COMPOSITES

SCIENCE AND
TECHNOLOGY
Composites Science and Technology 64 (2004) 1863–1874
www.elsevier.com/locate/compscitech

Environmental effects on the mechanical properties

of glass-fiber epoxy composite tubular specimens
Fernand Ellyin *, Rachel Maser
Department of Mechanical Engineering, Advanced Composite Materials Engineering Group, University of Alberta,
Edmonton, AB, Canada T6G 2G8
Received 9 July 2003; received in revised form 28 January 2004; accepted 29 January 2004
Available online 8 April 2004

Abstract

The main objective of this experimental investigation was to study the effects of moisture absorption and exposure to elevated
temperature on the mechanical properties of glass fiber reinforced epoxy composite tubes. Filament wound tubular composite
specimens were immersed in distilled water at two different temperatures for approximately four months and their moisture ab-
sorption was recorded. The rate of moisture absorption was greater for the group of specimens immersed in distilled water at an
elevated temperature (50 °C) than those at room temperature (
20 °C). Multi-axial tests were performed subsequently, and it was
observed that for all biaxial stress ratios, strength and stiffness decreased to some extent with the presence of moisture and increasing
temperature. Strains parallel and transverse to the fibers at functional failure (leakage) showed little variation in the presence of
moisture and temperature with the exception of pure hoop loading. Microscopically, the greatest water damage was apparent in the
matrix and at the fiber–matrix interface where there was less resin adhesion to the fibers with increasing water temperature. Fiber
strength was also negatively affected, possibly due to either leaching out of the glass fibers interface layer or glass fiber embrittle-
ment, and less effective bonding and load distribution at the fiber–matrix interface.
Ó 2004 Elsevier Ltd. All rights reserved.

Keywords: Distilled water; A. Glass fibers; B. Mechanical properties

1. Introduction This paper will consider the effects of immersion in

distilled water at two temperatures, on the mechanical
Polymer matrix composite materials offer many su- properties of filament wound glass fiber reinforced ep-
perior properties in comparison with conventional ma- oxy composite tubulars. Using distilled water, an ag-
terials, among them high strength to weight ratio, gressive form of moisture because of the lack of
superior corrosion resistance and the ability to tailor the impurities and ions, the worst possible scenario of water
material system to a particular requirement. The use of damage will be simulated. In the case of buried pipes
fiber reinforced polymer matrix composites in pressure and storage vessels in ground however, the concentra-
retaining structures such as pressure vessels and piping tion gradient would be less steep, therefore, the damage
systems has been confined mostly to low pressure ap- through moisture absorption would be gradual as dis-
plications when no internal liner is used. The reason for tilled water encourages diffusion to proceed at a much
this limited use is mainly due to the uncertainty with higher rate. Similarly, the ground may somewhat buffer
respect to the long term reliability of polymer matrix the seasonal variations in the external temperature, the
composites under natural environment (moisture and structure will experience a range of temperatures
temperature) and multi-axial loading. throughout the years and at different locations. Here the
exposure to moisture and temperature is simulated by
*
Corresponding author. Tel.: +1-780-492-3598; fax: +1-780-492-
submerging tubular specimens in distilled water at room
2200. temperature (
20 °C) and 50 °C temperature. Further-
E-mail address: fernand.ellyin@ualberta.ca (F. Ellyin). more, tubular structures are often subjected to loads

0266-3538/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compscitech.2004.01.017
1864 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874

