Professional Documents
Culture Documents
Butterfield (1969)
Butterfield (1969)
A THEORETICAL SINGLE-SIZE
STUDY OF THE PRESSURES DEVELOPED IN A SILO CONTAINING PARTICLES IN A REGULAR PACKING R. BUTTERFIELD
INTRODUCTION
RESULTSof measurements of the pressures exerted on the walls and bases of model [1, 2] and full-size [3-5] silos containing granular materials demonstrate conclusively that: (i) The pressures exerted on the walls and base of a silo containing a given depth of a specific fill material may have a range of values depending upon whether the silo falling has been very recently placed, standing in place for some time or is being removed from the lower part of the silo and that the maximum wall pressures realized may be more than double the minimum pressure [2, 5] at the same point. (ii) No theory has yet been offered which either explains this behaviour adequately or provides the engineer with a reasonable indication of the possible range of these pressure fluctuations in any silo. Theoretical contributions towards the calculations of the pressure distributions have been made by JANSSEN[6], AIm~Y [7], REIMBERT[5], JAKY [8], CAQUOT[9], all of whom arrive at formulae which give single-valued pressure distributions for a given depth of fill and no indication of the amount by which the actual pressures may possibly deviate from the predicted value. The formula most commonly used is that due to JA~SSEN [6] p -- - y R (1 -- e_KHtanO, /R) tan ~'
227
228 where p ~, R
R. BUTTERFIELD
= pressure on the silo wall at a depth ---=H below the surface of the fill ----- bulk density of the fill material = internal area of the silo internal perimeter of the silo
tan~' = coefficient of friction mobilized between the silo walls and the fill material K ---- ratio of horizontal to vertical pressure exerted by the fill.
However, the designer is left with the difficult problem of choosing a suitable K-value and there is little doubt that the values of K, of the order of0.33-0-5, which are used by analogy with the active and at rest earth pressure coefficients, represent more nearly a lower bound solution for the wall pressures rather than a suitable design value. A study of the all too numerous failures of silos [3] confirms this statement and this same reference summarizes the Russian and proposed German design regulations for cement silos which specify that the Janssen value for ~' = 30 should be multiplied by a factor of about 2.0. This curve then provides an envelope of all measured maximum wall pressures. This report is an attempt to indicate, by analysis of an idealized model, how such a spread of K-values may arise and tentative values of the upper and lower bound pressure distribution envelopes are developed. An 'arching factor' is then introduced in order to analyse intermediate cases of the pressure distributions and these solutions are compared with the available published experimental data on silo pressures.
THE T W O - D I M E N S I O N A L M O D E L
Earlier applications of models of regular packings of disks and spheres to problems related to the mechanics of granular materials have been made by JENKIN [10] to retaining walls, TROLLOPE[11, 12] to rockfill dams and ROWE [13] to shear strength theory. In the following, the interparticle force distributions in the upper and lower bound solutions and the arching factor concept are due to TROLLOPE[11, 12]. The silo (of width ----- D) is considered to be filled with very small uniform disks in a regular packing as in Fig. l(a), in this case the 'packing angle' a = 60 . Figure l(b) shows an assembly in which a is not 60 . All formulae will be developed in terms of a general a-value.
S no. in B unils
uts(
/ (o)
229
Throughout the development of the theory the 'smooth' disks used have the special property of NOT developing frictional forces between disks but frictional forces a r e developed at the silo walls. If the assembly of disks is used to model a fill material of bulk density 7 then, from Fig. l(b), for a model area 'p' disks by 's' disks, each of weight ---- w, which represents a prototype silo area of height H and breadth B, all of unit thickness, then for any a-value
p.s.w = 7.B.H.
For any small modelled prototype area both p and s will be numerically very large. Within this restriction of the disks being infinitely small when compared with the size of an elemental silo area, either the disk diameter or the disk weight can be specified arbitrarily. Let the disk diameter = d always be such that s = B, then
pw = 71t
(I) (2)
In general there will be six forces acting on any smooth disk in the assembly as shown in Fig. 2(a), of which three will be unknown and the problem is therefore 'indeterminatek However, the number of unknown reactions can be reduced to two and the horizontal forces acting in the model minimized by considering the case shown in Fig. 2(b), where contact between the disks has been assumed to be broken in the horizontal rows [11]. A further extreme case can be produced as shown in Fig. 2(c) where again the number of unknown reactions is reduced to two by assuming contact to be broken at the lower right-hand disk in each cell, this configuration maximizes the horizontal stresses. The case shown in Fig. 2(b) is described by TaOLLOPE [11] as the 'no arching' condition and Fig. 2(c) as the 'full arching' condition to the left, a mirror image of this condition exists and can be described as 'full arching' to the right. The stress distributions throughout the silo, due to the fall self-weight, can be simply determined for these three cases if the coefficient of friction mobilized at the wall is given a specific value.
