ ~w cos = ~ fi. (25) ~ 0.999). and ~ Figure 8 for the cellulose 87.4° (sin ~ acetate parameters Figure 8, the two regimes are clearly seen. The dotted corresponds curve takes place at about usual tn between the I x 10~~ s; here, the transition to the regimes m/s, good predicted by equation (25), ~w m 27 m/s. in agreement with the value ~w m 35 Increasing tn by a factor of 10 Ii-e-, decreasing km by a factor of100) leads to a decrease of factor of10 (dashed line in Fig. 8). The corresponding ~w m 3.5 m/s is a speed ~~~ by a failure typically obtained in experiments or simulations. The possible of brake consequences for simulated granular flow have been discussed by Schifer and Wolf [30]. This is In the illustrated in log-log plot = of = = DE JOURNAL 18 PHYSIQUE I N°1 o-i 0.001 0. I I e+01 I e+02 1e+03 v' Fig. t~ = Braking 8. I x 10~~ s, efficiency dashed line: as tn I of function a = x 10~~ initial relative velocity for ~ 87.4°. = Dotted line: s. Conclusion 5. specific and generic properties of force schemes used in the Molecular Dygranular materials. Some of the forces exhibit a quite realistic behaviour namics Which of the force schemes is the most normal and oblique impacts, others less so. in approbe priate for simulations answered in general; this depends very much on the specific cannot geometry, particle density, mean flow velocity, etc. In any case, we strongly recommend testing force schemes, and paying special influence of the corresponding attention various out to the free the flow Complementary discussion for free impact properties. parameters to our on ii. e. applying to rapid granular flow), Sadd et al. [49] have conducted an investigation on the laws for the case of dense properties of several contact systems with long-lasting contacts. approach to granular dynamics is its versatility While the general merit of the soft-sphere and ability to simulate also very dense and /or quasi-static inaccessible systems la regime to hard-sphere simulations), there are also some generic problems which we shortly described in effect and brake failure are especially predominant the preceding section. Both the detachment stiffness kn of the grains is lower than in real systems (or, equivalently, when the when the therefore be taken in normal collision should simulations time tn is too large). Special care to We have discussed simulation rule out the of presence of any of the two effects. Acknowledgments We thank M. Louge for providing fully acknowledge support by the 577/1-1. us with the Deutsche original experimental Forschungsgemeinschaft data from [25]. We through grant grate- No. Wo SCHEMES FORCE N°1 SIMULATIONS IN 19 References iii [2] [3] Williams J-C-, Powder Technol. 15 (1976) 245. Knight J-B-, Jaeger H-M- and Nagel S-R-, Phys. Rev. Lett. Savage S-B- and Sayed M., J. Fl~tid. Mech. 142 (1984) 391. [4] Schick [5] Baxter A-A-, Verveen and K-L- Nat~tre Behringer R-P-, Fagert G-W-, 251 T. (1974) and 70 (1993) 3728. 599. G-A-, Phys. Johnson Rev. Lett. 62 (1989) Eds (Else- 50 (1994) 2825. [6] [7] Tejchman J. and K-L-, Johnson [8] Cundall P-A- and [9] Walton O-R-, in: Amsterdam, vier, [10] [iii [12] [13] [14] [15] [16] ii?] [I8j [19] [20] Gudehus Contact Technol. 76 (1993) 201. (Univ. Press, Cambridge, 1989). O-D-L-, Gdotechniq~te 29 (1979) 47. G., Powder Mechanics Strack of hlechanics Granular Media, J. T. Jenkins and M. Satake, 1983). P.K. and Werner B-T-, Powder Technol. 48 (1986) 239. Thompson P-A- and Grest G-S-, Phys. Rev. Lett. 67 (1991) 1751. Gallas J-A-C-, Herrmann H-J- and Sokolowski S., Physica A 189 (1992) 437. Ristow G-H-, J. Phys. I France 2 (1992) 649. Technol. 71 (1992) 239. Tsuji Y., Tanaka T. and Ishida T., Powder Taguchi Y.-h., Phys. Rev. Lett. 69 (1992) 1367. Zhang Y. and Campbell C-S-; J. Fl~tid Mech. 237 (1992) 541. P6schel T., J. Phys. II France 3 (1993) 27. Lee J. and Herrmann H-J-, J. Phys. A 26 (1993) 373. llelin S., Phys. Rev. E 49 (1994) 2353. Luding S., ClAment E., Blumen A., Rajchenbach J. and Duran J., Phys. Rev. °E Haff R1762. [21] Allen M-P- Tildesley D-J-, Computer and Simulation of Liquids (Clarendon Press, Oxford, 1987). [22] Drake T-G- and Shreve R-L-, J. Rheol. 30 (1986) 981. [23] Drake T-G-, J. Geophysical Research 95 (1990) 8681. [24j Drake T-G-, J. Fl~tid Mech. 225 (1991)121. [25] Foerster S-F-, Louge M-Y-, Chang H. and Allia K., Phys. Fluids 6 (1994) l108. 1<uwabara G. and I<ono Il., Jap. J. Appl. Phys. 26 (1987) 1230. [26] Goldsmith W., Impact (Edward Arnold Publ., London, 1960). [27] [28] Bridges F-G-, Hatzes A. and Lin D-N-C-. Nat~tre 309 (1984) 333. [29j Sondergaard R., Chaney K. and Brennen C-E-, J. Appl. Mech. 57 (1990) 694. [30] Schifer J. and Wolf D-E-, Phys. Rev. E 51 (1995) 6154. H., J. fir die [31j Hertz [32] Landau L-D. and reine Lifschitz angew. u. E-M-, Math. Lehrbuch 92 (1882) der Verlag, Berlin, 1989). [33] Lee J., Phys. Rev. E 49 (1994) 281. [34] Taguchi Y.-h., J. Phys. II France 2 (1992) 2103. [35] Luding S., ClAment E. and Blumen A., Rajchenbach (1994) [36] [37] [38] [39] [40] Brilliantov J. Duran and N-V-, Spahn F., Hertzsch J-M- and P6schel T., and Braun It-L-, J. Rheoi. 30 (1986) 949. O-It- Adams M-J- and Briscoe Mehta, Ed. (Springer, Mindlin Maw Physik, Vol. VII J., Phys. (Akademie- Rev. E 50 4113. Walton A. 136. Theoretischen N., It-D- and Barber B-J-, New Deresiewicz J-R- and m: Granular Matter: York, 1994). H., J. Appl. Mech. Fawcett J-N-, Wear 38 preprint An 20 (1953) (1976) 101. (1995). Interdisciplinary 327. Approach, JOURNAL 20 [41] [42] [43] [44] [45] [46] [47] [48] [49] DE PHYSIQUE I N°1 N., Barber J-It- and Fawcett J-N-, J. Lubr. Technol. 103 (1981) 74. K., J. Mech. Phys. Solids 26 (1978) 139. P0schel T., J. Phys. I France 4 (1994) 499. P6schel T. and Herrmann H-J-, Europhys. Lett. 29 (1995) 123. Melin S., Inn. J. Mod. Phys. C4 (1993) l103. Itistow G-H- and Herrmann H-J-, Phys. Rev. E 50 (1994) Its. Systems, E. Guazzelli and L. Oger, Eds. (Kluwer, Walton O-It-, in: Mobile Particulate Dordrecht, 1995). McNamara S. and Young W-R-, Phys. Fluids A 4 (1992) 496. Sadd M-H-, Tai Q. and Shukla A., Inn. J. Non-Linear Mechanics 28 (1993) 251. Maw Walton
© Copyright 2025 Paperzz