University of Southern California Dissertations and Theses

Vacuum-bag-only processing of composites

(USC Thesis Other)

Vacuum-bag-only processing of composites

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

VACUUM-BAG-ONLY PROCESSING OF COMPOSITES

by

Shad Thomas

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

May 2009

Copyright 2009 Shad Thomas

ii
Dedication
This work is dedicated to my wife Sophea who stood by me through this endeavor.
Completion of this project would not have been possible without her support.

iii
Acknowledgements
First, I would like to thank my research advisor, Professor Steven R. Nutt for giving
me the opportunity to work in this new and exciting area of research. I would like to
thank him for his support and guidance in my work.
I would also like to thank my other committee members Professor Edward Goo and
Professor Charles Sammis for their time to serve on my committee and review my
dissertation. My appreciation is given to Warren Haby, the lab manager, for his advice
and encouragement.
Finally, I would like to express my appreciation to Cytec and Airbus for supporting
this project. My appreciation is extended to Jack Boyd and Chris Bongiovanni for their
recommendations.

iv
Table of Contents
Dedication

Acknowledgements

List of Tables

List of Figures

Abstract

Chapter 1: Introduction:
1.1: VBO Prepreg Development
1.2: Chapter 1 References

Chapter 2: Background
2.1: Ultrasound Method
2.2: Chapter 2 References

Chapter 3: Measurement of Resin Flow in VBO Prepregs by Ultrasonic
Imaging
3.1: Abstract
3.2: Introduction
3.3: Experimental
3.4: Results and Discussion
3.5: Conclusions
3.6: Chapter 3 References

Chapter 4: In Situ Estimation of Through-Thickness Resin Flow Using
Ultrasound
4.1: Abstract
4.2: Introduction
4.2.1: Summary of VBO Processing
4.2.2: In Situ Methods for Measuring Resin Flow
4.3: Experimental
4.4: Results and Discussion
4.5: Conclusions
4.6: Chapter 4 References
ii
iii
vi
vii
ix
1
1
4
5
5
8
9
9
9
10
12
14
15
16
16
16
16
17
21
26
33
35

v
Chapter 5: Temperature Dependence of Resin Flow in a Vacuum-Bag-Only
Process
5.1: Abstract
5.2: Introduction
5.3: Experimental
5.4: Results and Discussion
5.5: Conclusions
5.6: Chapter 5 References

Chapter 6: Effect of Fabric Architecture on Through-Thickness Permeability
in Multi-ply Laminates
6.1: Abstract
6.2: Introduction
6.3: Experimental
6.4: Results and Discussion
6.5: Conclusions
6.6: Chapter 6 References

Chapter 7: Modeling Resin Flow in VBO Prepregs
7.1: Abstract
7.2: Introduction
7.2.1: Permeability Models
7.2.2: Derivation of Flow Time
7.2.3: Method of Analysis
7.2.4: Materials
7.3: Results and Discussion
7.4: Conclusions
7.5: Chapter 7 References

Chapter 8. Recommendations for Future Works

Bibliography

37
37
37
39
42
54
55
57
57
57
60
63
72
74
76
76
76
78
80
82
83
84
93
95
97
98

vi
List of Tables
Table 4.1

Table 4.2

Table 4.3

Table 5.1

Table 5.2

Table 6.1

Table 6.2

Table 7.1

Table 7.2

Table 7.3

Table 7.4A

Table 7.4B

Table 7.5

Table 7.6

Table 7.7

Elliptical Parameters

Lamination Process Parameters

Flow Properties at 70
o
C Determined by C-Scan

Flow Properties at 50°C, 60°C, 70°C, & 80°C Determined by C-
Scan

Viscosity Data

Fabric weights and resin percentages

Resin Infusion Time of the Laminates (Minutes)

Fabric weight

Infusion times (minutes) measured for plain-weave fabric

Infusion times (measured in minutes) for 10-layer laminates at
65 °C

Values of φ
f
, L, D, and ΔP used in Equations 13 – 15

Values of T and μ used in Equations 13 – 15

Values of C
o
for each fabric

Calculated Resin Infusion Times (minutes)

V
z
, K
z
, K
c
, and (L
e
/L) for the 10-layer laminates made from each
fabric

21
22
30
48
48
62
69
83
84
84
85
85
88
88
89

vii
List of Figures
Figure 2.1

Figure 2.2

Figure 3.1

Figure 3.2

Figure 3.3

Figure 4.1

Figure 4.2

Figure 4.3

Figure 4.4

Figure 4.5A

Figure 4.5B

Figure 4.5C

Figure 5.1

Figure 5.2

Figure 5.3

Figure 5.4

Figure 5.5

Figure 5.6

Figure 5.7

Figure 5.8A

Gate set up for pulse-echo method

Initial condition of the fiber-film assemblies

Diagram of the vacuum bagging assembly

C-scans of the preliminary work on measuring resin flow in fabric

Microstructures of the initial and final conditions of resin flow
during the preliminary work

Diagram of elliptical fiber tow with coordinates

Schematic diagram of the vacuum bagging process for measuring
resin flow in fabric

C-scan images at 70
o
C of a vacuum bagged ply

Normalized reflected signal as a function of time at 70
o
C

Full vacuum for 3 minutes at 70
o
C

Full vacuum for 9 minutes at 70
o
C

Full vacuum for 15 minutes at 70
o
C

C-scan images at 50
o
C of a vacuum bagged ply

C-scan images at 60
o
C of a vacuum bagged ply

C-scan images at 70
o
C of a vacuum bagged ply

C-scan images at 80
o
C of a vacuum bagged ply

Normalized reflected signal as a function of time at 50°C, 60°C,
70°C, 80°C

X & Z Direction Flow Rates Vs. Inverse Viscosity at 50°C, 60°C,
70°C, and 80°C

Resin film laminated to the top of the fabric prior to impregnation

Full vacuum, 10 minutes, 60°C

6
7
12
13
14
20
24
27
28
31
31
32
44
44
45
45
46
49
49
50
viii
Figure 5.8B

Figure 5.8C

Figure 5.9

Figure 6.1

Figure 6.2

Figure 6.3A

Figure 6.3B

Figure 6.3C

Figure 6.4A

Figure 6.4B

Figure 6.4C

Figure 6.4D

Figure 6.5A

Figure 6.5B

Figure 7.1

Figure 7.2

Figure 7.3

Figure 7.4

Figure 7.5

Figure 7.6

Full vacuum, 20 minutes, 60°C

Full vacuum, 30 minutes, 60°C

C-scan images at 65°C of a 10-layer vacuum bagged laminate

Fabric architectures

4-Harness Satin Weave C-scans at 65 °C

Reflected intensities of the PW and 4-HSW samples

Reflected intensities of the 2 x 2 and 4 x 4 twill samples

Average reflected intensities of all samples

Macro and micro porosity in plain-weave laminate

Micro porosity in plain weave laminate

Ideal tow nesting in plain weave laminate

Non ideal tow nesting in 2 x 2 twill laminate

Nesting of tows in adjacent plies

Un-nested tows in adjacent plies

Diagram of the one-dimensional flow process

Images of the fabric architectures

Logarithmic relationship between measured infusion time t and
calculated value of η from Equation 14

Relationship between viscosity and temperature

Determination of the interaction parameter of plain-weave fabric

Cross-section of the fibers used to calculate the values of S, the
surface area per unit volume of fibers and ε, the fraction of
porosity

51
51
53
62
67
68
68
69
70
70
71
71
72
72
80
83
86
87
88
91

ix
Abstract
Ultrasonic imaging in the C-scan mode in conjunction with the amplitude of the
reflected signal was used to measure flow rates of an epoxy resin film penetrating
through the thickness of single layers of woven carbon fabric. Assemblies, comprised of
a single layer of fabric and film, were vacuum-bagged and ultrasonically scanned in a
water tank during impregnation at 50°C, 60°C, 70
o
C, and 80°C. Measured flow rates
were plotted versus inverse viscosity to determine the permeability in the thin film, non-
saturated system. The results demonstrated that ultrasonic imaging in the C-scan mode is
an effective method of measuring z-direction resin flow through a single layer of fabric.
The permeability values determined in this work were consistent with permeability
values reported in the literature. Capillary flow was not observed at the temperatures and
times required for pressurized flow to occur. The flow rate at 65°C was predicted from
the linear plot of flow rate versus inverse viscosity.
The effects of fabric architecture on through-thickness flow rates during
impregnation of an epoxy resin film were measured by ultrasonic imaging. Multilayered
laminates comprised of woven carbon fabrics and epoxy films (prepregs) were fabricated
by vacuum-bagging. Ultrasonic imaging was performed in a heated water tank (65
o
C)
during impregnation. Impregnation rates showed a strong dependence on fabric
architecture, despite similar areal densities. Impregnation rates are directly affected by
inter-tow spacing and tow nesting, which depend on fabric architecture, and are indirectly
affected by areal densities.
A new method of predicting resin infusion rates in prepreg and resin film infusion
processes was proposed. The Stokes equation was used to derive an equation to predict
x
the impregnation rate of laminates as a function of fabric architecture. Flow rate data
previously measured by ultrasound was analyzed with the new equation and the Kozeny-
Carman equation. A fiber interaction parameter was determined as a function of fabric
architecture. The derived equation is straight-forward to use, unlike the Kozeny-Carman
equation. The results demonstrated that the newly derived equation can be used to
predict the resin infusion rate of multilayer laminates.

1
Chapter 1: Introduction
1.1: VBO Prepreg Development
Thermosetting resins used in high-performance composite laminates can be generally
classified as addition cured and condensation cured. During condensation curing, volatile
reaction gasses are produced, and the vapor pressure of these gasses exceeds atmospheric
pressure. Consequently, porosity will result in the cured laminates from the evolved
gasses unless vaporization is suppressed. One reason autoclave processing is widely used
to cure composite laminates is to apply sufficient pressure to the prepregs (greater than
the vapor pressure of any gas evolved during curing) to suppress the evolution of
volatiles.
A new kind of prepreg has been introduced that does not require autoclave pressures.
Unlike conventional prepregs, VBO prepregs (for Vacuum Bag Only), employ thermoset
resins that are addition-cured, and thus, no gasses are evolved during the curing reaction.
Furthermore, these addition-cured resins flow under a pressure gradient of 1 atm or less,
and require only modest pressures for processing (thus the term VBO).
The evolution of VBO prepreg processing can be traced through a scant body of
literature. Palmer developed a vacuum bag process that allowed a resin film to infuse
into and saturate a fibrous reinforcement, flow through a porous release film, saturate a
bleeder ply, and outgas under vacuum bag pressure [1]. Parts were subsequently cured in
an autoclave. Johnson and Newsam modified Palmer’s process by using a release liner
that was impermeable to liquid resins but not gasses [2]. The parts were oven-cured
under vacuum pressure. This small but significant modification resulted in an early VBO
process.
2
Thorfinnson and Biermann established the importance of prepreg vacuum channels
for producing laminates free of porosity [3]. Their work focused on a class of epoxy
resin prepregs developed in the early 1980s. These prepregs were based on an epoxy
formulation designed for curing either at 177°C (using a traditional autoclave process), or
at lower temperatures using a VBO process. For example, one such prepreg (Cycom
950-1, American Cyanamid Company) was designed for curing by autoclave processing
at 177
o
C or by VBO processing at 121
o
C. Both processes resulted in laminates with
roughly equivalent properties [4]. In related work, Brietigam et al reported preliminary
development of a VBO epoxy resin for composite materials [5]. The processing time of
the VBO prepregs was reduced by as much as 75% compared to traditional autoclave-
processed prepregs. In addition, the mechanical properties of the VBO laminates, after a
post-cure thermal treatment, were equivalent to those of laminates processed in the
autoclave [5].
Two common methods have been employed to produce VBO prepregs. One method
involves partial impregnation of a fabric or unidirectional fibers with a resin film. In this
method, the partial impregnation is practiced to leave dry spaces or channels in the
prepregs. During subsequent compaction, these channels facilitate airflow in a direction
parallel to the laminate plane [6, 7, 8]. The second method involves use of a perforated
resin film that results in a prepreg with gaps in the resin film. This method allows gas to
flow out of the laminate in directions parallel and perpendicular to the plane of the
prepreg [9]. Recent published reports indicate that the former edge-breathing method is
the more common method of producing VBO laminates [6, 7, 8, 9].
3
Regardless of the method used to facilitate out gassing of the prepregs, R&D and use
of VBO prepregs will undoubtedly increase in the next several years. Growth of the
composites industry is largely due to extensive use of composites by Boeing and Airbus
in their current and next generation of passenger aircraft. Airbus plans to increase
production to fifty airplanes per month in 2009. The aerospace composites market is
expected to quadruple within the next twenty years [10]. Autoclave processing of
composites is too slow and expensive to meet the current and future production rates.
Out-of-autoclave processing such as the VBO method will be one of the primary
aerospace composites manufacturing methods used in the near future.

4
1.2: Chapter 1 References
[1] Palmer, RJ. Resin impregnation process. US Patent # 4,311,611, January 19, 1982.

[2] Johnson, FC, Newsam, SM. Method of manufacturing articles from a composite
material. US Patent # 4,562,033, December 31, 1985.

[3] Thorfinnson, B, Biermann, TF. Degree of impregnation of prepregs – effect on
porosity. 32nd International SAMPE Symposium, 1987; April 6-9:1500–1509.

[4] Hirschbuehler, KR. An advanced composite resin offering flexibility in processing
conditions. 37th International SAMPE Symposium, 1992; March 9 – 12:452–461.

[5] Breitigam, WV, Bauer, RS, May, C. Novel processing and cure of epoxy resin
systems. Polymer 1993; 34 (4):767–71.

[6] Hartness, JT, Xu, GF. Resin composition, a fiber reinforced material having a
partially impregnated resin and composites made therefrom. US Patent # 6,139,942,
October 31, 2000.

[7] Xu, GF, Repecka, L, Mortimer, S, Peake, S, Boyd, J. Manufacture of void-free
laminates and use thereof. US Patent # 6,391,436, May 21, 2002.

[8] Hartness, JT, Xu, GF. Resin composition, a fiber reinforced material having a
partially impregnated resin and composites made therefrom. US Patent # 6,565,944, May
20, 2003.

[9] Steele, M, Corden, T. New prepregs for cost effective out-of-autoclave tool and
component manufacture. SAMPE Journal 2004; 40 (2): 30–34.

[10] Michaels, K. Market trends: aerospace composites market will quadruple by 2026.
High Performance Composites. 2007; January 1.

5
Chapter 2: Background
2.1 Ultrasound Method
Ultrasound imaging is routinely used for non-destructive inspection of materials.
Sound waves in the 1-25 megahertz range are used for non-destructive imaging,
depending on the material and its thickness. Ultrasound emitted from a piezoelectric
transducer interacts with the material being inspected and then is detected when it exits
the material. A change in material properties causes an abrupt change in the amplitude of
the wave observed on an oscilloscope or computer monitor. Ultrasound in the megahertz
range does not travel through air or vacuum. Defects such as cracks, voids, and porosity
impede transmission of sound waves used for inspection. Two common techniques used
for inspection are transmit-receive and pulse-echo or transmit-reflect [1,2].
In the transmit-receive method, ultrasound waves emitted from a piezoelectric
transducer travel through the material being inspected and are detected on the opposite
side. The transmitter and receiver are in line with each other and are on opposite sides of
the material inspected. If the sound waves were detected on the same side of the
material that they were transmitted from, then the signal would be attenuated twice, once
on the entrance path, and again on the reflection path. The transmit-receive method is
preferred for thick materials that absorb up to 50% of the incident ultrasound signal as the
sound waves pass through the sample. Another common method of ultrasound inspection
is the pulse-echo method. The pulse-echo method is simpler to set up and operate than
the transmit-receive method and is preferred for thin materials.
When sound waves travel through a material and reach an interface with a different
material, a fraction of the energy will be transmitted through the second material and the
rest of the energy will be reflected back through the first. The fractions of transmitted
and reflected energy depend on the acoustic impedance of materials on both sides of the
interface. In the pulse-echo method, sound waves travel through the material inspected,
reflect at its back side, exit at the entrance point, and travel back to the transducer. The
transducer serves as both the transmitter and receiver in the pulse-echo method. If the
properties of the material under inspection are constant throughout, then a signal of
constant intensity is reflected back to the transducer and no defects are observed. If the
properties are not constant during the inspection, then the intensity of the reflected signal
will change, indicating non-uniform properties. Signal is recorded by setting a time gate
to record all reflections that occur between the top reflection of the sample and the
bottom reflection of the base plate. This is illustrated in Figure 2.1.

Internal
reflections

Back
reflection
Top
reflection

Amplitude
Time

Figure 2.1: Gate set up for pulse-echo method.
The ultrasound signal is attenuated by the materials used in the vacuum bagging
process. All of the materials except the fiber-film assemblies used in the vacuum-bagging
process have constant properties. Therefore, the only material that causes the ultrasound
signal to change with time is the fiber-film assembly. The initial condition of the fiber-
6
film assembly is shown in Figure 2.2. An increase in the ultrasound signal is observed as
resin gradually flows into the fabric during the impregnation cycle, reducing the porosity.
This change is detected by ultrasonic imaging in the pulse-echo mode. The signal stops
changing when the resin stops flowing. An in situ method of measuring the flow of resin
in a porous material such as carbon fabric was described and used in this research project.

100 μm
Carbon Fabric
Resin Film
Figure 2.2: Initial condition of the fiber-film assemblies.

7
8
2.2 Chapter 2 References
[1] ASM Handbook, Nondestructive Evaluation and Quality Control, 1990, V. 17, 231 –
261.

[2] Krautkrämer J. and Krautkrämer K., Ultrasonic Testing of Materials, 1990, 5th ed.,
167 – 221.

