University of Southern California Dissertations and Theses

Sustainable manufacturing of out-of-autoclave (OoA) prepreg: processing challenges

(USC Thesis Other)

Sustainable manufacturing of out-of-autoclave (OoA) prepreg: processing challenges

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Sustainable Manufacturing of Out-of-Autoclave (OoA) Prepreg:
Processing Challenges

By

Daniel June Kim

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Chemical Engineering)

May 2017

Copy Right 2017 Daniel June Kim
ii

Epigraph
“Therefore wisdom must plainly be the most finished of the forms of knowledge. It follows that
the wise man must not only know what follows from the first principles, but must also possess
truth about the first principles. Therefore, wisdom must be intuitive reason combined with
scientific knowledge--scientific knowledge of the highest objects which has received as it were its
proper completion.”

Aristotle, Nicomacheian Ethics

iii

Acknowledgements

It has been quite a journey. I have spent roughly eight years in graduate school
consummating in three Masters and a Ph.D. degrees. First, I would like to thank Professor Lynden
A. Archer for teaching me how to be a graduate student in engineering/science field. He was not
only a great scientist but also a great advisor. Despite holding numerous positions and guiding
tens of Ph.D. students & post-docs under his guidance, he took time to hold individual meetings
with me, a master’s student, to discuss my research progress every week or more. My dream of
getting a doctorate degree started with him and I was ready by the time I moved onto USC.
I would like to thank my thesis advisor, Professor Steven R. Nutt for all his support and
guidance throughout the completion of this work. He has given me ample opportunities to grow as
a researcher. I have learned how to approach a scientific problem and to navigate through the
problem to arrive at a solution that I can be proud of. I feel blessed to find a research area that I
am passionate about and I am confident that with the problem solving skills that I have acquired
under his guidance; I will be able to handle any scientific and work-associated challenges in the
future.
I am also blessed to have worked with Dr. Timotei Centea from the beginning to end of
my doctorate work at M.C. Gill Composite Center. He supported me as a second advisor within
the group. He always challenged me with new perspectives on my research analysis and those
invaluable insights lead my work to the next level. He always took time to discuss even the little
details of my research work with me and I have learned a lot from the discussion. I would also like
to thank all members of M.C. Gill Composite Center who helped me on various aspects of my
graduate school life. I will always cherish my experience here.
iv

I had an opportunity to have extensive internship experience at NASA’s Johnson Space
Center during my Ph.D. studies. There, I was blessed to meet my mentor, Jeremy B. Jacobs, who
was one of the brightest person I have ever worked with. He taught me how to be professional at
workplace and he challenged me on areas that I must improve on. Under his guidance, I was able
to accomplish a project before my departure back to USC to complete my thesis.
Lastly, I would like to thank all my friends and family who were always there for me. I
would not have completed the study without their unceasing support.

v

Table of Contents

Epigraph……………………………………………………………………….……. ………...ii
Acknowledgement…………………………………………………………………. ….…iii
List of Tables…………………………………………………………………………… …..…xi
List of Figures..…………………………………………………………………………… ………xiii
Abstract …………………………………………………...……………………………… …….xix
Chapter 1. Introduction..………………………………………………………………… …..…1
1.1. Motivation…………...……………....…………………………………… …..…1
1.2. Resin Formulation….……………………..…………………………………… …..…6
1.3. OoA Processing (Cure Kinetics & Viscosity) Models…………….…….. ………10
1.4. Cure Monitoring Techniques...…………………………………………………. ………14
1.5. Voids Formation………………………………………………….………… …….…17
1.6. Resin Flow………………….……………………………………….……… …..…18
1.7. Resin Recycling….……………………..………………………….…………… …...…20
Chapter 2. Experimental Method...………………………...…………………………… …...…23
2.1. Differential Scanning Calorimetry (DSC)..…………………………………… …..…23
2.2. Rheology…………………………………..…………………………………… …..…24
2.3. Dielectric Analysis (DEA)………………..…………………………………… ………26
2.4. Thermogravimetric Analysis (TGA)…….....…………………………………… …..…27
2.5. Coulometric Fischer Titration……………...…………………………………… …..…28

vi

Chapter 3. Out-time Effects on Cure Kinetics and Viscosity for an Out-of-Autoclave
(OoA) Prepreg: Modelling and Monitoring………………………………………. ………30
3.1. Introduction………………………………...…………………………………… …..…30
3.2. Experimental Procedure………………………………………………………… …..…33
3.2.1. Materials…………………………………..……………………………… …...…33
3.2.2. Sample conditioning……………….....…………………………………… …...…33
3.2.3. DSC measurements .......……………...…………………………………… ………34
3.2.4. Rheometry measurements………...………………………………… …...…34
3.2.5. DCM measurements………………..…...…………………………… …….…35
3.3. Model Framework…….....…………………………....………………………… …..…36
3.3.1. Cure kinetics model development using DSC measurement…..…….…… …..…36
3.3.2. Viscosity model development using rheometry measurement…………… …...…38
3.3.3. Cure kinetics and viscosity correlation development using
DCM measurement………………………………………………………… ……...39
3.4. Results and Discussion……….……………….…..…………………………… ………40
3.4.1. Cure kinetics measurement and modelling using DSC…….……...…….… …..…40
3.4.2. Viscosity measurement and modelling using rheometry.…….……..….… …..…44
3.4.3. Cure kinetics and viscosity measurement and correlation using C
DCM………………………………………………………………………… ……...47
3.5. Conclusions………………..……………...…………………………………… ………52

vii

Chapter 4. In-Situ Cure Monitoring of an Out-of-Autoclave Prepreg: Effects of
Out-time on Viscosity, Gelation and Vitrification……………………….……………. ….….53
4.1. Introduction………………………………...…………………………………… …..…53
4.1.1. Background…………………………………………….……….…… …..…53
4.2. Experimental Procedure………………………………………………………… …..…55
4.2.1. Materials…………………………………..……………………………… ………55
4.2.2. Modulated differential scanning calorimetry (MDSC)………………… …..…56
4.2.3. Rheometry………….......……………...………...………………………… …..…56
4.2.4. Dielectric analysis……………………………………………………. …..…57
4.3. Principles of Dielectric Analysis………………..…...………………………… …..…57
4.4. Determination of Physicochemical Parameters……………………………… …..…59
4.4.1. Out-time monitoring………………………..……………………………… …....…59
4.4.2. Gelation point and minimum viscosity……………………….………… …..…59
4.4.3. Vitrification point…………………..……………………….………… …..…60
4.5. Results and Discussion.………………………………………………………… …...…61
4.5.1. Out-time monitoring (ex-situ vs in-situ)…....…………………………… ………61
4.5.2. Gelation point and minimum viscosity (ex-situ vs in-situ)….….………… …..…65
4.5.3. Vitrification point (ex-situ vs in-situ)…..…………….………….………… …..…68
4.5.4. Manufacturing considerations...………..……………………….………… …..…72
4.6. Conclusions…………..………………………………………………………… …..…75

viii

Chapter 5. Modelling and Monitoring of Out-time and Moisture Absorption Effects on
Cure Kinetics and Viscosity for An Out-of-Autoclave (OoA) Prepreg…………… ……78
5.1. Introduction………………………………………………………………… ……78
5.2. Experimental Procedure………………………………………………………… …..….80
5.2.1. Sample conditioning: humidity & out-time control……...……………… ……80
5.2.2. Modulated differential scanning calorimetry (MDSC)..….…………… …....81
5.2.3. Rheometry………………………..………..……………………………… …..….81
5.2.4. Dielectric analysis (DEA)…………………………………………… …....82
5.3. Model Framework...…………………………………………………………… ……82
5.3.1. Cure kinetics model development...……..…..…………………………… ……83
5.3.2. Viscosity (ƞ) model development………..……………………..………….. …..….85
5.3.3. Ionic conductivity (σ) model development………..………………..... …..…87
5.4. Results and Discussion……..…………………………………………………… …..….88
5.4.1. Out-time characterization………………………………...……………… …….88
5.4.2. Cure kinetics modelling………………………………………………… …....92
5.4.3. Viscosity modelling………..……………………………………………… …..….94
5.4.4. Conductivity modelling – viscosity monitoring………………………… …..….96
5.4.5. Viscosity controlled cure cycle development...………………………..... ………98
5.5. Conclusions………….……..…………………………………………………… …..….101

ix

Chapter 6. Processability of DDS Isomers Cured Epoxy Resin: Effects of Amine/Epoxy
Ratio, Humidity and Out-time………………………………..…………………… …..103
6.1. Introduction………………………………………………………………… …..…103
6.2. Experimental…………………………………………………………….……… …....106
6.2.1. Materials……………………………..……………………………… …..…106
6.2.2. Sample conditioning: humidity & out-time control..………………… …..…108
6.2.3. Modulated differential scanning calorimetry (MDSC)…………………… …..…108
6.2.4. Rheometry…………………………………………………………… …..…109
6.3. Theoretical Background…...…………………………………………….……… …....109
6.3.1. Cure kinetics model………………....……………………………… …..…109
6.3.2. Viscosity model…………………………………..………………… …..…116
6.4. Results and Discussion………………………………………………….……… …..….119
6.4.1. Out-time characterization…………....……………………………… …..…119
6.4.2. Cure kinetics evolution and modelling…………....………………… …..…125
6.4.3. Viscosity evolution and modelling…….……..…………………..… …..…127
6.4.4. Gelation point…………………………………………………………… …..…130
6.4.5. Glass transition temperature…………………………………………… …..…132
6.4.6. Resin flow control………….…………………………………………… …..…133
6.5. Conclusions.…………………………………………………………….……… …..….137

x

Chapter 7. Effective Cure Cycle Development via Flow Optimization and Advanced
Cure Environments…………………………………………………………………. …….139
7.1. Introduction………………………………………………………………… ……...139
7.2. Experimental…………………………………………………………….……… ……...141
7.2.1. Materials……………………………..……………………………… ……...141
7.2.2. Heated tool……………………………………....………………...… ……...141
7.2.3. Image analysis………………………………………………………... ……...142
7.3. Theoretical Background……….………………………………….……… ……...143
7.3.1. Viscosity model……………………...……………………………… ……...143
7.3.2. Flow model………………………………………....………………… ……...145
7.4. Results and Discussion………………………………………………….……… …..….147
7.4.1. Flow modelling……………………....……………………………… …...…147
7.4.2. Void content analysis……………….…………....………………… …...…149
7.5. Conclusions.…………………………………………………………….……… …..….161
Chapter 8. Conclusions and Future Work………………………………………… …..…163
8.1. Contributions…………………………………………………………….……… …..….163
8.2. Broader Implications………………………………………………….……… …..….167
8.3. Recommendations for Future Work…………………………………….……… …..….168
References………………………………………………………………………….. ……170

xi

List of Tables

Table 1-1. Cure monitoring techniques.……………………………………….……. ……...16
Table 3-1. Parameters for cure kinetics models…………………………………... ………43
Table 3-2. Parameters for viscosity models…...…………………………………... ………46
Table 3-3. Parameters for DCM models…………………………………………... ………49
Table 4-1. Parameters for ηr and σr during isothermal cure (log(η) on day 0 = 0.88, 0.48,
and 0.27 Pa· s at 93 ˚C, 107 ˚C, and 121˚C respectively and log(σ) on day 0 = -5.93, -5.53
, and -5.20 S/m at 93 ˚C, 107 ˚C, and 121˚C respectively)………………………..... ………68
Table 4-2. Parameters for α-relaxation analysis during isothermal cure at 93 ˚C, 107 ˚C,
and 121˚C on out-time…………………………………………………………....... ………71
Table 5-1. Parameters for cure kinetics models (where r = RH in fraction and to =
out-time in days)....................................................................................................... ….…84
Table 5-2. Parameters for viscosity & conductivity models (where r = RH in fraction and
to = out-time in days………………………………………………………………. ……..86
Table 6-1. Parameters for cure kinetics model for 33DDS (a/e = 0.6) (where rh = rh in
fraction and to = out-time in days)............................................................................ ……..112
Table 6-2. Parameters for cure kinetics model for 33DDS (a/e = 0.8)………….……... ……113
Table 6-3. Parameters for cure kinetics model for 44DDS (a/e = 0.8)………….……... ……114
Table 6-4. Parameters for initial degree of cure (α0) input to cure kinetics model and
viscosity model (where α0,a = actual initial degree of cure (used as an input to viscosity
model) and α0,f = fixed initial degree of cure (= 0.0015, used as an input to cure kinetics
model))………......................................................................................................... ….…115
xii

Table 6-5. Parameters for viscosity model for 33DDS (a/e = 0.6) (where rh = rh in
fraction and to = out-time in days)………………….……………………………... ……117
Table 6-6. Parameters for viscosity model for 33DDS (a/e = 0.8) (where rh = rh in
fraction and to = out-time in days)………………….……………………………... ……118
Table 6-7. Parameters for viscosity model for 44DDS (a/e = 0.8) (where rh = rh in
fraction and to = out-time in days)………………….……………………………... ……119
Table 6-8. Glass transition temperature (Tg) for all resin systems where Tg,121C and Tg,150C
are Tg for resins cured at isothermal dwell of 121 ⁰C and 150 ⁰C respectively and Tg, ꝏ is
Tg for fully cured resin………….………………….……………………………... ……133
Table 7-1. Parameters for viscosity & conductivity models (where r = RH in fraction and
to = out-time in days)…………….………………….……………………………... ……145
Table 7-2. Effective flow number (NFL,eff) for out-life (day 28) sample at manufacturer’s
recommended cure cycle………...………………….……………………………... ……151

xiii

List of Figures

Figure 1-1. VBO layup schematic [5]..……………………………………………. ……...2
Figure 1-2. Schematic of cure cycle.………………….…………………………... ………4
Figure 1-3. Schematic of viscosity evolution [16]..…………………………………... ………5
Figure 1-4. Epoxy resin composition…..…………………………………………... ………7
Figure 1-5. Resin mixture component structures – 4,4’-diaminodiphenyl sulfone
(44DDS), 3’-diaminodiphenyl sulfone (33DDS), triglycidal p-aminophenol (TGAP),
tetraglycidyl-4,4’-methylenebisbenzenamine (TGMDA), and polyethersulphone
(PES)………………………………………………………………………………. ……..9
Figure 1-6. Epoxy resin reaction…..…..…………………………………………... ………10
Figure 1-7. Schematic of the OoA prepreg consolidation process, showing unit cells
consisting of aligned fibers, surrounding resin, micro-voids within the tow cores and
macro-voids within resin-rich regions. CPT indicates cured ply thickness [50]…... ………17
Figure 2-1. Rheometer geometries and suitable sample types [70]……................. ………25
Figure 2-2. Phase angle illustration [70]………………………….……................. ………26
Figure 3-1. Cure kinetics measurement and model prediction of dynamic ramp and
isothermal dwell at day 0, 28 and 49. Upper row - Cure rate versus cure time: (a)
Isothermal dwell at 93°C (b) Isothermal dwell at 121°C (c) Dynamic ramp. Bottom row -
Degree of cure versus cure time: (d) Isothermal dwell at 93°C (e) Isothermal dwell at
121°C (f) Dynamic ramp.......................................................................................... ……..42

xiv

Figure 3-2. Viscosity measurement and model prediction of dynamic ramp and
isothermal dwell at day 0, 28 and 49 (a) Isothermal dwell at 93°C (b) Isothermal dwell
at 121°C (c) Dynamic ramp....................................................................................... ………45
Figure 3-3. DCM measurement and prediction of isothermal dwell at day 0, 28 and 49.
Upper row: Degree of cure versus cure time (a) Isothermal dwell at 93°C (b) Isothermal
dwell at 121°C. Bottom row: Cure rate versus cure time (c) Isothermal dwell at 93°C (d)
Isothermal dwell at 121°C......................................................................................... ………48
Figure 3-4. DCM measurement and prediction of dynamic ramp: (a) Day 0 sample.
without thermal adjustment (b) Day 0 sample with thermal adjustment (c) Degree of
cure versus cure time at day 0, 28 and 49 (d) Cure rate versus cure time at day 0, 28
and 49………………………………………………………………………………. ………50
Figure 3-5. DCM measurement and viscosity correlation. Upper Row - Isothermal dwell
at day 0, 28 and 49: (a) Viscosity versus cure time at 93º C (b) Viscosity versus cure time
at 121º C. Bottom Row – Dynamic ramp: (c) Viscosity versus cure time at day 0 with
thermal adjustment demonstration (d) Viscosity versus cure time at day 0, 28 and
49…………………………………………………………………………………… ………51
Figure 4-1. (A) Total heat of reaction (∆HT) and quadratic fit (dotted line) on out-time
(B) B-stage glass transition temperature (Tg,0) and quadratic fit (dotted line) on out-time
(C) Conductivity (log(σ)) as function of frequency (log(f)) and out-time at 30˚C (D)
Conductivity (log(σ)) versus out-time at fixed frequency (log(f) = 0.21 Hz, 2.08 Hz,
and 4.16 Hz) at 30˚C.................................................................................................... ………63

xv

Figure 4-2. (A) Conductivity (log(σ)) versus cure time on isothermal dwell at 121˚C on
day 0, 28 and 49 (corresponding tgel points are drawn in arrow) (B) Gelation time (tgel)
versus out-time on isothermal dwell at 93 ˚C, 107 ˚C, and 121˚C (Closed symbol:
dielectric analysis and Open symbol: rheometry). Isothermal dwell at 93 ˚C, 107 ˚C,
and 121˚C: (C) Reduced viscosity (ηr) versus out-time (D) Reduced conductivity (σr)
versus out-time……………………………………………………………………. ……..66
Figure 4-3. (A) Dipolar contribution (εd”) of dielectric loss (ε”) as function of frequency
(f) and out-time on isothermal dwell at 121˚C on day 0, 21, 35, and 49 (B) Frequency
(log(fmax)) corresponding to dipolar loss peak (εd,max”) during isothermal dwell at 121˚C
on day 0, 7, 14, 21, 28, 35, 42, and 49 (C) MDSC results on vitrification time (tvit) versus
out-time on isothermal dwell at (C) 93˚C, and (D) 121˚C........................................ ……….70
Figure 4-4. Minimum viscosity (ηmin) as function of ramp rate and out-time on
isothermal dwell at (A) 93˚C, and (B) 121˚C. Gelation time (tgel) as function of ramp rate
and out-time on isothermal dwell at (C) 93˚C, and (D) 121˚C…........................................ ……….74
Figure 5-1. (a) Weight percent of absorbed water (H2O) versus out-time at different
humidity conditions (b) Total heat of reaction (ΔH) and (c) B-stage glass transition
temperature (Tg,0) and on out-time........................................................................... ……89
Figure 5-2. (a) σmix versus out-time; (b) σmix and resin conductivity (σr) versus initial
degree of cure (α0); and (c) σmix versus α0................................................................. ……91

xvi

Figure 5-3. Cure kinetics measurement. α versus cure time under isothermal dwell at
121 °C for resins conditioned at: (a) RH = 30% for day 0, 21 and 35, (b) RH = 60% for
day 0, 21, 35 and 49, and (c) RH = 90% for day 0, 21 and 35. α versus cure temperature
under dynamic ramp for resins conditioned at RH = 60% for day 0, 21, 35 and
49……………………………………………………………………………….…… ….….93
Figure 5-4. Viscosity (ƞ) measurement. ƞ versus cure time under isothermal dwell at
121 °C for resins conditioned at: (a) RH = 30% for day 0, 21 and 35, (b) RH = 60% for
day 0, 21, 35 and 49, and (c) RH = 90% for day 0, 21 and 35. ƞ versus cure temperature
Under dynamic ramp for resins conditioned at RH = 60% for day 0, 21, 35 and
49…………………………………………………………………………………… …..….95
Figure 5-5. (a) critical signal detection demonstration using σ evolution (b) σ versus cure
and (c) tgel versus out-time (closed symbol: rheometry and open symbol:
DEA)………………………………………………………………………….……… …..….97
Figure 5-6. Schematic of flow enhanced cure cycle development method: Path 1. Target
ƞ (ex. 500 Pa· s); and Path 2. Target minimum viscosity (ƞmin)………..…….……… ….…99
Figure 6-1. Resin mixture component structures – 4,4’-diaminodiphenyl sulfone
(44DDS), 3,3’-diaminodiphenyl sulfone (33DDS), triglycidal p-aminophenol (TGAP),
tetraglycidyl-4,4’-methylenebisbenzenamine (TGMDA), and polyethersulphone
(PES)………………………………………………………………………………. ….…107
Figure 6-2. Weight percent of absorbed water (H2O) change on out-time and rh for: (a)
33DDS (a/e = 0.8) and 33DDS (a/e = 0.6); and (b) 33DDS (a/e = 0.8) and 44DDS (a/e =
0.8)………………………………………………………………………………….. ……..121

xvii

Figure 6-3. Initial degree of cure (α0,a) change on out-time and rh for: (a) 33DDS (a/e =
= 0.8) and 33DDS (a/e = 0.6); and (b) 33DDS (a/e = 0.8) and 44DDS (a/e = 0.8)... ……123
Figure 6-4. Initial glass transition temperature (Tg,0) change on out-time and rh for: (a)
33DDS (a/e = 0.8) and 33DDS (a/e = 0.6); and (b) 33DDS (a/e = 0.8) and 44DDS (a/e =
0.8)…………………………………………………………………………………. …….125
Figure 6-5. Representative cure kinetics measurement and model prediction of
isothermal dwell and dynamic ramp.………………………………………………. …….127
Figure 6-6. Representative viscosity (η) measurement and model prediction of
isothermal dwell and dynamic ramp.……………………………………….………. ……130
Figure 6-7. Gelation time (tgel) measurement method and change under isothermal dwell
(a) G’ and G” versus cure time for 33DDS resin with a/e = 0.8 (b) tgel versus out-time for
33DDS resin with a/e = 0.6 & 0.8 (c) tgel versus out-time for 33DDS resin and 44DDS
resin with a/e = 0.8……………………………………………………………….. ……132
Figure 6-8. η model prediction for 33DDS (a/e = 0.8, rh 90%) at day 30: (a) fixed ramp
rate to incremental dwell temperatures and (b) incremental ramp rate to the same dwell
temperature…………………………………………………………………………. ……..135
Figure 6-9. Effective flow number (NFL,eff) at various cure conditions for 33DDS (a/e =
0.8, rh 90%) at day 30……………...…………………………………………….…. ……136
Figure 7-1. Block diagram of the lamination process.………………………………. ……..140
Figure 7-2. In-house developed heated tool [144]……........................................... ……..142
Figure 7-3. Representative viscosity (η) model prediction for: (a) day 0 and day 28; and
(b) day28……………………………………………………………………………. ……148
Figure 7-4. Inverse viscosity (η
-1
) evolution plot of Figure 7-3 (b)…...………….. ……..149
xviii

Figure 7-5. Optical micrograph of uncured laminates at: (a) day 0 and (b) day 49. ……150
Figure 7-6. Optical micrograph of laminates cured under 2 ⁰C/min to 121 ⁰C hold at: (a)
day 0 (NFL,eff = 542.1 Pa
-1
) and (b) day 42 (NFL,eff = 10.2 Pa
-1
)……………………… …….152
Figure 7-7. Optical micrograph of laminates at day 42 cured under: (a) 2 ⁰C/min to 121
⁰C hold (NFL,eff = 10.2 Pa
-1
) and (b) 20 ⁰C/min to 160 ⁰C hold (NFL,eff = 40.5 Pa
-1
).. ……153
Figure 7-8. Optical micrograph of laminates at day 49 cured under: (a) 25 ⁰C/min to 160
⁰C hold (NFL,eff = 21.1 Pa
-1
) and (b) 35 ⁰C/min to 170 ⁰C hold (NFL,eff = 24.1 Pa
-1
).. ……154
Figure 7-9. Optical micrograph of laminates at day 56 cured under: (a) 1 ⁰C/min to 121
⁰C hold (NFL,eff = 1.8 Pa
-1
) and (b) 35 ⁰C/min to 170 ⁰C hold (NFL,eff = 11.6 Pa
-1
)… ……..155
Figure 7-10. Representative processing map of NFL,eff vs dwell temperature vs heating
rate (red: capability of conventional oven & teal: capability of heated tool) at day
49……………………………………………………………………………………… …….156
Figure 7-11. Void map of: (a) void content (%) vs out-time (days) vs NFL,eff (Pa
-1
) and (b)
macro-void content (%) vs out-time (days) vs NFL,eff (Pa
-1
)………………………. …….157
Figure 7-12. Void map of: (a) void content (%) vs flow-time (min) vs NFL,eff (Pa
-1
) and
(b) macro-void content (%) vs flow-time (min) vs NFL,eff (Pa
-1
)……………………… ……158
Figure 7-13. Viscosity profiles and optical micrographs of day 42, 49 and 56 samples
subjected under various cure cycles…………………………………………………. ……159
Figure 7-14. Inverse viscosity profiles and optical micrographs of day 42, 49 and 56
samples subjected under various cure cycles…..…………………….………..…… …….160
Figure 7-15. (a) void content (%) vs NFL,eff (Pa
-1
) and (b) macro-void content (%) vs
NFL,eff (Pa
-1
)…………………………………..……………..…………………..…… ……161

xix

Abstract

Out-of-Autoclave (OoA) prepreg allow the manufacturing of low-porosity, high-
performance composite structures through flexible, cost-effective vacuum-bag-only (VBO)
processing. These are cured in conventional ovens to produce autoclave quality parts with
low void contents (< 1% by volume for aerospace applications). In the absence of high
consolidation pressure, such as that imparted by an autoclave, voids are suppressed by
evacuating entrapped air and vaporized moisture through a partially impregnated
microstructure made up of both dry and resin-rich regions. During VBO processing, dry
regions are infiltrated by surrounding resin to form a uniform and ideally void-free
microstructure. The rate of infiltration and the quality of OoA laminates are therefore strongly
influenced by the cure kinetics and viscosity evolution of the infiltrating resin.
During pre-cure operations, the resin state can be affected by environmental factors
such as out-time and ambient humidity. Extended out-time can advance the degree of cure
and viscosity of the resin and potentially prevent full infiltration of the fiber bed during cure.
Exposure to ambient humidity generally leads to the rapid absorption of moisture. The
dissolved water can then evolve during cure, causing the nucleation and growth of voids with
internal pressures that exceed that which is imparted to the resin by vacuum bag-only
processing. Moreover, moisture absorption has also been shown to affect process-relevant
properties, such as degree of cure and viscosity, by acting as both catalyst and solvent for the
amine-epoxy reaction and accelerating cross-linking. The present thesis focuses on
xx

developing a flow-enhanced cure cycle to limit flow-induced defects during OoA prepreg
processing.
First, a representative OoA resin (CYCOM
®
5320-1, Cytec Industries Inc.) was
characterized using conventional Modulated Differential Scanning Calorimetry (MDSC) and
rheometry methods, and predictive models that comprehensively captured out-time effects
on cure kinetics and viscosity in process conditions were developed. Polymerization at
ambient temperatures altered the initial degree of cure of the resin as well as the evolution of
cure rate and viscosity at elevated temperatures. Subsequently, in situ dielectric monitoring
experiments were used to determine the evolution of key dielectric properties during the same
temperature and out-time conditions, and effective correlations were developed to obtain
degree of cure and viscosity from these properties. Together, these methods were shown to
constitute complementary methods for predicting and monitoring in-situ the “instantaneous”
degree of cure, cure rate, and viscosity evolution of composite prepreg.
Second, effective methods for monitoring out-time under ambient conditions,
identifying critical physicochemical events (minimum viscosity, gelation, and vitrification)
and their dependence on out-time (and out-life) for an OoA prepreg resin during cure were
investigated. Conventional ex-situ methods (MDSC and rheometry) were used to collect
benchmark data. Subsequently, in-situ dielectric analysis was conducted to gain a more
detailed understanding of the physical phenomena involved, and to develop a means for
detecting such phenomena in real-time. The results showed that while the out-life may
correspond to the onset of pervasive porosity, it does not constitute a fundamental shift in
xxi

resin properties. Rather, the processing-critical resin transitioned (minimum viscosity,
gelation and vitrification) gradually but significantly changed during the out-life specified
for the material. The results demonstrated that dielectric analysis can be used as a stand-alone
technique to monitor both the extent of out-time at ambient temperature as well as the major
effects of out-time during cure.
Third, accurate process models that comprehensively capture out-time and humidity
effects on cure kinetics and viscosity in process conditions were developed. DEA was used
to further develop a method for accurate process monitoring. The results indicated that out-
time and moisture absorption primarily affect the initial degree of cure and consequently
influence the course of cure kinetics and viscosity evolution during cure. Tangibly, these
changes decrease flow time, gelation time and reaction end time, and potentially complicate
manufacturing. The production of high-quality parts using OoA prepreg therefore requires
out-time and humidity control and/or appropriate thermal control to ensure adequate flow
time and to fully impregnate the prepreg during processing. In addition, it was demonstrated
that DEA allows accurate monitoring of initial degree of cure, gel-time, and reaction-end
time for the entire range of relative humidity studied, confirming the robustness of the
proposed methodology.
Fourth, three aerospace grade resins were formulated to investigate the effects of
variation in DDS isomers, amine to epoxy (a/e) stoichiometric ratio, out-time, and moisture
absorption on processing characteristics and cured resin properties. Conventional
thermochemical and thermomechanical methods (MDSC and rheometry) were used to collect
xxii

benchmark data. Regardless of the resin system, the results showed that out-time increases
initial degree of cure and more so with moisture absorption which influence the cure kinetics
and viscosity evolution during cure. These effects were shown to shorten flow level and time
which could potentially lead to insufficient resin flow during composite manufacturing. Then,
accurate process models were developed that comprehensively captured out-time and
humidity effects on cure kinetics and viscosity for each resin systems. Among the three resin
systems investigated, meta-substituted 33DDS with a/e = 0.8 exhibited the highest cure rate
and the lowest flow time. This behavior was attributed to the increase in collision number
between amine and epoxy compared to a/e = 0.6 and the lack of delocalization of the lone
pair of electrons on nitrogen compared to para-substituted 44DDS. Conversely, 44DDS with
a/e = 0.8 exhibited the slowest cure rate with the highest flow time. In addition, 44DDS based
resin exhibited the highest glass transition temperature due to having lower configurational
entropy than 33DDS based resin. The results offered practical insights regarding resin
formulation and prepreg processing, particularly that: (1) 33DDS-based resin is more
susceptible to aging and flow time reduction, albeit with faster processing time; (2) lower a/e
will require longer dwell at higher temperature to complete the etherification reaction, albeit
with more flow time; and (3) 44DDS resin offers longer flow time with higher glass transition
temperature, yet with substantially longer processing time.
Finally, the predictive viscosity model was used to develop a flow enhanced cure
cycle that can limit flow-induced defects by optimizing the viscosity profiles of aged samples,
effectively extending out-life. Effective flow number was defined as the integral of the
inverse viscosity profile until gel point to quantify the amount of resin flow during OoA
xxiii

prepreg processing. The model predicted that resin flow can be enhanced by implementing
fast ramp rates and high dwell temperatures. In-house developed heated tool was used to
achieve rapid heating and accurate temperature control. The results showed that optimizing
the effective flow number via temperature control enhances resin flow during cure thereby
extending out-life of the sample.
Overall, the work presented here contributed to OoA prepreg processing in four ways.
First, accurate predictive models that comprehensively captured out-time and moisture
absorption effects on cure kinetics and viscosity in process conditions were developed.
Second, in-situ cure monitoring methods that complements the predictive models were
developed. Third, foundational knowledge for aerospace grade DDS isomer cured resins with
varying amine to epoxy stoichiometric ratio subjected to pre-cure conditions that are largely
unavoidable in practice with prepreg processing were investigated. Finally, a flow-enhanced
cure cycle was developed to limit flow-induced defects during prepreg processing and to
extend out-life of the prepreg.