which result in multi-axial stress states. To simulate this, properties of the weakest link, the matrix, will dominate
the exposed (aged) tubes will be tested under four dif- for most cases of fiber reinforcement.
ferent in-plane bi-axial stress ratios to assess the influ- Finally, it has been extensively shown that E-glass
ence of water absorption on the mechanical properties. can be susceptible to moisture and temperature damage
in the form of stress corrosion cracking (SCC) due to ion
diffusion on the surface of the fibers [7,8]. Jones [4] has
documented the phenomenon of environmental SCC in
2. Background water. Leaching of ions such as Ca2þ , Al3þ and Fe3þ can
lead to a reduction in the load carrying diameter of the
The effects of an aqueous environment and variable fibers. In this experimental investigation, the fibers were
temperature on the properties of prepreg glass-fiber stressed during the filament-winding process [9] and
epoxy resin coupon specimens have been documented in once the thermosetting epoxy resin cures, the fibers re-
[1] but few investigations have covered environmental main tensioned for their life duration.
effects on filament-wound tubular specimens, e.g., see The aims of this paper are: (1) to monitor the mois-
[2,3]. Although in both cases the composites consist of ture absorption of glass-fiber reinforced epoxy com-
epoxy resin and E-glass fibers, the difference lies in the posites at two different temperatures; (2) determine in
geometry and manufacturing of each specimen group what manner the mechanical multi-axial properties are
and the number of the diffusion surfaces. In the case of affected; (3) to microscopically observe the damage in-
the coupon specimens, moisture can diffuse into the duced by moisture absorption on the fracture surfaces of
specimen through six edges while there are only two the test specimens and finally; (4) to establish a corre-
surfaces (interior and exterior) on a tubular specimen. lation between moisture absorption, temperature and
Also, the moisture will diffuse faster along the fibers on mechanical behavior of the composite tubulars.
the cut-edges of the coupons. In the case of the tubes,
the moisture would first have to penetrate through the
matrix to the fiber–matrix interface and then it could 3. Specimen preparation
diffuse quickly along the length of the fibers. Thus, the
rate of moisture absorption in the tubular specimens is In this investigation, the test geometry consisted of a
much slower leading to a lower saturation moisture lay up of six layers of ½603 T , the angle being measured
content. Another difference is the fact that the coupon with respect to the central axis of the tube. The same
specimen fibers are not under tension while those of the manufacturing and testing procedures and equipment as
tubular specimens are tensioned during the filament in [9] was used. The specimens were manufactured in-
winding process. However, tensile and compressive house using a filament winding process. Tows of glass
stresses are present in both forms of specimens as a re- fibers from eight creels were run through a resin bath
sult of residual stresses introduced during the curing containing EPON 826 epoxy resin mixed with EPI-
cycle. CURE curing agent 9551 hardener. The tows ran to the
The most significant result of moisture absorption in filament-winding machine, which guided the tows as
composite materials is plasticization of the matrix and they were wound around a chrome-plated mandrel with
thus a lowering in glass transition temperature, Tg , a tension of 26.69 N. Winding angle and speed were
rendering the matrix softer and weaker [4,5]. In addi- both controlled by a computer software program. The
tion, non-uniform swelling of the matrix can lead to the curing cycle consisted of 60 min at 80 °C, followed by
introduction or relief of stresses. It should also be 150 min at 120 °C and a cooling period of 120 min
mentioned that the coefficients of thermal expansion of ramping down to 20 °C. There was a 15 min ramping
the matrix and fibers differ by an order of magnitude period between each temperature change. While curing
[4,6]. Thermal residual stresses introduced during cool- in the oven, the tubes were rotated to maintain even
ing after the curing temperatures may be increased distribution of the resin. The composite tube was ex-
or alleviated by a temperature change in an aqueous tracted from the mandrel using a Tinius Olsen material
environment. testing system. Following extraction from the mandrel,
The strength of bonding at the fiber–matrix interface specimens were roughly cut into 203 mm (8 in.) length
has also been shown to degrade in the presence of using a band saw and then more precisely using a rotary
temperature and moisture. It has been hypothesized that cutting device. The approximate dimensions of the tu-
rather than diffusing through the matrix – a slow and bular specimens were a 40.6 mm outer diameter with an
temperature dependent process – moisture will diffuse inner diameter of 38.1 mm. The fiber volume fraction of
with more ease along the fibers and destroy the bond at the specimens was 70.8%.
the fiber–matrix interface [7]. The fiber–matrix interface In order to avoid the possibility of water being ab-
plays a pivotal role in composite strength by load sorbed through the cut ends, and potentially debonding
transfer. Without a strong bond at the interface, the the matrix from the fibers by running along the fiber–
 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874 1865

