DTj2

,K COPY
TECHNICAL
REPORT BRL-TR-3121
BIRL
AD-A224
885
CURE SIMULATION OF
THICK THERMOSETTING COMPOSITES
DTIC .,
1; 'ECTE TRAVIS A. BOGETTI
AU G 0 3 1990
JOHN W. GILLESPIE
UNIVERSITY OF DELAWARE
JULY 1990
APPROVED FOR PUBLIC RELEASE; DiSTRIBUTION UNLIMrTED.
U.S. ARMY LABORATORY COMMAND
BALLISTIC RESEARCH LABORATORY
ABERDEEN PROVING GROUND, MARYLAND
qo 00 o3 086
NOTICES
Destroy this report when it is no longer needed. DO NOT return it to the originator.
Additional copies of this report may be obtained from the National Technical Information Service,
U.S. Department of Commerce, 5285 Port Royal Road, Springfield, VA 22161.
The findings of this report are not to be construed as an ufficial Department of the Army position,
unless so designated by other authorized documents.
The use of trade names or manufacturers' names in this report does not constitute indorsement of
any commercial product.
- ~ ~ ~ ~ ~ ~ Fr AppNLA
FE
--- 7
, ,roved
REPORT DOCUMENTATION PAGE 0MB No. 0704-.0188
Public reprorting bujrden for this 401lection of ir.formhuion Is estimated to vea qeIhu e repnit indludig "h time forrviewi n intructions, $earthing costinirg datai ources,
gatherinrg nd maitainintg the data neeided, and 4ompiotng Sire Iew n , th 0ilc-lilon of informrration. Send co mnsriarolng thiS "burders estimate or any other aspect of this
collection of information, rnc ding suqqetiont for rucng hris ouracer. to .VV~hington Headquarters Services. Directorate for Informration Operations and Reports. I21I I S f*Hrron
Davis Highway. Suite ,1204, Arilfigon, I. V 2202.4302, .rnd to ihitOffice of Management 4nd Budget, Peperwork Reduction Project (0704.0 1S8).Washington, DC 20503.
I. GECY SEONL (eae ban) 2 RPOT nfltS r3.RE-PORTT TYPE AND DATES COVERED
1. GECYUSEONY Leve lak)_July
1990
Final
Jan 89-Apr 90
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS
Cure Simulation of Thick Thermosetting Composites
6. AUTHOR(S)
Travis A. Bogetti
PR: 1L1662618AH80
John W. Gillespie.-Jr.*
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
REPORT NUMBER
9. SPONSORING/ MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/ MONITORING
US Army Ballistic Research Laboratory
AGENCY REPORT NUMBER
ATTNr SLCBR-DD-T
Aberdeen P~roving Ground, MD 21005-5066 BRL-TR-3121
Ii. SUPPLEMENTARY NOTES
John Gillespie, Jr. is. a senior scientist at the Center for Composite
Materials and Assistant Professor of Mechanical Engineering, Univ. of Delaware.
12&. DISTRIBUTION /AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE
Approved for public release; distribution is unlimited.
13. ABSTRACT (Maximum 200 words)
~-"The potential for enhanced performance characteristics of composite materials over conven-
tional metals, such as high strength- to-weight ratio, high stiffness, corrosion resistance and low
radar signature, makes them extremely attractive for a wide range of military applications. The
curing process of thick-section thermosetting composites is critical to the quality and in-service
performance2 of the finished component, In this paper, an investigation into the two-dimensional
cure simulation of thick thermuosetting composites is; presented. Temperature and degree of cure
distributions within typical glass/polyester and graphite/epoxy structural elements of arbitrary
cross-section (ply-drop and angle bend) are analyzed to provide insight into the non-uniform cur-
ing process unique to thick-sections, An incremental, transient finite difference solution scheme
isimplemented to solve the pertinent governing equations and boundary conditions, Correlation
between experimentally measured and predicted through- the- thickness temperature profiles in
glass/polyester laminates are presented for various arbitrary temperature cure cycle histories.
Spatial gradients in degree of cure are shown to be strongly dependent oil part geometry, ther-
mal anisotropy, cure kinetics and the temperature cure cycle. These spatial gradients directly
influence the quality and in-service performance of the finishe~d component by inducing warpage
and residual stress during the curing process. / .. * .
14. SUBJECT TERMS
15S. NUM8ER OF PAGES
-'Thermnoset Composites; Thick Laminates; Cure Simulation; Anisotopic Heat 62
Conduction; Boundary Fitled Coordinates; Finite Difference Theory , 7e 16. PRICE CODE
-17. S .ECURITY CLASIFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. LIMITATION OF ABSTRACT
OF REPORT Of THIS PAGE OF ABSTRACT
UNCLASSIFIED UNCLASSIFIED UNCLASSIFIED SAR
NSN 7540-01-280-51S00
Standard Forrm 298 (Rev 2-89)
Prescribed by A5NSI Sid Z39.18
UNCLASSIFIED
2512
INTENTONALLY LEFT BLANK.
Contents
1 Introduction 1
2 Background 2
3 Analysis 4
3.1 Assum ptions . . .. . . . . .. . . . . . . . . . . . .. . . . . .. . .. . . . . . . . . . 4
3.2 Heat Conduction Equation ....................... . ........ 4
3.3 Temperature Boundary Conditions ................................... 5
3.4 Chem ical Kinetics .... .... .. . ... .. . .. . . ... . ... . .. .. . 7
3.5 Initial Conditions ...................... . . . . ... ....... 8
3.6 Boundary Fitted Coordinate System Transformation Technique ............ 9
3.7 Alternating Direction Explicit Finite Difference Method ..................... 11
3.8 Degree of Cure Calculation ........ ................................ 15
4 Results and Discussion 16
4.1 Input Summary ......... ....................................... 16
4.1.1 Solution Details ........................................... 17
4.1.2 Mesh Input ....................... ........... 17
4.1.3 Thermal Properties ........ ................................ 17
4.1.4 Cure Kinetic Data ........ ................................. 19
4.2 Model Verification ......... ..................................... 21
4.3 Experimental Correlation ........ ................................. 21
4.4 Boundary Condition Effects ........................................ 24
iii
4.5 Thickness Effects ...... .. ......... . .. ...... . ....... .. .. 28
4.6 Cure Cycle Temperature Ramp Effects ................................. 34
4.7 Anisotropic Curing in Complex Shaped Composites ................... 34
5 Conclusions 43
A Appendix 49
A. . Finite Difference Approximations .................................... 49
A,l1. Central Differencing ........ ................................ 49
A-1.2 One-Sided Differencing .............................. . 49
A.1.3 Explicit Time Differencing Formula .............................. 50
A.2 Boundary-Fitted Coordinate System Transformation Coefficients ........... 50
A,2.1 Governing Equation Coefficients ...... ......................... 50
A.2.2 Boundary Conditions Coefficients ......................... 51
NTIS :&
[ 1) 1' t I /. w 3 - e
.J t, I C t C, II
By
DisIr ib .i t lo
hv,iihblbdif C(du.s
DIM",,c l
A-i
iv
List of Figures
1 Thermal Conductivity Transformation Between Ooordinate Systems ............. 6
2 Conceptual Representation of the Physical and Computational Domains ....... 10
3 Normal Derivatives on the Computational Domain Boundaries ................ 12
4 The Com putational M esh .... .. ... .. .... ... ........ ....... 14
5 Cure Simulation Mesh Geometries ....... ............................ 18
6 Verification of Temperature Solution in a Glass/Polyester Laminate ............ 22
7 Verification of the Degree of Cure Solution in a Glass/Polyester Laminate ...... 23
8 Temporature Comparision with Glass/Polyester Laminate 1 .............. 25
9 Temperature Comparison with Glass/Polyester Laminate 2 .............. 26
10 Temperature Comparison with Glass/Polyester Laminate 3 ................... 27
11 Typical Glass/Polyester and Graphite/Epoxy Temperature Cure Cycles ........ 29
12 Influence of (h/k)ff on Temperature Profiles in a Glass/Polyester Laminate . . .. 30
13 Influence of (h/k)
1
,f on Degree of Cure Profiles in a Glass/Polyester Laminate . .. 31
14 Influence of Thickness on Centerline Temperature Profiles in Glass/Polyester Lami-
nates ......... ............................................. 32
15 Influence of Thickness on Centerline Degree of Cure Profiles in Glass/Polyester Lam-
in a te s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 3
16 Temperatute Distributions in Glass/Polyester Laminates at Exotherrn ........ 35
17 Degree of Cure Distributions in Glass/Polyester Laminates at Exotherm ....... 36
18 Autoclave Temperature Cure Cycle Ramps ...... ....................... 37
19 Effect of the Temperature Ramp on Non-uniforrr Curing in a 2,54 cm Glass/Polyester
Laminate ........... .......................................... 38
V
20 Temperature Contours at Exotherm in a 2.54 cm Thick Graphite/Epoxy 900 Right
A ngle Bend . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. .. . . .. . . . 40
21 Condensed Glass/Polyester Cure Cycle ......................... . 41
22 Anisotropic Cu-'ng in a 2.54 cm Thick Glass/Polyester 901 Right Angle Bend .... 42
23 Anisotropic Curing in a Thick Glass/Polyester Ply-Drop .................... 44
vi
.... . . ...