nq
~Nq~
mq
Iq
(a)
(b)
(c)
(d)
230
R. BUTTERFIELD
Throughout the analyses of the extreme and intermediate cases q, mq and nq will be taken as the known reactions and Lq, Mq and Nq as the unknown reactions. Also, the particle weight w will be replaced by two equivalent forces q [Fig. 2(d)] which act in the a-direction where [11] 2q sina = w therefore from (1) ~,H
P'q -- 2sina"
(3a)
(3b)
The interparticle forces can be transformed into equivalent stresses by the following equations (4) which are readily obtained for any a-value from Figs. 3(a--c). From Figs. 3(c) and l(b)
Tx~=
and from (2)
(l - - m ) q .
cosa
(L -- M)q.cos
2d. cosa
2d. cosa
Txy = (l -- m ) q . c o s = ( L -- M ) q . c o s a .
F r o m Fig. 3(b) oy = (l + m)q. sina + w/2 = ( L -4- M ) q . sin -- w/2 2d.cosa 2d. cosa
Q
. Nq
i::"
~=o
M~I m ~ ~ Mq Lq /~ 0,,/2 t
(d)
mqq
(a)
T.
(c)
231
In general, when finally calculating stresses from b, m etc. the w/2 terms will be negligible, and from (2) % = (l -F m)q.sin~ = (L q- M)q.sin~. From Fig. 3(a)
Orx
(4b)
{(l -1- M)cosa q- 2N)q = {(L q- m)cos~ q- 2n}q 2d. sin~ 2d. sina
(4e)
The method of analysis consists of a progressive summation of the interparticle forces starting from the top horizontal fill surface layer of particles where in all cases I = m = n ---0. In the summation, forces arising at the silo walls are dealt with by the introduction of a 'force transmission factor' = t, Fig. 3(d), where
M ' = t.l.
Resolution of the forces in Fig. 3(d) gives the value of t, when N = 0, as t -= (tan~ -- tanff') (tana q- tan~')
(5a)
(5b)
where if' is the wall friction angle as defined earlier for Janssen's equation and therefore t is 'known' in a given problem, e.g. for if' = 30 and a = 60 (5b) gives t = . In the particular case where N = 0 equations (5a) and (4c) give the silo wall pressure as
~x~w,., =
(1 q- t).cos~.q tana
(4d)
L=lq-1 M=m-F1
N----n----0 The summation of interparticle forces is carried out in the Appendix for the calculation of wall pressures at any point X [Fig. 4(a)]. In order to allow for the effect of the 'transmission factor' t when the depth of fill exceeds B Y = D.tan, a 'characteristic depth' of fill H' is introduced where 2H'= D.tana. (7) (6)
232
R. BUTTERFIELD
(Crx)n_ yD
~(1 + t ) ( 1 2tanct L (1 -- t)
tw);, t <
J
(8b)
where the depth of fill H = W . 2 H ' = W. D tana, W being a positive whole number; from 5(b)
(ax) n= ~,O -(1--tw),t<l. (8c)
2tan~'
_/
~,ooo _ ::~o,
1~,,'2, ~%
-1. A;:
t ,/
j'q
0'2 0-4 0"6 q~/)'D 0"8
(b)
A non-dimensional form of this equation [Fig. 4(b)] plots as a series of straight lines which change slope at vertical intervals of 2H'. The dotted curve in Fig. 4(c) is Janssen's equation with K = 1/3 and these practically identical curves represent the lower bound solution for the model silo wall pressures when a = 60 and q~' = 30 . K = 1/3 corresponds to the active earth pressure coefficient for a fill material with apparent friction angle (fie) of 30 . Where ~c = the Coulomb 'apparent' friction angle for the fill material. For large depths of fill W = oo and equation 8(c) reduces to
t<l
~,D 2tan~'
(9)
233
L=l+m+2
M = 0 N = n + 2cosa(m + 1). An analysis of the wall pressures developed is given in the Appendix which leads to the following equation, valid for H ~< H'. (ax)u= (10)
yH
tan2a
(1 + t ) (1 -- t)"
(11)
For H > H' there is no, further increase in ax (see Appendix) and (ox)w=
yD .(1 + t ) _
2tana (1 -- t)
yD 2tanff'
(12)
irrespective of the fill depth, i.e. identical with equation (9). These equations are plotted non-dimensionally, Fig. 5(b), and represent the upper bound solution for the wall pressures at any point .t". The lower bound solution is also shown in Fig. 5(b) and the region between the curves is the zone in which the actual wall pressures
~4 -1-
k,
0'4 ~D
0"6
0"8
(b)
Fio. 5. The 'full arching' solution and the zone of possible wall pressure distributions.