matrix interface, the ends were dipped in melted crys-

talline wax and left to cool. The wax used was of type
M-coat W-1, supplied by the Micro-Measurements
Group. The wax has a melting temperature of 75 °C, an
operating range from )20 to 65 °C and is known to
provide excellent water-immersion coating. In addition,
it does not react chemically with the epoxy resin matrix
nor does it degrade in an aqueous environment. The one
disadvantage (in retrospect) is that the wax although
firm, is not solid and therefore can be dislodged without
much effort. This problem was greatest in the 50 °C bath
because the wax was relatively soft. The tubes were
Fig. 1. Schematic of mounted test specimen [19].
immersed in the upright position in water bath and with
every weighing session a minute amount of wax was left
on the bottom of the water bath and on the scale-
weighing surface. In water at 20 °C however, the wax were bonded and the rubber bladder in place, larger
was harder and therefore quite effective. grips were mounted to bolt the specimen into the multi-
In total, 18 tubular composite specimens were im- axial testing machine. In effect, the small aluminum end
mersed in distilled water baths – half in water at room tabs glued to the specimen provide protection for the
temperature and the other half in water at 50 °C. The specimen from the larger steel grips. In this manner the
specimens were immersed for a period of approximately stress concentrations that would otherwise be induced in
four months. The baths were topped up with water the specimen are minimized during the mounting of the
regularly to compensate for the evaporation of water. grips. Fig. 1 shows a schematic drawing of the mounted
Since the specimens immersed at 50 °C appeared to have specimen. For a description of the test facility, see [10].
reached saturation sooner, they were tested first and the As the rubber bladder is pressurized, the oil sur-
specimens in water at room temperature were allowed to rounding it increases the pressure on the wall of the pipe.
soak longer. At the weepage point (functional failure), cracks are
formed throughout the wall of the tube and these cracks
propagate from the interior to the exterior of the specimen
and oil leaks from the tube surface. Functional failure was
4. Testing procedure defined as a drop in the internal volume of the specimen of
1% equivalent to 2 ml. A subsequent loss of oil followed as
After removal from the water bath, specimen prepa- the oil surrounding the rubber bladder was pushed out
ration began by removing the wax on the ends of the and the rubber bladder expanded until it pressed up
tubes. The presence of any grease or oils would hinder against the walls of the tube. The remainder of the test was
the bonding of the end tab adhesive, Scotch Weld DP- dependent upon the properties of the fibers. The rubber
460. The extremities of the pipes were lightly sanded to bladder continues to be pressurized until the fibers fail and
promote adhesion, rinsed with acetone and then dried. the rubber bladder bursts, setting off an intensifier volume
The aluminum end tabs were sand blasted and then loss alarm and stopping the test. Structural failure is de-
washed with acetone to remove all debris. The end tabs fined as the loss in load carrying capacity while functional
were heated prior to assembly with the tubular speci- failure is characterized by the specimen’s inability to carry
mens and then the adhesive was injected through port pressure – essentially to function as a pressure retaining
holes. Heating was required to remove any moisture structure.
from the surface of the end tabs and render the adhesive Although the specimens were immersed in water at 50
less viscous. At the very least, the specimens were cov- °C for four months, they were tested at room tempera-
ered with a wet cloth at all times during end tab bonding ture. The duration of the test was too short to affect the
to prevent them from drying out. A temporary moisture results. It is to be noted that most of the damage in-
chamber was created to prevent moisture loss from any curred while the specimens were immersed and this
exposed surface on the specimen. process is irreversible. Permanent water damage, dem-
The end tabs were allowed 24 h to cure; following onstrated by the yellowing of the matrix in the speci-
this, a rubber bladder was inserted into the specimen mens immersed in water at 50 °C was observed after
and filled with oil. Surrounding the rubber bladder, immersion, which was not reversed during testing at a
more oil was added to remove all air in the interior of lower temperature.
the specimen. The purpose of the rubber bladder was to The biaxial loading ratios tested were: pure axial
allow functional and structural failure loads of the [0H:1A], pure hoop [1H:0A], [2H:1A] and [3H:1A]. The
specimen to be measured. Once the aluminum end tabs testing machine was controlled by a software program
1866 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874

designed in-house that controlled the ramping, loading 0.35

rate, etc. Constant loading rates were maintained at 0.3

Moisture Content (%)