List of Tables
1 Generalized Boundary Condition Coefficients . * 20
2 Thermal Properties for GlIass/ Polyester and Graphite/Epoxy Composites .. .. ..... 20
3 Cure Kinetic Parameters for Glass/Polyesiter and Graphite/Epoxy Composites . 20
vii
INTE~'1ONALLY LEFI' BLANKC.
viii
Acknowledgements
This work was funded by the Army Research Office University Research Initiative Program, The
authors are grateful for their financial support.
LX
INTENTIONALLY LEFT BLANK.
1 Introduction
A fundamental understanding of the relationships between processing and the overall quality and
in-service performance of thick-section thermosetting composite structures is needed. The devel-
opment of residual stresses, for example, is strongly influenced by processing history. Residual
stresses can have a significant effect on Lhe mechanics and performance of composite structures by
inducing warpage or initiating matrix cracks and delaminations [1,2,3,4,5].
Processing concerns associated with thermosetting composites become increasingly important
for components of appreciable thickness [6,7,8]. Perhaps the most familiar concern is an increase
in internal temperature resulting from the irrevei .ible exothermic chemical reaction of the ma-
trix phase, Liberated heat is slow to dissipate by conduction and may potentially raise internal
part temperatures to levels risking material degradation. A second concern relates to the complex
temperature and degree of cure gradients that develop in thick-sections during the curing process
[9,10,11]. These gradients induce non-uniform curing within the part that may ultimately lead to a
reduction in the overall quality and in-service performance of the finished component. Non-uniform
curing can result in incomplete consolidation of the part which, in turn, may lead to uti',esirable
volume fraction gradients and entrapped volatiles or voids [12,13). Complex cure gradients increase
the potential for process-induced warpage and matrix-microcracking and diminished residual prop-
erties of the structure.
The objective of this study is to gain a fundamental understanding of the curing process unique
to thick thermosetting composites parts of arbitrary cross-section, A two-dimensional anisotropic
cure simulation analysis is presented which accounts for thermal and chemical interactions asso-
ciated with the cure, Several typical glass/polyester and graphite/epoxy structural elements of
arbitrary cross-section are analyzed to provide insight into the non-uniform curing process of thick-
sections. Spatial gradients in tcmperature and degree of cure, unique to thick-section composites,
are shown to be strongly dependent on part geometry, thermal anisotropy, the chemical cure kinetics
and the thermal boundary conditions (cure cycle and tooling). Correlation with experimental mea-
surements of through-the-thickness temperature profiles in glass/polyester laminates are presented
for several arbitrary temperature histories.
2 Background
The autoclave curing process of thermosetting composites has been the subject of numerous in-
vestigations [6,12,14,15,16,17]. From early efforts to fabricate thick-section composites, various
undesirable effects were encountered that lead to poor part quality. Consequently, most studies
have sought to understand the curing process on a fundamental level. Studies of the curing process
have focused on the thermal and chemical interactions, degree of cure profiles, viscosity behav-
ior, void formation and growth, and resin flow phenomena occurring in the composite under the
application of a specified temperature and pressure cure cycle history [12,13,18]. Other investiga-
tions are more empirically oriented, citing general experiences encountered in the fabrication and
manufacture of thick-section thermoset composite parts [7,8]. A review of some of the literature
associated with the processing of thermosetting composites is now presented. The survey presented
here is not intended to be exhaustive, but, those authors cited are felt to have made significant
contributions in the area.
Early investigations by Levitsky and Shaffer [9] focused on temperature and degree of cure
gradients that develop in chemically reacting isotropic systems. Their one-dimensional an:ilytical
solution with prescribed temperature boundary conditions enabled them to investigate the inihence
of various reaction kinetic variables on the curing process. They extended their work to sho%, the
significant influence of temperature and degree of cure gradients on the development of stress, in
isotropic materials induced by non-uniform curing [19,20,21].
Loos and Springer [12] developed a comprehensive one-dimensional simulation model to describe
the curing process of flat plate unidirectional AS4/3501-6 graphite/epoxy composite laminates. The
model integrated submodels which describe the fundamental mechanisms associated with the cur-
ing process such as the thermo-chemical interactions, resin flow and void formation. Governing
equations describing Ihe curing process are solved with an implicit finite difference method. Tem-
perature, degree of cure, resin flow and void size, among other processing variables are predicted as
2
a function of the autoclave pressure and temperature cure cycle history. Experimental verification
of the model was performed and results are in good agreement with simulated predictions,
Kays [6] has conducted a comprehensive three year investigation on the processing issues unique
to large area thick-section laminates. The baseline material system was unidirectional AS4/3501
graphite/epoxy. Cure simulation models were developed and used in the investigation, Vari-
ous autoclave procedures, cure monitoring and non-destructive evaluation (NDE) techniques for
thick-section laminates were developed and evaluated. Contributions towards the development of
a generic methodology for processing thick-section laminates were made. Interesting observations
reported in the study were the development of microcracks and delaminations under certain process-
ing conditions, indicating the importance of processing on the cure and performance of thick-section
composites. Although ply-drop geometries were included in the study, cure simulation was limited
to a one-dimensional through-the-thickness analysis.
Efforts to optimize cure cycles for the large scale manufacture of thermosetting resin compos-
ites have been attempted, [14,16,17]. Computer-aided curing systems, utilizing cure simulation
submodels and control feedback systems, were developed. The studies focused on reducing coin-
posite manufacturing cost while improving part quality on a reproducible basis,
Bogetti and Gillespie [10] recently conducted a fundamental study of process-induced residual
stress in thick-section thermosetting composites. A one-dimensional cure simulation model is cou-
pled to an incremental stress analysis. A constitutive model is proposed to describe r material
behavior during cure that includes chemical hardetning, thermal and cure shrinkage effects. Residual
stresses are shown to be strongly influenced by gradients in temperature and degree of cure.
While cure simulations models for thermosetting composites are generally based on one-dimensional
through-the-thickness assumptions, the cure of arbitrarily shaped thick-sections is significantly in-
fluenced by part geometry and anisotropic heat transbier necessitating a two-dimensional analysis.
The present investigation contributes to the processing rcience of thick-section structural elements
encountered in the practical application of thermosetting composites, The two-dimensional anal-
ysis developed is a prerequisite for studying the evolution of process-induced residual stress and
deformation in arbitrarily shaped thick-sections, This work, therefore, represents an important
3
step towards achieving our ultimate goal of building in quality, long life, predictable and reliable
performance, durability and lower cycle costs of thick-section thermosetting composite structures
for future Army systems.
In the following section the problem formulation is discussed. The governing equations and
boundary conditions describing the thermal and chemical interactions associated with the two-
dimensional anisotropic curing process are presented. Solution of the pertinent equations utilizing
the boundary fitted coordinate system (BFCS) transformation technique in conjunction with the
alternating direction explicit (ADE) finite difference method is then described.
3 Analysis
3.1 Assumptions
The analysis assumes that gradients in temperature and degree of cure normal to the cross-section
of the part geometry are negligible. This assumption is valid for large area parts where edge effects
are minimal. All thermal and cure kinetic material pararheters are assumed constant, independent
of time and temperature. It is also assumed that no resin flow or part thickness reduction occurs
during the curing process. The no-resin flow assumption is reasonable for the large area, net resin
or low bleed thermosetting composite systems investigated in this study.
3.2 Heat Conduction Equation
Temperature solutions are based on Fourier's heat conduction equation for two-dimensional, tran-
sient anisotropic heat transfer with constant material properties and an internal heat generation
source term. The equation is well established and is referenced here for completeness as [22]:
+ .. 2T+ k 2T+ . 0T= PC O
for T(xz) in (1)
where D5 is the domain of interest defined in an orthogonal (x, z) coordinate system. The term q
represents internal heat generation and k,., k.;, k., are the effective anisotropic thermal conduc-
4
tivities, p is the density, and cp is specific heat of the composite. T and t are absolute temperature
and time, respectively,
The coordinate directions in equation (1) are defined in a fixed (x, z) global coordinate system.
Fiber-reinforced composites exhibit anisotropic thermal properties defined in a principle coordinate
system with coordinate axes parallel and perpendicular to the fiber direction. Fiber orientation will
generally vary with respect to the global coordinate system in an arbitrary shaped geometry. The
effective anisotropic thermal conductivities in equation (1) are bAsed on the second order tensor
transformation of the principle thermal conductivities given by:
k. = n2 "12 -ran k3 (2)
krx -mn rnn mn 2 - n 113
In equation (2), m = cos(O),n = sin(O) and kll, k3 and k
13
are the longitudinal, transverse
and cross-term thermal conductivities of the composite in its principle (1,3) material coordinate
system, respectively. Fiber orientation within the domain is assumed coincident with the curvilinear
coordinate system, (q,4), shown in Figure (1). In our simulations, the fiber orientation is coincident
with j7 = constant grid lines in the domain. This definition of fiber orientation requires that layers
witlin the laminate conform to the tool surface profile and do3s not accommodate laminate stacking
sequence variations.