Roc 6/2--H
234
R. BUTTERFIELD
developed must lie. Again, for the special case of ~ = 30 , the dotted curve in Fig. 5(b) is Janssen's equation for K = 3, where
K= Kp--
1 + sin6 I sins
-
In practice the likely pressure distributions will lie between the 'no arching' and 'full arching' conditions and these can be investigated by introducing an 'arching factor' = A. This concept was first suggested by TROLLOPE [11] although the following definition of A differs slightly from his 'arching factor'. If we define A such that M = ( 1 - A____) (m -4- 1) L (1 + A) (l + 1) (13)
where -- 1 <~ A ~< + 1, then the general case of all particles making contact [Fig. 2(a)] can be solved by the methods used above for given values of A. This empirical factor A takes account of the fact that the state of particles in the silo will not normally correspond to either the 'full arching' or the 'no arching' condition. Furthermore since the situation will be additionally complicated by the presence of interparticle frictional forces it seems unlikely that any particular A value curves will always apply for any one fill material in a given silo and a considerable scatter of experimental results would be expected. However, when correlated with sufficient experimental evidence, the 'arching factor' presents a convenient way of reducing the range over which the maximum and minimum pressures occurring might be expected to extend. The following general expressions for L, M and N are obtained by resolution of forces in Fig. 2(a)--see Appendix. L= M= (1 + A ) ( l + l ) ( l + m + 2 ) {(1 + A) (I + I) + ( 1 - - A) (m + I)} (1--A)(m+l)(l+m+2) {(1 + A) (l + l) + ( 1 - - Z) (m + l)} {(1 + A) (1 + 1) -- (1 -- A) (m -4- 1)}] (m -- t) + (l + m + 2) {(1 + A) (I + I) + ( 1 - - A) (m + O}J (14)
N=n+cosa
= n + cosa {(m -- l) -- ( M -- L)}. For A = 0 these reduce to the no arching case, equations (6). For A = + 1 these reduce to the full arching to the left case, equations (10). For A = - - I these reduce to the full arching to the right case, which is the mirror image of A = +1. Values of A in the range 0 to + 1 are considered in the left-hand half of a silo and symmetry of behaviour about the silo centreline is assumed. [This symmetry could conceivably break down with asymmetrical silo discharge arrangements when, as an extreme case, arching across the whole silo width could occur. Equation (11) would then apply to a depth of H = 2H'.]
235
For values of A between these limits the algebra involved in the summation process becomes exceedingly clumsy and therefore a body of particles, 96 wide across the silo half breadth, was modelled in a digital computer and a program developed to carry out the summation processes on the computer. Figures 6 and 7 show ,~,/(~,D) -- HID curves for a range of values of the parameters a, t and ,4, these results can be summarized as follows: 1. The peak *x/(~,D) value is a function of ' only, equations (9) and (12). 2. Variation of a has a twofold effect: (a) it causes a change in scale of the graphs along the HID axis, as shown at the HID = tana, 2 tana etc. points Co) from equation (Sb) variation of a affects t. Thus, ' = 30 and a = 60 gives t = 0.50, whereas for a = 63, t = 0.55. These effects tend to nullify each other and small variations of a have little influence on the *x/(~,D) values. 3. Variation of t causes a significant change in wall pressures at greater depths, but for a given a-value (e.g. 60 ) a large change (from 36 to 17) in $' is needed to produce large (from 0.4 to 0-7) changes in t.
ax/~D
0 0"2 04 0"6 Q:B I0
236
R. BUTTERFIELD
o~/~,D
0 0.4 0.4 I,~
I
I0
_
o
a -- 6 o * +--9.Z~.17 ~, i-~n~
Kp = I/K, /
the variation of
case.
