You are on page 1of 23

Journal of

Composite Materials http://jcm.sagepub.com/

Process Modeling of Composite Materials: Residual Stress


Development during Cure. Part I. Model Formulation
S.R. White and H.T. Hahn
Journal of Composite Materials 1992 26: 2402
DOI: 10.1177/002199839202601604

The online version of this article can be found at:


http://jcm.sagepub.com/content/26/16/2402

Published by:

http://www.sagepublications.com

On behalf of:

American Society for Composites

Additional services and information for Journal of Composite Materials can be found
at:

Email Alerts: http://jcm.sagepub.com/cgi/alerts

Subscriptions: http://jcm.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
Citations: http://jcm.sagepub.com/content/26/16/2402.refs.html
2014
>> Version of Record - Jan 1, 1992

What is This?

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
Process Modeling of Composite
Materials: Residual Stress Development
during Cure. Part I. Model Formulation
S. R. WHITE
Department of Aeronautical
and Astronautical Engineering
University of Illinois at Urbana-Champaign
Urbana, IL 61801

H. T. HAHN
Department of Mechanical, Aerospace
and Nuclear Engineering
University of California at Los Angeles
Los Angeles, CA 90024
(Received July 1, 1991)
(Revised April 16, 1992)

ABSTRACT: In the production of composite parts the pertinent processing parameters


are time, temperature, and pressure. A judicious choice of these three parameters pro-
duces composites which are fully cured, compacted, and of high quality. Slight deviations
from the recommended processing conditions can result in unacceptable quality. One of
the most significant problems in the processing of composites is residual stresses.
Processing-induced residual stresses can be high enough to cause cracking within the
matrix even before mechanical loading. This microcracking of the matrix can expose the
fibers to degradation by chemical attack. Strength is adversely affected by residual stresses
since a pre-loading has been introduced.
The topics considered and discussion presented in this paper have been chosen to ad-
dress the issue of understanding how residual stresses develop during processing and how
they can be predicted. A process model has been developed which can be used to predict
the residual stress history during the curing of composite laminates. This model includes
the effects of chemical and thermal strains and assumes the material to exhibit linear, vis-
coelastic behavior. A phenomenological model is used to predict the degree of cure history
during the cure cycle. Mechanical properties are allowed to develop based on a functional
dependence on the cure state (degree of cure) and the transverse compliance is taken as the
only time-dependent compliance. Simultaneous application of the cure kinetics and a vis-
coelastic stress analysis yields the residual moments and curvatures for unsymmetric
cross-ply laminates. An experimental correlation is provided in an accompanying paper.

2402

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2403

KEY WORDS: composites, processing, curing/cure monitoring, polymers, manufactur-


ing/fabrication/processing, residual stresses, process modeling.

1. INTRODUCTION
N THE PRODUCTION of composite parts the pertinent processing parameters
Iare time, temperature, and pressure. A judicious choice of these three parame-
ters produces composites which are fully cured, compacted, and of high quality.
Slight deviations from the recommended processing conditions can result in
unacceptable quality.
The standard process cycle for polymer matrix composites (PMCs) is a two-
step cure cycle. In such cycles the temperature of the material is increased from
room temperature to the first dwell temperature and this temperature is held con-
stant for the first dwell period ( ~ I hour). Afterwards, the temperature is in-
creased again to the second dwell temperature and held constant for the second
dwell period (2-8 hours). After the second dwell, the part is cooled down to
room temperature at a constant rate. Since there are two dwell periods, this type
of cure cycle is referred to as a two-step cure cycle. The purpose of the first dwell
is to allow gases (entrapped air, water, or volatiles) to escape the matrix material
and to allow the matrix to flow, facilitating compaction of the part. Thus, the vis-
cosity must be low during the first dwell. Typical viscosity versus temperature
profiles of polymer matrices show that as the temperature is increased, the vis-
cosity of the polymer decreases until a minimum viscosity is reached. As the
temperature is increased further, the polymer begins to cure rapidly and the vis-
cosity increases dramatically. The first dwell temperature must be chosen judi-
ciously to allow the viscosity of the resin to be low while keeping the cure to a
minimum. Isothermal viscosity versus time profiles are useful in determining the
pot life of the polymer: the maximum length of time at a specific temperature for
the polymer to remain fluid-like. The first dwell time must be within the pot life
of the polymer at the dwell temperature.
The purpose of the second dwell is to allow crosslinking of the polymer to take
place. It is here that the strength and related mechanical properties of the com-
posite are developed. To characterize the exothermic cross-linking reaction of a
thermosetting polymer matrix, a thermal cure monitor technique such as isother-
mal Differential Scanning Calorimetry (DSC) is commonly used. A typical DSC
trace for a thermosetting polymer is shown in Figure 1. As the DSC trace ap-
proaches a flat line, the crosslinking reaction is nearing completion. As the tem-
perature is increased the reaction rate increases and the time to complete the
reaction decreases. Two competing priorities take place in the choice of the sec-
ond dwell temperature. First, a low temperature is desirable for ease of manufac-
turing and to reduce the effects of thermal mismatch between fiber and matrix
materials. Second, the processing time should be as short as possible for eco-
nomic considerations. Since low temperatures require longer dwell times, these
two concerns must be compromised.
Laminated composites are most commonly processed in an autoclave. In this
case vacuum and pressure are applied to the part during processing. Typically,