The rotation angle, 0, is defined by the local orientation between the fiber direction and the
global (z, z) coordinate system. Transformation at a node in the domain is graphically illustrated
in Figure (1) where the orientation angle between the principle (1, 3) and global (x, z) coordinate
systems is defined,
3.3 Temperature Boundary Conditions
A generalized temperature boundary condition formulation is used to permit flexibility in the sim-
ulation of the autoclave curing process. Either Dirichlet, Neumann or Robin boundary conditions
may be enforced on the domain boundaries. The generalized boundary condition is expressed
5
zz
x
Figure 1: Thermal Conductivity Transformation Between Coordinate Systems
6
mathematically as;
a-A + bT, + c1(t) = 0
for T(xz) on ah (3)
where Ob represents the domain boundary or part surface. The surface boundary temperature is
T, and fA is the outward unit normal to the domain surface. The coefficients a, b, and c define
the effective heat transfer across the domain boundaries. The expression T(t) in equation (3) is
interpreted as either the ambient autoclave temperature cure cycle or the actual boundary surface
temperature, depending on the values of a,b and c specified. Table (1) summarizes the three
possible boundary conditions obtainable from this generalized formulation.
The Neumann or insulated boundary condition is used when symmetry conditions are imposed
to conserve computation time by reducing the number of nodes required for the simulation. The
Dirichlet or prescribed temperature boundary condition is most useful when experimental transient
part surface temperatures are known. In the presence of tooling or a bag assembly, these temper-
ature profiles are not known apriori and may be very different from the cure cycle. In this case,
the Robin boundary condition is used where (h/k).ji defines the effective heat transfer coefficient
quantifying the heat flux from the actual part surface to the environment. This effective boundary
condition accounts for all thermal resistance associated with the tooling and bag assembly.
3.4 Chemical Kinetics
The term 4 in equation (1) represents the instantaneous heat generation per unit volume of material
and is introduced to account for the exothermic chemical reaction associated with the curing process.
It is a cure rate dependent term which is evaluated throughout the domain at every time step during
the incremental solution.
The degree of cure, ax, at a material point is defined as the ratio of the cumulative heat liberated
from the chemical reaction, 11(t), to the total heat of the reaction, H.. This is mathematically
expressed as:
c = !(t)(4
'7 (4)
The heat liberated at any point in time, t, is expressed in integral form:
fo p dt
dq/dt is the rate of heat generation from the cure reaction. The total b-at of reaction is similarly
expressed as:
11,=[ !(.!)dt (6
r o p dt
where t] is the time for complete reaction. Heat generation at a material point is identically zero
for t> t1. Equations (4), (5) and (6) are combined to yield an expression for the cure rate:
da 1= (d) 1 (7)
dt -pH,. dt H
'The cure rate, dot/dt, is a time and temperature dependent function of the reacting material system
and is typically determined empirically with isothermal Differential Scanning Calorimetry (DSC)
techniques. The cure rate is often expressed in terms of time, temperature and degree of cure in
some form of the Arrhenius rate equation. Details of the experimental procedures for determining
cure rate expressions and the total heat of reaction are found elsewhere [23,24]. Once da/dt and 1,.
for tkhe material are known, the heat generation term in equation (1) is straightforwardly obtained
by rearrangemer.t of equation (7):
dia
ql = pgrj (8)
3.5 Initial Conditions
The model formulation is flexible enabling arbitrary initial temperature, Ti, and degree of cure, ai,
distributions throughout the domain at the start of the simulation to be specified. Uniform initial
conditions are specified in all simulations presented in this study:
T(x,z) :T, in b at t=0
a(r,z) =a in D at t.- 0 (9)
Ti and ai are taken to be the ambient temperature and zero, respectively.
8
3.6 Boundary Fitted Coordinate System Transformation Technique
The boundary fitted coordinate system (BFCS) transformation technique is used to solve the gov-
erning equation (1) and generalized boundary condition (3). The BFCS technique is a mapping
technique in which coordinates in a physical curvilinear coordinate system, (z, z), are transformed
into a computational rectangular coordinate system, (, q), Motivation for using the technique
in the present investigation is its advantage to easily accommodate complex shaped geometries.
Once tiIe governing equations and boundary conditions are transformed into the computational
domain, a straightforward finite difference solution technique is applied. Solutions are obtained in
the computational domain and subsequently mapped back into the physical domain through the
correspondence of nodes in the respective planes. A conceptual representation of the relationship
between the two coordinate systems is illustrated in Figure (2), where a one-to-one correspondence
between nodes in each coordinate system is noted. The transformation from the physical domain
into the computational domain is based on the Poisson equation.
The BFCS technique is well documented [25,26], so derivation and details of the coordinate
transformation procedure are omitted for brevity, Utilization of the BFCS technique in this inves-
tigation involves both a mesh generation procedure and coordinate transformation of the governing
equation (1) and boundary condition (3). All the finite difference meshes presented in this investi-
gation were produced with the mesh generation code TGMESH developed by Gilmore [27],
Transformation of the governing equation (1) through the Poisson equation into the computa-
tional domain yields:
C 2T A 0
2
T " 9
2
T C T CT CT
+ t 7 +A LT +A OT+ A
4
--+ As- E =PC 07 (10)
where the spatial coordinates in the computational domain are and ij, Finite difference formulas
for the spatial and time derivatives appearing in equation (10) are presented in the Appendix, The
coefficients, A&, are functions of the effective thermal conductivities, kxxl kzz and k,,, and spatial
derivatives in the physical domain and are defined in the Appendix.
Transformation of the generalized boundary condition (3) is also required, Details of the trans-
formation prncedure are presented elsewhere [28]. Consequently, only the final expressions are
9
z
x
Physical Curvilinear Coordinate System
I
Jacoblan
Transformation
- --
Computational Rectangular Coordinate System
Figure 2: Conceptual Representation of the Physical and Computational Domains
10
presented here. Neumann and Robin boundary conditions are functions of the temperature gra-
dient on the domain boundary. The computational domain has four separate faces comprising its
entire surface as indicated in Figure (3). Each face or side of the domain has a unique expression
for the temperature gradient on the boundary surface, OT./Ohi, given by:
iT cT-J7T, on the fi(+) rc
- 7' on the f(+q) face
on ihe o . (-) face (11)
04 on ~the ii(-t
1
)fac
the coefficients a, and -y and the Jacobian coordinate transformation, J, are functions of the
spatial derivatives in the physical domain, (x, z), and are defined in Lhe Appendix, Substituting
equations (11) into equation (3) yields the expressions for the transformed generalized temperature
boundary conditions on each face of the computational domain:
a aT
-07'
3iI +bT. + cT(t) = 0 on the fi(+e) face
3 ,2/T + bTa + cT(t) = 0 on the Ai(+7
1
) face
a + bTa A cT(t) = 0 on the A(-) face (12)
a,,-'3 + bT, + cT(I) = 0 on the Ai(- q) face
Transformations on temperature, (i.e. T, and T(t)), are not required since scalars transform identi-
cally. Equations (10) and (12) constitute the transformed heat transfer equation and temperature
boundary conditions governing the transient temperatures distributions in the composite during the
curing process. The solution to the system of resulting finite difference equations is now presented.
3.7 Alternating Direction Explicit Finite Difference Method
The Alternating Direction Explicit (ADE) finite difference method [20] is employed in the solution of
equations (10) and (12). The ADE method is preferred over an implicit method because substantial
reduction in computation time is realized, The ADE method does not require the inversion of the
11
1o
Figure 3: Normal Derivatives on the Cornputational Doinain BLotiLdaries
12
temperature coefficient matrix encountered in the implicit approach, It is also preferred over a
fully explicit method since it has been shown to be unconditionally stable with time step size, and
is less sensitive to computational errors [29].
The ADE finite difference method applies a fully explicit approach twice on the computational
domain for each time step increment. One sweep through the finite difference mesh is made in
a systematic forward direction where a pseudo-temperature solution, ui,j, at each interior node is
defined explicitly in terms of adjacent nodes, For the same time step, another sweep is made in
exactly the reverse direction to obtain a second p'uodo temperature solution, vi,j. The temperature
at node i, j, (Tij), is then computed at each interior node for the current time step as the arithmetic
average of the two pseudo-temperature solutions:
__,, U= 4 Vi j (13)
2
The computational mesh of nodal dimensions m by n is illustrated in Figure (4), Central
differencing formulas are employed at all interior nodes, while forward and backward differencing
formulas are used on the boundaries nodes, All finite differencing formulas are documented in the
Appendix,
Following the detailed procedure described by Barakat and Clark [29], the expression for uij at
the time step t + At for an interior node during a forward sweep is given by:
,Ai =(1/((pcp/At) + Al + A
2
))i~j [ (pcp/At),juLj + ij +
~ ~ APJuJ
1
-u + ~)+
(A/4-~ -uI~~ - + u+..)