4. The graphs show that the o:,/(TD) values are extremely sensitive to the value of A, that the 10 per cent arching case is hardly distinguishable from the 100 per cent arching case, and that systematic arching throughout the silo of only 1-2 per cent has a very appreciable effect on the wall pressures. 4' and A are therefore seen to be the key parameters. VALUES OF THE PARAMETERS q~', A AND a Before comparing the theoretical curves with published results of measured pressures in silo walls, it may be of some value to enumerate the likely range of values of, and the factors affecting, the parameters if', A and a. There are numerous published values [1-5] of the coefficients of sliding friction between fall materials and silo walls. For most common fill materials they appear to range between extremes of 17 and 30 (i.e. for a = 60 a t-range of 0 . 7 0.5. It is not known what proportion of the full s6' value is, in fact, mobilized at the wall but since the general tendency is for the fill to settle inside the silo, it is suggested that a large proportion of the full 4' value could be mobilized. In a heterogeneous fill material there is no exact counterpart of a. However, it would seem possible that there may exist within the fill at rest an equivalent particle structural skeleton in which case the equivalent a-value is likely to be of the order of(45 + i6c/2). Thus a = 60 is retained as a convenient and possibly realistic value in all the plotted curves. The value of A is variable and generally unknown, although from the silo designer's point of view it would seem that the upper limiting value of + 1 is worth consideration. Experiments by TROLLOPE[1 I, 12] on a model embankment show that (i) arching can indeed develop in accordance with a theory based on this same concept and (ii) the arching action increases with increasing basal distortion of a 'dishing' type.
237
It is of interest at this stage to investigate qualitatively the patterns of relative particle movement which will tend to develop the two extreme values of A. Figure 8(a) shows the type of relative particle movement which would tend to produce the 'no arching' case and Fig. 8(b) that which would produce 'full arching' to the left. Figure 8(c) shows how, in a silo with rigid walls and a flexible base, the correct relative particle movement occurs for 'full arching' to develop symmetrically about the silo centreline [i.e. the Fig. 8(c) condition will tend to maximize wall pressures and minimize base pressures]. This type of arching is relatively well known in other fields in connexion with the under-registration of diaphragm activated soil stress gauges [15, 16] and TERZAGm'S [17] famous 'trap door' experiments. A precisely similar pattern will tend to be produced when a silo is being emptied from the bottom, which has, initially, the same effect as allowing a large base deflexion of the silo to develop. The 'no arching' case, Fig. 8(d), represents minimum wall pressures and maximum base pressures; this is obviously a more stable condition than 8(c) and will represent the 'long term' at rest trend of the pressure distributions. The pattern of relative particle movement for this condition is encouraged as in Fig. 8(c) by relatively flexible silo walls and a rigid base. Figure 9 shows the slip lines developed in a two-dimensional model silo where this type of wall distortion has occurred.
(a]
/ /
I
(c)
\ \
(d)
FIo. 8. Relative particle movements tending to produce the 'full arching' and 'no arching' conditions.
238
R. BUTTERFIELD
cr:,/yD
0-~ 0'2 0"3 0"4 05 06
07
O'R
FIe. 10 (and Table 1). Comparison of theoretical curves with wall pressure measurements in silos [1].
TABLE 1. Co~m,~aSON OF rHr.OS~TICALCURVESwrrH WALLPRE~URE
r,m~tr~n~N'rs n~ o ~ Curve No. 1 2 3 4 5 5 6 7 8 9 10 Tests by Ketchum Jamieson Jamieson Jamieson Jamieson Jamieson Bovey Lufft Lufft Pleissner Pleissner
smos rxoM KerCHUM [1] Silo material Timber Timber Steel Timber Steel Steel Timber Concrete Concrete Concrete Timber Silo 'diameter' lft 12ft 6in. Ift lft lft lft 12ft 6in. 24ft 11ft 3in. 9ft 5ft
Wheat Sand
THE THREE-DIME~SIONAL SILO A similar analysis to the f o r e g o i n g suggests t h a t for a given d e p t h o f fill in a square silo the wall pressures r e a l i z e d w o u l d be exactly h a l f t h o s e d e v e l o p e d in the t w o - d i m e n s i o n a l m o d e l . I n the f o l l o w i n g c o m p a r i s o n s o f t h e o r e t i c a l a n d m e a s u r e d wall pressures a similar r e l a t i o n s h i p is a s s u m e d to exist for a circular silo, b y a n a l o g y with J a n s s e n ' s e q u a t i o n which gives i d e n t i c a l results for s q u a r e a n d circular silos o f similar R-value. Since the c a l c u l a t e d t w o - d i m e n s i o n a l base pressures are a l m o s t exactly u n i f o r m across the silo base, these w o u l d r e m a i n u n a l t e r e d in the square silo case, a n d the calculated results are o b t a i n e d f r o m e q u a t i o n s d e v e l o p e d for the t w o - d i m e n s i o n a l case.
