Polymer Testing 9 (1990) 363-377

Reducing the Void Content and its Variability

in Polymeric Fibre Reinforced Composite
Test Specimens Using a Vacuum Injection
Moulding Process

Freddy Y. C. Boey
Nanyang Technological Institute, Nanyang Avenue, Singapore 2263

(Received 7 March 1990; accepted 5 April 1990)

ABSTRACT
To obtain acceptable accuracy and consistency in the mechanical testing
of polymeric fibre reinforced composities, it is important that both the
fibre alignment and the void content be precisely controlled. In
particular, a high void content can greatly affect both the variability and
the value of the mechanical strengths of the composite. This paper
describes the use of a simple vacuum injection moulding process to
produce fibre reinforced composites that have a consistently low void
content, of about 1.3%, and a correspondingly high and consistent
flexural strength.

INTRODUCTION

To obtain acceptable accuracy and consistency in the mechanical testing

of polymeric fibre reinforced composites, it is important that both the
fibre alignment and the void content be precisely controlled. Thermo-
setting fibre reinforced composite test specimens are usually fabricated,
using either a wet lay up process or a prepreg approach. In either
approach, it is necessary that a vacuum system be used to evacuate
voids occuring in the matrix resin. This is important since variations in
the percentage of voids can cause considerable differences in the
mechanical strengths of the final composite material. 1-3
The use of prepreg material is often costly and involves difficult
transportation and storage problems, particularly if the source company
363
Polymer Testing 0142-9418/91/$03.50 1991 Elsevier Science Publishers Ltd, England.
Printed in Northern Ireland
364 Freddy Y. C. Boey

is overseas. The wet lay up approach, although less expensive, involves a

very tedious laying up process caused mainly by the relatively high
viscosity of the resin. To minimise void content, t each layer of fibre
cloth is usually rolled on, thus making the process rather time
consuming. More significantly, misalignment of the fibres and uneven
thickness often results during the lay up, often causing a considerable
amount of scrap. This is a particular liability for the production of test
specimens, where the fibre alignment has to be accurately known, and
the thickness should be consistent.
The vacuum injection moulding (VIM) process on the other hand can
result in consistently low void content, accurate fibre alignment and
thickness, since the initial fibre lay up is not preformed with any resin
application. Once the fibre lay up is accomplished, a vacuum bagging is
put over the mould and lay up, along with the usual release cloth,
breather material, etc. Vacuum is then effected, thereby holding the
fibres in place, as the resin is pulled into the fibre network by the
vacuum. The process is also faster as a result, since the injection is
carried out automatically. Additionally, direct contact with toxic or skin
irritant resin systems are kept to a minimum.
The present paper describes how a Simple VIM process can be used
to produce flat test specimens for mechanical testing, which have
consistently low void content, as well as good fibre alignment.
In any polymeric composite process, the presence of voids are
inevitable and also signficantly adverse. A high void content indicates
that the resin has not thoroughly embedded the fibres, resulting in the
following:

1. A weaker interfacial strength due to inadequate adhesion,

resulting in the reduction of the composite strength and stiffness
properties.
2. Mutual abrasion of the fibres, leading to fibre damage and
fracture.
3. Crack initiation and crack growth due to void coalescence.

One of the main problems in achieving strength maximisation and

consistency in polymeric composites is the existence of voids in the final
product. While the problem is generic to almost all polymeric matrices,
including thermoplastics, this paper focuses on the problems of voids
mainly for an epoxy based matrix. The initial occurrence of voids can
come from many sources, amongst themP

1. Dissolved volatiles within premix resins.

2. Bubbles evolved from the mixing process itself.
 Reducing void content and its variability using VIM 365

3. Voids created by the presence of various e m b e d d e d particles as

well as loose fibres.
4. Air pockets trapped during lay up of the fibres and resin.
5. Ply terminations.

