Accepted Manuscript

closed form expression for residual stresses and warpage during cure of com-
posite laminates

M. Abouhamzeh, J. Sinke, K.M.B. Jansen, R. Benedictus

PII: S0263-8223(15)00644-3
DOI: http://dx.doi.org/10.1016/j.compstruct.2015.07.098
Reference: COST 6679

To appear in: Composite Structures

Please cite this article as: Abouhamzeh, M., Sinke, J., Jansen, K.M.B., Benedictus, R., closed form expression for
residual stresses and warpage during cure of composite laminates, Composite Structures (2015), doi: http://
dx.doi.org/10.1016/j.compstruct.2015.07.098

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 CLOSED F ORM EXPRESSION FOR R ESIDUAL STRESSES AND
WARPAGE DURING CURE OF COMPOSITE LAMINATES
M. Abouhamzeh a, 1, J. Sinke a , K.M.B. Jansen b , R. Benedictus a
a
Faculty of Aerospace Engineering, Delft University of Technology, Kluyerweg 1, 2629 HS Delft, The Netherlands

b
Faculty of Industrial Design Engineering, Delft University of Technology, Landbergstraat 15, 2628 CE Delft, The Netherlands

Abstract
In this paper, an analytical model is presented to predict the warpage and residual stresses
developed during curing of Fibre Metal Laminates (FMLs). The model is generally applicable
to other composite materials. The change of stiffness (E modulus) and also the chemical
shrinkage of the material during cure is implemented in the model. The developed closed
form solution is based on the Classical Laminate Theory (CLT) and an additional term
depending on the cure shrinkage. The development of residual stresses can be predicted
during cure for both free and constrained (moulded) cure of composite laminates. The input
material parameters are determined and used in the model to predict the elastic response of
FMLs upon manufacturing. Predictions are compared to simulations with experimental data
already obtained for this kind of material. The needed improvements for prediction of the
residual stresses and final distortions of FMLs are outlined as the future works.

Keywords: Composites, Fibre Metal Laminates, GLARE, Cure shrinkage, Warpage,

Residual stress

1 Introduction

Fibre Metal Laminates (FMLs) are hybrid materials consisting of alternating metal and
composite layers. FMLs have found applications in structural parts of aircraft like fuselage
panels and leading edges of tail planes. In recent years, extended research is done on design
aspects of FMLs [1-5]. However, the effects of curing processes are not yet predicted for
these materials. Fuselage panels made of FMLs encounter deviations in shape and residual

1
Corresponding author: Tel.: +31 15 2788673; Fax: +31 15 2781151.
E-mail address: [email protected]; [email protected];
stresses after cure and post-cure processes. The shape deviations need to be compensated
during assembly and the residual stresses produced in manufacturing and assembly may
reduce the mechanical performance regarding the load capacity, fatigue, residual strength and
damage tolerance of the component. A predictive model is needed to adapt the mould to
produce accurate panels. In the first phase, primary modelling and experiments on FMLs with
non-symmetric lay-up of prepreg layers were performed from which the results are already
published in previous papers [6,7]. The research methodology was also presented in ECCM16
[8].

The cure process of thermosetting polymers induces residual stresses and warpage when
used in layered materials with different properties. These cure-induced stresses add up to the
well-known thermal stresses arising during cooling down of the composite. An indication of
these cure induced stresses is the observation that the bending strain free temperature of a
non-symmetric FML always deviates considerably from the cure temperature [7]. Therefore, a
better prediction model is needed to include both thermal and cure stresses. Jansen et al.
[9,10] have developed an analytical approach for determining the warpage of electronic
packages as well as thermoset coating layers on substrates. They included the cure strain
together with the evolution of the epoxy E modulus in the analysis of materials with isotropic
layers.

The analytical solution presented here is generally applicable to any thermoset composite
material. The material can be layered made of isotropic or anisotropic layers. It is assumed
that the composite is cured above its glass transition temperature and therefore it is in its
rubbery state with negligible viscoelastic relaxation during curing. Modelling is based on
Classical Laminate Theory (CLT) incorporating the effects from the whole cure cycle
including curing and cool-down parts. The evolution of the material properties and the
developing cure shrinkage are included during the curing of a composite laminate. The
stiffness of the prepreg layers is presented as a function of the degree of cure. Therefore, the
residual stresses are obtainable during cure of any composite laminate.

After presenting the analytical model developed for this purpose, the input material
parameters are derived for glass fibre reinforced aluminium laminates (GLARE) which is
used in this study.

For the purpose of predicting the residual stresses after cure cycles of GLARE panels, the
material properties of the constituents were needed to be determined. Some material
characterization procedures have already been followed. Thermo-Mechanical Analysis
2
(TMA) is carried out on the epoxy material (FM-94) and time/temperature dependency of the
stiffness parameters are determined using Dynamic-Mechanical Analysis (DMA).
Accordingly, the thermo-elastic and thermo-viscoelastic response of the polymeric part of
GLARE has been already determined [11].