1

Chapter 1. Introduction

1.1. Motivation
Composite materials based on carbon fiber reinforced polymers (CFRPs) have
attracted significant interest as structural materials for primary aerospace structures, as well
as high performance sporting goods, and marine and wind energy structures. CFRPs are
multi-component materials that can comprise lightweight structures with high specific
mechanical (i.e. strength to weight ratio) properties and excellent fatigue life [1-2]. Thus,
CFRPs offer opportunities for improved fuel efficiency and reduced greenhouse gas
emissions. However, manufacturing of CFRPs based composite parts can be costly and time
consuming. Additionally, the demand for composites use in aircrafts is expected to quadruple
in the next decade. Therefore, more efficient and cost-effective manufacturing methods are
required to meet the growing demand.
Traditionally, structural composites for aircraft parts begin with layers of prepreg or
carbon fiber beds pre-impregnated with uncured resin. Individual prepreg layers are usually
between 0.1 ~ 0.5 mm thick when unprocessed, and contain approximately 30 ~ 40 % resin
by weight. These layers of prepreg are then stacked driven by end-use application and
vacuum bagged on a tool and placed in a pressure vessel or autoclave for processing [3-4].
Vacuum bagging comprises of enclosing the laminate within an assembly of consumables
designed to ensure adequate breathing (i.e. air and volatile evacuation) during cure (Figure
1-1). Then the bag is placed in autoclave where temperature is raised while the autoclave is
2

pressurized (up to 6 atm) while vacuum is pulled in the bag. The applied pressure conforms
the laminate to the shape of the tool and suppresses porosity by maintaining volatiles in liquid
state, which is the main defect in the prepreg-based manufactured parts. Finally, the cure step
is defined by temperature and pressure cycles.

Figure 1-1. VBO layup schematic [5]

Autoclave manufacturing is robust and well understood [6-9]. However, it incurs
high capital investment on equipment acquisition and operations cost. Autoclave
manufacturing is therefore inefficient and expensive. Also, autoclaves limit size of the part
to be within the inner diameter of the vessel. This size limitation also puts constraints on
other end of the spectrum where small parts are to be cured and processed in large vessel
thereby wasting energy. Therefore, autoclave cure method is not a viable option to meet the
growing demand for composites use in aircrafts.
3

In order to meet the growing demand, a new generation of Out-of-Autoclave (OoA)
processing method has been introduced targeting production of autoclave quality parts using
vacuum bag only (VBO) consolidation [10-12]. By taking the process out of the autoclave,
faster production with lower capital and processing cost are possible. In addition, high
pressure induced defects such as honeycomb core crush can also be avoided. However,
removing high pressure “safe guards’ that suppress the volatiles into solution and compact
and conforms the laminates into the tool shape poses new problems. Therefore, thorough
understanding of relation between material properties and processing parameters on defects
and porosity formation is required.
Critical analysis required then is an improved understanding of resin infiltration or
flow mechanisms during VBO processing [13-15]. Resin infiltration is influenced by both
cure temperature and chemical reaction. A traditional cure cycle for an aerospace-grade
epoxy resin consists of a combination of heating ramps (0.5°C/min to 3°C/min) and
isothermal dwells (121°C to 177°C), with dwell times on the order of hours (Figure 1-2).
Here, 121°C dwell mainly induces epoxy-amine reaction and 177°C dwell induces
etherification reaction. Such cure cycle ensures 100% cure of available reacting groups in the
resin.

4

Figure 1-2. Schematic of cure cycle

Resin flow is typically studied by investigating resin viscosity evolution during cure.
Figure 1-3 depicts viscosity evolution of the resin subjected to a standard cure cycle. The
viscosity evolution is governed by two competing effects: direct thermal effects and cross-
linking of the resin. Initially, during heat-up, the rate of cure is minimal and the viscosity
decreases due to increased molecular mobility [16-18]. Then, as the rate of cure increases,
the effect of cross-linking and increased molecular size begins to dominate, and the viscosity
increases. Therefore, the cure cycle must be designed to ensure adequate resin flow during
VBO processing.

5

Figure 1-3. Schematic of viscosity evolution [16]

During pre-cure operations, the resin state can be affected by environmental factors
such as out-time and ambient humidity. Extended out-time can advance the degree of cure
and viscosity of the resin and potentially prevent full infiltration of the fiber bed during cure
[19-22]. Exposure to ambient humidity generally leads to the rapid absorption of moisture.
The dissolved water can then evolve during cure, causing the nucleation and growth of voids
with internal pressures that exceed that which is imparted to the resin by vacuum bag-only
processing. Moreover, moisture absorption has also been shown to affect process-relevant
properties, such as degree of cure and viscosity, by acting as both catalyst and solvent for the
amine-epoxy reaction and accelerating cross-linking [21, 23-24].
Overall, the main defect-governing parameter during VBO processing is the
6

thermomechanical (cure-viscosity-temperature) changes of the resin. Insufficient resin flow
cause voids formation which is detrimental to mechanical performance of the laminates. Pre-
cure operations such as out-time and moisture absorption further deters resin flow promoting
voids formation. Therefore, clear understanding and predictive modelling of resin flow
during VBO processing is essential to limit flow-induced defects.

1.2. Resin Formulation
Among many thermoset resin types, epoxy-based resins have long been prevalent in
the aerospace industry because of its superior thermal and mechanical performance and ease
of processing [25-29]. Epoxy-based resins are composed of reactive components (epoxy
resins and cure agents/hardeners) and additives (tougheners, accelerators, flame retardants,
etc.). With great latitude in epoxy formulation, cost, processing, and performance (thermal
and mechanical) requirements can be optimized to meet a wide variety of application needs.

7

Figure 1-4. Epoxy resin composition

Amine based cure agents are the most diverse group of epoxy curing agents. Other classes of
cure agents such as anhydride and phenolic resins are also used in some applications yet the
easiness of processing and the breadth of performance imparted from amine based cure
agents are unparalleled. Generally, choice of amine based cure agents come in three folds: 1)
cost 2) processing requirements and 3) performance requirements. This thesis work focuses
on aerospace grade resin where multifunctional epoxy resins blended with aromatic diamines
and thermoplastic tougheners are used to meet the performance requirements. However, its
processing requires many considerations before use.
Processing requirements can be met by varying: (a) epoxy resins and amine based
cure agent types, (b) amine to epoxy (a/e) stoichiometric ratio, and (c) mixing conditions
(pot-life). On the other hand, the inter-relation of thermal and mechanical performance
complicates performance optimization. In general, formulations are best chosen to provide
8

thermal performance, which usually depends on glass transition temperatures (Tg), that are
slightly higher than the expected service temperature to maintain rigidity and surface
hardness [29]. However, not all desired mechanical properties (elongation, hardness,
modulus, strength and toughness) are optimized based solely on Tg. For example, choosing a
Tg much higher than necessary will decrease material toughness, while formulating resins at
a/e stoichiometric ratio 1 will tend to maximize Tg, and under-cure or off-stoichiometric
formation may provide more flexibility. a/e stoichiometric ratio of 0.6 ~ 0.9 are therefore
commonly used in aerospace applications [30].
The most commonly used aerospace grade aromatic amines are: tetra-functional
diamines 3,3’-diaminodiphenyl sulfone (33DDS) and 4,4’-diaminodiphenyl sulfone
(44DDS); shown in Figure 1.5. These two isomers of DDS are 3,3’-DDS (33DDS), which
features a meta substitution, and 4,4’-DDS (44DDS), which has a para substitution. Studies
have shown that these isomers result in different resin processing and cured properties due
to different energy dissipation mechanisms [31-32]. The 33DDS-based resins generally
exhibit greater flexibility than 44DDS-based resins because of higher configurational entropy.
This flexibility allows the polymer chains to rearrange at lower temperatures, eliminating
free volume to form more tightly packed amorphous networks (or less free volume), thereby
resulting in lower Tg values than 44DDS-based resins. In addition, the a/e stoichiometric ratio
has been shown to have complex effects on cured resin properties, where optimal thermal
and mechanical properties are obtained at different a/e stoichiometric ratios due to phase
separation within the resin [33]. Lower a/e stoichiometric ratios generally require longer
dwells at higher temperature to complete etherification caused by excess epoxy groups.
9

Figure 1-5. Resin mixture component structures – 4,4’-diaminodiphenyl sulfone (44DDS),
3,3’-diaminodiphenyl sulfone (33DDS), triglycidal p-aminophenol (TGAP), tetraglycidyl-
4,4’-methylenebisbenzenamine (TGMDA), and polyethersulphone (PES)

Epoxy – amine reactions are mainly represented by step growth polymerization [34].
Epoxy groups react with primary (1⁰) and secondary (2⁰)amines and anilines. When a/e is
equal to or greater than 1, epoxy-hydroxy or etherification reaction do not take place. The
reactivity of the amine based cure agents increase with its nucleophilic-ity: aliphatic >
cycloaliphatic > aromatic. Therefore, aliphatic amines are used in low temperature curing
applications such as adhesives and coatings. Anilines are used in high temperature cure
applications such as aerospace and automotive applications. Epoxy and amine reaction
produces hydroxyl groups which catalyze the epoxy and amine reaction by forming a
trimolecular complex. This facilitates the nucleophilic attack of the amine thus the overall
reaction becomes autocatalyzed. When a/e is less than 1, etherification reaction takes place
10

usually at higher temperature. Figure 1-6 summarizes all reactions taking place during epoxy
– amine reaction [35].

Figure 1-6. Epoxy resin reaction

1.3. OoA Processing (Cure Kinetics & Viscosity) Models

The key requirement for OoA processing of low porosity parts is the removal of air
entrapped during lay-up as well as cure induced volatiles which have been suppressed with
pressure provided by autoclave [36-38]. Taking the process out of autoclave, to minimize
porosity, OoA prepreg offer breathable feature where it forms partially impregnated
11

microstructures consisting of both resin-rich and dry areas. It is this dry area that is
relatively permeable in the initial stages of processing and allow gas migration towards the
laminate boundaries; then, they are infiltrated by resin to produce a void-free
microstructure. The flow then nearly halts upon reaching the gel phase where a pervasive
network is formed [37-43].
To allow gases to escape from the laminate's dry areas into the bag, OoA bag
assemblies must include permeable boundaries that connect the laminate to the breather
cloth. For in-plane gas evacuation, these paths may take the form of dry fiberglass strands,
cork or other edge breathing arrangements placed at the laminate edges. For through-
thickness air evacuation, perforated release film is used to separate the top laminate surface
from the breather. However, in-plane permeability has been consistently found to be a few
orders of magnitude greater than that of through-thickness direction as is apparent from the
design of prepreg [37-38]. The only way to improve through thickness permeability is
sparse dispersion of resin rich regions but it has not been studied yet.
The cure cycles used for OoA prepreg processing feature three important
differences relative to autoclave processing [44 - 46]. First, as the consolidation and
air/volatile evacuation depends solely on VBO process, a maximum bag pressure of 28 in
Hg or greater is required. Thus, vacuum bad leak proof test needs more attention. Second,
an initial room-temperature vacuum hold is recommended to evacuate gases entrapped
during laminate lay-up. This hold may range from a few hours for small parts to more than
16 h for larger components and structures featuring complex geometry. Lastly, OoA dwell
temperatures generally range from 93°C to 121°C, and are therefore lower than the
12

traditional 177°C used for epoxy-based autoclave prepreg. This decrease achieves two
aims: (1) epoxy and amine reaction separation from etherification reaction; and (2) it
ensures a resin viscosity profile that allows enough time for air evacuation through the dry
fiber regions but full prepreg impregnation before gelation, and it may also mitigate void
growth by limiting the void gas pressure.
Additional post cure at 177°C completes etherification reaction as we generally
have more epoxy than amine groups to maximize after cure glass transition temperature or
the part service temperature. These process descriptions offer useful guidelines and suggest
that successful OoA processing depends on several concurrent phenomena. However, they
provide few details on the fundamental physics that may be involved during manufacture.
Kratz et al. [17] characterized the cure kinetics, viscosity and glass transition
temperature behavior of two common OoA epoxy matrix resins. For the cure kinetics model,
they combined the models proposed by Kamal and Sourour [18] and by Cole et al. [39] into
the following expression:
𝑑𝛼
𝑑𝑡
= 𝐾 1
𝛼 𝑚 1
( 1 − 𝛼 )
𝑛 1
+
𝐾 2
𝛼 𝑚 2
( 1 − 𝛼 )
𝑛 2
1 + exp ⁡ ( 𝐷 ( 𝛼 − ( 𝛼 𝐶 0
+ 𝛼 𝐶𝑇
𝑇 ) ) )
(1-1)
Where Ki is the Arrhenius temperature dependency:
𝐾 𝑖 = 𝐴 𝑖 e x p ⁡ (
− 𝐸 𝐴𝑖
𝑅𝑇
) (1-2)

13

For the viscosity model, they resorted to the model used by Khoun et al. [40], a
modification of the model proposed by Castro and Macosko [41]:
μ = 𝜇 1
+ 𝜇 2
(
𝛼 𝑔𝑒 𝑙 𝛼 𝑔𝑒 𝑙 − 𝛼 )
𝐴 + 𝐵𝛼 + 𝐶 𝛼 2
(1-3)
where μ i is the Arrhenius temperature dependency:
𝜇 𝑖 = 𝐴 𝜇𝑖
exp ⁡ (
𝐸 𝜇𝑖
𝑅𝑇
) (1-4)
For the glass transition temperature, they used the common DiBenedetto
Model [42]:
𝑇 𝑔 − 𝑇 𝑔 0
𝑇 𝑔 ∞
− 𝑇 𝑔 0
=
𝜆𝛼
1 − ( 1 − 𝜆 ) 𝛼 (1-5)
In the above equations (1-1 ~ 1-5), R is the universal gas constant. In the cure
kinetics model, EAi is the activation energy of the resin; D is a diffusion constant; α C0 is the
critical degree of cure at absolute zero; α CT accounts for the latter's increase with
temperature; Ai, mi and ni are constants. In the viscosity model, Eμi is the viscosity activation
energy; αgel is the degree of cure at gelation; and A, B, C and Aμi are constants. In the glass
transition temperature model, Tg is the glass transition temperature corresponding to a
degree of cure α, Tg1 and Tg0 are the glass transition temperatures of fully cured and uncured
resin, respectively, and λ is a shape parameter used as a fitting constant. These models may
be used to predict the evolution of these properties for any temperature cycle.
Semi-empirical model developed by Kratz et al. imply that resin viscosity or flow
is highly dependent of the cure cycle selection. That is, the cure cycle can be “tuned” to
maximize flow time, thus minimizing flow-induced defect formation. Kratz et al.
tentatively defined flow time as time the resin spent below 100 Pa· s during cure which
14

persisted from 1 ~ 2 hours during conventional isothermal dwell. However, more in-depth
study is needed to determine (1) what constitutes flow time and (2) if the flow time can be
extended with cure cycle optimization.

1.4. Cure Monitoring Techniques
Cure monitoring is conducted by measuring quantifiable physicochemical
properties that represent instantaneous material state [19-21, 47-49]. Table 1-1 lists cure
monitoring techniques that have been extensively studied which are categorized by the
nature of the measurement method.
Thermal monitoring techniques, such as differential scanning calorimetry (DSC)
and differential thermal analysis (DTA), measure heat flow or heat capacity difference
between the sample and the inert reference [17, 20]. These techniques have set the
benchmark for cure monitoring as the analysis is relatively easy and the accuracy has been
determined well. However, disadvantage of thermal techniques is that they cannot be used
in-situ in industrial environments since they require use of small samples in idealized
conditions. In many cases, cure under such environment was found to be different from the
actual parts being cured. That is, actual part experiences temperature gradient, exothermic
heat buildup, etc. which are not captured in ideal conditions. Another problem is that the
small dimensions reduce the scale of heat transfer effects that may alter the actual
temperature of the curing part. However, thermal analysis can be used as a tool to test
validity of other monitoring techniques.
15

The principle of optical methods for cure monitoring lies in absorption of radiation
by molecular groups [43]. Typically, high energy radiation produced by laser is used as an
input and its response through the material during cure is measured and analyzed. In this
regards, optical methods use an external field to carry out the response however the
magnitude and the duration of the applied field is substantially greater than that of methods
presented in the last categories in Table 1-1. The excess of energy provided to the
molecular groups gives rise to several phenomena such us oscillatory movements (FTIR),
radiation emission (RS) or excitation to a higher electronic state (Fluorescence). All these
techniques use optical fibers as guiding media. Sophisticated software that usually comes
with the instrument is used for the identification of molecular groups in the spectra.
The most significant feature these optical techniques offer are that they are
molecular groups specific thus we can measure and analyze temporal evolution of these
groups during cure. Therefore, mechanistic approach to cure kinetics model can be built
which is usually more robust than any other semi-empirical models. Furthermore, cost
effectiveness and robustness offered by optical fibers certainly help industries to adopt the
technique for online cure monitoring improvements concerning the cost and robustness of
the optical fibers that carry the signal could render optical techniques applicable in
industrial environment. A drawback is that data analysis is not straightforward since the
responses from different molecular groups can overlap and also that the measuring area is
localized. Another problem in the analysis of experimental data is the definition of a
baseline needed for the calculation of conversion.

16

Table 1-1. Cure monitoring techniques
Method Techniques
Thermal Monitoring
Differential Scanning Calorimetry (DSC)
Differential Thermal Analysis (DTA)
Optical Methods
Fourier Transform Infrared Spectroscopy (FTIR)
Raman Spectroscopy (RS)
Fluorescence
Response To Small
External Field
Ultrasound
Dielectric Analysis (DEA)
Nuclear Magnetic Resonance (NMR)

The cure monitoring techniques that responds to small external field are the most
promising techniques of all. Ultrasound uses acoustic, DEA uses electric [47-48], and NMR
uses magnetic source as the source and measures resulting response during cure. The
different molecular species respond to this excitation in different time scales (relaxation
times). The obtained signal encompasses the responses of all the molecular species present
in the material. Therefore, the measurement provides macroscopic as well as microscopic
information about the system. Of the methods, DEA is the technique of interest in this work
thus will be discussed throughout the study. The principles of DSC are presented in Chapter
2.1 as a part of the description of the calorimeter used for cure kinetics experiments.

17

1.5. Voids Formation
There are three main sources of voids formed from VBO processing which are
entrapped air, evolved volatiles and insufficient resin flow [48-49]. Briefly, composite
laminates start with lay up of multiple prepregs. These are then vacuum bagged and vacuum
hold or debulk is applied to remove entrapped air. Subsequently, temperature is elevated to
allow resin crosslinking and flow to ideally produce a void free part. This process is
schematically shown in Figure 1-7.

Figure 1-7. Schematic of the OoA prepreg consolidation process, showing unit cells
consisting of aligned fibers, surrounding resin, micro-voids within the tow cores and
macro-voids within resin-rich regions. CPT indicates cured ply thickness [50]

18

Air can be entrapped between prepreg layers during layup process along with the
macro-voids already present in a prepreg. Majority of these air pockets are typically removed
during room temperature debulk given that adequate bagging for facile edge breathing is
done. Epoxy resins used in prepreg typically have very low volatile contents. However, as
discussed in Chapter 5, epoxy resins can absorb moisture and more so at higher relative
humidity environment. Therefore, these water has to be removed during debulk cycle and
processing.
Given adequate bagging and vacuum level, the main source of void in VBO
processing is flow-induced. This problem is unlikely with fresh prepreg or prepreg with low
out-time where sufficient flow level and time are expected with prepreg manufacturer’s
recommended cure cycle. However, out-time is unavoidable in industrial settings, as the lay-
up and preparation of large structures often takes several days to weeks. Out-time has been
shown to cause ambient temperature polymerization/cross-linking of the resin, which
adversely affects tack and drape and can lead to pervasive and unacceptably high void
contents in cured parts [5, 51-53]. Furthermore, the chemical reactions taking place during
room temperature cure may differ from those at elevated temperatures. Therefore, material
age must be carefully monitored.

1.6. Resin Flow
During OoA prepreg processing, B-staged thermosets softens and flows through
fiber beds and then stops at gelation. Thus, the resin exhibits viscoelastic characteristic
19

transitioning from viscous (liquid-like), gelation and to elastic (solid-like) behavior. During
the flow period, the thermosets are assumed to be Newtonian signified by linear relation
between shear stress and shear rate. Two supporting arguments have been made in the past
that supports the assumption. First, the flow occurs during early stages of cure where the
resin degree of cure is low. Therefore, it is likely that the resin remains a viscous fluid during
this stage and this behavior is expected to deviate as the resin approaches gel point. Second,
creeping flow is expected in high fiber volume preform where low shear rates is likely.
Newtonian fluid assumption has been widely used and investigated to model resin flow in
the past [54-60].
A squeezing flow geometry have been widely investigated for characterizing resin
flow during lamination [54-56]. Here, the prepreg layup is sandwiched between stainless
steel plates into the Instron test frame, where a linear variable differential transformer is used
to measure the movement of the cross heads. While the plates are heated, a constant force is
applied and measured by the load cell. Assuming that the resin flow can be approximated as
Newtonian before it gels, a characteristic flow quantity called flow number (NFL) has been
defined as follows:
0.5
2
1
4
0
16
1 ( ) 1 100
3
gel
t
p
o
Fl
o
Fh
N t dt
R






 

   
 


 


(1-6)
where ρp is the resin density, ρo is the prepreg density, F is the lamination press force, h0 is
the initial stack thickness, R is the effective radius of the resin.
20

The main variable controlling the flow quantity is:
1
.
0
()
gel
t
Fl eff
N t dt 



(1-7)
where NFL,eff is the effective flow number, and higher NFL,eff means more resin flow. With the
predictive viscosity model, there are two controlling variables that affect viscosity evolution:
heating rate and dwell temperature.

1.7. Resin Recycling
Even though resin recycling is not the focus of this thesis, several collaboration work
has been investigated. Please refer to these publications for further information [61-63].
Historically and currently, CFRPs at End-Of-Life (EOL) end up in landfills or incineration.
Though landfill is a relatively cheap way to dispose of CFRPs, it is the least preferred method,
as CFRPs are non-biodegradable [64]. Incineration is another commonly used waste disposal
method, which requires a large amount of heat and generates polluting emissions [64]. The
European legislation has passed laws forbidding landfill disposal of CFRPs [65], urging the
European composites industry to seek effective recycling solutions to the recycling problem.
The EOL Vehicle Directive (2000/53/EC) requires re-use and recovery of at least 95% of the
average weight of a car by 2015 [66]. Recycling is gaining more attention as environmental
legislation is becoming stricter, and many countries are expected to enact laws regulating
composite waste over the coming years. Thus, interest in effective recycling method that
generates valuable products is strong.
21

Various recycling methods have been studied focusing on recovering carbon fibers
from CFRPs by decomposing the polymer matrix using different techniques. Three routes of
recycling methods have been extensively studied: mechanical grinding, thermal processing,
and chemical treatment [64].
Mechanical recycling typically involves the use of cutting, crushing or other
mechanical processes to reduce the CFRPs into smaller pieces. Mechanical grinding
produces resin-rich powders and fibers of various lengths with resin, which can be used as
fillers or reinforcement for lower grade composites. However, clean carbon fiber of valuable
length cannot be recovered from this process. Mechanical grinding has been studied both on
carbon fiber and glass fiber reinforced composites, but more extensively on glass fibers.
Companies have attempted to upscale the method into industrial scale, but the method was
deemed unfavorable due to the low value of the recovered products compared to the facility
and energy costs incurred and the negative impacts on environment [64-65].
Thus, a thermal process was developed to recover incombustible carbon fibers from
the composites by decomposing the resin into lower molecular weight components [67].
Thermal processes include pyrolysis, fluidized-bed pyrolysis and pyrolysis assisted with
microwaves [64], among which pyrolysis is the most studied thermal treatment approach. In
a pyrolysis process, CFRPs are heated in the absence of oxygen. Under inert atmosphere, the
polymer matrix volatilizes into gas phase, and carbon fibers can then be separated and
recovered. The decomposed lower molecular weight organic components can be burned in a
combustion chamber, allowing the energy to be reused as supplementary energy source for
22

pyrolysis, thereby constructing the process loop. However, the thermal process generally
operates at a very high temperature, ranging from 450°C to 700°C depending on the resin,
and the thermal decomposed resin often leaves char on the fiber [67].
Chemical treatment is a relatively new recycling method that uses solvents and
catalysts as reactive medium to decompose polymer matrix into oligomers or monomers
under near- or supercritical pressure. Recovered carbon fibers are clean and can be recycled.
In chemical treatment, it is possible to adjust the properties of the reaction medium and
control the reaction by changing solvent, catalyst type, temperature and pressure [64]. Thus,
among recycling methods, chemical treatment has received the most attention. Supercritical
fluids provide unique reaction medium properties between gas and liquids, which has high
mass transfer coefficient and diffusivity [67-69]. Generally, high temperature and pressure
are required to reach near- or supercritical conditions due to the hydrogen bonds existing in
the solution. Under such severe conditions, reactors made of special alloys are required to
withstand the high pressures and the corrosion from strong chemicals. As a result, the facility
cost associated with supercritical chemical treatment can be steep. Thus, chemical treatment
under atmosphere pressure has become a new research focus in the recycling industries.

23

Chapter 2. Experimental Method

Basic principles of the key experimental techniques used in this thesis are
presented in the following sections. Details on the thermoset resin specific experimental
techniques will be introduced in subsequent chapters.

2.1. Differential Scanning Calorimetry (DSC)
Differential Scanning Calorimetry (DSC) is a thermo-analytical/thermo-chemical
tool which combines: 1. A calorimetry which measures the heat into and out of a sample; and
2. A differential calorimetry which measures the heat of sample relative to a reference (or an
empty pan). Combined, DSC can measure heat flows associated with physi-co-chemical
transformations of materials state. In DSC testing, samples (typically ~ 10 mg) are usually
enclosed in a disposable aluminum pan and the heat flow difference to the empty pan is
recorded as:
     
  
     
     
Sample Reference
dH dH dH
dt dt dt
(2-1)
where H is enthalpy or heat in units of energy. The process is called endothermic if ΔdH/dt
> 0, i.e. heat flow into the sample is higher than the reference. On the other hand, the
process is called exothermic if ΔdH/dt < 0, i.e. heat flow into the sample is lower than the
reference.
24

Standard DSC technique poses a few limitations such as temperature overlap of
physi-co-chemical transitions of complex materials and averaging of the heat flow rate of
overlapping processes. As a solution, Modulated DSC (MDSC) has been extensively used
to characterize complex materials. In MDSC, modulated (sinusoidal) heating rates are
applied on top of the linear heating rates which allow further analysis of the sample such as
heat capacity (Cp), glass transition temperature (Tg), etc. MDSC technique can be expressed
in the equation as follows:
( , )
p
dH dT
C f T t
dt dt

(2-2)
where dH/dt is the total heat flow, f(T,t) is the kinetic component of the total heat flow
(calculated from dH/dt and Cp components), and Cp· dT/dt represents reversing heat flow
component. By using MDSC technique, Cp, Tg, and melting can be analyzed using reversing
signal; and crystallization and enthalpy recovery at Tg can be analyzed using non-reversing
signal.

2.2. Rheology
Rheology is the study of the deformation and flow of material [70]. There exist two
types of rheometer: strain- and stress-controlled. A strain-controlled rheometer applies a
controlled strain to the material and measures its stress response and vice versa. The type of
sample that can be run in a rheometer is dependent on the geometry. Parallel plate is designed
25

to work on a wide range of samples from low viscosity liquids to soft solids which is usually
used to characterize resins, for example, the changes in viscosity that occur during a
thermoset cure procedure. Sample sizes will vary, but the instrument requires that the gap
between the plates be entirely filled with the test material, and the recommended sample
thickness is 0.5 - 2 mm. Cone and plate geometry can be useful when characterizing
thermoplastics as it provides uniform shear rate across the sample.

Figure 2-1. Rheometer geometries and suitable sample types [70]

Most polymers exhibit both viscous (liquid-like) and elastic (solid-like) behavior
under deformation. Such material is called ‘viscoelastic material’. Dynamic measurement
using a strain-controlled rheometer works by applying an oscillatory stress to the material.
The applied deformation is sinusoidal with the defined amplitude and frequency and the
response of the material is measured. The difference in input and output signal waves are
recorded to track phase angle (δ) of the material where δ of 0° means a perfectly elastic
26

material and a δ of 90º means a perfectly viscous material. Therefore, most real materials
have δ somewhere in between 0° and 90º .

Figure 2-2. Phase angle illustration [70]

The measured complex stress and strain signals in dynamic experiments allow
calculation of several key properties of the material. The complex modulus (G*), composed
of storage modulus (G’) and loss modulus (G”), and complex viscosity (η*) can be calculated.
These variables allow examination of cure and flow behavior of polymers.

2.3. Dielectric Analysis (DEA)
Dielectric Analysis (DEA) enables researchers to investigate a broad range of
phenomena from fast moving small ions in liquid polymers to slow moving molecules in
vitrified polymers [42-43]. Such range of physi-co-chemical processes are accessible by the
27

large frequency range (10
-6
Hz – 10
12
Hz) of DEA. The measurement is accomplished by
applying time-varying voltage to a material and measuring its time-varying current or
charge. The resulting time-varying current has a phase lag or δ just like the resulting strain
in the rheology experiment. Then macroscopic relaxation function in the frequency domain
called complex permittivity (ε*) can be measured and tracked. ε* is composed of
permittivity (ε’) and dielectric loss (ε”). The ε* sums the contributions from electrons,
atoms, molecules and molecular groups that react to the applied field.
In the case of cure monitoring, the main contribution comes from migrating
charges and dipolar relaxations. ε” measures energy loss associated with dipolar relaxation
and ionic conduction and the relation can be expressed as:
""
0
2




d
f
(2-3)
where σ is the ionic conductivity, f is the frequency, ε0 is the permittivity of free space and
εd” is the contributions from dipoles. The main difference between the rheology and DEA
measurement comes from σ which allows probing of thermoset curing to the extent that
may not be accessible by the torque limited rheometry.

2.4. Thermogravimetric Analysis (TGA)
Thermogravimetric analysis (TGA) is a thermo-analytical technique used to measure
volatile content and or degradation temperature of materials. TGA measures weight loss of a
sample as a function of temperature. TGA testing works by tracking the weight of a sample
28

(with a high degree of precision) as a function of temperature. As the temperature increases,
volatile components and absorbed moisture will evolve from the sample, causing its weight
to decrease. TGA testing to very high temperatures provides information on decomposition,
corrosion kinetics and oxidation behavior. A standard TGA test outputs weight % change in
the sample, plotted against temperature.
TGA alone can only measure weight loss as a function of temperature. Many weight
loss curves look similar and many volatiles evolve at similar temperatures, so without
additional equipment (Mass Spectrometer, FT-IR) detailed information on the species
evolving during testing is difficult to obtain.

2.5. Coulometric Fischer Titration
Titration unit has the ability to run liquid samples as well as solid samples [71-72].
To run a solid sample, which is the usual type of testing done in our lab, the drying oven
attachment is used. A coulometric titration unit is suitable for measuring moisture contents
of 1 wt % and less. Samples should be 1 g or smaller. Solid samples must have dimensions
that will fit inside the aluminum sample boats used for testing.

29

Titration utilizes the Bunsen Reaction between iodine and sulfur dioxide to
determine the amount of water, by weight, in a sample. The Karl Fischer reaction is the
following:

In this reaction water and iodine are consumed in a 1:1 ratio. Titration is done using a
chemical reagent. Once all of the water is consumed the amount of excess iodine is detected
by an electrode in the titration unit. The amount of water is calculated based on the
concentration of iodine in the reagent, and the amount of reagent consumed during the test.

