You are on page 1of 13

Powder Technology 145 (2004) 190 – 202 www.elsevier.

com/locate/ owtec

!ilo m"sic and silo #"a$e% gran"lar &low'ind"ced vi(ration
)enson *. +"itea, !handon -. ."inna, !an$aran !"ndaresana,/, *. *esava 0ao(

!chool o& 1ngineering and 2 lied !cience, Princeton 3niversity, Princeton, 45 06544, 3nited !tates (7e artment o& 8hemical 1ngineering, 9ndian 9nstit"te o& !cience, )angalore, 9ndia 0eceived : ;cto(er 200<= received in revised &orm 5 +ay 2004= acce ted 21 5"ly 2004

2(stract 2cceleration and so"nd meas"rements d"ring gran"lar discharge &rom silos are "sed to show that silo m"sic is a so"nd resonance rod"ced (y silo #"a$e. 9n tall and narrow silos, the latter is rod"ced (y stic$–sli &riction (etween the wall and the gran"lar material. -or the discharge rates st"died, the occ"rrence o& &low "lsations is determined rimarily (y the s"r&ace ro erties o& the gran"lar material and the silo wall. The meas"rements show that the "lsating motion o& the gran"lar material drives the oscillatory motion o& the silo. 7 2004 1lsevier ).>. 2ll rights reserved.
*eywords% !ilo #"a$e= !ilo m"sic= !tic$–sli &low= 0esonance= 8ree = ?ran"lar discharge

1. 9ntrod"ction The discharge o& gran"lar materials &rom silos is o&ten characteri@ed (y vi(rations or "lsations o& the silo, termed dsilo #"a$eT, and a lo"d noise, termed dsilo m"sicT A1–9B. )oth o& these are "ndesira(le as silo #"a$e may ca"se str"ct"ral &ail"re, and silo m"sic is a so"rce o& noise oll"tion. 3n&ort"nately, the n"mero"s con&licting st"dies "(lished in the literat"re A1–9B do not give the silo designer a sim le model to "nderstand the hysical rocesses that ca"se the "lsations and to g"ide the silo design or modi&ication that wo"ld revent the "lsations or at least minimi@e their e&&ect. The "r ose o& this st"dy is to investigate the ca"se o& the noise and the "lsations, and the interaction (etween the motion o& the gran"lar material and the motion o& the str"ct"re. !everal st"dies o& the discharge o& gran"lar material &rom silos have noted &l"ct"ations in discharge rate and the rod"ction o& noise and vi(ration A1–9B. The to o& the gran"lar material has (een o(served to move in discrete ste s, altho"gh the discharge &rom the (ottom o& the silo was contin"o"s A1,CB. -or smooth'walled, tall, narrow

silos, "lsations occ"rred d"ring (oth mass and miEed &low. The "lsations were o(served to sto at a critical height o& gran"lar material in the silo A2,10B. +ethods s"ggested &or reventing "lsations incl"de ro"ghening the walls in the transition @one (etween the ("n$er and the ori&ice A2,4,10B and the lacement o& inserts along the silo walls A1B. 9n an early st"dy, Philli s ACB o(served the motion o& sand in a t"(e which had a glass &ace and was closed at the lower end (y a &lat (ottom having a central ori&ice. Ihen the ori&ice was o ened, the sand in the " er art o& the t"(e moved downward intermittently in Her$s. Philli s noted, (when the &low (egins, a c"rio"s rattling so"nd is heard which changes to a distinct m"sical note.. Je also did eE eriments in which the t"(e was &irst artly &illed with merc"ry and then &illed with sand. ;nce again, the &ree s"r&ace o& the sand descended intermittently when the merc"ry was allowed to &low thro"gh the ori&ice. Je o(served that the length o& the col"mn o& sand increased (y a(o"t 2K d"ring the dstic$T hase. -"rthermore , the motion o& the gran"lar material ca"sed the wall o& the t"(e to vi(rate. Th"s, (oth silo m"sic and silo #"a$e occ"rred in his eE eriments, and he s"ggested that the stic$–sli motion o& the sand may (e res onsi(le &or these henomena.

/ 8orres onding a"thor. Tel.% D1 C09 256 456<= &aE% D1 C09 256 0211. 1'mail address% s"ndarF rinceton.ed" (!. !"ndaresan). 00<2'5910/G ' see &ront matter 7 2004 1lsevier ).>. 2ll rights reserved. doi%10.101C/H. owtec.2004.0:.00<

).*. +"ite et al. / Powder Technology 145 (2004) 190–202


!ome recent st"dies have s"ggested that the "lsations are intensi&ied (y a resonant interaction (etween the gran"lar material and the silo str"ct"re A<,5,10,11B. Jow' ever, in one o& these st"dies, TeHchman A10B also noted that the magnit"de and resence o& the &low "lsations were in&l"enced (y environmental &actors, s"ch as tem erat"re and electrostatic e&&ects, which s"ggests that while resonant interaction can intensi&y the "lsations it is not the only re#"irement &or "lsations to occ"r. ;lvare@ and 8lementL A12B have also o(served that the h"midity, an environmental &actor, can have a strong in&l"ence on the sliding motion o& a slowly "shed gran"lar col"mn. They &o"nd that article and wall &rictional ro erties are im ortant in determining stic$–sli motion in gran"lar systems. Jardow et al. A1B cond"cted eE eriments in a silo whose nat"ral &re#"encies were signi&icantly greater than the "lsation &re#"ency. 9n these eE eriments, "lsations clearly occ"rred even in the a(sence o& resonance. These a"thors ro osed that the motion o& the silo was driven (y the ra id acceleration and deceleration o& the gran"lar material in the (in section, which were ca"sed (y the stress &l"ct"ations in the gran"lar material in the ho er section. 2s the gran"lar material in the ho er region de&orms, there are eriods where the mass o& gran"lar material in the (in is not s" orted and the (ed colla ses in a downward ste creating a large im "lse which sha$es the silo str"ct"re. These a"thors o(served "lsations d"ring core &low in a silo that was C m high, 0.C m dee , and 1.2 m wide, and hence the &low $inematics were considera(ly di&&erent &rom those in tall narrow silos. Iensrich A6,9B ro osed that these "lsations are d"e to com ression and dilation waves in the gran"lar material, which are created (y stic$–sli motion (etween the gran"lar material and the silo walls. Jowever, "lsations have also (een o(served in &"nnel &low ("n$ers, where the gran"lar material at the walls does not sli d"ring discharge A1B. Iensrich A6,9B has s"ggested that the "lsation creation mechanism is entirely di&&erent in &"nnel &low, ("t does not give evidence to s" ort his conHect"re. -inally, +oriyama and 5im(oMs A4B &indings s"ggest that the magnit"de o& the "lsations is determined (y how the gran"lar material changes &rom a com ressed state in the ("n$er to a dilated state in the ho er. They also &o"nd that the li$elihood o& a silo discharging with "lsations was de endent on the method "sed to &ill the silo. They did not ro ose a hysical mechanism to eE lain their o(servations. The aim o& this st"dy, which is largely eE erimental, is to o(tain a mechanistic "nderstanding o& silo m"sic and &low "lsations. Thro"gh a com(ination o& so"nd, (ed height, and acceleration meas"rements, it is shown that silo m"sic is driven (y the stic$–sli "lsating motion o& the gran"lar material d"ring discharge and is associated with a so"nd resonance in the air col"mn a(ove the (ed. 7i&&erent wall and gran"lar materials have (een "sed to ro(e their role on &low "lsations and silo m"sic d"ring silo discharge.