9
Chapter 3: Measurement of Resin Flow in VBO Prepregs by Ultrasonic Imaging
3.1: Abstract
Ultrasonic reflectivity and imaging in the C-scan mode was used to measure the flow
rate of resin in a low-pressure, low-temperature, vacuum-bag-only (VBO) prepreg at
constant temperature and pressure. The prepreg was vacuum-bagged and scanned in a
water tank at 60
o
C. Resin flow was monitored continuously by C-scan images, and
complete impregnation (cessation of resin flow) was determined by when the last image
stopped changing. The results demonstrate that ultrasonic reflectivity and imaging in the
C-scan mode is an effective method for measuring resin flow through a single layer of
fabric. Comparison of ultrasonic and microscopy images demonstrated consistency and
the effectiveness of ultrasonic imaging as an in situ process diagnostic for monitoring
impregnation.
3.2: Introduction
Composite materials are commonly used in applications that require high strength and
light weight. The aerospace industry continues to be the main user of composites.
Traditional fabrication of high-strength composite parts is done in an autoclave above
100
o
C. The purpose of the high pressure in the autoclave is two-fold. First, the high
pressure drives the resin into the fiber tows of the prepreg. Second, the pressure in the
autoclave during the impregnation and cure cycles exceeds the vapor pressure of any
entrapped gas or volatile material evolved during the cure process, producing a composite
free of porosity. Autoclaves are considerably more expensive than ovens of the same
size due to the high-pressure capabilities of the autoclave compared to the oven.
10
VBO composites were developed as a low cost alternative to autoclave cured
composites. Composites made from VBO prepregs are typically vacuum bagged and
oven cured at temperatures below 100
o
C followed by a post cure between 175
o
C to
205
o
C. The post cure is designed to increase the glass transition temperature [1, 2, 3, 4]
Ultrasonic imaging in the C-scan mode is routinely used to detect internal defects
such as porosity and delaminations. Ultrasound in the megahertz range is attenuated by
low-density regions of the composite, allowing defects to be detected. Prepregs, by
design, are very porous in their initial conditions. Porosity decreases as resin
impregnates the fiber tows of the prepregs. Ultrasound imaging techniques have been
used to measure resin flow rates in both RTM and VBO processes [5, 6]. In this paper
we will demonstrate the method of measuring resin impregnation of a prepreg by
monitoring ultrasonic reflectivity.
3.3: Experimental
Experiments were conducted using plain-weave carbon fabric with 3000 fibers per
tow (density = 196 grams per square meter) and an epoxy resin film (Cytec 5215, density
= 102 grams per square meter). The resin film was laminated to the fabric at 1500 Pa and
50
o
C for 2 seconds. A small section was cut from the laminated sample, then cured,
mounted, and polished. The penetration depth of the resin in the cured prepreg was
measured microscopically.
Next, the prepreg was vacuum bagged with the resin film on top of the fabric as
shown in Figure 3.1 (not to scale). An aluminum plate free of scratches and other visual
imperfections was used as the base plate for vacuum bagging.
11
Ultrasonic imaging of the vacuum bagged prepreg was performed in an aluminum
water tank (60 cm long x 45 cm wide x 15 cm deep). The A-wave and scanning
parameters for ultrasonic imaging were set up on a flat region of the vacuum bagged
assembly near the prepreg. A support fixture was used that allowed the vacuum bagged
assembly to be removed and re-inserted into the water tank at the same position. An area
of 127 mm x 127 mm near the center of the prepreg was scanned in 12 minutes at 20
o
C to
document the initial condition.
The vacuum bagged assembly was removed from the water tank while the water was
heated. Four electric hot plates under the water tank were used to heat and maintain the
water at 60
o
C ± 1
o
C. The intensity of the signal from the transducer decreased as the
temperature increased. Once the temperature stabilized, the gain was increased to bring
the intensity of the A-wave back to its value at 20
o
C. The vacuum bagged sample was re-
inserted into the water tank after the temperature stabilized. Flow of the resin through the
carbon fabric was measured at 60
o
C ± 1
o
C. The scans began at the lower left corner and
ended at the upper left corner of the prepreg. Twelve minutes was required to complete
one C-scan cycle. The C-scans were repeated until the reflectivity of the image changed
by 1% or less, indicating that resin stopped flowing. A mechanical pump was used to
apply a constant pressure of one atmosphere to the vacuum bagged prepreg during the
flow measurements. When flow stopped, the vacuum bagged sample was removed from
the water tank, quenched in cold water, vented, and cured. The cured prepreg was then
cross sectioned, mounted, polished, and examined microscopically to determine the
extent of resin flow into the fiber tows.

a
12

b d e
g c f

Figure 3.1: Diagram of the vacuum bagging assembly for measuring resin flow by
ultrasonic imaging. a. Tacky tape, b. Aluminum plate, c. Vacuum bag, d. Prepreg, e.
Fiberglass breather, f. Vacuum valve, g. Thermocouple.
3.4: Results and Discussion
Ultrasonic images of the initial condition and the conditions at 2 – 14 minutes, 16 –
28 minutes, and 30 – 42 minutes are shown in Figure 3.2. The two-minute delay between
each scan was due to the time required to save the image and start the next scan. The
plain-woven fabric is not uniform in thickness. Gaps between the fiber tows appear
green and thin edges of the fiber tows appear yellow in the initial condition. The initial
condition of the prepreg was porous due to the resin film lightly laminated to the top of
the carbon fabric, indicated by the predominant red color. Flow of resin into the fabric
can be seen in the next three images as the color gradually transforms from red to yellow
to green. Thicker regions absorb more of the ultrasound signal than thinner regions,
appearing yellow in the final image. Thus, the C-scan image of the fully impregnated
fabric is not entirely green. The final C-scan image in Figure 3.2 indicates that the resin
stopped flowing at approximately 30 minutes. The reflectivity reached a maximum value
at 30 minutes and remained constant during the rest of the scan, indicating that the resin
stopped flowing.
13
Microstructures of the initial and final conditions of the prepreg (before and after
impregnation) are shown in Figures 3.3A and 3.3B. The prepreg lightly laminated to the
top of the fabric is shown in Figure 3.3A. The prepreg fully impregnated with resin is
shown in Figure 3.3B. Figures 3.2A and 3.3A correspond to each other, and Figures
3.2D and 3.3B also correspond to each other.

Figure 3.2: A. Initial Condition (upper left), B. 2 – 14 minutes (upper right),
C. 16 – 28 minutes (lower left), D. 30 – 42 minutes (lower right).

2 cm

100 μm 100 μm
Figure 3.3: A. Initial condition, resin film on top of fabric (left), B. Final
condition after 42 minutes, resin film fully impregnated the fabric (right).
3.5: Conclusions
Ultrasonic reflectivity and imaging in the C-scan mode can be used to determine
when prepregs are fully impregnated with resin during VBO processing. Laminates
containing more than one layer can be imaged by C-scan during the infusion process to
determine when resin stops flowing. The vacuum bagging procedure for ultrasonic
imaging must be modified from the normal production procedure by eliminating the
upper release liner, breather cloth, and pressure plate so only vacuum bag film is above
the prepreg. Data analysis on the large area scans (12.7 cm x 12.7 cm) is complicated by
the long time (12 minutes) to complete the scans.

14
15
3.6: Chapter 3 References
[1] Xu G.F., Repecka L., and Boyd J., SAMPE International Symposium, 1998; 43: 9-19.

[2] Mitani K. and Wakabayashi K., International SAMPE Symposium, 2001; 46: 2293-
2302.

[3] Jackson K. and Crabtree M., International SAMPE Symposium, 2002; 47: 800-807.

[4] Repecka L. and Boyd J., International SAMPE Symposium, 2002; 47: 1862-1875.

[5] Adison R.C., McKie A.D.W., Liao T.L.T., and Ryang H.S., IEEE Ultrasonics
Symposium, 1992; 783-786.

[6] Thomas, S, Nutt, SR, and Bongiovanni, C. In situ estimation of through-thickness
resin flow using ultrasound. Submitted to Composites Science and Technology.

16
Chapter 4: In Situ Estimation of Through-Thickness Resin Flow Using Ultrasound
4.1: Abstract
Ultrasonic imaging in the C-scan mode was used to measure the flow rate of an epoxy
resin film penetrating through the thickness of a single layer of woven carbon fabric.
Assemblies, comprised of a single layer of fabric and film, were vacuum-bagged and
ultrasonically scanned in a water tank during impregnation at 70
o
C. The permeability of
the fabric was calculated using Darcy’s law. The results demonstrated that ultrasonic
imaging in the C-scan mode is an effective method of measuring z-direction resin flow
through a single layer of fabric. Comparison of ultrasonic and microscopy images
yielded consistent results and demonstrated the effectiveness of ultrasonic imaging as an
in situ process diagnostic for monitoring through-thickness impregnation and flow rates.
4.2: Introduction
4.2.1: Summary of VBO Processing
Traditional processing of high-strength polymer matrix composite materials requires
autoclave consolidation and curing at high pressures and temperatures (pressures of
several atmospheres and temperatures above 100°C). A new class of vacuum-bag-only
(VBO) prepregs has been introduced for processing at much lower pressures and
temperatures. VBO prepregs are intended to yield autoclave-quality parts without the use
of an autoclave, and are thus a member of the family of out-of-autoclave (OOA)
techniques for composite manufacture. In the VBO process, composite laminates are
produced from prepregs by vacuum-bag consolidation followed by curing in an oven at
atmospheric pressure. Like resin film infusion (RFI), the VBO prepreg is produced by
impregnating dry fabric with a resin film. However, unlike RFI, the VBO prepreg is
17
produced by partial impregnation, leaving vacuum channels for air escape during
subsequent consolidation in a vacuum bag. Eliminating autoclave processing simplifies
the manufacturing process and greatly reduces operational and capital equipment costs.
One purpose of the autoclave is to apply pressure greater than the vapor pressure of
volatile components. Preventing gasses from evolving from the resin during the cure
cycle produces a dense structure free of porosity. In RFI and VBO processing, unlike
autoclave processing, there is no high pressure to force volatile gasses to remain
dissolved in the resin, and the maximum applied pressure is only 1 atm. Therefore, all of
the gasses evolved must be removed prior to resin curing, and the cure process cannot
produce volatile reaction products. The maximum fiber volume fraction of a composite
laminate fabricated from VBO prepregs is limited by the low pressure that is
characteristic of the process. Fiber volume fractions s up to 65% for unidirectional
composites and 52% - 57% for composites made with fabrics have been achieved [1].
4.2.2: In Situ Methods for Measuring Resin Flow
Understanding resin flow and impregnation of the reinforcing medium is critical to
the manufacture of consistent composite laminates free of porosity and flow-related
defects. Although resin flow measurements are generally not performed in prepreg
processes, they are routinely performed in resin transfer molding and related liquid
molding processes. For example, fiber optic sensors, used in the transmission or
reflection modes and embedded at various locations in the layers of reinforcing material,
have been used to measure resin flow in resin transfer molding (RTM) processes [2, 3].
Optical fibers can be fragile, however, and fiber optic sensors are not as sensitive to resin
flow inside fiber tows, so alternative approaches are often employed.
18
Resin flow in RTM processes is often measured by employing a transparent mold,
allowing one to visually monitor and record the advance of resin flow fronts as the mold
is filled [4, 5, 6, 7, 8]. In this approach, the flow rate through and around the stack of
fabric or plies is measured directly, while the flow rate through the fiber tows is measured
indirectly by a non-linear pressure-time profile [4]. The video recording method works
well for measuring the bulk flow rate in in-plane RTM processes, in which the mold is
initially dry (provided also that the mold is transparent). However, video recording
methods are not useful for measuring flow rates in prepregs, where impregnation occurs
primarily in the through-thickness direction.
Microstructural analysis of composites produced by RTM has also been used to study
impregnation kinetics and to provide a basis for model simulations [9]. This work
provided the basis for a permeability model for RTM processes [4], as well as two-
dimensional elliptic flow models [10, 11]. Attempts to measure transverse flow have met
with limited success. Wu, Li, and Shenoi demonstrated a simple method of measuring
transverse flow in which fibrous material was wound in a coil and resin was injected
through the center [12]. By observing the flow front, the flow rate and permeability
could be determined. However, if gaps existed between fiber tows, as in most woven
fabrics, the rapid flow of resin through the gaps prevented determination of flow rates
and permeability. Using a different approach, transverse or through-thickness flow
measurements have been performed by pumping fluid through a porous sample at a
prescribed rate. Permeability was determined from a linear plot of the pressure drop as a
function of flow rate [13, 14].
19
An alternative approach to measuring resin flow during impregnation involves the use
of ultrasonic imaging, which has been used in the transmission mode to study the flow
front of resin in an RTM process [15]. The samples were cured at different stages of the
impregnation process and imaged to determine the flow pattern in an RTM process. In
other work, ultrasonic transmission with air coupling was used to measure the
impregnation and fill rate of the preform in a resin transfer molding process [16]. The
ultrasonic transmitter and receiver were aligned on opposite sides of the mold. The mold
fill rates determined ultrasonically were equivalent to those determined visually. This
ultrasonic transmission method is well-suited to situations in which the transducers
cannot be immersed and do not require immersion in a coupling fluid. The ultrasound
signal resolution achieved with air coupling is generally inferior to liquid coupling,
however [17].
In the present work, we utilize ultrasonic imaging and reflectivity to monitor the
impregnation of fabrics in situ. (Reflectivity is the amplitude of the reflected signal.) An
assembly consisting of fabric, resin film, and vacuum bag is immersed in a heated water
tank while performing C-scan imaging to produce a density map or profile of the scanned
material. The method relies on the acoustic contrast provided by scattering from wet and
dry regions of the fabric. As the resin flows into the fabric and permeates the fiber tows,
the volume of the material system decreases and the density increases [18]. The volume
and density of the material stop changing when the resin stops flowing, and this is
detected in sequential C-scan images, which cease to change when fiber wetting is
complete and resin flow ceases. The procedure provides a non-invasive, in situ
inspection method for monitoring resin impregnation of fiber tows.
The capabilities and limitations associated with this technique are considered and
analyzed. The primary intent is to demonstrate the utility and limitations of the method
for monitoring through-thickness resin flow in RFI, VBO prepregs, and other non-
autoclave processes – not to investigate the effects of process parameters or to detect
micro-porosity. (Future work will address the influence of process parameters on resin
flow, such as temperature, multiple layers of prepreg, and fiber architecture). The
ultrasound data are used to calculate permeability using Darcy’s law.
Darcy’s law relates the flow rate of a viscous fluid to the permeability of the porous
material. The flow rate V (m/s), permeability, K (m
2
), pressure gradient, ΛP (Pa/m), and
viscosity μ (Pa*s) are related by
P
V
K
∇
=
μ

Flow into the fiber tows is two-dimensional. The cross-section of the fibers tows is
approximately elliptical and the relationship between elliptical and Cartesian coordinates
is well established [19, 20]. A diagram of an elliptical fiber tow is shown in Figure 4.1.

20

p
-L +L a
b
Figure 4.1: Diagram of elliptical fiber tow with coordinates.
The sum of distance from the positive and negative values of the foci (L) to any point (p)
on the surface of the ellipse is equal to 2a. The major and minor coordinates of the
ellipse (a,b) and their relation to the Cartesian coordinates (x,y) are defined as follows.
) cosh( ξ L a = , ) sinh( ξ L b = , ) cos( η a x = , and ) sin( η b z =
ξ= ( Lp + + Lp − )/2L = a/L, η = ( Lp − - Lp + )/2L
Values of the elliptical parameters are listed in Table 4.1.

Table 4.1: Elliptical Parameters
a (mm) b (mm) ±L (mm)
ξ
η X (mm) Z (mm)
0.825 0.075 ± 0.822 1.00 0.60 1.05 0.55

4.3: Experimental
The materials selected for the study included a plain-weave carbon fabric (elastic
modulus and tensile strength of fibers = 231 GPa and 3650 MPa) containing 3000 fibers
per tow (density = 194 grams per square meter) and an epoxy resin film (Cycom 5215,
density = 102 grams per square meter). A single layer of resin film on a single ply of
woven carbon fabric was used in all trials. A two-step process was employed in which the
resin film was first attached to the fabric by lamination, which was followed by
impregnation in a vacuum bag assembly. The lamination procedure involved removing
the release liner from the resin film, placing the resin film and protective paper layer on
top of the fabric, placing the laminate on a Teflon
®
coated flat surface preheated to 50°C,
applying a pressure of 1500 Pa (15mbar) for 2 seconds, rapidly cooling to room
temperature, and removing the protective paper. The lamination process was followed by
21
22
the second step of the process - full impregnation by vacuum bagging at 70°C.
Preliminary experiments were conducted to determine the depth to which the resin film
penetrated the carbon fabric during the lamination step (prior to impregnation).
Lamination: Sections of resin film, 254 mm x 254 mm, were cut from a large roll and
lightly pressed onto a piece of plain-weave carbon fabric of equal dimensions at room
temperature. Next, lamination was performed using all combinations of the process
parameters shown in Table 4.2. An alternative lamination procedure involved ironing the
resin film onto the fabric, although pressure, temperature, and time were not determined
during the ironing process.
Laminated samples were cured, sectioned, and polished prior to microscopic
examination to determine the resin penetration depth during lamination. When the resin
film was laminated to the fabric at 1500 Pa (15 mbar) and 50°C for 2 seconds, resin flow
through the fabric and into fiber tows was minimal. These conditions were used for
lamination of all samples. In samples thus produced, the resin film was lightly laminated
to one side of the 0.254 mm thick carbon fabric.

Table 4.2: Lamination Process Parameters
P: (Pa) 1500 15000 150000
T: (
o
C) 50 60 70
t: (s) 2 10 30

Capillary Flow Measurements: The release liner was removed from a 100 x 100 mm
sheet of resin film that was lightly pressed onto a 100 x 100 mm piece of carbon fabric at
23
room temperature. The sample was placed in an oven for 20 minutes preheated to 70
o
C
without applied pressure. The samples were cured, sectioned, mounted, polished, and
examined. Microscopic examination revealed that the resin did not flow into the fabric,
indicating an absence of capillary flow.
Impregnation: RFI samples were vacuum-bagged, out-gassed, and heated under full
vacuum of 100 kPa (1000 mbar) for 3, 9, and 15 minutes in the water tank at 70°C. Flow
of the resin was halted by immediately removing the samples from the heated water tank,
venting the vacuum-bagged samples, and quenching in cold water. The samples were
cured, and polished sections were examined by light microscopy.
Measuring Flow: Ultrasonic imaging was performed to measure the extent of resin flow
through a single-ply RFI sample during the impregnation process. The experimental
configuration consisted of a thin epoxy resin film laminated to a single ply of woven
carbon fabric. The fabric rested on an aluminum base plate, and the fabric-film assembly
was sealed in a vacuum bag, as shown in Figure 4.2.