(A 4)ij(ul + 1j - ui-Aj) + (A-2)i) (14)
-- lJ' ( !j+ I - uit'j +,-All
The uij's at time step t + At on the right hand side of equation (14) all have previously been
computed during that same time step, (i.e. those with subscripts (i - lj), (i,j - 1), (i + 1Ij -- 1)
and (i - 1,j - 1)). Similarly, vij at time step t + At for anl interior node during a be ckward sweep
is given by:
v" = (1/( ((pc/At) + Al + A.
2
))ij, [ (pcp/ t)i,.ivj + 1,1j ,
13
114
i ni
(n-l) . ..-
j -__
1 2 3 (m-l)mr
F'igure 4: 'rhe Complutati ovial Mesh
14
( vf,j + v-_,.) + (A2)i,j(Vit4 - v,
1
j -+- v,J
1
) +
( .4Iq\tot+t t I\fl!+At + )t
(A4/2)1j(vi+$J - v0-IJ +" ,(A
5
/2) ,j(v4 i vt.. -)] (15)
where tLe vij's at time step t + At on the right hand side of equation (15) are those previously
computed during the same time step, (i.e, those with subscripts (i + 1,j), (i,j + 1), (i - 1,j + 1)
and (i - l,j - 1)).
Boundary nodes are treated in the usual fully explicit fashion since no time derivatives appear in
the transformed generalized temperature boundary condition expressions. An explicit representa-
tion for the boundary node temperature, T., in terms of adjacent nodes is obtained by substitution
of the appropriate one-sided three-point differencing formulas presented in the Appendix into the
generalized boundary condition equations (12).
The heat generation term, 4!j, in equations (14) and (15), is computed from equation (8),
rewritten here as:
PH, do)(16)
4L(Lt ij
where (da/dI)Ij is the instantaneous cure rate at node (ii) evaluated at time step t. Instantaneous
cure rates are computed based on instantaneous temperature and degree of cure through an em-
pirical rate expression. Temperature distributions at time t + At are obtained by solving the finite
difference equations (14) and (15) and substituting into (13) in conjunction with the appropriate
explicit representations of the boundary temperatures defined in equation (12).
3.8 Degree of Cure Calculation
The degree of cure over each time step increment during the simulation is based on the degree of
cure at the previous time step and the instantaneous cure rate at the current time step. The degree
of cure at node (i, j) for time step t + At is computed from the relationship:
at~at t (dt+At
at = a!,j + (\",)t+A At (17)
15
The value of alj is known since it was computed at the previous time step increment and (da/dt)AJt
is also known, from the empirically determined cure rate expression,
Temperature and degree of cure distributions throughout the domain are computed at each time
step incremeni. as a function of the cure cycle temperature history. A computer code was written
to perform the numerical calculations. Results of this investigation are discussed in the following
section.
4 Results and Discussion
The cure simulation analysis developed is used to gain a fundamental understanding of the curing
process unique to thick-section composites. A summary of the input is presented first, Verification
of the cure simulation analysis is demonstrated by comparison with available exact solutions for
simulated temperature and degree of cure profiles. Predicted temperature profiles are compared to
experimentally measured values within several 2.54 cm thick glass/polyester laminates exposed to
various temperature histories. The influence of the tooling on the curing process is demonstrated
through the generalized boundary condition formulation. The influence of laminate thickness and
the "temperature cure cycle ramp on non-uniform through-the-thickness curing are illustrated to
demonstrate the complex curing phenomena unique to thick-section composites. Several typical
glass/polyester and graphite/epoxy structural elements of arbitrary cross-section are analyzed to
show the strong dependence of temperature and degree of cure gradients on geometry and material
thermal anisotropy.
4.1 Input Summary
Required input data for the cure simulation analysis includes; (1) solution details including the
boundary and initial conditions, (2) the finite difference mesh representation of the part geometry,
(3) thermal properties and (4) a cure kinetic description of the composite material system.
16
4.1.1 Solution Details
The desired cure cycle thermal history, T(t), is segmented into descrete time step increments, At.
Since the ADE method used in this investigation is unconditionally stable with the time step size,
little difficulty in obtaining a converged solution was encountered. In general, time step increments
between one and two seconds proved to be sufficiently small to yield converged solutions for all the
cure simulation results presented.
The transient temperature boundary conditions for the analysis are based on the specified
temperature cure cycle history and the effective heat transfer coefficients a, b and C defined in the
generalized boundary condition discussed previously. Initial conditions are defined in equation (9).
4.1.2 Mesh Input
The part geometry is discretized into a suitable finite difference mesh consisting of rn by n nodes,
The physical (x, z) coordinates of each node in the mesh are used to evaluate the spatial derivatives
appearing in transformed governing equations. The finite difference meshes employed in this inves-
tigation were generated with a computer software code TGMESH[27] that is based on the BFCS
technique, Three geometries typical of structural composite components are used in this investiga-
tion; a flat plate, a 90" right angle bend and a ply-drop, The finite difference mesh representations
of these geometries are shown in Figure (5). The flat plate and right angle bend geometry are
of constant thickness equal to 2.54 cm. The ply-drop geometry consists of a 1.27 cm flat section
joined to a 2.54 cm flat section by a sinusoidal transition region. Fiber orientation in each geom-
etry is assumed coincident with the (r} = constant) grid lines. Solution accuracy is influenced by
refinement of the finite difference mesh. Mesh refinement was chosen to yield converged solutions
in all the cure simulations performed in this investigation.
4.1.3 Thermal Properties
The thermal properties used in this investigation for the glass/polyester and graphite/epoxy com-
posites are summarized in Table (2). The cross-term thermal conductivity, k
13
, in the principle
17
(a) Flat Plate Geometry Mesh
(b) 900 Right Angle Bend Geometry Mesh
(c) Ply-Drop Geomet~y Mesh
Figure 5: Cure Simulation Mesh Geometries
coordinate system of both materials is identically zero.
4.1.4 Cure Kinetic Data
The complete description of the cure kinetics for the composite includes the total heat of reaction
and a description of the rate of reaction as a function of temperature and degree of cure. Reaction
rate expressions for the glass/polyester and graphite/epoxy material systems are different in form
due to the inherent differences in the overall order of the reaction kinetics.
The glass/polyester composite consists of CYCOM 4102 polyester resin, manufactured by the
American Cyanamid Corporation, and a woven roving E-glass plain weave fabric reinforcement
containing approximately 6 yarns per inch. The reaction rate expression for the glass/polyester
system is second-order overall [24]:
S= Acezp(-AE/RT)a'cn(1 - a)fc (18)
dt
R is the universal gas constant and T is absolute temperature. The exponents mr and n,, the
pre-exponential coefficient, A,, the activation energy, AE,, and the total heat of reaction are listed
in Table (3).
The graphite/epoxy composite contains Hercules Corporation's 3501-6 resin, reinforced with
unidirectional AS4 graphite fibers, The reaction rate expression for the graphite/epoxy system
follows a markedly different form [12]:
da
T = (ki + k
2
a)(1 - a)(0,47
- a)
for (a < 0.3)
S= k3(I - a) for (a > 0.3)
(19)
kj, k
2
and k
3
are defined by the Arrhenius rate expressions:
kt = Aiexp(-AEi/RT)
k2 = A
2
exp(-AE
2
/RT) (20)
k
3
= Aaerp(-AE
3
/RT)
19
Dirichlet (prescribed) a = 0 b = I c = -1
Neumann (insulated a =1 b = 0 C = 0
Robin (convective) a = 1 b = (h/k)o
0
c = -(h~k).of
Table 1: Generalized Boundary Condition Coefficients
p [kg/mr] c (kJ/(W.C)] k33 [kW/(tn -. C)] k
11
/k3
3
Glass/Polyester 1.89x10 1.26 2.163x10
4
2
Graphite/Epoxy 1.52x10
3
9.42x10-
1
4.457x10-
4
1,5,10
Table 2: Thermal Properties for Glass/Polyester and Graphite/Epoxy Composites
Glass/Polyester [24].
me 0.524
nc 1.476
A,[min.-
1
] 3.7x101
2
AE,[J/mol] 1.674x10
5
H,[kJ/kg] 77.5
Graphite/Epoxy [12]
Al[min.-'] 2,102x10
9
A
2
[[min.-'] -2,014x1Q
A3[min.-'] 1.960x10
5
AEl [J/reol] 8.07x10
4
AE
2
[J/mol] 7.78x10
4
AE
3
[J/mol] 5.66x104
H,[kJ/kg] 198.9
Table 3: Cure Kinetic Parameters for Glass/Polyester and Graphite/Epoxy Composites
20
The pre-exponential coefficients Al, A
2
and A
3
, the activation energies, AE
1
, AE
2
and AE
3
, and
the total heat of reaction for the graphite/epoxy composite are summarized in Table (3).