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2404

Figure 1. Typical DSC thermogram for a polymer matrix composite.

vacuum isapplied to the part during the initial dwell to help facilitate removal of
entrapped gases. Vacuum is discontinued after the increased viscosity of the
matrix nullifies any benefit from the pressure differential experienced by the
matrix. Thereafter, pressure is applied to help in consolidation and ensuring
fiber-matrix interaction. Pressure is removed either after significant crosslinking
or after completion of the process cycle. Certainly, the proper choice of tempera-
ture and pressure histories is vitally important in the production of high quality
composite parts.
One of the most critical parameters in the process cycle is the second dwell
temperature. Its choice is largely material dependent. A certain minimum tem-
perature must be reached before the crosslinking reaction begins. Demands for
increased performance have led to the development of several high-temperature
resins recently. These high-temperature resins retain good mechanical properties
at elevated service temperatures. Processing of these resins requires higher cure
temperatures than conventional epoxy-matrix composites. One of the problems
encountered when processing at higher temperatures is the inducement of resid-
ual stresses.
The constituent materials of a composite react differently to the changes in en-
vironmental conditions encountered during processing. Chemically, the reinforc-
ing fibers are affected very little during the process cycle. The polymer matrix on
the other hand will contract during crosslinking by as much as 6% [1] in thermo-
sets. As well as chemically-induced deformations, there are thermally-induced
deformations during processing. The reinforcing fibers show very little thermal
deformation during cooldown along the axis of the fiber due to their highly
oriented structure. The polymer matrix on the other hand has a higher thermal
expansion coefficient, typically an order of magnitude or more. After process-
ing, the composite must be well-bonded and continuous, therefore these deforma-

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2405

tions are balanced internally within the composite and residual stresses are
induced.
Processing-induced residual stresses can be high enough to cause cracking
within the matrix even before mechanical loading [2]. This microcracking of the
matrix can expose the fibers to degradation by chemical attack. Strength is
adversely affected by residual stresses since a &dquo;pre-loading&dquo; has been intro-
duced [3].
The effects of residual stresses can be seen visually in the production of com-
posite materials with unsymmetric layups. With a symmetric layup such as
[0/90]f the resulting laminate will remain flat after production since the in-plane
strains are balanced through the thickness. With an unsymmetric layup such as
[0/90] the laminate will experience out-of-plane deformations upon cooldown
due to the variation in strains across the thickness. This curvature development
can be directly related to the magnitude of the induced residual stresses. Other,
more exotic layups such as [0/60/90] will produce a spiral shape after produc-
tion.
The processing of composites at lower temperatures should reduce thermally
induced residual stresses. However, a certain minimum temperature must be
reached before crosslinking reactions can commence. Also, at lower tempera-
tures the cure reaction will take longer and may not proceed to completion. This
will not only adversely affect the mechanical performance of the material, but
also may not be economically feasible. Stress relaxation during the second dwell
period can reduce chemical shrinkage effects. The viscoelastic response of poly-
mers is greater at higher temperatures. However, these higher temperatures will
increase thermally induced residual stresses. Increased pressure during cool-
down could reduce thermal contraction and thus, reduce residual stresses. How-
ever, too much pressure could adversely affect the strength of the composite after
processing through pressure-induced damage. The optimization of processing
parameters is a complicated and interrelated problem and one in which little
research has been done. Once a process cycle is chosen the resulting changes in
mechanical properties must be monitored. Significant degradation of the proper-
ties of interest: strength, stiffness, fracture toughness, etc., would not be accept-
able.
It is clear that there is a need for the development of an optimization scheme
for the minimization of residual stresses in the processing of composite materials.
To accomplish this objective, a residual stress process model should incorporate
viscoelastic material response, chemical and thermal shrinkage effects, and
mechanical property development during cure. Previous works have been either
elastic models [3-9] or have restricted the analysis to the cooldown phase of the
cure cycle [10-13]. Additionally, most of the analyses [3-5,10-13] are limited to
either thermally induced strains solely, or chemically induced strains solely. Sev-
eral investigators [14-16] have tried to assess the relative importance of chemical
shrinkage in residual stress development. These works are largely experimental
in nature and utilize simple, elastic beam models for the most part. The analysis
techniques employed in these works tends to overpredict the influence of chemi-
cal strains in residual stress development. It is believed that chemical shrinkage