24

7
6
4
5
2
3
1
Figure 4.2: Diagram of the vacuum bagging process for measuring resin flow by
ultrasonic imaging. The assembly consists of: (1) aluminum base plate, (2) sealant tape,
(3) Teflon
®
film, (4) breather cloth, (5) RFI sample with resin film on top of the fabric,
(6) vacuum bag film, and (7) vacuum valve.

The resin film comprised 34% by weight of the fabric-resin system. The assembly was
immersed in a heated water tank designed for ultrasonic C-scan imaging, and imaging
was performed to monitor the impregnation while vacuum of 100 kPa (100 mbar) was
applied to the assembly. A commercial ultrasound system (Physical Acoustics Ultrapac
II) was used for the ultrasonic imaging. A 0.5 inch diameter (12.7 mm) 10 MHz focused
transducer (focal length 38.1 mm, focal point 1 micrometer) was used for scanning in the
reflected mode. The beam diameter at half a wavelength is 24 micrometers. The gates
were set to record signal between the top of the sample and bottom of the base plate. A
constant signal was produced from the base plate; so that any change in the signal
25
resulted from a change in the sample. The amplitude of the signal was set to 90% of the
maximum on the base plate.
First, the assembly was placed in a water tank (60 cm long x 45 cm wide x 15 cm
deep) at 20
o
C to measure the initial condition. The A-wave scanning parameters for
ultrasonic imaging were established on a flat region of the vacuum bagged assembly near
the ply. The signal amplitude was plotted against time. A support fixture was used for
the vacuum bagged assembly that allowed removal and re-insertion in the water tank at
the same position. A 25.4 mm x 25.4 mm area near the center of the ply was scanned in
the C-scan mode in one minute at 20
o
C to document the initial condition. The assembly
was removed from the water tank after the initial condition was recorded.
In preparation for flow measurements, the water tank was heated to the desired
temperature and maintained to within ± 1
o
C. The vacuum-bagged sample was re-inserted
into the heated water tank. Flow of the resin through the carbon fabric was measured at
70
o
C at three-minute intervals until the C-scan image was unchanged from the previous
interval (see below). Measurements of the initial condition at 20
o
C and flow of the resin
at 70
o
C were repeated on three samples. A constant pressure of 100 KPa was applied to
the vacuum-bagged plies during the flow measurements using a mechanical pump. A
thermocouple was embedded under the vacuum-bagged sample. The samples reached 70
o
C in approximately 40 seconds after immersion into the heated water tank and the
temperature remained constant during the flow measurements.
The amplitude of the reflected signal was monitored and plotted as a function of time
during impregnation. The cessation of resin flow was determined by comparing
successive scans, and when the reflectivity of the last two consecutive scans changed by
less than one percent, resin flow was taken to be complete. Permeability was then
calculated from Darcy’s law, the general form of which is given by
P
V
K
∇
=
μ
.
The pressure gradient ΛP, in two dimensions is approximated as ΔP/x + ΔP/z [11, 14].
The C-scan images are affected by multiple factors, including density and thickness
variations of the fabric, porosity, temperature, and the gain setting. For example, the
amplitude of the signal decreases as the temperature increases. Once a constant
temperature is reached, the gain must be increased to an appropriate setting. If the gain is
not set appropriately, then the gradual change in the ultrasound reflectivity caused by
permeating resin will not be detected. Note that the amount of porosity at any given time
is not determined from the C-scans, nor was that the intent of this investigation. Instead,
the purpose was to measure average flow rates by determining when the resin flow
ceased.
4.4: Results and Discussion
Representative C-scan images recorded at 70
o
C are shown in Figure 4.3. Red
indicates a reflected signal intensity of approximately 10%, yellow - approximately 50%,
and dark green - approximately 90%. Intermediate colors between red and yellow
indicate reflected intensities between 10% and 50%, whereas intermediate colors between
yellow and dark green indicate reflected intensities between 50% and 90%. The
uniformly spaced green regions of the initial scans correspond to inter-tow gaps in the
woven fabric. The resin flows relatively quickly between tows and fills these gaps
26
(macro-flow), then begins permeating the fiber tows, saturating the tow edges first, where
the tows are thinnest (the tows are oblate ellipses in cross-section). After penetrating
through the inter-tow gaps, the resin flows simultaneously from the top and bottom of the
fabric.
The initial condition of Figure 4.3 shows that the fabric is not entirely uniform.
Thicker regions appear red, thinner regions yellow, and gaps between the fiber tows
appear light green. The individual fiber tows are naturally thickest at the centers and
thinnest at the edges. The non-uniformity of thickness and density of the fabric is
apparent throughout the series of scans. Thicker regions absorb more of the ultrasonic
signal than thinner regions. Therefore, one should not expect the C-scan of a fabric fully
impregnated with resin to appear dark green. In addition, regions of higher density
reflect more of the signal than regions of lower density. Thus, C-scan images provide a
qualitative measure of resin flow and determination of when resin permeation ceases.
1 cm
3 minutes
Initial condition
9 minutes
18 minutes

Figure 4.3: C-scan images at 70
o
C of a vacuum bagged ply.
27

Flow at 70
°
C: Reflectivity measurements at 70°C show that the resin stopped flowing
after 15 minutes in samples 1 and 3 and after 18 minutes in sample 2, yielding an average
flow time of 16 minutes ( ± 1.73 minutes). Figure 4.4 shows plots of normalized
reflectivity versus time for the three samples. The greatest change in reflectivity takes
place during the initial few minutes, when inter-tow “macro” flow occurs. Subsequently,
the resin penetrates the tows (“micro” flow), resulting in gradual increases in ultra-sound
reflectivity. Impregnation was considered complete when the reflectivity in sequential
scans varied by less than 1%.
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
90.0
100.0
0 3 6 9 12 15 18 21
Time (minutes)
Normalized Reflectivity
70C, #1
70C, #2
70C, #3

Figure 4.4: Normalized reflected signal as a function of time at 70
o
C

The three profiles are similar, with only minor variations. Variability of the flow rate is
attributed to non-uniform distribution of the fibers in the fabric, a consequence of fabric
distortion during RFI sample fabrication. The standard deviation of the flow time is
28
approximately 11% of the mean value. Based on the 95% confidence interval for a
sample size of 3, ( x ± 4.303s
m
), the resin should completely fill the fabric within 16 ± 4
minutes. Achieving full impregnation of the fabric in the shorter time interval is unlikely
because complete permeation of the fabric was not observed in less than 15 minutes.
Furthermore, the resin viscosity increases with time, albeit quite slowly. (The gel time is
5-6 hours at 70°C, so the change in viscosity in the first 20 minutes was assumed to be
negligible.)
Flow rates measured from the reflectivity data are summarized in Table 4.3.
Permeability values for the fiber tows calculated from Darcy’s law are also summarized
in Table 4.3. Note that these values are not necessarily equivalent to bulk permeability
values. Bulk permeability cannot be determined for a single layer, because the Teflon
®

film on the bottom of the carbon fabric provides a flow resistance different from an
adjacent ply of carbon fabric. However, bulk permeability in laminates made from
prepregs is less relevant than tow permeability, primarily because the region surrounding
the prepregs is saturated with resin when the laminate is laid up and fiber nesting reduces
the likelihood of contiguous through-plane porosity. While bulk permeability is
somewhat important for air removal through resin bleeding, air is normally removed from
the laminate by application of vacuum during vacuum bagging. Thus, vacuum-bag
processing of prepregs relies on removing all of the air and vapor from the laminate
before the vacuum channels are sealed off by flowing resin.

29
Table 4.3: Flow Properties at 70
o
C Determined by C-Scan.
Property # 1 # 2 # 3
x σ
Flow Time (min.) 18 15 15 16 1.73
V
x
(10
-7
m/s) 9.72 11.7 11.7 11.0 1.14
V
z
(10
-7
m/s) 5.09 6.11 6.11 5.77 0.60
K
x
(10
-13
m
2
) 2.56 3.08 3.08 2.91 0.30
K
z
(10
-14
m
2
) 7.03 8.43 8.43 7.96 0.81
Viscosity at 70
o
C = 25.1 Pa*s, ΔP = 100000 Pa

Figures 4.5A and 4.5B shows polished sections of vacuum-bagged samples removed
from the 70
o
C water bath after 3 and 9 minutes. The original position of the resin film
was on one side (top) of the fabric. Dark areas between the fibers indicate porosity. The
images show that the porosity is predominantly in the central regions of the tows,
indicating that the resin flowed from the sides as well as from the top. The resin flowed
into the tows from the top, but simultaneously, the resin flowed through inter-tow gaps,
and along the opposite (lower) side of the fabric, encapsulating the tows.
Microscopic observations revealed that some of the fiber tows were fully
impregnated with resin after 3 minutes at 70
o
C. A fully impregnated fabric representative
of vacuum-bagged samples removed from the 70
o
C water bath after 15 minutes is shown
in Figure 4.5C. The number of fiber tows fully impregnated with resin gradually
increased with time at both temperatures. As noted above, the resin completely
impregnated the fabric after 16 minutes at 70
o
C. Tow permeability values in the x and z
directions were 2.91 x 10
-13
and 7.96 ×10
-14
m
2
at 70
o
C, and the average was 1.53 x 10
-13

m
2
. The data are summarized in Table 4.3.
30

Figure 4.5A: Full vacuum for 3 minutes at 70
o
C

Figure 4.5B: Full vacuum for 9 minutes at 70
o
C

31

Figure 4.5C: Full vacuum for 15 minutes at 70
o
C

Comparison of the microscopy with the C-scan images revealed qualitative agreement
between the two imaging modes. Both image sets revealed that the resin gradually
flowed into the fiber tows until flow ceased after 16 minutes at 70
o
C. Furthermore, both
image sets also show that the fiber distribution in the fabric was not uniform. However,
the two imaging techniques are complementary rather than redundant. C-scan imaging
provides a non-destructive in situ method of monitoring resin permeation and
determining when resin flow ceases. The microscopy images, on the other hand, furnish
insight into the patterns of resin flow within and between tows. While the resolution is
superior to the ultrasound images, these images are effectively “post mortem”.
The values of tow permeability for the woven carbon fabric determined in this work
are similar to values reported previously. For example, the tow permeability values, K
z

and K
x
, for the plain woven fabric measured here were 7.96 ×10
-14
and 2.91 ×10
-13
m
2
,
32
33
while values determined by model-based calculations were 3.25 ×10
-14
m
2
and 2.47 ×10
-
13
m
2
[21]. The porosity of the fabric used for the model-based calculations was 21.6%,
whereas the porosity of the fabric used in this study was 33%. Porosity of the fabric is
the void space between the fibers before impregnation by resin. Material with a greater
porosity should have a greater permeability.
The similarity in measured and reported values of permeability is encouraging, yet it
may also be fortuitous, given the significant differences between the two studies.
Accurate determination, whether by measurement or calculation, presents a challenge.
For these reasons, experimental process diagnostics in general, and in situ ultrasonic
imaging in particular, are useful and necessary tools for measuring permeability,
understanding impregnation, and validating model predictions.
4.5: Conclusions
Ultrasonic imaging and reflectivity in the C-scan mode was used to measure the flow
rate of epoxy resin through a single layer of plain-woven carbon fabric. The ultrasonic
reflectivity signals were used to determine when resin flow through the fabric ceased,
while microscopy observations provided insight and details of how the resin flowed
through the fabric. The microscopy images clearly revealed that the resin did not flow
uniformly through the fabric from top to bottom. The information in the C-scans was
consistent with the microscopy images, and the measured permeability value was
consistent with permeability values for woven carbon fabric determined by other methods.
We have demonstrated that conventional C-scan ultrasonic imaging and reflectivity
can be used to monitor through-thickness resin flow in fiber arrays. In situ measurement
of resin impregnation and consolidation in thicker, multi-layered laminates should be
34
possible with the methods described in the present work. Furthermore, the method can be
used to determine the effects of different process parameters on fabric impregnation and
to better understand the processing science underlying this new class of low-pressure,
VBO prepregs. The method can also be applied to monitor impregnation and
consolidation during the production of large, flat composite parts, provided they are
impregnated, consolidated, and cured at temperatures below the maximum service
temperature of the transducer.
Ultrasound imaging, together with ultrasound reflectivity, is especially useful as a
process diagnostic for composite VBO prepregs, primarily because these materials are
amenable to low-pressure vacuum-bag processing and do not require high-pressure
autoclaves. Employing this tool for process monitoring may reduce defect rates and
eliminate the need for final NDE inspection after parts are cured, further reducing
production time and cost.

35
4.6: Chapter 4 References
[1] Cytec Engineered Materials, unpublished data.

[2] Klosterman, DA, Saliba, TE. Development of an on-line, in-situ fiber-optic void
sensor. Journal of Thermoplastic Composite Materials 1994; 7: 219–229.

[3] Drapier, S, Monatte, J, Elbouazzaoui, O, Henrat, P. Characterization of transient
through-thickness permeabilities of non crimp new concept (NC@) multiaxial fabrics.
Composites: Part A, Applied Science and Manufacturing 2005; 36: 877–892.

[4] Kuentzer, N, Simacek, P, Advani, SG, Walsh, S. Permeability characterization of dual
scale fibrous porous media. Composites: Part A: Applied Science and Manufacturing
2006; 37: 2057–2068.

[5] Louis, M, Huber, U. Investigation of shearing effects on the permeability of woven
fabrics and implementation into LCM simulation. Composites Science and Technology
2002; 63: 2081–2088.

[6] Wu, X, Li, J, Shenoi, RA. Measurement of braided perform permeability. Composites
Science and Technology 2006; 66: 3064–3069.

[7] Rodriguez, E, Giacomelli, F, Vázquez, A. Permeability-porosity relationship in RTM
for different fiberglass and natural reinforcements. Journal of Composite Materials, 2004;
38 (3): 259–268.

[8] Lekakou, C, Johari, MAK, Norman, D, Bader, MG. Measurement techniques and
effects on in-plane permeability of woven cloths in resin transfer moulding. Composites:
Part A, Applied Science and Manufacturing 1996;27:401–408.

[9] Griffin, PR, Grove, SM, Russell, P, Short, D, Summerscales, J, Guild, FJ, and Taylor,
E. The effect of reinforcement architecture on the long-range flow in fibrous
reinforcements. Composites Manufacturing 1995;6(3-4):221-235.

[10] Adams, KL, Russel, WB, and Rebenfeld, L. Radial penetration of a viscous liquid
into a planar anisotropic porous medium. International Journal of Multiphase Flow
1988;14(2):203-215.

[11] Chan, AW and Hwang ST. Anisotropic in-plane permeability of fabric media.
Polymer Engineering and Science 1991;31(16):1233-1239.

[12] Wu, X, Li, J. A new method to determine fiber transverse permeability. Journal of
Composite Materials 2007;41(6):747-756.

36
[13] Scholz, S, Gillespie, JW, Heider, D. Measurement of transverse permeability using
gaseous and liquid flow. Composites: Part A 2007;38:2034-2040.

[14] Elbouazzaoui, O, Drapier, S, Henrat, P. An experimental assessment of the saturated
transverse permeability of non-crimped new concept (nc2) multiaxial fabrics. Journal of
Composite Materials 2005;39;(13):1169-1193.

[15] Pearce, N, Guild, F, and Summerscales, J. A study of the convergent flow fronts on
the properties of fibre reinforced composites produced by RTM. Composites: Part A
1998;29:141-152.

[16] Stöven, T, Weyrauch, F, Mitschang, P, Neitzel, M. Continuous monitoring of three-
dimensional resin flow through a fiber preform. Composites: Part A 2003;34:475–480.

[17] ASM Metals Handbook, Desk Edition, 1998, 2nd ed., 1283.

[18] Williams, CD, Grove, SM, and Summerscales, J. The compression response of fibre-
reinforced plastic plates during manufacture by the resin infusion under flexible tooling
method. Composites: Part A 1998;29:111-114.

[19] Fu-quan, S and Ci-qun, L. The transient elliptic flow of power-law fluid in fractal
porous media. Applied Mathematics and Mechanics 2002;23;(8); 875-880.

[20] Wallstrom, TC, Christie, MA, Durlofsky, LJ, and Sharp, DH. Effective flux
boundary conditions for upscaling porous media equations. Transport in Porous Media
2002;46:139-153.

[21] Wang, Y and Grove, SM. Modelling microscopic flow in woven fabric
reinforcements and its application to dual-scale resin infusion modeling. Composites:
Part A: Applied Science and Manufacturing 2008;39:843-855.

