You are on page 1of 9

Powder Technology, 78 (1994) 5-13 5

Agglomerate strength measurement using a uniaxial confined


compression test

M.J. Adams
Unilever Research, Port Sunlight Laboratory, Quarry Road East, Bebington, Wiral, Memyside L63 3JW (UK)

M.A. Mullier” and J.P.K. Seville**


Department of Chemical and Process Engineering, Universi~ of Surrey, Guildford, GlJ2 5XH (UK)

(Received July 27, 1991; in revised form October 12, 1993)

Abstract

In order to measure the strength of an agglomerated product it is convenient to compress a bed of the agglomerates
using a piston in a rigid cylinder; this is known as a confined uniaxial compression test. A simple first-order
lumped-parameter analysis of this compression process is presented, treating the system as purely dissipative
and applying the Mohr-Coulomb macroscopic failure criterion. This enables average single agglomerate strengths
to be deduced from the initial deformation behaviour of the bed under comparatively low loads. Agglomerates
with a range of strengths have been formed from quartz sand using varying amounts of a polyvinylpyrrolidone
binder. These have been tested both in the uniaxial compression test and individually, by compression between
parallel platens. The results show agreement with the proposed theory over an order of magnitude in single
particle crushing load. The load-deformation curve for the compression experiments is also consistent with
Kawakita’s equation and it is shown that, over a certain range of strain, the load-deformation equation developed
here and that due to Kawakita take approximately the same form; the agreement enables a physical interpretation
to be made of one of the parameters in Kawakita’s equation.

Introduction parameter from the the whole under


compression. This most easily achieved experimentally
and other particle formation pro- using a piston in a cylinder (Fig. in which case the
cesses it is often necessary to monitor strength of becomes of uniaxial confined compression.
product rapidly reproducibly. This difficult step then to relate observed macroscopic
han- behaviour to single particle properties in a
dling is to be avoided. conventional method is quantitative way. full solution to this problem would
so-called ‘Brazilian test’, in which single particles require the model relating
crushed between platens load required to the the distribution
cause fracture is recorded. However, in batch of of single particle failure stresses within bed.
particles formed under nominally identical conditions this problem is so complex that can only
there is always a wide variation in fracture loads (see, for
measured in this way. Consequently, many particles and Barnes [l]). Here ap-
must be tested a reliable average to be obtained. proximate first-order lumped-parameter approach which
A further complication is that the
loads are small (typically equivalem to a
grams’weight) so that the with which they
can may be high.
A simple alternative method consists of replacing
single particle with bed of similar particles
inferring some average single particle strength
cross-
tion
*Present address: Gillette Management 452 Basingstoke A0
Road, Reading, Berks OQE, UK.
whom correspondence should be addressed. Fig. Confined uniaxial test -

0032-5910/94/$07.00 0 Elsevier Sequoia. rights reserved


0032-5910(93)02777-8
6

nevertheless enables average single agglomerate failure Eventually, as the applied pressure is increased, the
stresses to be deduced from simple uniaxial compression single particles lose their separate identities and bulk
experiments, which has obvious practical advantages. deformation occurs. Much work on compaction has
Heckel[2] adopted the above approach and was able concentrated on this final stage of the process, because
to describe the pressure-volume relationship of metal this is of most importance in tableting. The work
powders in terms of two lumped parameters. He sub- presented here focuses upon the initial deformation
sequently demonstrated that one of these parameters under low loads, because it is this region of the load/
was directly related to the yield stress of the particles porosity relationship which yields most information
[3]. The application of this work to a wide range of about single particle properties.
powders has been undertaken by many workers with The approach taken here is a macroscopic one.
perhaps the most comprehensive study being carried Therefore, it should be stated at the outset that it is
out by Roberts and Rowe [4]. They greatly extended not possible to deduce anything in detail about the
the value of the technique by introducing the effects strength of the particles in terms of a fracture mechanics
of particle ductility, particle size and deformation. How- analysis. The authors are engaged in a parallel study
ever, they and others have observed that there are of the fracture mechanics of individual agglomerates;
significant deviations from the Heckel relationship for the details of this work are described elsewhere [lO-121.
brittle materials. While this does limit the value of the It is argued below that the breakdown of agglomerates
technique for agglomerates with solid binders, a more due to the compressive stress fields experienced in
serious concern is that the failure stress for a brittle uniaxial compression is most appropriately prescribed
porous material in compression is a function of the by conditions on a failure surface which may be defined,
principal stresses acting on its periphery. For example, for example, by eqn. (1).
the failure surface for a sandstone rock has recently
been proposed to be of the following form [5]: Theory