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2406

occurs when the matrix is highly viscoelastic and mechanical properties are quite
low. In this case their contribution to residual stresses would be negligible.
The topics considered and discussion presented in this paper have been chosen
to address the issue of understanding how residual stresses develop during pro-
cessing and how they can be predicted.
2. MODEL FORMULATION

2.1 Cure Kinetics


A kinetic model describes the changes of species composition with time under
the influence of temperature. In general, kinetic models may be phenomenologi-
cal or mechanistic. A phenomenological model captures the main features of the
reaction kinetics ignoring the details of how individual species react with each
other. Mechanistic models on the other hand are obtained from balances of spe-
cies involved in the reaction; hence, they are better for prediction and interpreta-
tion of composition. However, due to the complexity of thermosetting reactions,
phenomenological models are the most popular for these systems.
Phenomenological models for the reaction rate have the general formula [17]

where a is the fractional conversion of the reactive group (degree of cure), t is


the reaction time, K is a reaction rate constant, and f(a) represents some function
of a.
Most commonly K is considered to have an Arrhenius temperature depen-
dence :

In Equation (2), A is the frequency factor, R is the universal gas constant, DE is


the activation energy, and T is the absolute temperature.
As the chemical reaction progresses, the reaction rate decreases. Eventually,
the reaction rate approaches zero as the completion of the reaction nears. In the
absence of vitrification, the degree of cure will reach unity at the completion of
the reaction. With these concepts in mind, a functional form of f(a) is taken as

Combining Equations (1)-(3) yields the following general expression for reaction
rate:

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2407

This equation includes the particular cases of an 11 th order reaction model [17]
with g (a) = 1 and the autocatalytic model [17] with g (a) I + Ccx where C =

is the autocatalysis intensity. In general, the form of g(a) must be determined


from experimental data.
Various researchers have modeled the cure kinetics of thermosetting resins in
the past. Bogetti and Gillespie [6] modeled the response of an unsaturated poly-
ester resin as
. ,

Loos and Springer [18J showed that Hercules 3501-6 epoxy resin could be
modeled by

Here K,, Kz, and K, are Arrhenius rate constants given in Equation (2) with ap-
propriate activation energies and frequency factors. Another epoxy resin system,
Fiberite 976, was modeled by Dusi et al. [19] and by Han and Lem [20] as

Again, K, and K2 are Arrhenius rate constants.


In the present work, Equation (5) is used to describe the reaction rate. This
form of reaction rate is chosen to provide a proper balance between computa-
tional complexity and material versatility. Specific values of the parameters are
provided in a companion paper [21].
To obtain the cure kinetic parameters K, /12, and n, cure rate and cure state must
be measured simultaneously. This is most commonly accomplished by thermal
analysis techniques such as Differential Scanning Calorimetry (DSC). Isothermal
DSC testing has been widely used to model the cure kinetics of polymers
[17,19,20,22,23J. Several isothermal temperatures are used to develop the temper-
ature dependence of the kinetic parameters. At a constant cure temperature, the
degree of cure and cure rate are determined by DSC analysis. The kinetic param-
eters are then obtained for the cure temperature used. With the temperature de-
pendence of the kinetic parameters known, the degree of cure can be predicted
for any temperature history by integration of Equation (5).
In an isothermal DSC test the heat flow to or from the sample is monitored over
time (Figure 2). As the crosslinking reaction commences the DSC monitors the
release of energy and the heat flow is mapped in the exothermic region. After a

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2408

Figure 2. Isothermal DSC data reduction.

short period of time the heat flow reaches its maximum and begins to decrease
gradually until reaching the baseline (zero heat flow) upon completion of the
reaction. The total area under the exotherm is the heat of reaction at the test tem-
perature, HT . After completion of the isothermal cure, the sample is dynamically
scanned from room temperature to the highest cure temperature and the resulting
exotherm measured to obtain the residual heat of reaction, //R. The summation
of HT and HR is the ultimate heat of reaction, H~ . If the isothermal cure tempera-
ture is high, then the residual heat of reaction will be small and HT is approx-
imately equal to H. The final degree of cure at the end of the isothermal cure
is found from

Similarly, the area under the exotherm up to time t * is H and the degree of cure
at time t * is

To obtain the cure rate at any time, the height of the exotherm from the baseline,
dHldt, is divided by the ultimate heat of reaction:

For isothermal DSC testing Equations (9) and (10) are used in data reduction

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2409

to produce degree of cure and cure rate histories. With these histories, solution
of Equations (2) and (5) yields the kinetic parameters of interest. The degree of
cure at any time given a prescribed temperature history may be obtained by inte-

grating Equation (5):

