_ o; E=:_D_<_p+xy(_D_ - DXXe_~® e_.)v_O_lDX" <_ 01,
I~I_~:_D___p-~-,(__D)vDID~_<0; ~=:_D_ <_ p-Tr1( ~ - O=e_x ® e_x)VDlD =
>_
01.
(lla)
(1 lb)
The geometrical meaning of these domains in the macroscopic stress space is as follows. Gy+ (resp. G f ) is obtained by first homothetically contracting GI with homotheticity ratio p+ (resp. p - ) . The part o f boundary o f the contracted domain where the outward normal D has positive (resp. negative) component along 2?xx coincides with a part of the boundary of Gf+ (resp. G f ) . The remaining part of boundary belongs to the cylinder with generatrices parallel to S = and tangent to the contracted domain at the points where the outward normal D is perpendicular to S xx (see Fig. 2).
l ~ik(i,k~x,x) t
G~ \
~
/
(1:)' I:) --"~"=o)
V"- ~'~ O+O'Im+
..~:.:.:.G. !::::::"~:~'.:.-:i::
,G;
_
"'
Fig. 2. Graphical construction of the domains Gf+ and Gf (see eqns (1 la,b)) in the macroscopic stress space. The dotted area is the intersection of these two domains (G7 n (:;7).
750
A. TALIERCIO
Thus, the scalar inequality eqn (10) is equivalent to the inclusion G h°m C (G7 N
G£). In order to improve the upper bound just defined, periodic failure mechanisms featured by rigid relative movements between blocks in the RVE will be considered. Only mechanisms with planes as slip surfaces will be considered; the planes will be supposed not to cross any fiber, their intersection with RVE will be denoted by S and their normal by ns. Recalling eqn (3') and denoting by V the jump in the velocity field across the plane of failure, for the considered mechanism the term (Tr(d)) is given by <~(d)> = [Sl~m(~s;_V). For further developments, it is convenient to associate any mechanism, with failure plane perpendicular to _ns, with a domain in the macroscopic stress space defined as follows:
c*(_~s) = I__sl__s:__D-< ~ m ( P ) l, where
z,~(D) = ~,~(es; V) = +oo
if D
=
_V (~ _ns;
otherwise.
IV. 1. Rectangular array If the fiber array is rectangular, a mechanism of the kind described above, with failure plane passing only through the matrix and perpendicular to ey (or _ez), can be defined for any percentage of reinforcement-see Figures 3a,b. Let Vy (resp. Vz) be the relative velocity between the two rigid blocks in movement and letsSy (resp. Sz) be the surface in the failure plane. Since D = e_y ~ Vy ISy I (resp. D = e z ® Vz ISz ]), by virtue of the definition of 7rh°m, eqn (3), the two following inequalities hold for any ~/ ( _< ~r) :
~hom(p = e_z Q ~ ) <- ~m(e_z; ~ ) .
n~
"~ Vy
¢)
Fig. 3. Periodic failure mechanisms for rectangular RVEs, featured by relative rigid movement of blocks.
Macroscopic strength domain of a fiber-reinforcedcomposite material
751
These inequalities amount at the inclusion between domains in the macroscopic stress space
G h°m C Gm(e_y) n G*(ez). Another failure mechanism featured by the required periodicity consists o f two parallel slip planes crossing each RVE, located at opposite sides with respect to the fiber and with traces in the (y, z) plane intersecting the edges of the RVE at their midpoints (see Fig. 3c). The unit vector perpendicular to these planes has the form n~ = +_e_yc~+_ ezS~. The triangular wedge of RVE with n, as inward normal is supposed to have relative velocity V~ with respect to the central block in which the fiber is located, so that by periodicity the triangular wedge with n~ as outward normal is required to have relative velocity - ~ with respect to the central block. Furthermore, in order to have failure planes passing only through the matrix, the fiber volume fraction ~ must not be greater than 7r/4s~c~ (i.e., ~ _< a-/8 = 0.39 for square RVEs). If this restraint is fulfilled,
7fh°m(D : V~ ~ n_13)~ 7fm(_V~;n~) that is, in terms of domains
G h°m C G*(n_/~). As a conclusion, an upper bound to G h°m for composites with rectangular array of fibers is given by the domain GI = Gf+ n G f n G*, where
G* = G* (ey) n G* (ez) n G*(_na) = G*(ey) n G*(ez)
if ~ < lr/4s~c,
otherwise.