The parallel friction-element model


Consider a bed of particles confined within a rigid
where J2 is the second invariant of the deviatoric stress die (Fig. 1). The compressive energy applied to the
tensor and 1, is the first invariant of the stress tensor. bed can be accommodated by elastic or plastic defor-
The Heckel analysis ignores the frictional constraints mation, interparticle frictional sliding, and plastic or
at the container wall which attenuate the transmission brittle failure. The cyclic pressure-strain curves obtained
of stress through a bed. This is valid if the bed height from the particles studied in the current work show
is relatively small compared with the bed diameter. If that the elastic strain (and hence the stored elastic
this were not the case, a three-dimensional approach energy) is negligible. Thus the system can be regarded
to the problem would be required (see, for example, as entirely dissipative. Here, it is useful to lump together
Isherwood [6]). Kawakita and Ludde [7] list 15 other interparticle sliding and individual particle fracture
equations which have been used to express bed void which are, from the point of view of the macroscopic
fraction as a function of applied load without allowance assembly, both internal friction processes. The particles
for frictional effects. Most of the equations are essen- used in the current work were themselves agglomerates
tially empirical although it is claimed that some of the of smaller sub-particles and were coarse (greater than
parameters that arise within these constitutive rela- 1 mm) and non-cohesive. Thus, their tapped bulk density
tionships are related to single particle properties. was virtually independent of applied pressure until the
Seelig and WulfT [8] have argued that compaction threshold pressure for particle failure was reached. In
is the result of several distinct processes: initial com- other words, since little spatial rearrangement is possible
paction occurs due to particle rearrangement, followed under applied load without particle breakdown, almost
by elastic and plastic deformation and fracture of single all the internal friction can be ascribed to particle
particles. Cooper and Eaton [9] considered that frag- failure rather than interparticle sliding. The problem
mentation was unimportant and based their model on is to model the extent to which the incremental failure
a two-stage mechanism consisting of large void filling work increases as the strain proceeds.
by rearrangement and small void filling by distortion. Two-dimensional computer models of granular ma-
The result of their analysis is a deconvolution of the terials under load e.g. [l, 131 and experiments with
experimental pressure-volume data into two curves photoelastic particles [ 141 have consistently shown con-
which characterise the two mechanisms. It is clear that centrated contact force lines propagated through the
compaction does involve several mechanisms and that material in the general direction of the major principal
the relative importance of each will depend on the stress (Fig. 2). Many of the particles in the structure
properties of the particles under examination. are in fact under little or no load and are therefore
Fig. 2. Force transmission lines in a compressed bed of particles
[l] (the width of the lines indicates the relative magnitude of
the local stress; principal stress vertical).

Fig. 3. Parallel column compression model.