37
Chapter 5: Temperature Dependence of Resin Flow in a Vacuum-Bag-Only Process
5.1: Abstract
Ultrasonic imaging in the C-scan mode was used in conjunction with the amplitude of
the reflected signal to measure the temperature dependence of resin flow rate in single
layers of woven carbon fabric. The RFI samples were vacuum-bagged and scanned in a
water tank at 50°C, 60°C, 70°C, and 80°C. The measured flow rates were plotted versus
inverse viscosity to determine the permeability in the thin film, non-saturated system.
The permeability values determined in this work were consistent with permeability
values reported in the literature. Capillary flow was not observed at the temperatures and
times required for pressurized flow to occur. The flow rate at 65°C was predicted from
the measured flow rates, and then measured in a 10-layer laminate. The investigation
demonstrates that ultrasonic imaging in the C-scan mode in conjunction with the
amplitude of the reflected signal is an effective method for measuring resin flow through
fabric.
5.2: Introduction
Understanding resin flow and impregnation of the reinforcing medium is critical to
the manufacture of composite laminates free of porosity and flow-related defects.
Although resin flow measurements are generally not performed in prepreg processes,
they are routinely performed in resin transfer molding processes. For example, fiber
optic sensors, used in the transmission or reflection modes and embedded at various
locations in the layers of reinforcing material, have been used to measure resin flowing in
the resin transfer molding (RTM) and resin film infusion processes [1, 2, 3]. However,
38
optical fibers are fragile, and fiber optic sensors are not sensitive to resin flow inside fiber
tows, so alternative approaches are often employed.
One approach to measurement of resin flow during impregnation involves the use of
ultrasonic imaging. Stöven et al used ultrasonic transmission to measure the
impregnation and fill rate of the preform in a resin transfer molding process. The
ultrasonic transmitter and receiver were aligned on opposite sides of the mold [4]. The
mold fill rates determined ultrasonically and visually were equivalent. This ultrasonic
transmission method is well suited to situations in which the transducers cannot be
immersed and do not require immersion in a coupling fluid. However, the ultrasound
signal resolution achieved with liquid coupling is superior to air coupling due to higher
frequency capabilities [5].
In the present work, we utilize ultrasonic imaging to monitor the impregnation of
fabrics in situ. In this method, an assembly consisting of fabric, resin film, and vacuum
bag is immersed in a heated water tank while performing C-scan imaging to produce a
density map or profile of the scanned material [6]. The method relies on the acoustic
contrast provided by scattering from wet and dry regions of the fabric. As the resin flows
into the fabric and permeates the fiber tows, the volume of the material system decreases
and the density increases. The volume and density of the material stop changing when
the resin stops flowing, and this is reflected in C-scan images and the amplitude of the
reflected signal, which cease to change when impregnation ceases. The procedure
demonstrates a noninvasive, in situ inspection method for monitoring resin impregnation
of fiber tows.
39
Darcy’s law, given below, relates the flow rate of a viscous fluid to the permeability
of the porous material. The flow rate in the fiber tows is V (m/s), permeability is K (m
2
),
pressure gradient is ∇P (Pa/m), and viscosity is μ (Pa*s).
V = K ∇P/ μ 1.
In RTM processes, pressure vs. time is plotted and the slope is related to K. The
viscosity of the resin decreases as temperature increases, and the flow rate increases with
temperature. A plot of V vs. 1/ μ should also be linear over the temperature range with
the slope equal to K ∇P.
Darcy’s Law was originally intended to describe saturated flow in a porous granular
medium [7] (in saturated flow, the interstitial spaces are filled with the flowing liquid).
In contrast, non-saturated flow occurs when the porous material is initially dry or does
not contain the liquid within its porous structure. In composite processes such as resin
film infusion and resin transfer molding, the dry fabric is initially in the non-saturated
condition. On the other hand, composite prepregs involve both saturated and unsaturated
flow because they generally contain regions of dry fabric or vacuum channels by design,
and these dry regions may also be considered non-saturated. Darcy’s law is routinely
used to determine permeability in materials where saturated and unsaturated flow occurs.
The objectives of the present work are to demonstrate the usefulness of ultrasound
imaging to measuring flow of resin through woven fabric and to determine average
permeability from flow rate measurements over a temperature range.
5.3: Experimental
The materials selected for the study included a plain-weave carbon fabric (Cytec plain
weave T-300), and an epoxy resin film (Cycom 5215 produced by Cytec). The fabric
40
was characterized by 3000 fibers per tow, fiber elastic modulus and strength of 231 GPa
and 3650 MPa, and a mass per area of 194 grams per square meter. The resin film had a
mass per area of 102 grams per square meter. A two-step process was employed in
which the resin film was first attached to the fabric by lamination, which was followed by
impregnation in a vacuum bag assembly [6]. The lamination procedure involved
removing the release liner from the resin film, placing the resin film and protective paper
layer on top of the fabric, placing the laminate on a teflon-coated flat surface preheated to
50°C, applying a pressure of 1500 Pa for 2 seconds, rapidly cooling to room temperature,
and removing the protective paper. Vacuum bagging and full impregnation at 50°C,
60°C, 70°C, and 80°C followed the lamination process. The manufacturer’s
recommended processing temperature of the resin was 65°C.
Resin viscosity at temperatures used in this work was measured at Cytec on a TA
Instruments ARES rheometer. Capillary flow measurements were performed on
laminated samples at 50°C for 60 minutes, 60°C for 40 minutes, 70°C for 20 minutes,
80°C for 10 minutes, and 90°C for 5 minutes without applied pressure. The samples
were cured at room temperature for 30 days after the heating cycles. Next, the samples
were sectioned, mounted in acrylic, polished, and examined microscopically.
Microscopic examination showed that the resin did not flow in the absence of applied
pressure during the heating cycles at 80
o
C and below, but resin did flow into the fiber
tows during the heating cycle at 90
o
C. Capillary flow occurred at 90
o
C, but not at 80
o
C
and below. Capillary flow does not occur in the test conditions used here.
Ultrasonic imaging was performed to determine when resin stopped flowing through
a single layer of fabric during the impregnation process, as described in recent work by
41
Thomas et al [6]. The general experimental configuration consisted of a thin epoxy resin
film laminated to a single layer of woven carbon fabric (RFI sample). Continuous strips
of breather cloth extended from approximately 1 cm under the edges of the RFI sample to
the vacuum valve, creating a breather path to allow evacuation of air. The fabric, with
the resin film on top, rested on an aluminum base plate, and was sealed in a vacuum bag.
The RFI sample contained 34% by weight of the resin film. The vacuum-bagged
assembly was immersed in a heated water tank designed for ultrasonic C-scan imaging,
and imaging was performed to monitor the impregnation while vacuum was applied to
the assembly.
First, the assembly was placed in a water tank (60 cm long x 45 cm wide x 15 cm
deep) at 20°C to measure the initial condition. A commercial ultrasound system
(Physical Acoustics Ultrapac II) was used for the ultrasonic imaging. A 12.7 mm
diameter 10 MHz focused transducer (focal length 38.1 mm) was used for scanning in the
reflected mode. The beam diameter at half a wavelength is 24 micrometers. The gates
were set to record signal between the top of the sample and bottom of the base plate. A
constant signal was produced from the base plate, so that any change in the signal
resulted from a change in the sample. The amplitude of the signal was set to 90% of the
maximum on the base plate. The A-wave scanning parameters for ultrasonic imaging
were established on a flat region of the vacuum bagged assembly near the ply. A support
fixture was used for the vacuum bagged assembly that allowed removal and re-insertion
in the water tank at the same position. A 25.4 mm x 25.4 mm area near the center of the
ply was scanned in the C-scan mode in one minute at 20°C to document the initial
42
condition. The assembly was removed from the water tank after the initial condition was
recorded.
In preparation for flow measurements, the water tank was heated and maintained at
the predetermined temperature of 50°C, 60°C, 70°C, or 80°C ± 1°C. The vacuum bagged
sample was re-inserted into the heated water tank. Flow of the resin through the carbon
fabric was measured at 50°C at five-minute intervals, 60°C at five-minute intervals, 70°C
at three-minute intervals, and 80°C at two-minute intervals until the C-scan image was
unchanged from the previous interval. Measurements of the initial condition at 20°C and
flow of the resin at 50°C, 60°C, 70°C, and 80°C were performed a total of three times at
each temperature on new samples. A constant pressure of 100 KPa ± 0.3 KPa was
continuously applied to the vacuum bagged samples during the flow measurements using
a mechanical pump. Pressure was measured by a digital vacuum gauge. The temperature
of the samples was measured with a thermocouple embedded under them. Once the
samples reached the temperature of the water tank, the temperature of the samples
remained constant during the duration of the experiment. Flow of the resin was
considered complete when the amplitude of the reflected signal of consecutive C-scan
images changed by one percent or less. Flow rate was plotted as a function of inverse
viscosity at 50°C, 60°C, 70°C, and 80°C. Permeability was calculated from the slope of
the straight line. The flow rate at 65°C was measured on a 10-layer laminate and
compared to that predicted from the flow data measured at 50°C, 60°C, 70°C, and 80°C.
5.4: Results and Discussion
C-scans: C-scan images from one of the three samples at each temperature are shown
in Figures 5.1 – 5.4. Red indicates a reflected signal intensity of approximately 10%,
43
yellow - approximately 50%, and dark green - approximately 90%. Intermediate colors
between red and yellow indicate reflected intensities between 10% and 50%, whereas
intermediate colors between yellow and dark green indicate reflected intensities between
50% and 90%. The uniformly spaced green regions of the initial scan correspond to
inter-tow gaps in the woven fabric. The resin flows relatively quickly through these gaps
(macro-flow) and begins permeating the fiber tows from the tow edges, where the tows
are thinnest (the tows are oblate ellipses in cross-section). The resin, after penetrating
through the inter-tow gaps, simultaneously flows from the top and bottom of the fabric.
The fabric is not entirely uniform, as evidenced from the initial conditions of Figures
5.1 – 5.4. Thicker regions appear red, thinner regions yellow, and gaps between the fiber
tows appear light green. The fiber tows are naturally thickest at the center of the tows
and thinnest at the edges. The non-uniformity of thickness and density of the fabric is
apparent throughout the series of scans. Thicker regions absorb more of the ultrasonic
signal than thinner regions. Therefore, one should not expect the C-scan of a fabric fully
impregnated with resin to be dark green. In addition, regions of higher density reflect
more of the signal than regions of lower density. Thus, C-scan images provide a
qualitative determination of when resin permeation ceases.
44

Initial Condition
15 Minutes

1 cm
30 Minutes 45 Minutes
Figure 5.1: C-scan images at 50
o
C of a vacuum bagged ply.

Initial condition
10 minutes
20 minutes 30 minutes
1 cm
Figure 5.2: C-scan images at 60
o
C of a vacuum bagged ply.
45

1 cm
Initial condition
3 minutes
9 minutes 18 minutes

Figure 5.3: C-scan images at 70
o
C of a vacuum bagged ply.

6 minutes
Initial condition
2 minutes
1 cm
4 minutes
Figure 5.4: C-scan images at 80
o
C of a vacuum bagged ply.

The amplitude of the reflected signals as a function of time for all samples is shown
in Figures 5.5. The resin stopped flowing when the amplitude of the reflected signal of
the last two images changed by one percent or less. The resin viscosity increased with
time, albeit quite slowly. (The gel time was 5-6 hours at 70°C and 3-4 hours at 80°C, so
the change in viscosity during the flow measurements was assumed to be negligible.)

0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
80.00
90.00
100.00
0 5 10 15 20 25 30 35 40 45 50
time (minutes)
normalized reflected signal
50C
60C
70C
80C

Figure 5.5: Normalized reflected signal as a function of time at 50°C, 60°C, 70°C, 80°C

Flow rate measurements were used to calculate the permeability of the fiber tows in
the z and x directions. Flow in the z direction is through the thickness of the fabric, while
flow in the x direction is perpendicular to the fiber tows but parallel to the plane of the
46
47
fabric. Flow in the y direction is parallel to the fibers and parallel to the plane of the
fabric, but is not considered. Flow rates measured from C-scan images and reflectivity
measurements are summarized in Table 5.1. Flow rates were determined from the z and x
dimensions divided by the flow time. Values of z and x of the fiber tows were
determined previously as 0.55 mm and 1.05 mm [6]. The center of the fiber tow was
assumed to be the last region filled by resin.
Flow rate as a function of inverse viscosity is shown in Figure 5.6 for x and z
directions. The viscosity data is listed in Table 5.2. A maximum standard deviation of
10.3% was calculated for the flow rate measurements, indicating the reproducibility of
the ultrasound data. The linear plots show that Darcy’s law is valid for non-saturated
flow through the fabric. Note that Darcy’s law was developed for one-dimensional
saturated flow of water through a tall column of wet sand. The sand was saturated with
water to remove air prior to flow measurements [7]. In this work, however, the initial
condition of the fabric was dry, not saturated with resin, and the porous material was
fibrous, not granular. (The resin film was lightly laminated to one side of the fabric so
that initially, nearly all of the fabric was dry, as shown in Figure 5.7.) A column of
particles uniform in size and shape may be considered isotropic, whereas woven carbon
fabric is orthotropic, if not anisotropic.
The slopes of the lines for x and z direction flow in Figure 5.6 are equal to the
products of K ∇P. A constant pressure was maintained by continuous operation of the
vacuum pump during impregnation. The pressure gradients in the z and x directions are
proportional to the flow lengths. Measurable pressure drops are often observed over
large flow distances, as in the case of resin transfer molding processes. No measurable
48
pressure drop was observed over the short flow distances in this work, indicating that the
pressure was constant. Therefore, the pressure gradients were approximated as ΔP/z and
ΔP/x [6, 8, 9]. Values of z and x of the fiber tows were determined previously as 0.55
mm and 1.05 mm, and ΔP was 100 kPa [6]. The permeability values, K
z
and K
x

determined from the linear plots are 5.5 x 10
-14
m
2
and 2.1 x 10
-13
m
2
. The difference in
permeability’s is due to the relative dimensions of the fiber tows and the different
pressure gradients in the z and x directions.
Table 5.1: Flow Properties at 50°C, 60°C, 70°C, & 80°C Determined by C-Scan.
Property at 50
o
C Sample 1 Sample 2 Sample 3 Average Stan. Dev.
Flow Time (minutes) 50 50 45 48.3 2.9
Z Flow Rate (10
-7
m/s) 1.83 1.83 2.04 1.90 0.12
X Flow Rate (10
-7
m/s) 3.50 3.50 3.89 3.62 0.22
Property at 60
o
C Sample 1 Sample 2 Sample 3 Average Stan. Dev.
Flow Time (minutes) 35 30 30 31.7 2.9
Z Flow Rate (10
-7
m/s) 3.06 3.06 2.62 2.89 0.25
X Flow Rate (10
-7
m/s) 5.83 5.83 5.00 5.53 0.48
Property at 70
o
C Sample 1 Sample 2 Sample 3 Average Stan. Dev.
Flow Time (minutes) 18 15 15 16 1.7
Z Flow Rate (10
-7
m/s) 6.11 6.11 5.09 5.73 0.59
X Flow Rate (10
-6
m/s) 1.17 1.17 0.97 1.09 0.11
Property at 80
o
C Sample 1 Sample 2 Sample 3 Average Stan. Dev.
Flow Time (minutes) 6 6 6 6 0
Z Flow Rate (10
-6
m/s) 1.53 1.53 1.53 1.53 0
X Flow Rate (10
-6
m/s) 2.92 2.92 2.92 2.92 0

Table 5.2: Viscosity Data
Temperature (°C) 50 60 65 70 80
Viscosity (Pa*s) 570.2 109.8 50.8 25.1 7.9

80C
80C
70C
60C
50C
X = 2E-05a + 3E-07
R2 = 0.9992
Z = 1E-05a + 2E-07
R2 = 0.9992
0.00
0.50
1.00
1.50
2.00
2.50
3.00
0 0.05 0.1 0.15
Inverse Viscosity 1/(Pa*s)
x and z direction flow rates (E-6 m/s)
X
Z

Figure 5.6: X & Z Direction Flow Rates Vs. Inverse Viscosity at 50°C, 60°C, 70°C, and
80°C.

100 μm
Figure 5.7: Resin film laminated to the top of the fabric prior to impregnation.
49
Polished sections of vacuum-bagged samples were removed from the 60°C water bath
after 10, 20, and 30 minutes, as shown in Figures 5.8A-C. The original position of the
resin film was on one side (top) of the fabric. Dark areas and voids between the fibers
indicate porosity. The images show a random distribution of porosity, indicating that the
resin did not flow uniformly from top to bottom. If the resin flow was uniform from top
to bottom, a clear flow-front would be visible in Figures 5.8A and 5.8B, and voids would
occur only ahead of the flow front. The resin flowed into the tows simultaneously from
the top, bottom, and sides. The resin completely permeated the fabric after 30 minutes at
60°C.

Porosity
100 μm

Figure 5.8A: Full vacuum, 10 minutes, 60°C

50

Porosity
100 μm
Figure 5.8B: Full vacuum, 20 minutes, 60°C

100 μm

Figure 5.8C: Full vacuum, 30 minutes, 60°C

51
52
Comparison of the microscopy with the C-scan images revealed qualitative agreement
between the two imaging modes, although close inspection indicated that the two
techniques are complementary rather than redundant. Both image sets revealed that the
resin gradually permeated into the fiber tows until flow ceased after 30 minutes at 60°C.
However, C-scan imaging provided a non-destructive in situ method of monitoring resin
permeation and determining when resin flow ceases. The microscopy images, on the
other hand, furnished insight into the patterns of resin flow within and between tows.
Although these images were effectively “post mortem” (and destructive) the resolution
was superior. The resin flow patterns shown in the microscopy images of Figure 5.8A-C
are representative of the resin flow patterns of the other samples at 50°C, 70°C, and 80°C.
The values of tow permeability for the woven carbon fabric determined in this work
are similar to values reported previously. For example, the tow permeability values, K
z

and K
x
, for the plain woven fabric measured here were 5.5 ×10
-14
and 2.1 ×10
-13
m
2
, while
values determined by model-based calculations were 3.25 ×10
-14
m
2
and 2.47 ×10
-13
m
2

[10]. The porosity of the fabric used for the model-based calculations was 21.6%,
whereas the porosity of the fabric used in this study was 33%, determined from polished
sections. Porosity of the fabric is the void space between the fibers before impregnation
by resin. Kuentzer et al measured through-thickness tow permeability values of 1.30
×10
-14
m
2
and 2.30 ×10
-14
m
2
for a woven carbon fabric [11], while those measured here
were ~ 5.5 ×10
-14
m
2
. Note that the methods used in the present work are quite different
from the methods used by Kuentzer et al. In addition, the carbon fabric used in the
present work had an area density of 194 grams per square meter, while the area density of
carbon fabric used in the previous work was 56% greater (305 grams per square meter).
Factors that affect permeability include but are not limited to fabric architecture, fiber
arrangement within the tow, fabric porosity, and fiber waviness [12, 13, 14, 15, 16].
The resin viscosity was 50.75 Pa*s at 65 °C, and a flow time of 24 minutes was
predicted from the linear plots in Figure 5.6. Ultrasound images of a 10-layer laminate
vacuum-bagged and scanned at 65 °C are shown in Figure 5.9. Ultrasound imaging
indicated that the laminate was fully impregnated and resin stopped flowing after 24
minutes, as predicted. The laminate was immediately removed from the 65 °C water tank
after 24 minutes, vented, and cured without applied pressure. Microscopic examination
also revealed that the laminate was fully impregnated.

