Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
NASA Contractor Report 177958
NASA-CR-177958 19850026930
CHEMOVISCOSITY MODELING FOR THERMOSETTING RESINS - II
T. H. Hou
PRC Kentron, Inc.
Aerospace Technologies Division Hampton, Virginia 23666
Contract NASl-18000
August 1985
NI\SI\ Natronal AeronautiCS and Space Adrrun,stratlon
Langley Research Center Hampton, VirglnlB 23665
111111111111111111111111111111111111111111111 NF00712
: '
.f , . .' )
~ ..... , , • , • 11 , J ,- i
L : T j 1985
I ) _.., [ ,..... I I ::: _ - L to(
I=~ -{ )r,IJA:;A
• I-::;'N Vlt/GIN',\
https://ntrs.nasa.gov/search.jsp?R=19850026930 2020-02-22T20:22:25+00:00Z
I 2 Government Acceulon No 1 Report No NASA CR-177958
ChemoV1scoS1ty Model1ng for Thermosett1ng ReS1ns - II
5 Report Oate August 1985
6 Performing Orglnlzltlon Code
7 Author(s) 8 Performing OrglnlZltlon Report No
r..-_T_._H_._H_O_u ______________________ """""I 10 Work Unit No 9 Performlnll Organization Name and Address
PRC Kentron, Inc. Aerospace Technolog1es D1v1s1on 3221 N. Armlstead Avenue
11 Contract or Grant No
NASl-18000 t-_H_a_m_p_t_o_n_,_VA __ 2_3_6_6_6 __________________ "'"""i 13 Type of Repon and Period Covered
Contractor Report 12 Sponsorlnll Agency Name and Address
Natlonal Aeronautlcs and Space Adm1n1strat1on Washlngton. DC 20546
15 Supplementary Notes Langley Technical Monltor: R. M. Baucom F1nal Report
16 Abstract
14 Sponsoring Agency Code 505-33-33-01
A new analyt1cal model for simulat1ng chemov1scosity of thermosett1ng reS1n has been formulated. The model 1S developed by mod1fying the well-established W1lliams-Landel-Ferry (WLF) theory 1n polymer rheology for thermoplastlc materlals. By assumlng a llnear relationship between the glass transltlon temperature Tg{t) and the degree of cure a{t) of the resin system under cure, the WLF theory can be mod1f1ed to account for the factor of reactlon t1me. Temperature-dependent functlons of the mod1fled WLF theory constants C1{T) and C2 (T) were determlned from the lsothermal cure data of Lee, Loos, and Sprlnger for the Hercules 3501-6 reSln system. Theoretlcal predlct10ns of the model for the resin under dynam1c heat1ng cure cycles were shown to compare favorably w1th the exper1mental data reported by Carpenter. Th1S work represents progress toward establishlng a chemov1scos1ty model Wh1Ch 1S capable of not only descr1bing V1SCOSlty profiles accurately under varlOUS cure cycles, but also correlat1ng V1SCOSlty data to the changes of phys1cal propert1es assoc1at~d w1th the structural transformations of the thermosett1ng reSln systems dur1ng cure.
17 Key Words (Suggested by Author(sl) Thermoset, Chemoviscos1ty, Autoclave, WLF Theory, Cure
18 Distribution Statement UnclaSSlf1ed - Unlimlted
19 Security Olulf (of thIS report) UnClaSSlfled
20 Security CIIUlf (of thIS Plge) UnClaSS1fled
Subject Category 24
21 No of PIgIS
37 22 Price
A03
N-305 For sale by the National Technical Information SerVice, Springfield, Virginia 22161
Chemoviscosity Modeling for Thermosetting Resins - II
ABSTRACT
FOREWORD
I. INTRODUCTION
LIST OF SYMBOLS
II. THEORY
Ill. EXPERIMENTAL DATA
Degree of Cure a(t,T)
ChemoVlscosity n(t,T)
TABLE OF CONTENTS
. . . . . . . . . . . . . . . .
· . . . . . . . . . . . . . . · . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . · . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
IV. MODEL PREDICTIONS AND DISCUSSION . . . . . . . . . Isothermal Case
Dynamlc Heatlng Case
V. CONCLUSIONS
REFERENCES
. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . APPENDIX - GENERALIZED NEWTON'S METHOD ••
LIST OF FIGURES . • • • • • • • • • •••••••
LIST OF COMPOSITES PROCESSING LABORATORY RESEARCH REPORTS •
PAGE
iii
v
1
3
4
7
7
9
11
11
13
18
19
20
21
22
This Page Intentionally Left Blank
ABSTRACT
A new analytical model for simulating chemoviscosity of thermosetting
reS1n has been formulated. The model is developed by modifying the
we1l-establ1shed W1ll1ams-Landel-Ferry (WLF) theory in polymer rheology for
thermoplastic mater1als. By assum1ng a linear relat10nship between the glass
transition temperature Tg(t) and the degree of cure a(t) of the resin system
under cure, the WLF theory can be modified to account for the factor of
react10n time. Temperature-dependent functions of the modified WLF theory
constants C1(T) and C2(T) were determ1ned from the lsothermal cure data of
Lee, Loos, and Spr1nger for the Hercules 3501-6 reS1n system. Theoretical
pred1ctions of the model for the resin under dynam1c heating cure cycles were
shown to compare favorably with the experimental data reported by Carpenter.