0.463 kPa/s for pure axial loading and 4.626 kPa/s for
0.25
the remaining loading ratios. Testing times took any-
where from approximately one (for 50 °C water im- 0.2
mersed axial tests) to three hours (for room temperature
0.15 Water 20C Experimental Data
water immersed [3H:1A] tests).
0.1 Water 20C Fickian Model
The in-house developed testing software collected the
Water 50C Experimental Data
raw data output from the multi-axial tests. The data 0.05
output consisted of time in seconds; axial force, internal Water 50C Fickian Model
0
pressure and four strain readings from two 90° strain 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0
gage rosettes with a gage length of 1/1600 , all in volts. 1/2 1/2
Time h

Fig. 2. Fickian diffusion model fitted to water absorption of ½603 T

5. Moisture absorption behavior glass fiber epoxy resin tubular specimens immersed in water at two
temperatures.

Moisture absorption of the tubular composites as a

function of time was measured according to mass gain, Mention should be made of a discoloration noted in
using the following formula: the matrix of the specimens immersed in water at 50 °C.


mðtÞ  mdry Visual comparison revealed that the matrix of the un-
MðtÞ ¼  100; ð1Þ tested specimen in 20 °C water retained its white opaque
mdry
coloring while the matrix of the untested specimen im-
where mdry is the initial mass of the specimen with the mersed in 50 °C water darkened and developed a yellow
ends coated in wax, mðtÞ is the mass of the specimen tinge. This change can be attributed to exposure to an
after immersion for time t in hours and MðtÞ is the elevated temperature that altered the chemical and
percent moisture content of the specimen as a function physical properties of the matrix through a variety of
of time. The average initial mass of the tubular speci- processes including hydrolysis which leads to lowering
mens with the ends coated in wax was 73.25 g. In total, of T g; matrix softening, discoloration and damage [4,5].
4 g of wax were used to coat the ends of the specimens, A Fickian model can be used to describe the moisture
or 2 grams of wax per cut end. Over the course of the absorption data, as shown in Fig. 2. If the initial tem-
experiment, the specimens in water at room temperature perature and moisture content in a specimen are uni-
gained an average of 0.17 g or 0.23% water mass. The form, and if the geometry allows for the assumption of a
tubes immersed in water at 50 °C gained 0.21 g or one-dimensional diffusion through the thickness (thin
0.30% water mass. In contrast coupon specimens gained section), then the percent moisture content (by weight),
0.68% water mass in ambient temperature. This no- MðtÞ is given by [14]:
table difference is due to the number of the diffusion
surfaces, two surfaces for the tubular specimens versus MðtÞ ¼ GðM1  Mi Þ þ Mi ; ð2Þ
six surfaces for the coupon specimens. This was dis- where
cussed earlier in Section 2. The moisture absorption of
8 X1
1
the tubular specimens as a function of the square root of G¼1 exp½ð2j þ 1Þ2 p2 ðDt=h2 Þ;
time is plotted in Fig. 2. p j¼0 ð2j þ 1Þ2
2

Moisture absorption is characterized by the migra- ð3Þ

tion of molecules down the concentration gradient, " #
 0:75
which occurs through diffusion. The glass fibers are Dt
thought to be susceptible to leaching in distilled water,  1  exp  7:3 2 ; ð4Þ
h
and fibers tensioned in highly acidic or basic solutions
have shown considerable amounts of SCC [4,11–13]. where Mi , M1 and MðtÞ are the percent moisture con-
The properties of the matrix can also change signifi- tents of the tubular specimens prior to immersion, at the
cantly during exposure to an aqueous environment. saturation state and at a given time t, respectively. In
Furthermore, the water can diffuse between the fibers Eqs. (3) and (4), G is a time dependent parameter de-
and the matrix, weakening or destroying the bond at the scribing the extent of diffusion, h is the specimen
fiber–matrix interface. As mentioned earlier, the ends thickness and the diffusion coefficient, D, can be defined
were coated in wax to prevent the moisture entering in as the constant of proportionality between the diffusion
this manner. However, once the water penetrates as far flux and the concentration gradient in Fick’s first law.
as the fiber/matrix interface it is possible it to diffuse The magnitude of the diffusion coefficient, D is pro-
along the interface area. portional to the rate of diffusion and is a function of
 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874 1867