The curing process of GLARE products consists of three stages: heating to the cure
temperature, isothermal curing and cool-down to ambient temperature. During cool down
process in composite laminates, contraction or shrinkage occurs in all constituents of the
material. Besides the major part of residual stresses that are produced during the cooling part
of the cure cycle, residual stresses are developed during cure (before cool-down) due to the
“curing effects” including:

- Evolution (increase) of the epoxy stiffness with degree of cure

- Cure (chemical) shrinkage of the epoxy that in the fibre direction is restrained by the
fibres but will produce strains in the directions normal to the fibres

The focus of this paper is on the solution to estimate the residual stresses and distortion
during the cure cycle of FMLs considering the above mentioned curing effects. Results are
formulated for both free curing (without mould) and constrained curing in a vacuum bag on a
mould in an autoclave. The analysis presented in this paper is a novel approach on FMLs
which can be extended to any composite laminate. The final curvature of some non-
symmetric panels are calculated. The accuracy of the model predictions are shown by
comparing with numerical FE simulations. Discussions are given on the residual stresses and
warpage of non-symmetric laminates describing the necessity of more accurate prediction of
residual stresses during cool-down. Further improvements of the model are also mentioned
and discussed.

2 Theoretical formulation

The formulation given in this section applies to thin orthotropic laminates based on CLT
assumptions [12] useful for thin composite laminates. Two dimensional (2D) constitutive
equations for an elastic orthotropic material are written as:

      

  = 
 ∙ 〈
  −
  −
  〉
   

3
 Strain changes due to change in temperature and degree of cure. At the same time, the E-
modulus of the matrix evolves and the stiffness components of the composite material
increase. In order to incorporate the cure shrinkage effects, the analysis is first performed for
incremental strain steps dε and subsequently integrated to allow both an increasing stiffness
Qij of the different composite layers and an increasing cure shrinkage strain dεcure.
Incremental strains are assumed to change linearly through the thickness:

  = ̅ + ! ∙ "

(1)

̅ is the vector of incremental midplane strains and " is the vector of curvature
components of the laminate.

Force resultants can be written as:

#  = # . + # %  
# %   = # &. + # 
(2)

Where force resultant # %   is due to thermal deformation and cure (chemical) shrinkage
of the thermoset resin.

2.1 Case 1: free curing

Here, it is assumed that the laminate is cured without any constrained, so no mould is used
and the laminate is free to deform during curing of the resin, adhesive, etc.

Equilibrium of forces requires that the incremental mechanical force resultant be zero during
cure:

# . = ∑56 (
+78 1
12
*
[ +
] 
-̅ − . ° /+ 0
! +
234 (3)

[* + ][! − !9 " ]! = 0

12
∑56
+78 (1
234

. ° / is incremental initial strain due to cure and/or temperature change and [*+ ] is the
+

transformed stiffness matrix for layer k (see Figure 1). The parameter !9 is the distance
between the neutral plane and the midplane. NL is the number of layers and Zk-1 and Zk denote
the z-coordinate of the bottom and top surface of each layer k from the midplane surface of
the laminate.

4
 Figure 1: Layup of the composite laminate

Eq. (3) should be valid independently for any arbitrary curvature " . Therefore, the two
summation terms in Eq. (3) must both vanish:

56 12
+
; < [* + ] -̅ − . ° / 0 ! = 0
+78 1234

56 12

; < [*+ ]! − !9 ! = 0

+78 1234

Accordingly, the midplane strains ̅ and the distance !9 can be determined as:

2
∑@A *2
2B4[= ].>? / ∙2
̅ =
°

∑@A *2
2B4[= ]∙2
(4)

D2
∑@A
2B4 (D [=* 2 ]C>C 4 @A
∑ [=* 2](C2EGC234
E
)
!9 = @A
234
∑2B4[=* 2]∙2
= E 2B4
∑@A *2
=
2B4[= ]∙2
(5)

∑@A *2
2B4[= ] 1E2
∑@A *2
2B4[= ]∙2

where hk is each layer thickness. In Eq. (5) and the equations below, we define the
parameters IJ+ and IK+ as:

IJ+ = !+ J − !+G8 J

IK+ = !+ K − !+G8 K

5
 In a similar manner, the equilibrium of incremental mechanical moment resultant can be
written as:

L. = ∑56 (
IM
*+
+78 I [ ] -̅ + (! − M−1 (6)
+
!9 )"−. °/ 0 !! = 0

Accordingly, the incremental curvature " can be obtained, using the already determined
values for ̅ and !9 :

" = 6 ∙
2 2
(7)
∑@A *2 @A * 2 @A * 2 @A * 2
2B4[= ]P>? Q 1E2 ∙∑2B4[= ]2 G∑2B4[= ]1E2 ∙∑2B4[= ]P>? Q 2
° °

R%(S)

, with TUV(W) defined as:

56 56 56 J

TUV (W ) = 4 ;[*+ ]IK+ ∙ ;[* + ]ℎ + − 3 [;[* + ]IJ+ \

+78 +78 +78

The incremental curvature "  can be due to thermal or cure strains. Therefore, the total
curvature can be divided into two parts during cure, one due to the chemical shrinkage strains
and one due to the temperature change.