Th1S work represents progress toward establ1shing a chemoviscosity model which
lS capable of not only descr1b1ng v1scosity prof1les accurately under various
cure cycles, but also correlat1ng V1SCOS1ty data to the changes of physical
propert1es assoc1ated w1th the structural transformations of the thermosetting
reS1n systems during cure.
iii
This Page Intentionally Left Blank
FOREWORD
This work, performed at NASA - Langley Research Center (LaRC), was
supported by the LaRC Polymeric Materials Branch (PMB) under Contract
NASl-l8000. ThlS is a progress report on an ongoing research project toward
the chemoviscosity modeling for thermosetting resins, and ;s monltored by
R. M. Baucom (PMB). Technlcal guidance provlded by Dr. N. T. Wakelyn (PMB)
and W. T. Freeman (PMB) with regard to the uses of LaRC IAF computing
faCllltles and helpful discussions with Or. J. A. Hlnkley (PMB) during the
course of thlS study are gratefully acknowledged.
v
CHEMOVISCOSITY MODELING FOR THERMOSETTING RESINS - II
I. INTRODUCTION
There are two dlfferent basic approaches in chemoviscosity modeling of
thermosettlng reSln. One approach, tYPlcally represented by the work of
Roller [lJ, lS to emplrically formulate a rate equation which relates the
change of chemovlscosity with reactlon tlme. For a reSln system cured under a
dynamlc heatlng cure cycle (as commonly encountered in autoclave processlng
for composlte materlals), such an approach usually Ylelds a model Wh1Ch 1S
1nadequate in descrlblng accurately the nonlinearity of chemoviscoslty as a
functlon of reactlon time. As TaJlma and Crozler [2J pointed out, such a
modellng approach renders ltself to the llmitatlon of batch-spec1fic. The
model parameters cannot be readily related to the chemical and rheologlcal
propertles of the reactlng system as well.
The second approach is based upon a modlflcation to the well-establlshed
V1SCoslty-temperature relationship eXlstlng ln polymer rheology for
thermoplastlc materlals. The parameters in such an equation can be expressed
in terms of polymerlzation k1netlcs, and the chemoviscosity profiles as a
function of reactlon tlme can then be modeled for thermosetting reSln systems.
The appllcablllty of the modlfied Wllllams-Landel-Ferry (WLF) theory [3]
ln chemovlscoslty modellng for thermosettlng reSln has been studled by TaJima
and Crozler [2,4J, Aplcella et al. [5J, and Hou [6J among others. It has been
extenslvely documented ln the llterature that temperature-dependent ViscoSlty
of thermoplastic material can be accurately descrlbed by the WLF equatlons
w1th1n 1000 K above its glass trans1tion temperature (Tg). For a
thermosetting resin system under cure, 1t is reasonable to assume that Tg(t)
of the material is always within 1000 K lower than the cure temperature. The
WLF equat10n should then be appl1cable to every instant state of the resin
system under cur1ng. The central theme in mod1fying the WLF theory to
descr1be the chemov1~cosity of thermosettlng reS1n 1S to take 1nto account the
react10n t1me factor.
In an earl1er paper [6J, two assumptions were proposed, namely, that the
rate constant of reaction, kT' at any temperature T, 1S (i) diffusion
controlled and 1S, therefore, inversely proportional to the viscosity
nT(t) of the reactlve med1um, and (11) d1rectly proport1onal to the rate
of change of glass transition temperature Tg(t). The modified WLF equat10n
becomes a f1rst order ord1nary nonl1near differential equation. Numer1cal
solutions have also been shown to compare favorably w1th the experimental
results for several thermosett1ng systems under 1sothermal and dynamic heat1ng
cure condit10ns. It has been concluded that the flexibil1ty demonstrated by
the mod1f1ed WLF theory [6J can be conveniently explo1ted to establ1sh an
analyt1cal model with h1gh degree of accuracy for the chemoviscosity of any
thermosett1ng resin system under var10US cure cycles. The phys1cal
slgn1f1cances of the mater1al parameters selected for the model were, however,
d1ff1cult to extract for the particular resin system under investigat10n.
As a next step toward the 1mprovement of the chemoviscosity model1ng for
a thermosetting resin system, a new mod1ficat1on of the WLF theory 1S proposed
1n th1S report. Comparisons between the theoretical predictions and
experimental data for chemov1scos1ty of a reS1n system (Hercules 3501-6) under
both lsothermal and dynam1c heating cure cond1t1ons will be d1scussed.
2
li st of Symbols
a,b,c,d
B
DSC
R
T
t
t.t
WLF
nT(t)
a(t)
Materlal parameters (min-I) speclfied in Table 1
Constants used in Eq. (5)
Materlal parameter specified ln Table 1
Temperature-dependent WLF constants defined by Eqs. (7) and
( 8)
Differentlal Scanning Calorimeter
Actlvation energles (J/mol) speclfied in Table 1
Materlal constants (min-I) deflned ln Eqs. (3) and (4)
Universal gas constant (=1.987 Cal/gm-mole/oK)
Temperature (OK)
Cure temperature (OK), Tc = Tg{m)
Glass transition temperature of the reSln system
under cure (OK)
Glass transition temperature of the resins system at
completion of cure (OK)
Glass transltion temperature of the uncured resin g
Cure time (minutes)
Tlme (mlnutes) deflned in Eq. (6)
Tlme required for completion of cure (minutes)
Time interval (minutes) used ln solving Eq. (9)
Wllliams-Landel-Ferry
Chemoviscosity at temperature T (poises)
Degree of cure of the thermosetting resin durlng cure
as determlned by DSC thermal analysis
3
II. THEORY
The W1lliam-Landel-Ferry (WLF) theory [3] is given by
(1)
C1 and C2 are two mater1al constants, and nT represents v1scosity at
temperature T of the glven thermoplastic mater1al Wh1Ch possesses a glass
trans1tion temperature Tg• The normal use of the WLF equat10n for
thermoplastic materials requ1res that Tg be constant wh1le the temperature T
1S var1ed for the specif1c polymer under study. However, during cure of
thermosett1ng resins, the monomers are initially polymerized and crosslinks
are formed later. Th1S is a system where Tg{t) 1S changing and the curing
temperature T 1S held constant (In an isothermal cure case, for example). The
glass trans1tion temperature Tg r1ses cont1nuously and may eventually
approach the curing temperature. Over the entire curing cycle, the material
structure actually undergoes cont1nuous phase transformations from a low
molecular weight llQU1d to a h1gh molecular we1ght polymer1c melt, and
eventually transforms to a crossl1nked network. It 1S reasonable to assume
that Tg{t) of the mater1al 1S always lower than the cure temperature T, and
that (T-Tg{t)) 1S always w1th1n lOOoK. The WLF theory {Eq. (I)) should then
4
be applicable to all of the different polymeric structure phases during cure.
Before Eq. (l) can be applied to describe the chemoviscosity of the curing
resin, however, modlfications need to be made in order that reaction time, t,
of the thermosetting resin system lS properly taken into account. This can be
accompllshed as follows:
A simple assumption that Tg{t) lS a linear functlon of the degree of
cure a{t) of the thermosetting reSln can be made, WhlCh gives:
Tg{t) = TO + (T _TO) a(t) 9 c g
(2)
where T~ = Tg(O) lS the glass transitlon temperature of the unreacted reSln
system, Tc lS the cure temperature ln OK, and t is the reaction time in
minutes. Equation (2) should satisfy the boundary conditions, namely,
T = TO a = 0 at t = 0 g g'
and
where tf lS the tlme required for completlon of cure. The degree of cure
a(t) for the resin system can be determlned by means of thermal analysis. It
is assumed that the degree of cure at tlme t lS equal to the fractlon of heat
released as measured by a OSC up to time t for the thermosetting materlal
under cure. The total heat of reaction can be determined for the experlments
uSlng dynamlc temperature scanning. The degree of cure a(t) can then be
5
determ1ned for dlfferent isothermal cure cond1tions [7,8J. The a's thus
determ1ned are funct10ns of curlng temperature and tlme.
Equations (1) and (2) const1tute a mod1fled WLF theory for simulat1ng
chemoviscosity of thermosett1ng reS1n. Among the f1ve mater1al parameters, C1
and C2 can be determined as a funct10n of temperature T by fitting w1th
lsothermal cure n(a) data, nT = 10 13 poises lS given by Gordon and 9
Slmpson [9J for many polymer1c mater1als at T = Tg, and T~ can be determlned
by DSC thermal analys1s. In an 1sothermal cure case, Tc of Eq. (2) 1S equal
to cure temperature. In the dynam1c heating case, however, Tc lS d1ff1cult
to spec1fy exactly and 1S the only adJustable var1able 1n the present mod1fied
WLF theory.
6
III. EXPERIMENTAL DATA
For the purpose of evaluating model pred1ctions of this modified WLF
theory, experimental data on Hercules 3501-6 epoxy reS1n reported in the
llterature by Lee, Laos, and Springer [8] and Carpenter [10] were selected.
These data will be summarlzed br1efly.