temperature [1]. The magnitude of D can be found from 6. Multi-axial testing results
the absorption curve, Fig. 2, in which the quantity in the
second parenthesis on the RHS is the slope of the linear 6.1. Stress–strain behavior
part of the part of the absorption curve as given in Eq.
(5) The average stress–average strain of axially loaded
 2  2 tubular specimens immersed in 20 and 50 °C distilled
h M2  M1
D¼p pffiffiffiffi pffiffiffiffi : ð5Þ water for about four months, is depicted in Fig. 3. It is
4M1 t2  t1 seen that the immersion in 50 °C water causes a decrease
The diffusion coefficients used in the plot in Fig. 2 of the stress and strain at failure. It is also noted from
were found to be 0.427 and 0.379 m2 /h, for the speci- Table 1 that the initial (elastic) modulus in the fiber
mens immersed in water at room temperature and those dominated loading, pure hoop, is not affected by the
in water at 50 °C, respectively. The moisture absorption immersion in water at 20 and 50 °C, see the first row in
data of the tubular specimens in water at 50 °C fits the Table 1. In contrast, in the matrix dominated loading,
Fickian diffusion model with a saturation moisture i.e., pure axial, one notes decrease in the elastic modulus
content of 0.30% however, there appears to be a certain due to water absorption of the matrix, hence plastici-
amount of data scatter at the latter stage for the speci- zation of it. In the other two biaxial stress ratios, a
mens in water at room temperature in Fig. 2. The combined effect is noted.
moisture content of the specimens immersed in 20 °C Fig. 4 shows a plot of the average applied stress
water rose to 0.238% while the Fickian model pre- versus the measured strain when tested in a two hoop to
dicted a saturation moisture content of 0.22%. It is to be one axial stress ratio. This is the so-called pressure vessel
noted that the final change in mass due to water ab- loading, in a pressurized closed end tube with no de-
sorption was only 0.18 g, which is a small value, there- formation constraint. It is seen that there is very little
fore difficult to quantify with a high degree of accuracy. variation in the stress–strain behavior of the glass fiber
For example, for the last four data points in a time epoxy resin tubular specimens in the two temperatures
period of two weeks, the mean moisture content was until functional failure in the axial direction. In the hoop
0.236% with a standard deviation of 0.00386%. This led direction the trends remained similar up to a stress of
us to conclude that within the measuring accuracy a 235 MPa and a strain of 0.0045. It is noted from Table
steady saturation level was either achieved or was nearly 1, that within the experimental accuracy, the elastic
achieved. moduli are approximately the same for the specimens

70.0
300.0
60.0 Specimens in Water at 20C-Hoop
Direction
Axial Stress (MPa)

250.0 Specimens in Water at 20C-Axial

50.0 Direction
Hoop Specimens in Water at 50C-Hoop
Stress (MPa)

40.0 200.0 Direction

Specimens in Water at 50C-Axial
[ 0H : 1A] 150.0
30.0

20.0 Specimens in Water at 20C 100.0 Axial

10.0 50.0 [ 2H :1A]

Specimens in Water at 50C
0.0 0.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.2 0.4 0.6 0.8 1 1.2
Strain (%)
Axial Strain (%)
Fig. 4. Averaged stress vs. average strain behavior of ½603 T glass
Fig. 3. Averaged stress vs. average strain behavior of ½603 T glass
fiber epoxy resin tubular specimens loaded at [2 hoop:1 axial] im-
fiber epoxy resin tubular specimens in pure axial loading immersed in
mersed in distilled water at two temperatures.
distilled water at two temperatures.