The use of a vacuum bagging system helps to reduce the existing

amount of voids. It has, however, only a limiting effect, due to two
reasons.
Firstly, the viscosity of the resin changes during the cure process,
initally decreasing in value as the temperature increases, and then
increasing very quickly due to the effect of gelation, resulting in
extensive crosslinking within the material. The effect of the vacuum is
at a maximum when the viscosity is lowest; once it increases in value,
especially during gelation, the vacuum suction would become increas-
ingly ineffective. Obviously then, it is important to effect the vacuum
suction until the initiation of gelation. This is one of the reasons for
studying the gelation process for the composite.
The second limiting reason for the vacuum is because to the effect of
partial pressures. Industrial vacuum pumps are usually mechanical
pumps, and these usually experience a degree of back pressure, limiting
their extraction of volatiles from the fluid. The a m o u n t of voids
occurring within the curing c o m p o n e n t varies according to the process
parameters. As the temperature increases, nucleation of dissolved
volatiles, particularly on e m b e d d e d particles, as well as expansion or
even coalescence of existing voids, occur. In particular, because of the
very small critical nucleus size for the formation of stable hetero-
geneous nucleation of water, absorbed moisture within both the fibres
and the resin would result in significant void formation. 4"5
While it is not practical industrially to attempt to eliminate the
absorbed water in both fibres and resin prior to curing, a more practical
approach is to control the growth of these nucleii. A detailed analysis of
the kinetics of void growth can be found in Refs. 5-8. To further
decrease the void content, a combination of pressure and vacuum can
be used, with the pressure usually applied externally to vacuum bagging
from an autoclave vessel. Using such a combination, the void content
can be brought down from 5 to 3%, when the mechanical strengths
within the composite material becomes less variable. 1.3

EXPERIMENTAL PROCEDURE

Figure 1 shows a schematic illustration of the mould set up. The mould
was made of a flat 10 m m thick alumium plate, with a 5 m m deep and
366 Freddy Y. C. Boey

- 50 800
0
In

RESIN
INLETS
0
0
O3

RESIN
CHANNEL .L
. . . . . . . . . . . . . . . . II1" . . . . . . . . . . . . . .

t t t

]q V- --~-s-
I I I I II
S
4..
II I I i I
,, . I II

HOSETAIL FITTING
(ENLARGED)

Fig. 1. Schematic illustration of the fiat mould.

 Reducing void content and its variability using VIM 367

(
E

\ i

\l

O
E

7
O

E
~ t~

~4
~n ~Z

0
368 Freddy Y. C. Boey

10 m m wide resin channel milled all around it for the resin intake. Each
side was given an excess of 25 m m for sealing the bagging sheet. Four
resin inlets were made, allowing for even and thorough flow of the
resin.
Figure 2 shows the schematic illustration of the whole vacuum
injection moulding process. A vacuum probe connected the vacuum
pump to the vacuum bagging. The resin trap was necessary to ensure
that no resin penetrated into the vacuum pump. The probe used was a
simply fabricated aluminium disc with an adhesive tape connection to
the vacuum bagging (Fig. 3). A commercially available quick discon-
nect probe can also be used, 9 although the former simple probe was

NYLON PERFORATION
VACUUM BAGGING

SECTION A-A

Fig. 3. Illustration of the hosetaii fitting vacuum probe.

 Reducing void content and its variability using VIM 369

found to suffice. For clarity, only a one-way vacuum probe is shown on

Fig. 2. In practice, it was found that a multiprobe system attached to
the same single vacuum p u m p gave a faster and more even inpregnation
of the resin.
Figure 2 also shows the lay up sequences. A perforated release film
separated the fibre mat from the sacrifical breather layer. It also helped
to ensure that the finished surface was flat and unwrinkled, which could
result from the folding of the nylon bagging during evacuation. The
breather layer served both to provide a vacuum line into the resin, as
well as to soak up any excess resin. A sufficiently thick layer was found
necessary to avoid the resin from thoroughly soaking the breather
layer, thereby choking up the vacuum line. A disposable nylon bagging
was then used to effect the vacuum. While the nylon bagging used was
relatively cheap, it also sometimes contained pinhole defects, which
considerably lowered the effective vacuum. The nylon bagging was
checked for pin hole defects using soap solution. This was done before
resin intake, since repair of the bagging was more easily done prior to
resin intake. To eliminate this problem, reusable silicone rubber sheets
could also be used, 10 although this would be more costly.
The lay up process started with a release agent layer sprayed on to
the mould surface. (Frekote 700 proved to be particularly good in
yielding a smooth surface finish). The required layers of fibre mats
were then carefully laid on. Because there were no resin applications,
adjustment of the fibre mats to obtain precise alignment was easy. A
small amount of resin was used to tack down the fibre mats or tapes
accurately. This was in contrast to the normal wet lay up process, where
the resin is immediately rolled on to each of the fibre layer, after which
realignment was very difficult. A perforated release film (Teflon sheet)
was added between the breather and the fibre mats to reduce the flow
of the resin into the breather. A resin trap in series to the vacuum line
ensured that resin did not get into the vacuum pump. A mechanical
pump capable of going down to 0.1 bar was used, although in practice
the achievable vacuum was at most 0-2 bar. The reason for this is
because industrial mechanical pumps usually experience a degree of
back pressure caused by the partial pressure of the volatiles in the
resins system. Vacuuming was carried out right until resin gelation
occured. After this, due to the extensive crosslinking, resulting in a
very viscous resin, the vacuum became ineffective, and so was turned
off.
The time to gelation d e p e n d e d on the resin system used, as well as on
the curing temperature. Hot curing resins have lower viscosities than
cold curing resins before gelation, and therefore impregnated the fibres
370 Freddy Y. C. Boey