30

Chapter 3. Out-Time Effects on Cure Kinetics and Viscosity for an Out-of-Autoclave
(OoA) Prepreg: Modelling and Monitoring

3.1. Introduction
Out-of-autoclave (OoA) prepreg allow the manufacturing of low-porosity, high-
performance composite structures through flexible, cost-effective vacuum-bag-only (VBO)
processing [75, 76]. In the absence of high consolidation pressure, such as that imparted by
an autoclave, voids are suppressed by evacuating entrapped air and vaporized moisture
through a partially impregnated microstructure made up of both dry and resin-rich regions.
Initially, the dry regions form a permeable vascular network of vacuum channels that allows
in-plane gas transfer under vacuum. During processing, they are infiltrated by surrounding
resin to form a uniform and ideally void-free microstructure. The rate of infiltration and the
quality of OoA laminates are therefore strongly influenced by the cure kinetics and viscosity
evolution of the infiltrating resin.
These resin properties are generally assumed to depend on the temperature cycle
used during cure, and are commonly described using semi-empirical analytical models [73,
75-81]. However, such equations ignore the potential influence of out-time prior to cure. Out-
time is unavoidable in industrial settings, as the lay-up and preparation of large structures
often takes several days to weeks. Recently, out-time has been shown to cause ambient
temperature induced polymerization/cross-linking of the resin, which adversely affects tack
and drape and can lead to pervasive and unacceptably high void contents in cured parts [74,
31

82-84]. Furthermore, the chemical reactions taking place during room temperature cure may
differ from those at elevated temperatures. Thus, out-time may affect not only the initial
degree of cure, but also the evolution of cure kinetics and viscosity during subsequent cure.
Altogether, accurate methods for monitoring and predicting the influence of out-time on key
resin properties are necessary.
Resin thermo-chemical properties have been previously characterized using
differential scanning calorimetry (DSC) [75, 77-83] and Fourier transform infrared
spectroscopy (FTIR) [80]. In addition, the influence of cure on thermo-mechanical properties
has been measured using rheometry [75, 79] and dynamic mechanical analysis (DMA) [87].
Generally, DSC is the most widely accepted means of characterizing the cure kinetics of
thermoset composite matrices, as it is based on the evolution of heat generation as a function
of cure time and temperature. Similarly, rheometry is commonly used to define the
composite’s gel point and measure the viscosity evolution in process conditions. To
characterize unknown resins, data is acquired in both dynamic ramp and isothermal dwell
temperature conditions, and used to develop various phenomenological models with several
fitting parameters [75, 79]. However, both methods are carried out ex-situ, on small-scale
samples, and in idealized conditions, rather than on the actual full-scale parts being
manufactured.
Recent developments in on-line cure monitoring techniques such as dielectric
analysis (DEA) [79-81, 86-87], ultrasonic monitoring [88], Raman infrared spectroscopy [79]
and optical fiber methods [88] have been demonstrated. Among them, DEA is the most
32

promising candidate for its non-invasiveness, high sensitivity, and potential wealth of
obtainable data. DEA relies on measurement of the mobility of ionic species in the thermoset
resin during cure by creating a resistive-capacitive circuit, which can be scrutinized and
potentially correlated to properties such as degree of cure, viscosity, gelation time and
vitrification. In particular, the evolution of imaginary impedance maximum (
"
max
Z ), which
directly relates resistivity levels (i.e. dissipative processes) to the amount and mobility of
charged species, has been shown to correlate to DSC signals (i.e. heat generation) [80-81,
86].
The ambient aging of autoclave prepregs has been studied using DSC [74], Micro-
CT [74], photoacoustic spectroscopy [83] and infrared spectroscopy [84]. Furthermore, the
effect of out-time on the thermochemical properties of an OOA prepreg resin has also been
investigated [82]. These studies have identified various phenomena at the resin chemistry
level. However, two critical needs have yet to be addressed - models that can accurately
predict the degree of cure and viscosity evolution for any cure cycle at any out-time, and
experimental methods capable of carrying out real-time, in-situ process monitoring on full-
scale parts for quality diagnostics and, in extremis, adaptive process control.
The present study seeks to develop accurate methods for predicting and monitoring
the effects of out-time on resin properties. To this effect, both traditional and emerging
methods are investigated. First, DSC and rheometry measurements obtained from neat resin
samples are used to understand and model the evolution of cure kinetics and viscosity for
various out-times and cure cycles. In addition, DEA measurements are carried out on prepreg
33

samples, and correlations are developed to allow resin cure to be accurately monitored in-
situ using dielectric data.

3.2. Experimental Procedure

3.2.1. Materials
For this study, I selected a commercially available OoA prepreg consisting of a
woven eight harness satin (T650-35 8HS) carbon fiber fabric and a toughened epoxy resin
(CYCOM
®
5320-1, Cytec Industries Inc.). The manufacturer’s stated out-life is 30 days at
ambient temperature [90]. Neat resin film was used for cure kinetics and viscosity
characterization using DSC and rheometer, and prepreg was used for the same purpose with
the DCM system. For both neat resin and prepreg, out-times and freezer times prior to use
were consistent and negligible.

3.2.2. Sample conditioning
Neat resin samples were pre-cut to nominal specifications of 10 mg for DSC tests
and 3 x 3 cm for rheometry tests. Prepreg samples measuring 2.5 x 2.5 cm were similarly
prepared for DCM tests. All samples were stored and aged separately in unsealed containers
at stable ambient conditions (21±2°C, 51±5% relative humidity).
34

3.2.3. DSC measurements
DSC measurements were performed with a TA Instruments Q2000 under a constant
nitrogen flow of 50 mL/min. For each measurement, 10 ± 2 mg of neat resin was sealed in
aluminum hermetic DSC pans (Tzero, TA Instruments). Dynamic ramps were conducted by
heating the DSC cell from -60º C to 280º C at a constant rate of 1.7º C/min, and under a
temperature modulation of ±0.5º C/min. The total heat of reaction (∆HT) of the resin was
determined by integrating the heat flow evolution from these measurements. Isothermal
dwell measurements were conducted at three different temperatures spanning the
manufacturer’s recommended cure range (93°C, 107°C and 121°C). The temperature in the
DSC cell was rapidly brought to the isothermal dwell temperature and held constant until
negligible heat flow was measured within the sample. After all isothermal tests, the DSC cell
was cooled to 20º C, then heated to 280º C at a constant heating rate of 1.7º C/min to measure
the residual heat of reaction (∆HR).

3.2.4. Rheometry measurements
Viscosity measurements were performed using 25 mm aluminum parallel plates in a
rheometer (TA Instruments AR2000). All tests were conducted under constant oscillatory
shear at a frequency of 1 Hz and at strain of 0.25%, within the linear viscoelastic regime. Pre-
cut neat resin samples were sandwiched between aluminum parallel plates and compressed
35

to a gap of 0.5 mm, and excess resin was manually trimmed to prevent boundary effects.
Dynamic ramps were conducted by heating at 1.7º C/min to 260º C, and isothermal dwells
were performed by heating at 10º C/min to 93°C, 107°C and 121°C and holding until a
stopping condition was reached. For both dynamic and isothermal tests, the stopping
condition was defined as 90% of the machine-specified maximum torque (200 mN∙m), to
ensure that measurements extended as close to the gel point as feasible. In addition, the same
limit was used as an initial condition for high out-times samples, which exhibited higher
room temperature viscosity (η) due to extensive cure.

3.2.5. DCM measurements
DCM measurement was performed using a dielectric monitoring system (DETA
SCOPE
TM
, ADVISE E.E., Greece). The system utilizes coplanar electrodes to generate
fringing electric field lines that penetrate into the dielectric material and allow non-
destructive measurements and one-sided access. Because the penetration depth was up to 100
µm, a glass fiber layer was added to the tool-side surface of the prepreg samples to insulate
the sensor from the conductive carbon fibers. DCM tests were performed within an
instrumented, integrally heated test cell. First, laminated prepreg samples 30 mm thick (with
glass fiber layer) was placed on top of the dielectric sensor, within a thin picture-frame spacer
containing a thermocouple. The assembly then was sandwiched between top and bottom
sections with embedded heating cartridges, which applied controlled and uniform heating to
the sample. Note that while the test cell was used for convenience, the DCM instrument
36

includes identical sensors that can be embedded at the surface of oven or autoclave tool plates,
or within liquid injection molds.
During each experiment, a 10 V excitation voltage was applied and twenty-five
frequencies were scanned, logarithmically spaced over a frequency window of 1 Hz to 1
MHz. From the applied voltage and scanned frequency, circuit analysis was used to determine
the evolution of the mobility of charged species within the resin as a function of time.
Dynamic ramp runs were conducted at 1.7º C/min up to 220º C (machine specified maximum
temperature) and isothermal dwells were conducted by heating at 10º C/min to 93°C, 107°C
and 121°C, and held constant until no significant change in the signals could be observed.

3.3. Model Framework

3.3.1. Cure kinetics model development using DSC measurement
To model cure kinetics, the total heat of reaction (∆HT) was first determined from
dynamic ramp data from the fresh sample. Assuming that the cure rate is directly proportional
to the heat flow measurement, the cure rate is defined as [73, 75-81, 86]:
1
T
d dH
dt H dt



(3-1)
where α is the degree of cure, dα/dt is the cure rate and dH/dt is the heat flow measured from
37

DSC. Integration of dα/dt versus time then yields α as a function of time (α(t)), which ranges
from 0 to 1, or fully-uncured to fully-cured.
Phenomenological cure kinetics models are widely used for the generally complex
cure reactions of epoxy matrices, where linear polymerization, branching and cross-linking
take place concurrently [73, 75-81]. Among these, a model was developed by Kratz et al. [73]
for the antecedent of the resin under study. This model accounts for the interplay between
kinetics- and diffusion- controlled reactions, and was modified to capture the out-time effects
through the following form:

0, ,
( ( ( )))
1,3 2,4
(1 )
(1 )
1 exp
jj
ii
j C j CT j
mn
j mn
i i j DT
ij
K
d
w K w
dt
  






  


(3-2)

,
exp
An
nn
E
KA
RT





where n =i , j (3-3)
where Kn is the Arrhenius temperature dependent term, An is the Arrhenius constant, EA,n is
the activation energy, mi and ni are reaction order-based fitting constants, Dj is the diffusion
constant, T is the temperature, αC0 is the critical degree of cure at absolute zero, and αCT
accounts for the increase in critical degree of cure with temperature. To account for the
changes associated with ambient temperature cure, which induces both time- and magnitude
shifts in the cure rate profile, the weight factors wi and wj, as well as the initial degree of cure
α0 were allowed to vary with out-time. A reference value of wi = wj = 1 was used as the
reference value for the fresh resin samples.
38

3.3.2. Viscosity model development using rheometry measurement
To model the viscosity evolution during cure, a phenomenological model used by
Khoun et al. [49] was modified to capture the out-time effects. The model includes an
additional polynomial term accounting for the viscosity at gelation, shown below:
1 1 2 2
()
de
gel A B C
gel
ww


  




(3-4)
exp
i
i
i
E
A
RT







, 1,2 i  (3-5)
where wi is a weight factor that depends on out-time, ηi is the Arrhenius dependent viscosity
component, Aηi is the Arrhenius constant, Eηi is the viscosity activation energy, αgel is the
degree of cure at gelation, and A, B, C, d and e are fitting constants. The model uses a degree-
of-cure profile obtained from a DSC experiment or a cure kinetics model as an input for α.
The viscosity evolution is governed by two competing effects: direct thermal effects and
cross-linking. Initially, during heat-up, the rate of cure is minimal and the viscosity decreases
due to increased molecular mobility. Then, as the rate of cure increases, the effect of cross-
linking and increased molecular size begins to dominate, and the viscosity increases. The
shift from thermal to cross-linking effects is captured by the d and e terms.

39

3.3.3. Cure kinetics and viscosity correlation development using DCM measurement
Skordos et al. [80] and Kazilas et al. [81] proposed a correlation between degree of
cure and dielectric properties that relies on the mobility of charged species in the resin,
measured as imaginary impedance (Z”) versus frequency along the cure profile. They
reported that the evolution of imaginary impedance maxima (
"
max
log Z
) during cure follows
the development of α, as obtained from DSC. The correlation relates
"
max
log Z
to α and T by:
"
max 11 12 2
log ( ) Z c c T c    
(3-6)
where c 11 and c12 are fitting constants determined from isothermal dwell measurements; and
c2 is
"
max
log Z
at α = 0. Differentiating equation (3-6) with respect to time yields:
" " "
max max max
log log log d Z Z Z d dT
dt dt T dt





(3-7)
"
max
11 12 12
log
()
dZ d dT
c c T c
dt dt dt

    (3-8)
where dT/dt is the heating rate term that allows modelling of non-isothermal conditions. A
dynamic ramp model can be expressed using the two previously obtained constants (c 11 and
c12). Skordos et al. [80] and Kazilas et al. [81] reported that the evolution of logZ
max
"
along
the cure profile exhibits competing effects similar to those occurring during the evolution of
η. Before the onset of cure, the rate of cure is minimal and
"
max
log Z
decreases due to
40

increased ionic mobility. As the rate of cure increases, the effect of cross-linking and
increased molecular size slows ionic mobility, and
"
max
log Z
increases. However, near the end
of the cure reaction, the thermal effect returns, and
"
max
log Z
begins to decrease, as little
further reaction takes place.
Using modifications similar to those made on cure kinetics and viscosity modelling,
equation (3-6) was modified to account for out-time effects as follows:
"
max 1 11 2 12 2
log ( ) Z w c w c T c    
(3-9)
where w1 and w2 are weight factors depending on out-time. The dα/dt profile was obtained
by differentiating equation (3-9). For this study, fitting constants and weight factors are
determined by correlating equation (3-9) with the cure kinetics model developed using DSC.
However, the model is not inherently necessary, as the constants may be obtained by relating
DCM data directly to DSC data. The α profile determined from equation (3-9) was used to
predict η by equation (3-4).

3.4. Results and Discussion

3.4.1 Cure kinetics measurement and modelling using DSC
Cure kinetics measurements and corresponding predictive model results obtained
41

using equation (3-2) are shown in Figure 3-1 (for the same time-temperature cycle). The
determined model parameters are provided in Table 3-1. Although not shown, isothermal
dwell data at 107°C was also taken into account. As described in section 3.1, out-time causes
both time-based (horizontal) and magnitude (vertical) shifts of the cure kinetics profile, along
with α0 increases. These effects are visible in Figure 3-1, where the top row shows that the
shape of the cure rate profile changes from 0 to 49 days of out-time, and the bottom row
shows that α0 increases as a consequence. Time-based shifts are more apparent at low
temperature (93°C) and during the dynamic ramp. In contrast, the magnitude of the cure rate
at high temperature (121°C) remains nearly unaffected during isothermal cure over a similar
time period. In this case, the cure rate is largely affected by the dynamic ramp due to the slow
heating rate and the high extent of room-temperature cure.

42

Figure 3-1. Cure kinetics measurement and model prediction of dynamic ramp and
isothermal dwell at day 0, 28 and 49. Upper row - Cure rate versus cure time: (a)
Isothermal dwell at 93°C (b) Isothermal dwell at 121°C (c) Dynamic ramp. Bottom row -
Degree of cure versus cure time: (d) Isothermal dwell at 93°C (e) Isothermal dwell at
121°C (f) Dynamic ramp

43

Table 3-1. Parameters for cure kinetics models
Parameter Value Parameter Value
A1 (s
-1
) 1.48 x 10
7
A3 (s
-1
) 6.39 x 10
7

E A1/R (K) 1.02 x 10
4
EA3/R (K) 8.94 x 10
3

m1 0.17 m3 1.65
n1 19.3 n3 16.6
A2 (s
-1
) 8.3 x 10
4
A4 (s
-1
) 9.8 x 10
4

E A2/R (K) 8.54 x 10
3
EA4/R (K) 7.1 x 10
3

m2 0.70 m4 1.66
n2 0.87 n4 3.9
D2 97.4 D4 63.3
αC0,2 -1.6 αC0,4 -0.60
αCT,2 (K
-1
) 5.7 x 10
-3
αCT,4 (K
-1
) 3.0 x 10
-3

Out-time parameters (t=days)
α0 = 5.98 x 10
-5
t
2
+2.5 x 10
-3
t +0.005
w1 = -5.34 x 10
-3
t
3
+ 2.40 x 10
-1
t
2
– 2.67t + 1
w2 = -2.74 x 10
-5
t
3
+ 2.24 x 10
-3
t
2
– 0.05t + 1
w3 = 3.21 x 10
-5
t
3
- 1.62 x 10
-3
t
2
+ 2.74t + 1
w4 = 1

44

Consistent (and predictable) α0 increases with out-time are observed for all cure
cycles, and can be described by the relation:
0
( 0) ( )
( 0)
TT
T
H day H out time
H day

   


(3-10)
Out-time induced polymerization/cross-linking causes α0 to increase. The thermoset resin
cure can be perceived as a blend of fast- and slow-occurring reactions, kinetics-driven and
diffusion-driven, respectively. Thus, out-time, where the reaction is induced at low
temperature, mainly affects kinetically driven reactions. Time-based cure rate shifts are more
apparent during low temperature (93°C) cure than during dynamic ramp conditions. High-
temperature (121°C) cure on the other hand is relatively unaffected, as sufficient heat is
provided to overcome the energy barriers associated with slow cure. Because of these
complex phenomena, previously developed cure kinetics models cannot accurately model
out-time effects. In contrast, the model developed here incorporates weight factors that
provide additional “degrees of freedom” that can capture cure rate shifts, while time-based
shifts are accounted for by the α0 values determined from equation (3-10). Thus, the present
cure kinetics models capture out-time effects accurately over the entire range of conditions
studied.

3.4.2 Viscosity measurement and modelling using rheometry
Viscosity measurements and results obtained from the predictive model using
45

equation (3-4) are shown in Figure 3-2 for the same time-temperature cycle. The viscosity
model parameters are provided in Table 3-2. The predicted viscosities accurately match the
measured values, even for long out-times. The sole notable deviations occur in dynamic
conditions, at high out-times and low temperatures (below 90°C), where the viscosity model
underestimates the measured values. In this regime, thermal effects dominate over cross-
linking (see Figure 3-1C), and the out-time-dependent weight factors used in the equations
may not fully capture the influence of ambient exposure. The first weight factor of the
modified model remained constant with out-time (w1 = 1), which acts as a baseline for
viscosity.

Figure 3-2. Viscosity measurement and model prediction of dynamic ramp and isothermal
dwell at day 0, 28 and 49 (a) Isothermal dwell at 93°C (b) Isothermal dwell at 121°C (c)
Dynamic ramp

46

Table 3-2. Parameters for viscosity models
Parameter Value Parameter Value
Aη1 (Pa· s) 4.52x10
-9
A 14.1x10
-6

E η1/R (K) 7.59x10
3
B 53.7
Aη2 (Pa· s) 1.73x10
-
14

C
-44.96
E η2/R (K) 1.24x10
4
d -0.13
αgel 0.66 e -0.11
Out-time parameters (t=days)

w1 = 1

w2

= 2.9 x 10
-5
t
3
- 2.3 x 10
-3
t
2
+ 2.8 x 10
-2
t +1

The second part of the modified model exhibits a complex dependence on out-time.
Indeed, w2 was found to decrease after 14 days, implying that the overall viscosity decreased
with out-time. Because the measurements clearly show that viscosity increases due to
ambient exposure and cure, we conclude that the agreement between viscosity predictions
and measurements can be attributed in part to the interplay with, and accuracy of, the
modified cure kinetics model and α input.

47

3.4.3 Cure kinetics and viscosity measurement and correlation using DCM
As described previously, the nature of the DCM correlation requires that α is first
correlated with
"
max
log Z
from isothermal dwell data, and dα/dt is determined from the
derivative. Correlation results are plotted in Figure 3-3 along with predictions of the cure
kinetics model, and the correlation parameters are provided in Table 3-3. In the correlation
data, α first decreases as
"
max
log Z
decreases due to thermal effects during heating. This non-
physical effect can be easily neglected by eliminating the data range during which the rate of
cure is negative. When the temperature reaches cure conditions (~90º C) and the rate of cure
increases, the degree of cure begins to increase, as expected. Past this point, the degree of
cure correlation closely matches the measured values for all out-times. The calculated dα/dt
exhibits small fluctuations due to scatter in the degree of cure measurements. However, the
scatter can be reduced by increasing the sampling interval or by data smoothing.
48

Figure 3-3. DCM measurement and prediction of isothermal dwell at day 0, 28 and 49.
Upper row: Degree of cure versus cure time (a) Isothermal dwell at 93°C (b) Isothermal
dwell at 121°C. Bottom row: Cure rate versus cure time (c) Isothermal dwell at 93°C (d)
Isothermal dwell at 121°C.

49

Table 3-3. Parameters for DCM models
Parameter Value
C11 (log(Ω)) 21.38

C12 (log(Ω)) -0.035
C2 (log(Ω)/K) 6.96
Out-time parameters (t=days)

w1

= 1.8 x 10
-4
t
3
- 1.6 x 10
-2
t
2
+ 3.7 x 10
-1
t +1

w2

= 2.9 x 10
-4
t
3
- 2.6 x 10
-2
t
2
+ 6.0 x 10
-1
t +1

In Figure 3-4, data from the dynamic ramp correlations are plotted alongside model
predictions. As shown in Figure 3-4A, the decrease in α is more pronounced than in the
isothermal data. The difference is attributed to the prolonged time required to reach cure
conditions (~90º C). In such cases, an effective alternate metric for the onset of cure can be
defined. During the
"
max
log Z
decrease period, we assume that α = α0, and cure is assumed to
start only after an increase in
"
max
log Z
is observed. When this “thermal adjustment” is
implemented, results from the dielectric measurement correlations closely match the
predictive model results at all out-times, as shown in Figure 3-4.

50

Figure 3-4. DCM measurement and prediction of dynamic ramp: (a) Day 0 sample without
thermal adjustment (b) Day 0 sample with thermal adjustment (c) Degree of cure versus
cure time at day 0, 28 and 49 (d) Cure rate versus cure time at day 0, 28 and 49

The α values obtained from the
"
max
log Z
correlation were substituted directly into
equation (3-4) to obtain viscosity data. The results are plotted in Figure 3-5 along with
predictive model data obtained using the cure kinetics model as input. The upper row plots
show isothermal dwell measurements, where the viscosity correlation generates values
several orders of magnitude greater than those obtained from rheometry. This discrepancy
can be attributed to the nature of dielectric measurement, wherein ionic mobility is monitored
instead of the actual mechanical response.
51

Figure 3-5. DCM measurement and viscosity correlation. Upper Row - Isothermal dwell at
day 0, 28 and 49: (a) Viscosity versus cure time at 93º C (b) Viscosity versus cure time at
121º C. Bottom Row – Dynamic ramp: (c) Viscosity versus cure time at day 0 with thermal
adjustment demonstration (d) Viscosity versus cure time at day 0, 28 and 49

The importance of the thermal adjustment is specifically highlighted in the bottom
row plots of Figure 3-5. Without thermal adjustment, prediction from dielectric measurement
over-predicts viscosity during periods where thermal effects dominate. The difference is not
as significant isothermal conditions, where the heating ramps (10 º C/min) to dwell
temperatures are relatively short. With thermal adjustment, however, the viscosity evolution
52

is accurately predicted from dielectric measurement.

3.5. Conclusions
This work investigated feasible, effective methods for quantifying the effects of out-
time on an OoA prepreg resin. First, a representative OoA resin was characterized using
conventional DSC and rheometry methods, and predictive models that comprehensively
captured out-time effects on cure kinetics and viscosity in process conditions were developed.
Polymerization at ambient temperatures altered the initial degree of cure of the resin as well
as the evolution of cure rate and viscosity at elevated temperatures. Subsequently, in-situ
dielectric monitoring experiments were used to determine the evolution of key dielectric
properties during the same temperature and out-time conditions, and effective correlations
were developed to obtain degree of cure and viscosity from these properties. Together, these
methods constitute complementary methods for predicting and monitoring in-situ the
“instantaneous” degree of cure, cure rate, and viscosity evolution of composite prepreg. Used
together, the methods offer a means to improve process effectiveness and efficiency in
composite manufacturing.
It should be noted that the rich dataset obtained from dielectric monitoring offers
additional opportunities for detecting and identifying key resin properties and transitions, and
for further evaluating the effect of out-time. These issues will be discussed in Chapter 4.

53

Chapter 4. In-Situ Cure Monitoring of an Out-of-Autoclave Prepreg: Effects of Out-
Time on Viscosity, Gelation and Vitrification

4.1. Introduction
Recent studies have shown out-of-autoclave (OoA) prepreg processing via vacuum-
bag-only (VBO) consolidation to be a viable and potentially cost-effective alternative to
autoclave cure [91-93]. OoA prepregs are cured under a much lower maximum compaction
pressure of 101,325 Pa (1 atm), and voids are suppressed by evacuating entrapped air and
volatiles through a microstructure that is initially only partially impregnated (by design), and
that is infiltrated by surrounding resin during cure [91]. The physicochemical properties of
the resin thus govern the rate at which this microstructure evolves, and must be understood
and appropriately controlled to ensure successful part manufacture. Therefore, the adverse
effects that derive from prolonged out-time, or room temperature exposure time before cure
are arguably even more critical in OOA processes than in autoclave processing.

4.1.1. Background
Out-time causes polymerization/cross-linking of the resin at ambient temperature,
adversely affecting tack and drape and potentially leading to unacceptable porosity levels in
cured parts due to inhibited flow [93-96]. For OoA prepregs, out-time was shown to cause
pervasive porosity once the out-life was exceeded. [94]. Therefore, accurate methods to
monitor out-time and to predict the effects of out-time on the key resin properties are
54

necessary. Investigators have measured out-time of prepregs using differential scanning
calorimetry (DSC) [93] and analyzed the resin’s physicochemical parameters (minimum
viscosity, gelation, and vitrification) using DSC [97], rheometry [98-99] and dynamic
mechanical analysis (DMA) [100]. Among these, DSC and rheometry techniques have been
particularly useful. However, DSC measurements are mass-specific, and the exact amount of
resin within a prepreg sample is difficult to quantify. Similarly, rheometry requires neat resin
film for viscosity data. In addition, both tests are conducted ex-situ, on small samples, in
idealized conditions.
Dielectric analysis is both non-invasive and highly sensitive to degree of
polymerization, and is thus appealing [99, 101-103]. Recent studies have shown that
dielectric analysis can be used to identify physicochemical transformations in prepregs
during cure, highlighting the possibility of distinguishing both micro- and macroscopic
information using appropriate signal processing. The minimum viscosity and gelation state
of the resin were identified as a maximum and an inflection in ionic conductivity, respectively,
during isothermal cure. Furthermore, vitrification can be detected by tracking α-relaxation
time from the dipolar contribution of the dielectric loss at a frequency of 0.1 Hz. These
correlations can be used to understand specific changes in the physicochemical parameters
and process-critical moments associated with out-time. However, such systematic studies
have not yet been reported.
In a recent study, Kim et al. [104] investigated the effects of out-time on the rate of
cure and viscosity evolution of an OOA prepreg and developed accurate predictive models
55

from DSC and rheometer data. Furthermore, we proposed accurate correlations between
dielectric behavior and resin cure kinetics and viscosity, thus demonstrating the potential for
an in-situ process diagnostic to dynamically monitor cure. In the present study, we analyze a
large thermochemical, thermomechanical and dielectric dataset to specifically determine the
effects of out-time on several resin properties of interest to composites processing. The two
key objectives of the study are: (1) to establish correlations between dielectric properties and
the key processing transitions of minimum viscosity, gelation and vitrification; and (2) to
investigate the effects of out-time and out-life on these moments.

4.2. Experimental Procedure
This experimental procedure was previously described in chapter 3, but the current
chapter considers a second, previously undiscussed set of data.

4.2.1. Materials
We selected an OoA prepreg consisting of eight harness satin (8HS) carbon fiber
fabric and a toughened epoxy resin (CYCOM
®
5320-1, Cytec Engineered Materials Inc.).
Prepreg was used for dielectric analysis, while neat resin film was used for modulated DSC
(MDSC) and rheometry. The specified out-life at ambient temperature was 30 days [106],
and samples were stored below -12˚C before use. The samples were then conditioned at
21±2°C and 51±5% relative humidity for 0 to 7 weeks.
56

4.2.2. Modulated differential scanning calorimetry (MDSC)
MDSC was conducted under a nitrogen purge (TA Instruments Q2000). Applying
sinusoidal temperature modulation on top of the linear temperature ramp allows signal
separation of the total heat flow into reversing and non-reversing components. The reversing
component of heat flow depends on heat capacity and heating rate, while the non-reversing
kinetic component directly indicates the cure exotherm [93, 97]. For each dynamic
measurement, a constant temperature ramp from -60 º C to 280 º C at a rate of 1.7 º C/min with
a temperature modulation of ±0.5 º C/min was applied. For isothermal measurements, dwells
were performed at 93 °C, 107 °C and 121 °C, and a temperature modulation of ±0.5 º C/min
was applied over the dwell period.

4.2.3. Rheometry
Rheological measurements were performed (TA Instruments AR2000) using 25 mm
aluminum parallel plates at a gap setting of 0.5 mm. All tests were conducted under constant
oscillatory shear at frequency of 1 Hz and at strain of 0.25%, within the Newtonian plateau
regime. Isothermal dwells were performed by heating at 10 º C/min to 93 °C, 107 °C and
121 °C and holding until 90% of the machine-specified maximum torque (200 mN∙m) was
required for shear displacement.

57

4.2.4. Dielectric analysis
Dielectric measurements were conducted with a dielectric monitoring system (DETA
SCOPE
TM
, ADVISE E.E., Greece). Coplanar electrodes were used as the sensor, which
generated fringing electric field lines (~100 µm) penetrating into the dielectric material (the
epoxy resin in the prepreg), thus enabling non-invasive measurements. To insulate the sensor
from the conductive carbon fiber bed, thin (10 µm) glass fiber layers were first placed over
the sensor. Laminated prepreg samples 30 mm thick were placed within a thin picture-frame
spacer containing a thermocouple. Finally, top and bottom plates with embedded heating
cartridges were used to uniformly heat the sample. A sinusoidal voltage of 10 V and a
frequency scan over the range from 1 Hz to 1 MHz were used for each measurement. The
comparison of the input and the return signals was automatically carried out in real-time
within the system software. The isothermal dwell heating profile was the same as that used
for rheometry measurements, with heating at 10º C/min to 93°C, 107°C and 121°C.

4.3. Principles of Dielectric Analysis
The electrical response of a thermoset resin during cure is better understood in terms
of the two major immitance functions: complex impedance (Z*) and complex permittivity
(ε*). The complex impedance, Z*, relates directly to electric circuitry, and correlates well
with cure kinetics and viscosity evolution during cure [104-105]. Permittivity, ε*, on the other
58

hand, permits detailed analysis of the physicochemical changes during cure through complex
analysis [99, 101-102]:
* ' "
   
(4-1)
where ε’ is a real part and ε” is an imaginary part of ε*. ε” can be further divided into two
parts for the thermoset resins as [99, 101-102]:
" " "
cd
   
(4-2)
where εc” is the conductivity contribution and εd” is the dipolar contribution. The former
term, εc”, dominates at low frequencies through a linear drop. Subtraction of εc” from ε”
yields εd”, and allows analysis of the dipolar relaxation of the thermoset resin.
Ionic conductivity (σ) is also related to ε” by [99, 101-102]:
0
2" f     
(4-3)
where ε0 is the free space permittivity of 9.85 pF/m [103]. The frequency dependence of σ is
useful because it relates to the length scale-dependent (microscopic or macroscopic) response
of the ionic mobility: the motion of ionic species over large distances (low frequency) reflects
macroscopic viscosity, and vice versa for microscopic viscosity.

