Introduction
As compression molding of chopped fiber reinforced ther- ranged in a randomly woven network in the plane of the sheet.
mosetting compounds has grown in commercial importance in This arrangement of fibers makes the material anisotropic, so
recent years, a need has developed for a model describing the that its mechanical response is substantially different from
flow of such materials as they are formed in the mold cavity. that of a polymer melt.
Most efforts to date in this area have been motivated by suc- All existing models for the flow of SMC assume it to be an
cessful models for the flow of thermoplastic melts in injection incompressible isotropic fluid. Inertia is neglected and a no-
molding. Although the molding compound is squeezed be- slip boundary condition is imposed at the cavity surface. Silva-
tween cavity surfaces in compression molding rather than Nieto, Fisher, and Birley (1980) were the first to propose a
forced through a gate onto a fixed cavity as in injection model under these assumptions, based on isothermal, Newto-
molding, the flow in both cases is confined to a relatively nar- nian lubrication theory. Tucker and Folgar (1983) later used
row cavity, which suggests a similar approach. However, the this model with a finite element method to calculate the flow-
sheet molding compounds (SMC) typically used in compres- front progression in a rectangular charge. These calculations
sion molding have a distinguishing characteristic which correlated well with their experiments on modelling clay and
dramatically influences the qualitative aspects of their flow. single-layer charges of SMC. In a later paper, Lee, Folgar, and
These compounds are filled with a relatively large volume Tucker (1984) generalized this model to include a power-law
fraction of chopped fibers (at least 25 percent) which are ar- viscous response, but showed (again by means of finite ele-
ment calculations) that the flow-front progression is insen-
sitive to the value of the power-law exponent. In the same
Contributed by the Applied Mechanics Division for publication in the JOUR-
paper, they reported experimental results that showed a
NAL OF APPLIED MECHANICS. significant change in the flow-front progression as the initial
Discussion on this paper should be addressed to the Editorial Department, charge thickness was increased. This observation is not consis-
ASME, United Engineering Center, 345 East 47th Street, New York, N.Y. tent with the isothermal lubrication model, which predicts that
10017, and will be accepted until two months after final publication of the paper
itself in the JOURNAL OF APPLIED MECHANICS. Manuscript received by ASME the flow-front progression is independent of the instantaneous
Applied Mechanics Division, February 7, 1985. cavity thickness. The authors reasoned that this difference
PLAN VIEW
CHARGE
(a) (b)
Fig. 3 Black and white charges used to investigate the kinematics of
flow; (a) alternate black and white layers, and (b) separate black and
white regions (from Barone and Caulk, 1985)
;: == g
2 6 8 10 em
(a)
, t
. .:..: .. -
2 4 8 8 10 em
(b)
Fig. 4 Stages of deformation at two different closing speeds for a six-
layer charge with alternate black and white layers (from Barone and
Caulk, 1985)
I-,~,-:~~~
I
II
I~~ 2 4 6 6 10 em
[ Jrhl2
-1112
(J3/3 dX3]
,/3
+ [ - p + (J33]':ITzI2 = O. and the unit tangent vector Aa by vaA a = 0 and AaA a 1. On
a free boundary, we require that the stress resultant vanish and
The second of these is trivially satisfied by the symmetry of the that g(xa,t) represent the motion of a material surface. These
problem. With the additional definitions conditions may be expressed as
hl2 Jhl2 (-Po a/3+n a{3)v/3=O, }
p= J- hI2pdx3,na/3= (Ja/3 dX 3, on fa (15)
-hl2
g,,+vag,a=O.
(11)
On a fixed boundary, the normal velocity vanishes and we
fa = (Ja3)
x3~h/2
= (Ja31X3= -hl2 ' assume that the resultant shear caused by friction with the ver-
the first equation in (10) reduces to tical edge of the cavity may be neglected. These two conditions
are expressed by
V V
dafl= ( a,P + g,a) (17)
We now assume that nbb is affected by the presence of the
because the material remains isotropic in this plane.
fiber parallel to a in direct proportion to the product
Therefore, we abandon any three-dimensional description
of the compound and model it instead as a two-dimensional Oa.bl. (24)
extensible sheet undergoing a squeezing motion governed by Note that the value of (24) depends only on the direction and
(12), (13), and the boundary conditions (15) and (16). We not the sense of both unit vectors. This assumption implies
assume that nafi is a linear function of dafi, which, to be pro- that the fiber bundle contributes to the principal stress
perly invariant, must also be an isotropic function of dal). The response in proportion to: (f) the degree of relative motion be-
most general form for such a dependence is tween the fiber and the effective homogeneous continuum in
nali=h(\dyy5alj+2ndafl), (18) the direction parallel to the fiber, and (if) the cosine of the
angle between the fiber and the principal directions of the
where X and /J. are scalar coefficients and the cavity thickness h stress resultant n . In addition, we assume that this
afj
is introduced in (18) to give X and p units of viscosity. In mechanism accounts for the entire material response
general, X and JX depend on the transverse temperature represented by the constitutive equation (18).
distribution, which, because of the condition (3), is indepen-
We now return to the actual compound with its random
dent of xa. This temperature solution is available (Barone and
distribution of fibers and assume that the principal stress
Caulk, 1979), but we do not make explicit use of it in this
response nbb at any point xa is given by a simple average over
paper.
all directions a in the plane. Accordingly, we assume that
It is interesting to note that although the stress response for
an incompressible linear viscous fluid is expressed in terms of (25)
a single viscosity coefficient, the representation (18) for the
n
bb= \ndm \a-b\dO,
IT JO
two-dimensional stress resultant may in general have two. This
is because the planar dilatation daa is never zero during flow, a where
fact which follows from the incompressibility condition (13) a = cos0e [ + sin0e2, (26)
expressed as and a depends on the resin properties and the volume fraction
daa + h/h = Q. (19) of fiber.
Consider two simple cases for da$:
But this expression also implies that datX:ji = 0, so that when
the constitutive equation (18) is substituted in the equilibrium d$=-(h/2h)8aP, (27)
equation (12), the resulting expression
-P,a + iihva^ + 2fa=Q (20) h/h ; a = /3 = l,
(28)
does not include X. The only other place that X appears in the 0 ; a,i3^1,
governing equations is in the boundary condition (15), on the
which both satisfy the incompressibility condition (19). The
advancing flow front. With (18) and (19), this boundary con-
first is an equibiaxial deformation in the xx -x2 plane and the
dition reduces to
second is a one-dimensional deformation parallel to the x{
-(P+\h)+2nhdcll3vavIJ = 0 onr0. (21) axis. In both instances e, is a principal direction and from (22)
Now since (P + \h)a = P a , the coefficient X may be ab- and (26)-(28)
sorbed into the definition of the pressure resultant and the d$ = - (A/2/;) , d%> = - (h/h)cos2d. (29)
solution for va determined independent of X. One must still From a combination of (18) with (25)-(29), we obtain
reckon with X, however, since the other coefficient /* is deter-
mined by measuring the pressure exerted on the SMC during 2a r ""/2
flow. In the remainder of this section we motivate a relation- \V= (h/2h)cosd d6=-(\ + n)h, (30)
ship between X and JJ. by considering the principal mechanism w J o
2fv f xfl/ -2/2
that produces the resultant stress response na0.
nf,= (h/h)cosi6d6=-(\ + 2ix)h. (3D
Consider a single fiber bundle oriented parallel to a given 7T J o
direction in the *, - x2 plane specified by the unit vector a =
The first term in both (30) and (31) comes from the assumed
aaea. Now define response mechanism (25) and the second term from the con-
don adaaa (22) stitutive equation (18). By performing the integration in (30)
as the extensional rate of deformation parallel to the fiber. and (31), and then eliminating a between the two equations,
Next, consider the extensional component of the stress resul- we determine the simple relationship between X and /z:
tant along one of its principal directions specified by the unit X = 2^. (32)
vector b = ie, i.e.,
Substituting this relationship in the constitutive equation (18),
i = bnapbp- (23) we obtain
Thin Charge Approximation Coulomb Friction. In the absence of material resistance, the
governing equations for Coulomb friction become
To further compare the different constitutive assumptions
for the friction response, it is useful to consider the case when VP + 2(Kc/h)Pu = 0,
the material resistance may be neglected relative to friction. (62)
Vv + /i//i = 0.
Based on the nondimensional equations (46) and (51) in the
previous section, this approximation would be appropriate for By taking the curl of (62),, it is easy to show that
very thin charges. For simplicity, we also confine attention in VXu = 0 or M , 2 - M 2 , =0. (63)
this section to the case when the entire boundary is free. Let s be a measure of arc length along a streamline. Then since
Hydrodynamic Friction. Neglecting the material resistance, u is the unit tangent vector to the streamline,
the governing equations for hydrodynamic friction reduce to
~ = "cpUg = ue<au& = ( ) , = 0, (64)
VP + 2KH\ = 0,
(52) so that every streamline must be straight. From (62), the
pressure gradient is always parallel to the streamlines, and
subject to the condition P = 0 on the free boundary (2). Tak- since the boundary is a line of constant pressure, 5 the
ing the divergence of (52), and combining the result with (52)2 streamlines must intersect the boundary at right angles. Hence
yields for P the streamlines are uniquely determined by the geometry of
V2P=2nHh/h. (53) the region. The magnitude of the velocity along these
streamlines may be found by integrating (62)2. Again, the
Therefore, the pressure resultant satisfies Poisson's equation deformation history turns out to be independent of KC, h, and
and vanishes on the boundary. The velocity field can be com- h. The details of this analysis are given in the Appendix,
puted by substituting the solution for P into (52)!. It is in- where we again obtain an analytical solution for an ellipse. In
teresting to note that the specific values KH, h, or h do not af- this case, the result is simply
fect this velocity field apart from a change in time scale; hence
the deformation history resulting from the solution to (52) (ff 2 -& 2 ) 1 / 2 = a 2 e = constant. (65)
depends only on the initial geometry of the charge. By comparing (61) with (65), it is clear that Coulomb friction
Consider an elliptical charge with major and minor axes 2a causes the ellipse to approach a circle more rapidly than
and 2b, respectively. Then the initial boundary of the region is hydrodynamic friction.
specified by
4 - LAYER
I - LAYER
Fig. 9 Comparison of experimental flow-front progression ( ) in
one and four-layer elliptical charges with the analytical result ( ) ,
given by (61), which was obtained for hydrodynamic friction, neglecting
material resistance
Acknowledgment
Thanks are extended to Ronald Henderson from General
Motors Advanced Engineering Staff for preparing the charges
and molding the samples used in this study.
References
Barone, M. R., and Caulk, D. A., 1979, "Effect of Deformation and Ther-
moset Cure on Heat Conduction in a Chopped Fiber Reinforced Polyester Dur-
ing Compressing Molding," Int. J. Heat Mass Transfer, Vol. 22, pp.
1021-1032.
I - LAYER Barone, M. R., and Caulk, D. A., 1985, "Kinematics of Flow in SMC,"
Polymer Composites, Vol. 6, pp. 105-109.
Fig. 10 Comparison of the experimental flow-front progression Lee, C. C , Folgar, F., and Tucker, C. L., 1984, "Simulation of Compression
( ) in one and four-layer elliptical charges with the analytical Molding for Fiber-Reinforced Thermosetting Polymers," Journal of Engineer-
result ( ) , given by (65), which was obtained for Coulomb friction, ing for Industry, Vol. 106, pp. 114-125.
neglecting material resistance Lee, C. C , and Tucker, C. L., 1983, "A Simulation of Nonisothermal Com-
pression Molding," Society of Plastics Engineers ANTEC.
Lee, S. J., Denn, M. M., Crochet, M. J., and Metzner, A. B., 1982, "Com-
pressive Flow Between Parallel Disks: I. Newtonian Fluid with Transverse
material resistance, which should be appropriate, at best, for Viscosity Gradient," Journal of Non-Newtonian Fluid Mechanics, Vol. 10, pp.
very thin charges. The analytical solution (61) for 3-10.
hydrodynamic friction is compared to the experimental results Marker, L. F., and Ford, B., 1977, "Flow and Curing Behavior of SMC Dur-
in Fig. 9 and the corresponding solution (65) for Coulomb ing Molding," Modern Plastics, Vol. 54, pp. 64-70.
Naghdi, P. M., 1972, "The Theory of Shells and Plates," Handbuch der
friction is compared to the same results in Fig. 10. Since the Physik, Truesdell, C , ed., Springer Verlag, Berlin, pp. 425-640.
degree of approximation in these solutions is directly propor- Silva-Nieto, R. J., Fisher, B. C , and Birley, A. W., 1980, "Predicting Mold
tional to the value of h, we should see a relatively rapid con- Flow for Unsaturated Polyester Sheet Molding Compounds," Polymer Com-
vergence of the experimental results to the analytical solution posites, Vol. 1, pp. 14-23.
Truesdell, C , and Toupin, R. A., 1960, "The Classical Field Theories,"
as the material thickness is reduced by a factor of four. This Handbuch der Physik, Vol. III/A, Fliigge, A., ed., Springer Verlag, Berlin, pp.
seems to be true only for hydrodynamic friction (Fig. 9). 226-902.
Therefore, we conclude from this evidence that hydrodynamic Tucker, C. L., and Folgar, F., 1983, " A Model of Compression Mold Fill-
friction is the best assumption among the alternatives ing," Polymer Engineering and Science, Vol. 23, pp. 69-73.
considered.
Discussion A P P E N D I X
In this section we discuss the relationship between certain In this appendix, we determine the general solution to the
results obtained as part of this work and previous analysis of approximate equations (62), which are valid for Coulomb fric-
the flow of SMC by Tucker and Folgar (1983). These authors tion with negligible material resistance. We also obtain an ex-
modelled SMC as an isotropic Newtonian fluid using the usual plicit solution for an elliptical free boundary.
lubrication approximations, which neglect inertia and normal Recall that the streamlines corresponding to (62) must be
stresses. This approach gives a parabolic velocity profile straight and intersect the free boundary at right angles. Each
through the thickness of the cavity and yields for the govern- streamline emanates from a stagnation point inside the region
ing equations: located by its intersection with one or more other streamlines.
In general, the locus of these stagnation points will be a con-
. h2 nected curve which may have one or more branches.
V2p=12[ih/h3 , v= Vp, (66)
12ft To construct a general solution, we introduce a system of
\-KS2
This solution will not be valid for t > 0 unless the region re-
mains elliptical as the flow progresses. For this to be true, the
so that the incompressibility equation may be written in the boundary velocity 0417) must identically satisfy (15)2 when
form the major and minor axes extend consistent with 0417):
1 / h \ b2 ( h \ r 1 / b2 \~\
fl= 6= ( 18)
Vy + h/h=
dv
+
KV h
+- ^ =0 (-48) ~(x)T' -(T-)T~(-^-)J- ^
ds2 1 KS2 h This can be verified by a straightforward calculation. Now
or from (58) and (A1S)