",^ =
2 2
∑@A * 2 `a
2B4[= ]_ d 1E2 ∙∑@A *2 @A * 2 @A * 2 `a
2B4[= ]2 G∑2B4[= ]1E2 ∙∑2B4[= ]_ d 2
° °
S ,&
6 (S f,& f W, e
`b,`c `b,`c

g g TUV(W)
(8)

+
>?°
Where, h>S,>& i means that differentiation is made with respect to degree of cure q (instead

of time) or temperature T. The integral limits qt or Tt are the desired degree of cure or
temperature, respectively, at the corresponding cure time and q0 or T0 are the starting values.

Note that the above equation can be split in separate thermal and cure induced parts if first
the conversion during heating remains below gelation level, qgel and second during cooling,
there is no further reaction; two conditions that are mostly met.

6
 Therefore, Eq. (8) is valid for all parts of the cure cycle including heating, isothermal cure
and cool-down. For instance, in the isothermal cure part, Tt is equal to T0 (Tcure) and therefore,
there will be no thermal strain and the only source of curvature will be the cure strain. If all
cure occurs at the isothermal cure temperature, then Eq. (8) can be further simplified by
separating it in a term depending only on cure shrinkage (by integration only over q) and a
term depending only on the thermal contraction effects during cooling (by integration only
over T). If curing is not isothermal, calculations are done for both the change in degree of cure
and temperature and the resulting curvature, strains and stresses are summed up. Since, both

[*+ ] and . ° / are direct functions of q and only indirect functions of time, the integration
+

is performed over degree of cure (q) instead of cure time (t). If the cure shrinkage strain is
assumed to be linear, then the above equation can be simplified. More descriptions are given
in the next section about the input parameters of analysis.

For the cool-down part, assuming that there is no curing reaction, the curvature is
>?°
determined similar to Eq. (8) just by replacing the h i = α(T). Then, we will obtain:
>&

κkl,mnoo =

kwxy ∑tu qr r tu q r tu q r tu q r r
rB4[p ]α(k) sEr ∙∑rB4[p ]lr G∑rB4[p ]sEr ∙∑rB4[p ]α(k) lr
6 (k dT
R%(S)
(9)
z

In the above equation, {(e)+ is the vector of the instantaneous coefficients of thermal
expansion of layer k in different directions and the integration is made from cure temperature
until ambient temperature.

2.2 Case 2: constrained curing

In the previous section, Equations (8) and (9) are written in case that there is no constraint
during curing of the composite. If the vacuum bag is attached to a solid mould, then both
midplane strain and curvature are constrained during cure and warpage is prevented. We,
therefore, consider the assumption " = ̅ = 0 in Eq.’s (3) and (6) and following Eq.
(2), recalculate the incremental force and moment due to the incremental cure shrinkage as

+
[* + ]. ° / !
12
# . = − ∑56
+78 (1 234 (10)

7
 +
[* + ]. °/ !!
|2
L. = − ∑56
+78 (| 234

We should now calculate the force and moment at the end of cure, before cooling, by
integration of Eq. (10):

+
〈[*+ ] h
S}~` ,&}~` >?
#%>
.
= − ∑56
+78 (Sg ,&g i ∙ ℎ+ 〉 W, e
°

>S,>& (11)

+
L%>
.
= − ∑56 (
S}~` ,&}~`
+78 Sg,&g 〈 *
[ +
] h
>?°
i ∙ ZJ€〉 W, e
>S,>& (12)

Since the shrinkage strain is negative, the positive reactions #%>

.
and L%>
.
are applied to
the mould. After opening the mould, the panel is released and, in case of non-symmetric
layup, will deform driven by the force and moments derived above.

Hereafter, we replace (̅ and dκ) by (∆̅ and ∆κ) denoting the changes in inplane shrinkage
strain and curvature, respectively due to curing after the mould release. If we use Eq. (12) in
Eq. (3), we will get the equilibrium of forces at the moment of mould release as:

[* + ][∆̅ − 0]! + ∑56 [* + ][! −

12 12
∑56
+78 (1 +78 (1
234 234 (13)

!9 ∆" ]! + #%>

.
=0

Again, to be valid independently for any arbitrary ", we will obtain similar to the free
curing case for the neutral plane:

∑@A [=*2 ]sEr

!9 = ∑2B4
@A *2
2B4[= ]∙2
(14)

And also, we should have: ∑56 *+

+78 [ ][∆̅ − 0] ∙ ℎ + + #%>
.
= 0 , from which the change in
the inplane shrinkage strain can be calculated as:

‚}ƒ„.
5}~`
∆̅ = − *2
…∑@A
2B4[= ]∙2 … f}~`
(15)

In a similar manner, Eq. (12) can be used in Eq. (6) to get the equilibrium of moments when
the mould is released:

8
 [* + ][∆̅ + (! − !9 )∆" − 0] !! +
|2
∑56
+78 (| 234 (16)

L%>
.
=0

If Eq.’s (14) and (15) are used in Eq. (16), the change in curvature ∆κ can be determined for
the constrained curing problem after opening the mould as follows:

∆" †,%‡
+
Š ∑56 [*+ ] _  ° d I ∙ ∑56 [*+ ]ℎ 
S}~` ,&}~`
‰ +78 W, e J+ +78 + Œ
(S ,& ‰ Œ W, e
g g ‰ 
+
Œ
‰− ∑56 [* +
]I ∙ ∑ 56 * +
[ ] _ d ℎ Œ
°

ˆ
+78 J+ +78 W, e +
‹
=6
|TUV(W)| }~` (17)
Where, W%> = 1 and e%> = e are the degree of cure and temperature at the end of curing
before cooling. If the cure is in the isothermal part, integration can be taken over degree of
cure instead of time. Compared to the free curing case, Eq. (8), the only difference in the
expressions for the curvature is that the integration is here done only on the numerator; the
denominator of Eq.