59

4.4. Determination of Physicochemical Parameters

4.4.1. Out-time monitoring
Ex-situ methods for monitoring out-time were investigated by tracking changes to
the B-stage glass transition temperature (Tg,0) and the total heat of reaction (∆Hrxn), as
measured by MDSC. Glass transition manifests as an inflection point in the change in specific
heat within the reversing heat flow, while the heat of reaction is directly obtained by
integrating the non-reversing heat flow [93].
The method developed in this study for monitoring out-time in-situ involves
measurements of the conductivity σ at ambient temperature (chosen here as 30°C for stability)
via dielectric analysis. The thermoset resin contains significant amounts of ionic impurities
introduced during resin preparation. Thus, the resin transformation from liquid-like states
towards solid-like states during out-time will cause a decrease in measurable σ, even at
ambient temperature. In contrast to rheological measurements, dielectric analysis involves
measurement of ionic species motion, thus avoiding the limitations associated with material
viscosity/stiffness near gelation.

4.4.2. Gelation point and minimum viscosity
The minimum viscosity, ηmin, of a resin during cure is critical due to its influence on
60

resin infiltration, part impregnation and void minimization. Using the isothermal cure profile,
the ηmin measured from rheometry was associated with the maximum ionic conductivity
(σmax), the highest ionic mobility point during cure, from dielectric analysis.
Gelation refers to the moment when an “infinitely large” cross-linked network of
resin molecules is formed during cure. For epoxies, gelation is often defined as the moment
when viscosity (η) reaches 10 kPa∙s after passing minimum viscosity (ηmin) [99, 102]. In this
range, η increases quickly as cure effects dominate over thermal effects, and rheometric
measurements are constrained by the limits of machine torque. For dielectric measurements,
gelation was defined as the inflection point in the plot of log(σ) versus cure time. This point
corresponds to the maximum rate of σ change, which corresponds to the maximum decrease
in ionic mobility during cure [102].

4.4.3. Vitrification point
Vitrification is defined as the point at which the polymer chains become closely
packed, cooperative motion cannot occur due to insufficient volume, and the rate of cure
drastically decreases [102]. Therefore, vitrification is a good guideline to define the minimum
duration of the isothermal portion of a cure cycle. Using MDSC, studies have shown that the
inflection point of the heat capacity (Cp) during the isothermal dwell period corresponds to
the vitrification point [97].
Dielectric analysis, on the other hand, tracks α-relaxation of the resin during cure.
61

The α-relaxation process is interpreted as a dipolar loss peak (εd,max ) along the frequency-
time scale. That is, assuming that insignificant change occurs during each frequency scan,
εd,max  and the corresponding frequency (fmax) is tracked and extrapolated to 0.1 Hz, which
closely matches the vitrification point determined from MDSC [101-102]. While this
empirical finding may differ for specific resin systems, it allows one to track the evolution of
bulk relaxation time ( τb) of the resin during cure, defined as follows [107]:
1
max
(2 )
b
f 


(4-4)

4.5. Results and Discussion

4.5.1. Out-time monitoring (ex-situ vs in-situ)
The total heat of reaction (∆Hrxn) and the B-stage glass transition temperature (Tg,0)
measured from MDSC are displayed in Figure 4-1A and 4-1B. A quadratic fit captures both
data series accurately, where:
2
1 1 1 rxn
H A t B t C     
(4-5)
2
,0 2 2 2 g
T A t B t C    (4-6)
where t is the time in days, and the values of constants are as follows: A1 = 0.03 J/g· day
2
, B1
62

= 1.28 J/g· day,

C1 = 513.64 J/g, A2 = 0.01 °C/day
2
, B2 = 0.34 °C/day,

C2 = 1.31 °C. ∆Hrxn
measures the exothermic heat release during resin cure. Thus, the decrease in ∆Hrxn with out-
time in quadratic manner indicates that the polymerization/cross-linking at ambient
temperature accelerates with out-time. The trend is explained by the autocatalytic nature of
the epoxy cure reaction. The change in Tg,0 with out-time reportedly exhibits a linear
dependence until the resin cure temperature is exceeded, at which point vitrification occurs
[93]. However in the current study, Tg,0 increased in a quadratic manner with respect to the
out-time. Note that Tg,0 is a function of the amount of unreacted end-groups. Thus, a decrease
in the amount of unreacted end-groups due to out-time increases Tg,0 and decreases segmental
mobility. Furthermore, a quadratic dependence of ∆Hrxn on out-time leads to the same
expected trend in Tg,0.

63

Figure 4-1. (A) Total heat of reaction (∆HT) and quadratic fit (dotted line) on out-time (B)
B-stage glass transition temperature (Tg,0) and quadratic fit (dotted line) on out-time (C)
Conductivity (log(σ)) as function of frequency (log(f)) and out-time at 30˚C (D)
Conductivity (log(σ)) versus out-time at fixed frequency (log(f) = 0.21 Hz, 2.08 Hz, and
4.16 Hz) at 30˚C

From dielectric analysis, logarithmic conductivity (log(σ)) as a function of
logarithmic frequency (log(f)) and out-time at 30˚C is displayed in Figure 4-1C. The
measurements are averaged over 100 repetitions and the resulting error bars are drawn on
64

each data point. As described in Section 4.3, log(σ) is a strong function of the resin η, where
the resin transformation from a liquid-like to a solid-like state during out-time restricts the
mobility of ionic species and thus results in a σ decrease (i.e., σ is inversely proportional to
η). At out-times less than 28 days, log(σ) is relatively insensitive to frequency at low
frequencies and linearly sensitive at higher frequencies, qualitatively echoing the viscoelastic
response of the resin observed during rheological frequency sweeps [107]. The (insensitive)
low-frequency plateau region decreases with out-time and eventually disappears. After 28
days, log(σ) decreases at log(f) values greater than 4.16 Hz, yet due to fluctuations in
measured values (resulting from the viscous resin due to cure), further analysis in this region
was disregarded. Furthermore, although the log(σ) versus log(f) comparison qualitatively
portrays the viscoelastic response of the resin, the different nature of the measurements (ion
mobility vs mechanical response) precluded a direct quantitative correlation.
To illustrate the frequency dependence of σ, log(f) was fixed at 0.21 Hz, 2.08 Hz,
and 4.16 Hz. The corresponding log(σ) versus out-time chart, shown in Figure 4-1D, captures
the transition of the motion of ionic species from long distance (low f, macroscopic η) to
short distance (high f, microscopic η). At low f, where σ reflects macroscopic η, log(σ)
decreases linearly. However, from mid to high values of f, σ does not change significantly
until day 21 and 28, respectively. Such behavior requires a microscopic interpretation: the
concentration of ionic species is sufficiently low that polymerization/cross-linking (up to 21
and 28 days) does not significantly reduce ionic mobility, resulting in the observed plateau.
The low-frequency macroscopic conductivity, most closely related to rheological viscosity,
can be related to out-time as:
65

 
33
log B t C    
(4-8)
where t is time in days and the constants are B3 = 0.05, C3 = 8.64. This linear trend extends
from values of f of 1 Hz up to 10 Hz.

4.5.2. Gelation point and minimum viscosity (ex-situ vs in-situ)
A representative plot of σ evolution during isothermal dwell is shown in Figure 4-
2A, where log(σ) versus cure time is displayed for an isothermal dwell at 121˚C on days 0,
28 and 49. The log(σ) increases during heat-up, when thermal effects dominate over cure
effects, and the mobility of ionic species increases relative to the initial condition. After
reaching the cure temperature, cure effects begin to dominate over thermal effects, and log(σ)
decreases. The magnitude of log(σ) decreases throughout cure for increasing out-times,
showing, as expected, an inverse trend to the η evolution. The gelation time (tgel), starting
from cure onset, is taken to be the inflection point, calculated by taking the second derivative
of the curve for dielectric measurements, and 10 kPa∙s for rheological measurements.
66

Figure 4-2. (A) Conductivity (log(σ)) versus cure time on isothermal dwell at 121˚C on day
0, 28 and 49 (corresponding tgel points are drawn in arrow) (B) Gelation time (tgel) versus
out-time on isothermal dwell at 93 ˚C, 107 ˚C, and 121˚C (Closed symbol: dielectric
analysis and Open symbol: rheometry). Isothermal dwell at 93 ˚C, 107 ˚C, and 121˚C: (C)
Reduced viscosity (ηr) versus out-time (D) Reduced conductivity (σr) versus out-time

The tgel versus out-time is shown in Figure 4-2B for isothermal dwells at 93 ˚C, 107
˚C, and 121 ˚C (where closed symbols indicate dielectric measurements and open symbols
67

denote rheological measurements). The comparison shows close correlation between the two
methods, and a high degree of accuracy for a linear relation of the form:
1 2 3 4
( ) ( )
gel
t a a T t a a T     (4-9)
where the constants are a1 = 0.157 min/day, a2 = 19.855 min/day·°C, a3 = 1222 min, and a4
= 9.508 min/ °C and T is temperature in °C.
The minimum viscosity point is identified as ηmin and σmax during isothermal cures,
as both ηmin and σmax occurred simultaneously, regardless of out-time, for the same isothermal
dwell temperature. Normalized viscosity (ηr = ηmin(t) / ηmin(day0)) and normalized
conductivity (σr = σmax(t) / σmax(day0)) versus out-time are plotted in Figure 4-2C and 4-2D.
Both ηr and σr increase quadratically with out-time and more rapidly with increasing dwell
temperature. The fit parameters are shown in Table 4-1. Since ηmin or σmax is associated with
the highest resin flow rate (all other factors being equal), the results demonstrate that
dielectric analysis can be used to determine the potential moment of maximum infiltration
flow rate, and to obtain microstructural information.

68

Table 4-1. Parameters for ηr and σr during isothermal cure (log(η) on day 0 = 0.88, 0.48,
and 0.27 Pa· s at 93 ˚C, 107 ˚C, and 121˚C respectively and log(σ) on day 0 = -5.93, -5.53,
and -5.20 S/m at 93 ˚C, 107 ˚C, and 121˚C respectively)
Equation (t = days)
2
12
log( ) 1
r
b t b t    

2
12
log( ) 1
r
c t c t    

Parameters
Temp b1 (day
-2
) b2 (day
-1
) c1 (day
-2
) c2 (day
-1
)
93°C 5.64 x10
-4
0.019 4.45 x10
-5
1.33 x10
-3

107°C 9.56 x10
-4
0.026 3.47 x10
-5
1.97 x10
-3

121°C 2.22 x10
-3
0.007 3.47 x10
-5
2.28 x10
-3

4.5.3. Vitrification point (ex-situ vs in-situ)
Figure 4-3A shows representative dipolar contributions to dielectric loss (log(εd”))
as functions of logarithmic frequency (log(f)) and cure time at various out-time during
isothermal dwell at 121˚C. Typically, the maximum dipolar loss peak (εd,max”) occurs earlier
with increasing out-time. The frequencies (log(fmax)) corresponding to log(εd,max”) are plotted
versus cure time during isothermal dwell at 121˚C from day 0 to 49 (Figure 4-3B). Linear
69

fits corresponding to each out-time point indicate that the slope remains the same regardless
of out-time, but the intercept changes. Intercepts and slopes from different isothermal dwell
measurements are shown in Table 4-2. The trend shows that for a given out-time, both the
intercept and the absolute value of the slope increase with isothermal dwell temperature. The
behavior can be interpreted in light of the evolution of the polymer chain relaxation during
cure. Equation 4-4, which expresses the inverse relationship of fmax and bulk relaxation time
(τb), suggests that the value of τb at which vitrification occurs (tvit) will be shorter for higher
isothermal dwell temperatures.

70

Figure 4-3. (A) Dipolar contribution (εd”) of dielectric loss (ε”) as function of frequency (f)
and out-time on isothermal dwell at 121˚C on day 0, 21, 35, and 49 (B) Frequency
(log(fmax)) corresponding to dipolar loss peak (εd,max”) during isothermal dwell at 121˚C on
day 0, 7, 14, 21, 28, 35, 42, and 49 (C) MDSC results on vitrification time (tvit) versus out-
time on isothermal dwell at 93 ˚C, 107 ˚C, and 121˚C

71

Table 4-2. Parameters for α-relaxation analysis during isothermal cure at 93 ˚C, 107 ˚C, and
121˚C on out-time

Intercept
(t = days)
Slope
log(fmax)
at tvit
τvit
(s)
93˚C -0.23t + 21.83 -0.05 ± 0.001 -7.02 ± 0.12 1.74 x10
6

107˚C -0.21t + 23.19 -0.08 ± 0.002 -4.39 ± 0.22 4.42 x10
3

121˚C -0.18t + 24.20 -0.15 ± 0.004 -3.64 ± 0.28 8.17 x10
2

The vitrification time (tvit) versus out-time obtained from MDSC is shown in Figure
4-3C for isothermal dwell measurements. The plot confirm that tvit decreases with out-time
and with higher isothermal dwell temperatures. In our case, tvit is related to temperature and
out-time as
22
1 2 3 1 2 3
( ) ( )
vit
t d d T d T t e e T e T       
(4-10)
where T is in °C, t is in days, and the constants are d1 = 0.01 min/day, d2 = 1.15 min/day·°C,
d3 = 72.13 min/day·°C
2
, e1 = 0.42 min, e2 = 106.38 min/day·°C, and e3 = 6869.9 min/day·°C
2
.
The values of tvit decrease more rapidly at lower cure temperatures. As expected, since out-
time affects the initial degree of cure in a quadratic manner, the time required for fresh resin
to reach the initial degree of cure induced by out-time is more sensitive at lower isothermal
dwell temperatures.
72

Previous studies [99, 102-103] have identified the point of vitrification by
determining the cure time at which εd,max” occurs and recording the corresponding fmax. The
value of fmax was then related to tvit determined from MDSC. In those studies, tvit was closely
related to the fmax value of 0.1 Hz ( τb ~ 1.59 seconds using Equation 4), regardless of the
isothermal dwell temperature. However, for the present system, fmax at tvit increased markedly
with dwell temperature (and consequently, τb decreased). Also, while out-time decreased tvit,
τb was largely unaffected. This result supports the assertion that by determining τb at tvit (τvit)
as a function of temperature, one can determine the effective end of cure cycle on-line,
regardless of out-time. In the present case, τvit is related to temperature as
2
()
1
exp
kT
vit
k 


(4-11)
where τvit is in sec, k1 = 9.56· 10
16
sec, k2 = 0.27/°C and T is in °C. The relationship is resin-
specific, as expected from the results of other studies [9, 12-13].

4.5.4. Manufacturing considerations
The primary manufacturing consideration associated with out-time is part quality.
By considering the effects of out-time on viscosity and gelation, situation-appropriate cure
cycles can be developed. Figures 4-4A and 4-4B show maps of ηmin as a function of out-time
and ramp rate for 93 °C and 121 °C dwell temperatures, respectively. As expected, higher
temperatures enable lower viscosities. However, other dependencies also exist. At 93 °C (A),
the ηmin increases continuously with out-time, and this behavior is relatively insensitive to
73

ramp rate. However, at 121°C, at high ramp rates, the same property exhibits a low viscosity
plateau between 10 days and 30 days. This difference allows low viscosities to be achieved
at high out-times. For 30 days of out-time, a 3 °C/min ramp to 121 °C enables 13 Pa s to be
achieved, compared to, remarkably, the same 13 Pa· s at 10 days. In contrast, a 3°C/min ramp
to 93°C leads to a 30 day minimum viscosity of only 119 Pa· s, versus 41 Pa· s for the same
cure cycle and 10 days. Thus, for this resin, prepreg subjected to high out-times should be
rapidly heated to high temperatures to counteract the influence of out-time.

74

Figure 4-4. Minimum viscosity (ηmin) as function of ramp rate and out-time on isothermal
dwell at (A) 93 ˚C, and (B) 121˚C. Gelation time (tgel) as function of ramp rate and out-time
on isothermal dwell at (C) 93 ˚C, and (D) 121˚C

Figures 4-4C and 4-4D show the gel time as a function of out-time and ramp rate for
93°C and 121°C dwells, respectively. The former is associated to long tgel values due to low
cure rates, but is weakly dependent on ramp rate: at 30 days, at 0.5°C/min, the tgel is only 25%
longer than at 3°C/min. Conversely, the latter enables faster gelation but shows strong ramp
rate dependence: the 30 day tgel is twice as long at 0.5°C/min as at 3°C/min. Furthermore, at
75

121°C, the tgel is also less sensitive to increasing out-time. For all cure cycles, faster gelation
enables faster cure. For a given part, reducing elevated temperature processing time by one
(or a few) hours may be negligible, given that out-time is measured in weeks. However, in
situations where the curing vessel availability is rate-limiting, faster cure may enable more
parts to be manufactured per day.

4.6. Conclusions
In this chapter, effective methods (1) for monitoring out-time under ambient
conditions, and (2) for identifying critical physicochemical events (gelation, vitrification and
maximum flow) and their dependence on out-time for an OoA prepreg resin during cure were
investigated. First, conventional ex-situ methods (MDSC and rheometry) were used to collect
benchmark data. Subsequently, in-situ dielectric analysis was conducted to gain a more
detailed understanding of the physical phenomena involved, and to develop a means for
detecting such phenomena in real-time.
In Chapter 3, a method to predict and monitor the instantaneous degree of cure, cure
rate, and viscosity evolution of composite prepreg by dielectric cure monitoring in-situ were
presented. The results of the current study demonstrate that dielectric analysis can be used as
a stand-alone technique to monitor the extent of out-time at ambient temperature as well as
the major effects of out-time during cure. Ultimately, this information can be used to design
efficient cure cycles that ensure that the part has been fully cured, or control cure in real-time.
Therefore, dielectric monitoring methods offer a means to improve process effectiveness and
76

efficiency in composite manufacturing.

Appendix
A1. Gelation time (tgel) as a function of out-time (t) and isothermal dwell temperature (T):
1 2 3 4
( ) ( )
gel
t a a T t a a T    

where the constants are a1 = 0.16 min/day, a2 = 19.86 min/day·°C, a3 = 1222 min, and a4 =
9.51 min/ °C and T is temperature in °C.

A2. Vitrification time determined from MDSC (tvit) as a function of out-time (t) and
isothermal dwell temperature (T):
22
1 2 3 1 2 3
( ) ( )
vit
t d d T d T t e e T e T       

where T is in °C, t is in days, and the constants are d1 = 0.01 min/day, d2 = 1.15
min/day·°C, d3 = 72.13 min/day·°C
2
, e1 = 0.42 min, e2 = 106.38 min/day·°C, and e3 =
6869.9 min/day·°C
2
.

77

A.3 Vitrification time determined from dielectric analysis (τvit) as a function of out-time (t)
and isothermal dwell temperature (T):
2
()
1
exp
kT
vit
k 



where τvit is in sec, k1 = 9.56· 10
16
sec, k2 = 0.27/°C and T is in °C.

78

Chapter 5. Modelling and Monitoring of Out-time and Moisture Absorption Effects
on Cure Kinetics and Viscosity for an Out-of-Autoclave (OoA) Prepreg

5.1. Introduction
Over the past decade, increasing use of composites has created a demand for
manufacturing methods with lower costs and improved processing efficiency. As a result,
OoA prepreg processing has been introduced and implemented in industrial settings [108].
In the absence of high consolidation pressure, voids are suppressed by evacuating entrapped
air and vaporized moisture through a partially impregnated microstructure consisting of both
dry and resin-rich regions. Initially, the dry regions form a permeable vascular network of
vacuum channels that allow in-plane gas transport. During processing, these channels are
infiltrated with the goal of forming a fully saturated, defect-free microstructure. The rate of
infiltration and the quality of OoA laminates are therefore strongly influenced by the cure
kinetics and viscosity evolution of the infiltrating resin.
The manufacture of high quality, defect-free parts by OoA processes requires the
development of defect mitigation strategies and technologies, including in situ process
monitoring [109-112]. The accurate determination of the resin state prior to and during cure
is particularly important, as resin properties govern in-process phenomena and have a
substantial impact on final part quality and performance for all OoA processes. During pre-
cure operations, the resin state can be affected by environmental factors such as out-time and
ambient humidity. Extended out-time can advance the degree of cure and viscosity of the
resin and potentially prevent full infiltration of the fiber bed during cure [111-113]. Exposure
79

to ambient humidity generally leads to the rapid absorption of moisture. The dissolved water
can then evolve during cure, causing the nucleation and growth of voids with internal
pressures that exceed that which is imparted to the resin by vacuum bag-only processing
[113]. Moreover, moisture absorption has also been shown to affect process-relevant
properties, such as degree of cure and viscosity, by acting as both catalyst and solvent for the
amine-epoxy reaction [114-115] and accelerating cross-linking. Water absorption can also
disrupt hydrogen bonds among segments of the polymers and cause swelling of the polymer,
thus increasing the diffusion coefficient of water within the polymer and reducing the glass
transition temperature [114]. In addition, water can act as a catalyst, facilitating epoxy-amine
and etherification reactions [115]. Therefore, the combination of out-time and moisture
absorption lead to permanent change in the resin state prior to cure which affect cure kinetics
and viscosity evolution during cure.
The specific objectives of this study were to develop:
1. A methodology for tracking the resin degree of cure prior to cure of prepreg
subjected to humidity as well as out-time using in situ, real-time dielectric
monitoring.
2. Efficient models for cure kinetics and viscosity that capture out-time and
moisture absorption effects during consolidation.
3. A novel method for monitoring cure kinetics and viscosity capable of
accounting for both out-time and moisture absorption.
Finally, to demonstrate the utility of these methods, predictive models for resin
80

properties were used to understand the influence of out-time on resin viscosity evolution
during consolidation, prior to gelation, when resin flow takes place. For OoA prepreg with
exposure time exceeding the manufacturer’s specified out-life, two viscosity controlled (or
flow-enhanced) cure cycles were proposed to limit flow-induced defects using adaptive
heating and the model. The non-standard cure cycles were shown to theoretically increase
‘flow-time’ spent at targeted viscosity levels.

5.2. Experimental Procedure
The detailed protocols associated with sample preparation as well as thermochemical,
thermomechanical and dielectric measurement methods have been described elsewhere [111-
112], and are briefly described below. In this study, we investigated the additional nature and
effects of humidity conditioning on the properties of an OoA prepreg resin, and developed
efficient cure kinetics and viscosity modelling and monitoring methods.

5.2.1. Sample conditioning: humidity & out-time control
An OoA prepreg consisting of eight-harness satin (8HS) carbon fiber fabric and a
toughened epoxy resin (CYCOM
®
5320-1, Cytec Industries, USA) was used for analysis [9].
The initial out-time of all material was 0 days, and samples were stored frozen prior to testing.
Two types of samples were prepared: prepreg laminates were used for DEA, and neat resin
was used for DSC, rheometry, and TGA. All samples were stored in humidity chambers
81

created with saturated salt solutions, where accurate control of equilibrium vapor pressure is
possible [117], with RH levels of 30 and 90% for 0 to 5 weeks and of 60% for 0 to 7 weeks
and tested weekly.

5.2.2. Modulated differential scanning calorimetry (MDSC)
Dynamic and isothermal measurements were performed on a differential scanning
calorimeter (TA Instruments Q2000). For each dynamic measurement, a constant temperature
ramp from -60 °C to 280 °C at a rate of 1.7 °C/min was applied. For isothermal measurements,
dwells were performed at 93, 107, and 121 °C for an RH level of 60% to build accurate cure
kinetics models and at 121 °C for RH levels of 30 and 90% to quantify effect of moisture
absorption on cure. Both dynamic and isothermal dwells were subjected to temperature
modulation of ±0.5 °C/min.

5.2.3. Rheometry
The same dynamic and isothermal dwell tests were performed on a parallel-plate
rheometer (TA Instruments AR2000ex). The termination condition consisted of 90% of the
machine-specified maximum torque (200 mN· m).

82

5.2.4. Dielectric analysis (DEA)
Dielectric signals were collected using a (DETA SCOPE
TM
, ADVISE E.E., Greece).
Prepreg samples were placed in a test cell consisting of the dielectric sensor, resistive heaters,
and a monitoring and control system. Measurements were carried out under the isothermal
and ramp conditions described above by applying a sinusoidal voltage of 10 V , scanning
twenty-five selected frequencies (f) ranging from 1 Hz to 1 MHz. The measured signals were
automatically converted to a useful immitance function—complex permittivity (ε*)—which
was subsequently converted to ionic conductivity (σ). Studies [110, 118] have shown that σ
exhibits strong correlation to viscosity (η), yet no accurate modelling has yet been published.

5.3. Model Framework
In a previous study, we developed models for cure kinetics and viscosity (for this
resin system) capable of accounting for the cure cycle temperature profile as well as for
prepreg out-time using ‘weight factors’ [111-112]. In this study, we develop efficient cure
kinetics and viscosity models to support analysis of flow phenomena before resin gelation
during material processing while accounting for absorbed moisture and out-time effects.
Specifically, we focus on modelling up to 150 °C to separate the amine-epoxy reaction from
the etherification reaction (i.e. post cure) [120]. In addition, we develop a cure monitoring
method using ionic conductivity as a metric to correlate to the resin viscosity and cure
kinetics (from the conversion using equation 5-3).

83

5.3.1. Cure kinetics model development
To model cure kinetics, the total heat of reaction (∆H0) was first determined from
dynamic ramp data for the fresh sample. Assuming that the cure rate (dα/dt) is directly
proportional to the heat flow (dH/dt) measurement, the cure rate is defined as the measured
dH/dt divided by ∆H0. Integration of dα/dt versus time then yields α as a function of time
(α(t)), which ranges from 0 to 1, or fully-uncured to fully-cured. Then, a phenomenological
model developed by Kratz et al. [119] accounting for the interplay between kinetics- and
diffusion-controlled reactions was adapted and modified to capture out-time and moisture
absorption effects:
22
11
0
2
1 ( ( ( )))
(1 )
(1 )
1 exp
C CT
mn
mn
DT
dK
K
dt
  
  



  

(5-1)
,
1,2
exp
Ai
ii
i
E
KA
RT





(5-2)
where Ki is the Arrhenius temperature dependent term, Ai is the Arrhenius constant, EA,i is
the activation energy, mi and ni are reaction order-based fitting constants, D is the diffusion
constant, T is the temperature, αC0 is the critical degree of cure at absolute zero, and αCT
accounts for the increase in critical degree of cure with temperature. The first term from
equation (5-1) describes an Arrhenius type autocatalytic reaction, while the second adds a
diffusion factor to account for the shift from kinetics to a diffusion-controlled reaction as α
increases. To account for the changes associated with ambient temperature cure and moisture
absorption, which induce both time and magnitude shifts in the cure rate profile, the initial
84

degree of cure (α0) and the variables Ai, mi, ni, D, αC0, and αCT were defined as the general
form f(r, to) = g(r)to
2
+ h(r)to + i(r) , where g(r), h(r) and i(r) are second-order polynomials,
r is RH, and to is out-time. The model parameters determined are provided in Table 5-1.

Table 5-1. Parameters for cure kinetics models (where r = RH in fraction and to = out-time
in days)
α0 = (-5.6r
2
+9.3r+0.9)x10
-3
to+(4.06r
2
-2.52r-0.3)x10
-2

A1 (x10
6
s
-1
) = (-2.1r
2
+2.1r+2.7)to+(-62.5r
2
+81.6r-14.2)
A2 (x10
4
s
-1
) = (-5.0r
2
+4.9r+3.3)to+(-17.9r
2
+22.2r-9.8)
m1 = (2.1x10
-3
)to
2
+(-1.6x10
-2
r
2
+1.7x10
-2
r+9.1x10
-3
)to+(-7.6x10
-2
r
2
+0.1r+0.9)
m2 = (1.1r
2
+1.7r+3.0)x10
-3
to
2
+(0.53r
2
+0.81r+0.25)to+(8.6x10
-2
r
2
+0.16r+2.02)
n1 = (2.2x10
-3
)to
2
+(0.2r
2
-0.22r-7.5x10
-2
)to+(0.54r
2
-0.65r+10.6)
n2 = (2.2x10
-3
)to
2
+(4.8r
2
-5.4x10
-2
r-4.2x10
-2
)to+(-0.13r
2
+0.21r+3.74)
D = (0.17r
2
-0.18r-0.03)to
2
+(1.64r
2
-1.82r+2.84)to+(-5.0r
2
+1.05r+43.1)
αC0 = (-1.1x10
-3
)to
2
+(3.33x10
-2
r
2
-4.7x10
-2
r+3.66x10
-2
)to+(-4.61x10
-2
r
2
+0.11r-1.42)
αCT (x10
-3
K
-1
) =(2.87x10
-3
)to
2
+(-2.17x10
-2
r
2
-4.22x10
-2
r+7.18x10
-2
)to+(-
0.16r
2
+0.05r+5.20)
EA1/R (K) = 9.01x10
3
EA2/R (K) = 7.54x10
3

85

5.3.2. Viscosity (ƞ) model development
The viscosity of curing epoxy resin is affected by two competing effects. Heating
the resin increases molecular mobility, thereby decreasing viscosity, but also eventually
increases molecular size, increasing viscosity. To capture these phenomena, a
phenomenological model developed by Khoun et al. [121] was adapted and modified to
capture out-time and moisture absorption effects:
1
12
1
()


  




de
A B C
(5-3)
1,2
exp








i
i
i
i
E
A
RT
(5-4)
where ηi is the Arrhenius dependent viscosity component, Aηi is the Arrhenius constant, Eηi
is the viscosity activation energy, α1 is the degree of cure at gelation, and A, B, C, d and e
are fitting constants. The first term from equation (5-3) takes an Arrhenius type form, and
the second term is added to account for the rapid viscosity increase near gelation point.
Here, the variables Aηi, Eηi, α1, B, and d were defined as f(r, to) = g(r)to
2
+ h(r)to + i(r) to
account for changes associated with ambient temperature cure as well as moisture
absorption. In addition, the current model focuses on viscosity before gelation, at which
point deviations from Newtonian behavior are expected (gelation can be taken as ƞ ~ 10
86

kPa· s [111, 120] or σ inflection point [111-112]) and flow drastically reduces. The model
parameters are provided in Table 5-2.