4.2 Model Verification
Cure simulation results were verified for accuracy by comparing predicted temperature and degree
of cure profiles with available exact solutions. Temperature predictions are based on an analytical
one-dimensional transient temperature solution [22]. Temperature profiles in a glass/polyester
laminate of thickness t=2.54 cm, initially at 00C and exposed to constant boundary temperatures
of 00C at z=0(bottom surface) and 100C at z=e(top surface), were compared at z=t/4, z=t/2
and z=3U/4 locations. The cure simulation mesh utilized is the flat plate geometry illustrated in
Figure (5), employing insulated boundary conditions on the sides to isolate one-dimensional heat
transfer effects. Excellent agreement with the exact solution is shown in Figure (6), providing
confidence in the accuracy of the cure simulation analysis.
Degree of cure profiles obtained from the analysis were compared to independent results based
on the cure rate constitutive relations. Degree of cure profiles in a glass/polyester flat laminate
under uniform, isothermal conditions are compared in Figure (7). Results are in excellent agreement
which confirms the accuracy of this portion of the analysis.
4.3 Experimental Correlation
Experimental temperature distributions within 2.54 cm thick flat plate glassy/polyester laminates
processed under various arbitrary autoclave temperature cure cycle historiec were correlated with
the cure simulation predictions. Thermor.ouples were embedded within the laminates at various
locations through the thickness and temperature distributions were monitored during the autoclave
cure cycle. The laminates, 15.24 cm by 15.24 cm by 2,54 cm thick in dimension, were constructed by
stacking together approximately 42 plys of prepreg. The assembly was placed on a 0.635 cm thick
aluminum caul plate, topped with a single layer of bleeder cloth and surrounded by an aluminum
dam to prevent transverse resin flow and compaction, The entire assembly was placed within a vac-
uum bag, sealed to the caul plate with tacky tape and drawn to one atmosphere of vacuum pressure,
21
8- I i I I 1 I I
6 I14 0/0 '''
w /"
1/42
1-1
4-
"OX
31 4Analytic
V 3Solutions
31/4
- - - --
2 10
A . A ' *A " _ 0 i
2 A
d A p ASimulated
. AResults
TIME (SECONDS)
Figure 6: Verification of Temperature Solution in a Glass/Polyester Laininate
22
120 C
1.o I*
-,-
- . ---- ". .
., "110
C
',
0.8 C A
cc
41000c
o 06 : ,I
I-
LL -m- Analytic
UJ 0
... Solutions
cct
.10
0D
1
?
w x
0.20.4
.. 0.2 ' 0 Simulated
a0 X40CX
Results
0.0 - 1__ _1 _ __1 ___1
TIME (MINUTES)
Figure 7: Verification of the Degree of Cure Solution in a Glass/Polyester Laminate
23
The laminates were then placed into the autoclave and subjected to various prescribed tempera-
ture cure cycle histories. During the cure, temperature profiles at the various locations within the
laminate were monitored with an IBM personal computer linked to a K EITHLEY data aquisition
system. Thermocouple readings were sampled every 30 seconds to ensure accurate temperature
measuruments during points in the cure cycle when large temperature gradients develop.
The thermal properties and cure kinetic input data presented previously for the glass/polyester
system and the flat plate mesh geometry were used in the correlation. A parametric study identified
the effective heat transfer coefficient, (h/k).
11
, to be 87[1/m] on the top surface and
1
25[1/m] on
the bottom surface of the laminate assembly in contact with the aluminum tool surface. In addition,
insulated boundary conditions were employed on the sides to isolate through the thickness effects.
Comparison of numerical and experimental transient temperature profiles at the center of the
laminates subjected to three different cure cycles are presented in Figures (8), (9) and (10), Good
agreement in this experimental study offers additional independent validation of the cure simulation
analysis.
4.4 Boundary Condition Effects
The presence of the tooling, bag assembly and the internal autoclave environment can profoundly
influence the heat transfer to the composite and may ultimately alter the curing process within
the part, The net effect is that the surface temperature variation of the part may be significantly
different than the prescribed cure cycle temperature. While convective heat transfer due to heat flow
within the autoclave is not modeled explicitly, flexibility with the generalized boundary condition
formulation enables the influence of the effective heat transfer to the part on hie curing process
to be investigated, In the limit as (h/k)1 "-- oo, the actual surface temperature of the part
approaches the prescribed cure cycle temperature, T(t), As (h/k).
1
-, 0, an insulated boundary
*%ondition is approached, The generalized boundary condition formulation enables the influence of
various tools and bagging assemblies on the curing process to be investigated.
The following example denmontrates the, influence (h/k),
1
f, can have on the curing process, A
2.54 cm thick glans/polyester laminate was subjected to cure cycle ternperature history indicated
24
120
100 AUTOCLAVE
w
S80
SIMULATION
40
"EXPERIMENTAL
0
50 100 150
TIME (MINUTES)
Figure 8: Temuperature Comparision with Ghass/Polyester [,alalinlato 1
25
100
AUTO CLAVE
ui
I -
~60 SIMULATION
4.
EXPERIMENTAL
50 100 150 200 250
TIME (MINUTES)
Figure 9: Tlmpoerature Comuparisou with G4 lass/Polyostor Laini nato 2
26
120 I - I I
100
"_ AUTOCLAVE
~80
SSIM SIMULATION
60 0
600
, EXPERIMENTAL
40
50 100 150 200
TIME (MINUTES)
Figure 10: ni, eratLure Comparison with C4lasss/ Iolyvter lazuiii aL, 3
27
in Figure (11) with specified (h/k),ef coefficients on the top and bottom surfaces of 100[1/rn] and
500[1/m], respectively, The flat plate mesh with insulated boundary conditions on the sides was
used to isolate through-the-thickness efrects, Resulting temperature profiles at the surface and
center of the laminate are shown in Figure (12), Th'eli, higher (h/k)ej] coefficient perrmits a more
rapid transfer of heat into the laminate, causing the laminate to heat up and cure faster. This also
allows more heat to escape from the laminate during the exotherm, resulting in an overall lower
exotherm.
Corresponding degree of cure profiles, shown in Figure (13), demonstrate the significant influ-
ence (h/k)ej] can have on the curing process. Lower (h/k),ff increases degree of cure gradients
during the exotherm. The effective heat transfer across the laminate surface plays an important
role in the development of residual stress and warpage during processing [10],
4.5 Thickness Effects
The thickness of the laminate will strongly influence the curing process, The effect of increasing
laminate thickness on the temperature and degree of cure profiles is examined in glass/polyester
lam.pates between 1.38 cm and 5,08 cm thick, Flat plate mesh geonmetries of similar nodal density to
the flat plate mesh illustrated in Figure (5) were employed, Prescribed temperature boundary coii*.
ditions defined by the glass/polyester cure cycle temperature history illustrated in Figure (11) were
imposed on the top and bottom laminate surfaces, Insulated boundary conditions were enforced
on the sides to isolate through-the-thickness processing effects. Predicted centerline temperature
and degree of cure profiles are illustrated in Figures (14) and (15), respectively,
Fen'mperature distributions at 164 minutes into the cycle, shown in Figure (16), demonstrate that
while the 2.54 cm and thinner laminates are exothermuing, the 5.08 latniinate is still heating up aLt this
point in the cycle, Similar gradients in the degree of cure distributions at 164 minutes are shown in
Figure (17). The degree of cure gradients ate most severe at this poina in the 2.54 cmi laminate, The
interior of the 2,54 cmr laminate is essentially cured. In contrast, the 5.08 cm laminate is essentially
uncured at the interior, At a later point in the cure cycle, the thicker laminate will exotferin anid
its distributions will reverse shape similar to the thinner laminates, ultin-ately developing more
28
200- I ,
177 C Graphite/Epox"
175-
-
Cure Cycib-
I
150- I
126 C
100 90 C
cc
w Glass/Polyester
50
Cure
Cycle
25
2
C
0I -
0 100 200 300 400
TIME (MINUTES)
Figure 11: Typical Glass/Polyester and Graphite/Epoxy Temperature Cure Cycles
29
140- - I
120 --- z-0, 100
z -0. 00
- - 1/2, 100
t10 - - z1l/2, 500
0,
,'.O'er
1PL,
80-
;00;
0 - 0
40-
0 100 200 300 400
Time (minutes)
Figure 12: Influence of (h/k)ef on Temperature Profiles in a Glass/Polyester Laminate
30
1.0- * I I
Z-0, 100 I /
I -- - - 2 -0, 500
I
/
- z-I/2, 100 I -
0.6 " z 1/2, 500
0.6-
o i
aIi
0 0.4-
"o 1
0.2 - "
0.0.
/
0.0- - ... ....
100 200 300
Time (minutes)
Figure 13: Influence of (h/k),ff on Degree of Cure Profiles in a Glass/Polyester Laminate
31
4,.
- 1.38 am
140- - - 1.85 am
"
"- 2.34 0m
%
120- -"--5.06
am '
"
80 - --
so
10
106 00200
3
40 -
20-
100 200 300
Time (minutes)
Figure 14: Influence of Thickness on Centerline 'Temperature Profiles in Glass/Polyester Laminates
32
1.0-
- 1.38cm |
1.865m am
0.6
2.54:1
m
/ I
-- -5.08 am
:~ I
0.6
o I,
h..