(17) should be evaluated at the end of cure and not integrated over the entire curing time.

Note that as before, Eq. (16) simplifies considerably if all curing is in the isothermal stage
such that the expression can be split in a thermal cooling and a cure shrinkage part.

2.3 Resin shrinkage during cure

Thermoset resins are known to shrink during polymerisation. Each time a monomer reacts
and bonds to the network, it is restricted in its movements and loses a certain amount of free
volume. For all monomer molecules, this amount of loss of free volume is likely to be
identical. Therefore, the loss of free volume (volumetric shrinkage) must be proportional to
the degree of cure. The linear relation ( (W ) = 
&
∙ W) is used by many researchers [13-
15], where 
&
is the total volumetric shrinkage of the epoxy which is here assumed to be
about 3%.
8
The linear shrinkage strain of the resin can be found as:  (W) = ‘1 +  (W) − 1 ≅ K  (W)
for small amount of 
&
. The equations can be further simplified by taking the linear shrinkage
strain as:

9
  (W) = , ∙ W
(18)

, where εc,max denotes the linear cure shrinkage after full conversion. Then we get κ“”“ =
κ• + κkl for both free and constrained curing. κ• is the term included in the current
work with an analytical expression for composites that was not available before in the
literature.

3 Material Characterization

The input parameters of the model formulated in the previous section should be determined
by experimental measurements and/or typical assumptions for the resin type used in the
laminate. This includes the relation between degree of cure and temperature, variation of
stiffness of the laminate during cure and chemical shrinkage of the epoxy adhesive during
cure. In the following subsections, assumptions and measurements are described in detail.

3.1 Cure characterization of GLARE

The parameters of the kinetics model for curing of epoxy FM-94 has already been reported
by the authors [11]. Cure rates are calculated from Differential Scanning Calorimetry (DSC)
measurements and fitted to the Kamal-Sourour reaction rate equation [16] with parameters
given in Table 1:

W G™šœ
= M— U ˜ ›& W  (1 − W)%
– (19)

In the above equation, R is the universal gas constant and T is the temperature in Kelvin. EA
is the activation energy, k0 is a coefficient and m and n are two constants.
Table 1: Cure kinetic parameters for epoxy FM-94 [11]

k0 [1/s] 3.52 E+06

EA [J/mole] 6.75 E+04
m [-] 0.558
n [-] 2.508
Htot [J/g] 134.9

10
 The degree of cure at any time t may be calculated by integration of the cure rate. The
detailed description of the results can be found in ref. [11]. Corresponding values for
temperature and degree of cure during the typical cure cycle of GLARE is shown in Figure 2.

125 1
120

110
0.8
100

90

0.6
Temperature [° C]

Degree of cure [-]

80

70

0.4
60

50

Temperature 0.2
40
Degree of cure

30

20 0
0 20 40 50 60 80 100 120 150 200 250
Cure time [min]

Figure 2: Temperature and conversion during the cure cycle of GLARE [11]

In order to determine the gelation point of the epoxy, measurement of the stiffness is needed
during cure. Other methods are also used for this purpose like DSC [17] but the most
convenient method is using Dynamic Mechanical Analysis (DMA) to measure the response of
the material to the applied force versus temperature and time [18]. Usually, a Rheology test is
done on the resin material which is a special kind of DMA in liquid state. The setup consists
of two parallel plates that apply a shear torque to the adhesive. The temperature profile is
defined and the point (time) at which the torque abruptly changes is designated as the gelation
time. Figure 3 shows the Rheology measurement data for epoxy FM-94 indicating the change
of the viscosity of the material versus time. As it can be seen in the figure, the gelation time is
designated as tgel=117sec. in order to get the gelation degree of cure, the cure kinetics data is
needed for the same ramp heating rate of 1°C /min (Figure 4).

11
 Figure 3: Rheology data for epoxy FM-94 under ramp heating (1°C /min) [19]

200 1

150

Temperature
Temperature [° C]

Degree of cure [-]

Degree of cure

100 0.5

50

0 0
0 20 40 60 80 100 120 140
Cure time [min]

Figure 4: Temperature and conversion during ramp heating (1°C /min) of epoxy FM-94

Correlating the temperature at which the viscosity abruptly increases (Figure 3) to the
corresponding degree of cure in Figure 4 yields the gelation degree of cure equal to 0.8 which
means that the adhesive does not carry load until the cure reaches 80% of its ultimate value. It
is obvious that the earlier the gelation point is, more residual stress is produced during the
isothermal cure. It should be noted that the value of 0.8 is in fact too high; the gelation degree
of cure should fall between 0.4-0.6 for epoxies [20].