Degree of Cure a{t,T)
Lee et al. [8] performed thermal analyses on the epoxy resin system using
a Different1al Scanning Calorimeter (DSC). The degree of cure a and the rate
of degree of cure da/dt were determined from the results of the isothermal
scanning experiments. Eight d1fferent temperatures were selected, and the
results are reproduced 1n Figure 1. It is noted from the f1gure that both
h1gher values of the degree of cure and rate of degree of cure can be obtained
at higher cure temperatures. The rate of degree of cure 1S a decreasing
funct10n of cure t1me t, and eventually levels off at a low rate value (< 0.1)
when a h1gh degree of cure 1S reached. The follow1ng equations were found to
describe these temperature-dependent data accurately:
for a < 0.3 (3)
da df = K3 (l-a) for a > 0.3 (4)
where
7
KI = Al exp (-l1E I/RT)
K2 = A2 exp (-l1E 2/RT)
K3 = A3 exp (-l1E 3/RT)
Values of constants in Eqs. (3) and (4) are shown in the Table below.
Comparisons between the measured values of da/dt and the values calculated by
Eqs. (3) and (4) are also shown 1n F1gure 1.
Table 1. The values of the constants of Eqs. (3) and (4) for Hercules 3501-6 Resin System [8J
B = 0.47
Al = 2.101 X 10 9 m1n- 1
A2 = -2.014 X 10 9 min- I
A3 = 1.960 X 10 5 m1n- 1
l1E 1 = 8.07 X 104 J/mol
l1E 2 = 7.78 X 104 J/mol
l1E 3 = 5.66 X 104 J/mol
Eqs. (3) and (4) can be integrated prov1ded that the temperature is kept
constant. The following relat10nships between a and t are obtained:
a K2 t = K; Ln (1 + K7 a) - b Ln (I-a) - c Ln (1- f) (5)
t = 1 Ln (I-a) + t -K; 0.7 c (6)
8
where
a = K2 (B-l)/d 2
b = - (K 1+K 2)/d
c = (K 1+K 2B)/d
d = (K 1+K 2) [K 1B2+(K 1-K 2) B-K 1 ]
and tc is the t1me given by Equat10n (5) for a = 0.3.
ChemOV1scoS1ty n(t,T)
ChemOV1scoS1ty of the reS1n under an lsothermal cure cond1t1on were
measured by Lee et. al. [8] on a Rheometr1cs parallel d1SC and plate
mechan1cal spectrometer Model 605. The rad1us of the disc was 25 mm and the
gap between the d1SC and the plate was 0.5 mm. A frequency of 1.6 Hz was
employed. For a glven cure temperature, the n(t) measured can be related to
a(t) obtained from DSC thermal analys1s. The results are plotted 1n F1gure 2
for three d1fferent lsothermal cure cases. It lS again noted that at a h1gher
cure temperature, h1gher values of the degree of cure can be reached. For a
glven a, the decrease 1n V1SCOS1ty 1S attributed to the temperature effects.
ChemOV1scos1t1es of the 3501-6 epoxy resin cured under dynam1c heat1ng
cond1t1ons were reported by Carpenter [10]. Carpenter measured chemov1sco
slt1es at d1fferent rate of 1ncrease 1n temperatures and reported the measured
values as a funct10n of cure t1me. The experimental data shown here are for
reS1ns cured at constant heat1ng rates of 10 K/min (f1gure 5), 2.17 and 5°K/min
(F1gure 6) and a cure cycle (F1gure 8) slm1lar to the one typ1cally found 1n
autoclave processes for compos1te materials.
9
The procedure used for model evaluations will be as follows:
1. Model parameters of Eqs. (1) and (2) will be determlned first by the
experimental data of Lee et al. [8] under isothermal cure conditlons.
Temperature-dependent materlal constants C1(T) and C2 (T) can thus be
determined.
2. The chemoviscoslty model n(t,T) establlshed by isothermal cure data under
procedure 1 wlll then be made to follow varlOUS dynamic heating cure
cycles T(t). The viscosity values thus calculated will be compared with
the experimental data reported by Carpenter [10].
10
IV. MODEL PREDICTIONS AND DISCUSSION
Isothermal case
Gordon and Simpson [9] pointed out that glass-forming substances
generally have the same shear viscosity of nT = 10 13 poises at their g
Tg, and this has also been conf1rmed for polymer systems. By specifying TO
T8 = 3000 K (27°C) 1n Eq. (2), the model (l.e., Eqs. (l) and (2)) can then be
f1tted to the experimental data n(a) as shown in Figure 2 provided that proper
values of material constants C1 and C2 are selected. C1 and C2 determined for
different temperatures are tabulated below, and the model predictions are
drawn as solid llnes 1n the figure.
Table 2. Values of material constants C1 and C2 1n the modi fled WLF Theory for Hercules 3501-6 reS1n system as determined by the isothermal cure data.
Tc(OK) C1 C2(OK)
375 21.1 53.0
400 17.1 33.6
425 18.5 46.0
The model is capable of descr1bing the nonl1near relationship between
V1SCOS1ty, n, and the degree of cure, a, as shown in Figure 2, especially for
11
the case of h1gher cure temperature. Th1S 1S cons1dered a slgn1f1cant
improvement over the emp1r1cal 11near model used by Lee, Loos, and Spr1nger
[8]. An arrhen1us type of plot 1S arb1trar11y chosen for C1 and C2 as shown
in F1gure 3. The data 1S rather scattered. Due to the (unfortunate)
11m1tat1on on size of the exper1mental data and the w1de range 1n temperatures
covered by the dynam1c heat1ng cure exper1ments (as w11l be d1scussed later),
a 11near least squares flt was chosen. The straight lines are represented,
respectively, by:
log C1 = 186.02 (liT) + 0.8084 (7)
and
log C2 = 212.16 (liT) + 1.106 (8)
It 1S noted that 1n the analysis descrlbed above, the glass trans1tlon
temperatures Tg(t) for the varYlng molecular structures of the reSln cured
at dlfferent tempreatures are chosen as reference temperatures. C1 and C2 are
therefore treated as "unlVersal constants" ln the present context. Val ues of
C1 and C2 reported ln Table 2 are close to the values of 17.44 and 51.6,
respect1vely, once reported by Will1ams et ale [3] as the unlversal constants
for all polymer1c materials sytems. Eqs. (7) and (8) wlll still be used,
however, to account for temperature-dependences of C1 and C2 1n chemovlscoslty
slmulatl0n under dynam1c heat1ng cure cond1tlons.
The theoretical assumptlons as descrlbed above, WhlCh Yleld close values
for the WLF constants C1 and C2 , respectlvely, among three d1fferent
12
1sothermal cur1ng temperatures, are also supported by the parallellsm among
experimental data observed 1n Figure 2 for a < 0.3. Data of higher values of
both n and a are d1fficult to be measured, and could be considered less
accurate.
Dynam1c Heat1ng Case
In the case of dynamic heating cure conditions, 1ntegrations of Eqs. (3)
and (4) are very difficult to perform. If one assumes that within any t1me
1nterval lI,t = t2 - t l' the temperature Tc 1S uniform, and 1 s equal to
(Tl+T 2)/2, where Tl and T 2 are temperatures at t 1 and t2 respectively as
def1 ned by the cure cycle, Eqs. (3) and (4) become:
lI,t = t 2 - tl
(9)
where
a K2 f (a,Tc ) = K2 Ln (1 ~l a) - b Ln{1 - a) - c Ln (1 - f), for a < 0.3
and
f (a, Tc) = 1 Ln (1 - a) + t - K3 0.7 c for a > 0.3
13
The parameters a, b, c, and tc are the same as defined before in Eqs. (S)
and (6). The value of the increased degree of cure a2 1n Eq. (9) can be
calculated by the Generalized Newton's Method. Deta1ls of the solution
schemes are summarized 1n the Appendix.
Values of a(t} determined by the scheme descr1bed above depend heavily on
the t1me 1nterval ~t chosen. Theoret1cally more accurate a(t} can be
determined numerically when a smaller ~t 1S used. On the other hand, a
smaller ~t could increase dramatically the numer1cal round-off errors. A
series of ~t values were selected, and a(t} was calculated from Eq. (9) for a
cure cycle w1th a constant heat1ng rate of SOK/min. The results are plotted
1n F1gure 4. It can be seen that numer1cal round-off errors become dom1nant
when ~t < O.OS m1n. It is also noted that numerical values of a(t} calculated
for ~t = O.S and 1.0 minute, respect1vely, are almost 1ndistingu;shable.
Consequently, ~t = O.S m1nute was chosen for all calculations discussed below.
Chemov1scosities of the Hercules 3S01-6 epoxy resin system under dynamic
heating cure cond1t10ns were reported by Carpenter [10J. Data for these
constant rates of heating experiments are reproduced in F1gures Sand 6. The
heat1ng rates are 1.0, 2.17, and S oK/min respect1vely. All cure cycles
started from 3000 K (27°C). Temperature-dependent C1 (T) and C2(T) as def1ned
by Eqs. (7) and (8) are followed 1n the slmulation. Values of the rest of the
parameters used 1n the model are tabulated below:
14
Table 3. Values of model parameters used 1n simulat1ng chemov1scosity under rate of heating cure condit1ons
Heat1ng rate nT (poises) TO (0 K) T c (0 K) (0 Klm1 n) g g
1 10 13 300 415
2.17 10 13 300 430
5 10 13 300 450
The model predictions are shown as solid curves in F1gures 5 and 6. It is
noted that the theoret1cal calculations compared favorably with the
exper1mental data for all three cases.
Theoret1cally, Tc 1S equal to Tg(~) at T + ~ or a + 1 (see Eq. (2)).
Under the dynam1c heat1ng cure cond1tions, however, Tc 1S d1ff1cult to
spec1fy. Values of Tc (In Table 3) 1n the above calculations were
selected arb1trar1ly such that model pred1ct1ons are 1n reasonably good
agreement w1th the exper1mental data. Nevertheless, h1gher values of Tc
(Table 3) resulting from the slmulat10n of higher rates of heating
experiments, are 1n agreement w1th phys1cal observat1ons for thermosett1ng
res1ns.
