October 1965 Effect 01Yolyaxial Stress States on Failure Strength of Alumina Ceramics 519

E. A. Maguire and J. J. Green, “Magnetic Properties of M. A. Gilleo and S. Geller, “Magnetic and Crystallographic
Yttrium-Gadolinium-Aluminum-Iron Garnets,” J . Appl. Phys., Properties of Substituted Yttrium-Iron Garnct, 3YL03.xM203 .-
33 [3, SUppl1 1373-74 (1962). (56- x)Fez03,”Phys. Rev.,110 [l]73-78 (1958).
J. 0. Artman and P. E. Tannenwald, “Measurement of
G. R IIarrison and L. R. Hodges, Jr., “Temperature Stable Susceptibility Tensor in Ferrites,” J . Appl. Phys., 26 [9] 1124-32
Microwavc Hybrid Garncts,” ibid., 1375-76. (1955).

Effect of Polyaxial Stress States on Failure Strength

of Alumina Ceramics
by L. J. B R O U T M A N a n d R . H. C O R N I S H
IIT Research Institute, Chicago, Illinois

The effect of polyaxial stress fields on the brittle Table I. Specimen Design and Loading Conditions
fracture strength of polycrystalline alumina was for Various Stress States
investigated through the use of thin-walled Specimen
cylinders. Combinations of internal pressure, length*
Stress state Type of Loading (in.)
external pressure, and axial loads produced
stress states of tension-compression, tension- Uniaxial tension Internal pressure
tension, and compression-compression. The open-end cylinder
Biaxial tension
failure envelope was generated for these stress (2:I) Internal pressure
states. The results indicated that biaxial tensile closed-end cylinder
stresses reduced the strength of the material; (4:3) Internal and external
however, the tensile strength increased at least pressure closed-
end cylinder
50% when a compressive stress existed normal Uniaxial com- End-loaded cylinder
to the tensile direction. Compression strengths pression
as high as 640,000 psi were measured for a Biaxial com- External pressure
biaxial compressive stress state. pression (2: 1) closed-end cylinder
Compression- Internal pressure
tension compressive end
load
I. Introduction * Cylinder dimensions were: OD, 1.8000 in., and wall thickness,
and theories for brittle materials i n polyaxial states
D ATA
of stress are vitally important. T o conceive of an
operational structure in which a simple state of stress can be
0.100 =I=0.004 in.

postulated, let alone observed in practice, is almost impossible.

This investigation was designed t o measure the effects of
polyaxial stresses on t h e fracture behavior of polycrystalline
alumina. * Experimental fracture d a t a were taken for several specimen. All the cylinders had the same diameter and wall
ratios of biaxiality, and a n effort was made t o correlate these thickness but the cylinder length was varied according
d a t a with existing theory. t o t h e stress state t o achieve uniform stresses and t o minimize
A thin-walled cylinder was chosen for experimental deter- end effects. Table I summarizes the geometries used.
mination of t h e failure strength envelope for alumina. Every T h e selection of a single-specimen geometry avoided t h e size
combination of stress state could be created in this single and stress gradient effects which could arise in comparing
results from specimens of different basic geometries, surface
properties, and residual stresses. The envelope of failure
stresses was developed, one quadrant at a time, using dif-
Presented a t the Sixty-Seventh Annual Meeting, The American ferent test equipment or methods when necessary, b u t in all
Ceramic Society, Philadelphia, Pa., May 4, 1965 (Basic Science
Division, No. 31-B-65). Received November 5, 1964; revised experiments specimens as nearly identical as possible were
copy received March 24, 1965. used.
This research was done in the Solid Mechanics Division of the
I I T Research Institute under the United States Air Force 11. Experimental Procedures
Contract No. AF 33(657)-10697 and was under the direct
supervision of the Ceramics and Graphite Branch, Air Force (I) Uniaxial Tensile Strength
Materials Laboratory, Research and Technology Division, Air
Force Systems Command, Wright-Patterson Air Force Base, Accurate measurement of the tensile strength was necessary
Ohio. t o establish a n important point on the failure envelope which
The writers are, respectively, senior research engineer and served as a reference for other strength measurements.
assistant dircctor, Solid Mechanics Division, IIT Research T h e method chosen for measuring tensile strength and the
Institute.
* A1995, manufactured by Western Gold and Platinum Co., equipment developed was similar t o that described b y Sed-
Belmont, Calif. lacek and Halden.’ A tensile stress was created by internal
520 Journal of The American Ceramic Society-Broutman and Cornish Vol. 48, No. 10
(A1 SIMPLE TENSION

Pump
J

(61 SIMPLE COMPRESSION

no*--

(C) BIAXIAL TENSION - TENSION (2:l)

0-Ring Seals

Bonded End Plugs

no Test Specimen

Fig. 2. Schematic diagram of equipment for biaxial tension

(0) ElAXlAL TENSION - TENSION (4:3) experiments.

Biaxial compression strengths were determined by using

(E) ElAXlAL COMPRESSION - COMPRESSION (2:l) closed-end cylinders placed in an external hydrostatic pres-
sure environment. The maximum hoop compression stress
created at the inside surface of the cylinder wall was twice
the longitudinal compressive stress. The radial compressive
stress in the wall was equal to the external pressure at the
(Fl ElAXlAL TENSION - COMPRESSION outside surface and decreased to zero at the inside surface.
A 150,000-psi pressure vessel with a 2-in. bore was used to
produce the external pressures required for failure.

( 3 ) BiaxiaZ TensiZe Strength

Internal pressurization of a hollow cylinder with closed
Fig. 1. Uniaxial and biaxial stress specimens. P = hydro-
static pressure, Q = direct axial load, U T ~= tensile stress, ends results in a 2: 1 (hoop:longitudinal) tensile stress state
uc = compressive stress, ur = radial stress, Pint = internal in the wall of the cylinder. Aluminum plugs were ceinentcd
pressure, and P,,t = external pressure. Values of specimen into the ends of the alumina cylinders. The length of the
diameter and wall thickness were selected so that ul is small plug needed to contain the pressure was calculated by assum-
compared to ut.
ing a value for the shear strength of the adhesive bond.
End plugs of three lengths (ll/d, l l / z , and 2 in.) were used
in the investigation. This was done to determine the effect
pressurization of an open-end thin-walled cylinder or ring of the end plug on the failure strength of the cylinder. The
(Fig. l ( A ) ) . A rubber bulb (neoprene), which conformed to cylinder and plugs were assembled by coating the plugs with
the internal dimensions of the test specimen, was used as an epoxy adhesive and inserting the plug into the cylinder
the pressure diaphragm. The pressure was generated with a allowing the excess adhesive to be squeezed out to the top
10,000 psi hand pump and was recorded by a 7500 psi Heise of the plug.
(graduations of 10 psi) pressure gage. No axial force or To produce tensile stress states other than 2 : l in ductile
restraint was applied directly to the specimen; therefore a materials, cylinders can be simultaneously end loaded and
pure state of tension existed throughout the specimen. A internally pressurized. In a brittle material such as alumina,
more complete dcscription of the tensile test method has however, the perfect alignment required for end loading
been given elsewhere (especially in tension) renders this method untenable. I t
was decided that in the present study tensile stress states
( 2 ) Compression Strength could best be generated by combinations of internal and
The uniaxial compressive strength of brittle materials is external pressures as illustrated in Fig 1. A schematic
difficult t o measure accurately. Specimen ends must be diagram of the arrangement used to produce one of these
flat and parallel and the friction force occurring a t the stress states (4:3) is shown in Fig 2 and the assembled test
specimen ends must be reduced to a minimum. In addition, specimen with the external pressure vessel is shown in Fig.
the high compressive strength of alumina ceramics requires 3. A 10,000 psi hand pump and a set of hydraulic jacks
end bearing plates of a harder material, such as tungsten placed between the platens of a hydraulic test machine were
carbide, to prevent indentation of the test specimen into the arranged to provide internal pressure P and external pressure
end plates. P / 2 . If the external pressure acted only around the cir-
Ryshkewitch3 investigated the compressive strengths of cumference of the cylinder, a 1:1 stress ratio would be
alu-mina ceramics being particularly careful to eliminate achieved. The alumina cylinder, however, is smaller in
premature failures from end restraiFt. He used a 1/4-in. diameter than the external pressure vessel so that the ex-
cube with ends ground parallel and polished. The loading ternal pressure exerted a force on the rim of the end plugs
platens were truncated pyramids of tungsten carbide and the producing an additional longitudinal tension in the wall.
specimen ends were lubricated with graphite and oil. The The actual stress ratio in the wall was more nearly 4:3
uniaxial compressive strength was 450,000 psi which is one of (longitudinal tension to hoop tension).
the highest values obtained for this type of test. The
strengths of the uniaxial compression specimens used in this (4) Tension-Compression Strength
investigation were lower than those obtained by Ryshke- The uniaxial tensile apparatus was combined with a
witch because of the difference in test specimen geometry. 120,000-lb hydraulic testing machine to produce a com-
October 1965 Efect of Polyaxial Stress States on Failure Strength of Alumina Ceramics 52 1
Table 11. Uniaxial Tensile Strength Results
Number Specimen Average Standard Coeficient
of length tensile deviation variation
of
specimens (in.) strength* (psi) (psi)

13 '/Z 28,798 2218 7.7

13 1 29,710 3180t 10.7
13 1 29,888 2271 7.6
* Loading rate is approx 2500 psi/sec.
t If the two lowest values in this series are dropped the average
for 11 specimens is 30,961 psi and the standard deviation is 1021
psi resulting in a coeficient of variation of 3.3%.

3600
I in
3 2 0 0 r otin

Fig. 3. External pressure vessel and closed-end alumina

cylinder used for biaxial tensile stress experiments.
2800

pression-tension stress s h t c in the cylindcr wall. The bottom

platen of the tensile apparatus rested on the machine bed
and the upper platen was fastened under the moving cross-
head. The hydraulic testing machine was used to apply end
-
v)
a
2400-

g 2000--
2
v)

m
-
J
loads to the cylindrical specimen. A constant stress ratio
loading path was not used as in the other combined stress
expcriments. Generally, two types of loading paths were 1200-
used: (1) An initial compressive end load was applied
with the testing machine followed by pressurizing the cylinder
to failure, and ('2) internal pressurization oi the cylinder was
followed by end loading to failure. During the investigation
incremental loadings were also used. The load sequence
was: internal pressure, end load, raise internal pressure, and
then end load to failurc. The results are given in Section 111. I
80 160 240 320 400
111. Results S t r o i n , p in./in.

( I ) Uniaxial Tensile Strength Measurements Fig. 4 . Uniaxial stress-strain data for alumina cylinders.
The tensile strengths were evaluated for three cylinder
lengths t o determine whether a size effect existed within
the range of gage lengths used in the investigation. The
strengths are summarized in Table I1 and are independent Table 111. Uniaxial Compression Strength Results
of speciincn length over the range l/z to 2 in. The approx-
Specimen Compressive Parallelism
iinatc relation between the tensile stress and the internal length Load strength of ends
prcssure for a thin-walled cylinder is (in.) (1b) (psi) (in.)

1 177,500 326,300 0.0002

u = P?/t (1) 1 102.500 188,400 .0002
u = stress in the wall 1 130;000 239,000 ,0003
P = pressure. 1 113,500 208,600 ,0002
B = mean radius. 1 95,000 174,600 ,0006
t = wnll thickness. 1 101,500 186,600 .0003
Avg. 220,580
Laink's equation for thickwalled cylinders is :
1/9 116.800 214.700 ,0001
li, 111:400 204 800 ,0001
;
~~~

1); 124 000 227; 900

'/2 101,000 185,700 ,0001
u = stress at any radius. '/z 94,000 172,800 .0001
ri = inner radius. '/Z 149,000 273,900 .0001
ro = outer radius. 1/9 123.000 226,100 .0001
r = radius. AGg. 215; 130

Equation (1) was used to calculate tensile stresses in the

cylinders (?/t = 8.5) as the difference between the two
equations is <0.5%,.
Stress-strain measurements were made on various cylinders passes through zero which indicates that the effect of pres-
to veriry the accuracy and reproducibility ol the test method surizing the rubber bulb inside the specimen was negligible.
and to further prove that no restraints were imposed on the
specimen by the test equipment (i.e. the ends were free to ( 2 ) Compression Strength Measwements
expand). Figure 4 shows pressure versus strain (l/& The uniaxial compressive strengths of the cylinders were
strain gages were mounted on the outside wall) for cylinders measured in a 300,000-lb capacity Olsen hydraulic testing
of two lengths. A straight line can be drawn through both machine, No significant difference was noted between the
sets of points and the calculated modulus based on Lame's compressive strength of cylinders and 1 in. long (Table
equation was 55 X 106 psi. The stress-strain curve also 111). On failure, the cylinders generally fragmented into
522 Journal o j The American Ceramic Society-Broutman and Cornish Vol. 48, No. 10
Table IV. Biaxial Compression Results a
u axial(10 PSI I
Specinien External Hoop, Axial
length pressur-e compressive compressive
(ill.) (psi) stress (psi) stress (psi)
~ ~ __
2 57,500 550,000 275 ,000*
2 55, 500 530,000 265, 000t 100
2 67,000 640,000 320,000$
2 63,600 606,000 303,000 $
I 66,000 630,000 315,000$ 2 00
Average Uniaxial Vaiue
* Steel bcaring plates.
t Aluriiiiia disk on steel used for bearing plate. 300
$ Tungstcn carbide bearing plate.
Highest Uniaxial Volue

400
Ryshkewitch Value

500
many small pieces. The difficulty of measuring the true
compressive strength of a brittle solid has already been
discussed. Since the biaxial compression and combined 600
tension experiments were also performed with straight hollow u (dpsi)
hoop
cylinders, it was necessary to use the same type of specimens
Tor uniaxial comprcssion measurements so that a basis for Fig. 5 . Experimental compression-compression quadrant of
comparison existed. failure envelope.
Strength results for a 2: 1 biaxial compressive stress state
obtained by external pressure are summarized in Table IV.
IXoop stresses and axial stresses were computed from the
following equations : Table V. Biaxial Tension Results
Snec-
imen Internal Hoop Axial End plug
(3) Stress length pressure tension tension length
ratio (in.) (psi) (psi) (psi) (in.)

2: 1 6 2920 24,800 12,400 2

2: 1 6 2840 24,100 12,050 2
u0 = hoop stress 2: 1 6 2600 22,100 11,050 2
2: 1 4 2720 23,100 11,550 1.25
6, = axial stress. 2: 1 4 2220 18,900 9,450 1.25
I-' = external pressurc. 2: 1 4 4350 31,000 18,500 1.25
2: 1 4 3400 28,300 14,450 1.25
The last three results in Table I V (tungsten carbide end 4:3* 6 6080 22,800 30,400 1.5
plates) are very close and show no difference between cylinders 4:3 6 4200 15,800 21,000 1.6
1 and 2 in long. The results are plotted in Fig. 5 and 4:3 4 4870 18,300 24,400 1.5
represent the compression-compression quadrant of the com- 4:3 4 5230 19,600 26,200 1.5
4:3 4 5050 19,000 25,200 1.5
bined stress failure envelope The dashed line in the figure 4:3 4 4350 16,300 21,800 1.5
was drawn through the experimental points. The biaxial 4:3 4 4480 16,800 22,400 1.5
compressive strengths were in some cases three times greater * External pressure also used in this experiment ( P e:t, = J'intP)
than the uniaxial compressive strengths of the same cylinder.
The uniaxial strength measured by Kyshkewitch3 was much
greater than the present measured uniaxial strengths but is
still much less than the biaxial compression strengths.
( 3 ) Biuxiul Tension Experiments A 6" Specimens
The results for the biaxial tension experiments are presented

40t Y
4:3 Load Path
0 4" Specimens
in Table V. Two stress ratios investigated were 2 : l and
4" Specimens Failed
4:3. The results, shown in Table V and plotted in Fig. 6, at Center
represent the tension-tension quadrant of the failure envelope.
The stresses in the wall for the 2: 1 stress state were calculated
using Lamb's equations. The hoop stress is given by equation
(7) or equation (2) and the axial stress is given by

The stresses in the wall lor the 4:3 stress state are calculated
by superimposing the stresses from the internal and external
pressure. For example the hoop stress is Stress ( i o ' p s i )

Fig. 6. Experimental tension-tension quadrant of failure

envelope. (Hatched area represents maximum principal
stress theory and width of hatched area represents standard
and the axial stress can be calculated from: deviation of uniaxial tensile strength.)

R = inner radius of external pressure vessel

The hatched area in Fig. 6 represents the maximum
The second term of equation (7) accounts for the tension principal stress theory and also the variability in tensile
exerted by the external pressure acting over the rim of the strength; the width of the area represents the standard
end plugs. deviation for the tensile strength of the cylinders. The
October 196.5 €ifat of Polyaxial Stress States on Failure Strength of Alumina Ceramics 523

Fig. 7. Biaxial tension cylinders, 4 in. long (4:3 stress ratio). Fig. 9. Polycrystalline alumina ring which failed in com-
( A ) Failure near an end plug, and ( B ) failure at center of gage pression-tension stress state.
length.

quadrant of the failure envelope. The hatched band shown

in the figure intercepts the tensile strength on one axis and
the compression strength on the other axis. The width of
the band represents the standard deviation of the strengths.
The band has been drawn to represent both the Grifhth
failure4 theory and the Coulomb-Mohr theory.G The points
shown in the figure represent two specimen lengths and two
types of loading paths. In general, the results indicate that
Loading Paths
the tensile strength increases as the compression stress in the
orthogonal direction increases. Tensile strengths as high
0 End load then pressurize to
foilure (I,* lenglh ) as 50,000 psi were observed for specimens with compressive
o End b o d lhen pressurize lo stresses of over 100,000 psi in the orthogonal direction. In
+"
faibre i length)
one case a tensile strength of 62,000 psi was observed for a
A Internal pressurize first ihen
end load to foilure(l"leng1h)
1/2-in. specimen. These specimens were first end loaded
0 Internal pressurize first then
to a desired level and then pressurized to failure. Low tensile
end load (4'' length) strengths were expected because of the end restraint resulting
A End load reoched machine from the high axial compressive loads. The arrows (Fig. 8)
1 copoc~ly; specimen did not foil indicate that the specimens had not failed before the load
A Torsion test ( r e f . 6 1 capacity of the testing machine was reached.
Figure 9 shows a specimen which failed by compression-
Pig. 8. Experimental compression-tension quadrant of tension stresses. This specimen resembled those which
failure envelope. failed by uniaxial hoop tensile stresses and its appearance
indicates that the compressive alignment was good. The
platens of the test equipment were always prealigned by
loading them with two parallel bars placed between the
averages o f the results for the 2 : 1 and 4:3 stress ratios fall platens.
short of this band; the dashed line in Fig. 6 passes through To further substantiate the effect observed in the com-
the average ol the results From this investigation. pression-tension quadrant, an increase in tensile strength with
The fractures usually occurred near one of the end plugs as increasing orthogonal compression, the result of a torsion
shown in Fig. 7 ( A ) . Figure 7 ( B ) ,however, shows a specimen test by Orowan6 is included in Fig. 8. The torsion strength
which failed in the center of the gage length and yet had a was measured on the same type of alumina (Wesgo A1 995)
strength very similar to those which failed near an end plug used in this study and the test conditions (strain rate, tem-
(Fig 6). This indicated that the stresses were uniform along perature, and atmosphere) were almost identical, Figure 8
the gage length and that the specimens which failed near an shows that the torsion strength (I : 1 tension-compression)
end plug were still valid. The bending stresses near the end is in very good agreement with the direct compression-tension
plugs for the specimens with the 4:3 stress ratio should be data obtained in this study.
less than those for thc 2: 1 specimens since the combination of
external and internal pressures will provide restraint from IV. Discussion
bending These results indicate that the biaxial tensile When comparing the experimental results obtained in this
strengths are indeed less than those predicted by the maxi- investigation to the available theories, it was assumed that a
muin principal stress theory. biaxial stress state existed and that the stresses were uniform
throughout the cylinder walls. As shown in Fig. 10, how-
( 4 ) Compression- Tension Measurements ever, internal or external pressure produces a triaxial stress
The results for the compression-tension experiments are state. Furthermore, the circumferential or hoop stress
presented in Fig. 8 which represents the compression-tension ag and the radial stress ur vary through the wall thickness.
524 Journal of The American Ceramic Society-Ijroutman and Cosnish Vol. 48, No. 10
-Griffilh Theory

I /
_ - - -Weibull Theory

P
+qJp33 uhoop

I *R

(11 Internal Pressure b) External Pressure

Fig. 10. Stress distribution in cylinder walls as a result of

c) Internal and External Pressure
.”
combined loading.

Figure 1O(a) shows tlie stress distribution through the wall

for a closcd-end cylindcr which is internally pressurized. Fig. 11. Plots of data for experimental and proposed failure
Thc stress distritiution is the same for the open-end cylinder envelopes.
except that u Z = 0. The hoop stress decreases from 8.5P
(P = prcssurc) a t tlic inside wall to 7.5P a t the outer wall,
indicating that failure will initiate from the inside wall unless
a miicli more scvcre flaw is present a t the outside wall. With
external pressure (Fig. 1 O(h))211 stresses are compressive and
again the maximum compressive hoop stress occurs a t the tensile strength. The measured increase in strength is not
inncr wall and thc radial compression is zero a t this point. as great as predicted by the Griffith theory.
The hoop stress varies from - 8.5P a t the inner wall to - 8.5P The experimental results shown in Fig. 11 definitely cor-
at the outer wall. Therefore it is liltely that failure will relate better with the Weibull theory (developed for hydro-
initiate from the inner wall ol‘ this cylinder also. A com- stone plaster) than with the Griffith theory. The increase in
bination o f inicrnal and external pressure producing a stress tensile strength with increasing compression applied in the
ratio o f 4 : 3 is shown in Fig. lO(c); the axial stress is the orthogonal direction is even greater than that predicted by
maximum principal stress and is uniform throughout the wall. the Weibull theory. The Weibull curve, shown as a dashed
Thc hoop tensile stress varies From 3.75Pi,,*a t the inner wall line in Fig. 11, was originally devcloped for a hydrostone
to 3.25Pin6 a t the outer wall. The eflect of the pressure plaster and among other things is dependent on the Weibull
combination is to reduce the stress gradient in the wall. theory parameter, m. In addition, an assumption of the
The inaxiinuin stresses, however, again occur a t tlie inside Weibull theory is that the material has an infinite compressive
wall indicating that failure is probably initiated here. strength. A cutoff has been indicated (Fig. 11) to represent
Although many theories are postulated to predict failure the finite compressive strength of the material. The data in
strengths for combined stresses, those most appropriate for the tension-tension quadrant also correspond more closely
brittle ceramics such as alumina are the Griffith theory4 and to the Weibull theory than to the Griffith or maximum prin-
the Weibull theory.7 The various theories have been sum- cipal stress theories.
marized by other investigators.8*o The Weibull theory for Acknowledgments
hiaxial stresses was applied to hydrostone plaster by Salmassy The support from the Rcsearch and Technology Division of
et d 7and the theoretical curve is shown in Fig. 11. This the United States Air Force, with L. L. Fehrenbacher as technical
theory predicts that the tensile strength increases as the monitor, is gratefully acknowledged. The program at IIT
compression stress in the normal direction increases, and also Rcsearch Institute was under the overall direction of N. M.
Parikh and W. Rostokcr and their assistance and guidance is
that the hiaxial tensile strength is lower than the uniaxial also acknowledged.
tensile strength.
The Griffith theory is also shown in Fig. 1I and can be References
expressed by the following equations : R. Sedlacek and R. A. Halden, “Method for Tensile Testing
of Brittle Materials,” Rev. Scz. Instr., 33, 298-300 (1962).
u1 = u2 = u0 for - 3 <a2
~ ~<al (8) R. H. Cornish and L. J. Broutman; pp. 291-325 in Studies
of the Brittle Behavior of Ceramic Materials by N. M. Parikh.
and Technical Documentary Report N o . ASD-TR-61-628, Part 111;
Contract No. AF33(657)-10697; 409 pp., June 1964.
(UI - ~ 2 + 8u,,(ul +
) ~ u2) = 0 for US <-3ul (9) Eugene Ryshkewitch, Oxide Ceramics; pp. 135-42. Ac-
ademic Press, New York, N. Y., 1960.
uo = uniaxial tensilc strength A. A. Griffith, pp. 55-63 in Proceedings First lnternational
ul,u2 = principle stresses. Congress of Applied Mechanics, 1924.
Burton Paul, “Modification of the Coulomb-Mohr Theory of
Tlic experimental results for the tension-tension and Fracture,” J . Appl. Mech., 28 [2] 259-68 (1961).
compression-tension quadrants are shown, along with the E. Orowan; pp. 177-213 in Ref. 2.
superimposed curves for the Griffith and Weibull theories 0. K. Salniassy, W. H. Duckworth, and A. D. Schwope,
(Fig. 11). Note that the ratio of uniaxial compressive “Behavior of Brittle-State Materials,” Wright Air Development
Center, WADC-TR-53-50, Part I ; Contract No. AF33 (038)-
strength to tensile strength is 7.5 compared to the ratio of 8 8682; pp. 49-74; June 1955.
predicted by Lhe Griffith theory. Although the compression- R. E. Ely, “Strength For Graphite and Two Types of Plastic
compression results are not shown in Fig. 11, as seen in Tubing Under Biaxial Stresses,” N A S A Doc., N63-22934, 34 pp.
Fig. 5, the biaxial (stress ratio 2: 1) compressive strength is (1963).
Joseph Marin, Mechanical Behavior of Engineering Ma-
approximately 22 times greater than the tensile strength. terials; pp. 116-77. Prenticc-Hall Inc., Englewood Cliffs,
From equation (9), however, this value should be 48 times the N. J , , 1962.