2. 0elated st"dies on stic$–sli &riction in gran"lar materials To eE lain the rationale in the choice o& eE erimental meas"rements, it is worth reviewing the generally "nder' stood $inematics o& the discharge o& gran"lar material &rom a (in or ho er and relating these to stic$–sli &riction in gran"lar materials. 1E eriments show that in a tall, &lat' (ottomed cylindrical (in, with walls having a lower &riction coe&&icient than the internal &riction angle o& the gran"lar material, there is a region o& l"g &low at the to o& the &"ll silo. 2s the silo em ties, the si@e o& the l"g &low region decreases, and event"ally all o& the &lowing material is in converging &low. The discharge rate &rom the (in is inde endent o& the height o& material in the (in, rovided the height is greater than a &ew m"lti les o& the diameter o& the ori&ice A1<B and scales as g 1/27 5/2, where 7 is the ori&ice diameter and g is the acceleration d"e to gravity. 0adiogra hic st"dies o& slow dense gran"lar &low in model ("n$ers show that velocity discontin"ities eEist at the transition &rom the (in to the ho er A14B. +eas"rements in a discharging ("n$er indicate that there is a dynamic arch at the transition where the nat"re o& the material &low changes &rom one witho"t de&ormation (a(ove the arch) to one where the material de&orms ((elow the arch) as it a roaches the ori&ice A15B. Press"re meas"rements A1,<,4B near the transition &rom the (in to the ho er indicate that there is also a stress discontin"ity A15B and that there can (e large "lsating stresses, which corres ond to the cyclical &or' mation and (rea$age o& the dynamic arch. This "lsating (ehavior only occ"rs &or dense assem(lies A15B and is very similar to silo #"a$e. The eE eriments that identi&ied the dynamic arch A15B were cond"cted in a ("n$er where the (in to ho er transition determined the location o& the dynamic arch. 9n a &lat'(ottomed silo (s"ch as the one "sed in o"r st"dy), the stagnant material adHacent to the ori&ice creates a ho erli$e region. Th"s, the discharge &rom a &lat'(ottomed silo can (e eE ected to show many o& the &eat"res o(served in ("n$ers. 9& the density o& the material in the l"g &low region a(ove the dynamic arch is high, it m"st dilate as it crosses the arch in order to de&orm in the ho erli$e region. 4as"no et al. A1CB have st"died stic$–sli motion in gran"lar materials "sing a sim le shear device with :0–110 2m glass (eads and 100–C00 2m sand. They o(served stic$– sli motion at low average sli rates, which (ecame contin"o"s at very large average sli rates. They also o(served that at very small driving velocities, the eriod o& stic$–sli &l"ct"ations was inversely ro ortional to the driving velocity. -or glass (eads, the system &l"ct"ated with a nearly constant eriod, while &or sand, the eriod varied stochastically. 2s the sliding velocity was increased, the eriod (ecame inde endent o& velocity, and &inally at large sliding velocities, the motion (ecame contin"o"s. 9n these sim le shear eE eriments, the s ring constant connecting the driving iston to the sliding mass was varied, and it was


).*. +"ite et al. / Powder Technology 145 (2004) 190–202

discharge o& gran"lar materials. 1Ecellent reviews on stic$– sli &riction can (e &o"nd in )owden and Ta(or A16B, *rim A19B, and )erman et al. A20B who disc"ss the vario"s ost"lated stic$–sli &riction mechanisms. The mechanism o& most relevance to this st"dy is adhesive stic$–sli &riction, which occ"rs when slowly wea$ening, time' de endent &orces eEist (etween sliding s"r&aces. The hy othesis in this st"dy is that adhesive stic$–sli &riction is the determining &actor in cyclical dynamic arch &ormation and (rea$age, which creates im "lses that drive the silo str"ct"re.

<. 1E erimental method 2l"min"m, lain steel, and acrylic t"(es, o en at the to and covered at the (ottom with a &lat acrylic late having a concentric ori&ice (see (elow &or details o& this late), were "sed as silos. 2 n"m(er o& eE eriments were cond"cted "sing silos resting on s" orting s rings (see -ig. 1), which in t"rn were attached to a steel &rame that was rigidly connected to the la(oratory walls. The silo was also e#"i ed with rollers and sliders, which were attached to the steel &rame. These allowed vertical oscillation o& the silos and restricted lateral motion. The s" orting s rings had s ring constants ranging &rom 4 to 22C5 4/mm. 1E eriments were also done "sing an al"min"m (loc$ in lace o& the s ring or sim ly (olting the silo directly to the s" orting steel &rameNthese con&ig"rations a&&orded the two largest nat"ral &re#"encies &or vertical silo oscillation re orted in this st"dy. Pro erties o& the t"(es and gran"lar materials are listed in Ta(les 1–<. Photogra hs o& the gran"lar materials, o(tained "sing a microsco e, are shown in -ig. 2a–c. The gran"lar materials did not eEhi(it s#"ea$ing or (ooming when sheared. The tem erat"re and h"midity were recorded in each eE eriment. The tem er' at"re varied (etween 20 and 25 68 (&rom one day to another), and the relative h"midity (etween 16K and 40K. 7"ring eE eriments with each t"(e and gran"lar material com(ination (which lasted a &ew ho"rs), the h"midity variation was within 5K, and the tem erat"re variation was within 2 68. The angle o& internal &riction o& each gran"lar material was estimated (y meas"ring the angle o& re ose o& the gran"lar material (etween two lane walls 1.9 cm a art. These val"es are resented in Ta(le 2. The angle o& wall &riction &or each gran"lar material and t"(e com(ination

-ig. 1. 1E erimental set" &or vertical acceleration and so"nd meas"re' ments. The n"m(ers indicate (1) s ring on ositioning slider, (2) accelerometer, (<) ositioning roller, (4) micro hone, and (5) ositioning slider.