Table 1
Elastic stiffness, E (GPa) of ½603 T glass fiber epoxy resin tubular specimens in three environmental conditions
Measured direction Axial Hoop
Environment Dry Water (20 °C) Water (50 °C) Dry Water (20 °C) Water (50 °C)
Pure hoop – – – 36.6 36.9 36.8
[2 Hoop:1 axial] 45.6 39.4 41.8 66.9 52.8 50.5
[3 Hoop:1 axial] 91.2 195.1 126.0 66.0 48.5 49.0
Pure axial 23.8 18.2 17.9 – – –
1868 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874

800.0 envelope in stress space is obtained by plotting the stress

700.0 values in Tables 2 and 3. Fig. 6 is a plot of the stress
Hoop Stress (MPa)

600.0 functional failure envelope. It is seen that no deterio-

500.0 ration of composite tubular performance is observed
400.0
[ 1H : 0A ]
under [2 hoop:1 axial] biaxial loading and significant
300.0 improvement under [3H:1A] loading. In the case of pure
200.0 Specimens in Water at 20C axial loading, a significant change is noted between the
100.0 Specimens in Water at 50C
dry and distilled water immersed specimens. Of more
0.0 significance are the strain values at which function fail-
0 0.5 1 1.5 2 2.5 3 3.5 ure occurred as listed in the last row of Table 3. It is seen
Hoop Strain (%) that water immersion in 20 °C significantly increases the
strain at the functional failure (0.6 vs. 0.4). The wide
Fig. 5. Averaged stress vs. average strain behavior of ½603 T glass
fiber epoxy resin tubular specimens loaded in pure hoop immersed in
variation of functional failure at the pure hoop loading
distilled water at two temperatures. cannot be attributed entirely to the deterioration of the
fibers. Due to constraints at the tube ends, the specimens
deform in a barrel shape; hence bending stresses are
immersed in 20 and 50 °C under 2 hoop:1 axial loading induced in the extremities of the tubes, and failures are
ratio. A trend similar to the axial loading case was ob- of a burst type (see Section 10).
served for the pure hoop loading [1 hoop:0 axial]. There
was very little variation in the stress–strain behavior in
20 and 50 °C immersed specimens as seen in Fig. 5.
Table 1, row 4, also indicated that there was no signif- 150
icant difference in the elastic modulus. The loading ratio, [2H:1A] [3H:1A]
Dry specimens
3 hoop:1 axial is very close to no axial deformation case,
Axial Stress [MPa]

Specimens immersed at 20oC

[0H:1A]
hence the high elastic modulus in the axial direction, 100 Specimens immersed at 50oC

Table 1, row 3.

50
7. Functional failure envelope [1H:0A]

It was mentioned earlier that in pressurized compo- 0

nents, a functional failure refers to the loss of pressure 0 200 400 600 800
Hoop Stress [MPa]
retention. Tables 2 and 3 list the hoop stress/strain, and
axial stress/strain values corresponding to the functional Fig. 6. Functional failure envelope of ½603 T glass fiber epoxy resin
failure under various stress ratios. A functional failure tubular specimens in three environmental conditions.

Table 2
Average hoop stress and strain at functional failure of ½603 T glass fiber epoxy resin tubular specimens in three environmental conditions
Environment Hoop stress (MPa) Hoop strain (%)
Dry Water (20 °C) Water (50 °C) Dry Water (20 °C) Water (50 °C)
Pure hoop 833.0 706.7 558.2 3.6 3.1 2.3
[2 Hoop:1 axial] 231.9 236.3 232.7 0.4 0.5 1.0
[3 Hoop:1 axial] 325.7 388.0 383.7 0.8 1.0 0.5
Pure axial 7.6 3.5 3.5 )0.1 )0.2 )0.1

Table 3
Average axial stress and strain values at functional failure ½603 T glass fiber epoxy resin tubular specimens in three environmental conditions
Environment Axial stress (MPa) Axial strain (%)
Dry Water (20 °C) Water (50 °C) Dry Water (20 °C) Water (50 °C)
Pure hoop 0.0 0.0 0.0 )4.5 )4.0 )2.8
[2 Hoop:1 axial] 111.3 114.3 112.6 0.4 0.4 0.2
[3 Hoop:1 axial] 107.8 129.5 128.1 0.1 0.1 0.2
Pure axial 70.6 60.9 49.5 0.4 0.6 0.3
 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874 1869

Dry specimens
dry specimens. The greatest spread in strains was noted
[2H:1A]
300 Specimens immersed at 20oC in pure hoop loading where the dry specimens registered
Axial Stress [MPa]