effectively redundant, as elegantly demonstrated by
Troadec et al. [15] for two-dimensional packings of
cylinders in which redundant elements were removed The single a&plomerate failure criterion
from the structure without affecting its macroscopic In order to develop a constitutive relationship it is
stress-strain behaviour. The propagation of stress in necessary to define a failure criterion for the agglom-
concentrated lines has led Dunstan et aZ. [16] to propose erates, since failure represents the primary event in
a new theory for slow dilatant plastic deformation of the compaction process. In the field of particle tech-
granular materials. They model their material as ‘a nology, the Rumpf [18, 191 model has traditionally been
changing collection of spatially dispersed crooked col- used to describe the strength of agglomerates. According
umns which are supported by distributed lateral forces to this approach the mean tensile strength per unit
arising from the minor principal stress and which trans- section area, a,, is given by
mit all the major principal stress through the material’.
This approach enabled them to calculate the energy g=$kH
m
P
stored and subsequently dissipated in the loading and
subsequent collapse of each transitory column. Their where E is the void faction, d, is the diameter of the
extensive plane strain tests on real granular materials particles making up the agglomerate, k is the coor-
show good agreement with this conceptual theory. dination number and H is the inter-particle force acting
A similar approach may be used to model confined at each contact point. Rumpf assumed that the particles
uniaxial compression. In effect, the bed is modelled as are monodisperse and spherical and that the bonds
a series of parallel columns (Fig. 3). Since the applied between them were of equal strength and randomly
stress is transmitted via relatively few preferred routes, distributed across the failure section. By summing the
the number of particles which are undergoing or are contributions of the individual bonds across the section
about to undergo fracture is also relatively small. As Rumpf also effectively assumed instantaneous failure
compression proceeds and particle fracture occurs, some at all points.
re-arrangement is permitted, load paths bifurcate and More recently it has been argued that agglomerates,
more particles become load-bearing. (In terms of Fig. like many other solids, will fail in tension by a process
3, the number of columns which are ‘active’ increases of crack propagation [lo, 11, 20, 211. It is then possible
as the compression proceeds.) This is a well-known to apply fracture mechanics techniques to determine
problem in comminution, where repeated packing rear- the strength. Such an analysis shows that the applied
rangements must occur in order to ensure that the stress is concentrated at flaws or cracks within a body;
loads on individual particles remain high (see, for for elastic systems the applied stress becomes singular
example, ref. 17). Any model must take the features at the crack tip. The stress field ahead of the crack
mentioned above into account. Here, they are intro- tip is proportional to a quantity known as the stress
duced by bringing an increasing proportion of all possible intensity factor, K, which increases with the loading.
columns into play as the compression proceeds. In practice, it has been shown that these linear elastic
8

fracture mechanics concepts have to be modified to where A* = k,nA and n is the total number of active
take account of inelastic processes occurring at the or load-bearing columns.
crack tip. The analytical procedures involved to de- Note that as the compression proceeds and agglom-
termine the critical value of K at fracture, the ‘critical erates fail, the columns of which they form a part may
stress intensity factor’, K,, for agglomerates have been become inactive (in the load-bearing sense) only to be
described elsewhere [21]. In short, Kc is a measure of reactivated at a higher strain. However, in this simple
the stress at which fracture occurs by crack propagation approach, the failure load F of each column is the
and is therefore an intrinsic material property. same, so that any increment in load implies an increase
For an isolated column or agglomerate, failure in in the total number of load-bearing columns, n. Thus,
uniaxial compression occurs by tensile crack opening
dP=7dA* (6)
[22], as in the well-known ‘Brazilian test’. In the ‘column
model’ of confined uniaxial compaction described here, or
however, local crack opening within a single agglomerate
resulting from the major principal stress acting in the dP=rdA*lA, (7)
axial direction would be constrained by the radial where dP is the increment in nominal pressure and A0
principal stress induced by neighbouring columns. Under is the cross-sectional area of the bed. The number of
these conditions, the particles would be constrained to load-bearing columns must increase with the applied
fail in oblique shear. Such failure is common in the strain - dhlh in order to support the load. The simplest
fracture of stone columns under a compressive load. assumption is then that the increment in new fracture
This was first recognised by Coulomb [23] who related area is directly proportional to the strain. This may be
the shear failure stress, 7, to the sum of the cohesive written in the following form in order to render the
strength r0 and the frictional stress Crp’ acting at the constant of proportionality, ka, dimensionless:
failure planes, thus
7=70+& (3) dA*=k* $A,,