1 cm
Initial condition 9 minutes
24 minutes 18 minutes
Figure 5.9: C-scan images at 65°C of a 10-layer vacuum bagged laminate.

53
54
5.5: Conclusions
The temperature dependence of flow rate for resin film flowing through a single layer
of woven carbon fabric was measured using ultrasonic imaging. The flow rate varied
linearly with inverse viscosity, and the slope of the line was equal to the product of K ∇P.
The linear relation between flow rate and inverse viscosity shows that Darcy’s law is
valid for non-saturated flow of a resin film through a thin porous material such as carbon
fabric. The results presented here are directly applicable to understanding flow in
prepregs and RFI processes. The flow behavior in individual layers of prepreg can be
predicted by Darcy’s law.
Comparison of ultrasonic and light microscopy images yielded identical results and
demonstrated the effectiveness of ultrasonic imaging as an in situ process diagnostic for
monitoring impregnation and flow rates. Experimental process diagnostics such as in situ
ultrasonic imaging are useful and necessary tools for measuring permeability,
understanding impregnation, and validating model predictions. Using the measured
permeability values, the flow rate at any temperature within the measured range can be
predicted with a high degree of certainty, given that the viscosity is known at that
temperature. In situ measurement of resin impregnation and consolidation in thick,
multi-layered laminates is also possible using the ultrasound methods described here.

55
5.6 Chapter 5 References
[1] Klosterman, DA, Saliba, TE. Development of an on-line, in-situ fiber-optic void
sensor. Journal of Thermoplastic Composite Materials 1994; 7: 219–229.

[2] Drapier, S, Monatte, J, Elbouazzaoui, O, Henrat, P. Characterization of transient
through-thickness permeabilities of non crimp new concept (NC@) multiaxial fabrics.
Composites: Part A, Applied Science and Manufacturing 2005; 36: 877–892.

[3]Antonucci, V, Giordano, M, Nicolais, L, Calabro, A, Cusano, A, Cutolo, A, Inserra, S.
Resin flow monitoring in resin film infusion process. Journal of Materials Processing
Technology 2003; 143-144: 687-692.

[4] Stöven, T, Weyrauch, F, Mitschang, P, Neitzel, M. Continuous monitoring of three-
dimensional resin flow through a fiber preform. Composites Part A: Applied Science and
Manufacturing 2003; 34: 475–480.

[5] ASM Metals Handbook, Desk Edition, 1998, 2nd ed., 1283.

[6] Thomas, S, Nutt, SR, and Bongiovanni, C. In situ estimation of through-thickness
resin flow using ultrasound. Composites Science and Technology 2008; 68: 3093-3098.

[7] Darcy, H. Les Fontaines Publiques de la Ville de Dijon. de Jussieu, Hist. de I’
Acadimie Royale de Sciences 1856; 1733: 351-8.

[8] Chan, AW and Hwang ST. Anisotropic in-plane permeability of fabric media.
Polymer Engineering and Science 1991; 31; (16): 1233-1239.

[9] Elbouazzaoui, O, Drapier, S, Henrat, P. An experimental assessment of the saturated
transverse permeability of non-crimped new concept (nc2) multiaxial fabrics. Journal of
Composite Materials 2005; 39; (13): 1169-1193.

[10] Wang, Y and Grove, SM. Modeling microscopic flow in woven fabric
reinforcements and its application to dual-scale resin infusion modeling. Composites:
Part A: Applied Science and Manufacturing 2008; 39: 843-855.

[11] Kuentzer, N, Simacek, P, Advani, SG, and Walsh, S. Permeability characterization
of dual scale fibrous porous media. Composites Part A: Applied Science and
Manufacturing 2006; 37: 2057–2068.

[12] Ma, Y and Shishoo, R. Permeability characterization of different architectural
fabrics. Journal of Composite Materials 1999; 33 (8): 729-751.

56
[13] Pearce, NRL, Guild, FJ, and Summerscsales, J. The use of automated image analysis
for the investigation of fabric architecture on the processing and properties of fiber-
reinforced composites produced by RTM. Composites Part A: Applied Science and
Manufacturing 1998; 29: 829–837.

[14] Dungan, FD and Sastry, AM. Saturated and unsaturated polymer flows:
microphenomena and modeling. Journal of Composite Materials 2002; 36; (13): 1581-
1602.

[15] Gutowski, TG, Cai, Z, Bauer, S, and Boucher, D. Consolidation experiments for
laminate composites. Journal of Composite Materials 1987; 21; (7): 651-669.

[16] Åström, BT, Pipes, RB, and Advani, SG. On flow through aligned fiber beds and its
application to composites processing. Journal of Composite Materials 1992; 26; (9):
1351-1372.

57
Chapter 6: Effect of Fabric Architecture on Through-Thickness Permeability in
Multi-ply Laminates
6.1: Abstract
Ultrasonic imaging was performed to measure the effects of fabric architecture on
through-thickness flow rates during impregnation of an epoxy resin film. Multilayered
laminates comprised of woven carbon fabrics and epoxy films, resin film infusion (RFI)
precursors, were fabricated by vacuum-bagging, and ultrasonic C-scan imaging was
performed on the laminates during impregnation in a heated water tank (65
o
C).
Impregnation rates showed a strong dependence on fabric architecture, despite similar
areal densities. Impregnation rates were directly affected by inter-tow spacing and tow
nesting, which depended on fabric architecture, and were indirectly affected by areal
densities.
6.2: Introduction
Interest in out-of-autoclave processing (OOA) has grown significantly in recent years
because of the increased versatility, low cost, and short cycle times relative to autoclave
processing. Out-of-autoclave processes such as resin film infusion (RFI), vacuum
assisted resin transfer molding (VARTM), and vacuum-bag-only processing (VBO)
generally require lower pressures than autoclave processing. In VBO processing,
pressure gradients of only one atmosphere, as opposed to several atmospheres in an
autoclave, are typical [1,2,3,4], and cycle times are typically much shorter. In VBO
processing, resin must flow through macro- and micro-channels of the reinforcing
material during the impregnation stage to produce a void-free part. To analyze the flow
process during impregnation, Darcy’s law is often used.
P
K
V ∇ =
μ

Darcy’s law relates the flow rate (V) of the resin to the permeability (K) of the reinforcing
medium, the pressure gradient ( P ∇ ) across the system, and the viscosity ( μ) of the resin.
Multiple factors influence permeability, including macro- and micro-porosity in the
laminate, fiber arrangement in the tows, flow direction, and fabric architecture [1 - 7].
For example, continuous macro-channels between layers can exist in loose weaves, but
these channels can be eliminated in tightly woven fabrics, limiting flow to pathways
through the fiber tows [3]. In such cases, in-plane permeability depends on multiple
factors, including inter-tow spacing, fabric crimp, and tow nesting. Greater fiber crimp
creates increased void space and reduces nesting between layers. Crimp of the fibers also
creates interstices within the tows, increasing the tow permeability [4, 5]. A third factor
affecting permeability is nesting, which arises during debulking, a common practice that
reduces the permeability by reducing the gap between layers. Nesting of tows in adjacent
layers reduces the size of the gaps between layers, and consequently reduces the in-plane
permeability. An additional factor affecting permeability is simply the number of layers
in the laminate – as these increase, the permeability reportedly decreases [7].
The multitude of process parameters that affect permeability highlights the need for a
predictive model for flow kinetics. The most common approaches for modeling in-plane
and tow permeability rely either on Darcy’s law [1, 3, 7] or the Kozeny-Carman equation
[2, 4, 6]. In general, these models predict that the unsaturated permeability is less than the
saturated permeability [1 – 7]. Mathematical models based on Darcy’s law [8 - 11],
geometrical descriptions of the fabric architecture [12 – 14], Stokes and Brinkman
58
59
equations [15 – 16], and the Kozeny-Carman equation [17 – 20] have been developed to
describe macro- and micro-flow in liquid injection molding processes. The majority of
flow modeling has been applied to resin transfer molding (RTM) processes, and has been
used for predicting in-plane permeability, not through-thickness permeability. In contrast
to RTM processes, resin flow in prepregs (and resin film infusion processes) is primarily
in the through-thickness direction, and requires different application of the models.
In the present work, laminates were fabricated by attaching a resin film to a sheet of
carbon fabric, and stacking ten such layers in a vacuum bag. This was intended to
simulate the conditions of VBO prepreg consolidation, and differed from most liquid
resin infusion processes in that macro-flow occurred primarily in the through-thickness
(“z”) direction. Macro-flow in the z-direction occurred by resin flow between layers and
through inter-tow gaps. In-plane (x,y) flow in this setup was nearly complete when the
laminate was laid up. Thus, macroscopic flow in the (x,y) plane was expected to depend
on how the layers were nested or meshed with each other. On the other hand, micro-flow
in the tows was expected to occur primarily in the z-direction perpendicular to the top
and bottom faces of the tow, and perpendicular to the edges of the tow. Micro-flow
parallel to the fibers was expected to be negligible because the length of the fiber tow was
several orders of magnitude greater than the width and height. Because of these factors,
and because capillary flow did not occur under the conditions used in this work, macro-
and micro-flow analysis was simplified for this fabric-resin system, affording new
insights into the impregnation process.

60
6.3: Experimental
The materials selected for the study included four carbon fabrics consisting of (1) a
plain-weave (density = 194 grams per square meter), (2) a 4-harness satin weave (density
= 187 grams per square meter), (3) a 2 × 2 twill (density = 194 grams per square meter),
and (4) a 4 × 4 twill (density = 281 grams per square meter). All of the carbon fabrics
were made from T300 fiber with 3000 fibers per tow. Epoxy resin films (5215-1, Cytec)
were selected with densities of 68 and 102 g/m
2
and were used as needed to achieve a
resin volume fraction of 54 -56 %. The fabric weights and resin fractions used are given
in Table 1, and the architecture of the woven fabrics is shown in Figure 6.1. The resin
film was attached to one side of the fabric by lightly laminating, as described previously
[21]. The lamination process was followed by the second step of the process - full
impregnation by vacuum bagging at 65°C.
Impregnation: Laminates of plain weave, 4-harness satin weave, 2 × 2 twill, and 4 × 4
twill fabrics containing 10 layers each were vacuum-bagged, out-gassed, and heated
under full vacuum for 12, 24, 36, 48, and 60 minutes in a heated water tank at 65°C.
Flow of the resin was halted by immediately removing the samples from the heated water
tank, venting the vacuum-bagged samples, and quenching in cold water. The samples
were cured at 60°C for 24 hours without pressure applied, and polished sections were
examined by light microscopy.
Measuring Flow: Ultrasonic imaging was performed during the impregnation cycle to
measure the extent of resin flow through the 10-layer laminates [21]. Prior to
impregnation, the vacuum-bagged assembly was placed in a water tank (60 cm long × 45
cm wide × 15 cm deep) at 20
o
C to measure the initial condition. The A-wave scanning
61
parameters for ultrasonic imaging were established on a flat region of the vacuum bagged
assembly near the ply. A support fixture was used for the vacuum bagged assembly that
allowed removal and re-insertion in the water tank at the same position. A 25.4 mm ×
25.4 mm area near the center of the ply was scanned in the C-scan mode in one minute at
20
o
C to document the initial condition. The assembly was removed from the water tank
after the initial condition was recorded.
In preparation for flow measurements, the water tank was heated to 65
o
C and
maintained to within ± 1
o
C. The vacuum-bagged sample was re-inserted into the heated
water tank. Flow of the resin through the laminates was measured at three-minute or
five-minute intervals until resin flow ceased. Measurements of the initial condition at
20
o
C and flow of the resin at 65
o
C were repeated on three samples of each fabric type. A
constant pressure of one atmosphere was applied to the vacuum-bagged plies during the
flow measurements using a mechanical pump.
Ultrasound reflectivity was monitored and plotted as a function of time during
impregnation. The cessation of resin flow was determined by comparing consecutive
scans, and when the reflectivity of the last two consecutive scans changed by less than
one percent, resin flow was taken to be complete.

Table 6.1: Fabric weights and resin percentages.
Fabric Type PW 4-HSW 2 × 2 Twill 4 × 4 Twill
Fabric Weight
(g/m2)
194 186 194 281
Volume of
Resin (%)
54.7 55.6 54.8 53.8

Plain Weave 2 x 2 Twill
4-HSW
4 x 4 Twill

Figure 6.1: Fabric architectures.

62
63
6.4: Results and Discussion
Representative C-scan images of the 4-harness satin weave laminate were recorded
during impregnation at 65
o
C, as shown in Figure 6.2. The images were recorded at
different time intervals, and show the progression of impregnation. In the images, red
indicates a reflected signal intensity of approximately 10%, yellow - approximately 50%,
and dark green - approximately 90%. Intermediate colors between red and yellow
indicate reflected intensities between 10% and 50%, whereas intermediate colors between
yellow and dark green indicate reflected intensities between 50% and 90%. The initial
conditions of the laminates were not uniform due to inherent non-uniformity of the fabric
and non-uniform tow nesting of the layers. Resin flow in the laminates also was not
uniform, as evidenced by the absence of periodic contrast in the images (Figure 6.2). The
C-scan images shown in Figure 6.2 are typical, and laminates made with the other fabrics
yielded similar C-scan images, which are not shown here.
The reflected intensities from the four laminate types are shown in Figures 6.3A and
6.3B, and the average reflected intensities are shown in Figure 6.3C. The similarity of
the curves in Figures 6.3A and 6.3B indicate the precision of the measurements, while the
average curves show the distinctive impregnation kinetics peculiar to each laminate type.
Note that the plain weave fabric showed the most rapid impregnation, while the 4 × 4
twill showed the slowest. The infusion times are shown in Table 2, and the fabric rank
was PW, 4-HSW, 2 × 2, and 4 × 4, in order of shortest to longest impregnation time. Note
that the deviations were typically 3% or less.
Comparison of C-scan and microscopy images revealed qualitative agreement
between the two imaging modes (see Figures 6.2 and 6.4A-D). Both image sets showed
64
non-uniform flow, in which some regions were impregnated earlier than others. Macro-
and micro-porosity existed in the early stages of resin flow, and the micro-porosity
persisted into the intermediate stages of flow, followed by full impregnation. This flow
pattern occurred in all four types of laminates, as shown in Figures 6.4A-D. Note that the
images showed that the initial distribution of tows within the fabrics was also not uniform.
Non-uniformity was due to inherent variations in tow density across the fabric and
inconsistent tow nesting, both of which are not uncommon.
Adjacent layers of fabric in the laminate can nest together to produce either a dense
structure (“ideal nesting”), or a structure with gaps between the layers (“non-ideal
nesting”), shown schematically in Figures 6.5A and 6.5B. The different types of nesting
are observed in practice (Figures 6.4C and 6.4D), and both structures are found in the
laminates at different locations. Variation in fiber density across the fabric layers and
inconsistent tow nesting created the non-uniform initial conditions and non-uniform resin
flow. For impregnation, there are advantages to the ideal tow nesting shown in Figure
6.5A, as opposed to non-ideal nesting. In regions where the tows are nested, the distance
for resin macro-flow is relatively short. Consequently, the fiber tows are quickly
surrounded by liquid resin, which penetrates the fiber tows from all directions. In these
regions, the laminate exhibits greater local fiber loading, and is generally free of porosity.
The degree of tow nesting is influenced by the fabric architecture. For example, ideal
tow nesting is more likely to occur in the plain weave laminates than in laminates made
from other fabrics, because there are more locations in the plain-weave structure for
nesting to occur. Peaks and valleys are created in the fabric every time the fiber tows are
woven over and under each other. The peaks and valleys provide opportunities for tow
65
nesting. Close nesting of laminate plies is considered desirable in VBO and RFI
processing because it minimizes the amount of resin required. On the other hand, if the
quantity of resin is insufficient to fill the gaps between layers, macro-porosity will result.
In contrast, in resin transfer molding (RTM) processes, the resin supply is not limited,
and gaps between layers provide macro-flow channels for the resin.
In addition to tow nesting, inter-tow gaps within the (x-y) fabric plane of each ply
also influenced impregnation kinetics, enhancing flow between layers and promoting a
dense structure. The gaps between fiber tows also allowed resin to surround the tows and
penetrate from all directions simultaneously. The plain-weave tows contained gaps that
were fairly uniform in size and distribution, and these existed at each tow intersection.
On the other hand, tow gaps in the 4-HSW fabric varied significantly in size, shape, and
distribution and approximately 23% of the fiber tows had no gaps. Gaps between the
tows of the 2 × 2 twill structure were smaller in size than those of the 4-HSW structure,
and approximately 31% of the fiber tows in the 2 × 2 structure had no gaps. The 4 × 4
structure showed no gaps between the fiber tows. The spatial distribution of the inter-tow
gaps in the fabrics correlated directly with impregnation times (Table 2), and affected the
patterns and kinetics of resin flow.
The tow waviness within fabrics also influenced macro-flow patterns during laminate
consolidation. Tow waviness is based on a qualitative assessment of how often (and the
extent to which) the tow is bent as it is woven around adjacent tows. Of the four fabrics
used, the plain weave tows exhibited the greatest tow waviness, followed by 4-HSW, 2 ×
2 twill, and 4 × 4 twill. Tow waviness creates gaps between overlapping tows, opening
channels for macro-flow.
66
Fabric weight (measured in units of g/m
2
) was also expected to affect the flow
kinetics during impregnation. Factors that affect fabric weight are tow size, fiber
diameter, density of fibers, and weave construction. For a constant tow size, fiber
diameter, and fiber density, fabric weight is a function of the weave construction. Based
on this metric, the 4-HSW laminates were expected to have the fastest infusion rates
(they had the lowest fabric weight), followed by the plain-weave laminates, the 2 × 2
twill weave laminates; and the 4 × 4 twill weave laminates. However, this was not the
case. Impregnation times for the PW laminates were roughly half those of the 4-HSW
laminates (see Table 6.2), despite having slightly greater fabric weight (see Table 6.1).
Furthermore, the plain-weave and 2 × 2 twill fabrics had identical fabric weights (194
g/m
2
) and resin content (55%), yet the infusion time for the 2 × 2 twill laminates was
approximately twice that of the plain weave laminates. If infusion times were directly
controlled by fabric weight, then the infusion rates of PW and 2 × 2 twill laminates would
have been similar. Thus, two fabrics with the same areal weight can have different
distributions of tows, and the distribution of tows influences the flow rates. In particular,
it appears that the distribution of macro-flow channels within the fabric is more important
than the absolute areal weight. Supporting this assertion is the observation that the 4 × 4
twill laminates exhibited to the slowest infusion rates, apparently because the tows were
tightly woven together, resulting in a lack of macro-flow channels. The results show that
impregnation rate is dependent on fabric architecture, and while the areal weight is
related to the architecture, the dependence on areal weight is not direct.
In summary, resin flow was directly affected by tow nesting, gaps between the tows,
and distortion of the tows, and indirectly affected by fabric weight. All of these
characteristics stem from the fabric architecture. Resin flow on the macro- and micro-
levels was aided by greater tow nesting, more numerous gaps between the tows, and
greater distortion of the fiber tows. The plain-weave fabric possessed all the
characteristics that assist resin flow, and thus exhibited the shortest infusion times. The
other three fabrics showed some but not all of the characteristics, and consequently
showed slower infusion kinetics. Note that the fabric characteristics that facilitate resin
flow do not necessarily optimize strength or other important properties. For example,
fabric crimp results in tow bending, which promotes nesting and opens flow channels
between the tows, but this also reduces the strength of the composite (as well as the fiber
loading). Likewise, drapability is reduced by tow bending, which can lead to defects,
particularly in curved laminates. These considerations indicate some of the tradeoffs that
arise when selecting fabrics for composite fabrication.
1 cm
Initial Condition 15 minutes
27 minutes 39 minutes