faster and more evenly. Hot curing inside a standard oven, or even in
an autoclave, is possible. For this project, all hot curing systems were
cured inside an autoclave (designed in our laboratory) which has a
built-in vacuum hose system.
Cold curing resins have higher viscosities and took a longer time to
thoroughly impregnate the fibre network. Experience showed that in
using cold curing systems, it was important to ensure that sufficient time
was available for the resin to penetrate into the fibre network
thoroughly before gelation commenced. If gelation is initiated during
resin intake, due to the rapid increase in the resin viscosity, resin intake
would rapidly decrease in speed, with the viscosity still increasing all
the time. Eventually, the vacuum line would get totally choked up. The
lag time to gelation initiation is often very sensitive to the amount of
catalyst or hardener added to the cold curing resin systems. A higher
amount of catalyst or hardener results in a faster cure time, but also a
shorter lag time to gelation initiation. It is, therefore, necessary to
optimise between the two time factors to obtain the ideal ratio of
catalyst/resin.
For the work reported here, the Eporez 15 system, which is a low
viscosity modified bisphenol A-epichlorohydrin epoxide resin, was
used with a T E T A curing agent. The system averaged about 0-5 m/h
for a single vacuum probe/double resin inlet system, which thoroughly
impregnated the flat panel, shown in Fig. 1, in less than 30 min. The
resin system started to gel significantly in 45 min. By using a double
vacuum probe/double resin inlet system, the time to total inpregnation
was reduced by about 50% to 13 min, thus allowing for the use of a
slightly higher amount of catalyst and shortening the curing time.
The fibre mats used were from an E glass cross woven (0, 90) fibre
system. Hot curing was effected in an autoclave (designed in our
laboratory) capable of 20 bars pressure, 50 mbar maximum vacuum and
a temperature of up to 200C. Both vacuum and pressure were
controlled with a Honeywell multicontroller. Details of the autoclave as
well as the lay up procedure can be found in Ref. 1. Although the
vacuum pump was capable of going down to 50 mbar, in practice, due
to the high partial pressure caused by the resin volatiles and the
relatively high resin viscosity, the lowest consistently attainable vacuum
was about 0.25 bar.
For all specimens, 24 plies of the cross woven fibre glass were used to
produce a panel of 2.5 mm x 300 mm x 600 mm dimensions, from which
specimens of dimensions 2-5 mm x 25 mm x 80 mm were cut out for
both the mechanical strength tests and the void content tests.
Determination of the void content was performed according to
 Reducing void content and its variability using VIM 371