where (Y is the pressure coefficient or coefficient of Thus


friction and P’ is the compressive stress. Due to the
generalisation by Mohr [24] that failure will occur when
dP=k,$
the shear stress in any plane’reaches a critical value,
this failure criterion is referred to as the Mohr-Coulomb
criterion. More generally it would now be known as The pressure P’ in the failure criterion (3) will now
one of a class of macroscopic stress criteria. be assumed to be proportional to the applied nominal
The theoretical basis of the above types of failure pressure, P, thus
criteria and their relationship with fracture mechanics P’=k$ (IO)
is described elsewhere [25]. Essentially, microcracks
grow but do not reach a critical state to result in gross The implied assumption here is that the entire stress
fracture because of the compressive stress field. Even- field scales with P.
tually, the micro-cracks coalesce and failure occurs by Combining eqns. (3) to (10) yields
net section rupture in shear. The failure force F mea-
sured in the axial direction will be proportional to the dP= - (7b+a’P) f
product of the failure stress and the cross-sectional
area of the fracture plane A, so that where
F=k12A (4) k, T()
7b= - (12)
where k, is a proportionality constant. In the context k3
of the ‘column’ model of the bed, the quantity F and
represents the maximum force that the column can
sustain; this corresponds to the failure force of the a’=k+ (13)
weakest granule. If this is summed over all ‘active’
Integrating eqn. (11) gives:
columns (i.e. those which contain a granule which is
currently failing), then the total force F* may be written P

as

+*=d*
s
PO
-=-
dlJ
Tb+a'P s
.h dh
ho
T (14)

where h, is the initial height of the bed.


9

Hence, the force-displacement behaviour was obtained for each


test.
ln(rb+a’P)-ln T;=(Y) In@&) (15) The single-particle crushing test was also carried out
Rearranging and introducing the natural strain E using the ‘JJ’ tester and consisted of a simple compres-
( = In&/h)) gives sion between platens at the same cross-head speed.
Because of the large statistical variation in fracture
InP=ln(r~/cu’)+cu’c+ln[l-exp(-a’e)] (16) strengths of single agglomerates, the data presented
This relatively simple equation provides a pres- here represent averages over ten measurements in each
sure-volume relationship for confined uniaxial compres- case.
sion. At large values of E, a plot of In P as a function
of E should be linear with a slope of (Y’and an intercept Results
of ln(rblcu’), from which T; can be calculated.
It is of interest to relate the quantity, rO’, to the Figure 4 shows a typical set of data obtained from
single agglomerate failure load measured in the simple the uniaxial confined compression test, plotted in the
compression test. In the diametric compression of spher- form of the natural log of the applied stress against
ical agglomerates, the failure zone scales with the the natural strain. As predicted by the theory, the plot
cross-sectional area of the agglomerates [22]. For the is linear at large values of strain, since under these
purposes of comparison, therefore, a single agglomerate circumstances the third term on the right of eqn. (16)
failure load, FcALc, has been defined as the product is small. Initial values of the intercept, ln(rO’l&), and
of the agglomerate maximum cross-sectional area and the gradient, cy’, of this line may be calculated from
the cohesive strength: the data obtained at large strains; in this case, the last
10% of data points were used. The value of (Y’ was
d2 nd: k3
F cALc- -a (17) then adjusted incrementally to give the best fit of the
4 r”=4 zb full form of eqn. (16) to the entire data set, on a least
where d, is the agglomerate diameter. Without con- squares basis. Finally, r,,’ was found from the intercept
sidering the details of the mechanisms leading to single and gradient of the linear portion of the final curve.
agglomerate failure it is impossible to proceed beyond Table 1 shows the calculated values of ro’ and (k,/
the crude form of eqn. (17) which is, in any case, k#‘oALc for a range of agglomerate diameters from
sufficient for the present purpose. 1440 to 2 580 pm and binder concentrations from 0.5
to 3.2% by weight. Agglomerates with less than 0.5%
binder were found to be too weak to give reliable
Experimental methods results.
Values of single particle crushing load, F, for the
The agglomerates to be studied were formed from same agglomerates are given in Table 2. The quantity
quartz sand using a polyvinyl-pyrrolidone (PVP) binder
of molecular weight 44 x 103. The sand was first sieved
to separate a size fraction between 45 and 180 pm. InP -
The particles were then wet-mixed with the binder in
aqueous solution, using a planetary mixer. The wet
mass was granulated in a rotating pan, and the ag-
glomerates were dried and sieved into the desired
fractions. w
For the confined uniaxial compression test, the ag- .
ii
glomerates to be tested were placed in a 37.2 mm
diameter cylinder and tapped to remove any large
5 -
packing irregularities. The bed depth was measured
using a vernier micrometer. An initial depth of about
18 mm was used throughout. A close-fitting piston was
then applied and the powder was compressed using a
‘JJ’ testing machine. In all the experiments reported
here the cross-head speed was maintained at 3.5 mm 0-I’ ” ” ” ’
0 5 10
per min. The applied force and the displacement were STRAIN x 100
measured using a load cell and a linear voltage dis- Fig. 4. In P vs. strain for sand agglomerates 2 360-2 800 pm
placement transducer respectively; these sensors were (3.2% binder; run 8)+data points; every tenth point shown -
interfaced to a computer so that a complete record of eqn. (16).
10

TABLE 1. Results of uniaxial confined compression tests

Test Binder Agglomerate (wwCALc*


no. COIYC. diameter, d, & m-*) (N)
(%) (pm)

6 0.5 2580 0.59 x lo6 2.7

12 1.0 2580 1.54 x lo* 8.1


4 2.0 1850 3.40 x lo6 9.2
3 2.0 2180 2.70 x lo6 10.1
5 2.0 2580 2.04 x lo6 10.7

11 3.2 1440 5.03 x lo6 8.2


10
9
8
3.2
3.2
3.2
1850
2180
2580
4.87
3.78
3.01
x
x
x
lo6
lo6
106
13.1
14.1
15.7
I . CONC2
3.2 % I

*From eqn. (17).

TABLE 2. Results of single particle crushing tests

Test Binder Agglomerate Crushing Flda2


no. cont. diameter, d, load, F (N m-‘)
(%) (pm) (N) Xl@
5 10 15

6 0.5 2580 0.7 f 0.2 1.05+0.3 Fexpt (N)

12 1.0 2580 3.1 f 0.9 4.7 f 1.4 Fig. 5. Comparison of values of calculated and measured single
particle fracture loads.
4 2.0 1850 5.7+0.8 16.7 k 2.3
3 2.0 6.9 + 1.9 14.5 rt4.0
5 2.0 2580 7.4* 1.3 11.1*2.0 to describe compression of a mass of powder. One of
11 3.2 1440 6.2* 1.1 29.9 f 5.3
the most widely used of these equations is that due
10 3.2 1850 6.9k1.8 20.2 + 5.3 to Kawakita and co-workers [7, 261:
9 3.2 2180 10.8 of:2.0 22.7 of:4.2
8 3.2 2580 14.0 f 2.9 21.Ok4.4 v,-v
-=- abl'
(18)
v, l+bP
(F/d:) should remain constant with changes in ag- where V, is the initial volume of the bed, V is the
glomerate size providing that the binder concentration current volume, P is the applied pressure and a and
remains unchanged. In fact, it is apparent from the b are constants. It is clear from eqn. (18) that as P
data of Table 2 that (F/&*) decreases somewhat with becomes large so that bP> 1,
increase in agglomerate size, suggesting, unsurprisingly,
that factors in the agglomeration process reduce the vo-v..
strength of the agglomerates as they become larger. a=V,
That the TV’values in Table 1 follow a similar trend where V, is the volume of the particles alone. Kawakita
is further evidence to support the analysis presented et al. [26] found that the values of the parameter a
above. they obtained from fitting eqn. (18) to experimental
Figure 5 compares values of (kZ/k3)FCALCderived data were in good agreement with the initial porosity
from the confined compression test and the measured of the powder bed. Combining eqns. (18) and (19) we
single particle fracture load. Despite the considerable have
scatter in the single agglomerate fracture loads, the
experimentally-determined value of the single particle p(!c-$)~$k$) (20)
crushing load is seen to be approximately proportional
to the single agglomerate failure load inferred from
the uniaxial confined compression test. which is one of the forms of his equation which Kawakita
recommends in order to obtain a value for the constant
b. Figure 6 shows a plot of a typical set of data from
Discussion the sand/PVP agglomerates, illustrating good agreement
with eqn. (20). Kawakita er al. [26] suggested that the
It is of interest to compare the predictions of eqn. constant (l/b), which has dimensions of pressure, is
(16) with those of the empirical equations often used related to the yield stress of the individual powder
11

Thus,
P v-v,

% In P= ln(T,‘/a’) + a’ In(V,/V) (22)


k-l

or

!c
p=$
a’ (1v
From Kawakita’s eqn. (20):

Equations (23) and (24) will give similar estimates for


P provided that their dependencies on strain, and
therefore current volume, V, are similar in the range
of interest. In fact, the voidage terms (l/cy’)(VO/v’
and (Vo-- V)I(V- V,), from eqns. (23) and (24), are

r-1
v, - v
indeed proportional over certain ranges of void fraction.
VO
For V, lV,,= 0.44, for example, the two terms are ap-
Fig. 6. Kawakita plot for 2 360-2 800 Frn sand agglomerates (1.0%
proximately equal in the range 0.9 > V/V,,> 0.6 if CY’= 4;
binder; run 12).
as the value of a’ increases, the range of void fraction
over which the two expressions are proportional de-
creases, so that for (Y’=20

(25)
over the range 0.98 > V/V, > 0.92. In this work, (Y’varied
between 10 and 30, but strains were confined to less
than lo%, so that eqn. (24) may be used, leading to
Tb= 0.7/b (26)
which is shown in Fig. 7, giving good agreement with
the data for this value of V,/V,.

.. , , , , , , , Conclusions
0
0 2 L 6
AGGLOMERATE SHEAR STRENGTH 7; (MRI Uniaxial confined compression of the agglomerates
Fig. 7. Comparison of Kawakita parameter, b, with agglomerate considered here is a purely dissipative process in which
shear strength, TV’, for all confined compaction experiments; eqn. strain can only be accommodated as the result of
(26). individual agglomerate failure. Here, the Mohr-
Coulomb macroscopic stress criterion is applied to the
particles. Values of l/b from this work are plotted in failure of individual agglomerates. The modelling prob-
Fig. 7 against values of TV’, obtained by fitting eqn. lem then reduces to describing (i) the way in which
(16) to the experimental data. The excellent linear the effective area over which failure occurs varies with
correlation confirms the link between Kawakita’s con- the extent of the overall deformation (i.e. the strain)
stant and the yield stress of the particles, which may and (ii) the way in which single agglomerate failure is
initially seem surprising because of the apparently dif- affected by the local normal stresses on the failure
ferent forms of eqns. (16) and (18). plane. In this analysis, extremely simple assumptions
Developing the comparison further, at large values have been made in each case, leading nevertheless to
of strain, eqn. (16) reduces to an expression for the load-deformation relationship
which is a good approximation to experimental be-
In P= ln(T,‘/a’) + a’6 (21) haviour. Moreover, this simple theory provides a means
Since E is here the natural strain it may be replaced by which the average shear strength of a single ag-
by ln(V,,lV), using the notation of eqn. (16). glomerate can be obtained by experiments on a bed
12