0.4
"o I
I I
7 II
I
0. .2-
0.0 -"...'
10o 200 300
Time (minutes)
Figure 15: Influence of Thickness on Centerline Degree of Cure Profiles in Glass/Polyester Lami-
nates
33
severe gradients. These results demonstrate the complex temperature and degree of cure gradients,
unique to thick-sections, which develop during the curing process. Bogetti and Gillespie [10] have
shown that these gradients have a profound influence on the evolution of process-induced stress
and deformation.
4.6 Cure Cycle Temperature Ramp Effects
The cure cycle uemperature ramp can strongly influence the temperature and degree of cure gra-
dients that develop during the cure. The effect of the temperature cure cycle ramp on the degree
of cure gradients is examined in a 2.54 cm glass/polyester laminate exposed to various cure cycle
temperature ramps. The flat plate mesh with specified temperature boundary conditions on the
top and bottom surfaces and insulated boundary conditions on the sides is used here, The laminate
was cured at 80'C for 180 minutes before being subjected to temperature cure cycle ramps of 0.0,
0.25, 0.5 and 1.0 *C/min,, illustrated in Figure (18), Predicted values of the degree of cure at
the centerline minus the degree of cure on the surface of the laminate, (ac - a
5
), are plotted in
Figure (19) as a function of cycle time for the various temperature ramps investigated.
With thiL particular laminate thickness, 2.54 cm, the isothermal cure under the 0.0[
0
C/nin.]
ramp reveals a steadily increasing positive value of (a, - a,), typical of an internal exotherm.
Above a critical temperature ramp, the surface temperature rises relative to the interior since heat
transfer by diffusion is low. Consequently, the surface temperature initiates the cure reaction. The
exothermic reaction accelerates the cure and as a result creates a solidification front that sweeps
the laminate from the outside towards the interior of the laminate. Curing from the outside to the
inside may potentially entrap voids and volatile by.products of the chemical reaction and enhance
warpage and residual stress development. Consequently, the temperature ramp can significantly
influence the quality and in-service performance of the part.
4.7 Anisotropic Curing in Complex Shaped Composites
The anisotropic curing process in arbitrary shaped geometries is now presented to build another
level of complexity in the cure simulation that enables realistic composite structures to be modeled.
34
0.5
0.3
z
o \ UNITS: CM
I /*
S0.1 5.08
w
j1.38 ~ 1.85 2.54
0
-
-40
0 .... - -
\ - "
90 100 110 120
TEMPERATURE (C)
Figure 16: Temperatute Distributions in Glass/Polye.ster Laminates at Exotherrnl
35
UJ ! 4
Co / . -i--
- ,.
UNITS: CM
w
/
1 - 8
1.85
2.54
zI
0' U., ,05" 5o8 ~ a8
-
\ / , --
- Id
DEGREE O F CURE(%
Figure 1.7: Degree ofCure Distributions in Glams/Polyester Laminates at Exotherin
36
140-
0.5 0.25
/ -- 'I - --
120- 1.0 /'. "
*oO / ,a. ,I
,100- -
h
.- - --
0.0.0
S60-
,
E
s
E
units:
C/minute
I
40-
20
0 -I
0 100 200 300 400
Time (minutes)
Figure 18: Autoclave Tempcrature Cure Cycle Ramps
37
0.5I I i
0.00 C/mmn
0.4 - - - 0.25 C/mmn
0,80 C/mln *
SC1.00
C/mln
0.3,
cn
t,\
2 0.2-
0.0"
0
-0.
.0.2
-0.2-
100 120 140 160 180 200 220 240 260 200 300
Time (minutes)
Figure 19: Effect of the Tlemperature Ramnp on Non-uniformi Curing in a. 2.54 cin Glass/Polye~ter
L~aminate
38
The effect of anisotropic heat transfer on the curing process in a 2,54 cm thick 90' right angle bend
graphite/epoxy structural element is examined. The angle bend mesh depicted in Figure (5) is
used. A specified boundary condition defined by the graphite/epoxy cure cycle temperature history
illustrated in Figure (11) was imposed on all four sides of the part, kiu/k
33
ratios of 1, 5 and 10,
typical of graphite/epoxy composites, were investigated.
Temperature contours are shown in Figure (20) at 145 minutes into the cure cycle, when the
maximum exotherm occurs. The isotropic case, kit/k33 = 1, exhibits through-the-thickness tem-
perature gradients resembling a one-dimensional curing process, In contrast, kjt/k
33
ratios of 5 and
10 demonstrate that temperature gradients at the bend are most severe at the inner radius and are
accentuated by increased thermal anisotropy, This simulation clearly illustrates the significance of
the anisotropic two-dimensional formulation used in our cure simulation analysis.
The curing process within a glass/polyester 900 right angle bend part is also examined. A
kii/k
3
a ratio of 2 is assumed in the simulation, Prescribed temperature boundary conditions
defined in Figure (21) are imposed on all four sides of the part. This cure cycle is similar to the
glass/polyester cycle defined in Figure (11) except the internmedia, temperature ramp to 90
0
C is
omit.ted.
'tremiperature and degree of cure contours at the time of maximrnum exotherm (165 minutes into
the cycle) are illustrated in Figure (22). The complex temperature gradients indicate "hot" regions
in each arm of the part geometry, moving inward towards the bend as they follow a sweeping cure
front. The exotherm here is occurring on the outer regions of the part, sweeping toward the interior,
The degree of cure contour at this point in the cycle indicates an outside to inside curing proces,
Severe gradients are noted with the surface fully cured and the interior uncured, This curing process
will inevitably entrap any voids or chemical by-products, reuult in incomplete consolidat6ion and
potentially induce severe stress and d(eformation in the part, during the curing process,
A final example examines the cure of a thick-section glass/polyoster ply-drop structural element
illustrated in Figure (5). Prescribed temperature boundary conditions defined in Figure (21) are
imposed on the top and bottom surfaces while insulated conditions are imposed on the sides, A
kj
1
/k
3 3
of 2 is used. Temperature and degree of cure contours at the time of mnaximumn exotherm
39
TEMPERATURE (C)
194,
192.
188.
186,
185.
b j/3 ,
184.
182.
180.
1 79.
(C) k
1 1
/k3 = 10.0
Figure 20: Temperature Contours at ExoLherni in a 2.54 er Th~1ick Graphlite/ Epoxy 900 Right
Angle Bond
40
126
120-
100-
40
.20
0 100 200 300
Time (minutes)
Figure 21: Coiidonued CGIaN,/Polyc,,toer Cure Cycle
41
140,
138.
135.
132.
TEMPERATURE
(C)
129.
127,
124.
121,
118M
116.
113.
110.
107.
105,
102. (a) Temperature Contour tt Exotherin
0.94
0.88
082 DEGREE OF CURE (%)
0.71
0.65
0,59
0,53 "x
0,47
0.41
0.35/-\,
0.30 /
0.24 )
0.18 "
0.12
(b) Degree of Cure Con1tour ati IxoLherin
Figure 22: Aniisotropic Curing in it 2.54 cm Thick Glass/ Polyemt,,r 9)0' Right, AMighI, olnd
42
(174 minutes) "re presented in Figure (23). The results identify a "hot" region in the transition area
with teimperatures exceeding the autoclave by 45C. The minimum temperature occurs within the
interior of the thicker section. The corresponding degree of nure cortour indicates severe gradients
in the thicker section where the exterior region is fully cured and ',.,e inlericr is only 12% cured.
In contrast, the thinner section of the ply-drop is completely cured at this portion of the .,roccss.
The cure front is elfectively sweeping from the thin section to the thick section.
The examples presented in this investigation clearly demonstrate the complex cure behavior
encountered in the manufacture of thick-section thermnoseL structural components. Complex tem-
perature and degree of cure gradients, accentuated by part geometry and anisotropic heat transfer,
demonstrate the significant contribution made towards the processing science of thick-sectioon ther-
mosets. The cure simulation analysis developed here provides the detailed prerequiste informatio'n
needed for predicting process-induced stress and deformation in thick-sections of arbitrary cross-
section. This achievement represents a critical step in our ultimate goal of optimizing the quality
and in-service performance of thermosetting structural components during the manufacturing pro-
cess,
5 Conclusions
A two-dimensional anisotropic cure simulation analysis was developed to study the complex curing
process within thick thermosetting composites of arbitrary cross-section. Complex gradients in
temperature and degree of cure were predicted as a function of the autoclave temperature history
and the influence of the tool on the curing process was demonstrated. Several typical glass/polyester
and graphite/epoxy structural elements of arbitrary cross-section were analyzed to provide ins'ght
into the non-uniform curing process unique to thick-sections. Spatial gradients in degree of cure
are shown to be strongly dependent on part geometry, thermal anisotropy, cure kinetics and the
autoclave temperature cure cycle. These spatial grazlients directly influence the quality and in-
service performance of the finished component by inducing warpage and residual stress during the
curing process. This work, therefore, represents an important step towards achieving our ultimate
goal of building in quality, long life, predictable and reliable performance, durability and lower
43
142.