Table 5-2. Parameters for viscosity & conductivity models (where r = RH in fraction and to
= out-time in days)
Aƞ1 (x10
-3
Pa· s) = (5.11r
2
-4.12r+2.23)to+(20.04r
2
-25.17r+8.46)
Aƞ2 (x10-
14
Pa· s) = (7.67x10
-4
)to
2
-(3.84x10
-2
)to+(-4.78x10
-2
r
2
-2.73x10
-2
r+0.5)
Eƞ1/R (x10
3
K) = (9.4r
2
-9.8r+4.1)x10
-3
to
2
+(-0.21r
2
+0.2r-0.14)to+(-
2.65r
2
+3.12r+1.97)
Eƞ2/R (x10
4
K) = (1.2r
2
-1.4r+0.2)x10
-3
to
2
+(-2.61r
2
+2.85r-0.23)x10
-2
to+(8.3x10
-
3
r
2
+8.5x10
-3
r+1.26)
α1 = (5.0x10
-3
)to
2
-(3.5x10
-2
r-1.67x10
-2
)to+(-8.93x10
-2
r+0.71)
B = (4.57x10
-2
r-8.9x10
-3
)to
2
+(-0.36r+0.29)to+(-1.38r+11.66)
d = (3.0x10
-3
)to
2
+(2.7r-6.7)x10
-3
to+(-3x10
-3
r+0.11)
A = 12.37 C = -0.43 e = -1.9x10
-3

L1 (x10
-4
S/m) = (-8.0r
2
+16.0r-6.0)x10
-4
to
2
+(4.11r
2
-6.3r+2.19)x10
-2
to+6.22x10
-2

L2 = (5.0r
2
-5.5r+1.1)x10
-3
to
2
+(-0.15r
2
+0.15r-2.97x10
-2
)to-0.79
L3 = (-7.3r
2
+8.2r-1.4)x10
-3
to
2
+(2.24r
2
-2.22r+0.37)x10
-1
to+0.68
L4 (x10
6
S/m) = (-0.6r
2
+1.2r+1.9)x10
-3
to
2
+(0.5r
2
-2.35r+12.24)x10
-2
to+1.83
L5 (x10
3
K) = (-1.67r
2
+1.2r+0.13)x10
-2
to
2
+(0.85r
2
-0.65r+0.14)to+1.71

87

5.3.3. Ionic conductivity (σ) model development
There are multiple ways to monitor cure using DEA. In previous work [111-112], we
proposed using imaginary impedance maxima, which measures ionic mobility and
concentration of charges, to monitor cure kinetics and viscosity evolution during cure.
Another approach consists of using dipolar contribution to dielectric loss to track the
evolution of the bulk relaxation time of the resin during cure. Both of these methods,
however, require significant analytical efforts. The evolution of ionic conductivity (σ) during
cure, however, provides rich information, including change in viscosity (η), minimum η (ηmin)
or maximum flow point, gelation time (tgel), and reaction end point (Figure 5-5) without
extensive analysis [109]. In particular, σ correlates closely with η, where increases in σ reflect
increased ionic mobility and thus decreased viscosity, and vice versa. Thus, accurate
modelling of σ up to the gelation point, where resin flow abruptly decreases, is useful. We
selected a mathematical form for σ based on the inverse relationship to η to describe the
evolution during cure:
23
()
5
14
exp( )




LL
mix
L
LL
T
(5-5)
where Li, i = 1 to 5 are fitting variables defined as functions of out-time and RH (Table 5-2),
and σmix represents the viscosity of a resin and water mixture. The model was developed
specifically to monitor viscosity evolution until gelation. Note that σ can be used as a separate
metric to model and monitor cure (section 5.4.4), but for the purpose of demonstrating the
correlation to viscosity, equation (5-5) was developed. Also, σ can be converted to α by taking
88

a reverse route from equation (5-5) to equation (5-3), thereby providing complete cure
monitoring capability.

5.4. Results and Discussion

5.4.1. Out-time characterization
Previous studies have shown that the primary volatile in OoA prepreg resins is
absorbed water, and that voids grow via diffusion of water from the surrounding resin [113].
To verify this, thermogravimetric analysis (TGA) was used to monitor mass stability during
heating, and results were compared to resin moisture contents measured by Coulometric
Fischer titration, during which prepreg samples were heated, and released moisture was
quantified using titration. Within the margin of error (< 5%), the weight percent values of
water contents measured by titration were equal to the values for total sample weight loss
determined by TGA. Figure 5-1(a) shows that the amount of water absorbed increased with
RH. At equilibrium, roughly 0.9 wt% water was absorbed for a RH of 90% while only 0.3
wt% was absorbed for a RH of 30%. Thus, at a molecular level, we expect that samples
conditioned at higher humidity levels will exhibit greater changes associated with the
catalytic/solvent effect.

89

Figure 5-1. (a) Weight percent of absorbed water (H2O) versus out-time at different
humidity conditions (b) Total heat of reaction (ΔH) and (c) B-stage glass transition
temperature (Tg,0) and on out-time

The effects of absorbed moisture on the total heat of reaction (ΔH) and the B-stage
glass transition temperature (Tg,0) during out-time were measured from MDSC data and are
displayed in Figures 5-1 (b) and (c). The results show that ΔH, which represents the
exothermic heat release during resin cure, decreases with out-time in a predictable manner.
Resin conditioned at higher RH exhibited a sharper decrease in ΔH, as absorbed water
facilitated cross-linking at the ambient temperature. This consistency allows prediction of the
initial degree of cure (α0), defined as a function of ΔH0 (the total heat of reaction at day 0)
and ΔH (the total heat released at various out-times and RH levels conditioned, Table 5-1),
resulting in a difference in α0 of up to 6% from RH of 30% to 90%.
The initial glass transition temperature Tg,0 exhibits RH-dependent changes with
out-time. Resin aged at 90% RH exhibited a decrease in Tg,0 at day 7 and had lower Tg,0
90

than the resin samples aged at 30% and 60% RH until day 21 (highlighted portion in Figure
5-1(c)), despite the progression towards a higher α. This difference is attributed to the
solvent effect dominating over the catalytic effect of water. Thus, tracking evolution of Tg,0
may not be a reliable way to track α0 for resin exposed to variable environmental
conditions. Overall, the combination of out-time and moisture absorption led to permanent
changes in the resin state prior to cure (i.e. increasing α0), which is expected to affect the
course of cure kinetics and viscosity evolution during cure.
Figure 5-2 (a) shows dielectric analysis (DEA) data expressed as logarithmic ionic
conductivity (log(σmix)) as a function of out-time at 30 º C. The subscript mix is used to denote
the mixture of water and resin. As DEA measures direct physical state of the prepreg, we
assumed that the measured conductivity can be expressed as:
   
mix r r w w
ww
(5-6)
where wr and ww are weight fraction of resin and water respectively, and σr and σw are ionic
conductivities of resin and water respectively (σw = 2 x 10-5 S/m at 30 º C). The reasoning for
describing conductivity as a weighted sum of resin and water was that (5-1) a weighted sum
is a simple, direct method for describing the properties of a composite, and (5-2) the low
moisture absorption level suggests that the nature of the resin has not been changed. With
this assumption and the known weight fraction of water in each sample (from TGA), σmix and
σr were plotted against α0 (Figure 5-2(b)). Although σw is four to five orders of magnitude
greater than σr at 30 ºC, its effect on the measured σ is limited by the low levels of absorbed
91

moisture in the resin. Therefore, the measurement of (log(σmix)) remains an effective method
of monitoring α0 (Figure 5-2(c)), despite the substantial effect of water on α0 during out-time.
This result indicates that accurate monitoring of out-time, even in challenging environments,
is possible due to the low moisture absorption of this OoA prepreg. However, this finding
also implies that the conductivity measured using dielectric analysis cannot be used to easily
determine the moisture content at a given out-time. For the case presented here, the relation
between (log(σmix)) and α0 is:
0
log( )  
mix
AB
(5-7)
where A is -11.31 S/m and B is -8.79 S/m.

Figure 5-2. (a) σmix versus out-time; (b) σmix and resin conductivity (σr) versus initial degree
of cure (α0); and (c) σmix versus α0

92

5.4.2. Cure kinetics modelling
Figure 5-3 shows representative cure kinetics measurements and corresponding
results of the predictive model obtained using equation (5-2). The cure kinetics model was
first developed using isothermal dwell data at 93, 107 and 121 °C and dynamic ramp data
with resin samples conditioned at RH = 60% from 0 to 49 days. Subsequently, isothermal
dwells at 121 °C and dynamic ramps were conducted on resins conditioned at RH = 30 and
90% from 0 to 35 days to capture the effect of moisture absorption during cure and determine
parameters in the cure kinetics equation. Confining the parameters to predictable patterns as
functions of out-time and RH (sec 5.3.1), the present cure kinetics model captured out-time
and humidity effects accurately over the entire range of conditions studied.

93

Figure 5-3. Cure kinetics measurement. α versus cure time under isothermal dwell at
121 °C for resins conditioned at: (a) RH = 30% for day 0, 21 and 35, (b) RH = 60% for day
0, 21, 35 and 49, and (c) RH = 90% for day 0, 21 and 35. α versus cure temperature under
dynamic ramp for resins conditioned at RH = 60% for day 0, 21, 35 and 49

Out-time induced cross-linking causes α0 to increase, and absorbed moisture can
further catalyze this reaction. Figure 3 shows that although the resin conditioned at higher
RH attained higher α0, the reaction end time was not affected as much. The result can be
understood by considering the reaction chemistry: the primary amine-to-epoxy reaction
occurs more rapidly than the secondary amine-to-epoxy reaction [121], as the former must
94

occur first in order for the latter reaction to take place, in addition to it having higher steric
hindrance. Thus, out-time and moisture absorption mainly affect the primary amine-to-epoxy
reaction, producing only small differences in reaction end time.
Conversely, from the standpoint of flow, a higher α0 and a shorter gel time [112],
which occurs prior to reaction end time, implies less time available for flow and full resin
impregnation during OoA processing. Therefore, the current work focuses on developing an
accurate model for α = f (α0, RH, to, t, T), as it will be used as an input to the ƞ model. However,
accurate α modelling also allows accurate prediction of cure rate (dα/dt).

5.4.3. Viscosity modelling
Figure 5-4 shows representative viscosity measurements and corresponding results
of the predictive model obtained using equation (5-3). As with the development of the cure
kinetics model, resin samples conditioned at RH = 60% were used to develop a benchmark
viscosity model, and resin samples conditioned at RH = 30 and 90% were modeled to capture
the effects of moisture absorption during cure and on viscosity equation parameters. Figure
5-4 shows that the viscosity model captured out-time and humidity effects accurately over
the entire range of conditions studied.

95

Figure 5-4. Viscosity (ƞ) measurement. ƞ versus cure time under isothermal dwell at 121 °C
for resins conditioned at: (a) RH = 30% for day 0, 21 and 35, (b) RH = 60% for day 0, 21,
35 and 49, and (c) RH = 90% for day 0, 21 and 35. ƞ versus cure temperature under
dynamic ramp for resins conditioned at RH = 60% for day 0, 21, 35 and 49

Generally, viscosity evolution for a cure cycle is governed by the competing effects
of cure (which increases viscosity) and heating (which decreases viscosity). Initially, as the
temperature increases and the resin degree of cure is low, the thermal effect is predominant,
and the viscosity decreases. As the degree of cure begins to accelerate and the temperature
approaches the isothermal dwell, the process becomes cure-driven, and the viscosity
96

increases at an increasing rate until gelation occurs. Such an effect is apparent in Figure 5-4.
In addition, the viscosity increases with ambient exposure, moisture absorption, and cure.
The result is especially important as viscosity directly affects resin flow, and the deviation
from Newtonian behavior is expected near the gel point (ƞ ~ 10 kPa· s [111, 120]), or the flow
stop point. In other words, the viscosity levels and resin flow times required to completely
wet fiber tows during processing are progressively limited by ambient temperature, moisture
absorption, and cure.
Therefore, the increase in viscosity due to ambient exposure and moisture absorption
is most likely the primary factor that determines the manufacturer’s stated out-life. In Section
5.5, the models developed for ƞ = f (α, RH, t, T) will be used to propose viscosity controlled
(or flow enhanced) cure cycles that can in principle extend the manufacturer’s stated out-life.

5.4.4. Conductivity modelling – viscosity monitoring
Figure 5-5 (b) shows representative conductivity measurements and corresponding
results of the predictive model obtained using equation (5-5). Resins were conditioned at RH
= 60% and used to develop a benchmark conductivity model, and resins conditioned at RH
= 30 and 90% were modeled to capture the effect of moisture absorption during cure and the
viscosity equation parameters. The model was developed using data up to the gelation point
(rheologically, ƞ ~ 10 kPa· s), which is defined as the inflection point in the plot of log(σ)
versus cure time. This point corresponds to the maximum rate of σ change, which
corresponds to the maximum decrease in ionic mobility during cure [111]. The resulting
97

model captured out-time and humidity effects accurately over the entire range of conditions
studied.

Figure 5-5. (a) critical signal detection demonstration using σ evolution (b) σ versus cure
and (c) tgel versus out-time (closed symbol: rheometry and open symbol: DEA)

Figure 5-5(a) depicts the maximum flow point or maximum conductivity (σmax),
gelation time (tgel), and reaction end point detection by monitoring conductivity evolution
during cure. In Figure 5-5(c), closed and open symbols correspond to traditional rheological
and DEA measurements, respectively. The tgel comparison shows close correlation between
the two methods with a high degree of accuracy for a quadratic relation of the form: tgel = f(to,
RH, T). Overall, the results demonstrate that DEA is an effective and robust method for
assessing material out-time prior to cure and for monitoring the evolution of key resin
properties at elevated temperatures. Furthermore, the model and methods presented here
eliminate post-processing of measured dielectric data [111-112] and allow direct and
98

efficient correlation of the measured conductivity to viscosity accounting for out-time and
moisture absorption.

5.4.5. Viscosity controlled cure cycle development
The predictive models developed here show that the influence of out-time and
absorbed moisture on the resin viscosity can be significant. Studies have found that these
undesired factors can lead to incomplete impregnation of the fiber bed [122]. To avoid these
issues, manufacturers specify an out-time limit (or an out-life) on material exposure. With
the accurate model for viscosity or conductivity developed in this study, it is possible to
develop a viscosity-controlled (or flow-enhanced) cure cycle that can potentially extend out-
life and limit flow-induced defects while minimizing cycle time.
The key assumption of this methodology is that as heat and time induce cure, a fast
ramp to a target viscosity is desirable, thus minimizing cure during the heat-up ramp [120].
Once the target viscosity level is attained, there is a competition between the cure effect,
which increases viscosity, and the thermal effect, which decreases viscosity, so a temperature
profile that maximizes the thermal effect can be specified. Figure 5-6 demonstrates two
methods for maximizing the thermal effect (and flow) for resin conditioned at RH = 60%.
The viscosity evolution at day 28 (~ out-life) was used as a benchmark that allows
manufacturing of quality parts (void content < 2% [122]), with one of the manufacturer’s
recommended cure cycles [117] (2.8 ⁰C/min to 121 ⁰C with a 3 h dwell). From this data, one
can determine a viscosity level and flow duration (defined here as time spent at or below the
99

target viscosity level) required for successful production.

Figure 5-6. Schematic of flow enhanced cure cycle development method: Path 1. Target ƞ
(ex. 500 Pa· s); and Path 2. Target minimum viscosity (ƞmin)

The first flow-enhanced method (Path 1 in Figure 5-6) targets a viscosity level of
500 Pa· s, which is below the gelation point (ƞ = 10 kPa·s or σ inflection point) and thus
maximizes the flow duration at this level. A demonstration of this approach using resin
conditioned at RH = 60% at out-time of 49 days, is shown. The method features a fast ramp
(~10 ⁰C/min) to 500 Pa· s, followed by a slow ramp (~0.9 ⁰C/min) to maximize flow duration
at viscosity level of 500 Pa· s or below. The method is developed by calculating the flow time
100

as a function of cure cycle using equation (5-3) and employing a simple gradient-based solver
to minimize the difference between the calculated flow time and a specified target (e.g., flow
duration at viscosity level of 500 Pa· s or below for ~47 min under the manufacturer’s
recommended cure cycle at the out-life). The results show that for an arbitrary target viscosity
and dwell time (strictly optimizing dwell time for viscosity at or below the target level), the
viscosity evolution of material with excessive out-time (49 days) can be rendered comparable
to that of the stated out-life.
The second flow-enhanced method (Path 2 in Figure 5-6) seeks to achieve a target
minimum viscosity level using a fast ramp (~10 ⁰C/min in this example) and subsequent
isothermal dwell (at 121 ⁰C). This method was developed in an attempt to maximize the
period of Newtonian flow.
The effective out-life of prepreg materials is reported to be determined by flow-
induced porosity caused by elevated viscosity [113, 122]. The case studies discussed above,
while simple, indicate that non-traditional cure cycles can counteract the effects of out-time
and thereby extend the effective out-life of aged material by leveraging the thermal effect
and lowering the viscosity profile during cure. The ramp rates required for these cure cycles
are not typically achievable during autoclave or oven cure. However, OoA materials can be
processed in a variety of curing environments, including heating blankets and heated tools,
which can significantly expand the thermal envelope available for processing. Finally, note
that the optimization process highlighted above does not yet account for cure cycle time.
However, models such as those developed in this work can be used within multi-objective
101

optimization algorithms that identify optimal solution fronts. Future efforts will combine
modeling with experiments within a parametric study targeting the two methods, to better
understand OoA prepreg consolidation phenomena and develop flow-enhanced cure cycles.

5.5. Conclusions
I have developed accurate process models that comprehensively capture out-time
and humidity effects on cure kinetics and viscosity in process conditions, and further
developed a method for accurate process monitoring using DEA. First, conventional ex-situ
methods (MDSC and rheometry) were used to collect benchmark data. Subsequently, in-situ
dielectric analysis was conducted to gain a more detailed understanding of the physical
phenomena involved, and to verify which phenomena can be detected in real-time. The
results indicate that out-time and moisture absorption primarily affect the initial degree of
cure and consequently influence the course of cure kinetics and viscosity evolution during
cure. Tangibly, these changes decrease flow time, gelation time and reaction end time, and
potentially complicate manufacturing. The production of high-quality parts using OoA
prepreg therefore requires out-time and humidity control and/or appropriate thermal control
to ensure adequate flow time and to fully impregnate the prepreg during processing. In
addition, I demonstrated that DEA allows accurate monitoring of initial degree of cure, gel-
time, and reaction-end time for the entire range of relative humidity studied in this chapter,
confirming the robustness of the previously proposed methodology. The study resulted in
practical insights, including (1) Tg measurements using DSC can be problematic for out-time
102

monitoring, since they can be affected by moisture absorption; (2) DEA is insensitive to
moisture absorption, rendering it effective for out-time measurements but unsuitable for
determining moisture contents; and (3) ΔH is a reliable/robust metric for both out-time and
moisture.
Viscosity measurements showed that out-time mainly increases resin viscosity and
limits resin flow during cure. Using the predictive model, we proposed non-standard cure
cycles that can potentially limit flow-induced defects by optimizing the viscosity profiles of
aged samples, effectively extending out-life. Validation studies on such flow-enhanced cure
cycles are in progress. Furthermore, while the models presented here were calibrated for a
specific resin system, the proposed equations and methods are applicable to many other
typical aerospace-grade epoxy resins with similar chemical formulations.
Overall, this chapter provides foundational knowledge for understanding and
optimizing consolidation phenomena as well as methods that can be used to predict material
properties or monitor them within advanced curing environments. Combined, they allow the
fabrication of low-defect parts under non-ideal material and process conditions.

103

Chapter 6. Processability of DDS Isomers Cured Epoxy Resin: Effects of
Amine/Epoxy Ratio, Humidity and Out-time

6.1. Introduction
Epoxy resins are widely used as matrix materials for prepreg or pre-impregnated
composite fibers which in turn are used in high-performance composites applications,
including primary aerospace structures. Upon curing, epoxy resins maintain the part shape,
protect the fibers from environmental degradation, and transfer loads to the fibers [123-
125]. Aerospace grade epoxy resins are typically composed of multifunctional epoxies,
aromatic cure agents, and tougheners, such as thermoplastics or liquid rubbers. The resins
are formulated to meet the desired processing characteristics (i.e., cure kinetics and
viscosity (η) evolution) and to yield the properties required of the cured resin, primarily
glass transition temperature (Tg) and mechanical property values. Among the variables
involved in resin formulation, cure agent type and amine-to-epoxy (a/e) stoichiometric ratio
have been shown to affect processing characteristics and cured resin properties the most
[126-128].
Epoxy resin cured with diaminodiphenyl sulfone (DDS), an aromatic cure agent, is
widely used for aerospace grade matrices [128-131]. The two isomers of DDS are 3,3’-DDS
(33DDS), which features a meta substitution, and 4,4’-DDS (44DDS), which has a para
substitution. Studies have shown that these isomers result in different resin processing and
cured properties due to different energy dissipation mechanisms [128-129]. The 33DDS-
104

based resins generally exhibit greater flexibility than 44DDS-based resins because of higher
configurational entropy. This flexibility allows the polymer chains to rearrange at lower
temperatures, eliminating free volume to form more tightly packed amorphous networks (or
less free volume), thereby resulting in lower Tg values than 44DDS-based resins. In addition,
the a/e stoichiometric ratio has been shown to have complex effects on cured resin properties,
where optimal thermal and mechanical properties are obtained at different a/e stoichiometric
ratios due to phase separation within the resin [127]. Lower a/e stoichiometric ratios
generally require longer dwells at higher temperature to complete etherification caused by
excess epoxy groups.
The production of high quality, defect-free composite parts requires defect
mitigation strategies. Especially, it is important to tailor cure cycles that yield the required
resin flow into fiber beds prior to gelation, the point at which resin flow stops. During pre-
cure operations, the resin state can be affected by environmental factors such as out-time and
ambient humidity [132-134]. Extended out-time can advance the degree of cure (α) and
viscosity (η) of the resin, potentially preventing full infiltration of the fiber bed prior to cure.
Exposure to ambient humidity during out-time leads to absorption of moisture. The absorbed
moisture can then evolve during cure, causing the growth of voids. In addition, moisture
absorption has also been shown to affect α and η of the resin by acting as both catalyst and
solvent for the amine-epoxy cure reaction [132].
There have been few studies relating the effects of variation in DDS isomers and a/e
stoichiometric ratio on resulting cured properties [128-131]. However, there have been no
105

studies reporting the effects of variation in DDS isomers, a/e stoichiometric ratio, out-time,
and moisture absorption on processing characteristics. In addition, cure kinetics and η
modelling accounting for these variables could offer insights into effects that could be useful
to develop defect mitigation strategies (e.g. flow enhanced cure cycle) for manufacturing
composite parts.
Therefore, in this work, three resin systems were created, varying DDS isomers and
a/e stoichiometric ratio, and these were subjected to several levels of out-time and humidity
conditioning. The key objectives of the study were to develop:
1. A methodology for quantifying the α prior to cure of each resin systems subjected to
humidity conditioning as well as out-time.
2. Models for cure kinetics and η that capture the effects of variation in DDS isomers,
a/e stoichiometric ratio, out-time, and moisture absorption.
In addition, process critical parameters were analyzed, including gelation time (tgel) and
minimum viscosity (ηmin), which influence resin flow time and a cured resin property, Tg.
We show that accurate process models that comprehensively capture out-time and
humidity effects on cure kinetics and η in process conditions for all three resin systems can
be developed. Variation in DDS isomers and a/e stoichiometric show tradeoff between cure
rate, resin flow time and Tg. Furthermore, we demonstrate that accurate process models allow
control of resin flow, thereby potentially mitigating flow-induced defects during prepreg
processing.
106

6.2. Experimental

6.2.1. Materials
The epoxy resin mixture components used are shown in Figure 6-1. Each of the epoxy
resin mixture compositions evaluated here contained two epoxy resins and a thermoplastic
polymer. The epoxy resins comprised a tetra-functional epoxy, tetraglycidyl-4,4’-
methylenebisbenzenamine (TGMDA; MY721, epoxy equivalent weight (EEW) ~ 112;
Huntsman Advanced Materials) and a tri-functional epoxy, triglycidal p-aminophenol
(TGAP; MY0510, EEW ~ 101; Huntsman Advanced Materials). A widely used impact
modifier, a thermoplastic toughening agent, comprised a functionalized polyether sulfone
(PES; SUMIKAEXCEL PES 5003P; Sumitomo Chemical Company). The curing agents
used in the resin mixtures were the aromatic amines: 4,4’-diaminodiphenyl sulfone (44DDS;
Aradur 9664-1, amine hydrogen equivalent weight (AHEW) ~ 62; Huntsman Advanced
Materials) or 3,3’-diaminodiphenyl sulfone (33DDS; Aradur 9719-1, AHEW ~ 62;
Huntsman Advanced Materials). Weight percent (wt%) ratio between TGMDA:TGAP was
fixed at 50:50, and PES comprised 15 wt% of the overall resin mixture components. Cure
agent type and amine-to-epoxy (a/e) stoichiometric ratio were varied to formulate following
three formulations: 1) 33DDS with a/e = 0.8; 2) 33DDS with a/e = 0.6; and 3) 44DDS with
a/e = 0.8. These formulations were chosen to yield the properties of a typical aerospace grade
epoxy resin [131].
107

Figure 6-1. Resin mixture component structures – 4,4’-diaminodiphenyl sulfone (44DDS),
3,3’-diaminodiphenyl sulfone (33DDS), triglycidal p-aminophenol (TGAP), tetraglycidyl-
4,4’-methylenebisbenzenamine (TGMDA), and polyethersulphone (PES)

Each resin mixture was prepared following the same procedure. TGAP and TGMDA
(50:50) were added to an aluminum container, and the container was placed into a convection
oven, where mixing began. The temperature was increased to ~110 ⁰C, and PES was added
and fully dissolved into the resin. Subsequently, the oven was cooled to 80 ⁰C, and the
specified stoichiometric level of 33DDS or 44DDS was mixed into the resin. Mixing
temperatures used herein were low enough to have negligible effect on epoxy amine cure.
Once a transparent solution was achieved, the solution was allowed to cool. The samples
were then stored in a freezer below -12˚C before use.

108

6.2.2. Sample conditioning: humidity & out-time control
The initial out-time of all resin mixtures was taken to be 0 days. All samples were
aged in humidity chambers containing saturated salt solutions, affording accurate control of
equilibrium vapor pressure. Chambers were maintained at 23 ± 1 °C and at relative
humidity (rh) levels of 30, 60 and 90% for 0 - 30 days. Subsequent testing was performed
within 10 days.

6.2.3. Modulated differential scanning calorimetry (MDSC)
MDSC measurements were conducted under a constant N2 flow of 50 mL/min (TA
Instruments Q2000). For each measurement, ~10 mg of resin was sealed in aluminum
hermetic pans (Tzero, TA Instruments). Non-isothermal cures were conducted by heating
the DSC cell from -60 ⁰C to 280 ⁰C at a constant heating rate of 3.0 ⁰C/min with a
temperature modulation of ±0.5 °C/min. The total heat of reaction (∆HT) of the resin was
determined by integrating the heat flow evolution from these measurements. Isothermal
dwells were performed at 121 and 150 °C for day 0 samples to build an accurate base for
cure kinetics models. Isothermal dwells were also performed at 150 °C for rh levels of 30,
60 and 90% for days 10, 20 and 30 to quantify the effects of moisture absorption and out-
time or aging on cure. After all isothermal tests, the DSC cell was cooled to 20 ⁰C, then
heated to 280 ⁰C at a constant heating rate of 3.0 ⁰C/min to measure the residual heat of
reaction (∆HR). The thermal stability of the resin was determined by thermogravimetric
analysis (TGA Q800, TA Instruments). Between 20°C and 280°C, the resin weight
109

decreased by less than 0.5%, excluding the amount water absorbed by the resin, indicating
that no resin degradation took place.

6.2.4. Rheometry
Viscosity (η) measurements were conducted using aluminum parallel plates in a
rheometer (TA Instruments AR2000). All tests were performed under constant oscillatory
shear at a frequency of 1 Hz and at strain of 0.25 %, within the linear viscoelastic (LVE)
regime of all resin systems. The resin samples were sandwiched between aluminum parallel
plates and compressed to a gap of 0.5 mm. Non-isothermal cures were conducted by
heating at 3.0 ⁰C/min to 280 ⁰C, and isothermal dwells were performed by heating at 10
⁰C/min to 121 and 150 °C for day 0 samples to build accurate base cure kinetics models,
and at 150 °C for rh levels of 30, 60 and 90% for day 10, 20 and 30 to quantify effects of
moisture absorption and aging on viscosity evolution. For both non-isothermal and
isothermal tests, the stopping condition was defined as 90 % of the machine-specified
maximum torque (200 mN∙m) to ensure that measurements extended to and beyond the gel
point as feasible.

6.3. Theoretical Background

6.3.1. Cure kinetics model
The following steps were taken to model cure kinetics. First, ∆HT was determined
from non-isothermal cure data by fully curing the day 0 sample. Then, assuming that a cure
110

rate directly proportional to the measured heat flow, the cure rate can be expressed as [135-
136]:
1
T
d dH
dt H dt



(6-1)
where α is the degree of cure, dα/dt is the cure rate, and dH/dt is the measured heat flow.
Integration of dα/dt versus time then yields α as a function of time.
0 , ,
( ( ( )))
(1 )
1 exp
ii
i C i CT i
mn
i
DT
K d
dt
  
 




(6-2)
where α is the degree of cure, dα/dt is the cure rate, and dH/dt is the measured heat flow.
Integration of dα/dt versus time then yields α as a function of time.
Epoxy cure reactions are complex, where linear chain extension and cross-linking
take place concurrently. Therefore, phenomenological cure kinetics models are generally
used to model epoxy cure reactions versus mechanical models [136-138]. In this study, a
phenomenological model [138] accounting for the interplay between kinetics-controlled and
diffusion-controlled reactions was modified to capture the effects of out-time and moisture
absorption, yielding the following form:
,
1,2
exp
Ai
ii
i
E
KA
RT





(6-3)
where Ki is the Arrhenius temperature-dependent term, Ai is the Arrhenius constant, EA,i is
the activation energy, mi and ni are reaction order-based fitting constants, D is the diffusion
constant, T is the temperature, αC0 is the critical degree of cure at absolute zero, and αCT
111

accounts for the increase in critical degree of cure with temperature. The numerator from
equation (6-2) describes an Arrhenius-type autocatalytic reaction, while the denominator
adds a diffusion factor to account for the shift from a kinetics-controlled reaction to a
diffusion-controlled reaction as α increases. To account for ambient temperature and moisture
absorption induced cure, which induce both time and magnitude shifts in the dα/dt profile, the
initial degree of cure (α0) and the variables Ai, mi, ni, Di, αC0,i, and αCT,i were defined as the
general form f(r, to) = g(r)to
2
+ h(r)to + i(r) , where g(r), h(r) and i(r) are second-order
polynomials, r is relative humidity, and to is out-time. The model parameters determined are
provided in Tables 6-1, 6-2, 6-3 and 6-4.

112

Table 6-1. Parameters for cure kinetics model for 33DDS (a/e = 0.6) (where rh = rh in
fraction and to = out-time in days)
A1 (x10
4
s
-1
) = (8.90rh
2
-10.8rh+6.51)x10
-2
to
2
+(-1.62rh+5.17)to+(5.0rh+146)
E A1/R (x10
-1
K) = (1.33rh
2
-1.47rh+59.2)x10
-2
to
2
+(-6.33rh
2
+7.37rh-3.18)to+(-
1.67rh-168)
m1 (x10
2
) = (-22.2rh
2
+27.6r-2.01)x10
-2
to
2
+(8.89rh
2
-11.0rh-11.9)x10
-1
to+(9.44rh
2
-
9.83rh-70.3)
n1 (x10
2
) = (-7.31rh+1.83)x10
-3
to
2
+(2.59rh-1.01)x10
-1
to+(-1.17rh
2
-1.58rh+8.27)
D1 = (-1.07rh+2.87)x10
-4
to
2
+(4.07rh-13.3)x10
-3
to+1.40
αC0,1 = (7.08rh
2
-8.15rh+1.96)x10
-1
to
2
+(-21.8rh
2
+26.2rh-6.73)to-49.7
αCT,1=1.10x10
-2
A2 (x10
4
s
-1
) = (28.9rh
2
-30.9rh+8.86)x10
-3
to
2
+(-3.22rh
2
-3.35rh-1.84)x10
-1
to+10.6
E A2/R (K) =-9.37x10
-2

m2 (x10
2
) = (-5.10rh-1.54)x10
-1
to+(-8.89rh
2
+8.67rh+76.9)
n2 (x10
1
) = 4.30x10
-3
to
2
+(28.3rh
2
-23.9rh+5.68)x10
-1
to+(-1.33rh+29.1)
D2 = (-19.8rh
2
+23.7rh-4.10)x10
-2
to
2
+(50.1rh
2
-58.6rh+7.40)x10
-1
to+(-6.67rh
2
-
8.0rh+51.5)
αC0,2=-7.72 x10
-5

αCT,2 (x10
3
) = (198rh
2
-260rh+6.50)x10
-4
to+1.66

113

Table 6-2. Parameters for cure kinetics model for 33DDS (a/e = 0.8)
A1 (x10
2
s
-1
) = (27.4rh-1.16)x10
-4
to
2
+(-100rh-1.80)x10
-3
to+(1.40)
E A1/R (x10
-3
K) = (4.98rh
2
-7.69rh+1.88)x10
-3
to
2
+(-15.0rh
2
+23.6rh-5.41)x10
-2
to-
1.44
m1 (x10
1
) = (6.81rh
2
-5.71rh+1.91)x10
-3
to
3
+(19.6rh
2
-20.7rh+1.71)x10
-2
to
2
+(16.1rh-
4.45) x10
-1
to-7.69
n1 (x10
1
) = (19.3rh
2
-23.2rh+6.98)x10
-3
to
2
+(-5.46rh
2
+6.01rh-1.78)x10
-1
to+(-1.33rh
2
-
1.43rh+3.54)
D1 = (28.2rh+5.63)x10
-3
to
2
+(-60.8rh-4.23)x10
-2
to+(-4.44rh
2
+5.67rh+1.72)
αC0,1 (x10
-1
) = (16.9rh
2
-16.7rh+3.96)x10
-3
to
2
+(-4.49rh
2
+4.07rh-1.0)x10
-1
to-4.68
αCT,1=1.07x10
-1

A2 (x10
4
s
-1
) = (12.5rh
2
-14.8rh+3.14)x10
-2
to
2
+(-4.26rh
2
+5.11rh-1.17)to+(5.56rh
2
-
7.0rh+13.7)
E A2/R (x10
2
K) = (4.80rh
2
-5.51rh+1.58)x10
-2
to
2
+(-9.24rh
2
+10.3rh-3.76)x10
-
1
to+(2.39rh
2
-2.85rh-8.8)
m2 (x10
1
) = (4.61rh
2
-5.37rh+1.10)x10
-2
to
2
+(-15.3rh
2
+17.6rh-3.96)x10
-1
to+7.38
n2 = (11.7rh
2
-14.0rh+3.10)x10
-3
to
2
+(-28.1rh
2
+38.2rh-8.63)x10
-2
to+1.94
D2 = (16.3rh
2.
-18.3rh+5.23)x10
-2
to
2
+(7.82rh-9.94)x10
-1
to+(-47.8rh
2
+56.0rh+43.7)
αC0,2 (x10
5
) = (-9.06rh
2
-10.3rh+1.03)x10
-2
to
2
+(1.17rh
2
-1.12rh-1.16)x10
-1
to+(-
2.61rh
2
+3.28rh-9.13)
αCT,2 (x10
3
) = (-4.97rh-4.17)x10
-3
to+1.77

114

Table 6-3. Parameters for cure kinetics model for 44DDS (a/e = 0.8)
A1 (x10
3
s
-1
) = (-3.25rh+1.13)x10
-2
to
2
+(9.60rh
2
-9.30rh+2.16)to+(4.0rh+15.3)
E A1/R (x10
-2
K) = (-28.8rh+8.20)x10
-2
to+(-5.0rh
2
+5.17rh-14.4)
m1 (x10
1
) = (8.66rh
2
-11.9rh+3.71)x10
-2
to
2
+(5.55rh-3.35)x10
-1
to+(-3.22rh
2
+3.53rh-
8.12)
n1 (x10
2
) = (3.85rh
2
-4.86rh+1.34)x10
-1
to
2
+(-8.72rh
2
+10.4rh-
2.64)to+(7.78rh
2
8.33rh+47.5)
D1 (x10
2
) = (13.5rh-3.43)x10
-2
to
2
+(5.64rh-2.18)to+(-8.33rh+136)
αC0,1 (x10
-1
) = (1.85rh
2
-2.45rh+1.30)x10
-3
to
2
+(8.07rh
2
-9.47rh+1.88)x10
-1
to-4.85
αCT,1=1.04x10
-1
A2 (x10
4
s
-1
) = (10.5rh
2
-10.4rh+1.31)x10
-3
to
3
+(-45.8rh
2
+43.0rh-4.03)x10
-
2
to
2
+(402rh
2
-334rh-2.60)x10
-2
to+6.81
E A2/R (x10
2
K) = (-42.6rh
2
+37.9rh-7.56)x10
-2
to
2
+(-19.0rh
2
+17.2rh-3.44)to+(-
3.89rh
2
+3.43rh-10.0)
m2 (x10
1
) = (-11.5rh+9.58)x10
-3
to
2
+(27.7rh
2
-31.7rh+5.16)x10
-1
to+(3.44rh
2
-
3.30rh+7.82)
n2 (x10
1
) = (-3.09rh+2.23)x10
-2
to
2
+(8.6rh-5.25)x10
-1
to+(6.11rh
2
-6.17rh+17.7)
D2 = (8.89rh
2.
-11.5rh+3.48)x10
-1
to
2
+(-3.24rh
2.
+4.19rh-1.27)x10
1
to+(4.33rh+52.7)
αC0,2 (x10
5
) = (-17.5rh+7.03)x10
-2
to
2
+(7.34rh-2.94)to+(3.53rh-9.14)
αCT,2 (x10
3
) = (-4.60rh-4.19)x10
-3
to+1.77

115

Table 6-4. Parameters for initial degree of cure (α0) input to cure kinetics model and
viscosity model (where α0,a = actual initial degree of cure (used as an input to viscosity
model) and α0,f = fixed initial degree of cure (= 0.0015, used as an input to cure kinetics
model))
α0,a (33DDS (a/e = 0.6)) = α0,f +(-75.4rh
2
+3.24rh+52.0)x10
-
6
to
2
+(41.3rh
2.
+13.8rh-1.78)x10
-4
to
α0,a (33DDS (a/e = 0.8)) = α0,f +(1.12rh
2
-2.39r+1.11)x10
-4
to
2
+(-1.74rh
2.
+9.37rh-
1.67)x10
-3
to
α0,a (44DDS (a/e = 0.8)) = α0,f +(-31.4rh
2
+30.0rh-3.68)x10
-
5
to
2
+(15.4rh
2.
+15.0rh+2.61)x10
-4
to

116

6.3.2. Viscosity model
The viscosity (η) evolution of epoxy resin during processing is affected by two
competing effects: temperature and cure. Heating the resin decreases viscosity due to
increased molecular mobility, but it also induces cure which increases viscosity due to
increase molecular size. To capture these phenomena, along with out-time and moisture
absorption effects on η evolution during cure, a phenomenological model developed by
Khoun et al. [139] was adapted and modified to yield the following expression:
1
12
1
()


  




de
A B C
(6-4)
1,2
exp








i
i
i
i
E
A
RT
(6-5)
where ηi is the Arrhenius dependent viscosity component, Aηi is the Arrhenius constant, Eηi
is the viscosity activation energy, α1 is the degree of cure at gelation, and A, B, C, d and e
are fitting constants. The first term in equation (6-4) takes an Arrhenius-type form, and the
second term is added to account for the rapid viscosity increase near the gelation point.
Here, the variables Aηi, Eηi, α1, A, B, C, d and e were defined as f(r, to) = g(r)to
2
+ h(r)to +
i(r) to account for changes associated with ambient temperature cure as well as moisture
absorption. The model parameters are provided in Table 6-5, 6-6 and 6-7.

117

Table 6-5. Parameters for viscosity model for 33DDS (a/e = 0.6) (where rh = rh in fraction
and to = out-time in days)
Aƞ1 (x10
2
Pa· s) = (-2.07rh-4.28)x10
-1
to+(-8.67rh+25.1)
Eƞ1/R (x10
2
K) = (48.3rh+60.0)x10
-2
to+(3.83rh
2
-4.08rh+122)
Aƞ2 (Pa· s) = (10.1rh-2.48)x10
-1
to
3
+(-43.7rh+8.05)to
2
+(400rh+13.1)to+326
Eƞ2/R (x10
1
K) = (-22.8rh
2
+28.8rh-7.81)x10
-3
to
2
+(5.83rh
2
-7.14rh+1.68)x10
-1
to+8.47
α1 (x10
1
) = (-24.2rh
2
+25.4rh-5.01)x10
-3
to
2
+(7.17rh
2
-7.04rh+1.39)x10
-1
to+4.53
A = (22.7rh-9.07)x10
-3
to
2
+(-9.55rh+3.82)x10
-1
to+14.2
B = (14.7rh
2
-17.1rh+3.80)x10
-1
to
2
+(-30.0rh
2
+33.9rh-7.63)to+(20.0rh
2
-23.7rh+60.7)

C = (-16.2rh
2
+19.1rh-4.44)x10
-1
to
2
+(34.6rh
2
-39.6rh+9.49)to+(-25.6rh
2
+31.7rh-
63.8)
d (x10
2
) = (2.10rh-3.30)x10
-2
to
2
+(-2.73rh+8.10)x10
-1
to+(1.72rh-7.51)

e (x10
5
) = (12.0rh-3.43)x10
-2
to
2
+(-5.11rh+1.48)to+(-1.67rh+137)

118

Table 6-6. Parameters for viscosity model for 33DDS (a/e = 0.8) (where rh = rh in fraction
and to = out-time in days)
Aƞ1 (x10
2
Pa· s) = (-14.3rh+5.26)x10
-2
to
2
+(4.29rh-1.15)to+(-11.8rh
2
+15.1rh-4.10)
Eƞ1/R (K) = (-7.10rh
2
+8.62rh-1.95)x10
-2
to
2
+(21.3rh
2
-25.8rh+5.77)x10
-1
to+1.41
Aƞ2 (x10
-1
Pa· s) = (-1.78rh
2
+1.97rh+4.63)to
2
+(4.44rh
2
-5.0rh-20.1)x10
1
to+1.93x10
3

Eƞ2/R (x10
1
K) = (15.0rh
2
-18.0rh+3.96)x10
-2
to
2
+(-4.19rh
2
+5.06rh-1.07)to+(-
1.78rh
2
+2.53rh+6.52)
α1 (x10
2
) = (-12.2rh+6.98)x10
-2
to
2
+(4.14rh-2.16)to+(-2.50rh+60.3)
A = (-9.94rh+2.16)x10
-3
to
2
+(29.3rh-5.17)x10
-2
to+14.1
B = (3.14rh-2.62)x10
-1
to
2
+(-11.4rh-10.0)to+(-1250rh
2
+1700rh-1.0)x10
-1
C = (-3.59rh+2.84)x10
-1
to
2
+(12.6rh-10.6)to+(124rh
2
-171rh+4.0)
d (x10
4
) = (-4.50rh
2
+5.58rh-2.46)to+(-6.83rh+18.1)

e (x10
2
) = (5.09rh-1.87)x10
-3
to
2
+(-15.1rh+2.76)x10
-2
to+(1.67rh
2
-2.17rh+1.71)

119

Table 6-7. Parameters for viscosity model for 44DDS (a/e = 0.8) (where rh = rh in fraction
and to = out-time in days)
Aƞ1 (x10
3
Pa· s) = (2.80rh+6.12)x10
-1
to
2
+(-5.67rh-28.5)to+(6.67rh+286)
Eƞ1/R (x10
2
K) = (-69.7rh+3.60)x10
-3
to
2
+(16.0rh+6.07)x10
-1
to+(-3.33rh+125)
Aƞ2 (x10
-2
Pa· s) = (-7.22rh
2
+9.50rh+3.84)to
2
+(2.06rh
2
-2.75rh-
2.0)x10
2
to+(5.0rh+262)x10
1

Eƞ2/R (x10
2
K) = (5.38rh
2
-6.45rh+1.45)x10
-1
to
2
+(-12.3rh
2
+16.2rh-3.26)to+(-
3.89rh
2
+5.17rh+64.0)
α1 (x10
2
) = (1.91rh-1.03)to+(-6.83rh+66.7)
A = (-7.29rh+1.84)x10
-3
to
2
+(16.4rh+1.87)x10
-2
to+13.0
B = (-14.4rh
2
+19.8rh-6.59)x10
-1
to
2
+(45.3rh
2
-60.0rh+19.3)to+(7.33rh+59.3)

C = (13.6rh
2
-18.8rh+6.27)x10
-1
to
2
+(-42.1rh
2
+56.0rh-18.3)to+(-8.0rh-54.3)
d (x10
3
) = (-2.17rh
2
+3.08rh-5.13)x10
-3
to
2
+(7.22rh
2
-9.83rh+20.3)x10
-2
to-1.61

e (x10
3
) = (16.4rh
2
-22.0rh+2.05)x10
-1
to
2
+(-511rh
2
+657rh+8.0)x10
-1
to+(-5.0rh-159)

6.4. Results and Discussion

6.4.1. Out-time characterization
Previous studies have shown that out-time or aging-induced cross-linking of the
epoxy resin causes α0 to increase, and the moisture absorbed during out-time can affect the
reaction by acting as both catalyst and solvent [132-134]. To verify this, TGA was used to
120

monitor mass stability during heating and results were measured. Figure 6-2(a) compares the
weight percent of absorbed water change with out-time for resin systems with the same cure
agent (33DDS) but different a/e stoichiometric ratio. The data show that the water uptake
increases with rh and with higher a/e stoichiometric ratio.
Figure 6-2(b) shows the water uptake as a function of out-time for formulations with
different isomeric cure agents but identical a/e stoichiometric ratios. The data show that water
absorption is a strong function of the amount of available amine groups in the resin. The
resins exhibit nearly the same water absorption level across all out-time values and rh
conditioning. Thus, combined, the results in Figures 6-2 indicate that amine has greater
affinity towards the hydroxyl group in water than to the epoxy group. Furthermore, water
absorption is a weak function of the ‘ambient temperature-induced’ primary degree (1⁰)
amine-epoxy reaction (or α0), which is expected to progress until the resin Tg approaches the
ambient temperature, where it vitrifies. At equilibrium, for resins with a/e = 0.8, roughly 1.1
wt% water was absorbed for a rh of 90%, while only 0.2 wt% was absorbed for the resin with
a/e = 0.6. Thus, at a molecular level, we expect that samples conditioned at higher humidity
levels will exhibit greater changes in α0 associated with the catalytic and solvent effect.

121

Figure 6-2. Weight percent of absorbed water (H2O) change on out-time and rh for: (a)
33DDS (a/e = 0.8) and 33DDS (a/e = 0.6); and (b) 33DDS (a/e = 0.8) and 44DDS (a/e =
0.8)

∆HT of the resin was determined by integrating the exothermic heat flow evolution
from DSC heating measurements where the resin was fully cured. The α0 was calculated
using the equation:
0
( 0) ( )
( 0)
TT
T
H day H out time
H day

   


(6-6)
where ∆HT is expected to decrease with out-time due to the decrease in the number of
reactants. Values of α0 are plotted in Figure 6-3. For the convenience of modelling cure
122

kinetics and viscosity evolution during cure, the α0 change on out-time can be fitted as:
0, 0, 0,
( , )
a f v o
t rh    
(6-7)
where α0,a is actual initial degree of cure (used as an input to the η model), α0,f is the fixed
initial degree of cure (= 0.0015, used as an input to the cure kinetics model), and α0,v is
variable degree of cure. The model parameters determined are provided in Table 6-4. The
results show that for all resin types, α0 increases with out-time in a predictable manner. Resin
conditioned at higher rh levels exhibited a sharper increase in α0, as absorbed water facilitated
cross-linking at the ambient temperature.
Figure 6-3(a) shows that 33DDS with higher a/e stoichiometric ratio ages more
rapidly, presumably due to higher collision numbers between available amine and epoxy
group. Resins with the same a/e stoichiometric ratio but with different isomeric cure agents
are compared in Figure 6-3(b), and the data show that 33DDS (with meta substitution)-based
resins age substantially more rapidly than the 44DDS (with para substitution)-based resins.
The accelerated aging occurs because the reactivity of the amine depends upon the
nucleophilicity of the amino group [129]. Both 33DDS and 44DDS isomers have the same
electron withdrawing sulphonyl group, where the only difference comes from the orientation
of the NH2 groups. Due to para-substitution, 44DDS has delocalization of the lone pairs of
electronson nitrogen but such resonance is not possible in meta-substitution, 33DDS.
Consequently, 33DDS is more reactive with epoxide groups than 44DDS.

123

Figure 6-3. Initial degree of cure (α0,a) change on out-time and rh for: (a) 33DDS (a/e = 0.8)
and 33DDS (a/e = 0.6); and (b) 33DDS (a/e = 0.8) and 44DDS (a/e = 0.8)

The effects of absorbed moisture on the B-stage or initial glass transition temperature
(Tg,0) during out-time were determined from MDSC data and are displayed in Figure 6-4.
Measurement of Tg,0 requires moderate heating of the sample, usually to slightly above
ambient temperature, making it useful for tracking out-time. Figure 6-4(a) shows that nearly
the same Tg,0 values are obtained for 33DDS-based resins conditioned at the same rh but with
different a/e stoichiometric ratios, except at day 30. The resin with higher α0 is expected to
have higher Tg,0. The results manifest the presence of two competing effects, where the
solvent effect of water decreases Tg,0, while the catalytic effect of water together with the
higher a/e stoichiometric ratio provide higher collision numbers, thus increasing Tg,0. These
124

competing effects are also apparent in Figure 4(b), where the 44DDS based resin aged at 60
& 90% rh exhibits a decrease in Tg,0 at day 10, despite progression towards a higher α.
When the 33DDS-based resin is fully cured, it will have greater configurational
entropy than the 44DDS-based resins, making the resin more flexible and resulting in lower
Tg. Out-time and rh conditioning effects here show that 44DDS-based resins exhibit higher
Tg,0 than the 33DDS-based resins until day 20, despite 44DDS based resin having lower α0
(see Figure 6-3(b)). With continued aging, the result reverses for day 30 samples conditioned
at rh = 60 & 90%, where α0 further increases quadratically and more rapidly against out-time
for the 33DDS-based resin relative to the 44DDS-based resin. Thus, tracking evolution of
Tg,0 may not be a reliable way to track α0 for a resin exposed to variable environmental
conditions. Overall, the combination of out-time and moisture absorption leads to permanent
changes in the resin state prior to cure, and this is expected to affect the course of cure kinetics
and viscosity evolution during cure.

125

Figure 6-4. Initial glass transition temperature (Tg,0) change on out-time and rh for: (a)
33DDS (a/e = 0.8) and 33DDS (a/e = 0.6); and (b) 33DDS (a/e = 0.8) and 44DDS (a/e =
0.8)

6.4.2. Cure kinetics evolution and modelling
Figure 6-5 shows representative cure kinetics measurements and the corresponding
predictive model obtained using equation (6-2). Note that α is an integration of cure rate over
time, and thus only the cure rate results are presented here. Cure kinetics data and the model
results for fresh resin and resin conditioned at rh = 90% for 30 days provide two extremities
in terms of the cure kinetics change/shift. Thus, these quantities have been selected and are
plotted in Figure 6-5, and the parameters for the cure kinetics models are shown in Table 6-
1, 6-2 and 6-3. The cure kinetics model for each resin system was first developed using
126

isothermal dwell data at 121 and 150 °C, and dynamic ramp data with fresh resin samples.
Subsequently, isothermal dwells at 150 °C and dynamic ramps were conducted on all three
resin systems conditioned at RH = 30, 60 and 90% from 0 to 30 days. These conditions were
selected to take into account the effects of moisture absorption and out-time during cure, and
to determine parameters in the cure kinetics equation. Confining the parameters to predictable
patterns as functions of out-time and rh, the present cure kinetics model captured out-time
and humidity effects accurately over the entire range of conditions studied.
Out-time induced cross-linking causes α0 to increase, and absorbed moisture can
further catalyze this reaction. In general, the thermoset resin cure can be perceived as a blend
of fast- and slow-occurring reactions, which are kinetics-driven and diffusion-driven,
respectively. Thus, out-time, where the reaction is induced at ambient temperature, is
expected to mainly affect kinetically driven reactions. Overall, time-based cure rate shifts are
more apparent during isothermal cure than during dynamic ramp conditions, as the reaction
temperature is relatively high. On the other hand, magnitude-based cure rate shifts with out-
time and rh conditioning are more apparent under dynamic ramp conditions. Both time- and
magnitude-based shifts are greater with higher a/e stoichiometric ratios within the same cure
agent, and weaker with para-substituted 44DDS-based resin, as it exhibits slower cure rates
than the meta-substituted 33DDS-based resin.
127

Figure 6-5. Representative cure kinetics measurement and model prediction of isothermal
dwell and dynamic ramp.

6.4.3. Viscosity evolution and modelling
Figure 6-6 shows representative η measurements and the corresponding predictive
model obtained using equation (4). As with the development of the cure kinetics model, fresh
resin samples were used to generate isothermal dwell data at 121 and 150 °C, and these data
were then combined with dynamic ramp data to develop a benchmark viscosity model.
Subsequently, resin samples conditioned at rh = 30, 60 and 90% from 0 to 30 days were used
128

to account for the effects of moisture absorption and out-time during cure and to determine
parameters in the η equation. Parameters for η models are shown in Tables 6-5, 6-6 and 6-7.
Figure 6-6 shows that the viscosity model captured out-time and humidity effects accurately
over the entire range of conditions studied.
Generally, η evolution for a cure cycle is governed by the competing effects of cure
(which increases η) and heating (which decreases η). Initially, as the temperature increases
and the resin degree of cure is low, the thermal effect is predominant, and the η decreases. As
the α begins to accelerate and the temperature approaches the isothermal dwell, the process
becomes cure-driven, and the η increases at an increasing rate until gelation occurs, where
the resin flow decreases drastically. Such effects are apparent in Figure 6-6. In addition, the
η also increases with ambient exposure, moisture absorption, and cure. The phenomena are
especially important because viscosity directly affects resin flow, and the deviation from
Newtonian behavior is expected near the gel point (G’ = G” [140]), or the flow stop point. In
other words, the η levels and resin flow times required to completely wet fiber tows during
prepreg processing are progressively limited by ambient temperature, moisture absorption,
and cure. Therefore, the increase in η due to ambient exposure and moisture absorption is the
primary factor that determines resin out-life, as specified by resin manufacturers.
Comparing the same 33DDS-based resin with different a/e stoichiometric ratios, a
lower a/e stoichiometric ratio offers a lower η profile when subjected to the same cure cycle,
while the 44DDS-based resin exhibits an even lower η profile. This behavior arises because
under the same cure cycle, cure progresses most rapidly in 33DDS with a/e = 0.8, and most
129

slowly in 44DDS with a/e = 0.8. Furthermore, this effect is more pronounced under dynamic
ramps. Therefore, in designing epoxy resins for prepreg, consideration must include the effect
of reaction speed on flow level and flow duration, as these are critical to full impregnation
during processing, as well as the effects of out-time and rh conditioning. As demonstrated in
a previous study [132], an accurate model for viscosity allows one to develop a viscosity-
controlled (or flow-enhanced) cure cycle that can in principle extend out-life and limit flow-
induced defects, while also minimizing cycle time for prepreg.

130

Figure 6-6. Representative viscosity (η) measurement and model prediction of isothermal
dwell and dynamic ramp.

6.4.4. Gelation point
For epoxies, gelation is often defined as the point where G’ and G” intersect [140]
(Figure 6-7(a)). At gelation, η increases quickly, cure effects dominate over thermal effects,
and deviation from Newtonian behavior is expected [141]. Thus, gelation is also the flow
stop point. In other words, the viscosity levels and resin flow times required to completely
131

wet fiber tows during processing are progressively limited by ambient temperature, moisture
absorption, and cure. Therefore, the increase in viscosity due to ambient exposure and
moisture absorption is the primary factor that determines the out-time limit or out-life for
prepreg.
As expected, Figures 6-7(b) & (c) show that for all resin systems, tgel or ‘resin flow
time’ under similar cure conditions decreases with both out-time and higher rh conditioning.
The 33DDS resin with a/e = 0.6 exhibits higher tgel than 33DDS with a/e = 0.8, as the reaction
is slower. Comparing para-substituted 44DDS resin and meta-substituted 33DDS resin, the
reaction of the 44DDS resin is substantially slower than the 33DDS resin and attains a lower
α0 during out-time and humidity conditioning. As a result, the 44DDS resin exhibits a tgel
nearly twice that of the 33DDS resin. Thus, from the perspective of a resin formulator,
44DDS and lower a/e stoichiometric ratios offer longer tgel and thus longer resin flow time.
The epoxy resin flow stops at the gelation point, yet the resin may still continue to react.

132

Figure 6-7. Gelation time (tgel) measurement method and change under isothermal dwell.
(a) G’ and G” versus cure time for 33DDS resin with a/e = 0.8 (b) tgel versus out-time for
33DDS resin with a/e = 0.6 & 0.8 (c) tgel versus out-time for 33DDS resin and 44DDS resin
with a/e = 0.8

6.4.5. Glass transition temperature
Glass transition or vitrification point is determined using MDSC from the inflection
133

point of the heat capacity (Cp) during the isothermal dwell period [141]. The results, tabulated
in Table 6-8, show that for a given resin system, during isothermal dwells, where
etherification reaction is unlikely to occur, higher a/e stoichiometric ratios resulted in higher
Tg values, as these formulations have more amine-epoxy bonds. However, due to
etherification at higher temperature, fully cured Tg (or Tg, ꝏ) reached nearly the same level.
Comparing para-substituted 44DDS resin and meta-substituted 33DDS resin, the Tg of
33DDS-based resin was lower, because 33DDS resin has greater configurational entropy,
making the resin more flexible and yielding lower Tg.

Table 6-8. Glass transition temperature (Tg) for all resin systems where Tg,121C and Tg,150C
are Tg for resins cured at isothermal dwell of 121 ⁰C and 150 ⁰C respectively and Tg, ꝏ is Tg
for fully cured resin.

33DDS (a/e = 0.8) 33DDS (a/e = 0.6) 44DDS (a/e = 0.8)
Tg,121C (°C) 145.9 139.9 148.3
Tg,150C (°C) 182.7 175.5 191.6
Tg, ꝏ (°C) 216.4 215.8 236.8

6.4.6. Resin flow control
The cure kinetics measurements, η measurements, and predictive models developed
in this study show that the influence of out-time and absorbed moisture on the resin viscosity
134

can be substantial. This, in turn, can cause incomplete impregnation of resin on fiber beds
during prepreg processing. However, with the predictive η model developed here, the
temperature cycle can be tuned to potentially develop a flow-enhanced or viscosity-
controlled cure cycle that can limit flow induced defects.
A squeezing flow geometry have been widely investigated for characterizing resin
flow during lamination [123, 142-143]. Here, the prepreg layup is sandwiched between
stainless steel plates into the Instron test frame, where a linear variable differential
transformer is used to measure the movement of the cross heads. While the plates are heated,
a constant force is applied and measured by the load cell. Assuming that the resin flow can
be approximated as Newtonian before it gels, a characteristic flow quantity called flow
number (NFL) has been defined as follows:
0.5
2
1
4
0
16
1 ( ) 1 100
3
gel
t
p
o
Fl
o
Fh
N t dt
R






 

   
 


 


(6-7)
where ρp is the resin density, ρo is the prepreg density, F is the lamination press force, h0 is
the initial stack thickness, R is the effective radius of the resin. The main variable controlling
the flow quantity is:
1
.
0
()
gel
t
Fl eff
N t dt 



(6-8)
where NFL,eff is the effective flow number, and higher NFL,eff means more resin flow. With the
predictive η model developed in this study, there are two controlling variables that affect η
135

evolution: heating rate and dwell temperature. Figure 6-8 compares η evolution at various
dwell temperatures and heating rates. The η evolution during cure is primarily affected by
cure temperature and α, where an increase in temperature leads to decrease in η, yet it also
accelerates cure, which increases η. Additionally, higher heating rate leads to a more rapid
decrease in η, allowing the resin to achieve lower η before the substantial effect of cure kicks
in. However, once the resin reaches cure temperature, the rapid cure effect is expected, thus
leading to reduction in resin flow time.

Figure 6-8. η model prediction for 33DDS (a/e = 0.8, rh 90%) at day 30: (a) fixed ramp rate
to incremental dwell temperatures and (b) incremental ramp rate to the same dwell
temperature
136

These competing effects are evident in Figure 6-8, where η model prediction for
33DDS (a/e = 0.8, rh 90%) at day 30 were subjected to various heating rates and dwell
temperatures. The results show that a higher dwell temperature leads to lower η, yet the faster
cure effect leads to a more rapid η increase (Figure 6-8(a)). Also, a more rapid heating rate is
shown to lead to lower η, although once the temperature reaches the dwell temperature, which
occurs at around the ηmin, η is shown to increase at the same rate regardless of thermal history.
Figure 6-9 depicts NFL,eff calculated using equation (6-8) at various cure conditions. The
purpose of this analysis was not to find a maximum NFL,eff , although it is apparent that the
predictive η model allows design of a ‘flow-enhanced’ cure cycle that is specific to the
particular resin system, aging, and thermal history.

Figure 6-9. Effective flow number (NFL,eff) at various cure conditions for 33DDS (a/e = 0.8,
rh 90%) at day 30

137

6.5. Conclusions
Three aerospace grade resins were formulated to investigate the effects of variation
in DDS isomers, a/e stoichiometric ratio, out-time, and moisture absorption on processing
characteristics and cured resin properties. First, conventional thermochemical and
thermomechanical methods (MDSC and rheometry) were used to collect benchmark data.
Regardless of the resin system, the results show that out-time increases initial degree of cure
and more so with moisture absorption which influence the cure kinetics and viscosity
evolution during cure. These effects, in turn, will shorten flow level and time which could
potentially lead to insufficient resin flow during composite manufacturing. Therefore,
limiting out-time and including humidity and thermal controls are generally required to
ensure sufficient resin flow.
Then, accurate process models were developed that comprehensively capture out-
time and humidity effects on cure kinetics and viscosity for each resin systems. Among the
three resin systems investigated in this study, meta- substituted 33DDS with a/e = 0.8
exhibited the highest cure rate and the lowest flow time. This behavior was attributed to the
increase in collision number between amine and epoxy compared to a/e = 0.6 and the lack of
delocalization of the lone pair of electrons on nitrogen compared to para- substituted 44DDS.
Conversely, 44DDS with a/e = 0.8 exhibited the slowest cure rate with the highest flow time.
In addition, 44DDS based resin exhibited the highest glass transition temperature due to
having lower configurational entropy than 33DDS based resin. The results offer practical
insights regarding resin formulation and prepreg processing, particularly that: (1) 33DDS-
based resin is more susceptible to aging and flow time reduction, albeit with faster processing
138

time; (2) lower a/e will require longer dwell at higher temperature to complete the
etherification reaction, albeit with more flow time; and (3) 44DDS resin offers longer flow
time with higher glass transition temperature, yet with substantially longer processing time.
Finally, the predictive viscosity model demonstrates a method to enhance resin flow during
prepreg processing which can potentially limit flow-induced defects.
Overall, this chapter provides foundational knowledge for aerospace grade DDS
isomer cured resins subjected to pre-cure conditions that are largely unavoidable in practice
with prepreg processing. The predictive models developed here can be used to enhance resin
flow. The models combined with the experimental results, they potentially allow adaptive
manufacturing of quality parts with non-ideal material and process conditions.

139

Chapter 7. Effective Cure Cycle Development via Flow Optimization and Advanced
Cure Environments

7.1. Introduction
Out-time is unavoidable in industrial settings, as the lay-up and preparation of large
structures often takes from several days to weeks. In chapter 3, it was shown that out-time
causes ambient temperature induced cross-linking of the resin. Predictive models that
comprehensively captured out-time effects on cure kinetics and viscosity in process
conditions were developed. In chapter 5, it was shown that moisture absorption also affects
process-relevant properties, such as degree of cure and viscosity, by acting as both catalyst
and solvent for the amine-epoxy reaction and accelerating cross-linking. Then, efficient
models for cure kinetics and viscosity that capture out-time and moisture absorption effects
during consolidation were developed.
Given adequate vacuum bagging and level (28 in Hg), there are two critical
components that represent lamination process which are viscosity and thickness change
during cure (Figure 7-1). Of the two, viscosity profile is controlled by the resin chemistry,
prepreg pre-conditions (i.e. out-time and humidity conditioning) and the cure profile. In other
words, with given pre-conditions, flow window can be tailored with chemistry of resin (i.e.
cure agent concentration and type) and temperature profile control.
140

Figure 7-1. Block diagram of the lamination process

The cure kinetics and viscosity measurements, and predictive models developed in
this thesis show that the influence of out-time and absorbed moisture on the resin viscosity
can be substantial. This, in turn, can cause incomplete impregnation of resin on fiber beds
during prepreg processing. Previous studies have also shown that the out-time increases the
resin viscosity and decreases its gelation time, thereby reducing the flow time for full fiber
tow impregnation [Chapter 3, 4]. Also, prepreg tack and compliance (or inverse of stiffness)
were shown to decrease with out-time which make ply collation difficult thus inhibiting ply
nesting. Combined, reduction in flow leads to increase in micro-porosity yet incomplete ply
nesting leads to decrease in macro-porosity.
Therefore, with the predictive viscosity models developed in Chapter 3 and 5, this
study seeks to tune the temperature cycle to potentially develop a flow-enhanced or viscosity-
141

controlled cure cycle that can limit flow induced defects.