IV.2. Hexagonal array Suitable periodic mechanisms featured by failure planes perpendicular to ey passing only through the matrix are possible provided that the trace o f any failure plane in the (y, z) plane intersects the skew edges o f the RVE at their midpoints (see Fig. 4a) and
b)
c)
Fig. 4. Periodic failure mechanisms for hexagonal RVEs, featured by relative rigid movement of blocks.
752
A. TALIERCIO
provided that any RVE be crossed by two parallel planes located at opposite sides with respect to the fiber. The failure planes pass through the only matrix if 7/_< 7r/Ssa/c~ (i.e., if ~/_< 7r%/-3/8 = 0.68 for regular hexagonal arrays). Denoting by Vy the relative velocity between the triangular wedge with ey as inward normal and the central block containing the fiber, by periodicity the velocity of the triangular wedge with _ey as outward normal relative to the central block must be -Vy. Similar considerations apply for mechanisms with failure planes perpendicular to _ez (Fig. 4b). These planes do not cross any fiber provided that ~/_< ~r/8c~/sa (i.e., ~ _< • -x/3/24 = 0.23 for regular hexagonal arrays). Furthermore, periodic are also mechanisms featured by two parallel failure planes per RVE, located at opposite sides with respect to the fiber and with traces in the (y, z) plane intersecting a skew edge o f the RVE and the adjacent edge parallel to the y-axis at the relevant midpoints (Fig. 4c). The equation of the unit vector perpendicular to these planes is n~ = _+eyC~_+ ezSa. In order to have failure planes passing through the matrix only, the fiber volume fraction 71 must not exceed ~l/2sac~ (that is, 0.68 for regular hexagonal fiber arrays). As pointed out in Section IV. l, denoting by V~ the relative velocity between the triangular wedge with na as inward normal and the central block containing the fiber, the velocity of the triangular wedge with n~ as outward normal relative to the central block must be -V~. As a conclusion, after having defined domains GTn(ey), G*(ez) and G * ( n a ) analogous to those defined for rectangular arrays, it is possible to state that G h°m C GI, where GI can be expressed as Gl = Gf+ n Gf- n G* and G*m = G*(ey) n G*(_ez) O G*(n~) if r / < 7r/8c~/s~ Gm(ey) O Gm(ne) if r/8c~/s~ < ~ < mini 7r/8s~/c~; 7r/2sec~l
G* (n~) if rc/8c~/s~ < rl < 7r/8s~/c~ and 3 -< 60 ° Gm*(ey) if ~r/2s~c~ < ~ < minlTr/8syc~;~e] and 3 -> 60 ° R 6 otherwise. V. APPLICATIONS The theoretical results presented up to here will now be applied to the evaluation of lower and upper approximations of the maximum solicitation that fiber reinforced composites subjected to particular states of stress can sustain. The material forming the matrix will be supposed to undergo Drucker-Prager strength criterion. This two-parameter criterion is an extension of Von Mises criterion; here it was chosen for the reasons discussed in Section I and because of the existence of experimental results showing that the ultimate strength of some polymeric matrices is well described by a Drucker-Prager criterion under some stress conditions (see, e.g., the results reported by HULL [1981]). As is well known, denoting by km the shear strength of the matrix and by ~m a parameter related to the "internal angle of friction" of the matrix, Drucker-Prager criterion reads (see, e.g. SALENqON [1983]):
O:mtr Om +
~/l (O=m:~=m -- ltr2 __am)--< km
and the relevant support function is
Macroscopic strength domain of a fiber-reinforcedcomposite material
Irm(D) =
=
km t r D
3arm
=
= +oo
if t r D >__OtmX/6(3_D:_D - tr2_D) = _ _ _
753
(12)
otherwise.
The parameters am, k,, defining Drucker-Prager criterion are related to the uniaxial strength values of the matrix by the relationships 1
-
a,.
-
a m -- am+
+, x/3 am + a..
km =
2
a m a i n+
+. x/3 am + am
Only the case am -> am+ (i.e., 0 _< ct,~ < l/x/-3) will be treated, for its greater interest in practical applications. When failure mechanisms with slip planes are considered, setting V @ _ns = D the condition tr D _ ct,,x/6(3D : D - tr 2 D) can be easily shown to be equivalent to 30/rn
_V'ns ~ x/1 - 3ot2m I N . In other words, finite values of a'm (and, consequently, significant upper bounds to ~rh°m) are obtained if the relative velocity _V between rigid blocks falls within the cone with opening angle 3'* = arccos3 otto/x/1 - 3at 2, having as axis the normal to the slip plane (see Fig. 5). Note that such cone degenerates if ctm > 0.5/4-3; thus, for Drucker-Prager matrices with excessively high compressive-to-tensile strength ratios (i.e. 0.5/x/3 < am --< l/x/3), no bound is obtained by this kind of mechanism. Drucker-Prager criterion reduces to Van Mises criterion if the matrix has equal strength in uniaxial compression and tension (am+ = am a m ~-- kmx[3, i.e., at,, = 0). The support function of Van Mises strength criterion can be deduced from eqn (12) noting that ctm ~ 0 implies tr D ---,0 if finite values for qrm (and consequently significant upper bounds to the macroscopic strength) are to be obtained. Thus (see also SAtENqON [1983]): =
r,,(D) = a , ~ ~ =+~
iftrD=0
iftrD~:0.
In the case o f Van Mises criterion, when failure mechanisms with slip planes are considered, setting D = _V ® _ns the condition tr __D= 0 is equivalent to _ns .1_ _V. In other words, for composites with Van Mises type matrix, the relative velocity between rigid blocks in movement has to be tangent to the failure plane. Also the material forming the reinforcing fibers will be assumed to undergo a DruckerPrager type criterion, defined by parameters Cry and ky. For the sake o f illustration, suppose that the ratios between the uniaxial strengths of fibers and matrix are such that + + af lain = a f / a m = r( >_ 1), so that eqns (9) under which the bilateral bounding of G h°m was obtained are fulfilled. Thus, o~y and Icy can be expressed in terms of am, km as otf=Ctm;
k f = r k m ( r > _ 1).
For composites with Drucker-Prager matrix, the following form for the domain Go, defined in general by eqn (7), is obtained:
754
A. TALIERCIO
Fig. 5. For failure mechanisms with slip planes passing through a Drucker-Prager matrix, the relative velocity V between rigid blocks in movement must form an angle 3' with the normal _ns to the slip plane not greater 2 of the outlined cone. {han the opening angle 3'* = arccos 3Otm/~/1 - 3orm
Gb = {__SIc~m(tr__S- o) + ~/~(_S:_S- ~trZ__S) + a ( ~ t r _ _ Z - S ~ x ) -
k m < 0, -
6-
+ ~a2 (13)
< a < 6+}
whereas the e q u a t i o n s o f the d o m a i n s defined in Section IV are
Gm(ey)=
S=
Odin~,yy -- kin~3 ~_4o~2
\
1
2
"1"-(Sxy)2"}- (SYZ)2 ~ 0 "
~ -- 40t m
(14a) Olin S z z
Gm(ez) =
=27
~
--
kin~3
4Otm z
+ %/12012(OlmfZZ~km/3~ 2 ,~
\
~ _ 4otZm
(,~,yz)2
] +
(~zx)2 +
I <0
;
(14b)
~mS"" -~ _ n ~ ) = __2: } - 4 ~
km/3 +
f l 2 o t 2 { ° t m E _I n n - k m2/ 3 )
\
2 -I- (S-,)2-l- ( ~ , n x ) 2 ~ 0 ;
~ --4Otm
S "n = SYYc~ +_ 2SYZcas~ + SZZs~; S "x = +_L'XYct3 +_ SZXs~; S"'
=
(14c)
+_(,rez _ ,re~)cese + syZ(c~ - s~) 1 ;
G f f q G 7 = [__SIO~mtr__S+ x/~ (__S:__S-- ~tr2__S) _< [1 + o ( r -
l)]km}.