&o"nd that the s ring constant in&l"enced (oth the "lsation &re#"ency and the critical driving velocity at which the "lsation &re#"ency (ecame inde endent o& the driving velocity. 4as"no et al. A1CB also o(served that a lengthy eriod o& slow vertical dilation receded ra id sli events in the hori@ontal direction. This dilation was meas"red (y ?em' inard et al. A1:B, who &o"nd that in the shear @one, articles clim( slowly over each other. ;nce the article is O5K o& a article diameter over the article (elow it, sli occ"rs and the to layer H"m s &orward (e&ore slowing down again and settling into another @one o& articles A1:B. 9n the eE eri' ments o& ?eminard et al. A1:B, the article vol"me &raction was indisting"isha(le &rom the random close' ac$ed vol' "me &raction o& C<K. These st"dies show that gran"lar materials can "ndergo stic$–sli motion, and that this can co" le with the mechanical system (&or eEam le, a mass–s ring system) in a com licated &ashion, which de ends on the system arameters. Together, these st"dies s"ggest that stic$–sli motion can occ"r in tall &lat'(ottomed silos d"ring the

Ta(le 1 T"(e ro erties T"(e material C0C1'TC al"min"m alloy Plain steel 8ast acrylic Pength (m) 1.6 1.6 1.5 97 (cm) C.<:-0.01 C.<6-0.01 C.<5-0.01 Iall thic$ness (cm) 0.51-0.1 0.5:-0.01 0.C4-0.01 !"r&ace &inish !mooth 0o"gh !mooth 8alc"lated lowest nat"ral &re#"ency (J@) 2C 25 12

).*. +"ite et al. / Powder Technology 145 (2004) 190–202 Ta(le 2 ?ran"lar material ro erties +aterial !" lier Particle si@e (2m) 450-50 460-C0 400-100 Particle density (g/cm<) 2.5-0.1 2.5-0.1 2.:-0.1


2ngle o& re ose (6) <4-1 2C-1 <<-1

8r"shed glass )allotini im act (eads Iashed and ignited !tandard ;ttawa sand

Potters 9nd"stries Potters 9nd"stries 1+7 !cience

-ig. 2. Photogra hs o& the gran"lar materials% (a) cr"shed glass, (() glass (eads, and (c) sand.

was estimated (y meas"ring the angle o& inclination o& the t"(e a(ove the hori@ontal at which the gran"lar material (egan to slide= these val"es are listed in Ta(le <. Ie co"ld not estimate the angle o& wall &riction &or the lain steel t"(e thro"gh s"ch eE eriments, as the gran"lar material (egan to slide over itsel& (e&ore it slid at the wall. To determine i& resonance (i.e., when the "lsation &re#"ency & is e#"al to a nat"ral &re#"ency o& the silo) was im ortant, the dominant nat"ral &re#"ency o& vertical silo oscillations, & v, was changed (y "sing di&&erent s rings (etween the silo and the steel &rame. To estimate the lowest nat"ral &re#"encies o& the em ty silo t"(e, we "sed the method o& 4aeem and !harma A21B with clam ed'&ree (o"ndary conditions. The lowest nat"ral &re#"ency esti' mates &or each t"(e are given in Ta(le 1.

2crylic lates with centrally located ori&ices (with diameters (etween 1.< and 2.5 cm) were (olted to a 1.21'$g al"min"m &lange, which was screwed on the (ottom o& the t"(es. To &ill the silo, the ori&ice at the (ottom o& the t"(e was &irst sealed with a iece o& d"ct ta e. The gran"lar material was o"red into the silo thro"gh a &"nnel laced at the to o& the t"(e. !tri ing away the d"ct ta e seal over the ori&ice initiated discharge. The mean discharge rate was meas"red "sing a sto watch. -or the acrylic t"(e, the height o& material in the silo co"ld also (e meas"red d"ring discharge to con&irm that the discharge rate was constant with time. To ens"re that the t"(es had reached a steady state o& wear, the gran"lar material was discharged several times thro"gh the same t"(e (e&ore &inal meas"rements were ta$en. !teady state wear was reached when re eata(le gran"lar material acceleration meas"rements co"ld (e ta$en. 2ccelerations were meas"red (oth in the gran"lar material and on the silo str"ct"re. >ertical accelerations inside the gran"lar material were meas"red "sing a "nidirectional *istler 6::4250 low'im edance ceramic shear accelerometer with an o"t "t sensitivity that deviated less than 1.5K &or &re#"encies (etween 10 J@ and 10 $J@. The accelerometer was em(edded a roE' imately 5 cm (elow the to s"r&ace o& the gran"lar material. This de th ens"red that d"ring discharge, the accelerometer was held " right (y the gran"lar material and was still shallow eno"gh that the acceleration co"ld (e meas"red &or the ("l$ o& the discharge. 2s the gran"lar material discharged, the accelerometer ca(le was care&"lly &ed into the silo to ens"re that the ca(le did not a&&ect the motion o& the accelerometer. This accelerometer had a range D/Q500 m/s2 and was acc"rate to within D/Q5 m/s2. 9t had a diameter o& 0.6 cm, a length o& 2.C cm, and a

Ta(le < 2ngle o& wall &riction &or vario"s silo wall and gran"lar material com(inations !ilo wall material R gran"lar material 8r"shed glass ?lass (eads !and

2crylica (6) 26 1: 25

2l"min"ma (6) << 1: <0

Plain steel -"lly ro"gh -"lly ro"gh -"lly ro"gh

The angles are acc"rate to within -26.


).*. +"ite et al. / Powder Technology 145 (2004) 190–202

mass o& 4 g, and hence was considera(ly larger than a sand grain. Jowever, the wide &re#"ency res onse allowed (etter time resol"tion o& the ("l$ gran"lar material acceleration than wo"ld (e ossi(le with smaller accel' erometers o& com ara(le cost. !ilo str"ct"ral vi(rations were meas"red "sing a *istler 6:6425 low'im edance ceramic shear accelerometer, which had a greater sensitivity ("t a smaller range than the accelerometer "sed to meas"re gran"lar material acceler' ations. To meas"re vertical accelerations, this accelerometer was waE'mo"nted on the &lange at the (ottom o& the silo. This accelerometer had a sensitivity that varied (y less than 0.5K &or &re#"encies (etween 10 J@ and C $J@. 9t had a range o& D/Q50 m/s2, an acc"racy o& D/Q0.5 m/s2, and a mass o& 21 g. The accelerometer o"t "t was sent thro"gh a *istler 5116)2 signal conditioner to a +eas"rement 8om "ting P89'72!1002 data'ac#"isition card on a 400'+J@ Penti"m 99 com "ter. The sam ling rate on the data ac#"isition card was 20 $J@. -or (oth accelerometers, the man"&act"rer' s" lied cali(ration was "sed to convert the accelerometer voltage o"t "t to acceleration. The accelerometers co"ld not (e "sed sim"ltaneo"sly (eca"se only one data ac#"isition system was availa(le. The ("l$ o& the so"nd meas"rements was ta$en in an a arat"s made &rom an acrylic t"(e, &or which the resonant &re#"ency &or vertical silo oscillations was not well controlled A22B. Jowever, several meas"rements were then re eated in the eE erimental set" "sed &or the acceleration meas"rements to chec$ that the same res"lts were o(tained. 9n these eE eriments, an omnidirectional ; tim"s <<'<02C la el micro hone with a constant am lit"de res onse &or a &re#"ency range (etween <0 J@ and 15 $J@ was "sed to collect the so"nd data thro"gh a so"nd card on a ersonal com "ter. 7"ring discharge, the so"nd was recorded, and a discrete -o"rier trans&orm o& 1 s o& so"nd data was "sed to determine the dominant &re#"ency as a &"nction o& time d"ring discharge. 9n the acrylic t"(e, the time at which the to o& the gran"lar material crossed a mar$ed height in the t"(e d"ring discharge was also recorded "sing a sto watch. -rom these meas"rements, the height o& the gran"lar material as a &"nction o& time since discharge started was &o"nd.

this t"(e. !ilo "lsations also did not occ"r when sand was discharged &rom the al"min"m t"(e, ("t did occ"r when sand was discharged &rom the acrylic t"(e. 2 &ew eE eri' ments with a smooth'walled galvani@ed steel t"(e A2<B showed that silo "lsations occ"rred when sand was discharge &rom this t"(e. This s"ggests that in tall and narrow silos, "lsations occ"r &or s eci&ic com(inations o& gran"lar material and wall material. This is in agreement with st"dies which show that stic$–sli &riction de ends on the com osition o& the sliding s"r&ace A16–20B.