239
~/~D
0.2
0"3
04
0"5
06
0.7
0.8
-w
A=O
0"01
0"02
0-04 'O
FIo. 11 (and Table 2). Comparison of theoretical curves with wall pressure measurements in concrete silos containing cement [3].
TABLE 2. COMPAmSON oF THEORETICAL CURVES WITH W A L L P R E S S t r ~ M E A S t m ~ m N ' r s IN CONCmmB
SILOSFROMC. & C.A. LIBRARYTRANSLATIONNO. 94 [3] Curve No. 1 2 8 9 10 13 14 Test Russia, October: filling Russia, October: emptying England, Rowe: filling England, Rowe: emptying Russia, Akmajansk: filling Russia, October: emptying Dutch, maximum loadings Silo diameter (m) I0" 0 10"0 9" 15 9" 15 8.1 10.0 9" 1
Four sets of published data are compared withthe curves obtained from the idealized model. 1. I~TCHUM (1929)[1] The numerous cases of measured wall pressures reported, mainly in silos containing gain, are unfortunately very suspect due to the crude pressure measuring devices used which, without exception, would tend to under-measure the pressures developed. Figure 10 shows the
240
R. BUTTERFIELD
zone in which 8 separate sets of quoted results fall for filling and emptying the silos. The theoretical curves are drawn for ~ = 60: if' = 17: t = 0.7 and A = 0, 0-01, 0.02, 0.04 and 1.0.
o~/TD
0
0"1
,j Ao,
\
~-- 6o*,~'- z3~*,,=o.6
Part elevation
(a)
(b)
Fio. 12. Detail of "Des Silos de Masni~res" and a comparison of theoretical curves with wall pressure measurements in a steel silo containing grain [5]. aeration was used during the discharge of the silos, which confuses the interpretation of the results and possibly justifies the rather low if' value used in the best fit theoretical curves.
241
2I ")
6"06 5,20 5 ~4 - 3 3
2"60
i
3.48 "- 2 . ~
r
= ""
6"0 4'8
/ 3 "6
--2-4
1.0
1.74 ~
~1.2
o.~--
0"8z
%
0
%
t1
2 Radius, in.
FIO. 13. Cmparisnfthcoretica1curvswithbasepressureresutsnamdestsicntainingsand[2] emptying of the silo and also theoretical curves for a = 60; ~' = 23; t = O. 6; ,4 = 0.01 and 1.0 which suggest that A approaches unity during emptying and a value < 0.01 during filling. 4. LENCZNER (1963) [2] As mentioned previously all the stresses throughout the idealized model can be calculated and in Fig. 13 the base pressures are plotted as full lines for a = 60 , t = 0.6, A = 0 and $ ' = 23 for various fall depths. The dotted lines are results published by LENCZN-ER[2] for a model silo containing sand. He quotes a ~' value of 25 for sand on steel. The theoretical curves in Figs. 10-13 were drawn on a 'best fit' basis and are not intended to justify the theory in detail but rather to suggest that the upper and lower bound solutions developed and the 'arching' concept m a y provide a rational guide to the mechanism behind the pressure fluctuations in silos, and a means of forecasting, semi-empirically, the likely range of these fluctuations.
Miffi n l 2
L) L~
"~ \
L
M3 -
" N ~ J
Lsl Fro. 14. Boundary forces for the full arching condition.
242
R. BUTTERFIELD
CONCLUSIONS
1. The study of the idealized model may have helped to explain the mechanism whereby the pressure distributions throughout a silo containing granular material can vary over a considerable range of values during filling and emptying. 2. In the absence of a rigorous solution to the problem, it is suggested that the concept of the A = 1 envelope may provide a more logical guide than at present exists to the value of the maximum short-term wall pressures which may occur in tall silos. Thus for any fill depth > H' above a given point the full 'arching' condition would give
O" X - -
~,D 4tanr'
as a possible maximum (,rx) value at that point. 3. The separation of 'silo-action' into a wall friction component and an arching contribution is thought to be an advance on previous analyses as also is the possibility of obtaining more detailed information on stress distributions throughout the silo. 4. Many more measurements on silos are needed before any theoretical predictions of pressure distributions can be used with confidence and it is hoped that the study may stimulate the publication of these.