A S T M D2734, u using a 70% nitric acid solution for digestion. For

statistical significance, each test was repeated a m i n i m u m of five
times, H each specimen being taken from a different, but independently
produced panel. Each test comprised of a 30 x 25 m m coupon, with the
density determined by weight ( + 0-001 g) and volume ( + 0.01 m m )
measurements. The specimen was then digested in 40 ml of nitric acid
in a beaker for 18 h at 23 C, the remaining fibres then being washed in
water and acetone, oven dried (150 C) and weighed. The fibre volume
was determined for each sample tested according to A S T M D3171.12
Details of the calculations to determine the percentage fibre volume
and the void content can be found in Refs 11 and 12. Four-point
bending tests were performed in accordance to ASTM-D790M, 13 using
a JJ tensile testing machine with a four-point bending test fixture
designed to be operated in tension. ~4 The reason for preference of a
bend test over a tensile test is that for composites, the failure m o d e and
therefore the value of the compressive and tensile failure stress differ. ~~
A bend test incorporates both modes in the loading. ~6 A four-point
bend test is used to avoid flaws created by the load indentation that
occurs in a three-point bend test. The length to thickness ratio was
specified at 32.
For statistical significance, each test was repeated a m i n i m u m of five
times, ~3 with each specimen taken from a different, but identically
produced composite panel. Each data point shown in the results
represents the averaged result. Hot curing was performed for 10 h at
120 C, with a 5 h postcuring at 140 C. For comparison purposes, five
panels that were conventionally hand laid-up were also tested, with the
same curing temperature profile indicated.

RESULTS AND DISCUSSION

Figure 4 shows the result of the void content and the flexural strength
versus the specimen position across the diagonal of the composite
panel, when only one probe was used, placed at the centre of the panel,
as shown in Fig. 1. The horizontal axis shows the specimen position,
with positions 1 and 8 at both ends of the diagonal, and the other
positions spaced equally along the diagonal. Variation in the percentage
void content was large, with a lower void content at the ends. This was
because the resin intake (Figs 1 and 2) came from the end perimeter of
the panel, so that the resin would have had a longer time to saturate
the fibres. Results for the conventionally hand laid-up panel are shown
for comparison (dashed line), averaging between 3 to 5% void content.
372 Freddy Y. C. Boey

~=)

0 0

(%) N ~ N O D ~IOA 0

0 O
o
< ..~~
:>
0
2:
I'
m ' o J e-~r-
m ~

I
I
/\ /', o

ed~

', /\/ - ~'~.~

= C

MM

E~

a~

.~ ~,

I I i I I I I I .~ m,
0 0 0 0 0

.= .
 Reducing void content and its variability using VIM 373

TABLE 1
Properties for the Three Processes

Property VIM process VIM process Normal

(one probe) (two probes) process

Fibre volume (%) 28 33 31

Apparent density (g cm 3) 1-27 1.48 1.42
Void content (%):
Average a 8.5 1.34 3-5
Standard deviation (.. n - 1) 8-90 0.28 0-90
Flexural stress (MPa):
Average a 83 121 111
Standard deviation (.. n - 1) 31.7 3-5 5.0

"For a total of 35 specimens.

Table 1 compares the percentage fibre volume, apparent composite

density, averaged percentage void and flexural strength with their
corresponding standard deviations. Earlier reports have shown that for
consistent results, a void content of at most 3% was necessary. 2"3 For
the panels produced using the VIM process, even with one probe at the
centre, the void contents at the ends were below 3 5: 0-5%. However,
unacceptably high void contents were recorded at the centre. Cor-
respondingly, the flexural strength achieved was also very poor. Table 1
indicates that the strength variation is unacceptably high.
Figure 5 shows the results for the VIM process, with two probes
placed 200 mm from each other, and also 200 mm from the ends of the
panel, both aligned at the middle of the panel width. Results for the
percentage void content indicate a consistent reading throughout the
panel, averaging only 1.4 + 0.5% with a standard deviation of 0-28%.
The averaged percentage void value for the normally processed panel
was 3.4 + 0.5%, with a standard deviation of 0-89%. The VIM process
produced both a lesser averaged void percentage as well as a lower
variation.
Figure 5 also shows the average flexural stress for both panels. The
normally processed panel averaged 110 MPa, with a standard deviation
of 5 MPa (Table 1). The corresponding values for the VIM panel were
120.5 and 3.5 MPa, respectively. Again, the strength obtained was both
slightly higher as well as more consistent throughout the panel.
Figure 6 shows the influence of the void content on the flexural
strength. Square symbols represent the results from the one-probe VIM
374 Freddy Y. C. Boey

(%) ~N~NOD ~IOA 0

'= "6

9
0
! l

o~0
H
>

< o

~d
> r~.d

!
! t9
<

I >
0 .- ~ ~
~ ~o

,~ ~-~.~

~ o
0
E~
(.) <
N~
X e~ i"2

Nx~
,_,.~
I
I I
I I I I !