4.1. !o"nd meas"rements -ig. < shows the so"nd am lit"de level as a &"nction o& time &or the discharge o& sand &rom the acrylic t"(e (no "nits are given, (eca"se altho"gh the am lit"de is a direct voltage reading &rom the micro hone that is linearly related to the so"nd deci(el level, (y moving the micro hone, di&&erent a(sol"te deci(el levels can (e recorded &or the same so"nd signal). The discharge lasted &or 51 s, and as shown in the &ig"re, silo m"sic occ"rred &or a roEimately hal& o& this time. -ig. 4 shows a ty ical ower s ectr"m &or the so"nd meas"rements d"ring discharge, determined (y analy@ing data o(tained over a 1's time interval. There are three ty es o& rominent ea$s. The &irst ea$ is at a &re#"ency o& a roEimately 40 J@, and it will (e shown later that this is the "lsation &re#"ency &or this artic"lar gran"lar material and silo com(ination. The second ea$ corres onds to the resonant &re#"ency o& the air col"mn a(ove the t"(e. This resonance is well doc"mented, and a good acco"nt can (e &o"nd in 0ayleigh A24B. 2t the time the data shown in the &ig"re was collected, this &re#"ency was 200 J@. The &act that this ea$ re resents a resonance &re#"ency is demon' strated in -ig. 5, which shows the #"arter wavelength corres onding to this &re#"ency as a &"nction o& time since

4. 0es"lts 9n what &ollows, so"nd meas"rements are shown &or sand discharging &rom the acrylic t"(e, and acceleration meas"re' ments are shown &or cr"shed glass and glass (eads discharging &rom the al"min"m t"(e. 2dditional so"nd and acceleration meas"rements are re orted in ."inn A22B and +"ite A2<B, res ectively. The variation o& the "lsation &re#"ency ( & ) with the dominant nat"ral &re#"ency &or vertical oscillations ( & v) was eEamined &or all t"(e and gran"lar material com(inations, eEce t &or the lain steel t"(e, as "lsations did not occ"r in

-ig. <. >ariation o& so"nd am lit"de with time d"ring discharge o& sand &rom an acrylic t"(e o& :.C cm o"ter diameter, wall thic$ness 0.< cm, and having an ori&ice o& diameter 1.9 cm% region 9Nno &low, region 99N&low with "lsations, and region 999N&low a&ter "lsations have ended.

).*. +"ite et al. / Powder Technology 145 (2004) 190–202


4.2. 7etermination o& the nat"ral &re#"ency &or vertical oscillations o& the silo To determine the dominant nat"ral &re#"ency &or vertical oscillations, & v, the silo was &illed with gran"lar material and the ori&ice closed. The (ase o& the &illed silo was then str"c$ with a so&t mallet, and the res"lting vertical acceleration d"ring &ree oscillations recorded. The val"e o& & v was &o"nd either (y "sing the largest ea$ in the ower s ectr"m o& the acceleration or (y co"nting the n"m(er o& &ree oscillations d"ring a s eci&ied time directly &rom the acceleration meas"rements. The two meas"re' ments gave essentially the same res"lts= however, when & v(<0 J@, co"nting the n"m(er o& oscillations in a s eci&ied time gave a more acc"rate meas"rement o& the nat"ral &re#"ency than locating the center o& the (road ea$ o(tained &rom the ower s ectr"m. !imilarly, when the & v4<0 J@, the ower s ectr"m was a (etter indicator o& the nat"ral &re#"ency, (eca"se an "nam(ig"o"s shar ea$ co"ld (e located, while the acceleration vs. time trace showed ra idly decaying oscillations which were not easy to co"nt. -or s ring constants, $(1000 4/mm, & vc& nS(1/ 2 )($/m)1/2, where m is the oscillating mass, and & n is the theoretical nat"ral &re#"ency &or a s ring mass system. -or $41000 4/mm, & v was signi&icantly less than & n, ossi(ly (eca"se o& &lange and t"(e de&ormations, which red"ced the e&&ective sti&&ness o& the system. This e&&ect was im ortant &or & v425 J@.

-ig. 4. Ty ical ower s ectr"m &or 1 s o& so"nd meas"rements d"ring silo m"sic when sand is discharged &rom an acrylic t"(e o& :.C cm o"ter diameter, wall thic$ness 0.< cm, and having an ori&ice o& diameter 1.9 cm% (9) the "lsation &re#"ency, (99) the dominant so"nd &re#"ency, and (999) the higher harmonics o& the dominant so"nd &re#"ency.

the (eginning o& discharge. The wavelength, $ a, is &o"nd &rom the relationshi $ aSc/& a, where c is the s eed o& so"nd in air, and & a is the &re#"ency o& the air col"mn. 2lso shown is the height o& the air col"mn a(ove the sand in the t"(e. This &ig"re shows that the dominant #"arter wavelength and the height o& the air col"mn are the same con&irming the resonant (ehavior. 9t is clear &rom the #"arter wavelength that this resonance corres onds to a standing wave mode with a node at the gran"lar material s"r&ace and an antinode at the o en end o& the t"(e (as the o en end o& the t"(e cannot (e a node). -ig. 4 also shows a n"m(er o& other ea$s at higher &re#"encies, which are sim ly the odd harmonics o& the &"ndamental (lowest) resonance &re#"ency o& the air col"mn.

4.<. 2cceleration meas"rements d"ring discharge -ig. C shows meas"rements o& the vertical acceleration o& the silo when cr"shed glass was discharged thro"gh a 1.9' cm ori&ice. The accelerometer was mo"nted on the (ase o& the silo. ;nce the &low started, there was a eriod o&

-ig. 5. >ariation o& the dominant #"arter wavelength ($ d/4) and the height o& the air col"mn (J a) with time d"ring discharge o& sand &rom an acrylic t"(e o& :.C cm o"ter diameter, wall thic$ness 0.< cm, and with an ori&ice o& diameter 1.9 cm.


).*. +"ite et al. / Powder Technology 145 (2004) 190–202

-ig. C. >ertical acceleration meas"rements on the (ase o& the al"min"m silo d"ring discharge o& cr"shed glass thro"gh a 1.9'cm ori&ice. The silo had a dominant nat"ral &re#"ency o& vertical oscillations o& 6 J@% region 9Nno &low, region 99N&low with "lsations, and region 999N&low a&ter "lsations have ended.

"lsations d"ring which the silo eE erienced large negative accelerations towards the earth. Jal&way d"ring the "lsa' tions, the magnit"de o& the negative "lsations s"ddenly do"(led. 2&ter the height o& the gran"lar material in the silo &ell (elow a critical level, the "lsations sto ed, and the silo str"ct"re eE erienced only small accelerations "ntil the &low ended. Ihile the "lsations occ"rred reg"larly, this do"(ling o& the "lsation magnit"de was not always re eata(le. 9t is not clear what changes in the &low res"lted in these changes in the magnit"de o& the "lsations, (eca"se the (asic set" was "nchanged &rom r"n to r"n. The close" o& the acceleration meas"red d"ring "lsa' tions (-ig. :) reveals that the eriods o& large negative accelerations were short com ared to the grad"al re(o"nd a&ter each "lsation. ;n this time scale, the "lsations had a very re rod"ci(le and steady &re#"ency, ("t the a(sol"te magnit"de o& the maEim"m acceleration varied &rom "lse to

-ig. 6. >ertical acceleration meas"rements made when the accelerometer was em(edded in the gran"lar materials A(a) cr"shed glass, (() glass (eadsB and allowed to translate with it d"ring discharge &rom the al"min"m silo% region 9Nno &low, region 99N&low with "lsations, and region 999N&low a&ter "lsations have ended. 9n (a), ori&ice diameter is 1.< cm= the dominant nat"ral &re#"ency o& vertical oscillations o& the &illed silo is 6 J@. 9n ((), ori&ice diameter is 1.9 cm= the dominant nat"ral &re#"ency o& vertical oscillations o& the &illed silo is C J@.

"lse. Ie &o"nd that & O<0 J@ &or discharge o& gran"lar material thro"gh a 1.<'cm ori&ice and thro"gh a 1.9'cm ori&ice, i.e., & was ro"ghly inde endent o& the discharge rate. -ig. 6a shows meas"rements o(tained with the accel' erometer ("ried in the gran"lar material. The &low con' ditions were the same as in -ig. C, eEce t that the ori&ice diameter was 1.< cm instead o& 1.9 cm. 2lso as in -ig. C, negative accelerations are towards the earth. -ig. 6a shows that large ositive accelerations occ"rred in the gran"lar material d"ring "lsations, while -ig. C shows that the silo eE erienced large negative accelerations. The two &ig"res show that d"ring each "lsation, the gran"lar material &ell a short distance and im acted the t"(e wall and &lange (ottom. -ig. 9a shows a close" view o& the acceleration re orted in -ig. 6a. 2 com arison o& -igs. : and 9a reveals that the "lsation &re#"ency in the gran"lar material is the same as the
-ig. :. 8lose" showing individ"al "lsations meas"red (y the accel' erometer on the silo str"ct"re &or the &low in -ig. C.

).*. +"ite et al. / Powder Technology 145 (2004) 190–202


re&lected in their ower s ectra= com are -ig. 10a and (. )oth s ectra have the same high'&re#"ency decay &or &re#"encies a(ove 1000 J@= however, &or &re#"encies (elow 1000 J@, the glass (eads have a larger n"m(er o& distinct harmonics than the cr"shed glass. The cr"shed glass ower s ectr"m is ty ical o& white noise with a high'&re#"ency c"to&&, while the glass (ead ower s ectr"m is ty ical o& a signal rod"ced (y a well'correlated eriodic ("t non' sin"soidal &"nction A25B. !"r risingly, all the ower s ectra &or acceleration meas"rements inside the gran"lar material &or all t"(e and gran"lar material com(inations that "lsated decayed &or &re#"encies a(ove 1000 J@. The high'&re#"ency c"to&& o& 1000 J@ was neither d"e to any limitation o& the accelerometer (which co"ld meas"re &re#"encies " to 10 $J@) or the lowest nat"ral &re#"ency o& the t"(e (which was varied in these eE eriments and did not a&&ect the high'

-ig. 9. 8lose" showing individ"al "lsations meas"red (y the accel' erometer em(edded in the gran"lar material &or the &low in -ig. 6% (a) close" &rom -ig. 6a, and (() close" &rom -ig. 6(.

&re#"ency with which the silo moves, s"ggesting that the motion o& the gran"lar material drives the motion o& the silo. -ig. 9a also shows that each "lsation was &ollowed (y a negative acceleration within the material and then a second large ositive acceleration, a&ter which the acceleration o& the gran"lar material was close to @ero "ntil the neEt "lsation. -ig. 10a shows a ower s ectr"m &or the acceleration meas"red d"ring 1 s o& "lsations in the cr"shed glass. 9t has a ea$ at & O<0 J@ &ollowed (y a &lat'(and region (etween 200 and 1000 J@, a&ter which the ower s ectr"m decays. -igs. 6(, 9(, and 10( are similar to -igs. 6a, 9a, and 10a ("t are &or glass (eads discharging thro"gh a 1.9'cm ori&ice (the acceleration s i$e seen in -ig. 6( (e&ore the discharge was initiated came a(o"t (eca"se o& an accidental ta o& the t"(e, and its e&&ect decayed well (e&ore discharge was started). The "lsations again sto ed at a critical height, and the individ"al "lsations can (e seen in -ig. 9(. The nat"re o& each "lsation &or glass (eads (-ig. 9() is a little di&&erent than &or the cr"shed glass (-ig. 9a), and this is

-ig. 10. Power s ectra &or 1 s o& the meas"rements in -ig. 6. The ower s ectra have (een averaged over 4 oints in &re#"ency to ma$e average trends clearer% (a) 20th second o& meas"rements o& -ig. 6a, and (() 15th second o& meas"rements o& -ig. 6(.


).*. +"ite et al. / Powder Technology 145 (2004) 190–202

o& the maEim"m acceleration. The threshold was adH"sted de ending on the ty e o& acceleration time gra h to ens"re the correct eriodicity was o(tained. 9n artic"lar, (y com aring -ig. 9a and (, one o(serves that i& the threshold is too low, a higher eriodicity wo"ld (e meas"red in some eE eriments, (eca"se the (a&tershoc$. wo"ld also (e incl"ded. !imilarly, i& the threshold is set too high, some #"a$es co"ld (e missed, (eca"se as shown in -ig. 6a and (, the ea$ am lit"de co"ld vary d"ring #"a$ing. The "lsation &re#"ency was determined &or each second o& &low "lsations and an average "lsation &re#"ency d"ring "lsating discharge o(tained. The standard deviation in the average &re#"ency meas"red d"ring a single discharge was ty ically less than 10K. Ihen & v(25 J@, & had no de endence on & v as shown in -igs. 11 and 12. -ig. 12 also shows that do"(ling the ori&ice diameter and hence increasing the discharge rate (y nearly a &actor o& C had a negligi(le e&&ect on the "lsation &re#"ency (do"(ling the ori&ice diameter gives a nearly siE&old increase in discharge rate in a silo o& constant cross'sectional area, (eca"se the discharge rate is ro or' tional to the ori&ice diameter to the ower 2.5 A1<B). Ihen & v425 J@, & had a ositive correlation with & v &or all

-ig. 11. >ariation o& the "lsation &re#"ency ( & ) o& the &illed silo with the dominant nat"ral &re#"ency &or vertical oscillations o& the silo ( & v) &or gran"lar materials discharging &rom the acrylic t"(e.

&re#"ency c"to&&). 9t may (e related to the t"(e diameter or to the article si@e and article density, which were not varied in the eE eriments. 9n -igs. :, 9a, and (, the maEim"m downward accel' erations o& the silo and articles are ro"ghly com ara(le, whereas the maEim"m " ward acceleration o& the gran"lar material is signi&icantly greater than that o& the silo. This s"ggests that d"ring each "lsation, the gran"lar material sli s ast the silo walls and is &orced to rest over a very short time eriod. This im act creates a shoc$ wave that travels thro"gh the gran"lar material and is recorded as the large " ward acceleration. The gran"lar material and silo then move together so that the res"lting accelerations are o& similar magnit"de.

4.4. 7e endence o& the "lsation &re#"ency o& the gran"lar material on the nat"ral &re#"ency o& vertical silo oscillations -igs. 11 and 12 show the variation o& the "lsation &re#"ency ( & ) &or di&&erent gran"lar material and silo wall com(inations as a &"nction o& & v (recall that & v was changed (y changing the s ring on which the silo was mo"nted). To determine & , the n"m(er o& ea$s er "nit time a(ove a certain threshold in the acceleration time data was co"nted. The threshold was determined (y loo$ing at the acceleration time trace and ic$ing a val"e a roEimately e#"al to a hal&

-ig. 12. >ariation o& the "lsation &re#"ency ( & ) o& the &illed silo with the dominant nat"ral &re#"ency &or vertical silo oscillations ( & v) &or gran"lar materials discharging &rom the al"min"m t"(e.

).*. +"ite et al. / Powder Technology 145 (2004) 190–202


gran"lar material t"(e wall com(inations that "lsated, eEce t &or the acrylic and cr"shed glass com(ination. The &ig"res show that glass (eads have similar &re#"ency (ehavior in the acrylic and al"min"m silos. 8r"shed glass has a lower "lsation &re#"ency in the al"min"m silo as com ared to the acrylic silo. 2s sand did not "lsate d"ring discharge &rom the al"min"m silo, no data oints are shown.

4.5. 8ritical height The critical height (J c) was ta$en as the height o& the gran"lar material a(ove the (ase o& the silo at which "lsations sto . The time at which this occ"rred was recorded &rom the acceleration meas"rements, and as the discharge rate was inde endent o& time, the critical height co"ld (e calc"lated. This method gave critical heights that were in agreement with direct meas"rements made &or the trans arent acrylic silo. -or & v(25 J@, J c did not vary (y more than 0.1 m when the ori&ice diameter and s ring constant were changed. -or & v425 J@, the variation o& J c with silo and gran"lar material ro erties was not closely eEamined. -or all the gran"lar materials, the val"es o& J c &or the acrylic silo were smaller than those &or the al"min"m silo (Ta(le 4). 9n the al"min"m silo, cr"shed glass had a sig' ni&icantly smaller val"e o& J c than glass (eads. 9n the acrylic silo, all three gran"lar materials had similar val"es o& J c. 2s &rictional ro erties can de end on stress level A1CB, an eE eriment was er&ormed in the al"min"m silo, where the to o& the gran"lar material was loaded with a $nown weight a&ter the silo had (een &illed. 2s eE lained in >anel et al. A2CB and ;varle@ et al. A2:B, s"ch a test may yield di&&erent stress transmission characteristics &or di&&erent eE erimental roced"res= nevertheless, it can also increase the stress level inside a gran"lar material. The weights were laced on to o& the gran"lar material in the &illed silo and away &rom the walls o& the silo. The val"es o& J c and & were calc"lated "sing time and acceleration meas"rements d"ring discharge. The eE eriments showed that & was inde endent o& the overload. -or glass (eads, J c did not vary with overload= however, &or cr"shed glass, J c decreased linearly with as the overload increased (-ig. 1<). 1E eriments in the acrylic silo gave similar res"lts.

-ig. 1<. >ariation o& critical height with overload &or glass (eads and cr"shed glass in the al"min"m t"(e when the dominant nat"ral &re#"ency o& vertical oscillations o& the &illed silo is 6 J@ and the silo has a 1.9'cm ori&ice.

5. 7isc"ssion 9n this section, a mechanism &or the rod"ction o& "lsations is s"ggested. The res"lts are then com ared with those o(tained in revio"s wor$ on "lsating gran"lar materials, and some s"ggestions &or &"rther wor$ are made. 5.1. 2 mechanism &or rod"cing silo #"a$e 3sing the (ac$gro"nd on stic$–sli &riction in gran"lar materials disc"ssed earlier, one can com are the eE erimen' tal o(servations in this st"dy with those in revio"s st"dies to #"alitatively eE lain the hysical mechanism &or stic$–sli motion. The dynamic arch which &orms in s"ch &lows is art o& a &orce chainNthat is, a article contact networ$ thro"gh which stresses are transmitted A26B. The arch is &ragile, and conse#"ently when the material (elow it has discharged eno"gh so that the arch is "ns" orted &rom (elow, a slow cree ty ically o(served in adhesive stic$–sli &low (egins. 7"ring this cree , the adhesive &riction &orces (ecome rogressively wea$er and wea$er, and event"ally the arch will (rea$. ;nce the arch colla ses, com lete sli occ"rs, a #"a$e is o(served, and a new arch is created. This #"a$e can set " str"ct"ral vi(rations o& decaying am lit"de that then colla se the newly &ormed arch= in this manner, a series o& sel&'s"stained "lsations res"lts. This is the "lsation rocess o(served in this st"dy, where the discharge rate is &ast eno"gh ((etween 1 and 6 cm/s) that it does not a&&ect the & "nli$e in IensrichMs st"dy A6,9B.

Ta(le 4 >ariation o& critical height with silo wall and gran"lar material ro erties !ilo wall material 2l"min"m 2l"min"m 2crylic 2crylic 2crylic

?ran"lar material 8r"shed glass ?lass (eads 8r"shed glass ?lass (eads !and

8ritical heighta (m) 0.9 1.< 0.6 0.C 0.:

The acc"racy o& the critical height data is -0.1 m.


).*. +"ite et al. / Powder Technology 145 (2004) 190–202

9n IensrichMs eE eriments A6,9B, the entire (ottom o& a cylindrical model silo was slowly lowered. There is no region o& converging &low in s"ch an eE eriment, ("t there is a region near the (ottom where the gran"lar material dilates as the iston descends. Jere, the arch may (e regarded as the (o"ndary (etween the dense and dilated material. -or the slow discharge rates eEamined (y Iensrich, cree and eEternal ert"r(ations did not determine & . 9nstead, the arch colla sed whenever the articles (elow the arch had dro ed eno"gh to lose contact with the arch. 8onse#"ently, it is entirely reasona(le that & was inversely ro ortional to the discharge velocity. Iensrich estimated the distance that articles at the (ase o& the silo moved (etween "lsations to (e O0.0< d (where d is the article diameter), which is com ara(le to the dilation distance o& O0.05 d re#"ired &or sli to occ"r in stic$–sli &low in sim le shear eE eriments with gran"lar materials A1:B.

5.2. 8om arison with revio"s wor$ Iensrich A6,9B o(served that the acceleration rod"ced (y each #"a$ing im "lse grew with distance traveled (y the wave carrying the in&ormation o& the im "lse &rom the dynamic arch to the to o& the silo. 9n contrast, the gran"lar material accelerations recorded in o"r eE eriments with the accelerometer at a &iEed de th (elow the &ree s"r&ace did not change a recia(ly as the (ed height decreased d"ring discharge. The gran"lar material accelerations meas"red (y Iens' rich A6,9B were less than 15 m/s2, while those meas"red in this st"dy were ty ically more than 100 m/s2. TeHchman A10B also o(served silo wall acceleration levels greater than 100 m/s2. 4onlinear e&&ects may (e res onsi(le &or the height inde endence o& the acceleration at the large accelerations seen in this st"dy. 9t is interesting to contrast the acceleration ower s ectra o(tained with cr"shed glass and glass (eads to see the e&&ect o& article sha e on gran"lar dynamics. The ower s ectra &or acceleration meas"rements in the glass (eads (-ig. 10() showed many harmonics o& & (e&ore the high'&re#"ency decay region was a roached. The ower s ectra &or cr"shed glass (-ig. 10a) showed only a &ew harmonics, &ollowed (y a (and'limited white noise region, and then a high'&re#"ency decay. This s"ggests that the glass (eads showed a highly correlated distri("ted res onse to sli which originated at the arch. 8r"shed glass had a signi&icantly less correlated res onse, #"ite ossi(ly (eca"se o& the heterogeneity in article sha e and article contacts (etween them. This is consistent with the s"ggestions (y +air et al. A29B that smooth ro"nd articles have &orce chains that are sta(le over a narrow range o& orientations, whereas ro"gh articles s"ch as the cr"shed glass have &orce chains that are sta(le over a wider range o& orientations. 8onse#"ently, the &orce chains in the glass (eads (rea$ in a highly correlated manner d"ring a "lsation, whereas those in the cr"shed glass (rea$ in a less'correlated manner.

2crylic s"r&aces are rone to stic$–sli motion A<0B. 2s acrylic is so&ter than all the articles "sed in this st"dy, the acrylic s"r&ace can (e eE ected to (e the dominant &actor in determining the adhesive relaEation time &or the stic$–sli motion, and indeed it was &o"nd that the "lsation &re#"encies &or all gran"lar materials are similar in the acrylic silo. The hy othesi@ed di&&erence in stress chain (ehavior (etween smooth and ro"gh articles s"ggested (y +air et al. A29B can also eE lain the di&&erence in the val"e o& the critical height when an overload is im osed on the gran"lar material. -or ro"gh articles, the critical height decreased linearly with im osed overload, whereas &or smooth articles, the critical height was inde endent o& the im osed overload. 2s the &orce chains in a gran"lar material com osed o& smooth s heres will have narrow direction' ality, the e&&ects o& the im osed overload will (e transmitted to the side walls o& the silo ra idly and will not a&&ect stress levels (etween the silo wall and the gran"lar material a signi&icant distance away &rom the overload. These &orce chains &orm a (ridge so that the ("l$ o& the overload is transmitted to the silo walls. -or ro"gh articles, the &orce chains will have a m"ch (roader directionality, as &riction and as erity interloc$ing allows ro"gh articles to transmit &orces in a variety o& directions witho"t &ail"re. )eca"se (ridging in the gran"lar material is less e&&ective, the im osed overload is not screened, and its e&&ects on the stress &ield can (e transmitted &"rther in to the gran"lar material. 8onse#"ently, the critical height decreases, (eca"se stresses at the arch are large and allow "lsations to occ"r &or a longer time d"ring discharge, in agreement with st"dies that sli –stic$ &riction is de endent on the local stress level A1CB. The 5anssen sol"tion &or the e&&ect o& an overload on the stress &ield in a gran"lar medi"m in a silo A1<B redicts that the e&&ect o& the overload on the stress &ield inside the gran"lar material decays &aster as the angle o& wall &riction is increased. 8r"shed glass has a larger angle o& wall &riction than the glass (eads, and so the &inding that an overload has more e&&ect on the cr"shed glass than the glass (eads does not agree with the redictions &rom the 5anssen sol"tion. 4edderman A1<B has s"ggested that the 5anssen sol"tion is not a good method &or redicting stress levels inside a gran"lar material when an overload is im osed. -"rther wor$ eEamining the shear and wall normal stresses in silos &or di&&erent sha es and distri("tions o& article si@es with varying overloads wo"ld hel in o(taining a ro riate constit"tive relations to descri(e gran"lar material stress &ields macrosco ically, an area which is the s"(Hect o& c"rrent de(ate A2C,2:B. +air et al. A29B &o"nd that article sha e in&l"ences gran"lar material sliding characteristics. This st"dy con&irms this &inding, (eca"se gran"lar materials made o& the same glass with similar si@es ("t di&&erent sha es had di&&erent "lsation &re#"encies in the al"min"m t"(e. 9n artic"lar, s"r&aces that are ro"gh are less li$ely to have stic$–sli

).*. +"ite et al. / Powder Technology 145 (2004) 190–202


&riction, (eca"se the as erities can loc$ and revent sli occ"rring at all contact oints. This arg"ment is consistent with o"r &inding that no "lsations occ"rred in the &"lly ro"gh lain steel t"(e. These res"lts are also in accord with &indings (y TeHchman A10B, +oriyama and 5im(o A4B, and 5ahagirdar A2B that to revent silo #"a$e, mass &low silos sho"ld have ro"gh walls. Jardow et al. A1B, whose st"dy did not eEamine a variety o& gran"lar materials, s"ggested that wall &riction was not the ca"se o& silo #"a$e, a &inding that this st"dy has shown is not always correct. 2s their st"dy was &or a core &low silo, they did not consider the ossi(ility that stic$–sli &riction can occ"r at sliding s"r&aces inside the gran"lar material, as shown in the st"dy (y 4as"no et al. A1CB, and not H"st (etween the gran"lar material and the silo wall. -inally, Jardow et al. A1B s"ggested that resonance is not always re#"ired &or silo #"a$e, while other st"dies have s"ggested that it is an im ortant &actor in am li&ying the am lit"de o& the "lsations A<,5,10,11B. ;"r st"dy does not give a concl"sive answer to this #"estion, as we have meas"red only the dominant nat"ral &re#"ency o& vertical oscillations o& the silo= care&"l meas"rements o& the nat"ral &re#"encies associated with radial vi(ration o& the &illed silo t"(e are re#"ired to ma$e a more de&initive statement, ("t these were not meas"red.

ac$nowledges &inancial s" ort rovided (y the 7ere$ Pidow senior thesis &"nd. This wor$ was s" orted (y the 3.!. 7e artment o& 1nergy 871'-82C'004T409:1.

A1B ). Jardow, 7. !ch"l@e, 5. !chwedes, 2n eE erimental analysis o& the dsilo #"a$ingT henomenon, Proc. ;& the <rd Iorld 8ongress on Particle Technology, )righton, 1ngland, 1996. A2B !. 5ahagirdar, 2n eE erimental st"dy o& so"nd emission d"ring gran"lar &low, 7e artment o& 8hemical 1ngineering, 9ndian 9nstit"te o& !cience, )angalore, 9ndia, 1999. A<B 5. *mita, !ilo as a system o& sel&'ind"ced vi(ration, 2!81 5. !tr"ct. 1ng. 111 (1965) 190. A4B 0. +oriyama, ?. 5im(o, 0ed"ction o& "lsating wall ress"re near the transition oint in a (in, )"l$ !olids Jandl. 6 (1966) 421. A5B +. 4iedostat$iewic@, 5. TeHchman, 1E erimental and theoretical st"dies on resonance dynamic e&&ects d"ring silo &low, Powder Jandl. Proc. 15 (1) (200<) <C. ACB 8.1.!. Philli s, 1lectrical and other ro erties o& sand, Proc. 0. 9nst. ?. ). 19 (1910) :42. A:B 5. TeHchman, ?. ?"deh"s, !ilo'm"sic and silo'#"a$e, eE eriments and a n"merical cosserat a roach, Powder Technol. :C (199<) 201. A6B 8.+. Iensrich, 1E erimental (ehavio"r o& #"a$ing in tall silos, Powder Technol. 12: (2002) 6:. A9B 8.+. Iensrich, 2nalytical and 4"merical +odeling o& ."a$ing in Tall !ilos, Ph7 thesis, 3niversity o& 4ewcastle, 2"stralia (2002). A10B 5. TeHchman, Technical conce t to revent the silo hon$ing, Powder Technol. 10C (1999) :. A11B 5. TeHchman, +. 4iedostat$iewic@, 0esonance e&&ects d"ring gran"lar &lows in silos, Proc. 9nternational 8ongress &or Particle Technology, 4"rem("rg, ?ermany, 2001. A12B ?. ;varle@, 1. 8lement, !low dynamics and aging in a con&inedL gran"lar &low, Phys. 0ev., 1 C6 (200<) 0<1<02. A1<B 0.+. 4edderman, !tatics and *inematics o& ?ran"lar +aterials, 8am(ridge 3niversity Press, 1992. A14B ?.I. )aEter, 0.P. )ehringer, T. -agert, ?.2. 5ohnson, Pattern &ormation in &lowing sand, Phys. 0ev. Pett. C2 (24) (1969) 2625. A15B +.?. Perry, +.-. Jandley, The dynamic arch in &ree &lowing gran"lar material discharging &rom a model ho er, Trans. 9nst. 8hem. 1ng. 45 (19C:) T<C:. A1CB !. 4as"no, 2. *"drolli, 2. )a$, 5.P. ?oll"(, Time'resolved st"dies o& stic$–sli &riction in sheared gran"lar layers, Phys. 0ev., 1 56 (2) (1996) 21C1. A1:B 5.8. ?eminard, I. Posert, 5.P. ?oll"(, -rictional mechanics o& wetL gran"lar material, Phys. 0ev., 1 59 (5) (1999) 5661. A16B -.P. )owden, 7. Ta(or, The -riction and P"(rication o& !olids, ;E&ord 3niversity Press, 1950. A19B 5. *rim, -riction at macrosco ic and microsco ic length scales, 2m. 5. Phys. :0 (9) (2002) 690. A20B 2.7. )erman, I.2. 7"c$er, 5.4. 9sraelachivili, 1E erimental and theoretical investigation o& stic$–sli &riction mechanisms, in% ).4.5. Persson, 1. Tosatti (1ds.), Physics o& !liding -riction, *"lwer 2cademic P"(lishers, 199C. A21B +.4. 4aeem, 8.). !harma, Prediction o& nat"ral &re#"encies &or thin circ"lar cylindrical shells, Proc. 9nst. +ech. 1ng. 8214 (2000) 1<1<. A22B !.-. ."inn, !ilo +"sic, )! thesis, Princeton 3niversity, Princeton, 4ew 5ersey, 3!2 (2002). A2<B ).*. +"ite, The e&&ects o& (o"ndaries on gran"lar and &l"id mechanics, +!1 thesis, Princeton 3niversity, Princeton, 4ew 5ersey, 3!2 (200<). A24B 5.I.!. 0ayleigh, The Theory o& !o"nd, vol. 99, 7over P"(lications, 1945.

C. 8oncl"sion This st"dy has shown that stic$–sli motion generates silo m"sic and silo #"a$e. !ilo m"sic is driven (y the stic$– sli "lsating motion o& the gran"lar material d"ring discharge and is associated with a resonance in the air col"mn a(ove the (ed. Ihen the "lsating motion disa ears, so does the silo m"sic. ;ver the range o& discharge rates st"died here (e#"ivalent to average velocities o& descent thro"gh the t"(e o& 1–6 cm/s), the "lsation &re#"ency was inde endent o& discharge velocity. )oth silo m"sic and &low "lsations sto ed a(r" tly when the (ed height &ell (elow a critical val"e. The critical height co"ld (e changed (y lacing an overload in the case o& cr"shed glass, ("t not in the case o& the smooth glass (eads. This may (e rationali@ed, altho"gh only s ec"latively at this oint, (y di&&erences in stress chain (ehavior.

2c$nowledgements Ie are grate&"l to Pro&essors 5. TeHchman and 2. !mits &or many hel &"l disc"ssions. The a"thors also wish to than$ 8. Iensrich &or roviding a co y o& his thesis in advance o& "(lication. **0 is very grate&"l to +r. P.T. 0agh"ram &or directing his attention to this ro(lem (y demonstrating the silo m"sic eE eriment in 199C. )*+ was s" orted (y a Princeton 3niversity -rancis 3 ton grad"ate &ellowshi and (y the 2&rican 9nstit"te &or +athematical !ciences. !-.


).*. +"ite et al. / Powder Technology 145 (2004) 190–202 A29B *. +air, *.+. -rye, 8. +arone, 9n&l"ence o& grain characteristics on the &riction o& gran"lar shear @ones, 5. ?eo hys. 0es. 10: ()10) (2002) 2219. A<0B !. )o"isso", 5.P. Petit, +. )ar#"ins, 4ormal load, sli rate and ro"ghness in&l"ence on the olymethylmethacrylate dynamics o& sliding% 1. !ta(le sliding to stic$–sli transition, Iear 214 (1996) 15C.

A25B P. )erge, R. Pomea", 8. >idal, ;rder Iithin 8haos, Iiley, 1964.L A2CB P. >anel, Ph. 8la"din, 5.'Ph. )o"cha"d, +.1. 8ates, 1. 8lement, 5.P.L Iittmer, !tresses in silos% com arison (etween theoretical models and new eE eriments, Phys. 0ev. Pett. 64 (:) (2000) 14<9. A2:B ?. ;varle@, 8. -ond, 1. 8lement, 2 giant overshoot e&&ect in theL 5anssen gran"lar col"mn, Phys. 0ev., 1 C: (200<) C0<02. A26B +.1. 8ates, 5.P. Iittmer, 5.P. )o"cha"d, P. 8la"din, 5amming, &orce chains, and &ragile matter, Phys. 0ev. Pett. 61 (9) (1996) 1641.