Figure 6.2: 4-Harness Satin Weave C-scans at 65 °C.
67
0
20
40
60
80
100
0 1020 30 4050 60
time (minutes)
normalized reflected intensity
PW-1
PW-2
PW-3
4HSW-1
4HSW-2
4HSW-3

Figure 6.3A: Reflected intensities of the PW and 4-HSW samples.
0
10
20
30
40
50
60
70
80
90
100
0 1020 3040 50607080
time (minutes)
normalized reflected intensity
2x2-1
2x2-2
2x2-3
4x4-1
4x4-2
4x4-3

Figure 6.3B: Reflected intensities of the 2 x 2 and 4 x 4 twill samples.
68
0
10
20
30
40
50
60
70
80
90
100
0 10 20304050 607080
time (minutes)
average normalized reflected intensity
PW
4HSW
2x2
4x4

Figure 6.3C: Average reflected intensities of all samples.

Table 6.2: Resin Infusion Time of the Laminates (Minutes)
Sample 1 2 3
x
σ
PW 24 24 24 24 0
4HSW 39 45 42 42 3
2 x 2 51 54 51 52 1.7
4 x 4 60 65 65 63.3 2.9

69

Porosity
200 µm
Figure 6.4A: Macro and micro porosity in plain-weave laminate

Porosity
200 µm
Figure 6.4B: Micro porosity in plain weave laminate.
70

200 µm
Figure 6.4C: Ideal tow nesting in plain weave laminate.

400 µm
Figure 6.4D: Non ideal tow nesting in 2 x 2 twill laminate.

71

Figure 6.5A: Nesting of tows in adjacent plies.

Figure 6.5B: Un-nested tows in adjacent plies.

6.5: Conclusions
Ultrasonic imaging in the C-scan mode was used to study the effects of fabric
architecture on the infusion rate of epoxy resin films in multi-ply laminates. The results
show that fabric architecture significantly affects cycle times for out-of-autoclave
processes such as resin film infusion (RFI) and VBO prepreg processes. The fabric
features that affected the resin flow rate included the extent of tow nesting between layers,
size and frequency of in-plane gaps between tows, and distortion of the fiber tows within
the fabric. All of these factors arose from the particular fabric architecture. Nesting of
72
73
the tows reduced the macro-flow length of the resin, reducing the impregnation time and
reducing or eliminating macro-porosity by reducing the distance between layers.
Laminates should be produced in a manner to promote tow nesting, if possible. Secondly,
gaps between the tows allowed resin to flow between layers and around the tows,
surrounding the tows. Larger and more numerous gaps exposed more surface area of the
tows, and resin penetrated from peripheral edges of tows. Infusion rates depended on
how fiber tows were arranged relative to each other, and was not directly related to fabric
weight.
However, commercial VBO prepregs will differ in several ways from the model
system considered here. For instance, most commercial VBO resin systems will be
toughened epoxies, and will thus possess substantially different rheological
characteristics. Also, and perhaps more significantly, some VBO prepregs will have
resin film laminated on both sides and partially impregnated. Such prepregs differ from
the materials used here in which resin film was laminated to only one side of the fabric
without partial impregnation. These simple but significant differences may or may not
affect the flow properties of the resin, and will require further investigation.

74
6.6: Chapter 6 References
[1] Ma, Y and Shishoo, R, Permeability characterization of different architectural fabrics,
Journal of Composite Materials 1999; 33; (8): 729-750.

[2] Shih, CH and Lee, LJ, Effect of fiber architecture on permeability in liquid composite
molding, Polymer Composites 1998; 19; (5): 627-639.

[3] Babu, BZ and Pillai, KM, Experimental investigation on the effect of fiber-mat
architecture on the unsaturated flow in liquid composite molding, Journal of Composite
Materials 2004; 38; (1): 57-79.

[4] Endruweit, A, McGregor, P, Long, AC, and Johnson, MS, Influence of the fabric
architecture on the variations in experimentally determined in-plane permeability values,
Composites Science and Technology, 2006; 66: 1778-1792.

[5] Pearce, NRL, Guild, FJ, and Summerscsales, J. The use of automated image analysis
for the investigation of fabric architecture on the processing and properties of fiber-
reinforced composites produced by RTM. Composites Part A: Applied Science and
Manufacturing 1998; 29: 829–837.

[6] Berdichevsky, AL and Cai, Z, Preform permeability predictions by self-consistent
method and finite element simulation. Polymer Composites, 1993;14(2):132-143.

[7] Scholz, S, Gillespie, JW, Heider, D. Measurement of transverse permeability using
gaseous and liquid flow. Composites: Part A 2007; 38: 2034-2040.

[8] Lekakou, C and Bader, MG. Mathematical modeling of macro- and micro-infiltration
in resin transfer moulding (RTM). Composites: Part A, Applied Science and
Manufacturing 1998; 29: 29–37.

[9] Dungan, FD and Sastry, AM. Saturated and Unsaturated Polymer Flows:
Microphenomena and Modelling. Jounal of Composite Materials 2002; 36; (13): 1581-
1602.

[10] Zhou, F, Kuentzer, N, Simacek, P, Advani, SG, and Walsh, S. Analytical
characterization of the permeability of dual-scale fibrous porous media. Composites
Science and Technology 2006; 66: 2795–2803.

[11] Simacek, P and Advani, SG. A numerical model to predict fiber tow saturation
during liquid composite molding. Composites Science and Technology 2003; 63: 1725–
1736.

[12] Wong, CC and Long, AC, Modelling variation of textile fabric permeability at
mesoscopic scale, Plastics, Rubber, and Composites, 2006; 35; (3): 101-111.
75
[13] Dungan, FD, Senoguz, MT, Sastry, AM and Faillaci, DA. Simulations and
experiments on low-pressure permeation of fabrics: part I-3D modeling of unbalanced
fabric. Jounal of Composite Materials 2001; 35; (14): 1250-1284.

[14] Senoguz, MT, Dungan, FD, Sastry, AM and Klamo, JT. Simulations and
experiments on low-pressure permeation of fabrics: part II-the variable gap model and
prediction of permeability. Journal of Composite Materials 2001; 35; (14): 1285-1321.

[15] Markicevic, B and Papathanasiou, TD. A model for the transverse permeability of
bi-material layered fibrous performs. Polymer Composites, 2003; 24; (1): 68-82.

[16] Ranganathan, S, Phelan, FR, and Advani, SG. A generalized model for the
transverse fluid permeability in unidirectional fibrous media. Polymer Composites, 1996;
17; (2): 222-230.

[17] Mayer, C, Wang, X, and Neitzel, M. Macro- and micro-impregnation phenomena in
continuous manufacturing of fabric reinforced thermoplastic composites. Composites:
Part A, Applied Science and Manufacturing 1998; 29: 783–793.

[18] Gebert, BR. Permeability of unidirectional reinforcements for rtm. Journal of
Composite Materials 1992; 26; (8): 1100-1133.

[19] Cai, Z and Berdivchesky, AL. An improved self-consistent method for estimating
the permeability of a fiber assembly. Polymer Composites, 1993; 14; (4): 314-323.

[20] Berdivchesky, AL and Cai, Z. Preform permeability predictions by self-consistent
method and finite element simulation. Polymer Composites, 1993; 14; (2): 132-143.

[21] Thomas, S, Bongiovanni, C, and Nutt, SR. In situ estimation of through-thickness
resin flow using ultrasound. Composites Science and Technology 2008; 68: 3093-3098.

76
Chapter 7: Modeling Resin Flow in VBO Composite Laminates
7.1: Abstract
A method for predicting resin infusion rates in prepreg and resin film infusion
processes was developed based on the Stokes equation. The method was used to predict
the impregnation rate of laminates as a function of fabric architecture. A resin infusion
rate equation was derived by equating the drag force on the fibers to the force induced by
the applied pressure on the laminate. Flow rate data previously measured by ultrasound
was analyzed with the derived equation. A fiber interaction parameter that was
dependent on fabric architecture was determined for four different types of carbon fabric.
The flow rate data was also analyzed using the Kozeny-Carman equation to determine
values of the Kozeny constant K
C
, and the tortuosity (L
e
/L). The results demonstrated the
utility of the newly derived equation for predicting resin infusion rates of multilayer
laminates. The utility of the equation was demonstrated by fitting with experimental data
for the four fabrics. The method can be used for predicting resin infusion rates in
conventional prepregs.
7.2: Introduction
Traditional processing of high-strength composite parts is performed in autoclaves,
which are expensive both to purchase and to operate. In addition, cycle times in
autoclaves are relatively slow compared to out-of-autoclave (OOA) processes, and thus,
high-volume production is not practical. In contrast, low-pressure out-of-autoclave
processes offer an attractive alternative, and are beginning to compete with autoclave
processes for production of composite parts. One OOA process in particular, the
vacuum-bag-only (VBO) process, is gaining acceptance in industry. The VBO process is
77
more cost effective than the autoclave process because it is amenable to different heat
sources, heat transfer rates, and lower consolidation pressures.
In the VBO process, parts are fabricated using specially formulated prepregs that can
be consolidated at low pressures (approximately 100 kPa) and cured in an oven. For such
out-of-autoclave processes, knowledge of the resin flow characteristics is required to
optimize the manufacture of high-quality parts. Resin flow in other low-pressure
processes has been investigated by experimental measurements and by computer
modeling, but little work has been published on the VBO process. Given the versatility
and potential cost reduction of the VBO process, it is important and necessary to develop
a method of predicting resin infusion rates.
Geometrical models have been developed to describe resin flow in plain-weave
fabrics. In most cases, these models have been employed to perform permeability
calculations using finite element or difference algorithms based on Darcy’s law [1- 3]. In
other cases, the Kozeny-Carman model has been used to determine permeability as a
function of the fiber arrangements in composites [4 – 11]. Regardless of which equation
is used, modeling is complicated by the fact that resin flow during fabric impregnation
can occur on two scales - macro-flow between layers, and micro-flow within fiber tows.
For example, in RTM processes, the resin flow rate does not drop to zero once the mold
is filled, due to delayed filling of the fiber tows [12-15]. Micro-flow within tows depends
on tow permeability, which has been modeled by setting the divergence of the resin flow
rate in Darcy’s law ( Λ·V) equal to a non-zero value or sink term. Generally, Λ·V = 0, but
delayed filling of the fiber tows (as in RTM processes) causes an increase in pressure,
which is accounted for by setting Λ·V equal to a non-zero sink term [12-15]. However,
during consolidation of prepregs, resin flow occurs primarily within tows, because the
region between layers is already filled with resin. In other words, dual-scale flow in
prepregs generally does not occur, and, only micro-flow is involved. The vast majority
of reports on OOA processes have dealt with RTM, and simulations have been based on
Darcy’s Law or the Kozeny-Carman equation.
7.2.1: Permeability Models
Darcy’s law relates the permeability of a porous material to the applied pressure
differential, the viscosity of the fluid flowing through it, the flow rate of the fluid, and the
flow distance. In this relation, the flow rate, V (m/s) is related to the permeability K (m
2
),
the pressure gradient, ΛP (Pa/m), and the viscosity, μ (Pa*s).
P
K
V ∇ =
μ
1.
This equation can be applied to the impregnation of composites, although slight
modification is required. The fabrics used in composites are comprised of woven bundles
or tows of fibers. Because the width of a fiber tow is typically much greater than the
thickness, flow is approximately one-dimensional and occurs through the thickness of the
tow. Thus, the appropriate one-dimensional form of Darcy’s law is used, written as
P
L q
K
Δ
=
μ
2.
where q is the linear flow rate (m/s) of the fluid, L is the distance the fluid flows (m), and
ΔP is the pressure gradient (Pa) [16-18]. Darcy’s law is used to determine permeability
from experimentally measured values of q, µ, L, and ΔP. Using Darcy’s law in
conjunction with the Kozeny-Carman equation, described below, the Kozeny constant,
K
c
, is then calculated and used for subsequent permeability predictions.
78
The Kozeny-Carman equation can be used to predict the permeability of a preform
from knowledge of the preform microstructure and an estimate of the Kozeny constant,
and thus provides an alternative means of determining permeability. The Kozeny-
Carman equation relates the permeability of a porous material (K) to the fraction of
porosity ( ε); and the surface area per unit volume of fibers, S (m
2
/m
3
), is given as
2 2
3
) 1 ( ε
ε
−
=
S K
K
c
. 3.
In the equation above, the Kozeny constant K
C
is unit-less and is often determined
experimentally by combining Darcy’s Law with the Kozeny-Carman equation. The
constant K
C
varies with flow direction, initial porosity in the preform, uniformity of fiber
packing, and fabric architecture. Experimentally determined values of K
C
range from 2.7
- 18, whereas values determined from computer models vary from 8 - 24 for transverse
flow [4-11]. K
C
is formally equal to K
o
(L
e
/L)
2
[19], where K
o
is the shape factor of the
fibers, L
e
is the effective path length that the resin travels through the prepreg, and L is
the thickness of preform. The shape factor, K
o
, is influenced by the flow direction
relative to the fiber orientation (random, parallel, or perpendicular), while the ratio (L
e
/L)
is the tortuosity of the fiber structure. The remaining factors in the Kozeny-Carmen
equation are readily obtained or determined. For example, the values of S and ε, the
initial porosity of the preform, can be calculated from images of the microstructure, while
the permeability can be predicted or experimentally measured. Once the shape factor is
known, the ratio (L
e
/L) can be determined. The shape factor for a sphere is unity, while
for all other shapes it is less than unity. The shape factor for non-spherical shapes is
calculated from the following equation, with D equal to the fiber diameter [19].
79
SD
K
o
) 1 ( 6 ε −
= 4.
The ratio (L
e
/L) will depend on the fabric architecture, since K
C
varies with fabric
architecture. The ratio (L
e
/L) can be determined for a given fabric architecture once the
permeability, initial porosity of the preform, fiber diameter, and shape factor are known.
Traditional methods of modeling flow of resin in fabric were described. In this work,
a resin infusion rate equation applicable to the VBO process is developed.
7.2.2: Derivation of Flow Time
When produced, prepregs are typically partially impregnated with resin, leaving
vacuum channels (regions of dry fibers) to facilitate escape of gas during final
impregnation and consolidation. Fiber tows typically exhibit lenticular cross-sections,
and resin flow occurs primarily in the short dimension of the tow. For this reason, one-
dimensional creeping flow through the thickness of the fiber tow is assumed. A diagram
of the one-dimensional flow process is shown in Figure 7.1.
80

Figure 7.1: Diagram of the one-dimensional flow process.
P
1

v=dL/dt
L
o

L
P
2

0 x
y
a
a
z
For the geometry described and the assumed conditions, the resin infusion time can be
derived. A drag force on a particle in a viscous fluid is proportional to the fluid viscosity,
flow rate, and particle diameter [20]. The viscous drag force F arises from the applied
pressure gradient ΔP and is defined in the following equations.
NF C L PL
o f y x
= − Δ ) 1 ( φ 5.