7.2. Experimental

7.2.1. Materials
For this study, a commercially available OoA prepreg consisting of a unidirectional
carbon fiber tape and a toughened epoxy resin (CYCOM
®
5320-1/IM7, Cytec Industries Inc.)
was used. All laminates had [0°]6s layup sequence. The unidirectional fiber bed has an areal
weight of 145 g/m
2
. The prepreg has a resin content of 33% by weight, and a manufacturer’s
specified out-life of 28 days [116]. All samples were stored in humidity chambers created
with saturated salt solutions, where accurate control of equilibrium vapor pressure is possible
[117], with relative humidity (RH) levels of 60% for 0 to 56 days and tested by 7 days. For
laminate manufacture, each cycle included an initial room temperature vacuum hold of
twelve hours to ensure that air was adequately extracted from the laminate. All laminates
were cured well beyond gelation point (η = 10 kPa· s) to ensure complete resin flow under
tested cure cycle.

7.2.2. Heated tool
The heated tool was designed and manufactured in-house to enable rapid heating up
142

to 35°C/min and accurate temperature control between 30°C and 200°C [144]. The tool
consists of six independent cells in a 10 cm X 10 cm matrix (Figure 7-2). Each cell is
composed of a heat sink, four 300 W cartridge heaters, and a cooling fan. Two thermocouples
are also embedded under flat aluminum tool plate connecting all six cells for temperature
measurement and control. Each cell’s heating and cooling are independently controlled via
solid state relays (DigiKey DRA1-MCX240D5) by a digital acquisition and control system
(National Instruments cRIO) and a virtual instrument (National Instruments LabView).

Figure 7-2. In-house developed heated tool [144]

7.2.3. Image analysis
Quality inspection of cured laminates (i.e. void area % calculation) were conducted
using light microscopy on the polished cross-sections. First, cured laminates were sectioned
and polished on a metallographic polisher (Buehler MetaServ) up to 4000 grids. A video
143

microscope (Keyence VHX-600) at both 100X & 200X magnifications was used to acquire
cross-sectional images of cured laminates. V oid area % were measured and calculated using
image analysis programs, Photoshop and ImageJ respectively.

7.3. Theoretical Background

7.3.1. Viscosity model
The viscosity of curing epoxy resin is affected by two competing effects. Heating
the resin increases molecular mobility, thereby decreasing viscosity, but also eventually
increases molecular size, increasing viscosity. To capture these phenomena, a
phenomenological model developed by Khoun et al. [40] was adapted and modified to
capture out-time and moisture absorption effects:
1
12
1
()


  




de
A B C
(7-1)
1,2
exp








i
i
i
i
E
A
RT
(7-2)
where ηi is the Arrhenius dependent viscosity component, Aηi is the Arrhenius constant, Eηi
is the viscosity activation energy, α1 is the degree of cure at gelation, and A, B, C, d and e
are fitting constants. The first term from equation (7-1) takes an Arrhenius type form, and
144

the second term is added to account for the rapid viscosity increase near gelation point.
Here, the variables Aηi, Eηi, α1, B, and d were defined as f(r, to) = g(r)to
2
+ h(r)to + i(r) to
account for changes associated with ambient temperature cure as well as moisture
absorption. In addition, the current model focuses on viscosity before gelation, at which
point deviations from Newtonian behavior are expected (gelation can be taken as ƞ ~ 10
kPa· s [10] or σ inflection point [10]) and flow drastically reduces. The model parameters
are provided in Table 7-1.

145

Table 7-1. Parameters for viscosity & conductivity models (where r = RH in fraction and to
= out-time in days)
Aƞ1 (x10
-3
Pa· s) = (5.11r
2
-4.12r+2.23)to+(20.04r
2
-25.17r+8.46)
Aƞ2 (x10-
14
Pa· s) = (7.67x10
-4
)to
2
-(3.84x10
-2
)to+(-4.78x10
-2
r
2
-2.73x10
-2
r+0.5)
Eƞ1/R (x10
3
K) = (9.4r
2
-9.8r+4.1)x10
-3
to
2
+(-0.21r
2
+0.2r-0.14)to+(-
2.65r
2
+3.12r+1.97)
Eƞ2/R (x10
4
K) = (1.2r
2
-1.4r+0.2)x10
-3
to
2
+(-2.61r
2
+2.85r-0.23)x10
-2
to+(8.3x10
-
3
r
2
+8.5x10
-3
r+1.26)
α1 = (5.0x10
-3
)to
2
-(3.5x10
-2
r-1.67x10
-2
)to+(-8.93x10
-2
r+0.71)
B = (4.57x10
-2
r-8.9x10
-3
)to
2
+(-0.36r+0.29)to+(-1.38r+11.66)
d = (3.0x10
-3
)to
2
+(2.7r-6.7)x10
-3
to+(-3x10
-3
r+0.11)
A = 12.37 C = -0.43 e = -1.9x10
-3

L1 (x10
-4
S/m) = (-8.0r
2
+16.0r-6.0)x10
-4
to
2
+(4.11r
2
-6.3r+2.19)x10
-2
to+6.22x10
-2

L2 = (5.0r
2
-5.5r+1.1)x10
-3
to
2
+(-0.15r
2
+0.15r-2.97x10
-2
)to-0.79
L3 = (-7.3r
2
+8.2r-1.4)x10
-3
to
2
+(2.24r
2
-2.22r+0.37)x10
-1
to+0.68
L4 (x10
6
S/m) = (-0.6r
2
+1.2r+1.9)x10
-3
to
2
+(0.5r
2
-2.35r+12.24)x10
-2
to+1.83
L5 (x10
3
K) = (-1.67r
2
+1.2r+0.13)x10
-2
to
2
+(0.85r
2
-0.65r+0.14)to+1.71

7.3.2. Flow model
A squeezing flow geometry have been widely investigated for characterizing resin
flow during lamination [123, 142-143]. Here, the prepreg layup is sandwiched between
146

stainless steel plates into the Instron test frame, where a linear variable differential
transformer (LVDT) is used to measure the movement of the cross heads. While the plates
are heated, a constant force is applied and measured by the load cell. Assuming that the resin
flow can be approximated as Newtonian before it gels, a characteristic flow quantity called
flow number (NFL) has been defined as follows:
0.5
2
1
4
0
16
1 ( ) 1 100
3
gel
t
p
o
Fl
o
Fh
N t dt
R






 

   
 


 


(7-3)
where ρp is the resin density, ρo is the prepreg density, F is the lamination press force, h0 is
the initial stack thickness, R is the effective radius of the resin. The main variable controlling
the flow quantity is:
1
.
0
()
gel
t
Fl eff
N t dt 



(7-4)
where NFL,eff is the effective flow number, and higher NFL,eff means more resin flow. With the
predictive η model developed in this study, there are two controlling variables that affect η
evolution: heating rate and dwell temperature. The η evolution during cure is primarily
affected by cure temperature and α, where an increase in temperature leads to decrease in η,
yet it also accelerates cure, which increases η. Additionally, higher heating rate leads to a
more rapid decrease in η, allowing the resin to achieve lower η before the substantial effect
of cure kicks in. However, once the resin reaches cure temperature, the rapid cure effect is
expected, thus leading to reduction in resin flow time.
147

7.4. Results and Discussion

7.4.1. Flow modelling
Representative model η evolution curve under 2 ⁰C/min to 121 ⁰C dwell of samples
aged for 0 and 28 (out-life) days are plotted in Figure 7-3(a). It can be clearly seen that η
increases due to ambient exposure and cure thus less flow is expected with aged samples
processed under same cure cycle. Effect of increasing ramp rate from 2 ⁰C/min to 20 ⁰C/min
and dwell temperature from 121 ⁰C to 160 ⁰C on η evolution of the out-life sample is shown
in Figure 7-3 (b). It can be seen that increasing ramp rate and dwell temperature leads to
lower minimum η (ηmin) thereby increasing low viscosity flow. Conversely, slower ramp rate
combined with lower dwell temperature causes more extensive cure in the course of the ramp
leading to decrease in low η flow. In addition, higher dwell temperature is expected to be
more beneficial for samples with higher out-time where advanced initial degree of cure (α0)
or η level is expected. However, significant reduction in flow time (defined here as the flow
duration between η = 10 kPa· s to 10 kPa· s, from heat up stage to the gelation point) results
with increase in ramp rate and dwell temperature. Gelation point is known to be iso-
conversional (αgel = 0.37) and not cure path dependent and is commonly defined as η at 10
kPa· s [10] or at G’ = G” [22] for aerospace grade resins which share similar formulation
[Chapter 6]. Both methods consistently resulted in gelation time with no more than 3 minutes
148

difference from each other. As can be seen in Figure 7-3 (b), flow time reduction of 66
minutes results from the cure cycle change from 2 ⁰C/min to 121 ⁰C hold to 20 ⁰C/min to
160 ⁰C hold.

Figure 7-3. Representative viscosity (η) model prediction for: (a) day 0 and day 28; and (b)
day 28

To evaluate the balance between flow level and flow time on maximizing resin
flow, inverse viscosity (η
-1
) evolution of day 28 samples under 2 ⁰C/min to 121 ⁰C dwell
and 20 ⁰C/min to 160 ⁰C dwell are plotted in Figure 7-4. Integration of η
-1
curve results in
NFl,eff = 63.6 and 281.8 Pa
-1
respectively. It is expected that as resin cures and makes
progression towards gelation point, a deviation from Newtonian fluid behavior is expected.
However, the latter portion of NFl,eff does not account for more than 5% of the total NFl,eff so
149

the whole curve was integrated to compute NFl,eff. The model results predict that lower
viscosity flow is beneficial in terms of promoting more resin flow during fiber
impregnation despite having substantially shorter flow time.

Figure 7-4. Inverse viscosity (η
-1
) evolution plot of Figure 7-3 (b)

7.4.2. Void content analysis
Representative optical micrographs of uncured laminates at day 0 and day 49 are
depicted in Figure 7-5 (a) and (b) respectively. Day 0 and 49 samples were cold cured using
ammonia. Ammonia bath was created where the low vapor pressure of aqueous-ammonia
create ammonia vapor atmosphere inside the container. Ammonia can facilely effuse
throughout the laminate via the air pathway and react with the epoxy resins at the room
temperature to prevent resin flow [145]. The micrographs show similar partially impregnated
150

prepreg microstructure at day 0 and day 49 which confirm that, there are no significant
changes to the unsaturated tow area at room temperature. Thus, the effect of out-time on the
initial prepreg structure is found to be negligible.

Figure 7-5. Optical micrograph of uncured laminates at: (a) day 0 and (b) day 49

Low to high end ramp rates and dwell temperatures of the manufacturer’s
recommended cure cycles and their corresponding NFl,eff at the out-life (day 28) are
summarized in Table 7-2. It is expected that lowest NFl,eff value from the out-life sample
guarantees minimum required NFl,eff value, if not being a conservative value, to manufacture
acceptable parts (<1% voids by volume).

151

Table 7-2. Effective flow number (NFL,eff) for out-life (day 28) sample with manufacturer’s
recommended cure cycle
Cure Cycle NFl,eff (1/Pa)
0.6 ⁰C/min to 93 ⁰C 23.8
1.7 ⁰C/min to 93 ⁰C 27.4
0.6 ⁰C/min to 121 ⁰C 36.5
2.8 ⁰C/min to 121 ⁰C 71.7

Optical micrographs of cured laminates at day 0 and day 42 (α0 = 0.21) subjected to
same 2 ⁰C/min to 121 ⁰C hold cure cycle are shown in Figure 7-6 (a) and (b) respectively.
Complete resin impregnation is observed with day 0 sample which has NFL,eff = 542.1 Pa
-1
.
However, void content of 1.4% (macro-voids = 1.3% & micro-voids = 0.1%) is obtained with
day 42 sample with NFL,eff = 10.2 Pa
-1
which is well below the assumed minimum required
NFL,eff = 23.8 Pa
-1
. Interestingly, only 0.1% of micro-voids are observed which is expected to
be the predominant type of voids in aged samples [5] while 1.3% of macro-voids are obtained.
Resulting flow time of 68.7 minutes with high macro-voids content suggest that most of tows
were impregnated before entrapped bubbles were able to migrate through edge breathing.

152

Figure 7-6. Optical micrograph of laminates cured under 2 ⁰C/min to 121 ⁰C hold at: (a)
day 0 (NFL,eff = 542.1 Pa
-1
) and (b) day 42 (NFL,eff = 10.2 Pa
-1
)

Substantial reduction in voids is observed in Figure 7-7 (b) where day 42 sample was
cure under 20 ⁰C/min to 160 ⁰C hold resulting in NFL,eff = 40.5 Pa
-1
which resulted in macro-
voids content of 0.1%. Calculated flow time for the cure cycle was found to be 17 minutes
which is 51.7 minutes shorter than the former cure cycle (2 ⁰C/min to 121 ⁰C hold). However,
the result shows that NFL,eff increase via cure cycle adjustment induce more resin flow and
thus is a better criterion combining flow level and time to quantify the total resin flow. Also,
the result further confirms that our presumed flow requirement of NFL,eff > 23.8 Pa
-1
produces
quality parts even with an aged sample well past its out-life.

153

Figure 7-7. Optical micrograph of laminates at day 42 cured under: (a) 2 ⁰C/min to 121 ⁰C
hold (NFL,eff = 10.2 Pa
-1
) and (b) 20 ⁰C/min to 160 ⁰C hold (NFL,eff = 40.5 Pa
-1
)

Further testing on resin flow enhancement via NFL,eff optimization is depicted in
Figure 7-8 where optical micrographs of cured laminates at day 49 (α0 = 0.27) subjected to
(a) 25 ⁰C/min to 160 ⁰C hold and (b) 35 ⁰C/min to 170 ⁰C hold are shown. The two cure
cycles result in NFL,eff = 21.1 Pa
-1
and 24.1 Pa
-1
with macro-voids content = 0.64% and
0.49% respectively. Thus, faster ramp with higher dwell temperature (or lower viscosity
profile or higher NFL,eff) is further proven to be effective on enhancing resin infiltration in
fiber beds. Although 25 ⁰C/min to 160 ⁰C hold resulted in NFL,eff < 23.8 Pa
-1
, acceptable
void content of < 1% was obtained which implies that the criterion is on the conservative
ends.

154

Figure 7-8. Optical micrograph of laminates at day 49 cured under: (a) 25 ⁰C/min to 160 ⁰C
hold (NFL,eff = 21.1 Pa
-1
) and (b) 35 ⁰C/min to 170 ⁰C hold (NFL,eff = 24.1 Pa
-1
)

Optical micrographs of cured laminates at day 56 (α0 = 0.32) subjected to (a) 1
⁰C/min to 121 ⁰C hold and (b) 35 ⁰C/min to 170 ⁰C hold are shown in Figure 7-9. At this
point, the ambient temperature cure has taken place substantially where NFL,eff > 23.8 Pa
-1
criterion could no longer be met within the heating capacity of our heated tool. That is,
given αgel = 0.37, α0 of day 56 sample only has 5% of cure left until it reaches the gelation
point. The sample subjected to 1 ⁰C/min to 121 ⁰C hold resulted in NFL,eff = 1.8 Pa
-1
and
macro-voids content of 0.3 % and micro-voids content of 7.2%. In other words, macro-
voids content actually decreased in comparison to samples with higher NFL,eff > 10 Pa
-1
but
micro-voids appeared at a great extent that wasn’t observed with samples with higher NFL,eff
> 10 Pa
-1
. We speculate that micro-voids facilitated removal of macro-voids in this case and
the further analysis will be revisited later. The sample subjected to 35 ⁰C/min to 170 ⁰C
hold resulted in NFL,eff = 11.6 Pa
-1
and only macro-voids content of 1.2 % was observed.
Therefore within 1.8 Pa
-1
< NFL,eff < 11.6 Pa
-1
, we speculate that transition from macro-
155

voids formation to micro-voids formation occurs. Furthermore, we speculate that adequate
flow time is required for all entrapped air to escape through tows and to edge breathing.

Figure 7-9. Optical micrograph of laminates at day 56 cured under: (a) 1 ⁰C/min to 121 ⁰C
hold (NFL,eff = 1.8 Pa
-1
) and (b) 35 ⁰C/min to 170 ⁰C hold (NFL,eff = 11.6 Pa
-1
)

Figure 7-10 plots theoretical processing map of NFL,eff vs dwell temperature vs
156

heating rate (red: capability of conventional oven & teal: capability of heated tool) at day
49. It can be seen that higher ranges of NFL,eff or resin flow can be achieved using the heated
tool versus conventional oven. The main difference is attributed to controlled heating rate
from 0.5 ⁰C/min to 35 ⁰C/min attainable from the heated tool versus maximum assumed
heating rate of 5 ⁰C/min attainable from conventional ovens.

Figure 7-10. Representative processing map of NFL,eff vs dwell temperature vs heating rate
(red: capability of conventional oven & teal: capability of heated tool) at day 49
157

In an attempt to analyze micro-voids formation, especially the one observed from
cured laminates at day 56 (α0 = 0.32) subjected to 1 ⁰C/min to 121 ⁰C hold with NFL,eff = 1.8
Pa
-1
, total void and macro-void content (%) vs out-time (days) vs NFL,eff (Pa
-1
) are plotted in
Figure 7-11. Figure 7-11 (a) shows that the void content is not a function of out-time but
NFL,eff. However, Figure 7-11 (b) clearly shows that macro-voids content alone decreases
with lower NFL,eff at day 56 suggesting that a transition from macro-voids formation to
micro-voids formation occurs around NFL,eff ~ 10 Pa
-1
.

Figure 7-11. Void map of: (a) void content (%) vs out-time (days) vs NFL,eff (Pa
-1
) and (b)
macro-void content (%) vs out-time (days) vs NFL,eff (Pa
-1
)
158

Total void and macro-void content (%) vs flow-time (min) vs NFL,eff (Pa
-1
) are
plotted in Figure 7-12. Figure 7-12 (a) shows that the void content is not a function of flow
time but NFL,eff. The result further confirms that the criterion for micro-voids formation is
unlikely to be associated with flow-time where 62.3 minutes of flow time was available
with day 56 & NFL,eff = 1.8 Pa
-1
sample. In addition, cured laminates at day 42 (α0 = 0.21)
subjected to 2 ⁰C/min to 121 ⁰C hold which had NFL,eff = 10.2 Pa
-1
and micro-voids content
of 0.1% also had flow time of 68.7 minutes.

Figure 7-12. Void map of: (a) void content (%) vs flow-time (min) vs NFL,eff (Pa
-1
) and (b)
macro-void content (%) vs flow-time (min) vs NFL,eff (Pa
-1
)

159

Therefore, the main criterion for micro-voids formation is likely to be associated
with low viscosity induced flow or flow level which influence NFL,eff more than flow time.
This effect is shown in Figure 7-13 and 7-14 where its shown that longer flow time with
lower flow level result in lower NFL,eff leading to micro-voids formation.

Figure 7-13. Viscosity profiles and optical micrographs of day 42, 49 and 56 samples
subjected under various cure cycles

160

Figure 7-14. Inverse viscosity profiles and optical micrographs of day 42, 49 and 56
samples subjected under various cure cycles

Overall, the current investigation confirms that higher NFL,eff generally leads to
lower void content in the cured laminates (Figure 7-15(a)). However, as can be seen in
Figure 7-15 (b), further investigation is needed to find a micro-void formation criterion
where NFL,eff < 11.6 Pa
-1
.

161

Figure 7-15 (a) void content (%) vs vs NFL,eff (Pa
-1
) and (b) macro-void content (%) vs
NFL,eff (Pa
-1
)

7.5. Conclusions
The present chapter investigated effect of temperature cycle variation to potentially
develop a flow-enhanced or viscosity-controlled cure cycle that can limit flow induced
defects. First, laminates (CYCOM
®
5320-1/IM7, Cytec Industries Inc.) were laid up with
[0°]6s sequence and aged for 0 to 56 days (out-life = 28 days). A flow quantity, NFl,eff (Pa
-1
)
was defined integrating inverse viscosity evolution during cure thus taking into account of
both flow level and time. Laminates with various out-time and cure cycle were tested and
their void contents were optically analyzed to ultimately develop a criterion for
manufacturing quality parts with samples well past the out-life. To attain high NFl,eff values,
162

in-house developed heated to tool capable of rapid heating up to 35°C/min and accurate
temperature control between 30°C and 200°C was used.
The results showed that higher heating rate to higher dwell temperature leads to a
more rapid decrease in η, allowing the resin to achieve lower η before the substantial effect
of cure kicks in. At the same time, the same cure cycle lead to flow time decrease. However,
NFl,eff analysis revealed that ηmin had greater influence on NFl,eff increase than flow time thus
favoring higher heating rate to higher dwell temperature. V oids analysis revealed strong
correlation between overall void content decrease with NFl,eff. No micro-voids were observed
above NFl,eff greater than 10.2 Pa
-1
but formation of macro-voids suggesting that entrapped air
needs evacuation time during resin flow. NFl,eff = 1.8 Pa
-1
exhibited decrease in macro-voids
content but substantial increase in micro-voids content from NFl,eff = 10.2 Pa
-1
. NFl,eff greater
than 20 Pa
-1
had only macro-voids with overall voids content less than < 1%.
Overall, this study suggest that NFl,eff can be used as an effective criterion to limit
flow induced defects. This allows use of materials long after its suggested out-life. Although
not investigated in this study, a higher NFl,eff criterion is expected with fabrics versus UD as
fabric requires more resin travel distance. In addition, a further investigation on micro-voids
formation criterion is suggested.

163

Chapter 8. Conclusions and Future Work

8.1. Contributions
The projects investigated in this thesis have led to several contributions on OoA
prepreg processing that have not been investigated before. In attempt to tackle processing
challenges associated with OoA/VBO manufacturing, efficient process modelling and
monitoring methods that are generic to aerospace grade resins were developed and used to
develop a flow enhanced cure cycle method that can limit flow induced defect. Specifics of
these contributions are divided into five parts as:
1. Accurate process modelling accounting for out-time and humidity (chapter 3, 4,
5 and 6). Effective methods for quantifying the effects of out-time and humidity on
an OoA prepreg resin were developed. A representative OoA resin was characterized
using conventional DSC and rheometry methods, and predictive models that
comprehensively captured out-time and humidity effects on cure kinetics and
viscosity in process conditions were developed. The study also identified critical
physicochemical events (gelation, vitrification and maximum flow) and their
dependence on out-time for an OoA prepreg resin during cure. Polymerization at
ambient temperatures altered the initial degree of cure of the resin as well as the
evolution of cure rate and viscosity at elevated temperatures and moisture absorption
was shown to further catalyze the reaction.

164

2. Process monitoring method accounting for out-time and humidity (chapter 3, 4
and 5). In-situ dielectric monitoring experiments were used to determine the evolution
of key dielectric properties during the same temperature and out-time and humidity
conditions, and effective correlations were developed to obtain degree of cure and
viscosity from these properties. Also, critical physicochemical events (gelation,
vitrification and maximum flow) and their dependence on out-time and humidity for
an OoA prepreg resin during cure were identified with dielectric monitoring. With the
process modelling, these methods constitute complementary methods for predicting
and monitoring in-situ the “instantaneous” degree of cure, cure rate, and viscosity
evolution of composite prepreg. Used together, the methods offer a means to improve
process effectiveness and efficiency in composite manufacturing.
3. Process modelling generality validation using commonly used aerospace grade
resin (chapter 5 and 6). Three aerospace grade resins were formulated to investigate
the effects of variation in DDS isomers, a/e stoichiometric ratio, out-time, and
moisture absorption on processing characteristics and cured resin properties. Same
conventional thermochemical and thermomechanical methods (MDSC and
rheometry) were used to collect benchmark data. The model result indicated that cure
kinetics and viscosity model developed in this thesis can be used on range of
aerospace grades resin systems.

165

4. Effects of variation in DDS isomers, a/e stoichiometric ratio, out-time, and
moisture absorption on processing characteristics (chapter 6). Regardless of the
resin system, the results showed that out-time increases initial degree of cure and
more so with moisture absorption which influence the cure kinetics and viscosity
evolution during cure. These effects were shown to shorten flow level and time which
could potentially lead to insufficient resin flow during composite manufacturing.
Therefore, limiting out-time and including humidity and thermal controls are
generally required to ensure sufficient resin flow.
Among the three resin systems investigated in this study, meta- substituted 33DDS
with a/e = 0.8 exhibited the highest cure rate and the lowest flow time. This behavior
was attributed to the increase in collision number between amine and epoxy compared
to a/e = 0.6 and the lack of delocalization of the lone pair of electrons on nitrogen
compared to para- substituted 44DDS. Conversely, 44DDS with a/e = 0.8 exhibited
the slowest cure rate with the highest flow time. In addition, 44DDS based resin
exhibited the highest glass transition temperature due to having lower configurational
entropy than 33DDS based resin. The results offered practical insights regarding resin
formulation and prepreg processing, particularly that: (1) 33DDS-based resin is more
susceptible to aging and flow time reduction, albeit with faster processing time; (2)
lower a/e will require longer dwell at higher temperature to complete the
etherification reaction, albeit with more flow time; and (3) 44DDS resin offers longer
flow time with higher glass transition temperature, yet with substantially longer
processing time. Finally, the predictive viscosity model demonstrated a method to
166

enhance resin flow during prepreg processing which can potentially limit flow-
induced defects.
Overall, this study provided foundational knowledge for aerospace grade DDS isomer
cured resins subjected to pre-cure conditions that are largely unavoidable in practice
with prepreg processing.
5. Flow enhanced cure cycle to limit flow induced defects (chapter 7). This study
investigated effect of temperature cycle variation to potentially develop a flow-
enhanced or viscosity-controlled cure cycle that can limit flow induced defects. First,
laminates (CYCOM
®
5320-1/IM7, Cytec Industries Inc.) were laid up with [0°]6s
sequence and aged for 0 to 56 days (out-life = 28 days). A flow quantity, NFl,eff (Pa
-1
)
was defined integrating inverse viscosity evolution during cure thus taking into
account of both flow level and time. Laminates with various out-time and cure cycle
were tested and their void contents were optically analyzed to ultimately develop a
criterion for manufacturing quality parts with samples well past out-life. To attain
high NFl,eff values, in-house developed heated to tool capable of rapid heating up to
35°C/min and accurate temperature control between 30°C and 200°C was used.
The results showed that higher heating rate to higher dwell temperature leads to a
more rapid decrease in η, allowing the resin to achieve lower η before the substantial
effect of cure kicks in. At the same time, the same cure cycle lead to flow time
decrease. However, NFl,eff analysis revealed that ηmin had greater influence on NFl,eff
increase than flow time thus favoring higher heating rate to higher dwell temperature.
167

V oids analysis revealed strong correlation between overall void content decrease with
NFl,eff.
Overall, this study suggested that NFl,eff can be used as an effective criterion to limit
flow induced defects. This allows use of materials long after its suggested out-life.

8.2. Broader Implications
Moving the composite manufacturing out of autoclave adds challenges in manifolds.
In addition to composites offering lightweight structures with high specific mechanical (i.e.
strength to weight ratio) properties and excellent fatigue life, VBO process enable further
increase in economic (cost + time) value. As the demand for composites use in aircrafts is
expected to quadruple in the next decade, efficient and cost-effective manufacturing methods
like VBO processing are required to meet the growing demand. VBO processing removes
high consolidation pressure imparted by an autoclave voids are suppressed by evacuating
entrapped air and vaporized moisture through a partially impregnated microstructure made
up of both dry and resin-rich regions. During VBO processing, dry regions are infiltrated by
surrounding resin to form a uniform and ideally void-free microstructure. The rate of
infiltration and the quality of OoA laminates are therefore strongly influenced by the cure
kinetics and viscosity evolution of the infiltrating resin. The five projects completed here
were aimed to investigate fundamental problems such as out-time, flow induced defects and
moisture absorption that are commonly encountered with VBO processing and to suggest
viable solutions through practical modelling and monitoring methods.
168

8.3. Recommendations for Future Work
Recommended future works include:
1. Cure monitoring capability using dielectric analysis (DEA) investigated in this thesis
can be further investigated to develop a ‘self-correcting cure control system’. As
variables such as ionic conductivity and dipole contribution to dielectric loss directly
represent cure state, these variables can be utilized to develop a cure monitoring
system that can self-correct itself to ensure sufficient resin flow during laminates
processing.
2. On the research level, further investigations of the flow enhanced cure cycle
development study on: 1. Finding micro-voids forming criterion; and 2. Effects of
flow enhanced cure cycle on cured mechanical properties are recommended. The flow
enhanced cure cycle method presented here was shown to be effective on decreasing
overall voids content yet a deeper investigation on micro-voids formation is still
needed. The work will potentially find a flow time criterion for tow impregnation.
Also, higher heating rates combined with dwell temperature used in the method may
have effect on mechanical properties of cured laminates as the commercial prepreg
may contain tougheners and additives.
3. Most of the work performed here are done on neat resin and on small flat panels. This
was to analyze fundamental problems and to model fundamental processes. Bigger
169

sized parts, complex geometries, and sandwich structures pose additional challenges
such as longer breathing and flow distance. In addition, general scale up operation in
real industrial settings add challenges such as uneven heating, production schedule,
etc.

170

References
1. Gardiner G. Out-of-autoclave prepregs: hype or revolution? High perform compos
2011.

2. Campbell. F.C. Manufacturing Technology for Aerospace Structural Materials.
Elsevier, London, U.K., 2006.

3. Repecka, L. and Boyd J., Vacuum-bag-only-curable prepregs that produce void free
parts, in 47th International SAMPE Symposium. 2002. p. 1862-1875.

4. Tavares, S.S., Michaud, V . and J.A.E. Manson, Assessment of semi-impregnated
fabrics in honeycomb sandwich structures. Composites: Part A, 2010. 41: p. 8-15.

5. Grunenfelder LK, Centea T, Hubert P, Nutt SR. Effect of room-temperature out-time
on tow impregnation in an out-of-autoclave prepreg. Compos Part A Appl Sci
Manuf 2013;45:119–26.

6. Vanlandingham, M.R., et al., Relationships between Stoichiometry, Microstructure,
and Properties for Amine-Cured Epoxies. Journal of Applied Polymer Science, V ol.
71, 699–712 (1999).

7. Thorfinnson, B. and Biermann, T. Production of void free composite parts without
debulking. In Proceedings of the 31st International SAMPE Symposium, Las Vegas,
NV , United States, 1986.

8. Thorfinnson, B. and Biermann, T. Degree of impregnation of prepregs - effects on
porosity. In Proceedings of the 32nd International SAMPE Symposium, Anaheim,
CA, United States, 1987.

9. Liu, L., et al., Effects of cure cycles on void content and mechanical properties of
composite laminates. Composite Structures, 2006. 73: p. 303-309.