,. 139.
136.
1 33.
130.
127.
124.
122. TEMPERATURE (C)
119.
116.
113.
107.
105.
102.
(a) Temperature Contour at Exotherm
0.94
0.89
0.84
0.78
0.73
0.67
0.62
0.56 DEGREE OF CURE(%
0.51
0.45
0.40
0.34
0.29
0.23
0.18
(b) Degree of Cure Contour at Exotherm
Figure 23: Anisotropic. Curing in a Thick Glass/Polyester Ply-Drop
44
cycle costs of thick-section thermosetting composite structures for future Army systems,
45
INTENTIONALLY LEFT BLANK.
46
References
[1] Hahn, H.T. and Pagano, N.JT Curing stresses in composite laminates. Journal of Composite
Materials, 9:91-105, 1975.
[2] Hahn, H.T. Residual stresses in polymer matrix composite laminates. Journal of Composite
Materials, 10:266.-277, 1976.
[3] Griffin, O.11. Three-dimensional curing stresses in symmetric cross-ply laminates with
temperature-dependent properties. Journal of Composite Materials, 17:449-463, 1983.
[4] Staugo, R.J. and Wang, S.S. Process-induced residual thermal stresses in advanced fiber-
reinforced composite laminates. Journal of Engineering for Industry, 106:48-54, 1984.
[5] Kau, H. and Petrusha, L.A. Dimensional stability and property gradients in thick srnc sections.
Technical Report GMR-6359, General Motors Research Laboratories, Warren Michigan, 1988.
[6] Kays, A. 0. Exploratory development on processing science of thick section composites. Tech-
nical Report AFWAL-TR-85-4090, Air Force Wright Aeronautical Laboratories, Wright Pat-
terson AFB, Ohio, 1985.
[7] Dickerson, G. E. Fabrication of thick laminates. Technical Report LAR-12019, NASA Tech
Brief, U.S. Patent No. 4,065,340.
[8] Meade, L. E. Fabrication of thick graphite/epoxy.wing surface structures, In 2Jth National
SAMPE Symposium and Exibition, Book 1, 1979.
[9] Levitsky, M. and Shaffer, B.W. The approximation of temperature distributions in hoinoge,
neous exothermic reactions. Tihe Chemical Engineering Journal, 5:235-242, 1973.
[10] Bogetti, T.A. and Gillespie, Jr., J. W. Process-induced stress and deformation in thick-section
thermosetting composite laminates. In 21st SA MPE Technical Conference, Atlantic City, N.J.,
September, 1989.
47
[11] McGee, S. H. Curing of particulate filled composites. Polymer Engineering and Science,
22:484, 1982.
[12] Loos, A, C, and Springer, G. S. Curing of epoxy matrix composites. Journal of Composite
Materials, 17:135-169, 1983.
[13] Woo, I. L., Ciriscioli, P. R., Dusi, M. R., and Peterson, D, R, Cure process rnodel of or-
ganic matrix composites in the presence of vapor release. Journal of Reinforced Plastics and
Composites, 5:3-8, 1986.
[14] Cambell, F. C,, Mallow, A, R,, Amuedo, K, C,, and Blase, G, A, Computer-aided curing of
composites, Technical Report IR-0355-5, McDonnell Douglas Corporation, 1985.
[15] Woo, I. L. and Springer, G. S. Microwave curing of composites. Journal of Composite Mate-
rials, 18:387-409, 1984.
[16] Dave, R., Kardos, J. L,, and Dudukovic, M. P, Procesd; modeling of thermosetting matrix
composites: A guide for autoclave cure cycle selection, In Proceedings of 'the American Society
for Composites First Technical Conference, pages 137-153, Dayton, Ohio, 1983,
[17] Mallow, A, R., Muncaster, F. R., and Campbell, F. C, Science based cure model for composites.
In Proceedings of the American Society for Composites First Technical Conference, pages 171-
186, Dayton, Ohio, 1986,
[18] Springer, G. S, Resin flow during the cure of fiber reinforced composites. Journal of Composite
Materials, 16:400-410, 1982.
[19] Levitsky, M. and Shaffer, B.W. Thermal stresses in chemically hardening elastic media with
application to the molding process. Journal of Applied Mechanics, pages 647-651, September
1974,
[20] Shaffer, B.W. and Levitsky, M. Thermoelastic constitutive equations for chemically hardening
materials. Journal of Applied Mechanics, pages 652-657, September 1974.
48
6
[21] Levitsky, M. and Shaffer, B.W. Residual thermal stresses in a solid sphere cast from a ther-
mosetting material. Journal of Applied Mechanics, pages 651-655, September 1975.
[22] Ozisik Necati M. Heat Conduction. Wiley Interscience, 1980.
[23] Morgan, R. J., Walkup, C. M,, and Hoheisel, T. H. Characterization of the cure of carbon
fiber/exopy composite prepregs by differential scanning calorimetry. Journal of Composites
Technology and Research, 7:17-19, 1985,
[24] Adams, D.C, Cure behavior of unsaturated polyester resin composites, Technical Report
CCM-88-16, Center for Composite Materials, University of Delaware, Newark, Delaware, 1988,
[25] Thompson, J. F., Thames, F. C,, and Mastin, C. W, Automatic numerical generation of
body-fitted curvilinear coordinate system for field containing any number of arbitrary two-
dimensional bodies, Journal of Computational' Physics, 15:299-319, 1974,
[26] Thompson, J, F., Waris, Z. A. U,, and Mastin, C. W. Body-fitted coordinate systems for nu-
merical solution of partial differential equations - a review, Journal of Computational Physics,
47:1-108, 1982.
[27] Gilmore, S. Tgmesh user's manual. Technical Report in press, Center for Composite Materials,
University of Delaware, Newark, Delaware, 1989,
[28] Trafford, D. Computational analysis and simulation of the mold filling process for hele-shhaw
flows. Technical Report CAE Report-3/86, Department of Mechanical and Aerospace Elngi-
neering, University of Delaware, Newark, Delaware, 1986.
[29] Barakat, H, Z, and Clark, J. A. On the solution of the diffusion equations by numerical
methods. Journal of Heat Transfer, pages 421-427, 1966.
49
INTENTIONALLY LEFT BLANK,
50
A Appendix
A.1 Finite Difference Approximations
Finite difference approximations to the partial derivatives used in the analysis are presented below,
In the following formulas, the temperature, T, and the spatial coordinates, x and z, are repre-
sented symbolically as f. In addition, subscripts tj and denote partial differentiation with respect
to the computational coordinates and the subscripts i and j refer to the nodal location on the
computational grid. Both central and one-sided differencing expressions are utilized.
A.1.1 Central Differencing
Three-point central differencing formulas are used at the interior nodes of the discretized domnain.
The first order derivatives are given by:
f = (fi+,, - fj
1
j)/MA (21)
and
"fn = (fij+i -- fij,-1)/2Aq (22)
Second order derivatives are given by:
f4 = (fi+lj - 2 fj tj)+(A)2 (23)
and
f, = (fi,3+l -
2
fi,j + f,,j-l)/(Aqv)
2
(24)
with the cross-term derivative:
f~n -" (fi+lj+ - fi-jj+j - fi+l,-i + fi-1,i-)/4AAq (25)
A.1,2 One-Sided Differencing
The three-point one-sided differencing formulas used to represent the temperature gradient:s on the
boundaries of the domain in the generalized boundary condition formulation are givuu below for
51
each face of the computational mesh. On the ( = 1) face, the normal derivative is approximated
by the forward differencing expression:
(f4)i,j = (-3fjj +
4
fi+l,j - fj+2,j)/A (26)
and on the ( = m) face by the backward differencing expression:
(fO)ij = (3ij -
4
f-i-4- fi-2,j)/A4 (27)
On the (q = 1) face, the normal derivative is approximated by tile forward differencing expression:
(fh)ij = (--
3
fid + 4fhj+j - fjj+
2
)/Aq (28)
and on the (n = n) face by the backward differencing expression:
(fn)i~j = (3fld -
4
fjd-1 - fij-2)/A4 (29)
A and Ail are identically 1 in the computational domain,
A.1.3 Explicit Time Differencing Formula
The explicit finite difference approximation for the time derivative of temperature, is givwn by:
T = (
7
1tA - T"(0'
OT 11 ! T.t' _T~ )/At
(30')
A.2 Boundary-Fitted Coordinate System Transformation Coefficients
The coefficients contained in the transformed governing equation and generalized boundary condi-
tion are listed below, Derivatives on the physical domain coordinates, x and -, with respect, t~o
and q are presented above,
A.2.1 Governing Equation Coefficients
The Ai coetlicients appearing in the transformed governing equation are given by:
A
1
= (k,,z2 - 2kxjx,,z,, + k,,x,)/J
2
(31)
52
A
2
= (k,,,z4 - 2k,,xtz4 +4 k
1 1
xt)/J
2
(32)
A
3
= (-2k,,.,z~zC - 2k,,,x~,'x + 2kx,~(x~z,? + x,?z()))/J
2
(33)
A
4
= (k..C' + k.,.,C
2
+ 2k.,,Ca3)/J" (.31)
A,= (kv.C
4
+ k..C
5
j + 2k~,.C
6
)/J
2
(35)
with the coifflcients (2j giveni by:
C
1
+ ~ (x
4
z,,J,, - z'nJil)/J (3(3)
= xn~ +(zx~~ zj~)/j
(:37)
C3 -~t Xc + (Xqz"J - xnn/ (38)
=4 qzz - Zz,
1
z + (Z(ZOJ
4
.-tj/ (:39)
CS =~~t - Z
11
Xcc + (X(Xl
1
Jj -l,?/
('10)
Q n= - xv + (x~ztJt? - z-,z
4
J4)/J (41)
The corresponiding Jacobiaris appearhig iii the Ci expressionis are giveii by:
J( X~Zr + (ZO
4
X, -t - (4:3)
A .2.2 Boundary Conditions Cooffkicentsg
"Thle coefflcietits used ini the iiorznal directiuin torrpertiture graidictits ire giveit by:
a = X + Y (45)
13 = X~ + YT lit6
xf 4, (el7)
INTENTIONALLY LF.I' BLANK.
54
No of Ne of
Cgpka Organiztin Cgia~ Qrgnnization
Office of the Secretary of Defense 1 Director
OUSD(A) US Army Aviation Research
Director, Live Fire Testing and Technology Activity
ATTN: James F. O'Bryon AMTN: SAVRT-R (Library)
Washington, DC 20301-3110 M/S 219-3
Ames Research Center
2 Administrator Moffett Field, CA 94035-1000
Defense Technical Info Center
ATTN: DTIC-DDA 1 Commander
Cameron Station US Army Missile Command
Alexandria, VA 22304-6145 AITN: AMSMI-RD-CS-R (DOC)
Redstone Arsenal, AL 35898-5010
1 HQDA (SARD-TR)
WASH DC 20310-0001 1 Commander
US Army Tank-Automotive Command
Commander ATrN: AMSTA-TSL (Technical Library)
US Army Materiel Command Warren, MI 48397-5000
ATTN: AMCDRA-ST
5001 Eisenhower Avenue I Director
Alexandria, VA 22333-0001 US Army TRADOC Analysis Command
ATTN: ATAA-SI.
Commander White Sands Missile Range, NM 88002-5502
US Army Laboratory Command
ATTN: AMSLC-DL (Chi& only) 1 Commandant
Adelphi, MD 20783-1145 US Army Infantry School
ATTN: ATSH-CD (Security Mgr,)
2 Commander Fort Bonning, GA 31905-5660
US Army, ARDEC
ATI'N: SMCAR-IMI-I (UIKI, ony) 1 Commandant
Picatinny Arsenal, NJ 07806-5000 US Army Infantry School
AMTN:
ATSH-CD-CSO-OR
2 Commander Fort Bonning, GA 31905-5660
US Army, ARDEC
ATTN: SMCAR-TDC 1 Air Force Armament Laboratory
Picatinny Arsenal, N.J 07806-5000 ATTN: AFATL/DLODL
Eglin AFB, FL 32542-50(X)
Director
Benet Weapons Laboratory Aberdeen Proving Ground
US Army, ARDEC
ATTN: SMCAR-CCB-TL 2 Dir, USAMSAA
Watervliet, NY 12189-4050 ATTN: AMXSY-D
AMXSY-MP, H, Cohen
Commander I Cdr, USATECOM
US Army Armament, Munitions ATTN: AMSTE-TD
and Chemical Command 3 Cdr, CRDEC, AMCCOM
ATTN: SMCAR-ESP-L ATTN: SMCCR-RSP-A
Rock Island, IL 61299-5000 SMCCR.MIU
SMCCR-MSI
Commander 1 Dir, VLAMO
US Army Aviation Systems Command ATTN: AMSLC-VL-D
ATIN: AMSAV-DACL
4300 Goodfellow Blvd.
St. Louis, MO 63120-1798
55
No. of No. of
uim oraiznain Q~ Oganization
Ray Garvey I University of Utah
Applied Technology Division ATN: Professor S. Swanson
Oak Ridge National Laboratory Salt Lake City, UT 84112
PO. Box 2003, NS-7294
Oak Ridge, TN 37831-7294 1 Virginia Polytechnic Institute & State
University
3 DTRC AITN: Professor M. Hlyer
ATTN: R. Rockwell Blacksburg, VA 24061.0214
W. Phyillaler
J. Corrado 1 The Pennsylvania State Universtly
Bethesda, MD 20084-5000 A'TTN: Professor T. Hahn
227 Hammond Bldg.
Naval Research Laboratory University Park, PA 16802
ATTN: 1. Wolock
4555 Overlook Avenue, Southwest 2 University of Illinois
Washington, DC 20375-5000 ATTN: Professor S.S, Wang
104 S. Wright Street
2 Director Urbana, IL 61801
Bonet Weapons Laboratory
US Army ARDEC I Dr. Steve Tsal
AMFN: L. Johnson Stanford University
0. Do Andrea Palo Alto, CA 94301
Watervllet, NY 12189-4050
-4 John W. Gillespie, Jr.
6 Lawrence Livenrmore National Laboratory Center of Composite Materials
ATTN: R.M, Christensen Department o' Mechanical Engineering
L. Chino University of Delaware
S. Do Teresa Newark, DE 19716
F. Magness
W. Feng 2 Wright Research & Dvelopment Center
J. Uapper ATTN: J. Whitney
P.O. Box 808 F. Abrams
Livermore, CA 94550 Dayton, OH 45433
5 Sandia National Laboratory, Livermore 2 Battelle PNL
Applied Mechanics Department ATTN: M. Smith
Engineer Design Division P.O. Box 999
ATrN: C.W, Robinson Richland, WA 99352
B. Benedetti
W. Kawahara 2 Olin Corlporation
P. Neilman ATTN: HI, Parkinson
D, Barnmman D. Marlow
Livermore, CA 94550 707 Berkshire Street
East Alton, IL 62024-1174
5 Army Materials Technology Laboratory
AIN: SLCMT-MEC, 3 Honeywell. Inc.
B. Halpin (2) A'FN: ,G. Campbell
W Haskell G. Stelmasic
D. Granville 1. Bode
S. Walsh 5640 SmiLtana Drive
Watetown, MA 02172-(X)01 Minnetonka, MN 55343
56
No. of
C2nia Organiatz~io
2 FMC Corporation
ATTN: Eric Weerth
Manager, Advance Structures and
Materials Dept.
1105 Coleman Avenue, Box 1201
San Jose, CA 95108
2 ARDEC CCAC
ATrIN: S. Mursalli
J. Hederich
Picatinny Arsenal, NJ 07806-5000
2 PM-TMAS
Picatinny Arsenal, NJ 07806-5000
2 DARPA
ATTN: J. Kelly
1400 Wilson Blvd
Arlington, VA 22209
57
INTE~fnoNALLY Lyyr BLANK.
58
USER EVALUATION SHEET/CIRANGE OF ADDRESS
This Laboratory und;lakcs a condnuing effort to improve the quality of the reports it publishes.
Your comments/answers to the items/question5 below will, aid us in our efforts.
1. BRL Report Number BRL-TR-3121 Date of Report July 1990
2. Date Report Received
3, Does this report satisfy a need? (Comment on purpose, related project, or other area of interest
for which the report will be used.)
4. Specifically, how is the report being used? (Information source, design data, procedure, source
of ideas, etc.)
5.' ,Has tl* information in this report led to any quantitative savings as far as man-hours or dollars
save$, opl ating costs avoided, or efficiencies achieved, ctc? If so, please elaborate,
"6. 'O" l~'l Comments. Whatfdo'you &ink should be changed to improve future reports? (Indicate
"chMgefttb organization, technical content, format, etc,)
Name
CURRENT Organization
ADDRESS
Address
City, State, Zip Code
7. If indicating a Change of Address or Address Correction, please provide the New or Cor(ralt
Address in Block 6 above and the Old or Incornrct address below.
Name
OLD Organization
ADDRESS
Address
City, State, Zip Code
(Remove this shect, iold as indicated, staple or tape closed, and mail,)
...........................................................
FOLD HERE --------------------------------------------------------------
DEPARTMENT OF THE ARMY
Director
No WAGE
U.S. Army Ballistic Reearch Latibriiory
SISAR
A'I'TN: SLCBR.DD.r
pN ku-
Abordeen Proving Ground, MD 21ut .5066
m T T
OFFICIAL BUSINESS
.,fK TAE
IBUSINESS REPLY MAIL
FIRST CLASS PERMIT No 0001, APG, MD I_ II_--
POSTAGE WILL BE PAID BY ADESSEE
Director
U.S. Army Ballistic Research Laboratory
ATTN: SLCBR-DD-T
Aberdccn Proving Ground, MD 21005-9989
-------.............................................. ------ FOLD lIERL- .................................................................