(14d)
Macroscopic strength domain of a fiber-reinforced composite material
755
V. 1. Bilateral bounding o f the macroscopic uniaxial strength Suppose that the composite is subjected to uniaxial tension (resp. uniaxial compression) acting in the ( x , y ) plane. Let 0(0 _ 0 < 90 ° ) be the orientation o f the uniaxial stress, S, to the fibers. Set, for brevity, So = sin 0 and Co = cos 0. T h e aim now is computing the functions S~(O) and 27~-(0) (resp. S~(O) and S;-(O)) constituting a lower and an u p p e r b o u n d , respectively, to the anisotropic macroscopic strength 27+(0) (resp. Z'-(0)) o f the composite. The lower b o u n d to the absolute values o f the tensile/compressive uniaxial strength is given by eqn (13) and reads: 27o-+(0) = s u p [27 10:m(27 -- (7) -['- ~jl (27 __ 0.)2 ..[_ 27(7S2 __ km ~ O; - 6 -
----- (7 ~--_ 6 + ]
0
that is
---6+ + 1--3---2 ~J ( ~ O t m 6 + S o 2 + k m ) 2 - ( 6 + ) 2 ( 1 - ' ~ 3 s 2 ) (1 - 3°t2)s2 1-6+$2 + o t m k m
=1=3 2
270±(0) =
-T-3~Otm
~/(1 - ~s~)(1 - 3~m2) ]
3 . 2 --74S0 3 2 "["
1 -
1 - 3c~2
(1
3"-'~-~2m ~ "~ ~So2-'~-_ )So
OtmO-So2 + km +
] km
if o>- 6 +
if - 5 _< o _< 5 +
- ( 6 - ) 2 1 - ~ s 2 (1 - 30~2)s 2
1 --3%m l
-T- 3
----
--~7
2
SO + o t m k m
1 -
3or 2
if o ~
-6-
where 17 - -
-1"3 0lm
km
1 -3a
2-
--
+
=
3 2 zso
1-3c~-
~sd3 2
x/1 - 3c~so 1 - 3~Zm - zSd32"
743s02
As for the upper b o u n d , note first o f all that G~,(e_z) does not impose any b o u n d to the uniaxial strength in the ( x , y ) plane, since 27= = 27yz = 27zx = 0. The definitions o f the other domains, eqns (14), yield 1. for rectangular RVEs with ~/ < 7r/4saca and for hexagonal RVEs with 7/ < mini 7r/8sa/ca; 7r/2saco} : I
27?(0) =km min
1 . (41 - 12t~2Co +_ 30lmSO)S 0 ' 1
-
--
. l+~(r-1)~.
+_ 3
(15a)
756
A. TALIERCIO
2. for rectangular RVEs with ~ > 7r/4s~c~ and for hexagonal RVEs with 7r/2soca < ~ < mini r/8s~/ce; ~e ] and 3 -> 60 ° :
S~(O)=kmmin
1 . 1 + r / ( r - 1) / " ( x / 1 - 12oe~Co +_ 3OemSo)So' + _ - - d ~ - i ~ ) '
(15b)
. for hexagonal RVEs with 7r/8s~/c~ < 71< 7r/2sac~ and 3 -< 60 ° :
S?(O) = kmmin
1
. l+~(r-1)
/"
(x/1 - 12c~2 x/1 - So2ca2 _ 3~,,SoC~)SoC ' _+~m---+-i-/~f~ )
(15c)
4. for hexagonal RVEs with fiber volume fraction different from those specified above, it is only possible to state that: 1+r/(rS?(O)
-
1)
++-Olm+ 1/ ~f3
kin.