0 o o o o 0
~D
~N
N e-
No
(~dN) SSZH~S q~H~xzqd
 Reducing void content and its variability using VIM 375

I
I
0
I
I
I GO
>
I ,D
e~
I p
I tO
Lll
I
/
/
I / ~E
/
'!=" Z
/
/ m O
O u
0 0 ,f,-
I ~ cZ
c'-i
H
/ z ~:
O ~
/
",,-' O
m m o
i m mm tD

i m ~<

> > 2;

Oq O
.,.-..,

I II I I I I I

0 0 0 0 0 0 0 0 O

>
(~dW) SS~H~S qv'c~X~qd
376 Freddy Y. C. Boey

process, triangles represent results for the two-probe V I M process,

while circles represent results from the normal process. The strength
increases fairly linearly with the decrease in void content, until about
3 - 4 % , when it became more consistent. Figure 6 clearly shows that
using a two-probe V I M process, good consistency in the strength results
can be obtained.

CONCLUSION

Composite test panels were produced using a vacuum injection mould-

ing process. Using a two vacuum probe method, the void content
throughout the panel was shown to be both lower and more consistent,
and the process time faster than that for a normally laid-up panel.

ACKNOWLEDEGMENT

The author would like to gratefully acknowledge the considerable help

given by Mr Ching Boon Khian.

REFERENCES

1. Boey, F. Y. C. & Lye, S. W., Effect of pressure and vacuum on the

autoclave curing of a thermosetting fibre reinforced composite. J. Mater.
Processing, 20 (6) (1990).
2. Stringer, L. G., Correlation between compression strength and void
content in hand layed up/vacuum bagged CFRP composites. RARDE
Internal Rep., Christchurch, UK, 1987.
3. Stringer, L. G., Optimisation of the wet lay up/vacuum bag process for
CFRP composites. Composites, 20 (5) (1989) 441.
4. Kardos, J. L., Dave, R. & Dudukovic, M. P., Voids in Composites. In
Proc. Symp. Manufacturing Science of Composites, ed. T. G. Gutowski,
ASME, Atlanta, 1988. p.41.
5. Kingery, W. D., Introduction to Ceramics, John Wiley, New York, 1960,
p. 291.
6. Subramaniam, R. S. & Weinberg, M. C., The role of convective transport
in the dissolution of a gas bubble. J. Chem. Phys., 72 (1980) 6811.
7. Tao, L. N., Dynamics of growth or dissolution of a gas bubble. J. Chem.
Phys., 69 (1978) 4189.
8. Tao, L. N., The general solution and its analyticity for growth or
dissolution of a gas bubble. J. Chem. Phys., 71 (1979) 3455.
9. Boey, F. Y. C., Development of an autoclave curing process for fibre
reinforced composites. Polymer Testing, 8 (6) (1989) 375.
 Reducing void content and its variability using VIM 377
10. Boey, F. Y. C., Development of a reusable vacuum bagging technique for
autoclave curing of fibre reinforced polymer composites, Experimental
Techniques, 13 (9) (1989) 21.
11. ASTM D2734-70 (Reapproved 1985), Void content of reinforced plastics.
ASTM Standards and References for Composite Materials. 1987, p. 441.
12. ASTM D3171-76 (Reapproved 1982), Fibre content of resin matrix
composites by matrix digestion, ASTM Standards and References for
Composite Materials, 1987, p. 41.
13. ASTM D790M (Reapproved 1985), Flexural tests for reinforced plastics.
ASTM Standards and References for Composite Materials, 1987, p. 360.
14. Teo, Y. M., Development of an autoclave process for fibre reinforced
composites. FYP Report 1989/90, MPE, Nanyang Technical Institute, p.
34.
15. Boey, F. Y. C., Microbuckling compression failure of a radiation induced
WP composite. Comp. Sci. Technol., 1990 (in press).
16. Boey, F. Y. C. and Teoh, S. H., Non linear creep rupture analysis of a
composite material. Mater. Sci. Technol., 1990 (in press).