of agglomerates, and this value is related to the single kl, k2,k, proportionality constants, -
particle crushing strength through a single empirical number of active columns (eqn. (5)), -
proportionality factor. ;: pressure, stress, Nme2
The form of any expression for the load-deformation P’ local value of normal stress on failure plane,
relationship under confined uniaxial compaction must Nmp2
include some measure of the strength of the individual V bed volume, m3
particles (in this case, agglomerates) being compressed. vo initial bed volume, m3
The analysis presented here clearly shows that the V/, final bed volume (eqn. (19)), m3
constant b in Kawakita’s equation is to be interpreted
as proportional to the reciprocal of the single particle Greek letters
failure stress, which is in agreement with the qualitative a parameter in Mohr-Coulomb failure cri-
arguments put forward by Kawakita et al. [26]. terion (eqn. (3)), -
ff’ =k,a (eqn. (13)), -,
E void fraction, -
% mean tensile strength (eqn. (2)), Nme2
Acknowledgements 7 failure stress, Nmp2
70 strength of agglomerate at zero normal load,
This work was carried out as part of a project Nmp2
sponsored under the Particulate Technology Specially rb =(k2/k3)~o (eqn. (12)), Nme2
Promoted Programme (S.P.P.) of the Science and En-
gineering Research Council. The grant-holder (Dr J.P.K.
Seville) wishes to acknowledge the assistance of Mr
L.J. Ford, the S.P.P. Coordinator. The authors also
wish to thank Professor R. Clift, Dr J.G. Petrie and References
Dr U. Ti.izi,in for useful discussions during the prep-
aration of this paper, and Dr C. Thornton of the C. Thornton and D.J. Barnes, Acta Mechanicu, 64 (1986) 45.
R.W. Heckel, Trans. Mefall. Sot. AIME, 221 (1960) 671.
Department of Civil Engineering of Aston University
R.W. Heckel, Trans. Metall. Sot. AIME, 221 (1961) 1001.
for the use of Fig. 2. R.J. Roberts and R.C. Rowe, Chem. Eng. Sci., 42 (1987) 903.
A.S. Khan, Y. Xiang and S. Huang, Int. 1 Plast., 7 (1991)
607.
D.P. Isherwood, in B.J. Briscoe and M.J. Adams (eds.),
TriboZogy in Particulate Technology, Hilger, Bristol, 1987, p.
List of symbols
234.
7 K. Kawakita and K.-H. Ludde, Powder Technol., 4 (1970/71)
A area of failure surface, m2 61.
A* =k+4, m2 8 R.P. Seelig and J. Wulff, Trans. Am. Inst. Min. Metall. Eng.,
cross-sectional area of bed, m2 166 (1946) 492.
A0
9 A.R. Cooper and L.E. Eaton, J. Am. Ceram. Sot., 45 (1962)
a,, 4, a2 parameters in eqn. (l), -
97.
a, b parameters in Kawakita’s equation (eqn. 10 M.A. Mullier, J.P.K. Seville and M.J. Adams, Chem. Eng.
(B)), -, m* N-’ Sci., 42 (1987) 667.
diameter of agglomerate, m 11 M.J. Adams, M.A. Mullier and J.P.K. Seville, in B.J. Briscoe
diameter of primary particle making up and M.J. Adams (eds.), TriboZogy in Particulate Technology,
Hilger, Bristol, 1987, p. 375.
agglomerate, m
12 M.A. Mullier, J.P.K. Seville and M.J. Adams, Powder Technol.,
F failure force, N 65 (1991) 321.
F* total force acting on test cell, N 13 P.A. Cundall, A. Drescher and O.D.L. Strack, Proc. IUTAM
F CALC calculated single agglomerate fracture load Conj, Delft, 1982, p. 355.
14 A. Drescher and G. de Josselin de Jong, 1. Mech. Phys. SoZids,
(eqn. (17)) N
20 (1972) 337.
H interparticle force (eqn. (2)), N
15 J.P. Troadec, D. Bideau and J.A. Dodds, Powder Technol.,
h bed height, m 65 (1991) 147.
11 first invariant of deviatoric stress tensor, 16 T. Dunstan, J.R.F. Arthur, A. Dali& 0.0. Ogunbekum and
Nmd2 R.K.S. Wong, Nature, 336 (1988) 52.
second invariant of deviatoric stress tensor, 17 B. Steenberg, Powder Technol., 37 (1984) 2.89.
J2
18 H. Rumpf, in W.A. Knepper (ed.), Agglomeration, Inter-
Nme2
science, New York, 1962, p. 379.
K stress intensity factor, Nrnp3” 19 H. Rumpf, in K.V. Sastty (ed.), AgZomeration 77, AIME,
KC critical stress intensity factor, Nm-‘” New York, 1977, p. 97.
k coordination number, - 20 M.J. Adams, J. Powder Bulk Solidrs Technof., 9 (1985) 15.
13

21 M.J. Adams, D. Williams and J.G. Williams, /. Muter. Sci., 24 0. Mohr, VIII Z., 44 (1900) 1524.
24 (1989) 1772. 25 B. Paul, in H. Liebowitz (ed.), Fruchrre, An Advanced Treatise,
22 C.E. Capes, Particle Size Enlargement, Elsevier, Amsterdam, Vol II: Mathematical Fundamentals, Academic Press, New
1980. York, 1968, pp. 315-496.
23 C.A. Coulomb, Mem. Math. Phys., Acad. Roy. Sci., Vol. 7, 26 K. Kawakita, I. Hattori and M. Kishigami, J. Powder Bulk
Paris, 1776, pp. 343, 382. Solids Technol., 1 (1977) 3.