12
 According to Figure 2, the prepreg starts to cure during the heating part. The temperature
change during this level of cure influences the temperature-dependent properties of the
polymer like the thermal expansion coefficient and especially the viscoelastic response (stress
relaxation). However, the degree of cure at the end of heating is about 0.24 which is much
below the gelation point (0.8). This means that gelation and the subsequent stress build-up
occurs within the isothermal cure part of the cure cycle. Therefore, thermal strains arising in
the heating part do not contribute to the residual stress development. Therefore, it can be
concluded that deformations and residual stresses produced before cool-down are due to the
cure shrinkage, increase of stiffness and stress relaxation (if exists) of the resin.

3.2 Resin stiffness during cure

Thermo-mechanical characterization of the resin material is needed since its mechanical

properties change with degree of cure and temperature. The properties which vary due to
polymerization include the elastic and viscous moduli. A linear relation between the stiffness
and degree of cure is assumed by some researchers [13] neglecting the glass transition
temperature, viscoelastic relaxation and temperature dependency. All the latter effects are
taken into account in some other works [21-23].

The need for full characterization of viscoelasticity depends on the thermoset system and
cure cycle parameters. If the thermoset used is cured above its glass-transition temperature,
dominant effects are from cure shrinkage and modulus increase during cure. Thermoset
systems are usually cured above the glass transition temperature since systems that cure at a
temperature below their glass transition will vitrify during curing, such that full conversion is
never reached [9].

Above glass-transition temperature, the material will be rubbery and its stiffness depends
only on cure polymerisation with negligible viscoelastic effects (stress relaxation) [24].
Therefore, if the resin is cured isothermally above Tg, the elastic modulus during cure can be
approximated by its equilibrium modulus which simplifies the cure modelling. The
equilibrium modulus of the resin can then be modelled using percolation theory [24,25]. A
simplified version of the theory is applied to the equilibrium shear modulus of the resin during
cure in the Martin and Adolf model [26,27]:

¢œ
W J − W 
J K
ž› (W) = ž› Ÿ J ¡ £¤¥ W > W 
^
1 − W 
(20)

13
 Where GRf=0.81GPa is the rubbery shear modulus in the final cured state, q is the degree of
cure and qgel is the gelation degree of cure. The properties of GLARE constituents are listed in
Table 2. For the S2-glass fibre, the mechanical data are from Hexcel product datasheets [28]
and the coefficient of thermal expansion property is from a military handbook [29]. The
properties for the epoxy FM-94 (Table 2) are from thermo-mechanical measurements [11].
Equivalent properties of the laminate can be calculated using the self-consistent
micromechanics equations listed in Appendix A.
Table 2: Material properties for GLARE constituents

S2-glass fibre [28] & [29]

E1= E2=E3 [GPa] G12=G13 [GPa] ν12=ν12 α1= α2 [1/˚C]

85 33.1 0.33 2.6×10-6

Epoxy FM-94 [11]

E1= E2=E3 [GPa] G12=G13 [GPa] ν12=ν12 α1= α2 [1/˚C]

21.7 0.81 0.33 90

Aluminium 2024-T3 [30]

E [GPa] G [GPa] ν α[1/˚C]

72.4 27.6 0.33 22×10-6

Table 3: Material properties for S2-Glass/FM94 prepreg layers

E1 [GPa] E2=E3 [GPa] G12=G13 [GPa] ν12=ν12 α1[1/˚C] α2[1/˚C]

From literature [30,31]

54 9.4 5.5 0.33 6.15×10-6 26.2×10-6

From measurements and using micromechanics equations

52.0 7.4 3.0 0.33 4.1×10-6 48.6

Using the value qgel=0.8 for epoxy FM-94 in Eq. (20) and Poisson ratio in the rubbery state
equal to 0.45, the development of stiffness of the prepreg layer of GLARE can be observed as
in Figure 5 after gelation of the epoxy.

14
 12

10

prepreg stiffness-Transverse E22 [GPa]

8

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Degree of cure [-]

Figure 5: Development of stiffness for a S2-glass/FM-94 epoxy prepreg

4 Numerical results and discussion

Results are obtained using the input data for the GLARE material and the analytical
formulation presented in this paper.

4.1 Validation of the model: comparison with numerical simulation

Numerical packages like ANSYS and ABAQUS are useful in multidisciplinary simulation
of composite laminates. However, they lack the capability of simulation of the curing process
with chemical shrinkage. For comparison purpose, only thermal cooling of the laminate is
solved by ANSYS. The results are compared with the ones from the current model, assuming
no chemical shrinkage occurs during cure. Consequently, the stiffness components of the
composite material change during cure but produce no curing stress or strain, therefore, the
only stress developed by the analysis would be the thermal stresses which are comparable to
the ANSYS results. Strips of 47×2.7 cm with layup (Al-0/90/0-Al-0/90/0-Al) are used. In
ANSYS, the temperature gradient (cooling from cure temperature to the ambient one) is
applied with the material properties corresponding to the complete cure of the prepreg. It can
be observed from Table 4 that there is good agreement between the prediction of thermal
stresses. The difference between ANSYS and our analytical solution is typically not more
than 10%. Therefore, we can conclude the model is working correctly.