The calculated degree of cure a{t) for all three cases 1S plotted 1n
F1gure 7. These calculated a(t) are based on the exper1mental data for the
resln under 1sothermal cure condlt1ons [8J, and are lndependent of the
arbltrarlly selected values of Tc d1scussed above. Values of da/dt go
15
through maXlma for all three cases. The arrow marks shown in the figure
denote the times when chemoviscositles approach inflnity during cure. These
arrow marks are located approxlmately at a(t) = 0.75 ~ 0.05, and are near the
lnflection pOlnts of the curves as well. Physical slgnificances of these
observatlons are unclear at the present time, and further research in the
future lS required.
Chemoviscosity was also measured by Carpenter [10] for the 3501-6 resin
system under a typlcal cure cycle used in autoclave processlng for composite
materials. The experimental data is reproduced ln Figure 8. Values of model
parameters used are the same as those lncluded in Table 3 and Eqs. (7) and
(8), except that Tc = 375°K. The theoretical predictlons, shown as solid
curves, are seen in reasonably good agreement with the experimental data. The
model, however, falls to describe the chemoviscosity "shoot-up" near the end
of the cure cycle. The reasons are unclear at the present time. It could be
due to the inadequate definitlons of C1(T) and C2 (T) by Eqs. (7) and (8); or
due to different reaction mechan1sms appeared 1n the h1gh cur1ng temperature
reg1me. Calculated values of a(t) are plotted in Flgure 9. The arrow mark
shown denotes the time when chemov1scosity "shoot-up" occurs. Features of the
a(t) curve discussed above for isothermal cure cases are again noticeable in
th1S f1gure.
Temperature ranges covered by Carpenter [10] in dynam1c heating cure
experiments, as shown by dotted lines 1n Figure 3, far exceed the ranges
covered by Lee et ale [8] in isothermal cure experiments (shown by solid lines
in the figure). Accurate deflnitions of the crucial temperature-dependences
of C1(T) and C2(T) needed for the model thus suffer from serious data
16
extrapolatlon errors. The results of the present lnvestigations, although
encouraglng, can only demonstrate the feaslbillty of the approach of modlfying
the WLF theory in chemoviscosity modeling for thermosetting resin. Detailed
studies such as the preclse deflnitlons of Tg{a) (Eq. (2», and the
functlons of C1{T) and C2 {T) over the full temperature range by experlmental
measurements for the reSln system under lnvestigation have to be pursued In
the future.
17
v. CONCLUSIONS
A new analytlcal model, based upon modiflcatlons of the WLF theory, has
been formulated. The model assumes that the glass transltlon temperature
Tg{t) of the reSln system under cure is llnearly related to the degree of
cure a{t) determined by DSC thermal analysis. Theoretlcal predlctlons of the
chemovlscoslty compare favorably wlth the experlmental data reported ln the
llterature for both lsothermal and dynamic heatlng cure condltions.
Conslderlng that data are available only ln the relatlvely narrow temperature
range used in lsothermal curing experlments by Lee, Loos, and Sprlnger [8J,
the agreement between theoretlcal predlctions and experlmental data reported
by Carpenter [10J for dynamlc heating cure cycles are partlcularly
encouraglng. Due to the limltations In Slze of the data avallable, the
successful analysls reported here has establlshed the feaslblllty of the
proposed approa~h WhlCh slmulates chemovlscoslty for thermosettlng reSln by
modlfYlng the wldely used WLF theory In polymer rheology.
18
REFERENCES
1. Roller, M. B.: "Characterization of the Time-Temperature-Viscosity
Behav10r of Cur1ng B-Staged Epoxy Resin," Polym. Engr. Sci.,.!,?, 406,
1975.
2. Taj1ma, Y. A.; and Croz1er, 0.: "Chemorheology of an Amine Cured Epoxy
Res1n," SPE ANTEC Conference Proceedings, 274, 1984.
3. Williams, M. L.; Landel, R. F.; and Ferry, J.D.: "The Temperature
Dependence of Relaxat10n Mechan1sms in Amorphous Polymers and Other
Glass-Forming Liquids," J. Amer. Chern. Soc., 72, 3701, 1955.
4. TaJima, Y. A.; and Crozier, 0.: "Thermokinetic Modeling of an Epoxy Resin
1. Chemov1scos1ty," Polym. Engr. Sci., Q, 186, 1983.
5. Apicella, A.; N1cola1s, L.; and Halp1n, J. C.: "The Role of the
Process1ng Chemo-Rheology on the Ag1ng Behav10r of H1gh Performance Epoxy
Mat ri ces," 28th Nat i ona 1 SAMPE Sympos i um, 518-527, 1983.
6. Hou, T. H.: "Chemov1scos1ty Model1ng for Thermosetting Res1ns - I,"
NASA-Langley Research Center, CR-172443, November, 1984.
7. Sourour, S.; and Kamel, M. R.: "01 fferenti al Scanning Calorimetry of
Epoxy Cure: Isothermal Cure K1netics," Thermochimica Acta, .!.i, 41-59,
1976.
8. Lee, W. I.; Loos, A. C.; and Springer, G. S.: "Heat of React1on, Degree
of Cure, and V1SCOS1ty of Hercules 3501-6 Resin," J. Compo. Matl., 16,
510-520, 1982.
9. Gordon, M.; and Slmpson, W.: "The Rate of Cure of Network Polymers and
the Superposition Pr1nc1ple," Polymer, ~, 383-391, 1961.
10. Carpenter, J. F.: "Process1ng Science for AS/3501-6 Carbon/Epoxy
Compos1tes," Technl cal Report NOOOI9-81-C-0184, McDonnell Ai rcraft Co.,
1983. 19
•
Appendix - General1zed Newton's Method 1n Solving a2 from Eq. (9)
Eq. (9) can be rearranged to become:
(a)
and
(b)
where g'(a2) = dg/da2, and f'(a2,Tc ) = df/da2' The Generalized Newton's
Method sets up the follow1ng recursion formula to solve for a2 1n Eq. (a):
(c)
prov1ded that g'(a2)/tO; where w 15 called the over-relaxat1on factor. A good
cho1ce for w can be determ1ned, 1n general, only by experimentation. A cho1ce
of w d1fferent from un1ty often can 1ncrease the convergence rate of the
rout1ne apprec1ably. In this study, W = 0.1 was used for all calculations.
Iterat10n5 are continued unt1l lan+1 - anl~ 10- 6 •
20
li st of F; gures
F1gure 1. Rate of degree of cure vs. degree of cure reported by Lee, Loos, and Springer [8J for Hercules 3501-6 reS1n system under 1sothermal cure cycles. Solid lines were calculated using Eqs. (3) and (4).
F1gure 2. Chemoviscosity as a function of degree of cure for Hercules 3501-6 resin system under isothermal cure cycles. Symbols represent experimental data reported by Lee, Laos, and Springer [8J. Solid lines were calculated by the model, Eqs. (1) and (2).
Figure 3. Temperature-dependences of the WLF constants C1 and C2 determ1ned by the isothermal cure data of Lee et ale [8J for the Hercules 3501-6 resin system.
Figure 4. Simulated degree of cure data for the Hercules 3501-6 resin system cured under constant rate of heating condition, 5°K/min. Various time intervals fit were used 1n solving Eq. (9).
Figure 5. Chemoviscosity of the Hercules 3501-6 resin system cured under constant rate of heating condition of 10 K/min. Circles denote experimental data by Carpenter [9J. Solid curve denotes the model predictions of Eqs. (1) and (2).
F1gure 6. Chemov1scos1ty of the Hercules 3501-6 resin system cured under constant rates of heating condit1ons of 2.17 and 5°K/min, respectively. Circles denote the experimental data by Carpenter [9J. Solid curves denote the model pred1ctions of Eqs. (1) and (2) •
F1gure 7. Simulated degree of cure for the Hercules 3501-6 reS1n system cured under constant rate of heat1ng conditions 1ndicated.
Figure 8. Chemov1scos1ty of the Hercules 3501-6 resin system cured under the cure cycle shown. Circles denote the experimental data by Carpenter [9J. SOlld curve was calculated by the model, Eqs. (1) and (2).
Figure 9. Simulated degree of cure data for the Hercules 3501-6 reS1n system cured under the cure cycle included.
21
Composite Processing laboratory Research Reports
Polymerlc Materlals Branch, Langley Research Center Natlonal Aeronautics and Space Admlnlstratlon
CPl1. "A Theoretical Study of Resln Flows for Thermosetting Materials
Durlng Prepreg Processing," NASA CR-172442, July, 1984, by T. H. Hou
(Kentron).
CPl2. "Chemovlscoslty Modeling for Thermosettlng Reslns - I," NASA
CR-172443, November, 1984, by T. H. Hou (Kentron).
CPl3. "Correlation of Dynamic Dlelectrlc Measurements with V1SCOSity 1n
Polymerlc Res1n Systems," 30th National SAMPE Sympos1um and
Exh1b1t1on, Anaheim, CA., March, 1985, pp 638-648, by D. Kranbuehl,
S. Delos, E. Y1, J. Mayer (W&M), T. H. Hou (Kentron), and W. P.
Wl nfree (LaRC).
CPl4. "On the Appllcations of WLF Theory to Thermosett1ng Resins," SPE
ANTEC, Washlngton, D.C., Aprll, 1985, pp 1253-1255, by T. H. Hou
(Kentron).
CPl5. "Chemov1scos1ty Model1ng for Thermosetting Reslns - II," NASA
CR-177958, August, 1985, by T. H. Hou (PRC Kentron, Inc.)
22
-'C ...... tS 'C .. W 0: ::> u lI-0 W W 0: (!) W 0 lI-0
0.1 400K o..5~450K
I o Data - Calculated
o oh:am.1
425K 465K
430K 470K
440K 475K
o O~LJ~~~~ OL--....L-~----='O.'-:'4---0:;::"'.6 0 0.2 0.4 0.6 DEGREE OF CURE.a
Flgure 1. Rate of degree of cure vs. degree of cure reported by Lee, Loos, and Springer [8J for Hercules 3,01-6 reSln system under lsothermal cure cycles. SOlld llnes were calculaterl using Eqs. (3) and (4).
23
uP
104 VI Q) VI ·0 c.. :;: c:: Q) u
c
103
006 Hercules 3501-6 resin (Lee, Loos & Springer)
- Model prediction
a, degree of cure
Flgure 2. Chemovlscoslty as a functlon of degree of cure for Hercules 3501-6 reSln system under lsothermal cure cycles. Symbols represent experlmental data reported by Lee, Loos, and Sprlnger [8J. Solid llnes were calculated by the model, Eqs. (1) and (2).
24
500 450 400 350 300 1oor------.-I-------.I-----------.-�------------~I-.
r -~~ -
~ ~ --~ --- -O ---50 r _--
~ I- _ - C2 --I.-- 0
-
-
-
20 -
--------0 __ -- -----
_---~1 ,.--
-
10~ _____ ~1 _____ ~1 _____ ~1 _____ ~1 ____ ~1 ______ ~1 ____ ~
lO l2 l4 l6 l8 10 12
1fT, oK x 103
Flgure 3. Temperature-dependences of the WLF constants C1 and C2 determlned by the lsothermal cure data of Lee et al. [8J for the Hercules 3501-6 reSln system.
25
14
a(t)
10°r---~----~----~----~--~----~----~ I T 1 I I I
10 -1 t-
D 0
10-5 I I I I
0 4 8 12 16 t, min
o o o
at, min 0.001 0.005 0.010
[). 0. 050 ~ 0.100 D 0. 500 0
0
<)
I
1000 5.000
10. 000
I
-
-
-
-
20 24 28
Flgure 4. Slmulated degr~e of cure data for the Hercules 3501-6 reSln system cured under constant rate of heatlng condltlon, 5°K/mln. Varlous tlme lntervals fit were used in solvlng Eq. (9).
26
8
7
6
~ 5 I 0 ->< VI Q) 4 VI
'0 Co
= C Q) u 3 c:
2
1
o
Hercu les 3501-6 0
Heating rate: 10K/min 0
0 Experimental data
- Model prediction 0
0
0 0
0
0
0
20 40 60 80 t, min
Flgure 5. Chemovlscoslty of the Hercules 3501-6 reSln system cured under constant rate of heatlng condltion of l°K/mln. Clrcles denote experlmental data by Carpenter [10]. Solid curve denotes the model predictions of Eqs. (1) and (2).
27
160
~ I 0 -)( VI C1) VI ·0 c..
:;::; c C1) u . c
8
Hercules 3501-6
7 0 Experimental data
- Model predictions
6 Heating rate: 2. 170K/min SOK/min
0 0
S
4
3
2
1
o 20 40 60 80 90 0 20 40 t, min t, min
F1gure 6. Chemov1scos1ty of the Hercules 3501-6 resin system cured under constant rates of heating connlt10ns of 2.17 and 5°K/m1n, respectlVely. C1rcles denote the exper1mental data by Carpenter [10]. S011d curves denote the model pred1ct10ns of Eqs. (1) and (2).
28
60
L 00 .-------J'{llr-------::-o-<;>-----~ Hercules 3501-6 0
.75
a(t) .50
.25
o
o 0 o
to .0 o 0
50K/min 0 o o
o o
o o
o o
o
o o
50
o
o o 2. 17oK/mln
o o
o
6 6
6
6666 6
6
6
6
66 LOOK/min 6
100 t, min
Flgure 7. Slmulated degree of cure for the Hercules 3501-6 reSln system cured under constant rate of heatlng condltlons indlcated.
29
..
150
y 0 ...-4
>< VI Q) VI
'0 c..
:;:; c: Q) u
c
T OK ,
8
Hercu les 3501-6 7 0
0 Experimental data
6 0 - Model prediction
5 o
4
3 o
2
1
0
700 600 500 400 300 200
0 20 40 60 80 100 120 140 t, min
Flgure 8. Chemovlscoslty of the Hercules 3501-6 reSln system cured under the cure cycle shown. Clrcles denote the experlmental data by Carpenter [10]. SOlld curve was calculated hy the model. Eqs. (1) and (2).
30
· 75 I-
a(t) .50 I-
.25 I-
o 'CIIDQ"()
Hercules 3501-6
° (")0
I
-
-
-
I
700~-------------~1-------------'-1--------/-"~-' /
600 ~ / -/.L:J2.7SoK/min
/ '----~--------
/// Cure cycle 400 I- _f'
/.LJ2. 7SoK/min 300 ~
500 >- -
--
200L-__________ ~1L_ __________ ~1 ____________ ~
o 50 100 t. min
Flgure q. Slmulated degree of cure data for the Hercules 3501-6 resin system cured under the cure cycle included.
31
150
End of Document