~
2.2 Residual Stress Model ,

Residual stresses have been experimentally investigated in the past [12,14,24-


27]. These works have used unsymmetric cross-ply laminates which are known
to develop curvature after processing. This curvature is directly related to the
level of residual stresses in the laminate. For the purpose of providing experimen-
tal correlation with the model output, the curvature development in unsymmetric
cross-ply strips is analyzed. In the elastic analysis the relationship between resid-
ual stress and curvature is developed. Subsequently, a viscoelastic model is de-
veloped based on the elastic solution.
2. 2. I ELASTIC ANALYSIS
The elastic stress-strain relations for orthotropic laminae under plane stress
conditions are

where v; are the stresses, E; are the strains, and Q;; are the lamina stiffnesses. In
Equation (12) the non-mechanical strains, e; , consist of chemical strains, e,‘. , and
thermal strains, e5

The total strains Ej at position z from the mid-plane can be found from the mid-
plane strains, 6j, and curvatures, k; , by

For the cross-ply [0&dquo;/90&dquo;]T laminate in Figure 3, the longitudinal and transverse
stresses in the individual plies are given by the laminated plate theory as [28]

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2410

Figure 3. Geometry of a cross-ply laminated plate.

In Equation (15) the superscripts 0 and 90 refer to the 0° and 90° plies, and the
overbar represents the material symmetry axis system. The subscripts 1 and 2
refer to the x, and X2 axis directions as depicted in Figure 3.

2.2.1.1 Anticlastic Curvature


For the freely warping laminate of Figure 3, the mid-plane strains and curva-
tures are calculated by requiring the stress resultants (N,,Nz) and the bending
moments (MI, M2) to be zero in the laminated plate theory [28] . The resulting
curvatures are given by

where

The major Poisson’s ratio is v,2 and E E22/E&dquo; is the ratio of the transverse
=

modulus to the longitudinal modulus. Equation (16) gives the induced curvature
for anticlastic deformation of an unsymmetric cross-ply laminate. The curvature
is dependent on the thermal and chemical strain mismatch (e, ~2), lamina -

mechanical properties (v,z, E) and the half thickness, h.

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2411

2.2.1.2 Cylindrical Curvature


Equation (16) is valid when the overall dimensions of the laminate are small so
that the deflections are small. However, for large laminates the deflections may
become too large to satisfy the assumptions of linear theory. In this case a large-
deflection plate theory must be used [24-26]. For long, narrow strips, the small-
deformation laminated plate theory can still be used with the modified condition
k2 = 0 in place of M2 = 0. The result is

where

To account for small thickness variations between different specimens, the


dimensionless curvature, he k~, is used. Note that he k~ depends directly on
E(e, - e2). Typically, E - 0.05 and Vl2 - 0.35 for PMCs. Thus, K~ is approx-
imately 0.6 and the dimensionless curvature is proportional to E(e, - e,).
2.2.1..3 Residual Moment Analysis
During the processing of composite materials in a hot press or an autoclave, the
laminate is usually kept flat until cure is complete. If the platen surfaces are
assumed frictionless, the effect of the constraints is to require that the curvatures
k, and k2 be zero. To develop the elastic solution under these constrained condi-
tions, the laminated plate theory may be used with conditions of N; 0 and =

ki = 0. The resulting mid-plane strains are given by

With the strains determined throughout the laminate the moment resultants can
be calculated to be

2.2.2 VISCOELASTIC STRESS ANALYSIS


Polymer materials are known to exhibit time-dependent mechanical behavior
especially at high temperature. It has been proposed that the chemical shrinkage
strains developed during curing of polymer matrix composites can be neglected
since they occur at high temperature when the stress relaxation effects are signifi-
cant [4]. For linear thermorheologically simple materials a single temperature-
dependent shift factor, aT(T ), can be used to predict the transient thermal re-
sponse [13] .

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2412

For viscoelastic materials, mechanical response is history dependent and in-


volves the use of reduced times, ~(t) and ~(7-). These reduced times can be found
from the shift factor as

Some authors have modeled the shift factor considering temperature dependence
only [10-13,29-31]. In their models either a modification of the Williams-Landel-
Ferry (WLF) equation is used, or a best fit to the data is generated according to
the general form of the equation:

In Equation (23), B, and B2 are constants to be determined from experimental


correlation.
During the process cycle the cure state and, consequently, the microstructure
of the material changes. Thus, the time-dependent mechanical properties must be
determined as a function of degree of cure. In the present study, the shift factor
is modeled according to Equation (23). Furthermore, the quasi-elastic method as
developed by Schapery [32] is used in the development of the viscoelastic resid-
ual stress model. The use of the quasi-elastic method is motivated by the fact that
the relaxation moduli are required in the viscoelastic analysis of residual stresses,
whereas the experimental characterization of composite materials is usually in
terms of the creep compliances. An excellent account of the development of the
quasi-elastic method is given in Reference [33]. The underlying restriction in the
application of the quasi-elastic method is that the compliance response of the
material shows little curvature when plotted versus log time [34,35]. Mathemati-
cally, this restriction stems from the approximation of the Laplacian weighting
function with the delta function 6(s - so) in the direct inversion method. The
spread of the weighting function (about 2 decades) means that if the compliance
response shows large curvature (over 2 decades), then the errors introduced by
approximating the weighting function by the delta function will be significant.
Harper [33] shows excellent agreement between the quasi-elastic method and
direct inversion for AS-4/3501-6 graphite/epoxy composite. For most graphite/
thermoset systems, the restrictions imposed by the quasi-elastic method are
satisfied.
When the quasi-elastic method is used, the viscoelastic resultant moment can
be approximated by substituting the time-dependent stiffnesses for elastic stiff-
nesses in Equation (21) and making use of the convolution integral. The results
are

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2413

where

Since Equation (24) is a history-dependent integral, the degree of cure, a(T), is


represented as a function of T and not the current time, t. Equation (24) is solved
by discretizing the time domain into n equal portions ~1t with the initial time
t, = 0 and the current time t&dquo; = t: ?

In A e is defined as the strain difference (e, - ez). From Equation


Equation (27),
(27) the resultant is seen to be history dependent. To solve this equation
moment
the time-dependent mechanical properties in F(cx. t) must be determined. Addi-
tionally, the non-mechanical strain histories e,(t) and e2(t) must be calculated.
2.3 Curvature-Moment Relations
Once the viscoelastic moment history has been determined, the final moment
value may be applied elastically as the platens open (hot pressing) or as the ap-
plied pressure is released (autoclave processing) to find the resulting cross-ply
curvature. This curvature can be measured experimentally after processing.
Thus, experimental correlation necessitates a relation between the residual mo-
ments in Equation (26) and cross-ply curvatures.
The elastic mid-plane strains and curvatures are related to the resultant forces
(N; ) and moments (M, ) through the laminate compliance matrix:

Since N; = 0 and M( = - M2 during cure, the pertinent compliances are 8&dquo;1


and 612. The 6 compliance matrix can be found from [28]

where a =
[A]-’ is the inverse of the in-plane stiffness matrix, D is the flexural
stiffness matrix, and B is the coupling modulus matrix. With knowledge of the

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2414

a, B, and D matrices after processing, residual curvatures can be determined by


applying the final residual moments from Equation (26) in Equation (28).
2.4 Mechanical Property Modeling
For accurate representation of the mechanical response of a polymeric com-
posite material during processing, the change in mechanical properties with cure
state must be ascertained. Additionally, the time-dependent behavior of the lami-
nate must be understood to model the viscoelastic residual stress development
during cure.
It is known that the only compliance which exhibits significant time-dependent
behavior is the transverse compliance, S22(t) [10-12,36]. Furthermore, the trans-
verse compliance can be modeled by a power law equation of the form [10,12]

The creep parameters D and q can be obtained through transverse creep experi-
ments. The initial compliance S22i is the elastic response of the material and can
be found from static tension testing at several cure states. In general, the creep
parameters D, q, and a, may be dependent on the cure state of the material. As
a first approximation the creep parameters are represented as linear functions of
the degree of cure, a..

The instantaneous response of the material is also dependent on the cure state.
The transverse modulus increases quickly after a certain minimum degree of
cure, a*, before asymptotically approaching a fully cured value. As such the
transverse modulus dependence on degree of cure is modeled according to Equa-
tion (33).

The longitudinal modulus and major Poisson’s ratio are assumed to be linearly
dependent on degree of cure as indicated from experimental testing [21].

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2415

With knowledge of the cure state of the material the mechanical properties
El,, Ezz, P12, D, and q can be found by solution of Equations (31) through (35).
The elastic compliances can now be found from

The time-dependent stiffnesses in Equation (25) can then be obtained


through in-
version of the compliance matrix following the quasi-elastic approximation:

2.5 Strain History Modeling


From Equation (13) the laminate non-mechanical strains are composed of
chemical strains associated with the chemical shrinkage during the crosslinking
reaction and thermal strains resulting from the thermal expansion or contraction
of the material. The longitudinal chemical strains are taken to be zero since the
fibers do not experience chemical strains during the process cycle. The transverse
chemical strains are influenced by the matrix shrinkage associated with the cross-
linking reaction. The matrix chemical shrinkage strains have been modeled
assuming a linear dependence on degree of cure [6]. Experimental results [21]
showed that a more accurate representation of the development of transverse
chemical strains during the cure cycle could be obtained from an exponential de-
pendence on degree of cure:

The final transverse chemical shrinkage strain is e2 and a&dquo; is the degree of cure
when chemical shrinkage is complete. The parameters (3; are empirical constants
obtained from chemical shrinkage characterization tests.