15
 Table 4: Comparison of thermal stresses in layers of GLARE panels (Layup: Al-0/90/0-Al-0/90/0-Al)

Layers Al 0° 90°
layers layers layers
Stresses [MPa]

s0° -Model 27.6 -76.3 22.3

s0°-ANSYS 26.3 -71.2 18.4

s90° -Model 11.0 20.1 -92.3

s90°-ANSYS 10.9 16.9 -85.4

4.2 Curing stresses due to chemical shrinkage and stiffness evolution

Another discussion can be made by comparing the results with the results available for non-
symmetric GLARE panels in a previous paper [7]. In the aforementioned paper, the bending-
strain free temperature of some panels were measured and used to determine the thermal
stresses instead of the cure temperature. In that paper, it was discussed that the difference
between Tcure and TBSF comes from the so called “curing effects” including chemical
shrinkage during cure and stress relaxation during cool-down. Hereafter, we call the stresses
and deformations during cure of the material as the curing stresses and the ones after
completion of cure and cool-down as cooling stresses and deformations.

If we calculate the curing deflection of a GLARE laminate with constrained conditions

during cure, using Eq. (17), they are much less than the cooling ones. The curing deflection of
a constrained cure process is the deformation of the laminate after it is removed from the
mould and just before cooling. It seems that the bending moment produced due to non-
symmetric curing stresses are not enough to curve the aluminium layers and consequently, the
curing deflection of the panel is much less than the cooling ones.

The stresses developed after curing, just before cooling, are however considerable specially
in the fibre layers. Results are gathered in Table 5 for a non-symmetric panel.

16
 Table 5: Stresses in layers of GLARE panel (Layup: Al-0/0/0-Al-90/90/90-Al)-all values are from the current model

Layers Al 0° 0° 0° Al 90° 90° 90° Al

Stresses [MPa]

scure0° -0.6 13.7 13.6 13.5 -1.2 58.5 58.4 58.4 -1.8

scool0° 33.6 -73.1 -76.5 -79.9 12.0 19.9 19.4 18.9 -9.6

scure90° -1.5 58.4 58.5 58.5 -0.9 13.6 13.7 13.8 -0.3

scool90° 1.5 19.4 19.9 20.4 23.0 -76.5 -73.1 -69.7 44.6

Panel deflection due to cure 0.4 mm

Panel deflection due to cooling 13.9 mm+

+ The small difference between this value and the value 13.4mm in ref. [7] is due to the difference
between the material data used there and the ones calculated here from micromechanics equations.

As it can be seen, the deflection due to cooling from cure temperature is 13.9mm. The
bending-strain-free temperature was measured for the same panel to be TBSF=101°C, whereas
the cure temperature was Tcure=120°C. Using this temperature (TBSF) instead of cure
temperature (Tcure) to calculate the thermal stresses, yields a final deflection of 10.7mm which
is much closer to the experimentally measured value of 10.8mm [7]. As mentioned, it can be
concluded that despite of the high values of stresses due to isothermal curing, the produced
deflection is much lower than during cool-down part. Consequently, we can say that the
although “curing effects” before cooling and just after the isothermal cure part produce
considerable residual stresses but contribute less to the deflection of a non-symmetric GLARE
panel. Thus, use of TBSF in determination of deflection of GLARE panels mainly
approximates the stress relaxation of the polymeric adhesive during cool-down. Using TBSF is
just an approximation of the curing effects since the panel is not essentially free of stress at
TBSF but only flat and free of bending strain (curvature).

It should be noted that, as discussed earlier, the viscoelastic relaxation of the epoxy is
negligible during cure but it can have a considerable effect during cooling after completion of
cure. This is due to the fact that the epoxy cools from a temperature above its glass transition
to room temperature, where temperature dependency of stiffness and stress relaxation are
important factors in development of thermal residual stresses. Consequently, the deflection
due to cooling which is mentioned in Table 5, overestimates the real ones by neglecting the
viscoelastic response of the polymer.
17
 The value of the gelation degree of cure (q gel) has a considerable effect on the residual
stresses developed during curing. If we take a more reasonable value of 0.4 for the gelation
degree of cure, the deflection after cure of the same GLARE panel as in Table 5 becomes
0.9mm which is more considerable comparing to the total deflection. We should again have in
mind that the cooling deflection would be much lower considering stress relaxation during
cooling.

To complete our discussion, we have made another specimen to exclude the effect of
aluminium layers. The laminate is actually a full composite strip with dimensions of
300×30mm. The laminate was put between two aluminium plates before putting in the
vacuum bag. Again, the deflection and layers’ stresses are calculated with the analytical
model and the model results are presented in Table 6. The value of the gelation degree of
cure, qgel =0.8 was considered. As it can be seen, considerable amounts of residual stresses are
produced during the curing process of the epoxy in the prepreg layers. The important point is
that the contribution of the curing part to the total amount of deflection, has increased. The
real deflection of the laminate was measured by a caliper and it was about 14.8mm. Note the
overestimation of “deflection due to cooling” because of the stress relaxation during cooling.
As it can be seen, the contribution of the curing stresses to the total deflection is near 10%.
This completes our previous discussion that despite development of large amounts of layer
residual stresses due to curing, the major part of the produced curvature in the non-symmetric
laminates is due to cooling, especially when metal layers are present in the material.