N =
y
f y x
L D
L L L
2
4
π
φ
6.
A F τ = 7.
a
V μ
τ = 8.
y
DL A π = 9.
dt
dL
V = 10.
2 2
25 . 0 D a
f
π φ = 11.
where P
1
= applied pressure, P
2
= opposing pressure, ΔP = (P
1
- P
2
) = pressure drop, F =
viscous drag force, τ = shear stress at the surface of the fibers due to flowing resin, μ =
viscosity of the resin, A = the surface area of the fibers, D = the fiber diameter, V = the
flow rate of the resin, N = the number of fibers = (volume of all fibers ÷ volume of one
fiber), a = the average fiber spacing in the fiber tows, C
o
= the fiber interaction
coefficient (described below), φ
f
= the fraction of fibers, L
x
L
y
is the area that the pressure
acts on, and t is the resin infusion time. Appropriate substitutions in Equation 5 yields
the following equations for the pressure gradient ΔP and the infusion time t.
81
dt
dL
Da
L C
P
f o
f
μφ
φ
4
) 1 ( = − Δ 12.
f
f
f o
D P
L C
t
φ
π
φ
μφ
2
2
) 1 (
4
− Δ
= 13.
Note that the fiber interaction coefficient, C
o
, is a function of fabric architecture, and
depends on the arrangement of the fiber tows relative to each other. C
o
increases as the
space between the tows decrease, and can be determined in the Method of Analysis
section.
7.2.3: Method of Analysis
The infusion time given in Equation 13 has the form of γ = m η + b, where γ = t, m =
C
o
, b = 0, and η is given by
f
f
f
D P
L
φ
π
φ
μφ
η
2
2
) 1 (
4
− Δ
= 14.
A plot of γ versus η for different values of viscosity should be linear with slope equal to
C
o
, the fiber interaction coefficient. Previously measured impregnation rates [18] were
plotted against η, and the derived (from Equation 13) and measured flow times were then
equated to determine C
o
, the interaction parameter. The derived flow equation was
generalized to different tow sizes.
Flow rate data on multilayer laminates were also analyzed by the Kozeny-Carman
equation. Values of the Kozeny constant, K
C
and the tortuosity, (L
e
/L), were calculated
from the Kozeny-Carman equation, and the tortuosity for the different fabric architectures
were compared to each other.
82
7.2.4: Materials
For the present work, four fabrics were selected, including plain weave, 4-harness
satin weave, 2 x 2 twill, and 4 x 4 twill weave fabrics. All of the fabrics were produced
from T300 carbon fiber and contained 3000 fibers per tow. Areal densities of the plain
and 2 x 2 twill weave fabrics were identical. The areal densities of the 4-harness satin
and 4 x 4 twill weave fabrics were 4% less and 45% greater than areal densities of the
plain and 2 x 2 twill weave fabrics. Areal densities of the fabrics evaluated in this work
are listed in Table 7.1 and images are shown in Figure 7.2.
Table 7.1: Fabric weight
Fabric Type PW 4-HSW 2 x 2 Twill 4 x 4 Twill
Fabric Weight (g/m2) 194 186 194 281

Plain Weave 2 x 2 Twill
4 x 4 Twill
4-HSW
Figure 7.2: Images of the fabric architectures.
83
7.3: Results and Discussion
As pointed out previously, the equation for the infusion time (Equation 13) has the
form of γ = C
o
η. Thus, an appropriate plot should yield C
o
values characteristic of the
different fabrics. To construct such a plot requires measured values for infusion times.
Tables 7.2 and 7.3 list measured values of the infusion times reported for plain weave, 4-
harness satin weave, 2 x 2 twill, and 4 x 4 twill fabrics [21].

Table 7.2: Infusion times (minutes) measured for plain-weave fabric.
Temperature (°C)
50 60 65 70 80
Infusion Time
(minutes)
48.3 31.7 24 16 6
Number of Layers
1 1 10 1 1

Table 7.3: Infusion times (measured in minutes) for 10-layer laminates at 65 °C [21].
Sample 1 2 3
x
σ
PW 24 24 24 24 0
4HSW 39 45 42 42 3
2 x 2 51 54 51 52 1.7
4 x 4 60 65 65 63.3 2.9

Values of the constants required to calculate the infusion time are listed in Tables
7.4A and 7.4B. Resin contents for the laminates were 54 - 55% by volume. The
measured infusion times (from Table 7.2) were plotted against calculated values of η for
84
the five temperatures listed, yielding the plot shown in Figure 3. Note that the plot shows
a logarithmic (not linear) relationship between γ and η, which is a consequence of the
non-linear relationship between viscosity (µ) and temperature, a plot of which is shown
in Figure 4. Because of the non-linear µ -T relationship, the equation for the infusion
time (Equation 13) was modified accordingly, resulting in the equation below.
ln( ) minutes
o
tC η = 15.
The modified equation above can now be used to determine the fiber interaction
parameter, C
o
. A plot of the measured infusion times of the plain-weave samples versus
the natural log of η is shown in Figure 7.5. The slope yields a value of C
o
for plain-
weave fabric of ~9.8. Interaction parameters for the three other fabrics were determined
by multiplying the value of C
o
for the PW fabric by the ratio of infusion times of the
multilayer laminates relative to that of the PW. The resulting C
o
values for these fabrics
are roughly 2-2.5× greater than that of PW fabric, as shown in Table 7.5.

Table 7.4A: Values of φ
f
, L, D, and ΔP used in Equations 13 - 15.
φ
f

L (m) D (m) ΔP (KPa)
0.67 5.5*10
-4
7*10
-6
100

Table 7.4B: Values of T and μ used in Equations 13 – 15.
T (°C) 50 60 65 70 80
μ (Pa*s) 570.2 109.8 50.75 25.11 7.86

85
t = 9.9771Ln( η) - 0.557
R
2
= 0.9985
0
5
10
15
20
25
30
35
40
45
50
0 20 40 60 80 100 120 140
η
measured infusion time t (minutes)

Figure 7.3: Logarithmic relationship between measured infusion time t and calculated
value of η from Equation 14.

Resin infusion times calculated from Equation 15 varied greatly for the four fabrics
(see Table 7.6), despite the fact that the fabric densities were nearly identical for three of
the fabrics. These differences arise from the different tow separations and distributions
of inter-tow gaps formed at four-tow junctions. For example, the 4 × 4 twill had the
greatest infusion time and exhibited no inter-tow gaps. On the other hand, the PW fabric
showed the most inter-tow gaps and yielded the shortest infusion time. The inter-tow
gaps in the plain-weave fabric were uniform in size and distribution, appearing at each
four-tow junction. In contrast, inter-tow tow gaps in the 4-HSW fabric varied
significantly in size, shape, and distribution, and approximately 23% of the four-tow
junctions had no inter-tow gaps. Gaps between the tows of the 2 × 2 twill structure
generally were smaller in size than those of the 4-HSW structure, and approximately 31%
86
of the four-tow junctions in the 2 × 2 structure had no inter-tow gaps. The 4 × 4 structure
showed no gaps between the fiber tows.
While the interaction parameters listed in Table 7.5 were determined for woven
carbon fabrics that contained 3000 fibers per tow, Equation 15 is general and can be used
to predict infusion times for fabrics with different fiber counts, also. The only parameter
in the equation that changes with tow size is L, the flow path length, and this is easily
measured from micrographs. In the next section, the flow data for multilayer laminates
are analyzed using the Kozeny-Carman equation and the tortuosity values of the fabrics
are compared.
0
100
200
300
400
500
600
320 330 340 350 360
temperature (K)
viscosity (Pa*s)

Figure 7.4: Relationship between viscosity and temperature.

87
t = 9.8125 Ln( η)
R
2
= 0.9982
0
5
10
15
20
25
30
35
40
45
50
01 23 45
natural log ( η)
measured infusion time t (minutes)

Figure 7.5: Determination of the interaction parameter of plain-weave fabric.

Table 7.5: Values of C
o
for each fabric.
Fabric PW 4-HSW 2 x 2 4 x 4
C
o
9.8 17.2 21.2 25.4
Normalized C
o
1 1.8 2.2 2.6

Table 7.6: Calculated Resin Infusion Times (minutes).
T (°C) PW 4-HSW 2 x 2 4 x 4
50 48 84 104 124
60 32 56 69 82
65 24 42 52 63
70 17 30 37 45
80 6 10 13 15
88
89

Values for the flow rate (V
z
), permeability (K
z
), Kozeny constant (K
c
), and ratio of
effective flow length-to-thickness ratio (L
e
/L), are tabulated in Table 7 for each of the
fabrics. The tabulated values show a dependence on the fabric architecture. The flow
rate and permeability decreased and the effective flow length increased as the amount of
inter-tow gaps of the fabric decreased. Recall that the plain-weave fabric had inter-tow
gaps that were consistent in size and distribution, 23% of the fiber tows of the 4-HSW
fabric had no inter-tow gaps, 31% of the fiber tows of the 2 × 2 structure had no inter-tow
gaps, and the 4 × 4 structure had no inter-tow gaps between the fiber tows at the 4-tow
junctions. Previously determined dimensions of the fiber tows [18] were used to
calculate the flow rates. Permeability was calculated from Darcy’s law, and the Kozeny
constant was calculated from the Kozeny-Carman equation. These values were required
to calculate (L
e
/L), described below.

Table 7.7: V
z
, K
z
, K
c
, and (L
e
/L) for the 10-layer laminates made from each fabric.
Fabric V
z
(m/s) K
z
(m
2
) K
c
(L
e
/L)
PW 3.8E-7 1.1E-13 2.2 1.6
4-HSW 2.2E-7 6.1E-14 3.8 2.1
2 x 2 1.8E-7 4.9E-14 4.8 2.4
4 x 4 1.4E-7 4.0E-14 5.8 2.6

90
The microstructural parameters required for calculating permeability via the Kozeny-
Carman equation were determined from micrographs of polished sections (see Figure 7.6
for example). The initial porosity in the fabric was assumed to be equal to the fraction of
resin after full impregnation, because the amount of resin penetration into the fabric
before impregnation was negligible. Porosity and surface area-to-volume ratio were
calculated from Figure 7.6, where the mean value of the porosity was determined to be
~33%. The cross-sections of the fibers were approximately elliptical, and the major and
minor axes were approximately 8 and 6 microns. A shape factor value of 0.86 was
calculated from a diameter of 8 microns, the larger of the dimensions of the fiber cross
section. Mean values for circumference (22.1 x 10
-6
m), and cross-sectional area (37.7 x
10
-12
m
2
) yielded a mean surface-area-to-volume ratio of 586200 m
-1
. Using these values,
the Kozeny constant was calculated for each type of laminate, and from these values,
values for the tortuosity (L
e
/L) were determined (Table 7.7).
The effective flow length L
e
depends on two factors - the arrangement of fiber within
the tows, and the distribution of gaps between tows. Weaving the fiber tows into
different architectural patterns may alter the arrangement of fibers within the tows, and
thus alter the effective flow length. Fewer inter-tow gaps results in an increase in
effective flow length because of the smaller surface area exposed to the resin. In both
cases, the fiber architecture influences the ratio of (L
e
/L).

10 µm
Figure 7.6: Cross-section of the fibers used to calculate the values of S, the surface area
per unit volume of fibers and ε, the fraction of porosity.

Note that if the values of the tortuosity (L
e
/L) are normalized relative to the value for
plain-weave fabric, and then squared, the numbers are equivalent to the normalized
values of C
o
. However, the tortuosity (L
e
/L) and fiber interaction parameter C
o
describe
different properties. The tortuosity describes the relative flow length of resin through the
fabric, whereas the interaction parameter is a constant of proportionality used to fit
experimental data to the flow rate expression (Equation 15). Both parameters depend on
the measured infusion times. Normalized values of C
o
and normalized squared values of
(L
e
/L) can also be determined by normalizing the measured infusion times relative to that
of the plain-weave 10-layer laminate.
The infusion rate equation derived in this work and the Kozeny-Carman (K-C)
equation share certain features, but they also exhibit some important distinctions. For
example, both equations require knowledge of physical parameters that are readily
91
92
measured or known in the design stage. For example, the parameters required for
Equation 14 and the Kozeny-Carman relation (Equation 3) are readily accessible, with
the exception of the Kozeny constant, K
C
. The shape factor can be estimated, but the
tortuosity (L
e
/L) cannot be determined without first determining the value of the Kozeny
constant. The Kozeny constant is usually determined by measuring the resin flow rate,
calculating permeability from Darcy’s law, equating Darcy’s law to the K-C equation,
and solving for K
C
. Thus, the problem intrinsic to the K-C equation is determining an
accurate value of K
C
for modeling permeability without experimentation.
The permeability calculated from the K-C equation depends strongly on the value of
K
C
, which in turn depends on the tortuosity for the fiber array. Recall that the tortuosity
is a function of the geometric configuration of the fibers, and higher values indicate
longer flow path lengths. Inter-fiber flow channels are often occluded because of fiber–
fiber contacts, as shown in Figure 7.6. When vacuum is applied, the compression of the
fabric reduces inter-fiber separations and closes some of the potential flow channels.
Occluded (and partly occluded) flow channels reduce the overall flow rate and
permeability, increasing the value of K
C
. This phenomenon is consistent with previous
reports [7, 22, 23, 24].
One can estimate the tow permeability by measuring the porosity fraction of the
preform and the circumference-to-area ratio (C/A) of the fibers, and subsequently
estimate the Kozeny constant, which is based on the degree of order of the fibers.
However, the accuracy of the estimate is ultimately limited by the uncertainty in K
C
, and
choosing the value of K
C
based on the appearance of the microstructure is admittedly
subjective. The magnitude of K
C
values reported in the literature varies from 2.7 to 24,
93
with a median value of approximately 13. If the median value of K
C
reported in the
literature is used to estimate the permeability, and the true value of K
C
is 2.7, the error in
the estimate can be up to 480% [(13/2.7)100% = 480%] of the true permeability.
The equation derived for the infusion time (Equation 15) provides a simple method
for determining the resin impregnation time of composite parts produced by the VBO
method. All of the parameters required in the equation are accessible or can be readily
measured. Even the interaction parameter, C
o
, can be experimentally determined for
specific fabrics, as demonstrated in the present work. The equation is general and can be
applied to different fabrics and part geometries, although further validation and
experiments will be necessary to determine the limits of utility and degree of accuracy.
7.4: Conclusions
A resin infusion rate equation was derived and fitted to experimental data for film
infusion of VBO prepregs. The new equation can be applied to predict the resin
impregnation rate of any fabric and tow size. Furthermore, all of the parameters used in
the rate equation are typically known at the initial design stage, the only exception being
the fiber interaction parameter, which is readily determined for the particular fabric
architecture. An additional attribute is that the rate equation does not require complex
computer codes, unlike other modeling methods.
The experimental lay-up of the resin films and fabrics used in this work was nearly
identical to the processing of vacuum-bag-only (VBO) prepregs. It was also similar to
the resin film infusion process (RFI). Therefore, both processes can be accurately
described by the infusion rate equation (Equation 15) described here. Additional work
94
with the VBO process using new commercial prepregs will be useful to validate the
infusion rate equation.
The method developed here greatly reduces the uncertainty involved in predicting the
infusion time using traditional approaches, particularly those that rely on the Kozeny-
Carman equation, which is frequently used to model permeability. However, the values
of the Kozeny constant (K
C
) reported in the literature vary by an order of magnitude,
and this large variation can result in errors up to about 480% if the incorrect value of K
C

is used in modeling permeability. Unfortunately, there are no clearly defined guidelines
that describe how to choose the correct value of K
C
for modeling permeability.
Limitations of the Kozeny-Carman equation are overcome with the method described in
this work.

95
7.5: Chapter 7 References
[1] Wang, Y and Grove, SM. Modeling microscopic flow in woven fabric reinforcements
and its application to dual-scale resin infusion modeling. Composites: Part A: Applied
Science and Manufacturing 2008; 39: 843-855.

[2] Verleye, B, Griebel, M, Klitz, M, Lomov, SV, Morren, G, Sol, H, Verpoest, I, and
Roose, D, Permeability of textile reinforcements: simulation, influence of shear and
validation. Composites Science and Technology 2008; 68: 2804-2810.

[3] Loix, F, Badel, P, Orgeas, L, Geindreau, C, and Boisse, P, Woven fabric permeability:
from textile deformation to fluid flow mesoscale simulations. Composites Science and
Technology 2008; 68: 1624-1630.

[4] Astrom, BT, Pipes, RB, Advani, SG. On flow through aligned fiber beds and its
application to composites processing. Journal of Composite Materials 1992; 26; (9):
1351-1372.

[5] Lam, RC, Kardos, JL. The permeability and compressibility of aligned and cross-
plied carbon fiber beds during processing of composites. Polymer Engineering and
Science 1991; 31; (14): 1064-1070.

[6] Bechtold, G. Composites Technologies for 2020. Proceedings of the fourth Asian-
Austrailasian conference on composite materials. 2004; 823-833.

[7] Xiaoming, C, and Papathanasiou, TD. On the variability of the Kozeny constant for
saturated flow across unidirectional disordered fiber arrays. Composites Part A: Applied
Science and Manufacturing 2006; 37: 836-846.

[8] Sullivan, RR and Hertel, KL, The flow of air through porous media. Journal of
Applied Physics, 1940; 11: 761-765.

[9] Bechtold, G and Ye, L, Influence of fibre diatribution on the transvers flow
permeability in fibre bundles. Composites Science and Technology 2003; 63: 2069-2079.

[10] Matteson, MJ and Orr, C, Filtration Principles and Practices, 1987, 2nd. ed., 181.

[11] Lubin, G and Peters, ST, Handbook of Composites, 1998, 2nd. ed., 422, 579.

[12] Pillai, KM, Modeling the unsaturated flow in liquid composite molding processes: a
review and some thoughts. Journal of Composite Materials 2004; 38; (23): 2097-2118.

[13] Parnas, RS and Phelan, FR, The effect of heterogeneous porous media on mold
filling in resin transfer molding. SAMPE Quarterly, 1991; 22; (2): 53-60.
96
[14] Pillai, KM and Advani, SG, A model for unsaturated flow in woven fiber
preformsduring mold filling in resin transfer molding. Journal of Composite Materials
1998; 21; (19): 1753-1783.

[15] Simacek, P and Advani, SG, Modeling resin flow and fiber tow saturation induced
by distribution media collapse in VARTM. Composites Science and Technology 2007;
67: 2757-2769.

[16] Darcy, H. Les Fontaines Publiques de la Ville de Dijon. de Jussieu, Hist. de I’
Acadimie Royale de Sciences 1856; 1733: 351-358.

[17] Elbouazzaoui, O, Drapier, S, Henrat, P. An experimental assessment of the saturated
transverse permeability of non-crimped new concept (nc2) multiaxial fabrics. Journal of
Composite Materials 2005; 39; (13): 1169-1193.