REFERENCES
1. I~TCHUM M. S. The Design of Walls, Bins and Grain Elevators, McGraw-Hill, N.Y. (1929). 2. LE~CZNERD. An investigation into the behaviour of sand in a model silo. Struct. Engr 41, 389-398 (1963). 3. LEoha-L~atDrF., BOLL K. and SPmDELE. The safe design of cement silos. Beton Stahlbetonb. 55, (3) 1-27 (1960). [Cement and Concrete Association Library Translation No. 94.] 4. gow~ R. E. An investigation into the cause of cracking in a reinforced concrete silo containing cement. Mug. Concr. Res. II, (32) 65-74 (1959). 5. REIMBERTM. and REIMBERTA. Silos, traitd thdorique et pratique, pp. 57-62, Editions Eyrolles, Paris (1959). 6. JANSSENH. A. Versuch tiber Getreidedruck in Silozellen. Z. Vet. dr. Ins. 39, (35) 1045-1049 (1895). 7. AmEY W. Theory of silos. Proc. Instn cir. Engrs 131, (1) 34%358 (189%98). 8. JAKY J. Pressure in Silos, Proceedings of the Second International Conference on Soil Mechanics and Foundation Engineering, Vol. 1, pp. i03-107 (1948). 9. CAQUOTA. La pression duns les silos. Proceedings of the Fourth International Conference on Soil Mechanics and Foundation Engineering, Vol. 2, pp. 191-195 (1957). 10. JENgrN C. F. The pressure exerted by granular material, an application of the principles of dilatency. Proc. R. Soc. A131, 53-89 (1931). 11. TROLLOPED. H. The Systematic Arching Theory AppLied to the StabiLity Analysis of Embankments, Proceedings of the Fourth International Conference on Soil Mechanics and Foundation Engineering, Vol. 2, pp. 383-388 (1957). 12. TROLLOP1~D. H. The Stability of Wedges of Granular Materials, Ph.D. Thesis, University of Melbourne (1956). 13. Rowe P. W. The stress dilatency relation for static equilibrium of an assembly of particles in contact. Proc. R. Soc. A269, 500-527 (1962). 14. Rowe P. W. Stress dilatency, earth pressures and slopes. Proc. Am. Soc. cir. Engrs 89, (SM3) Pt. I, 3%61 (1963). 15. KALLESTEN1US and BERGAUG. Investigation of soil pressure measurements by means of cells. Proc. R. T. Swed. Geotech. Inst. No. 12 (1956). 16. TROLLOPED. H. and LEE T. K. Measurement of Soil Pressures, Proceedings of the Fifth International Conference on Soil Mechanics and Foundation Engineering, Vol. 2, pp. 493--499 (1961). 17. TEgZAGm K. Stress Distribution in Dry and in Saturated Sand above a Yielding Trap Door, Proceedings of the International Conference on Soil Mechanics, Vol. 1, pp. 307-311, Cambridge, Mass. (1936). 18. Kvnpm R. Gravity flow of material in hoppers and bins--I. Int. J. Rock Mech. Min. Sei. 2, 25--41 (1965). 19. KVAPm R. Gravity flow of material in hoppers and bins--II. Int. J. Rock Mech. Min. Sci. 2, 277-304 (1965).
243
APPENDIX
Deduction of the Wall and Base Pressure Equations
The method used throughout these calculations is to carry out a summation process of the interparticle forces layer by layer starting from an upper horizontal surface layer where l, m and n will be known; usually zero. Then by using the relationships between l, m, n and L, M, N [e.g. equations (6), (10) or (14)] and equation 5(a), L, M a n d N c a n be calculated AT ANY POINTwithin the body of particles. These values of L, M and N when inserted into equations (4) provide the equivalent stresses acting at the particular point. 1. The 'No Arching' Case. A = O, by definition Case (a): Depth of fill above any point X is H, and H <<.2H' : Fig. 4(a) Equations (6) and (7) are repeated below for convenience
L=I+I
M=m+l
N=n=O
(6)
2 H ' = D.tana
(7)
At the rifling surface, Fig. 4(a), identified by '0', lo = 0 and ff Xis located at depth H, [i.e. p particle layers, below the surface, Fig. 1Co)I, then from equations (6)
Lu = p
MR = M ~ = t .p
} (6a)
N,,=0 (i) Wall pressures at X. From (4d): (oDR = and from (1) and (3) and (6a) above (1 + t).cosa. L s . q tana
i,e. o~ is a linear function of H for 0 ~< H ~< H'having, at H = 2H', the value
(o~)2u,
(8a)
e.g. a = 600 : t = gives (~x)2n' = ~ (Y. 2H') and t = 1 gives (~)2H' = 1/3(7.2H'). (ii) Base pressures at X. An argument exactly as above leads, via (4b): to
(O'y)H =
(1 + t) . Z ~ .
Case (b): General depth o f fill H is a multiple of 2H" (i) Wall pressures
i.e. H = W(2H') say, where W is integer .'. H = W.D.tana. Now 2. 7 H L H . q = ~----~' ( 1 + t + t 2 - ~ - ta " .. tw_l }
244
R. BUTTERFIELD
to the layer immediately below at every point such as B, Fig. 4(a), e.g. when IV --- 3, say. Fig. 4(e) 2~,H' (1 -- t w)
L~.q= sin~ " 1~ - " 7 ' t < 1
yD
{(l +t)(1--tw))
(1 - - t) --
),D(1--t w)
~-a~; " (8b)
(o:,)n
For W = 1 : (8b) gives (8a) as a special case. Also, for a n infinite depth of fill, W = oo, we have ~,D (1 + t) ?D (o~)~o -- 2tana (1 -- t) - - 2tan~" t < 1. (9)
Equation (8b) is plotted non-dimensionally in Fig. 4(b) and has a series of linear stages between IV = 1, 2, 3, etc. (ii) Base pressures. These can be calculated by reasoning exactly similar to (i) above; as a n example consider the centreline base pressure, Fig. 4(c), then where now
(H -- H')
2H'
The summation can be carried out as in (i): and leads to (~ry)n = ~,.D.tana
1 +_t=2tw'~ t
1 -- t j,
< 1.
Point Q
Point R
Substitution of these values of L, M and N in equations (4) leads to % u, and T~,. The solution for point R, whore R is fixed by dimensions measured from the surface, 0, is valid whether R happens to be o n the silo base X Y or not.
L=-l+m+2
M=0 N = n + 2cosa(m + 1). (10)
245
Case (a): H < H ' O) Wallpressures. As before, the wall pressures at any depth H can be calculated by a summation process
starting from the surface layer of particles which are unloaded .'. to = mo = no ----0 (Fig. 14). In the complete full arching case, by definition, M = 0, .'. M ' = 0, which means that the 'transmitted' forces at the wail cannot be equilibrated. This can be overcome by assuming that, in fact, M ' forces are transmitted only through the outer column of particles, shaded in Fig. 14. Thus the full arching condition will then exist in all except these outer particles. This represents the maximum possible arching condition and will tend to be very unstable for small 4' values. Reference to Figs. 14 and 2(c) will show how the 'L' values build up layer by layer in the following pattern. Denoting L for layers 1, 2, 3, etc. by LI, L2, L3 etc. and also noting that other than in the boundary column M = m = 0, we have LI -----1 and ml = 0. Throughout the analysis the unit increments arise where halfparticles occur along the boundaries (Fig. 14).
,'. MPl .', L 2
=
t.Lx
m2
= 12 + m2 + 2 where 12 = 1.2
.'. L= and L3
.'. L$
=1.t+2.2=13
= m 3 + 13 + 1 and m3 = 0 = 1.t + (2.2 + 1)and M'3 = t.L3 ----m4 = m, + 1, + 2 where !~ -----3.2 = 1.t 2 + (2.2 + 1)t + 4.2. = 1.t 2 + (2.2 + 1)t + (4.2 + 1)
=
L,
.', L 4
Similarly
L5 L7
L , = t P - - ~ 21 + ( 2 . 2 + 1 ) t ~ - ~ + ( 4 . 2 + 1 ) t ~ -~ + . . .
... +... which can be written L, -----{2(p - - p ) + 1} . t ~ --~ + {2(p - - p -- 2) + 1} . t ~ --~ + . . . {2(p-- 3) + 1}.t + { 2 ( p - - 1) + I}*
...+...
But t < 1 and since the particles are infinitely small compared to any finite depth H,p will be a very large number, i.e. p ~ oo '" (1----2) 2 1 +t+t a+t 3+...t +2t+2t 2+6t s+...