Table 6: Stresses in layers of a full composite panel (Layup: [0]3-[90]3)-all values are from the current model

Layers 0° 90° 0° 90° 0° 90°

Stresses [MPa]

scure0° 11.4 57.5 9.8 5.7 8.2 57.0

scool0° -14.4 29.6 -32.3 27.2 -50.1 24.8

scure90° 57.1 9.0 57.4 10.6 57.7 12.2

scool90° 26.0 -41.2 28.4 -23.3 30.8 -5.6

Panel deflection due to cure 1.4 mm

Panel deflection due to cooling 15.0 mm

18
5 Conclusions

An analytical solution was presented for estimation of curing residual stresses in laminated
composites. The only assumption was that there is no viscoelastic relaxation in the resin
during curing which is valid for composites cured above their final glass transition
temperature. The stiffness increase during curing of the epoxy is considered together with the
cure (chemical) shrinkage of the prepreg layers in different directions. It was seen that the
residual stresses developed in the fibre layers are considerable but the ones in the aluminium
layer are much less. The curvature and the deflection of the panel after cure, before cool-
down, is smaller compared to the ones developed during cool-down. For accurate prediction
of residual stresses in the layers, the current analysis with the developed model is necessary to
account for the curing-induced stresses . This is observed in both FMLs and full composites.
However, based on the comparisons with the previous experiments, it seems that the cooling
stresses produce the major part of the curvature when a source of non-symmetry exists in the
laminate. Therefore, the thermo-viscoelastic cool-down analysis for the accurate prediction of
cooling warpage is required which will be made in a future phase of this research.

19
Appendix A: Self-consistent micromechanics equations [13,14]

In this appendix, we list the equations used in the determination of the composite prerpeg
properties. In the following equations, subscripts 1,2 and 3 stand for the principal directions of
unidirectional laminae and f and m stand for fibre and matrix, vm is the fibre volume fraction
and k is the isotropic bulk modulus.
E8m
Km =
2 ∗ (1 − u8Jm − 2 ∗ u8Jm ∗ u8Jm )

E¬
K¬ =
2 ∗ (1 − u8J¬ − 2 ∗ u8J¬ ∗ u8J¬ )

4(u8J¬ − u8Jm ∗ u8Jm ) ∗ K m ∗ K ¬ ∗ GJK¬ ∗ (1 − vm )vm

E88 = E8m ∗ vm + (1 − vm ) ∗ E¬ +
(K m + GJK¬ ) ∗ K ¬ + (K m − K ¬) ∗ GJK¬ ∗ vm

(u8J¬ − u8Jm ) ∗ (K ¬ − K m ) ∗ GJK¬ ∗ (1 − vm ) ∗ vm

u8J = u8Jm ∗ vm + u8J¬ ∗ (1 − vm ) + Ÿ ¡
(K m + GJK¬ ) ∗ K ¬ + (K m − K ¬ ) ∗ GJK¬ ∗ vm

u8K = u8J

(G8Jm + G8J¬ ) + (G8Jm − G8J¬ ) ∗ vm

G8J = G8J¬ ∗
(G8Jm + G8J¬ ) − (G8Jm − G8J¬ ) ∗ vm

G8K = G8J

K ¬ ∗ (GJK¬ + GJKm ) + 2 ∗ GJK¬ ∗ GJKm + K ¬ ∗ (GJKm − GJK¬ ) ∗ vm

GJK = GJK¬ ∗
K ¬ ∗ GJK¬ + GJKm ) + 2 ∗ GJK¬ ∗ GJKm − (K ¬ + 2 ∗ GJK¬ ) ∗ (GJKm − GJK¬ ) ∗ vm
(

(K m + GJK¬ ) ∗ K ¬ + (K m − K ¬ ) ∗ GJK¬ ∗ vm
Kk =
(K m + GJK¬ ) − (K m − K ¬ ) ∗ vm

1
EJJ =
1 1
+ + u8J ∗ 8J
u
4 ∗ K k 4 ∗ GJK E88

EKK = EJJ

2 ∗ E88 ∗ K k − E88 ∗ EJJ − 4 ∗ u8J ∗ u8J ∗ K k ∗ EJJ

uJK =
2 ∗ E88 ∗ K k

α8m ∗ E8m ∗ vm + α8¬ ∗ E¬ ∗ (1 − vm )

α8 =
E8m ∗ vm + E¬ ∗ (1 − vm )

αJ = ˜αJ“lm + Nu8Jm ∗ α8m  ∗ vm + °αJ¬ + u8J¬ ∗ α8¬ ± ∗ (1 − vm ) − °u8Jm ∗ vm + u8J¬ ∗ (1 − vm )±