(15d)
In Figure 6 plots of So~ and Z'~ for composites with Drucker-Prager type fibers and matrix (with r = 5) are shown. Figure 6a refers to a square reinforcing array and Figure 6b to a regular hexagonal array. First of all note that, apart from the cases with greater o/m values, the gap between the two bounds is relatively small. However, note that in Figure 6 the fiber volume fraction was implicitly assumed to be sufficiently lower than the maximum one compatible with each kind of reinforcing array, so that the strictest upper bound was used (see eqn (15a)). The gap between bounds may tend to increase at some 0 for larger volume fractions. Anyway, for any am and whatever the geometry (3, rt) of the composite may be, the lower and the upper bounds coincide at 0 = 0, so that 1 + r/(r -
S0-+(0) = ~'1-+(0) =
1)
k m -+-'~n'-~ 1-/~f~ ( = S ± ( 0 ) )
(16)
is the actual macroscopic strength of the composite along the fiber direction. Note that S ± (0) = (1 - T/)o,~ + ~of-+ is apparently the weighted average of the uniaxial strengths of fiber and matrix, the weights being the relevant volume fractions. Thus, eqn (16) is a rigourous validation of a well-known semiempirical formula widely used in practice, usually called "rule of mixtures" (see, e.g. HASrnN [1983]). V.2. Bilateral bounding o f the macroscopic biaxial strength I f the strength properties of the composite have to be estimated under stress conditions not as simple as the uniaxial ones considered in Section V. 1, the solution of the problem is likely to be obtained numerically. As a rule, this can be done by fixing a radial path in the stress space and by determining the m a x i m u m norm of the macroscopic stress tensor compatible with Go or GI. The lower and upper bounds to the macroscopic strength domains for the state of stress considered are thus derived point by point. In order to illustrate how the model described here makes it possible the description of the macroscopic strength properties of fiber composites subjected to any state of stress, the lower and upper bounds to the biaxial strength of composites with Drucker-
Macroscopic strength domain of a fiber-reinforced composite material
757
a) I0
"
6i
M
6~
"~
4i 2a
....
0
_
,-4-
...~,. •. ~ . . . :~... ~ . . . ~ .. ~
-8-
-6i I -a!
,"-'~
-'ll
/
4 _,o.
- l
O
-
1
~
t 2
~ ~
° ,"- --"] /
-10 -IB
- 12'
-14,
-14'
-16
x,,,,,,,.
:1
::lf/ 't
Fig. 6. Bounds to the uniaxial tensile (27+) and compressive (27-) strength for composites with o f / o + = of~ore = 5 and Von Mises (Ctrn = 0) or Drucker-Prager (am = 0.1,0.2) matrix. Solid line is lower hound; dashed line is upper hound. (a) Square reinforcing array (8 = 45°; ~ < 0.39); (h) regular hexagonal reinforcing array (/3 = 60°; ~/< 0.68).
Prager type matrix will be now derived. Denote by 27/, 27//the two nonvanishing principal stresses and by 0 the orientation of 27/with respect to the fibers. Suppose that the principal stresses act in the (x, y ) plane. The lower bound to the biaxial strength is obtained by making use of the definition o f Go, eqn (13), noting that in the present case t r ~ = 27/+ 2,H, ~ : ~ = 272 + 272 and 27~ = 27/c2 + 2~us2. According to the procedure mentioned above, let A be a positive parameter (measuring the norm of =27)and let ~ be an angle varying between 0 ° and 360 ° in the plane (27/, 27//). Setting 27/= A cos ~( = Ac~), 27/i = A sin ~k( = As~), for any prescribed orientation 0 of the principal stresses to the fibers the maximum value for A (q~) fulfilling eqn (13) has to be found:
758
A. TALIERCIO
FindAo(dp) = supIAa [/~Olm(C~ -1-S¢a) -- OlrnO
+~
1/
l
A2(1-c4~s4,)-~Aa[co+so+3(c4~-s,~)(coZ-sZ)]+°2-km<-O;
- 6-< o< 6+/.
(17)
Numerical solution of problem (17) yields the lower bound sought. As for the upper bound, formed by the intersection of all the domains defined by eqns (14) or some of them, again note that in the case considered here G~,(e_z) does not furnish any significant bound (since S z': = S yz = S zz = 0). Also note that S yy : ~1 $2 "l- ,~11C2 and
S xy = (271 - Su)CoSo.
Following the same procedure as for the lower bound, for any prescribed orientation 0 three values for the parameter A(_> 0), denoted by Ay, Aa and Af, are obtained by finding the maximum value of A such that 271 and 27//are compatible with G~,(ey), G,~ (n~) and G~ f) G f , respectively. In this case, unconstrained maximization problems have to be solved and their solutions read km
3Otm(C~S2 + $4,Cif)
+ X/1 -- 12c~21c~ - s~lSoCo
km 3Otm(CoS ~ + S6Cff)C~ + X/1 -- 12etm 2 ~/(c~ - s~,)2c~s~ + (c4,s~ + s4,coZ)2s~c~ [1 + ~(r - 1)]km Af = (i(1
-- ¢~S~) + 2m(C¢ +S~)"
Finally, the pairs of principal stresses at which the upper bound to the ultimate biaxial strength o f the composite is reached are computed as St = A 1c~, S I / = A i s~, where 1. for rectangular RVEs with ~ < 7r/4s~c~ and for hexagonal RVEs with ~/ < mini ~/8s~/c,; 7r/2s~c~ I : A~ = minlAy, As; Af};
(18a)
2. for rectangular RVEs with 7r/4s~c~ < ~/and for hexagonal RVEs with r/2saca < < minlTr/8sa/ca;~e} and/3 _> 60°: A~ = minlAy; As];
(18b)
3. for hexagonal RVEs with 7r/8sa/c~ < 71 < ~r/2st~c ~ and/5 < 60°: A~ = minlA~; As];
(18c)
Macroscopic strength domain of a fiber-reinforced composite material
759
4. for other types of hexagonal arrays: A, = A/.
(18d)
Figures 7 and 8 show the bounds computed according to the procedure just described. Only three orientations (i.e. 0 = 0 °, 22.5 °, and 45 °) have been considered for clarity of representation. Figure 7 refers to composites with square RVEs, Figure 8 to composites with regular hexagonal RVEs (~ = 60 ° ). The case o f Von Mises type matrix is considered in Figures 7a and 8a, whereas Figures 7b and 8b refer to Drucker-Prager type matrix with a,, = 0.1. As a rule, the two bounds are very close. Again note that the strictest upper bound (computed according to eqn (18a)) was used, so that at greater volume fractions a larger gap might be obtained. VI. C O N C L U D I N G
REMARKS
The homogenization theory applied to limit analysis has proved to be an effective tool for obtaining lower and upper bounds to the strength of composites under any stress condition in a relatively easy manner. In particular, by means of the simple failure mech-
4 Et
2¸
a)
,(-
9
n-, ¢I
;
/
-2
IIIllllll,llllllllllllllllllli'llllllllllllllllllllllll'll -8 -4 0
8
4
12
r,,Ik.
c e e e o #=0
l.b.
oeeoe =0 *¢*** #=22.5 *,,*** =22.5 e"=--= #=45 aeaoe =45
u.b. l.b. u.b. l.b. u.b.
\'l
[.,I .
.
.
.
•x,~,-.~ ~ .
.
.
.
--4I,,,.,,,,,l,.,,..,,,l,..,,.,.,l.,.,,,,,,l,,,,,,,,,l.i,.,.,
-16
-12
-8
-4
. , ,,
0
4
8
,',,"I
12
/I
,
I
x
X lr
Fig. 7. L o w e r a n d u p p e r b o u n d s to the b i a x i a l stress d o m a i n s o f c o m p o s i t e s with s q u a r e r e i n f o r c i n g array (with /~ = 45 ° a n d 7/ < 0.39), for d i f f e r e n t o r i e n t a t i o n s 0 o f 271 to the fibers. (a) Von Mises m a t r i x ; (b) D r u c k e r - P r a g e r m a t r i x .
760
A. TALIERCIO
a)
4 ~
2.
~11II
O' I /
•
//'*
_ _~ x ~ l . .
-4
iiiiiii
¢.... . . . . .
IIIIIIIIIIIIIIIIIIIIIIIlllllllllllllllllll|lllllllll
-8
-12
0
-4
I
8
4
12
Y:,/k=
e e e e o ~=0
l.b.
oe~oe ***** ***¢*
u.b. l.b. u.b. 1.b. u.b.
:0 11=22.5 =22.5 ===-" 4=45 oe~ =45
~
1 b) ............ /
::]'"
~
.
. . . . -.
•
iiiiiiiillllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll
-18
-12
-8
-4
0
P.i/k=
4
8
12
.1i r<°,
Fig. 8. Lower and upper bounds to the biaxial stress domains of composites with hexagonal reinforcing array (with B = 60° and ~7< 0.68), for different orientations/9 of S I to the fibers. (a) Von Mises matrix; (b) Drucker-Prager matrix.
anisms for the RVE defined in Section IV, analytical equations for the upper bounds to the macroscopic strength exhibited by the composite under particular stress conditions were obtained (Section V). Provided that some conditions regarding the geometry of the composite and the strength properties of the components are fulfilled, these upper bounds do not excessively overestimate the lower bounds obtained by making use of the domain Go defined in Section III (and, in some cases, are coincident with the latter ones). If the geometry of the composite (defined by the percentage o f reinforcement 7/and by angle/3) is such that the discrepancy between the two bounds is excessive, an approach more complex than the one presented here might become necessary. First of all, periodic failure mechanisms for the RVE not as simple as those considered in Section IV could be taken into account, in order to improve the upper bound to the strength of the composite. Second, a structural analysis o f the RVE could be performed, in order to find with fair approximation the actual macroscopic strength of the composite, provided that a suitable computer program is available; however, this approach turns out to be quite cumbersome in terms of computational cost (as shown by similar nu-
Macroscopic strength domain of a fiber-reinforced composite material
761
merical studies performed by TURGEMAN and PASTOR [1987] for heterogeneous layered media and by MARIGO et ai. [1987] for perforated plates). Note that the approximations defined in Sections III and IV can be used for composites with components undergoing any kind o f strength criteria (provided that eqns (9) are fulfilled). If suitable experimental results are available, the definitions of Go and G~ can be specialized in order to properly take into account the actual strength properties of matrix and fibers. However, as mentioned in Section I, TAZlERCIO [1989] and DE BurIAN and TALIERCIO [1991] showed that, assuming that the matrix undergoes Von Mises strength criterion (which is a special case of that considered in Section V), the lower bound to the strength of the composite both under uniaxiai tension and under combined normal stress and shear is in many cases a close estimate of that measured by other authors on composite samples. Finally, it must be acknowledged that here the fiber-matrix interface was implicitly assumed to be indefinitely strong, or at least to be such that, in any case, failure never occurs along this surface. There are experimental results, however, that show that actually the fiber-matrix interface is the failure surface for composites subjected to particular load conditions. The bounds presented here can be quite easily modified if there is any experimental evidence for assuming that the interface has limited strength. In fact, homogenization theory applied to limit analysis allows also the formulation of macroscopic strength criteria taking into account interface strength criteria (see, e.g. DE BuHAN & TALIERClO [1988,1991]). Acknowledgements--This work was developed within the framework of a research program supported by the Italian Ministry for University and for Scientific and Technological Research (M.U.R.S.T.). The author is grateful to Professor Cesare Davini of Udine University (Italy) for his interest and most valuable suggestions.
REFERENCES 1950 1964 1965 1967 1970 1970 1971 1972 1974 1980 1981 1983 1983 1983a 1983b 1984 1987
HILL,R., "The Mathematical Theory of Plasticity," Clarendon Press, Oxford (GB). HASmN,Z., and ROSEN, B.W., "The Elastic Moduli of Fiber-Reinforced Materials," Trans. ASME, J. Appl. Mech., 31, 223. Azzi, V.D., and Ts~, S.W., "Anisotropic Strength of Composites," Experimental Mechanics, 5, 286. HOFFMANN,O., "The Brittle Strength of Orthotropic Materials," J. Compos. Mater., 1,200. McLAuor~IN, P.V., Jr., and BATTER~N, S.C., "Limit Behaviour of Fibrous Materials," Int. J. Solids Structures, 6, 1357. TVRRELLROCKAFELLAR, R., "Convex Analysis," Princeton University Press, Princeton, N.J. TsAI, S.W., and Wu, E.M., "A General Theory of Strength for Anisotropic Materials," J. Compos. Mater., 5, 58. McLAoGrtLIN, P.V., "Plastic Limit Behaviour and Failure of Filament Reinforced Materials," Int. J. Solids Structures, 8, 1299. Wu, E.M., "Phenomenological Anisotropic Failure Criterion," in B R O W , L.J., and KROCK, R.H. (eds.), Composite Materials, Vol. II, Academic Press, New York. HASmN,Z., "Failure Criteria for Unidirectional Fiber Composites," Trans. ASME, J. Appl. Mech., 47, 329. HULL,D., "An Introduction to Composite Materials," Cambridge Univ. Press. HASmN,Z., "Analysis of Composite M a t e r i a l s - A Survey," Trans. ASME, J. Appl. Mech., 50, 481. SALENqON,J., "Calcul ~ la Rupture et Analyse Limite," Presses de I'E.N.P.C., Paris. SUQtrET, P., "Plasticit6 et Homog6n6isation," Th~se d'Etat, Universit6 Pierre et Marie Curie, Paris. SUQUET, P., "Analyse Limite et Homog6n6isation," C.R. Acad. Sci., Paris, 296, S6rie II, 1355. BOErILER,J.P.,and RACLr~, J., "Failure Criteria for Glass-Fiber Reinforced Composites Under Confining Pressure," J. Struct. Mech., 13, 371. MARIOO,J.P., ML~,LO~, P., MICHEL, J.P., and SUQtr~T, P., "Plasticit6 et Homog~n6isation: Un Example de Pr~vision des Charges Limites d'une Structure H6t&og~ne P&iodique," J. M6c. Th. Appl., 6, 47.
762
1987 1988 1989 1991 1991
A. TALIERCIO
TURGEMAN,S., and PASTOR, J., "Comparaison des Charges Limites d'une Structure H6t6rog6ne et Homog6n6is6e", J. M6c. Th. Appl., 6, 121. DE BUI-IAN,P., and TAtIERCIO, A., "Crit6re de R6sistance Macroscopique pour les Mat6riaux Composites h Fibres," C.R. Acad. Sci., Paris, 307, S6rie If, 227. TALIERCIO,A., "Study of the Elastic and Ultimate Behaviour of Fiber Composite Materials and Composite Structural Elements" (in Italian), Ph.D. Thesis, Politecnico di Milano. DE BUI-IAN,P., SALENfON,J., and TALIERCIO,A., "Lower and Upper Bounds Estimates for the Macroscopic Strength Criterion of Fiber Composite Materials," in DVORAK,G.J. (ed.), Inelastic Deformation of Composite Materials, Springer-Verlag, New York, p. 563. DE BUHAN,P., and TALIERClO,A., "A Homogenization Approach to the Yield Strength of Composite Materials," Eur. J. Mech., A/Solids, 10, 129.
Dipartimento di Ingegneria Strutturale Politecnico di Milano Piazza Leonardo da Vinci, 32 20133 Milano, Italy
(Received 6 April 1991; in final revised form 17 September 1991 )