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2416

The thermal strains can be modeled using the longitudinal and transverse ther-
mal expansion coefficients. From experimental testing [21] these coefficients were
found not to be dependent on degree of cure. Therefore, the thermal strains are

where a; are the thermal expansion coefficients and To is the initial stress-free
temperature. The strain history during curing is calculated by substituting Equa-
tions (42) and (43) into Equation (13) and utilizing the temperature and degree of
cure histories. The dependence of transverse thermal expansion coefficient on
matrix cure state was not evidenced in the experiments and, therefore, was not in-
cluded in the analysis. This restriction may not be true for all material systems.

3. LAMCURE MODEL
The cure kinetics model and residual stress model are combined in an inte-
grated cure model designated LamCure. A flowchart of the model is presented in
Figure 4. The input module allows for user input of the temperature history and
selection of material systems. Number of dwells, dwell temperature, dwell times,
and cooling/heating rates are designated for the cure cycle investigated.

Figure 4. LamCure process model flowchart.

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2417

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2418

Next, the kinetics model is solved using a numerical integration proce-


cure
dure. Simpson’s Rule is used to solve Equations (5) and (11). From the prescribed
temperature history, a degree of cure history is generated. This degree of cure
history is then used as input for the mechanical property and strain history mod-
eling equations.
In the third module the residual moment modeling equations are solved by a
finite difference scheme according to Equation (27). The time-dependent me-
chanical properties are solved at each time step according to Equations (30)-(41).
The strain history is generated from Equations (42) and (43). The reduced times
are obtained from Equations (22) and (23). Finally, with reduced times, strain
histories, and stiffness histories, the residual moment history is calculated at in-
cremental time steps throughout the cure cycle from Equation (27). A table of the
required model input is summarized in Table 1.
4. CONCLUDING REMARKS
The problem of modeling the processing environment in composite materials
is a complex one. This complexity can be reduced by judicious choice of assump-
tions and simplifications. The model presented should be viewed as a tool to
understand how residual stresses are developed during processing and used in in-
vestigations of cure cycle modifications and optimization. The model is restricted
to thin laminates. This arises from the inherent assumption that the temperature
field is uniform though the thickness of the composite plate in the solution of the
cure kinetics equations. Additionally, the heat source should be of a uniform
nature as in the radiant heating of composites in an autoclave environment or for
hot presses where the platen surface area is sufficiently large in comparison to the
composite plate dimensions. The matrix system is restricted to thermosetting
polymers systems in which the cure kinetics can be modeled by an Arrhenius
type equation. Most epoxy and BMI structural composites can be modeled as
such. Additionally, the material should obey time-temperature superposition and
thus be thermorheologically simple. Linear viscoelastic behavior is assumed. All
assumptions inherent in lamination theory are inclusive in the model. Addition-
ally, moisture desorption during processing has not been considered.
Experimental verification of the model is presented in an accompanying paper
in which 8-ply unsymmetric cross-ply laminates are processed and tested. The
material system chosen is a graphite/BMI thermosetting composite system.
Model predictions for residual moment history and degree of cure history are
compared to experimental measurements.

NOMENCLATURE
.

[a] in-plane compliance matrix


ao, aI, a2 initial transverse modulus modeling parameters
aT(T) shift factor
A frequency factor
[B] coupling modulus matrix
Bl, B2 shift factor modeling parameters

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2419

[D] flexural stiffness matrix


Di uncured transverse creep coefficient
Df fully cured transverse creep coefficient
ej non-mechanical laminate strains
e’ laminate chemical strains
eTT laminate thermal strains
ez final transverse chemical shrinkage strain
E ratio of the transverse modulus to the longitudinal modulus (E,,IE,,)
E* uncured transverse modulus
EUf fully cured longitudinal modulus .