∗ α8
αK = αJ

20
 References
1. Asundi A, Choi AYN. Fiber metal laminates: An advanced material for future aircraft. Journal of
Materials Processing Technology. 1997; 63:384-94.
2. Alderliesten RC. Fatigue crack propagation and delamination growth in Glare. The netherlands:
Delft university of technology, 2005.
3. Homan JJ. Fatigue initiation in fibre metal laminates. International Journal of Fatigue. 2006;
28:366-74.
4. Sinke J. Development of fibre metal laminates: Concurrent multi-scale modeling and testing.
Journal of Materials Science. 2006; 41:6777-88.
5. Alderliesten RC, Benedictus R. Fiber/metal composite technology for future primary aircraft
structures. Journal of Aircraft. 2008; 45:1182-9.
6. Abouhamzeh M, Sinke J, Benedictus R. On the prediction of cure-induced shape deviations in fiber
metal laminates. Journal of Composite Materials. 2015; 49:1705–16.
7. Abouhamzeh M, Sinke J, Benedictus R. Investigation of curing effects on distortion of fibre metal
laminates. Composite Structures. 2015; 122:546–52.
8. Abouhamzeh M, Sinke J, Benedictus R. On the investigation of residual stress and shape deviation
development in manufacturing of GLARE. ECCM16-16th European Conference on Composite
Materials. Seville, Spain22-26 June 2014.
9. Jansen KMB, De Vreugd J, Ernst LJ. Analytical estimate for curing-induced stress and warpage in
coating layers. Journal of Applied Polymer Science. 2012; 126:1623-30.
10. Jansen KMB, Oeztuerk B. Warpage estimation of a multilayer package including cure shrinkage
effects. Ieee Transactions on Components Packaging and Manufacturing Technology. 2013; 3:459-66.
11. Abouhamzeh M, Sinke J, Jansen KMB, Benedictus R. Kinetic and thermo-viscoelastic
characterization of the epoxy adhesive in GLARE. Composite Structures. 2015; 124C:19-28.
12. Reddy JN. Mechanics of laminated composite plates and shells. New York: CRC Press, 2004.
13. Bogetti TA, Gillespie JW. Process-induced stress and deformation in thick-section thermoset
composite laminates. Journal of Composite Materials. 1992; 26:626-60.
14. White SR, Kim Y. Process-induced residual stress analysis of as4/3501-6 composite material.
Mechanics of Composite Materials and Structures. 1998; 5:153-86.
15. Ramos JA, Pagani N, Riccardi CC, Borrajo J, Goyanes SN, Mondragon I. Cure kinetics and
shrinkage model for epoxy-amine systems. Polymer. 2005; 46:3323-8.
16. Sourour S, Kamal MR. Differential scanning calorimetry of epoxy cure: Isothermal cure kinetics.
Thermochimica Acta. 1976; 14:41-59.
17. Gao J, Li L, Deng Y, Gao Z, Xu C, Zhang M. Study of gelation using differential scanning
calorimetry (dsc). Journal of Thermal Analysis. 1997; 49:303-10.
18. Dynamic mechanical analysis of polymers. http://www.tainstruments.com/: TA Instruments.
19. Fm 94 adhesive film-technical datasheet. http://www.cytec.com/products/fm-94-0: Cytec
Corporation.
20. Jansen KMB, Qian C, Ernst LJ, Bohm C, Kessler A, Preu H, et al. Characterization and modeling
of molding compound properties during cure.
21. Abou-Msallem Y, Kassem H, Jacquemin F, Poitou A. Experimental study of the induced residual
stresses during the manufacturing process of an aeronautic composite material. Research Journal of
Applied Sciences, Engineering and Technology. 2010; 2:596-602.
22. De Vreugd J, Jansen KMB, Ernst LJ, Pijnenburg JACM. Modelling of viscoelastic properties of a
curing adhesive. p. 241-51.

21
23. Ernst LJ, Jansen KMB, Saraswat M, Zhang GQ, Yang DG, Van 'T Hof C, et al. Fully cure-
dependent modelling and characterization of emc's with application to package warpage simulation. p.
23-7.
24. De Vreugd J, Jansen KMB, Ernst LJ, Bohm C. Prediction of cure induced warpage of micro-
electronic products. Microelectronics Reliability. 2010; 50:910-6.
25. Stauffer D. Introduction to percolation theory. London: Taylor and Francis, 1985.
26. Adolf D, Martin JE. Calculation of stresses in crosslinking polymers. Journal of Composite
Materials. 1996; 30:13-34.
27. Adolf DB, Chambers RS. A thermodynamically consistent, nonlinear viscoelastic approach for
modeling thermosets during cure. Journal of Rheology. 2007; 51:23-50.
28. Hexcel. Reinforcements for composites. Texas.
29. Military handbook (mil-hdbk-17-3f). In: Defense DO, editor.
30. Alderliesten RC, Homan JJ. Fatigue and damage tolerance issues of Glare in aircraft structures.
International Journal of Fatigue. 2006; 28:1116-23.
31. Krimbalis PP, Poon C, Fawaz Z, Behdinan K. On the prediction of induced residual stresses in
fibre metal laminates. In: 7th Joint Canada-Japan Workshop on Composites. Fujisawa, JAPAN:
Shonan Inst Technology. p. 125-32.

22