"te
etc rep
.t t to the power
etc
246
R. BUTTERFIELD
~2p + 1 ~ . . 2 =\l_t]l_t--2.2t
2p (1 -- t) 2p
Lp (1 - - t )
1 + 2t + 3t2 + 4a + . . . - - ~ 1 4t
(1 -- t) 2
p -- l.tP_~2 3)
(I - - t )
(1 + 3t)
(1 - - 0 2.
Furthermore, at the wall boundary n = 0 throughout, therefore from 4(d), (1) and (3) (crx)n = (l+t).cos~ ( , H tans " sina(-1 -- t) ~,H(l+3t) ) 2(1 -- t)2.p.sina (15)
as p ~c~ the second term -->0. Also from Fig. 5(a), if H > H ' the build up of 12, 14, Is, etc. used above no longer holds and in fact, from the symmetry of this figure when H > H ' n o further increases in I will occur, for H~< H ' (~)sis a linear function of H, thus ~H . (1 + t) (ax)n = tan2 ~ (1 -- t) e.g. a = 60 , t = gives (~x)H = 7H, and i f H = H ' = D]2. tan~ yD (1 + t) (12) (11)
which is identical with equation (9), i.e. in the 'full arching' case the wall pressure at fill depth = H" is equal to
the no arching wall presgure arising at very great fill depths. (ii) Base pressures. The full arching condition will minimize the base pressures. Consider a fall depth of H',
Fig. 5(a). In the body of the fall M = m = 0 therefore equation 10(a) reduces to L=l+2 and also from the symmetry of the force pattern 1 = L = M = m = 0 along the silo emtrdine. Thus at any point Y 2~,Hs ~,.Hs
~.H"
sma
Hence ~,, from 4(b). Case Co): H > H ' A study of the foregoing will show that the term Lt is multiplied by a factor t p which approaches zero as p-~ao, since t < 1 and therefore any additional depth of fill, increasing Lt, will cause no ~ in (~'~)B .'. for H > H " (~x)~ is constant at the equation (12) value. The above solution is plotted in Fig. 5(b). Similarly the base pressures will remain unchanged for H > H'. We now return briefly to equation (15) and calculate the errors involved in neglecting the second term i f p is not very large. e.g. if t ---- from (15)
(~)~ - - 2-i-G-r~ 2
247
Then i f p = 12, neglecting the second term above introduces a n error of approximately 21 per cent in the (o.)x value. Similarly if p = 50 p = 200 p = 600 error - 5 % error - 1% error - O"4 %
Since the following more complicated 'partial arching' solutions are carried out by modelling particles in a digital computer, it is of interest to note that a body of particles 200 no. deep by 100 no. wide between the wall and centreline of the silo is necessary to give a solution within about 1 per cent of the infinite number of particles solution. 3. The General Intermediate Arching Case Resolution of forces in Fig. 2(a) leads to
(L + M ) = (I + m + 2)
and (l + M ) cos a + N = (L + m) cosa + n.
The definition of A equation (13) is that L _ (l--A) (re+l) +A)" (l+l) "
Solving the three equations above for the unknowns L, M a n d N leads to equations (14) quoted earlier. These equations are so cumbersome that a solution in closed form as in Sections 1 and 2 above seems to be out of the question and as mentioned previously a step-by-step analysis was therefore performed in a digitalcomputer. 192 stores in the computer were used to represent l and m or L and M for one layer of 96 particles (s = 96), stretching from the silo wall to the silo centreline. Starting from the surface condition lo = me ----no -----0, for given A, t and a values, Lt and 3/1 were calculated for each particle in turn via equations (14). Then L2 and M2 were calculated similarly from Lt and M I until a depth H ' was reached in the silo. The number of layers involved = p is seen to be 192 = 2s as follows. The number of particles in 1)/2 is s, and by definition s = D[2; also tan a -----H'/(D/2) and from Fig. 1(b) p.d sin a = H ' . These three equations, together with (2) (viz. 2//. cos a ---- 1) solve, when s and p are large, to givep ----2s. By applying the L and M values for this 192nd layer as 1o and me to the surface layer again, the calculations were carried through for depths from H ' to 6 / / ' . This process was repeated for the following range of parameters with a ---- 60 t = 0.5, 0"6, 0.7 A = 0, 0"02, 0"04, 0"06, 0'08, 0"10. The results are discussed in the previous sections of the paper.