E,,; uncured longitudinal modulus


E,,i initial transverse modulus
F(a, t) time- and cure-dependent mechanical properties in calculation of re-
sidual moments
h laminate half thickness
HR residual heat of reaction
HT heat of reaction at an isothermal temperature, T
Hu ultimate heat of reaction
k~ laminate anticlastic curvature
k~ laminate cylindrical curvature
kj laminate curvatures
K Arrhenius kinetic constant
Ka mechanical property constant (for anticlastic curvature)
K~ mechanical property constant (for cylindrical curvature)
m cure kinetics exponent

Mi bending moments
n cure kinetics exponent
N; stress resultants
q~ fully cured transverse creep exponent
qi uncured transverse creep exponent
Q;; in-plane stiffness
R universal gas constant
Ra reaction rate
Sii longitudinal compliance
S12 shear compliance
longitudinal
S22(t) transverse compliance
S,,i initial transverse compliance
t time
t,I initial time
tn current time
T temperature
To initial stress-free temperature
a degree of cure
[a]] generalized in-plane compliance matrix
a* degree of cure at initial transverse modulus development
a
c
degree of cure when chemical shrinkage is complete

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2420

a; thermal expansion coefficients


[{3]] generalized coupling compliance matrix
Qi transverse chemical shrinkage strain modeling coefficient
{32 transverse chemical shrinkage strain modeling exponent
[6] generalized flexural compliance matrix
De non-mechanical strain difference (e, - e2)
AE activation energy
Att time increment
E; laminate strains
EJ laminate mid-plane strains
~ reduced time
VI2 major Poisson’s ratio
V12¡ uncured major Poisson’s ratio
VI2f fully cured major Poisson’s ratio
a; laminate stresses

Subscripts
0 0° plies
90 90° plies
1 x, coordinate direction
2 x2 coordinate direction
i uncured
f fully cured
ACKNOWLEDGEMENT
This research was supported by the Air Force Office of Scientific Research
under Grant AFOSR-87-0242 with Lt. Col. George Haritos as program director.
The support of the AFOSR is greatly appreciated.

REFERENCES
1. Yates, B., B. A. McCalla, L. N. Phillips, D. M. Kingston-Lee and K. F. Rogers. 1979. "The
Thermal Expansion of Carbon Fibre-Reinforced Plastics. Part 5: The Influence of Matrix Cur-
ing Characteristics," Journal of Materials Science, 14:1207.
2. Doner, D. R. and R. C. Novak. 1969. "Structural Behavior of Laminated Graphite Filament
Composites," 24th Annual Technical Conference, SPI, Paper
Inc., No. 2-d.
3. Hahn, H. T. 1984. "Effects of Residual Stresses in Polymer Matrix Composites," Journal of
Astronautical Sciences, 32(3):253.
4. Hahn, H. T. and N. J. Pagano. 1975. "Curing Stresses in Composite Laminates," Journal of Com-
posite Materials, 9:91.
5. Pagano, N. J. and H. T. Hahn. 1977. "Evaluation of Composite Curing Stress," Composite Mate-
rials : Testing and Design (4th Conf.), ASTM STP 617, p. 317.
6. Bogetti, T. A. and J. W. Gillespie, Jr. "Processing-Induced Stress and Deformation in Thick-
Section Thermosetting Composite Laminates," CCM Report 89-21, University of Delaware,
August, 1989.
7. Bogetti, T. A. and J. W. Gillespie, Jr. "Two-Dimensional Cure Simulation of Thick Thermoset-
ting Composites," CCM Report 89-23, University of Delaware, August, 1989.

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2421

8. Bogetti, T. A. and J. W. Gillespie, Jr. 1990. "Residual Stress and Deformation in Thick Lami-
nated Composites Undergoing Chemical Hardening and Shrinkage," Proceedings of the 45th SPI
Conference, Washington, DC, February, 1990.
9. Bogetti, T. A., J. W. Gillespie, Jr. and R. L. McCullough. 1989. "Influence of Processing on the
Development of Residual Stresses in Thick Section Thermoset Composites," Proceedings of the
36th Sagamore Army Materials Research Conference, Plymouth, MA, October 23-26, 1989.
10. Weitsman, Y. 1979. "Residual Thermal Stresses Due to Cooldown of Epoxy-Resin Composites,"
Journal of Applied Mechanics, 46:563.
11. Weitsman, Y. 1980. "Optimal Cooldown in Linear Viscoelasticity," Journal of Applied Mechan-
ics, 47:35.
12. Harper, B. D. and Y. Weitsman. 1981. "Residual Thermal Stresses in an Unsymmetric Cross-Ply
Graphite/Epoxy Laminate," American Institute of Aeronautics and Astronautics, Paper 81-0580.
13. Harper, B. D. and Y. Weitsman. 1985. "On the Effects of Environmental Conditioning on Resid-
ual Stresses in Composite Laminates," International Journal of Solids and Structures, 21:907.
14. Harper, B., D. Peret and Y. Weitsman. 1983. "Assessment of Chemical Cure-Shrinkage Stresses
in Two Technical Resins," AIAA/ASME/ASCE/AHS 24th Structures, Structural Dynamics and
Materials Conference, Lake Tahoe, NV, May 2-4, 1983, AIAA-83-0799-CP.
15. Penn, L. S., C. T. Chou, A. S. D. Wang and W. K. Binienda. 1989. "The Effect of Matrix
Shrinkage on Damage Accumulation in Composites," Journal of Composite Materials, 23:570-
586.
16. Hodges, J., B. Yates, M. I. Darby, G. H. Wostenholm, J. F. Clemmet and T. F. Keates. 1989.
"Residual Stresses and the Optimum Cure Cycle for an Epoxy Resin,"
Journal of Materials Sci-
ence, 24:1984.

17. Gonzalez-Romero, V. M. and N. Casillas. 1989. "Isothermal and Temperature Programmed Ki-
netic Studies of Thermosets," Polymer Engineering and Science, 29:295.
18. Loos, A. C. and G. S. Springer. 1983. "Curing of Epoxy Matrix Composites," Journal of Com-
posite Materials, 17:135.
19. Dusi, M. R., W. I. Lee, P. R. Ciriscioli and G. S. Springer. 1987. "Cure Kinetics and Viscosity
of Fiberite 976 Resin," Journal of Composite Materials, 21:243.
20. Han, C. D. and K.-W. Lem. 1983. "Chemorheology of Thermosetting Resins. I. The Chemo-
rheology and Curing Kinetics of Unsaturated Polyester Resin," Journal of Applied Polymer Sci-
ence, 28:3155.

21. White, S. R. and H. T. Hahn. 1992. "Process Modeling of Composite Materials: Residual Stress
Development during Cure. Part II. Experimental Validation," Journal of Composite Materials,
26(16):2423-2453.
22. Lee, W. I., A. C. Loos and G. S. Springer. 1982. "Heat of Reaction, Degree of Cure, and Viscos-
ity of Hercules 3501-6 Resin," Journal of Composite Materials, 16:510.
23. Ng, H. and I. Manas-Zloczower. 1989. "Kinetic Studies of a Composite Thermoset Cure Reac-
tion: Application in Pultrusion Simulations," Polymer Engineering and Science, 29:302.
24. Hyer, M. W. 1981. "Some Observations on the Cured Shape of Thin Unsymmetric Laminates,"
Journal of Composite Materials, 15:175.
25. Hyer, M. W. 1981. "Calculations of the Room-Temperature Shapes of Unsymmetric Laminates,"
Journal of Composite Materials, 15:296.
26. Hyer, M. W. 1982. "The Room-Temperature Shapes of Four-Layer Unsymmetric Cross-Ply
Laminates," Journal of Composite Materials, 16:318.
27. Hahn, H. T. 1981. "Warping of Unsymmetric Cross-Ply Graphite/Epoxy Laminates," Composites
Technology Review, 3:114.
28. Tsai, S. W. and H. T. Hahn. 1980. Introduction to Composite Materials. Lancaster, PA: Tech-
nomie Publishing Co., Inc., pp. 229-232.
29. Yeow, Y. T., D. H. Morris and H. F. Brinson. 1979. "Time-Temperature Behavior of a Unidirec-
tional Graphite/Epoxy Composite," in Composite Materials: Testing and Design (Fifth Confer-
ence), ASTM STP 674, S. W. Tsai, ed., American Society for Testing and Materials, p. 263.

Downloaded from jcm.sagepub.com at Scientific library of Moscow State University on February 11,
2014
2422

30. Wang, A. S. D., E. J. McQuillen and A. S. Ahmadi. 1975. "Analytical and Experimental Investi-
gation of Time-Temperature Creep of Graphite-Epoxy Composite Laminates," U.S. Naval Air
Systems Command Tech. Report.
31. Plazek, D. J. and I.-C. Choy. 1989. "The Physical Properties of Bisphenol-A-Based Epoxy Resins
during and after Curing. II. Creep Behavior above and below the Glass Transition Temperature,"
Journal of Polymer Science: Part B: Polymer Physics, 27:307.
32. Schapery, R. A. 1965. "A Method of Viscoelastic Stress Analysis Using Elastic Solutions," Jour-
nal of the Franklin Institute, Vol. 279.
33. Harper, B. D. 1983. "On the Effects of Post Cure Cool Down and Environmental Conditioning
on Residual Stresses in Composite Laminates," Ph.D. thesis, Texas A&M University, College
Station, TX, August.
34. Schapery, R. A. 1974. "Viscoelastic Behavior and Analysis of Composite Materials," in Mechan-
ics of Composite Materials, Vol. 2, G. P. Sendeckj, ed., Academic Press.
35. Cost, T. L. 1964. "Approximate Laplace Transform Inversion in Viscoelastic Stress Analysis,"
AIAA Journal, 2:2157.
36. Amijima, S. and T. Adachi. 1982. "Effect of Time on the Mechanical Behavior of Laminated
Composites," in Progress in Science and Engineering of Composites, T. Hayashi, K. Kawata and
S. Umekawa, eds., ICCM-IV, Tokyo, p. 811.

You might also like