[18] Thomas, S, Nutt, SR, and Bongiovanni, C. In situ estimation of through-thickness
resin flow using ultrasound. 2008;68:3093-3098.

[19] Carman, PC, Fluid flow through granular beds, Chemical Engineering Research and
Design 1937: 15A: 150-166.

[20] Fox, RW and McDonald, AT, Introduction to fluid mechanics, John Wiley and Sons,
1992, 4th ed., 438.

[21] Thomas, S and Nutt, SR, Effect of Fabric Architecture on Through-Thickness
Permeability in Multi-ply Laminates. Submitted to Composites Science and Technology.

[22] Dungan, FD and Sastry, AM. Saturated and unsaturated polymer flows: micro
phenomena and modeling. Journal of Composite Materials 2002; 36; (13): 1581-1602.

[23] Larson, RE and Higdon, JJL. Microscopic flow near the surface of two-dimensional
porous media. Part 2. Transverse flow. Journal of Fluid Mechanics 1987; 178: 119-136.

[24] Rodriguez, E, Giacomelli, F, and Vazquez, A. Permeability-porosity relationship in
RTM for different fiberglass and natural reinforcements. Journal of Composite Materials
2004; 38; (3): 259-268.

97
Chapter 8: Recommendations for Future Work
The vast majority of published work related to out-of-autoclave processes has been in
the resin transfer molding process. Very little work has been published on the vacuum-
bag-only (VBO) process. There is a lot of work to be done in the relatively new field of
VBO processing. Some recommendations for future work are listed below.
1. Work should be extended to commercially available prepregs.
2. Large area transducers capable of recording signal over the entire area of the
transducer are recommended to avoid the time lag due to scanning.
3. Transducers that can be used at least 20 °C above the processing temperatures of other
commercially available prepregs are also recommended.
4. A tank with a temperature controller and circulator should be used.
5. A liquid stable at least 20 °C above the hottest processing temperature of the selected
prepregs should be used in the new tank.
6. Work should be done to determine if a pressure change as a function of the number of
layers in laminates occurs during the period of resin flow in the VBO process.
7. Additional work on other fabric architectures and areal weights should be done to
evaluate the derived resin infusion rate equation.
8. Additional work should also be done with other types of resin films (examples:
toughened epoxies, bismaleimide, etc).

98
Bibliography
Adams, KL, Russel, WB, and Rebenfeld, L. Radial penetration of a viscous liquid into a
planar anisotropic porous medium. International Journal of Multiphase Flow
1988;14(2):203-15.

Adison R.C., McKie A.D.W., Liao T.L.T., and Ryang H.S., IEEE Ultrasonics
Symposium, 1992; 783.

Antonucci, V, Giordano, M, Nicolais, L, Calabro, A, Cusano, A, Cutolo, A, Inserra, S.
Resin flow monitoring in resin film infusion process. Journal of Materials Processing
Technology 2003; 143-144: 687-92.

ASM Handbook, Nondestructive Evaluation and Quality Control, 1990, V. 17, 231 – 261.
ASM Metals Handbook, Desk Edition, 1998, 2nd ed., 1283.

Åström, BT, Pipes, RB, and Advani, SG. On flow through aligned fiber beds and its
application to composites processing. Journal of Composite Materials 1992; 26; (9):
1351-72.

Babu, BZ and Pillai, KM, Experimental investigation on the effect of fiber-mat
architecture on the unsaturated flow in liquid composite molding, Journal of Composite
Materials 2004; 38; (1): 57-79.

Bechtold, G and Ye, L, Influence of fibre diatribution on the transvers flow permeability
in fibre bundles. Composites Science and Technology 2003; 63: 2069-79.

Bechtold, G. Composites Technologies for 2020. Proceedings of the fourth Asian-
Austrailasian conference on composite materials. 2004; 823-33.

Berdivchesky, AL and Cai, Z. Preform permeability predictions by self-consistent
method and finite element simulation. Polymer Composites, 1993; 14; (2): 132-143.

Breitigam, WV, Bauer, RS, May, C. Novel processing and cure of epoxy resin systems.
Polymer 1993; 34 (4): 767–71.

Cai, Z and Berdivchesky, AL. An improved self-consistent method for estimating the
permeability of a fiber assembly. Polymer Composites, 1993; 14; (4): 314-323.

Carman, PC, Fluid flow through granular beds, Chemical Engineering Research and
Design 1937: 15A: 150-66.

Chan, AW and Hwang ST. Anisotropic in-plane permeability of fabric media. Polymer
Engineering and Science 1991; 31; (16): 1233-39.

Cytec Engineered Materials, unpublished data.
99
Darcy, H. Les Fontaines Publiques de la Ville de Dijon. de Jussieu, Hist. de I’ Acadimie
Royale de Sciences 1856; 1733: 351-8.

Drapier, S, Monatte, J, Elbouazzaoui, O, Henrat, P. Characterization of transient through-
thickness permeabilities of non crimp new concept (NC@) multiaxial fabrics.
Composites: Part A, Applied Science and Manufacturing 2005; 36: 877–92.

Dungan, FD and Sastry, AM. Saturated and unsaturated polymer flows: microphenomena
and modeling. Journal of Composite Materials 2002; 36; (13): 1581-1602.

Dungan, FD, Senoguz, MT, Sastry, AM and Faillaci, DA. Simulations and experiments
on low-pressure permeation of fabrics: part I-3D modeling of unbalanced fabric. Jounal
of Composite Materials 2001; 35; (14): 1250-1284.

Elbouazzaoui, O, Drapier, S, Henrat, P. An experimental assessment of the saturated
transverse permeability of non-crimped new concept (nc2) multiaxial fabrics. Journal of
Composite Materials 2005; 39; (13): 1169-93.

Endruweit, A, McGregor, P, Long, AC, and Johnson, MS, Influence of the fabric
architecture on the variations in experimentally determined in-plane permeability values,
Composites Science and Technology, 2006; 66: 1778-92.

Fox, RW and McDonald, AT, Introduction to fluid mechanics, John Wiley and Sons,
1992, 4th ed., 438.

Fu-quan, S and Ci-qun, L. The transient elliptic flow of power-law fluid in fractal porous
media. Applied Mathematics and Mechanics 2002;23;(8); 875-80.

Gebert, BR. Permeability of unidirectional reinforcements for rtm. Journal of Composite
Materials 1992; 26; (8): 1100-33.

Griffin, PR, Grove, SM, Russell, P, Short, D, Summerscales, J, Guild, FJ, and Taylor, E.
The effect of reinforcement architecture on the long-range flow in fibrous reinforcements.
Composites Manufacturing 1995;6(3-4):221-35.

Gutowski, TG, Cai, Z, Bauer, S, and Boucher, D. Consolidation experiments for laminate
composites. Journal of Composite Materials 1987; 21; (7): 651-669.

Hartness, JT, Xu, GF. Resin composition, a fiber reinforced material having a partially
impregnated resin and composites made therefrom. US Patent # 6,139,942, October 31,
2000.

Hartness, JT, Xu, GF. Resin composition, a fiber reinforced material having a partially
impregnated resin and composites made therefrom. US Patent # 6,565,944, May 20, 2003.

100
Hirschbuehler, KR. An advanced composite resin offering flexibility in processing
conditions. 37th International SAMPE Symposium, 1992; March 9 – 12: 452–61.

Jackson K. and Crabtree M., International SAMPE Symposium, 2002; 47: 800.

Johnson, FC, Newsam, SM. Method of manufacturing articles from a composite material.
US Patent # 4,562,033, December 31, 1985.

Klosterman, DA, Saliba, TE. Development of an on-line, in-situ fiber-optic void sensor.
Journal of Thermoplastic Composite Materials 1994; 7: 219–29.

Krautkrämer J. and Krautkrämer K., Ultrasonic Testing of Materials, 1990, 5th ed., 167 –
221.

Kuentzer, N, Simacek, P, Advani, SG, and Walsh, S. Permeability characterization of
dual scale fibrous porous media. Composites Part A: Applied Science and Manufacturing
2006; 37: 2057–68.

Lam, RC, Kardos, JL. The permeability and compressibility of aligned and cross-plied
carbon fiber beds during processing of composites. Polymer Engineering and Science
1991; 31; (14): 1064-70.

Larson, RE and Higdon, JJL. Microscopic flow near the surface of two-dimensional
porous media. Part 2. Transverse flow. Journal of Fluid Mechanics 1987; 178: 119-36.

Lekakou, C and Bader, MG. Mathematical modeling of macro- and micro-infiltration in
resin transfer moulding (RTM). Composites: Part A, Applied Science and Manufacturing
1998; 29: 29–37.

Lekakou, C, Johari, MAK, Norman, D, Bader, MG. Measurement techniques and effects
on in-plane permeability of woven cloths in resin transfer moulding. Composites: Part A,
Applied Science and Manufacturing 1996;27:401–08.

Loix, F, Badel, P, Orgeas, L, Geindreau, C, and Boisse, P, Woven fabric permeability:
from textile deformation to fluid flow mesoscale simulations. Composites Science and
Technology 2008; 68: 1624-30.

Louis, M, Huber, U. Investigation of shearing effects on the permeability of woven
fabrics and implementation into LCM simulation. Composites Science and Technology
2002; 63: 2081–88.

Lubin, G and Peters, ST, Handbook of Composites, 1998, 2nd. ed., 422, 579.

Ma, Y and Shishoo, R. Permeability characterization of different architectural fabrics.
Journal of Composite Materials 1999; 33; (8): 729-51.
101
Markicevic, B and Papathanasiou, TD. A model for the transverse permeability of bi-
material layered fibrous performs. Polymer Composites, 2003; 24; (1): 68-82.

Matteson, MJ and Orr, C, Filtration Principles and Practices, 1987, 2nd. ed., 181.

Mayer, C, Wang, X, and Neitzel, M. Macro- and micro-impregnation phenomena in
continuous manufacturing of fabric reinforced thermoplastic composites. Composites:
Part A, Applied Science and Manufacturing 1998; 29: 783–93.

Michaels, K. Market trends: aerospace composites market will quadruple by 2026. High
Performance Composites. 2007; January 1.

Mitani K. and Wakabayashi K., International SAMPE Symposium, 2001; 46:2293.

Palmer, RJ. Resin impregnation process. US Patent # 4,311,611, January 19, 1982.

Parnas, RS and Phelan, FR, The effect of heterogeneous porous media on mold filling in
resin transfer molding. SAMPE Quarterly, 1991; 22; (2): 53-60.

Pearce, N, Guild, F, and Summerscales, J. A study of the convergent flow fronts on the
properties of fibre reinforced composites produced by RTM. Composites: Part A
1998;29:141-52.

Pearce, NRL, Guild, FJ, and Summerscsales, J. The use of automated image analysis for
the investigation of fabric architecture on the processing and properties of fiber-
reinforced composites produced by RTM. Composites Part A: Applied Science and
Manufacturing 1998; 29: 829–37.

Pillai, KM and Advani, SG, A model for unsaturated flow in woven fiber preformsduring
mold filling in resin transfer molding. Journal of Composite Materials 1998; 21; (19):
1753-83.

Pillai, KM, Modeling the unsaturated flow in liquid composite molding processes: a
review and some thoughts. Journal of Composite Materials 2004; 38; (23): 2097-2118.

Ranganathan, S, Phelan, FR, and Advani, SG. A generalized model for the transverse
fluid permeability in unidirectional fibrous media. Polymer Composites, 1996; 17; (2):
222-230.

Repecka L. and Boyd J., International SAMPE Symposium, 2002; 47: 1862.

Rodriguez, E, Giacomelli, F, and Vazquez, A. Permeability-porosity relationship in RTM
for different fiberglass and natural reinforcements. Journal of Composite Materials 2004;
38; (3): 259-68.

102
Scholz, S, Gillespie, JW, Heider, D. Measurement of transverse permeability using
gaseous and liquid flow. Composites: Part A 2007; 38: 2034-40.

Senoguz, MT, Dungan, FD, Sastry, AM and Klamo, JT. Simulations and experiments on
low-pressure permeation of fabrics: part II-the variable gap model and prediction of
permeability. Journal of Composite Materials 2001; 35; (14): 1285-1321.

Shih, CH and Lee, LJ, Effect of fiber architecture on permeability in liquid composite
molding, Polymer Composites 1998; 19; (5): 627-39.

Simacek, P and Advani, SG, Modeling resin flow and fiber tow saturation induced by
distribution media collapse in VARTM. Composites Science and Technology 2007; 67:
2757-69.

Simacek, P and Advani, SG. A numerical model to predict fiber tow saturation during
liquid composite molding. Composites Science and Technology 2003; 63: 1725–1736.

Steele, M, Corden, T. New prepregs for cost effective out-of-autoclave tool and
component manufacture. SAMPE Journal 2004; 40 (2): 30–34.

Stöven, T, Weyrauch, F, Mitschang, P, Neitzel, M. Continuous monitoring of three-
dimensional resin flow through a fiber preform. Composites Part A: Applied Science and
Manufacturing 2003; 34: 475–80.

Sullivan, RR and Hertel, KL, The flow of air through porous media. Journal of Applied
Physics, 1940; 11: 761-5.

Thomas S., Nutt S.R., and Bongiovanni C., Composites Science and Technology, 2008;
68, 3093-8.

Thomas, S and Nutt, SR, Effect of Fabric Architecture on Through-Thickness
Permeability in Multi-ply Laminates. Submitted to Composites Science and Technology.

Thomas, S, Bongiovanni, C, and Nutt, SR. In situ estimation of through-thickness resin
flow using ultrasound. Composites Science and Technology 2008; 68: 3093-98.

Thorfinnson, B, Biermann, TF. Degree of impregnation of prepregs – effect on porosity.
32nd International SAMPE Symposium, 1987; April 6-9: 1500–09.

Verleye, B, Griebel, M, Klitz, M, Lomov, SV, Morren, G, Sol, H, Verpoest, I, and Roose,
D, Permeability of textile reinforcements: simulation, influence of shear and validation.
Composites Science and Technology 2008; 68: 2804-10.

Wallstrom, TC, Christie, MA, Durlofsky, LJ, and Sharp, DH. Effective flux boundary
conditions for upscaling porous media equations. Transport in Porous Media
2002;46:139-153.
103
Wang, Y and Grove, SM. Modeling microscopic flow in woven fabric reinforcements
and its application to dual-scale resin infusion modeling. Composites: Part A: Applied
Science and Manufacturing 2008; 39: 843-855.

Williams, CD, Grove, SM, and Summerscales, J. The compression response of fibre-
reinforced plastic plates during manufacture by the resin infusion under flexible tooling
method. Composites: Part A 1998;29:111-4.

Wong, CC and Long, AC, Modelling variation of textile fabric permeability at
mesoscopic scale, Plastics, Rubber, and Composites, 2006; 35; (3): 101-11.

Wu, X, Li, J, Shenoi, RA. Measurement of braided perform permeability. Composites
Science and Technology 2006; 66: 3064–69.

Wu, X, Li, J. A new method to determine fiber transverse permeability. Journal of
Composite Materials 2007;41(6):747-56.

Xiaoming, C, and Papathanasiou, TD. On the variability of the Kozeny constant for
saturated flow across unidirectional disordered fiber arrays. Composites Part A: Applied
Science and Manufacturing 2006; 37: 836-46.

Xu G.F., Repecka L., and Boyd J., SAMPE International Symposium, 1998; 43: 9.

Xu, GF, Repecka, L, Mortimer, S, Peake, S, Boyd, J. Manufacture of void-free laminates
and use thereof. US Patent # 6,391,436, May 21, 2002.

Zhou, F, Kuentzer, N, Simacek, P, Advani, SG, and Walsh, S. Analytical characterization
of the permeability of dual-scale fibrous porous media. Composites Science and
Technology 2006; 66: 2795–2803.

Abstract (if available)

Abstract Ultrasonic imaging in the C-scan mode in conjunction with the amplitude of the reflected signal was used to measure flow rates of an epoxy resin film penetrating through the thickness of single layers of woven carbon fabric. Assemblies, comprised of a single layer of fabric and film, were vacuum-bagged and ultrasonically scanned in a water tank during impregnation at 50°C, 60°C, 70°C, and 80°C. Measured flow rates were plotted versus inverse viscosity to determine the permeability in the thin film, non-saturated system. The results demonstrated that ultrasonic imaging in the C-scan mode is an effective method of measuring z-direction resin flow through a single layer of fabric. The permeability values determined in this work were consistent with permeability values reported in the literature. Capillary flow was not observed at the temperatures and times required for pressurized flow to occur. The flow rate at 65°C was predicted from the linear plot of flow rate versus inverse viscosity.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Defect control in vacuum bag only processing of composite prepregs

PDF

Void evolution in vacuum bag-only prepregs

PDF

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

PDF

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

PDF

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

PDF

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

PDF

Polymer flow for manufacturing fiber reinforced polymer composites

PDF

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

PDF

Development of novel 1-3 piezocomposites for low-crosstalk high frequency ultrasound array transducers

PDF

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

PDF

Processing, mechanical behavior and biocompatibility of ultrafine grained zirconium fabricated by accumulative roll bonding

PDF

Molecular design strategies for blue organic light emitting diodes

Asset Metadata

Creator Thomas, Shad (author)

Core Title Vacuum-bag-only processing of composites

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Materials Science

Publication Date 03/02/2009

Defense Date 01/22/2009

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

Tag OAI-PMH Harvest,out-of-autoclave,prepreg,resin film infusion,ultrasound imaging,VBO

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Nutt, Steven R. (committee chair), Armani, Andrea M. (committee member), Sammis, Charles G. (committee member)

Creator Email s.s.thomas@att.net,slthomas@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1989

Unique identifier UC1316698

Identifier etd-Thomas-2447 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-147292 (legacy record id),usctheses-m1989 (legacy record id)

Legacy Identifier etd-Thomas-2447.pdf

Dmrecord 147292

Document Type Dissertation

Rights Thomas, Shad

Type texts

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

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu