Vous êtes sur la page 1sur 8

A Bar Impact Tester for Dynamic Fracture Testing of

Ceramics and Ceramic Composites

by L.R. Deobald and A.S. Kobayashi

ABSTRACT--A bar impact test was developed to study the objective of this paper is to describe such a dynamic
dynamic fracture responses of precracked ceramic bars, fracture-test procedure involving a bar-impact test.
AI208 and 15/29-percent volume SiCw/AI208. Crack-opening
displacement was measured with a laser-interferometric dis-
placement gage and was used to determine the crack velocity Experimental Procedure
and the dynamic stress-intensity factor, K~y". The crack
velocity and Ktay" increased with increasing impact velocity The bar-impact experiment consists of a 50.8-mm long,
while the dynamic-initiation fracture toughness, Kx.,, did not rectangular bar specimen which is impacted on its end by
vary consistently with increasing impact velocities. a 25.4-mm-long bar impactor of the same material. Ideally
the reflected tension wave from the free end of the speci-
men bar interacts with the incoming compressive wave and
Introduction generates a tensile stress pulse of 5.2-#s duration in the
Available configurations of new ceramics and ceramic middle of the bar specimen with the transit of approxi-
composites are often restricted to bar geometries of about mately 3/4 of the compressive pulse as shown by the
10 x 10 mm in cross section due to the historical pre- Lagrangian diagram of Fig. 1. In practice, the impact
cedence of using modulus of rupture (MOR) tests for typically produces a compressive stress wave of 1.6-#s
determining the mechanical properties of early experimental rise time propagating in both the impactor and the
materials. These MOR specimens are ideally suited for specimen. The superposition of the incoming ramping
single-edge notched, three-point-bend tests in determining portion of the compressive stress waves causes a tension
fracture toughness despite the inherent experimental pulse, which rises in about 0.8 #s and is sustained for
difficulties associated with the small dimensions. These another 3.6/~s, at the crack plane.
small three-point-bend fracture specimens are also The schematic of the experimental apparatus is shown
extremely sensitive to the interactions of reflected stress in Fig. 2. The specimen and impactor bars are held by
waves with the propagating crack tip as evidenced by the molded urethane holders which are mounted in two
copious fluctuations in the dynamic stress-intensity factor carriages. Both the specimen and the impactor carriage
during the dynamic fracture process. 1.2 Moderate fluctua-
tions in dynamic stress-intensity factor in relatively large
three-point-bend specimens of 75-mm height x 10-mm
thickness and 89-mm height x 9.5-mm thick are docu- Vo ~ Midplane Stress History
mented in Refs. 3 and 4 respectively, but not to the $Feeimen J -EV.._.~o +EVo
extent shown in Refs. 1 and 2. Such superfluous reflected 25.4 0 75.4 50.8~ 2CI b I 0 2C
: b
stress waves do not exist in the one-dimensional split-
Hopkinson tensile-bar test of Duffy and his colleagues. 5.6 J
However, this clean test configuration uses explosives for . iI 1.6 p,s
loading which requires special training and laboratory
facilities. Thus a need exists for a similarly clean test
which can be executed without the special facilities. The
t=8. i

L.R. Deobald is Sen&r Specialist Engineer, The Boeing Company,


Defense and Space Group, Seattle, WA 98124. A.S. Kobayashi (SEM
Fellow) is Professor, University of Washington, Department of Mechanical
o y 1
Engineering, FU-IO, Seattle, WA 91895. N
Paper was presented at the 1992 SEM Spring Conference on Experimental
Mechanics held in Las Vegas on June 8-11.
Original manuscript submitted." August 1, 1991. Final manuscript received: Fig. 1--kagrangian diagram for impact experiment, C6 =
February 26, 1992. 9.8 mm/~s

Experimental Mechanics 9109


run on guide rails. An air gun propels the impactor LIDG signal. The measured impact velocity was used to
carriage down the guide rails towards the stationary determine the level of sustained tension stress after the
specimen carriage. The ceramic bars impact well before initial stress rise had been measured.
the collision of the urethane holders. The resultant com-
pressive stress is proportional to the impact velocity
which is determined by a laser-velocity measurement Data-reduction Procedure
system.
The specimen geometry is shown in Fig. 3. The ceramic
Theoretical Preliminaries
materials tested and the corresponding nominal mechanical
properties are shown in Table 1. A sharp precrack with
an approximate initial length of 3 mm was generated by SEMI4NFINITE CRACK IN ANINFINITE DOMAIN
the single-edge-precracked-beam (SEPB) method 7 from a The dynamic-initiation fracture toughness, K~,, was
shallow 6hevron notch. This sharp crack is transparent to determined by using the solution to a stress wave impinging
the incoming compressive wave but reflects the tensile on a stationary crack. 9 The dynamic-initiation stress-
wave which will initiate dynamic crack propagation. The intensity factor, K , , , of a stationary semi-infinite crack in
crack-opening displacement (COD) at the location X was an infinite body and subjected to a square tension pulse
measured using the laser-interferometric displacement is given as 1~
gage (LIDG) system. 8 LIDG targets were mounted at the
precrack tip such that the reflective indentations, 0.4-mm Kz~(t,O)- _o C,(1-2v)),/2tt,,
_( . 2 O 2
(1)
apart, straddled the precrack as shown in Fig. 3. Also a
short strain gage of 1.6-mm gage length was mounted at
the midlength of the bar on the surface opposite to the where tj = time of fracture. For ramp type of tension
precrack. This strain gage was used to determine the pulse as shown in Fig. l, K~, can be determined by a
initial rise in the stress history and its relation to the simple superposition of the discrete stress values as 9

GuideR a i l ~ TopView ~-lmpa~or Carriage


AirGun
~ SpedmenCarriage

Z
~ ~ Absorber

I ~ ~ / - Ram Specimen Fig. 2 - - I m p a c t apparatus and


/ Impactor~ / air gun

ill \
CrossSectionProfileView

PhotornudtiplierTube , .
w/MagneticShield Filter,x632.8nm / L i g h t ProofCov~

LaserIntefferen~_~N.~ / J ./~r B/~deSli ~-----~W =7.24mm

...... \\1 I V ' x "~=~'§ q.~o= 6.38~ l ~ - -


Fig. 3 - - S p e c i m e n geometry
J I ---vo and LIDG technique

9 , ~ J [Opening
Crack ~ Precutin Foil Impact~/
Drawm8 Isnot toscale! .... ~ MicroHardnessIndents
///" ReflectiveFoil

50.8mm P" I-- 25Amm

110 9 June 1992


CI(1 - 2 v ) ) , n
/2 The bar configuration with a rectangular cross section
2 ]~ Ao, At, m (2) used in this study does accordingly violate the plane-strain
K " - v(1- - ~ ) ( ~- i=l
condition postulated in eqs (1)-(6). Fortunately the small
Poisson's ratio of the ceramic matrix, i.e., v = 0.22,
Ah is the difference from the fracture initiation time and introduced deviations of the order of five percent at the
the arrival of the stress pulse, Aa~, as shown in Fig. 4. most between the C~, the bar-wave velocity, C~, and the
The dynamic stress-intensity factor, K~~", for a semi- plate-wave velocity, C,. For consistency, the dilatational-
infinite crack propagating at a constant velocity is related wave velocity is treated as the bar-wave velocity, C~, in
to the dynamic energy-release rate as 9 this study.
Ki"(t) = rc2~
Cv~(~l
lD(V) G,]m (3)
DERIVATION OF CALIBRATIONPARAMETERS,xit AND

where A nondimensional parameter, ~/, was derived by the


simple inversion of the dynamic displacement equation
(1 - ~) ,, I-E_ where the crack-opening displacement (COD) can be
C2 = x]~ C~ - ( 1 - 2 v ) ( l + v ) VQ (4) represented as ~2
COD - 4(1 + v)K/y" ~ = ~ - , ~ 1 ( - ~ 21) 2 ^
for plane strain. E ~/,~r/Tr ~ D ( V ) )f(K~,r)
D(V) = 4/~1/~2 - (1 + ~)2 (5) (7)

~, = 1- vvc~ /3~ = 1- V*/C~ (6) where r = a - X. f ( K , r) is defined in Ref. 12, a is the


crack length, and X is the distance along the crack at
E
which the COD is determined. The nondimensional
/~ - 2(1 + v)
parameter, 't,, was then defined from eq (7) above as
V is the crack velocity, C1 and (72 are the dilatational and Kay'L, f d z - X - f ( V )
, = (4f")-l,,/r~
shear-wave velocities, respectively, E and # are the = E ' COD (8)
modulus of elasticity and the shear modulus, respectively,
and ~ is the mass density. Equation (3) has been used where
extensively in finite-element analysis of dynamic fracture
/3, (1 -/322)
and its use has been validated in a previous paper." f(v) -
(1 - v)D(V)
and
TABLE 1--MECHANICAL PROPERTIES E
ii ii
1 -v 2
E 0 Cb
Materials (GPa) ~ (g/cc) (mm//~s) For the near-field asymptotic solution of a semi-infinite
crack propagating at constant velocity in an infinite
AhO3* 99.5 percent 372 .22 3.89 9.78 plate, ~ = 1 and the theoretical value of ~I' becomes
AhOat 99.9 percent--HPt 387 .22 3.97 9.87 xvth,o = x / ~ 3 2 . As will be shown later through dynamic
15,percent v SiC,,/AhOat--HP 398 .23 3.85 10.17 finite-element analysis, ~I, remains essentially constant
during the short crack-propagation time in this small
29-percent v SiC,/AhOat--HP 408 .23 3.73 10.46 specimen.
A second nondimensional parameter, fl, was also
Cb is the bar-wave velocity derived based on an empirical relationship between the
energy consumed at the crack tip and the strain energy
flux into the crack-tip region as follows. 12
*Coors Ceramic Company, 600 9th Street, Golden, CO 80401, USA
tAdvanced Composite Materials Corporation, 1525 S. Buncombe a2(t) C1
Road, Greer, SC 29651, USA G, = k ( a - x ) ~ W V (9)
r pressed processing
where W is the specimen width and ~ is a proportionality
constant. The COD then can be rewritten as '2

COD
=
~
4 ~ / 2 ( a - X ) / w ( (1~ ,-( 1 -v~ )) D ( V ) ) .

f~K1,r) [g(V) ~ / ~ ( a - X ) W ] a ( t ) (10)

Rearranging eq (10), the nondimensional parameter,


ft, is defined as
X E t COD
fl = 4 x f 2 ~ f f X,~f~/ W =
(a - X ) We( t) f ( V)g( V)
[~ At, = b- (t~+ t 0 / 2
(11)
Fig. 4--K~d determination by stress-pulse
superposition where

Experimental Mechanics . 111


(C,C~D(V) ),,2 (12)
produced a sudden net tension stress. This tension pulse
g(V) = x,2(l -1- /))V3~i was modeled by the tension wave which reflected off the
rigid boundary along the crack plane as a tension wave in
the one-half finite-element model. This reflected tension
wave in the finite-element-method (FEM) model repre-
sented the tension wave passing from the back half of the
Dynamic Finite-element Analysis bar in a real test. The crack was then allowed to open at
the time the crack plane was in tension as predicted by
IMPACTED SEPB SPECIMEN Fig. 1. After a delay of about 0.6 /zs, which was deter-
mined from the COD measurement as will be described
A commercial finite-element code, ABAQUS*, was
later, a constant crack velocity was then prescribed and
used to simulate the dynamic fracture of the impacted
the energy release rate, G,, was calculated from the
ceramic bar. The purpose of this study was to determine
change in the total energy 1' in the model as the crack
the calibration parameters, ~I, and r , for the specimen
propagated.
configuration considered. Although the impact phen-
The FEM analysis justified the above use of the
omenon was not symmetric, the stress state and the stress-
theoretical solution for a semi-infinite crack which is
wave propagation becomes completely symmetric after the
either stationary or moving at a constant velocity and
stress wave has propagated about the length of the bar
impacted by a one-dimensional wave, in an infinite
and thus only half of the bar was modeled. Stress-wave
domain despite the finite geometry of the specimen. This
propagation was simulated in the finite-element model by
analysis consisted of prescribing the impulse (O,om~,,~ =
ramping the initial static compressive stress at the impact
100 MPa) of Fig. 1 on the impacted end of a FEM of the
end to zero over a 1.6-/~s duration. This loading created
a tension pulse which propagated towards the center of
whole specimen bar, without a precrack, and computing
the dynamic state of stress at the crack plane. Figure 5
the bar. As shown in Fig. 1, the two tension pulses, which
shows the wave front of the reflected tensile pulse, a,,
met at the crack plane at a post impact time of 8.6 t~s,
through the width of the specimen prior to impacting the
crack surface. Also shown is the secondary stress, a,,
*Hibbit, Karlsson and Sorensen, Inc., Providence, R t 02906-4402 USA.
which is perpendicular to the tensile pulse, at this instant.
The negligible distribution of the secondary stress, a,,
shows that a one-dimensional wave front is maintained at
this instant. Figure 6 shows the computed two stress
components at the strain gage (see Fig. 3), which is at the
140
midpoint of the specimen edge in front of the crack tip.
120 J The width-wise stress, a,, which should be zero, is
negligible in the plane-strain FEM analysis and again
100
justifies the use of the semi-infinite crack solution in this
80: o, data-reduction procedure.
9--II- .-D- t=9.0 bts
Gx
60 "4- ---O- t=9.5 + ]
DETERMINATION O F xl/c A N D ~ C
40 ~ ,.../x- t = 1o.o ] Gy
--o-- ---O-- t = 10.5 The parameter, t , listed as eq (10), was calculated in a
20~ series of dynamic finite-element analyses for the W =
-a~ ---V-- t = 11.0
8.89-mm specimen. The crack velocities varied from 1201
v v to 2222 m/s. Two different initial crack lengths, ao, and
-20 four different crack velocities were simulated. The results
o "021 d2" o::;" o'.,/" o.';' o16 o'.:;" oi;" o29' of the analyses are shown in Fig. 7. Immediately after the
Fractional Distance Across Crack Plane crack began to propagate, the nondimensional parameter,
Fig. 5--Computed stresses at the crack plane

2
150 13 V=1201 m/s, ao=2.667mm
0 V=1502 m/s, ao=3.1115 mm
100 + V=1821 m/s, ao=3AllSmm
1"5 1

50 /0e~ # V=2222 m/s, ao=3.1115 mm

1 oooo_OO
L~ i:ii~
0
-50 0.5- [][]

-I00-
0 . . . . i . . . . i . . . . i . . . . J . . . . i . . . . i . . . .

-150 0.3 0.4 0.5 0.6 0,7 0.8 0.9


' ' ' i , , , i ' ' ' i . , ' L . . . . . . . . .
2 4 6 S 1; ll2 14 (~ = a / W
Post Impact Time (Its) Fig. 7--Finite-element values of ft as influenced by
Fig. 6--Computed stresses at strain-gage location crack-tip location, W = 8.89 mm

112 9 June 1992


t , oscillated down, and assumed a constant value. From The crack-length history, determined by eq (13), is a
Fig. 7, a constant fl = 0.95 was defined as the calibration function of two functions, f ( V ) and g(V). Hence, the
constant, tic. This fl will not model correctly the dynamic- crack-length history, a(t), must be determined 'in an
fracture phenomenon at crack initiation where the crack iterative fashion. An initial guess was made of the crack
velocity jumps from zero to a high constant velocity used velocity, F, to calculate f ( V ) and g(V). a(t) was then
in the simulation. A similar finite-element analysis of the calculated with eq (13) and least-square fitted with a
specimen geometry with W = 7.24 mm yielded a calibra- straight line. The slope (crack velocity) was used to deter-
tion constant of f~c = 1.37. mine new estimates of f ( V ) and g(V). The process was
The calibration constant, 'I'c, was determined from a repeated until the crack velocity converged to the correct
plot of the nondimensional parameter, ,t,, shown in Fig. value, which is the measured crack velocity in this case.
8. The parameter, ,I,, was calculated from the finite- The final value o f f ( V ) was used with a(t) in eq (14) to
element analyses for four different crack velocities. As calculate K f y" (t), explicitly.
before with t , the parameter ,t, oscillated immediately The calibration constant, t o , was then used to deter-
after the crack velocity jumped from zero to the constant mine the crack length, a(t), from the measured stress
velocity. 9 remained constant before dropping off slightly history and COD history. Before using the numerically
as the cracks passed c~ = a / W = 0.85. A constant value determined t o , the procedure was verified by comparing
line was fitted to the data yielding the calibration constant, the actual finite-element crack history with a(t) deter-
,I% = 0.34. This value compares well with ~I, derived from mined by the procedure presented above for t c = 0.95.
the theoretical solution for a constant velocity crack in an This verification is shown in Fig. 9 for four different
infinite medium where ~Itheo = x~-/32 9 A speculation is crack velocities. Surprisingly, the large initial oscillations
made that the coarse mesh causes the FEM model to be in the data caused little error in the predicted crack
overly stiff and underestimates the COD, thus resulting in length. The error increased for two of the crack histories
a numerically determined ,I% which is slightly higher than as the crack approached the specimen boundary.
~Z,heo. A single finite-element solution was used to deter- The rate of convergence of this iterative procedure was
mine ,I% for the FEM model with W = 7.24 mm. The tested by using the COD and stress histories from the
result for the 7.24-mm wide model was ,I% = 0.35 which dynamic FEM simulation (V = 1502 m / s ) with t c =
is within three percent of ,1% for the 8.89-mm wide model. 0.95 to predict the crack history, f ( V ) and g ( V ) were
Although many FEM analyses were conducted to verify calculated with an initial value of V = 600 m/s. Only
the data-reduction procedure, only one analysis was re- four iterations were required to obtain the crack velocity
quired per specimen geometry. The resulting calibration of 1520 m / s which is within 1.2 percent of the true value.
constants, t c and ~I'c, are given in Table 2 for the two Again the numerically determined calibration constant,
geometries used here. The solution basically behaves as a 'I'c, was used with the FEM results to verify the method.
constant velocity crack in an infinite medium. The stress The K~ay" predicted using the above procedure was com-
wave reflection from the lateral surfaces apparently had pared with the K~ayn from the FEM model in Fig. 10. The
little effect on the solution until the crack tip approached large initial oscillations had only a slight effect on K,dy"
the specimen boundary. immediately following crack initiation. The new procedure
The crack-length history, a(t), using the experimental
midplane stress, a(t), was computed by the crack-opening
displacement, COD (t). K,ay"(t) computation requires the
use of a(t) and COD (t). These relations are shown in the TABLE 2--CALIBRATION CONSTANTS f~c AND ~Zc,
following as functions of the two nondimensional parameters. CALCULATED BY FEM
E 'COD(t) V) ) (13)
a(t) = X (1 + flcWa(t)f(V)g( W (mm) X (mm) ftc ,t,c

K,aY"(t) = ~c E'COD(t) (14) 8.89 2.5 0.95 0.34


f(V) ~/a~ - X 7.24 1.37 0.35

1 9 oO
[] V=1201m/s, ao=2.667mm
0.9- []
O V=1502 m / s , ao=3.1115 m m
8i :../.. d+ d ~176_
0.8- 7 .~'4 .,-'+* o~#~
+ V=1821m/s, ao=3.1115 m m ,, ~C_ o~O~-
"~ 0.7-
[] ~. V=2222m/s, ao=3.1115 m m 6 eN:
~0.6-
rq - - Theoretical Value
" 0.5-
~O -- Nominal Numerical Value E
"~4-
~. 0.4-
3- 0.95 FEM V (m/s)
0.3- - ~ ~ O b 8 - [] 1201
2- 0 . . . . 1502
0.2-
+ . . . . 1821
0.1- 1- t 2222

0 . . . . i . . . . i . . . . i . . . . i . . . . i . . . . i . . . . 0 .... i .... i .... i .... ~ .... i .... i .... i .... i'''

0.3 0.4 0.5 0.6 0.7 0.8 0.9 3.5 4 4.5 5 5.5 6 6.5 7 7.5
o~ = a / W F E M t i m e (}as)

Fig. 8--Finite-element values of ~I, as influenced by Fig. 0--Finite-element verification of a ( t ) , W =


crack-tip locations, W = 8.89 mm 8.89 mm

E x p e r i m e n t a l M e c h a n i c s ~ 113
overestimated K/y" beyond ot = 0.85 due to the drop off bined stress history is shown as a shaded area in Fig. 11
shown in Fig. 8. Hence, the solution would be considered and was used to calculate the experimental a(t).
valid to about c~ = 0.85. A typical photomultiplier-tube (PMT) output repre-
senting the COD variation with crack propagation in a
specimen (15-percent volume SiC,/AI203) impacted at
Experimental Results 8 m / s is shown in Fig. 12. The crack began to run shortly
As mentioned previously, an accurate measurement of after the arrival of the tension stress wave.
the impact velocity is important since it is used to calculate The fringe-motion history shown in Fig. 12 was com-
the stress level. Simple bar-wave theory predicts a square bined with the other PMT output to obtain the COD
stress pulse of magnitude a, = E Iio/2 Cb while the actual history for this test shown in Fig. 13. The COD data were
stress wave ramps up and oscillates around the stress level curve fitted with a higher order polynomial which was
predicted by the bar-wave theory as shown in Fig. 6. A used to analyze the data at the same time increments as
typical stress pulse, which is the first compressive pulse, the stress history.
measured with a single strain gage is shown in Fig. 11.
Other than the magnitude of the stress level, which ob- Crack-length H/story
viously will depend on the impact velocity, remarkable
agreement between the measured stress of Fig. 11 and the The procedure presented previously was used to obtain
computed stress of Fig. 6 is noted. Such agreement shows the crack-length histories of the ceramic specimens im-
again that the specimen bar is essentially in a uniaxial pacted at velocities of approximately Vo = 5.8, 8, and
state of stress and that the stress wave dispersions at the 10 m / s for each of the four materials. Figure 14 shows
specimen edges are minimal. The reflected tension pulse the crack length history for 15-percent volume SiCJA1203
was of the same shape except that the higher frequency where the average crack velocity, V, increased with in-
oscillations diminished. The initial ramping at the beginning creasing impact velocity, 1Io. The average crack velocities
of the measured wave was combined with the maximum for hot-pressed alumina and the two composite materials
stress based on the measure impact velocity. This com- showed similar behavior, but were 11-31 percent higher
for the lower density alumina.

30 Onsetof crack
propagation

o
25 epo~176176

~E20

~J
15-
t
5~ -- K I from FEM analysis

o K!determinedwith~e = 0.34
I r I I I I I I
0 .... i .... i .... i .... i .... ~ .... i ....
6 7 8 9 I0 1t " 12 13 14
0.3 0.4 0.5 0.6 0.7 0.8 0.9
Post ImpactTime(ps)
c~= a/W
Fig. 10-- Finite-element verification of Kzdy"(t), Fig. 12--Typical PMT signal representing LIDG fringe
W = 8.89 mm motion

30
250-
- - - ay(t) Calculated from Vo 25
with 1.6 ~ rise time
~y(t) Measured 2o

U
10-

\"A & D 5

0 , . . . . . . .
9 10 11 12 13
-2504
0 2 4 6 8 10 12 14 Post Impact Time (~ls)

Post ImpactTime (~s) Fig. 13--Typical COD history, Vo = 8 m/s, 15-percent


Fig. 11--Typical measured stress and ideal stress volume Si Cw/AI2 03

114 ~ June 1992


The initial crack lengths, ao, for 10 of the 12 specimens the crack plane since ~,'c and tic are based on the infinite
were determined by observing the fracture surface with a medium solution for a constant crack velocity. Both plots
binocular microscope. All materials exhibited a distinct of tic and 'I'c, which were obtained through dynamic
fracture initiation site with the exception of the 99.5-
percent dense alumina. For the last test of the 99.5-
percent dense alumina, dye penetrant was baked onto the
precrack area prior to ao determination. The initial crack TABLE 3 - - S U M M A R Y OF EXPERIMENTAL RESULTS
front was generally curved as shown in Fig. 3. The
measured initial crack length was thus taken as the average Vo V VlC~ K~c K,a Kzs
distance of ao measured at the side surfaces and in the Material m/s m/s MPa~/Cm MPa~/-mMPa~/-m
middle of the crack periphery.
5.8 1975 0.32 5.3 --
Am2O3 7.9 2317 0.37 3- 4 5.9 --
K, D e t e r m i n a t i o n 10.3 2605 0.42 5.2 12.4
The dynamic stress-intensity-factor, K/y", variation for Hot Pressed 5.7 1546 0.24 5.9 --
15-percent volume SiC,/A1203 bars impacted at three Ah Os 8.1 1830 0.20 4 6.8 14.9
velocities each is shown in Fig. 15. The fracture toughness 10.1 2465 0.39 5.8 13.6
increased with increasing impact velocities within each
15-percent vol. 5.8 1457 0.22 9.0 --
material group. The K/y" data were plotted as a function
Si Cw/AI= O3 8.0 1872 0.29 6 6.5 --
of the nondimensional crack length, ~ = a/W. K~y" 10.0 2210 0.34 6.7 14.2
increased rapidly to begin with, then increased more
slowly as the crack approached the boundary. 29-percent vol. 5.8 1536 0.23 6.5 --
Figure 16 is a plot of K/y~ versus the average crack Si Cw/Ah Oa 8.0 2383 0.36 7 5.7 --

10.1 2383 0.36 9.8 14.2


velocity, V, for all the materials in this study. The data-
reduction procedure probably masks detailed crack-
velocity fluctuations which is noted in Fig. 15.
The dynamic-initiation fracture toughness, K~d, was
calculated by eq (2). All results are summarized in Table
3. All specimens impacted at 10 m / s showed macroscopic 25

crack branching. Also, the hot-pressed alumina impacted t


at 8 m / s showed crack branching. The crack-branching
toughness, K ~ , is listed in Table 3. 2 0 ~t o D []
The crack-branching stress-intensity factor, K ~ , was t_ DD ODD O O O
determined by measuring the crack-branch length, a~, 15 ,r,,--Kib, V o = 1 0 m / s 9 D [] O o O
then choosing K ~ = K~ay" ( a = aJW) from the dynamic ~ 9 []o 0 0o~
stress-intensity-factor curves, e.g., Fig. 15. K,~ was 1 9 [] oo
( 1 0 q C]oOOO
typically 2 to 2.5 times the magnitude of K,~ which is
in accordance with previously reported results. ~a Sur- ~ "o~~176 o Vo=5.8m/s
1 oo
passing Kz~ was a necessary, but not a sufficient condition 5-[ 9 [] Vo=8.0m/s
for the crack to branch and is in agreement with some 1
9 Vo=10.0m/s
tests where K dr" exceeded K ~ without crack branching.
0 t .... i .... t .... t ........ t .... i ....
0.3 0.4 0.5 0.6 0.7 0.8 0.9
Discussions = a/W

The basic postulate for the procedure described above is Fig. 1 5 - - D y n a m i c stress-intensity factor for 15-
to neglect the effects of the boundaries perpendicular to percent volume, Si C J A h O3

9 9 i m i immi
OD 2,5"
@@ODe[] O
_ 0 @ [ ] [] 0 O0
,, l , , l , , . | . | | . . . | . | t . . ,
~@uoou OO o v voq ,of 9 ~ o o o o o o o o o o
5-
oOZeD[] O OOO
e~O 0o 1.5- OOCO0000000000000
,~-~4- ~ Du 000 o~ p,
3- 0 m ~ ~176176 Impact Ave. Crack > 1- Impact Velocities ( m / s )
5.8 8 10
Vel.(m/s) Vel.(m/s)
[] [] 9 A1203
o 5.8 1~2 0,5- A 9 9 A]203- HF
[] 8.0 18~ O O 9 15% SiCw/A1203
9 10.0 2~5 Jr ,~ X 29% $iCw/A[203
. . . . ~ , ~ , o:~ . . . . , . . . . , . . . .
.... * .... n .... i .... i .... i .... i .... i .... 5 10 15 20 25
9.5 10 10.5 11 11.5 12 12.5 13
Post Impact Time (ps) K1dyn (MPa ~ m )

Fig. 14--Typical crack-length histories, 15- Fig. 16--Influence of K,"y" on the crack velocity for
percent volume Si Cw/Ah 03 all materials

Experimental Mechanics 9 115


finite-element analysis, displayed an initial oscillation (2) A new method of analysis was developed for con-
immediately after the crack began to propagate. It is not verting the measured COD and stress to useful Ka y"
known if these oscillations are noise associated with the data. The technique provided the following.
numerical solution or if they indicate a real physical
phenomenon. - - T w o nondimensional invariant parameters, f~ and ~I,,
Before proceeding to the analysis of experimental data, were derived. The constants defined as f~c and ~c were
it was important to verify that the numerically deter- used to analyze any number of experiments of similar
mined calibration constants, flc and XI,c, would determine materials in a variety of impact velocities.
the correct crack-length history, a ( t ) , and K dvn a s deter- - - O n l y one dynamic finite-element analysis was needed
mined by the finite-element model. A verification, which per specimen geometry to analyze a series of tests. This
used [~c and ,1% and the known stress history and COD makes dynamic fracture study of ceramic materials
from the FEM analysis to calculate a ( t ) and K a y", was more economical.
thus conducted. In regard to KaY" , the initial oscillation
in ~I, did introduce a discrepancy of about ten percent (3) The toughening mechanisms for SiCw/A1203 com-
immediately after the crack began to propagate. However, posites were ineffective when the material was subjected
K ay" matched closely in the range of 0.5 < o~ = a ~ W < to severe dynamic fracture conditions.
0.85. The difference between ~theo = xf~-/32 and XI,c may
be due to an overly stiff finite-element mesh. (4) A unique relationship between average crack velocity
The LIDG technique has a sensitivity of 0.911 vm for and KaY" was not apparent for the crack velocities in the
this particular setup. The experiment was such that the range of 1400 to 2600 m/s.
COD was less than 1 #m at the time the crack began to
propagate. This caused uncertainty at the initial stages of (5) This study reaffirmed that K,b is 2 to 2.5 times K,d
the crack propagation and would have made it impossible and that the crack velocity at which branching occurred
to accurately determine ao or K,d. Instead, ao was measured was roughly 0.4 C2.
directly from the fracture surface and Kid was calculated
from the procedure presented. The crack-opening dis-
placement varied nearly linear with time. The variation
Acknowledgment
from a straight line matched the time the stress wave took This research was supported by the Office of Naval
to propagate from the crack to the free boundary and Research Contract No. 0004-87-K032. The authors express
back again. their sincere gratitude to Drs. Yapa Rajapakse and Steve
The crack velocities were nearly the same for the hot- Fishman for their patience and encouragement during the
pressed alumina and the two composites. This indicates course of this investigation.
that the toughening mechanisms in the composite were
ineffective under dynamic fracture. The higher crack
velocity in the lower density alumina was in accord with References
other published data.l, 2 1. Yang, K.H., Kobayashi, A.S. and Emery, A.F., "Dynamic Tough-
ness of Ceramic Composites," Car. Eng. ScL Proc., 9 (7-8), 795-802
Crack propagation initiated at the respective values of (1988).
KI~ for each material and reached values of K / y " as high 2. Takagi, Y. and Kobayashi, A.S., "'Further Studies on Dynamic
as four times K,d. A unique relationship between V and Fracture Response of Alumina and SiCw/AI20a Composite," Proc. of
K a y" was not apparent from the results of this experiment. Symp. on Elevated Temperature Crack Growth, ads. S. Mall and 7".
Nicholas, ASME MD-Vol. 18, 145-148 (1990).
Other researchers have reported similar results for ceramic 3. Kalthoff, J.F., Winkler, S. and Beinert, J., "The Influence of
materials. '. 2 Dynamic Effects in Impact Testing, "Int. J. Fract., 13, 528-531 (1977).
A great advantage of this technique was that the crack 4. Mall, S., Kobayashi, A.S. and Urabe, Y., "Dynamic Photoelastie
was loaded by a relatively 'clean' tension stress which and Dynamic Finite Element Analysis of Dynamic Tear Test Specimens,"
happens to be uniform across the crack plane and loads EXPERIMENTALMECHANICS, 16, 449-458 (1978).
5. Marehand, A., Duffy, J.F., Christman, T.A. and Suresh, S., "An
the crack in mode I opening only. The duration of the Experimental Study of the Dynamic Mechanical Properties of an Al-Si Cw
experiment was less than the time for the primary release Composite, "Eng. Fract. Mech., 30 (3), 295-315 (1988).
waves to return to the crack. The advantage of this 6. Suresh, S., Nakamura, T., Yeshurun, Y., Yang, K.-H. and Duffy,
experimental procedure is the lack of an expensive dynamic J.F., "'Tensile Fracture Toughness of Ceramics Materials: Effects of
Dynamic Loadings and Elevated Temperature," J. Amer. Car. Soc.,
finite-element analysis. The crack behaved as a crack in 73 (8), 2457-2466 (1990).
an infinite solid for the limited duration of the experiment. Z Nose, T. and Fujii, T., "Evaluation of Fracture Toughness for
Ceramic Materials by a Single-Edge-Precracked-Beam Method," J. Amer.
Car. Soc., 71 (5), 328-333 (1988).
Conclusions 8. Sharpe, W.N., Jr., ''A New Optical Technique for Rapid Deter-
mination of Creep and Fatigue Thresholds at High Temperatures,"
AFWAL TR 84-4028, Air Force Wright Aeronautical Lab., WPAFB
(1) A new type of experiment was designed to study the (1984).
dynamic fracture behavior of ceramic materials. The new 9. Freund, L.B., "Crack Propagation in an Elastic Solid Subjected to
General Loading-IlL Stress Wave Loading, "" J. Mech. Phys. Solids, 21,
test method provided several distinct advantages over 47-61 (1973).
current experimental methods. 10. Ravichandran, G. and Clifton, R.J., "'Dynamic Fracture Under
Plane Wave Loading, "Brown University Report No. AD-A178 594. U.S.
- - A predictable unidirectional stress pulse was suddenly Army Research Office, Rapt. No. ARO 223067EG (Dee. 1986).
11. Hodulak, L., Kobayashi, ,A.S. and Emery, A.F., "'A Critical
applied at the crack plane. Examination of a Numerical Fracture Dynamic Crack," Fracture
- - A sharp precrack provided a better estimate of Kxd. Mechanics: Twelfth Conference, ASTM STP 700, 174-188 (1980).
--Once set up, this experiment was relatively easy to 12. Deobald, L.R., "Experimental~Numerical Characterization of the
perform and was comparable to the popular instru- Dynamic Fracture Behavior of Ceramic Materials," PhD thesis submitted
to Univ. of Washington (June 1991).
mented Charpy impact test. 13. Ramulu, R. and Kobayashi, A.S., "Mechanics of Crack Curving
- - T h e method gave consistent results for a (t) and Kay". and Branching, ""Int. J. Fract., 27, 187-201 (1985).

116 9 June 1992

Vous aimerez peut-être aussi