10. Boyd J, Maskell RK. Product design for low cost manufacturing of composites for
aerospace applications. Proc SAMPE 2001 conf. Long Beach, CA: Society for the
Advancement of Material and Process Engineering; 2001.

11. Repecka L, Boyd J. Vacuum-bag-only-curable prepregs that produce void-free parts.
Proc SAMPE 2002 conf. Long Beach, CA: Society for the Advancement of
Material and Process Engineering; 2002.

171

12. Bond GG, Griffith JM, Hahn GL, Bongiovanni C, Boyd J. Non-autoclave prepreg
manufacturing technology. In: Proc 40th int SAMPE tech conf. Memphis, TN:
Society for the Advancement of Material and Process Engineering; 2008.

13. Mortimer S, Smith MJ, Olk E. Product development for out-of-autoclave (O.O.A.)
manufacture of aerospace structure. Proc SAMPE 2010 conf. Seattle, WA: Society
for the Advancement of Material and Process Engineering; 2010.

14. Lucas S, Howard S, Senger J. Vacuum bag only processing: improving prepreg out-
time and porosity for large composite structure. Proc SAMPE 2010 conf. Seattle,
WA: Society for the Advancement of Material and Process Engineering; 2010.

15. Hahn, GL, Bond GG, Fogarty JH. Non-autoclave (prepreg) manufacturing
technology: part scale-up with CYCOM 5320-1 prepregs. Proc SAMPE 2011 conf.
Long Beach, CA: Society

16. Hubert, P. and Poursartip, A. Review of flow and compaction modelling relevant to
thermoset matrix laminate processing. Journal of Reinforced Plastics and
Composites, 17(4):286-318, 1998.

17. Kratz, J., Hsiao K., Fernlund, G. and P. Hubert. Thermal models for mtm45-1 and
cycom 5320 out-of-autoclave prepreg resins. Journal of Composite Materials, 2011.

18. Kamal, M.R. and Sourour, S. Kinetics and thermal characterization of thermoset
cure. Polymer Engineering and Science, 13(1):59-64, 1973.

19. Kim, D., Centea, T., Nutt, SR. Out-time effects on cure kinetics and viscosity for an
out-of-autoclave (OoA) prepreg: modelling and monitoring. Compos Sci Technol
2014;100:63–9.

20. Kim, D., Centea, T., Nutt, SR. In-situ cure monitoring of an out-of-autoclave
prepreg: Effects of out-time on viscosity, gelation and vitrification. Compos Sci
Technol 2014;102:132–8.

21. Kim, D., Centea, T., Nutt, SR. Modelling and Monitoring of Out-time and Moisture
Absorption on Cure Kinetics and Viscosity for An Out-of Autoclave (OoA) Prepreg.
Compos Sci Technol 2017; 138:201-8.

22. Grunenfelder, LK., Centea, T., Huber,t P. and Nutt, SR. Effect of room-temperature
out-time on tow impregnation in an out-of-autoclave prepreg. Compos Part A: Appl
Sci Manuf 2013; 45:119–126

23. Powers, DA. Interaction of Water with Epoxy. Sandia Report: SAND2009-4405.
2009.
172

24. Blank, WJ., He, ZA. and Picci, M. Catalysis of The Epoxy-Carboxyl Reaction.
International Waterborne, High-Solids and Powder Coatings Symposium. New
Orleans, LA, Feb 21-23, 2001.

25. Min, B.G., et al., Cure kinetics of elementary reactions of a diglycidyl ether of
bisphenol A/diaminodiphenylsulfone epoxy resin: 2. Conversion versus time.
Polymers, 1993. 34: p. 4488-95

26. Min, B.G., et al., Cure kinetics of elementary reactions of a DGEBA/DDS epoxy
resin: 1. Glasstransition temperature versus conversion. Polymers, 1993. 34: p.
4908-12

27. Kim, D and Nutt, SR. Processability of DDS Isomers Cured Epoxy Resin: Effects of
Amine/Epoxy Ratio, Humidity and Out-time, Submitted to Journal of Applied
Polymer Science.

28. Pascault, J. P. General Concepts about Epoxy Polymers. WILEY-VCH Verlag
GmbH & Co. KGaA, Weinheim, 2010.

29. Amine Curing of Epoxy Resins: Options and Key Formulation Considerations
http://www.pcimag.com/articles/85952-amine-curing-of-epoxy-resins-options-and-
key-formulation-considerations

30. Bongiovanni, C. L. (Cytec Technology Corp.) et al. US 20100222461, September 2,
2010

31. Chang, T. D., Carr, S. H. and Brittain J. O. J. Polym. Eng. Sci. 1989, 22, 1213-1220.

32. Varma, I. K. and Satya Bhama P. V . J. Compos.Mater. 1986, 20, 410-422.

33. Vanlandingham, M. R., Eduljee, R. F. and Gillespie, J. W. J. Appl. Polym. Sci. 1999,
71, 699–712.

34. Tungare A. V ., Martin, G. C. and Gotro, J. T. Polym. Eng. Sci. 1988, 28, 1071-1075.

35. Pascault, J.P., Sautereau , H., Verdu , J., and Williams , R.J.J. ( 2002 ) Thermosetting
Polymers, Marcel Dekker, New York .
36. C. Ridgard. Out of autoclave composite technology for aerospace, defense and
space structures. In Proceedings of the SAMPE 2009 Conference of the Society for
the Advancement of Materials and Process Engineering, Baltimore, MD, United
States, 2009.
173

37. Ahn, K.J., J.C. Seferis, and J.C. Berg, Simultaneous measurements of permeability
and capillary pressure of thermosetting matrices in woven fabric reinforcements.
Polymer Composites, 1991. 12(3): p. 146-152.

38. Rudd, C.D., et al., Compaction and in plane permeability characteristics of quasi
unidirectional and continuous random reinforcements. Materials Science and
Technology, 1996. 12: p. 436-444.

39. Cole, K.C., Hechler, J.J. and Noel, D. New approach to modeling the cure kinetics
of epoxy amine thermosetting resins. 2. Application to a typical system based on
bis[4-(diglycidylamino)phenyl]methane and bis(4-aminophenyl) sulfone.
Macromolecules, 24(11):3098{3110, 1991.

40. Khoun, L.K., Centea, T. and Hubert, P. Characterization methodology of thermoset
resins for the processing of composite materials – case study: Cycom 890rtm epoxy
resin. Journal of Composite Materials, 44(11):1397{1415, 2009.

41. J.M. Castro and C.W. Macosko. Kinetics and rheology of typical polyurethane
reaction injection molding systems. In Society of Plastics Engineers Technical
Papers, volume 26, pages 434-438, 1980.

42. A.T. DiBenedetto. Prediction of the glass transition temperature of polymers: A
model based on the principle of corresponding states. Journal of Polymer Science
Part B: Polymer Physics, 25(9):1949-1969, 1987.

43. J. Mijovic, S. Andjelic, C.F.W. Yee, F. Bellucci, L. Nicolais “A study of reaction
kinetics by near-infrared spectroscopy. 2. Comparison with dielectric spectroscopy
of model and multifunctional epoxy/amine systems”, Macromolecules, 28 (1995)
2797 – 2806

44. Advanced Composites Group. ACG MTM 45-1 Matrix Resin.

45. Cytec Engineered Materials. CYCOM 5320 Information Sheet.

46. Cytec Engineered Materials. CYCOM 5320-1 Information Sheet.

47. J.P. Runt, J.J. Fitzgerald, “Dielectric Spectroscopy of Polymeric Materials:
fundamentals and applications”, American Chemical Society, Washington DC
1997,(ISBN: 0-8412-3335-7)

48. A.R.A. Arafath, G. Fernlund, and A. Poursartip. Gas transport in prepregs: Model
and permeability experiments. In Proceedings of the 17th International Conference
on Composite Materials, Edinburgh, Scotland, 2009.
174

49. J.L. Kardos, M.P. Dudukovic, and R. Dave. Void growth and resin transport during
processing of thermosetting-matrix composites. Advances in Polymer Science,
80:101-123, 1986.

50. Centea T, Grunenfelder LK, Nutt SR. A review of out-of-autoclave prepregs –
Material properties, process phenomena, and manufacturing considerations.
Compos Part A Appl Sci Manuf 2015;70:132–54

51. Grunenfelder LK, Nutt SR. Prepreg age monitoring via differential scanning
calorimetry. J Reinf Plast Compos 2012; 31(5):295-302

52. Jones RW, Ng Y and McClelland JF. Monitoring ambient-temperature aging of a
carbon-fiber/epoxy composite prepreg with photoacoustic spectroscopy. Compos
Part A: Appl Sci Manuf 2008; 39(6):965–971.

53. Ji KJ, Wei CY , Deng WH, et al. Evaluation of glass fibre/epoxy prepreg quality
during storage. Polym Polym Compos 2002; 10: 599–606.

54. Leider, Ind. Eng. Chem. Fundam., 13, 342 (1974)

55. Leider and Bird, Ind. Eng. Chem. Fundam., 13, 336 (1974)

56. Grimm, A.I.Ch.E. J., 24, 428 (1978)

57. Bird, Armstrong and Hassager (Dynamics of Polymeric Liquids, V ol 1. Fluid
Mechanics, 2nd ed. Wiley, New York, p. 21, 1987)

58. Macosko, Rheology Principles, Measurements, and Applications, VCH Publishers,
p. 270, 1994)

59. Stefan, Sitzber. K. Akad. Wiss. Math. Natur., Wien, 69, Part 2, 713 (1874)

60. Tungare, Martin, and Gotro, Polym.Eng. and Sci., 28, no. 16, 1071, (1988)

61. Yijia Ma, Daniel Kim & Steven R. Nutt, “Effect of Crosslink Density on The
Recyclability of Carbon Fiber Reinforced Polymers,” SAMPE 2016. Long Beach,
CA; 2016

62. Yijia Ma, Daniel Kim, Travis J. Williams & Steven R. Nutt, “Recycling of Carbon
Fiber Reinforced Polymers Using Chemical Treatment: Reaction Characterization
and Optimization,” SAMPE 2017. Seattle, CA; 2017

175

63. Yijia Ma, Daniel Kim & Steven R. Nutt, “Recycling of Carbon Fiber/Epoxy
Composites Using Chemical Treatment Under Atmospheric Pressure: Resin
Dissolution and Characterization”, In Preparation

64. Oliveux G, Dandy LO, Leeke GA. Current status of recycling of fibre reinforced
polymers: Review of technologies, reuse and resulting properties. Progress in
Materials Science. 2015 Jul 31;72:61-99.

65. EURECOMP recycling thermoset composites of the SST final summary report
(2012).

66. European Directive 2000/53/EC, on End of Life Vehicles.

67. Pickering SJ. Recycling technologies for thermoset composite materials—current
status. Composites Part A: Applied Science and Manufacturing. 2006 Aug
31;37(8):1206-15.

68. Hyde JR, Lester E, Kingman S, Pickering S, Wong KH. Supercritical propanol, a
possible route to composite carbon fibre recovery: A viability study. Composites
Part A: Applied Science and Manufacturing. 2006 Nov 30;37(11):2171-5.

69. Oliveux G, Dandy LO, Leeke GA. Degradation of a model epoxy resin by
solvolysis routes. Polymer Degradation and Stability. 2015 Aug 31;118:96-103.

70. Kamykowski, G. TA Training Seminars: Rheology. 2004.

71. Karl Fisher Titration: Good titration practice in Karl Fisher titration. Mettler-
Toledo 2011.

72. Karl Fischer Titration: One step water determination. Simple and secure. Mettler-
Toledo 2011.

73. Kratz J, Hsiao K, Fernlund G, Hubert P. Thermal models for MTM45-1 and Cycom
5320 out-of autoclave prepreg resins. J Compos Mater 2013; 47(3):341-352

74. Grunenfelder LK, Centea T, Hubert P, Nutt SR. Effect of room-temperature out-time
on tow impregnation in an out-of-autoclave prepreg. Compos Part A: Appl Sci
Manuf 2013; 45:119–126

75. Um MK, Daniel IM, Hwang BS. A study of cure kinetics by the use of dynamic
differential scanning calorimetry. Compo Sci and Technol 2002; 62(1):29-40

76. Du S, Guo ZS, Zhang B, Wu Z. Cure kinetics of epoxy resin used for advanced
composites. Polym Int 2004; 53:1343-47
176

77. Khoun L, Centea T, Hubert P. Characterization Methodology of Thermoset Resins
for the Processing of Composites Materials – Case Study: CYCOM 890RTM Epoxy
Resin. J Compos Mater 2010; 44(11):1397-1416

78. Lin RH, Hsu JH. In situ FT-IR and DSC investigation on the cure reaction of the
dicyanate/diepoxide/diamine system. Polym Int 2001; 50 (10):1073-81

79. Hardis R, Jessop JLP, Peters FE, Kessler MR. Cure kinetics characterization and
monitoring of an epoxy resin using DSC, Raman Spectroscopy, and DEA. Compos
Part A: Appl Sci Manuf 2013; 49:100–108

80. Skordos AA, Partridge IK. Determination of the Degree of Cure under Dynamic and
Isothermal Curing Conditions with Electrical Impedance Spectroscopy. J Polym Sci
Pol Phys 2004; 42:146-154

81. Kazilas MC, Partridge IK. Exploring equivalence of information from dielectric and
calorimetric measurements of thermoset cure – a model for the relationship between
curing temperature, degree of cure and electrical impedance. Polymer 2005;
46:5868-78

82. Grunenfelder LK, Nutt SR. Prepreg age monitoring via differential scanning
calorimetry. J Reinf Plast Compos 2012; 31(5):295-302

83. Jones RW, Ng Y and McClelland JF. Monitoring ambient-temperature aging of a
carbon-fiber/epoxy composite prepreg with photoacoustic spectroscopy. Compos
Part A: Appl Sci Manuf 2008; 39(6):965–971

84. Ji KJ, Wei CY , Deng WH, et al. Evaluation of glass fibre/epoxy prepreg quality
during storage. Polym Polym Compos 2002; 10: 599–606.

85. Akay M, Cracknell JG, Farnham HA. Detection of undercure in carbon-fibre/epoxy
composites by dynamic mechanical thermal analysis. Polym Polym Compos 1994;
2(5):317-322

86. Mijovic J, Kenny J, Maffezzoli A, Trivisano A, Bellucci F, Nicolais L. The
principles of dielectric measurements for in situ monitoring of composite
processing. Compo Sci and Technol 1993; 49(3):277-290

87. Pethrick RA, Hayward D. Real time dielectric relaxation studies of dynamic
polymeric systems. Prog Polym Sci 2002; 27:1983-2017

177

88. Maffezzoli A, Quarta E, Luprano, VAM, Montagna G, Nocolais L. Cure monitoring
of epoxy matrices for composites by ultrasonic wave propagation. J Appl Polym Sci
1999; 73(10):1969-77

89. Woerdeman DL, Flynn KM, Dunkers JP, Parnas RS. The Use of Evanescent Wave
Fluorescence Spectroscopy for Control of the Liquid Molding Process. J Reinf Plast
Comp 1996; 15:922-943

90. Cytec Engineered Materials. CYCOM 5320-1 Information Sheet.

91. Ridgard C. Out of autoclave composite technology for aerospace, defense and space
structures. SAMPE 2009. Baltimore, MD: Society for the advancement of materials
and process engineering; 2009

92. Mason, K. F. Autoclave Quality Outside the Autoclave? High-Performance
Composites. 2006; 14(2): 44-49

93. Grunenfelder LK, Nutt SR. Prepreg age monitoring via differential scanning
calorimetry. J Reinf Plast Compos 2012; 31(5):295-302

94. Grunenfelder LK, Centea T, Hubert P, Nutt SR. Effect of room-temperature out-time
on tow impregnation in an out-of-autoclave prepreg. Compos Part A: Appl Sci
Manuf 2013; 45:119–126

95. Jones RW, Ng Y and McClelland JF. Monitoring ambient-temperature aging of a
carbon-fiber/epoxy composite prepreg with photoacoustic spectroscopy. Compos
Part A: Appl Sci Manuf 2008; 39(6):965–971

96. Ji KJ, Wei CY , Deng WH, et al. Evaluation of glass fibre/epoxy prepreg quality
during storage. Polym Polym Compos 2002; 10: 599–606

97. Maistros G, Fontana QPV , Attwood D, Hudd JS. Use of modulated differential
scanning calorimetry to observe vitrification during epoxy resin cure. J Mater Sci
Lett 1997; 16: 273-275

98. Centea T, Kratz J, Hubert T. The effect of out-time and cure cycle on the
consolidation of out-of-autoclave prepregs. FPCM–11. Auckland, New Zealand:
11th International conference on flow processes in composite materials; 2012

99. Maistros GM, Partridge IK. Dielectric monitoring of cure in a commercial carbon-
fibre composite. Compo Sci and Technol 1995; 53:355-359

178

100. Stark W, Jaunich M, McHugh J. Cure state detection for pre-cured carbon-fibre
epoxy prepreg (CFC) using Temperature-Modulated Differential Scanning
Calorimetry (TMDSC). Polym Test 2013; 32(7): 1261-72

101. Mijovic J, Kenny J, Maffezzoli A, Trivisano A, Bellucci F, Nicolais L. The
principles of dielectric measurements for in situ monitoring of composite
processing. Compo Sci and Technol 1993; 49(3):277-290

102. Pethrick RA, Hayward D. Real time dielectric relaxation studies of dynamic
polymeric systems. Prog Polym Sci 2002; 27:1983-2017

103. Maistros GM, Partridge IK. Monitoring autoclave cure in commercial carbon
fibre/epoxy composites. Compos Part B: Eng 1998; 29B: 245-250

104. Kim, D., Centea, T., Nutt, S.R., Out-Time Effects on Cure Kinetics and Viscosity
for an Out-Of-Autoclave (OOA) Prepreg: Modelling and Monitoring, Compo Sci
and Technol (2014), doi: http://dx.doi.org/10.1016/j.compscitech.2014.05.027

105. Skordos AA, Partridge IK. Determination of the Degree of Cure under Dynamic and
Isothermal Curing Conditions with Electrical Impedance Spectroscopy. J Polym Sci
Pol Phys 2004; 42:146-154

106. Cytec Engineered Materials. CYCOM® 5320-1 Toughened Epoxy for Structural
Applications -- Out-of-Autoclave Manufacturing. 2010.

107. Kim D, Srivatava S, Narayaran S, Archer LA. Polymer nanocomposites: polymer
and particle dynamics. Soft Matter 2012; 8:10813-18

108. Mason, K. F., “Autoclave Quality outside the Autoclave?” High-Performance
Composites 14 (2006): 44-49.

109. Maistros GM, Partridge IK. Monitoring Autoclave Cure in Commercial Carbon
Fibre/Epoxy Composites. Compos Part B: Eng 1998; 29B:245–50.

110. Pethrick RA, Hayward D. Real Time Dielectric Relaxation Studies of Dynamic
Polymeric Systems. Prog Polym Sci 2002; 27:1983–2017.

111. Kim D, Centea T, Nutt SR. In-Situ Dielectric Monitoring of Out-of-Autoclave
Prepreg: Effects of Out-Time on Viscosity, Gelation and Vitrification. Compos Sci
Technol 2014; 102:132–138.

112. Kim D, Centea T, Nutt SR. Out-Time Effects on Cure Kinetics and Viscosity for an
Out-Of-Autoclave (OOA) Prepreg: Modelling and Monitoring. Compos Sci Technol
2014; 100:63–69.
179

113. Grunenfelder LK, Centea T, Hubert P, Nutt SR. Effect of Room-Temperature Out
time on Tow Impregnation in Out-of-Autoclave Prepreg. Compos Part A: Appl Sci
Manuf 2013; 45:119–26.

114. Powers DA. Interaction of Water with Epoxy. Sandia Report: SAND2009-4405.
2009.

115. Blank WJ, He ZA, Picci M. Catalysis of the Epoxy-Carboxyl Reaction.
International Waterborne, High-Solids and Powder Coatings Symposium. New
Orleans, LA, Feb 21-23, 2001.

116. Cytec Engineered Materials. CYCOM 5320-1 Information Sheet

117. Greenspan L. Humidity Fixed Points of Binary Saturated Aqueous Solutions. J Res
Natl Stand, Sec A 1977, 81A (1) 89-96.

118. Maistros, G. & Partridge, IK. “Dielectric monitoring of cure in a commercial
carbon-fibre composite.” Compo Sci and Technol 1995; 53:355-359.

119. Kratz J, Hsiao K, Fernlund G, and Hubert P. Thermal Models for MTM45-1 and
Cycom 5320 Out-of Sutoclave Prepreg Resins. J Compos Mater 2013; 47(3):341–
52.

120. Min BG, Stachurski ZH, Hodgkin JH. “Cure Kinetics of elementary reactions of a
DGEBA/DDS epoxy resin: 1. Glass transition temperature versus conversion.
Polymer 1993; 34: 4908-4912

121. Khoun L, Centea T and Hubert P. Characterization Methodology of Thermoset
Resins for the Processing of Composites Materials – Case Study: CYCOM 890RTM
Epoxy Resin. J Compos Mater 2010; 44(11):1397-1416

122. T. Centea, J. Kratz and P. Hubert. The Effect of Out-Time and Cure Cycle on the
Consolidation of Out-of-Autoclave Prepregs. 11th International Conference on Flow
Processes in Composite Materials. Auckland, New Zealand. July 9-12 2012.

123. Tungare A. V ., Martin, G. C. and Gotro, J. T. Polym. Eng. Sci. 1988, 28, 1071-1075.

124. Garcia, F. G., Soares, B. G., Pita, V . J., Sanchez, R. and Rieumont J. J. Appl. Polym.
Sci. 2007, 106, 1457–1471.

125. Swier, S. and Mele, S. S. J. Polym. Sci. Pol. Phys. 2003, 41, 594-608.

126. Vratsanos, M. S. and Farris, R. J. Polym. Eng. Sci. 1989, 29, 806-816.
180

127. Vanlandingham, M. R., Eduljee, R. F. and Gillespie, J. W. J. Appl. Polym. Sci. 1999,
71, 699–712.

128. Chang, T. D., Carr, S. H. and Brittain J. O. J. Polym. Eng. Sci. 1989, 22, 1213-1220.

129. Varma, I. K. and Satya Bhama P. V . J. Compos.Mater. 1986, 20, 410-422.

130. Min, B. G., Stachurski, Z. H. and Hodgkin. Polymer. 1993, 34, 4488-4495.

131. Bongiovanni, C. L. (Cytec Technology Corp.) et al. US 20100222461, September 2,
2010

132. Kim, D., Centea, T. and Nutt, S. R. Compos. Sci. Technol. 2017, 138, 201-208.

133. Jones, R. W., Ng Y . and McClelland J. F. Compos. Part A Appl. Sci. Manuf. 2008,
39, 965-971.

134. Grunenfelder, L. K., Centea, T., Hubert, P. and Nutt, S. R. Compos. Part A Appl.
Sci. Manuf. 2013, 45, 119-126.

135. Kazilas, M. C. and Partridge, I. K. Polymer. 2005, 45, 5868-5878.

136. Um, M. K., Im, D. and Hwang, B. S. Compos. Sci. Technol. 2002, 62, 29-40.

137. Du, S., Guo, Z. S., Zhang, B. and Wu, Z. Polym. Int. 2004, 53, 1343-1347.

138. Kratz, J., Hsiao K., Fernlund G. and Hubert P. J. Compos. Mater. 2013, 47, 341-352.

139. Khoun, L., Centea, T. Hubert, P. J. Compos. Mater. 2010, 44, 1397-1416.

140. Tung, C. Y . M. and Dynes, P. J. J. Appl. Polym. Sci. 1982, 27, 569–574.

141. Kim, D., Centea, T. and Nutt, S. R. Compos. Sci. Technol. 2014, 102, 132-138.

142. Leider, P. J. Ind. Eng. Chem. Fundam. 1974, 13, 342-346

143. Leider, P. J. and Bird, R. B. Ind. Eng. Chem. Fundam. 1974, 13, 336-341

144. Schuetze-Coburn A, Klimkzak S, Tsui V . Development of Active Tooling for
Composite Manufacturing. Proc. AIAA Reg. VI Student Conf., Sacramento, CA:
American Institute for Aeronautics and Astronautics; 2014.

181

145. Stephen J. Howard, “Evaluating Consolidation Dynamics – Study of Key
Consolidation Parameters in a Vacuum Bag Only (VBO) Cure,” SAMPE 2016.
Long Beach, CA; 2016

Abstract (if available)

Abstract Out-of-Autoclave (OoA) prepreg allow the manufacturing of low-porosity, high-performance composite structures through flexible, cost-effective vacuum-bag-only (VBO) processing. These are cured in conventional ovens to produce autoclave quality parts with low void contents (< 1% by volume for aerospace applications). In the absence of high consolidation pressure, such as that imparted by an autoclave, voids are suppressed by evacuating entrapped air and vaporized moisture through a partially impregnated microstructure made up of both dry and resin-rich regions. During VBO processing, dry regions are infiltrated by surrounding resin to form a uniform and ideally void-free microstructure. The rate of infiltration and the quality of OoA laminates are therefore strongly influenced by the cure kinetics and viscosity evolution of the infiltrating resin. ❧ During pre-cure operations, the resin state can be affected by environmental factors such as out-time and ambient humidity. Extended out-time can advance the degree of cure and viscosity of the resin and potentially prevent full infiltration of the fiber bed during cure. Exposure to ambient humidity generally leads to the rapid absorption of moisture. The dissolved water can then evolve during cure, causing the nucleation and growth of voids with internal pressures that exceed that which is imparted to the resin by vacuum bag-only processing. Moreover, moisture absorption has also been shown to affect process-relevant properties, such as degree of cure and viscosity, by acting as both catalyst and solvent for the amine-epoxy reaction and accelerating cross-linking. The present thesis focuses on developing a flow-enhanced cure cycle to limit flow-induced defects during OoA prepreg processing. ❧ First, a representative OoA resin (CYCOM® 5320-1, Cytec Industries Inc.) was characterized using conventional Modulated Differential Scanning Calorimetry (MDSC) and rheometry methods, and predictive models that comprehensively captured out-time effects on cure kinetics and viscosity in process conditions were developed. Polymerization at ambient temperatures altered the initial degree of cure of the resin as well as the evolution of cure rate and viscosity at elevated temperatures. Subsequently, in-situ dielectric monitoring experiments were used to determine the evolution of key dielectric properties during the same temperature and out-time conditions, and effective correlations were developed to obtain degree of cure and viscosity from these properties. Together, these methods were shown to constitute complementary methods for predicting and monitoring in-situ the “instantaneous” degree of cure, cure rate, and viscosity evolution of composite prepreg. ❧ Second, effective methods for monitoring out-time under ambient conditions, identifying critical physicochemical events (minimum viscosity, gelation, and vitrification) and their dependence on out-time (and out-life) for an OoA prepreg resin during cure were investigated. Conventional ex-situ methods (MDSC and rheometry) were used to collect benchmark data. Subsequently, in-situ dielectric analysis was conducted to gain a more detailed understanding of the physical phenomena involved, and to develop a means for detecting such phenomena in real-time. The results showed that while the out-life may correspond to the onset of pervasive porosity, it does not constitute a fundamental shift in resin properties. Rather, the processing-critical resin transitioned (minimum viscosity, gelation and vitrification) gradually but significantly changed during the out-life specified for the material. The results demonstrated that dielectric analysis can be used as a stand-alone technique to monitor both the extent of out-time at ambient temperature as well as the major effects of out-time during cure. ❧ Third, accurate process models that comprehensively capture out-time and humidity effects on cure kinetics and viscosity in process conditions were developed. DEA was used to further develop a method for accurate process monitoring. The results indicated that out-time and moisture absorption primarily affect the initial degree of cure and consequently influence the course of cure kinetics and viscosity evolution during cure. Tangibly, these changes decrease flow time, gelation time and reaction end time, and potentially complicate manufacturing. The production of high-quality parts using OoA prepreg therefore requires out-time and humidity control and/or appropriate thermal control to ensure adequate flow time and to fully impregnate the prepreg during processing. In addition, it was demonstrated that DEA allows accurate monitoring of initial degree of cure, gel-time, and reaction-end time for the entire range of relative humidity studied, confirming the robustness of the proposed methodology. ❧ Fourth, three aerospace grade resins were formulated to investigate the effects of variation in DDS isomers, amine to epoxy (a/e) stoichiometric ratio, out-time, and moisture absorption on processing characteristics and cured resin properties. Conventional thermochemical and thermomechanical methods (MDSC and rheometry) were used to collect benchmark data. Regardless of the resin system, the results showed that out-time increases initial degree of cure and more so with moisture absorption which influence the cure kinetics and viscosity evolution during cure. These effects were shown to shorten flow level and time which could potentially lead to insufficient resin flow during composite manufacturing. Then, accurate process models were developed that comprehensively captured out-time and humidity effects on cure kinetics and viscosity for each resin systems. Among the three resin systems investigated, meta-substituted 33DDS with a/e = 0.8 exhibited the highest cure rate and the lowest flow time. This behavior was attributed to the increase in collision number between amine and epoxy compared to a/e = 0.6 and the lack of delocalization of the lone pair of electrons on nitrogen compared to para-substituted 44DDS. Conversely, 44DDS with a/e = 0.8 exhibited the slowest cure rate with the highest flow time. In addition, 44DDS based resin exhibited the highest glass transition temperature due to having lower configurational entropy than 33DDS based resin. The results offered practical insights regarding resin formulation and prepreg processing, particularly that: (1) 33DDS-based resin is more susceptible to aging and flow time reduction, albeit with faster processing time

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Efficient manufacturing and repair of out-of-autoclave prepreg composites

PDF

Fabrication and analysis of prepregs with discontinuous resin patterning for robust out-of-autoclave manufacturing

PDF

Defect control in vacuum bag only processing of composite prepregs

PDF

Material and process development and optimization for efficient manufacturing of polymer composites

PDF

Chemical recycling of amine/epoxy composites at atmospheric pressure

PDF

Vacuum-bag-only processing of composites

PDF

Developing efficient methods for the manufacture and analysis of composites structures

PDF

In situ process monitoring for modeling and defect mitigation in composites processing

PDF

Void evolution in vacuum bag-only prepregs

PDF

In situ process analysis for defect control during composites manufacturing

PDF

Characterization, process analysis, and recycling of a benzoxazine-epoxy resin for structural composites

PDF

Processing and properties of phenylethynyl-terminated PMDA-type asymmetric polyimide and composites

PDF

Processing, thermal-oxidative stability and thermal cyclic fatigue of phenylethynyl-terminated polyimides

PDF

Polymer flow for manufacturing fiber reinforced polymer composites

PDF

Slurry based stereolithography: a solid freeform fabrication method of ceramics and composites

PDF

Comprehensive uncertainty quantification in composites manufacturing processes

PDF

Robust real-time algorithms for processing data from oil and gas facilities

PDF

The development of a hydraulic-control wave-maker (HCW) for the study of non-linear surface waves

PDF

Modeling and simulation of complex recovery processes

PDF

Kinetic studies of collisionless mesothermal plasma flow dynamics

Asset Metadata

Creator Kim, Daniel June (author)

Core Title Sustainable manufacturing of out-of-autoclave (OoA) prepreg: processing challenges

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Chemical Engineering

Publication Date 04/18/2017

Defense Date 03/09/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag composites,cure kinetics,dielectric analysis,epoxy resin,OAI-PMH Harvest,prepreg

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Nutt, Steven (committee chair), Armani, Andrea (committee member), Centea, Timotei (committee member), Williams, Travis (committee member)

Creator Email kim344@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-358200

Unique identifier UC11258102

Identifier etd-KimDanielJ-5209.pdf (filename),usctheses-c40-358200 (legacy record id)

Legacy Identifier etd-KimDanielJ-5209.pdf

Dmrecord 358200

Document Type Dissertation

Rights Kim, Daniel June

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA