You are on page 1of 7


Pl ast er of Paris as a model mat er i al f or br i t t l e

porous sol i ds
G. V E K I N I S * , M. F. A S H B Y , P. W. R. B E A U M O N T
Cambridge University Engineering Department, Trumpington Street, Cambridge CB2 1PZ, UK
Pl ast er of Paris is a br i t t l e, p o r o u s sol i d, easy t o shape, w h i c h has pot ent i al as a model mat er i al for
t he s t u d y of br i t t l e, por ous , sol i ds s uc h as cer ami cs, r ocks and cement . Thi s paper descr i bes t he
mec hani c al pr oper t i es of pl ast er of Pari s - mo d u l u s , st r engt h, f r act ur e t o u g h n e s s , etc. - as
a f u n c t i o n of por osi t y. The mat er i al is t hen used t o s t u d y t he i ni t i at i on and p r o p a g a t i o n of cr acks in
c o mp r e s s i o n , as a f u n c t i o n of por osi t y, st r ess st at e and st r ess c o n c e n t r a t i o n .
1. I n t r o d u c t i o n
Plaster of Paris (hydrat ed calcium sulphate) is a brittle
solid with fracture properties which resemble those of
cement, sandstone, and ot her porous ceramics. It can
be shaped easily, and used as a model material to
st udy the behaviour, under load, of porous solids
cont ai ni ng macroscopic cracks, holes and reinforce-
ment. The first part of this paper reports a st udy of the
properties of plaster of Paris - the modul us, tensile
and compressive strengths, modul us of rupture, and
fracture toughness as a function of the density (or
porosity) of the plaster.
The second part describes an example of an applica-
tion: the mechani sms of failure at a cylindrical hole or
spherical pore under simple and multiaxial compres-
sion. Stress is concent rat ed at a macroscopic hole in
an elastic solid. In uniaxial compression, tension ap-
pears at one pair of poles of the pore; but in most
biaxial compression states all principal stresses at and
near the pore are zero or compressive. Despite this,
fracture occurs at the pore surface in a porous or
mi cro-cracked mat eri al like plaster, causing i nward
spalling of mat eri al from the pore surface. The prob-
lem is known in the mi ni ng industry, when geostatic
loads cause i nward spalling of boreholes, and it is
probabl y the mechani sm by which fracture starts in
porous ceramics under compressive stress states. We
have used plaster of Paris samples to investigate the
i ni t i at i on and progress of cracking from holes and
pores, under a variety of stress states. The results are
related to observations and models for compressive
failure of brittle solids [1-16].
When plaster of Paris is mixed with water the reverse
reaction takes place: water is reabsorbed with the
format i on of gypsum. The reaction is exothermic and
results in a coherent mass of interlocking needle-
shaped gypsum crystals. The chemistry of the reaction
requires onl y 18.6 wt % water for rehydrat i on, but in
practice much more is used to give the fluidity needed
for casting. The excess water evaporates leaving con-
siderable porosity. The true density of the hemihy-
drate is about 2750 kg m- 3 and t hat of the di hydrat e
about 2320 kg m- 3, so a cont ract i on on setting would
be expected; but the arrangement of the crystals is
such t hat setting results in a slight expansion (about
The mechanical properties of plaster depend on
powder-t o-wat er ratio, curing time, t emperat ure and
pressure, and on post-cure heat treatment. All were
investigated [16] and a st andard procedure was ad-
opted. The CaSO4" 89 starting powder (British
Drug House) was mixed with distilled water, in the
ratio 100 : 62.5, removing all air t rapped in the suspen-
sion, and was cast into split rectangular (approximate
size 10 mm x 10 mm x 100 mm) or cylindrical (ap-
proxi mat e size 10mm di amet er x 80 mm height)
moul ds and allowed to cure for at least 7 days at 20 ~
Specimens of higher density t han t hat of the as-cast
specimens were obt ai ned by forcing water out of the
moul d under pressure, i mmedi at el y after casting. The
samples cont ai n small spherical pores (Fig. 1), prob-
ably due to t rapped water duri ng casting, which de-
crease in size with increasing bulk density of the
material (Fig. 2).
2. P l a s t e r o f Par i s: t h e m a t e r i a l
Plaster of Paris is calcium sul phat e hemi hydrat e,
CaSO4- 89 It is made by heat i ng gypsum between
120 and 160 ~
CaSO4" 2 H2 0 ~ CaSO4. 89 + 1 89 (1)
3. T h e m e c h a n i c a l p r o p e r t i e s o f p l a s t e r
o f Pa r i s
3. 1. Plain speci mens
A number of mechanical properties were measured as
a function of density. The properties included Young' s
* Present address: "Demokritos" National Centre for Scientific Research, Institute of Materials Science, 153 10 Ag Paraskevi Attikis, Athens,
0022-2461 9 1993 Chapman & Hall 3221
0 i I q 0
0.4 0.5 0.6 0.7 O.B
Relative density
Figure 3 Young' s modulus in uniaxial compression and modulus of
rupture in four-point bending as a function of relative density.
Figure 1 The structure of the plaster: (a) 50% relative density and
(b) 70% relative density specimens.
" o
'~ 100
0.4 0.8 0.5 0.6 0.7
Relative density
Figure 2 Macropore diameter as a function of relative density of
hydrated plaster of Paris.
modulus (by four-point bending, Ebend, and under
uniaxial compression, Ecomp), the modulus of rupture
under three- and four-point bending (MOR3 and
MOR4), the fracture toughness, Kit, by four-point
straight-edge-notched beam (SENB), the uniaxial ten-
sile strength, cyt, and the uniaxial compressive
strength, ac. In addition, the compressive strengths
under biaxial and hydrostatic conditions were also
measured. The results for the as-cast material are
summarized in Table I, which also lists data from
earlier published studies.
Young's modulus, E, was determined from the load-
ing curves in bending and uniaxial compression, as
described below.
Four-point bend tests were used to measure the
modulus of rupture (Table I). The beam specimens
(10 mm x 10 mmx 90 mm) were lightly ground using
1000 grade SiC paper prior to testing in order to
remove an outer skin which has a different morpho-
logy from the bulk of the specimens. Modulus of
rupture values were calculated using the well-known
MOR4 - (2)
where P is the fracture load (N), As is the difference
between major and minor spans, I is the second mo-
ment of area of specimen (bda/12, where b is the width
of specimen), and d is the height of specimen. ' All the
specimens displayed purely elastic behaviour up to
fast fracture. The variation of modulus and MOR4
with relative density and associated pore size is shown
in Fig. 3. Young's modulus for four-point bending
(Table I) was calculated using the elastic beam theory
dP AS 3
gbend - - dx 4bd 3 (3)
The fracture toughness measurements were carried
out on rectangular edge-notch beam specimens with
the same dimensions as those used for the bending
tests. The fracture toughness, Klc (Table I), was cal-
culated from the analysis of Strawley and Gross [19]
P As 3"C~ 1/2
KIc = b/d d 2(1 - a) 3/2 (4)
where ~ is a numerical constant, of order unity, which
depends on the ratio of notch depth, a, to specimen
height, d. The variation of K~c with relative density
and associated pore size is shown in Fig. 4.
Tensile tests proved difficult: the problem is that of
achieving alignment. Special rubberized split grips
were designed which provided effective gripping of the
cylindrical specimens without high hoop stresses [16].
Only results of specimens that fractured within the
TABLE I Properties of plaster of Paris (as cast, 62.5% water)
Present study Previous work
[t7, 183
Physical properties
Mechanical properties
Theoretical density (kg m- 3)
Density (kg m -3)
Total porosity content (%)
Mean diameter of spherical macropores (lain)
Mean grain size (Ixm)
Young's modulus (bending) (GPa)
Young's modulus (uniaxial compression) (GPa)
Modulus of rupture (four-point bending) (MPa)
Weibull modulus
Fracture toughness (SENB), (MPa m ~/2)
Uniaxial compressive strength (MPa)
Hydrostatic compressive strength (MPa)
Uniaxial tensile strength (MPa)
1170 _ 30
212 +_ 18
4.5 +_ 0.1
4.6 +_ 0.3
5.8 _+ 0.6
0.14 _+ 0.015
14.6 -t-_ 0.9
19.2 _+ 1.4
3.2 _+ 0.6
~, o:o
'.'- 0.05
F r a c t u r e
t ough?r "ce' ~
9 <>
0 i i I
0.4 0.5 0.6 0,7
Relative density
8 n ~ Hydrostatic
7 c o m p r e s s i o n
i n i t i o t i o
6 _
/ 9 ~ U n ( o x i o l U n i o x i a l
5~ ~ / M-~mpr~! Te / compressi on
~- Fracture
3 initiation
Figure 4 Fracture toughness, Kl~, and the tensile fracture strength
as a function of the relative density.
gauge length were accepted. The tensile st rengt h of the
as-cast mat eri al is given in Tabl e I. The vari at i on of ch
with relative densi t y and associated por e size is also
shown in Fig. 4.
Compr essi on specimens were pr epar ed by gri ndi ng
the ends parallel t o bet t er t han 5 Ixm c m- ~ and per-
pendi cul ar with respect to the sides t o bet t er t han 0.2 ~ .
To avoi d damage to the ends of the specimen and
reduce friction bet ween anvils and specimen surface,
paper shims were used in all tests. A typical compres-
sion curve is shown in Fig. 5 (lower curve). The initia-
tion stress for compressi ve fract ure and t he ul t i mat e
compressi on st rengt h are shown in Fig. 6 (lower pai r
of curves).
Equi-biaxial experi ment s were carri ed out bot h in
a n I n s t r o n a n d i n t he S E M c o mp r e s s i o n ri g, u s i n g
we dge gr i ps [ 16] . Hy d r o s t a t i c t es t s wer e c a r r i e d o u t
o n s h o r t c y l i n d e r s (10 mm d i a me t e r x 10 mm he i ght )
e n c l o s e d i n a l at ex r u b b e r s h e a t h a n d p r e s s u r i z e d i n
di s t i l l e d wa t e r i n a P T F E c a p s u l e i n a h i g h - p r e s s u r e
vessel ( Fi g. 7); t he a r r a n g e me n t c a n be u s e d u p t o
Figure 5 Load contraction curve for a specimen tested in uniaxial
compression (lower curve) and in hydrostatic compression (upper
3 -- Hydrostatic
g ' 0 ,=
Z compression ~ "~o J~
0 l l
0.4 0.5 0.6
Relative density
Figure 6 Fracture initiation stress (first
/ [ ]
[ 3 {3 [3 /
Ultimate hydrostatic /
compressive stress / / /
/ / /
/ ~ 0 0 / 0 1 1 0
D/ /'~;, Fracture
. / / / $ initiation stress
_~o . . . . . . . . . . =/
/ / O
o /r / /
Unioxiol ~ ~ o
c o m p r e s s i v e strength / / o / /
\ / n / ,
o ~- z ~ "- Fracture
~Uniaxiol i . . I / i o initiation stress
0'.7 0.8
non-linearity of the
stress-strain curve) and ultimate strength, as a function of relative
density, for uniaxial compression (lower curves) and for hydrostatic
compression (upper curves).
Wot er -
E.:-;: :.
Pr e s s u r e cell
~ - ' ~ ' .i.g s
!!i:!ji . ,
Figure 7 The PTFE cell used in the hydrostatic compression ex-
pressures of 1 GPa. All the specimens displayed an
elastic response up to the point at which fracture
initiated, detected by small steps in the loading curve.
A typical pressure volume curve is shown schemati-
cally in Fig. 5 (upper curve). The hydrost at i c collapse
pressure is plotted against density in Fig. 6 (upper
curves). Specimens removed from the cell after testing
had been compact ed to about three-quarters of their
original volume, and were very weak, crumbling easily
to a coarse powder.
Specimens were removed from the pressure vessel
af t er various pressures to examine the progressive
collapse of the contained porosity. Damage starts at
the surfaces of the pores. Collapse proceeds by fractur-
ing of large segments immediately adjacent to the
pores which fall inwards, filling the pore. The final
stages of collapse are characterized by a general
change of pore shape, with general crushing of the
surroundi ng material combi ned with large-scale frac-
turing. It appears t hat pore collapse takes place prior
to any bulk fracturing of t he material and proceeds
extensively before large-scale fracture of specimens.
3. 2. Sp e c i me n s c o n t a i n i n g c y l i n d r i c a l hol es
Uniaxial and biaxial tests were carried out on cubes
(approximate size 4.5 mm x 4.5 mm 4.5 mm) of fully
cured plaster of Paris with various densities. A single
cylindrical hole of diameter 0.5 mm was i nt roduced by
drilling at low speed. The specimens were prepared
and tested as before, under uniaxial and biaxial stress
states. The observations of pore collapse were made
using an in si t u SEM compression rig.
The stress for initiation of hole collapse and t hat for
bulk fracture initiation in simple compression were
measured for a range of densities. The results are
summarized in Fig. 8. The sequence of events taking
place duri ng uniaxial compressive failure of a cubic
specimen cont ai ni ng a drilled hole is illustrated in the
micrographs of Fig. 9. The compression direction is
2 0 [ S t r a i n s , .
~15[ ~/~-"~'~. /
10 ~ BUlk f r Q c t u r ~ ' ~ . ~ . . . f ~ . o
oE S I ~ St r e s s e s _ ~ f - - ' - ~ " ~ " ~ , ,L._
0.5 0.6 0.7
Reletive density
0.1 _~
Figure 8 Fracture initiation and hole collapse initiation stress and
strain as a function of relative density.
vertical. Some bulk fracture of the specimen precedes
hole collapse in simple compression. The hole t hen
collapses by the fracture of arc-shaped segments
ar ound the walls of the hole, particularly at areas of
highest compressive stress concent rat i on at the equa-
tor of the hole. The collapse and fracture process is
discontinuous. The l oad needed to continue the col-
lapse process decreases with increasing strain (as in
Fig. 5), al t hough the overall specimen collapse is
stable up to large strains.
Even a small degree of lateral const rai nt (i.e. biaxial
load) results in the simultaneous initiation of bulk
fracture and hole collapse. When the degree of con-
straint approached equi-biaxial loading, hole collapse
preceded the initiation of bulk fracture (Fig. 10). As
before, the collapse of the hole involves the inward
fracturing of arc-shaped segments from the walls. The
collapsed areas become more uniformly distributed
around the wall of the hole as an equibiaxial stress
state is approached. Hole collapse reaches an
advanced state before significant overall specimen
collapse. The collapse process is discontinuous, but
significantly more stable t han in the uniaxial case, i.e,
the l oad required to sustain the collapse process re-
mai ned const ant or decreased very slowly with in-
creasing strain (as in Fig. 5).
4 . D i s c u s s i o n
Models for the compressive cracking of brittle solids
[2, 14, 15] suppose t hat cracks initiate at flaws (the
flaws might be spherical pores or sharp cracks). The
models lead to an expression for the stresses for crack
initiation; it is a relationship between the axial stress,
0.1, and the radial stress, o3 = (~2, (c~1 > ~3), which
can be written as
O ' 1 = C l 0 " 3 - - C o ( 5 )
where the constants cl and Co depend on the nat ure of
the flaws. When cracks nucleate from spherical holes,
the const ant cl is predicted to be about 3.1-3.4 [15].
When, instead, they nucleate from cracks [2, 14] it
depends on the coefficient of friction, ~t, between the
sliding crack faces and ranges from 1-3.5 as ~t ranges
from 0-0.7. (Experimental results on crack initiation
in Westerly granite, quot ed in [20], suggest a value of
Figure 10 Progressive failure of CaS04 with a hole under con-
strained uniaxial compression; cr2/~1 ~ 0.2.
Figure 9 Progressive failure of CaS04 with a hole under uniaxial
cl of approximately 2.8 for both pores and cracks for
this material.) The constant c o depends on the fracture
toughness, K~, and the flaw size, 2a. For pores,
Co ~- 1. 6K~c / ( ~aLo) 1/2, where 2a is the diameter of the
pore divided by the pore radius). For sharp cracks,
c o = 3.1 Ki c / ( r c a) 1/2, where 2a is the length of the in-
clined crack. Both the form of Equation 5 and the
co = 3. 1Kl o/ ( r ~a) 1/2, where 2a is the length of the in-
clined crack. Bot h the form of equation 5 and the
constants are almost identical for bot h extreme types
of flaw; it may therefore be considered as a general
criterion for damage initiation in compression [21].
Final fracture involves crack-crack interaction. It is
treated in earlier publications [14, 15, 21].
For uniaxial compression cy3 = 0 and Equation
5 simplifies to
(5" 1 --
o- 1 - -
- - 1. 6Kl c / ( ~apLo) 1/2 (holes) (6a)
- - 3. 1Kl c / ( r c ac ) 1/2 (cracks) (6b)
where ap and ac are the dimensions of the pores and
cracks, respectively. A plot of uniaxial fracture initia-
tion stress (as determined by the first non-linearity of
the l oad-di spl acement curve) versus Kmc/ ( rca) 1/2
should be linear through the origin and yield a "best-
fit" value for Lo. The plot (Fig. 11) shows that the
fracture initiation model fits the observations well
with a calculated (best-fit) value for Lo = 1.5.
When hydrostatic stress is applied to a porous
body, shear stresses appear at and near the pores
(Fig. 12). The local stress state can be described by
a radial stress, ~,, and two equal tangential stresses,
0 0
, $ - -
- - -10
g - 2 0
= - 30
- 4 0
, < , c
10 20 30
K I C/ ( r oe) 1/2 ( MPo)
Figure 11 Uni axi al fract ure i ni t i at i on stress of pl ast er of Paris:
compar i s on wi t h model s [15, 21].
Q _
'=" - 4 0
9 g- r/q = 1,1
-50 2
- - 6 0
50 40 3'0 2'0 I b 0
(l-f) c o
Figure 13 The hydr ost at i c stress for fract ure i ni t i at i on pl ot t ed
agai nst (1 - f ) c o , fol l owi ng Equat i on 11. The line cor r esponds t o an
i ni t i at i on dept h r/a = 1.1.
/ \ \
/ / b \
i P o r e
/ I
I " - " _ ~ / l " o ' ~ I
\ /
\ F l a w /.~,..
. . , . ~\ ,--,
", /
N /
\ /
( o )
- #
1 - f
r = Q
- Cr
Figure 12 (a) Stresses ar ound a spheri cal por e subj ect ed to hydr o-
st at i c pressure, p, and (b) t hei r var i at i on wi t h di st ance from a pore,
r ( Equat i ons 7 and 8).
~o, where [223
b 3 ( r 3 - - a 3)
(3" r - - p r3 ( b 3 _ a3 ) ( 7 )
(3 0
b 3 ( a 3 + 2 r 3)
- p 2 r 3 ( b 3 - a 3)
( 8 )
3 2 2 6
Far f r om the por e all stress component s appr oach the
hydr ost at i c stress level, - p , but , whereas the radi al
component , %, decreases f r om 0 at the por e surface to
-- - p at r = b, the t angent i al component s increase
f r om a mi ni mum val ue = - 3 / 2 p / ( 1 - f ) (where
f = a3/ b 3 is the vol ume fract i on of pores in the mat er -
ial) at the por e surface, to the value - p at r = b, as
shown in Fig. 12.
Appl yi ng the fract ure nucl eat i on Equat i on 5 to the
case of hydr ost at i c compr essi on of por ous brittle ma-
terials, we find (because % is less compressi ve t han q0)
~0 = c l o t - Co (9)
where cl -~ 3.1 and Co ~- 1. 6Kl c / Oz apLo) 1/2 as before
for holes. Subst i t ut i ng for or and c~0 we obt ai n
b 3 (a 3 + 2 r 3) b 3( r 3 - a 3)
- P i n 2 r 3 (b 3 _ a 3) + c l p i , r 3 ( b 3 _ a 3 ) - Co
using f = a3/ b 3 and solving for Pi, (the hydr ost at i c
stress for fract ure initiation) we finally find
2r3/ a3(1 - f ) c o
Pi, = l + 2 c l - 2 ( c t - 1 ) r 3 / a 3 (11)
The equat i on shows t hat fract ure i ni t i at i on is easiest
at the surface of the hol e (r = a), and becomes rapi dl y
mor e difficult as r increases. The i ni t i at i on pressure,
Pi. , becomes infinite at r/ a = [(1 + 2 c l ) / ( 2 c l - 2)] 1/3
( -~ 1.2 for cl -~ 3.1), beyond which poi nt no fract ure
woul d be expect ed to initiate.
I n or der to compar e the hydr ost at i c compr essi on
model i nt roduced above (Equat i on 10) with the results
obt ai ned in this wor k the hydr ost at i c fract ure initia-
t i on stress, pl , , has been pl ot t ed versus (1 - f ) K ~ c in
Fig. 13. I t is clear t hat the results can be descri bed
quite well by a linear rel at i on t hr ough the ori gi n as
suggest ed by the model . Fr om this gr aph the "best -fi t "
value of r/ a is -~ 1.1 which lies within the "critical" r / a
value predi ct ed by the model and wi t h the val ue of L0
found for the uni axi al compr essi on case above. Thus
fract ure i ni t i at i on in this mat er i al occurs very close t o
the por e surface as predi ct ed by the model .
Finally, the overal l behavi our of the pl ast er of Pari s
is summar i zed by the surfaces of Fig. 14, which show
the combi nat i on of stresses requi red to cause crack
- 2 0
9 ~" - 4 0
- 6 0
T e n s i o n
- 6 0
/ / /
/ / o- ~ 50oA
/ - ~ 6 0 ~
. . . .
..) "
l o
~ - - 2o
u ~
- 4o
- 6 0
- 8 0 ' ' ' 2 ' - 8 0
- 8 0 - 40 - 0 10 - 8 0 - 60
(G) Ax i a l st r ess ( MPa) (b)
T e n s i o n
/ / / ,'/~
/ / . . . . . . >//
/ / ~ - ~ ' 5 0 o / o / /
/ / J - I 7 0 %
Compressi on
I i I i I
- 40 - 20
Ax i a l st r ess (MPa)
0 10
Figure 14 (a) Fr act ur e i ni t i at i on, and (b) final col l apse surfaces for pl ast er of Paris, for t hr ee relative densities.
initiation, at (a), and final failure, at (b), for t hree
relative densities.
5. Conclusion
Pl ast er of Paris can be used as a model mat eri al to
st udy fract ure phenomena in brittle por ous solids. To
do this, the propert i es of the pl ast er are needed: t hey
are det ermi ned and listed in this paper. The mat eri al is
t hen used to st udy spalling at holes, in multiaxial
stress states. The ease of fabri cat i on and mani pul at i on
make pl ast er an at t ract i ve model mat eri al for such
Acknowl edgement s
We t hank the technical staff of t he Mat eri al s Gr oup,
Ri chard Brand, Bri an Butler, Alan Hear er , Si mon
Marshal l and Oscar Skulskyj, who pr ovi ded invalu-
able technical support . The research was made pos-
sible t hr ough the financial suppor t of the Uni t ed
States Air For ce Office of Scientific Research under
grant number AFOSR-87-0307.
1. A. A. GRI FFI THS, in "Pr oceedi ngs of t he 1st I nt er nat i onal
Congr ess on Appl i ed Mechani cs", Delft (Delft Uni versi t y
Press, 1924) p. 55.
2. S. NEMAT- NASSER and H. HORI I , J. Geophys. Res. 87
(1982) 6805.
3. D. GRI GGS and H. HANDI N, (eds), "Rock Def or mat i on",
(Geol ogi cal Society of Ameri ca, Wa ve d Press, Bal t i more,
1960) Ch. 13.
4. !dem, Geol. Soc. Amer. Mem. 79 (1960) 347.
5. F. A. Mc CLI NTOCK and J. B. WALSH, in "Pr oceedi ngs of
t he 4t h US Congr ess on Appl i ed Mechani cs" (ASTM, Int er-
science, 1962) p. 1015.
6. S. A. F. MURRELL, in "Rock Mechani cs", edi t ed by
C. Fai r hur st , Pr oceedi ngs of t he 5th Sympos i um on Rock
Mechani cs ( Per gamon Press, Oxford, 1963) p. 563.
7. Idem, Geophys. J. R. Astron. Soc. 10 (1965) 231.
8. S . A. F . MURRELL and P. J. DI GBY, ibid. 19 (1970) 309.
9. Idem, ibid. 19 (1970) 499.
10. J . C. J AEGER and N. G. W. COOK, " Fundament al s of Rock
Mechani cs" ( Met huen, London, 1969).
I1. S. KOBAYASHI , J. Soc. Mater. ScL Jpn 20 (1971) 164.
12. J. B. NEWMAN, in "Devel opment s In Concr et e Techno-
logy-I", edi t ed by F. D. Lydon (Appl i ed Science, 1979).
13. Idem, Appl. Sci. 5 (1978) 151.
14. M. F . ASHBY and S. D. HALLAM, Acta Metall. 34 (1986)
15. C. G. SAMMI S and M. F. ASHBY, ibid. 34 (1986) 511.
16. G. VEKI NI S, M. F. ASHBY and P. W. R. BEAUMONT,
Cambr i dge Uni versi t y Engi neeri ng Depar t ment Repor t s
CUED/ C- MATS/ TR 148 (1989), CUED/ C- MATS/ TR 174
( Mar ch 1990),
17. A. DI NSDALE, "Pot t er y Science: Mat eri al s, processes and
pr oduct s", ( Hor wood, London, 1986).
18. F. SI NGER and S. S. SI NGER, "I ndust r i al Cerami cs" (Chap-
man and Hall, London, 1962).
19. J . E. STRAWLEY and B. GROSS, in "Cr acks and Fr act ur e",
Amer i can Society for Test i ng and Mat eri al s Special Techni cal
Publ i cat i on 601 (ASTM, Phi l adel phi a, PA, 1976) pp. 559 79.
20. M. L. KACHANOV, Mech. Mater. 1 (1982) 19.
21. M. F . ASHBY and C. G. SAMMI S, PAGEOPH 133 (1990)
22. R. J . ROARK and W. C. YOUNG, "For mul as for stress and
st rai n", 5th Edn (McGraw-Hi l l , New York, 1975).
Received 7 January
and accepted 4 February 1992