Specimens immersed at 50oC the largest strains both transverse and parallel to the
[3H:1A] fibers followed by the distilled water immersed room
200
temperature group and finally the specimens immersed
[0H:1A]
in 50 °C distilled water. However, as noted earlier the
100 failure in the pure hoop loading was of a burst type and
[1H:0A] influence by the end constraints.
If the pure hoop case is excluded, then one observes
0
that independent of the loading case and environment, a
0 200 400 600 800 1000
Hoop Stress [MPa]
transverse strain value of about 0.4% is a good repre-
sentation of the functional failure criterion. This is a
Fig. 7. Structural failure envelope of ½603 T glass fiber epoxy resin significant finding since in cross-ply coupon specimens
tubular specimens in three environmental conditions. immersed in distilled water at 20 and 90 °C and dry
environments, the transverse cracking initiated at 0.5%
8. Final failure envelope strain, see Ellyin and Rohrbacher [1].

The final failure envelope of the tubular specimens is

shown in Fig. 7. For the [2 hoop:1 axial] loading ratio, 10. Final failure modes
the specimens immersed in water at 50 °C failed at the
lowest stress while the dry tubes and those in water at In general it was observed that failure of the specimen
room temperature failed at hoop and axial stresses of immersed in water at 50 °C was predominantly charac-
450 and 225 MPa, respectively. In the [3 hoop:1 axial] terized by brittle fiber fracture rather than fiber pull-out.
case the tubes in water at room temperature had the The reversed was consistently true for the specimens
greatest ultimate strengths followed by the dry ones and immersed in 20 °C water.
those immersed in water at 50 °C. This figure essentially For a lay-up of ½603 T , failure in the hoop direction
confirms the observation in Fig. 5, which indicated that is primarily dominated by the fiber properties and a
the immersion in 50 °C distilled water affected the ulti- wide spread in the strains parallel to the fibers was ob-
mate strength in the fiber dominated loadings. However, served, Fig. 8. A factor possibly responsible for the
it is to be noted that the values of failure stresses in 50
°C water immersed tubular specimens, under the two
1
biaxial loading ratios, have been influenced by the end
Strain Transverse to Fibres [%]

0.5
tab constraints (see note in Table 4).
0
[0H:1A] [2H:1A]
-0.5 [1H:0A]
[3H:1A]
9. Strain envelope at functional failure -1
Dry specimens
-1.5
Specimens immersed at 20oC
Fig. 8 illustrates the relationship between the strains -2 Specimens immersed at 50oC
parallel and transverse to the fibers at functional failure -2.5
for various loading ratios. In pure axial and [2 hoop:1 -3
axial] loading all environment groups failed within a -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
narrow interval. For the [3 hoop:1 axial] loading sce- Strain Parallel to Fibres [%]

nario, strains transverse to the fibers were similar while Fig. 8. Strains parallel and transverse to fiber direction at functional
the specimens immersed in water at 20 and 50 °C ex- failure for ½603 T glass fiber epoxy resin tubular specimens in three
perienced greater strain parallel to the fibers than the environmental conditions.

Table 4
Final failure stresses of ½603 T glass fiber epoxy resin tubular specimens in three environmental conditions
Environment Hoop stress (MPa) Axial stress (MPa)
Dry Water (20 °C) Water (50 °C) Dry Water (20 °C) Water (50 °C)
Pure hoop 836.9 706.7 558.2 )1.3 )1.2 )1.1
[2 Hoop:1 axial] 480.5 490.0 308.1a 232.6 237.2 149.1a
[3 Hoop:1 axial] 620.6 799.9 459.3a 207.4 267.4 153.4a
Pure axial 7.6 3.5 3.5 71.2 49.5 49.5
a
Failed near end tab.
1870 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874

broad spectrum of strains parallel to the fibers is the

varying degree of fiber damage as a result of the end tab
constraint. A violent burst failure typifies failure of tu-
bular specimens loaded purely in the hoop direction. At
both water temperatures, brittle fiber fracture was the
main failure mechanism, however the specimens im-
mersed in water at 20 °C also showed considerable fiber
pull-out.
For a loading ratio of [2 hoop:1 axial], typical failure
is characterized by a localized burst near an end tab as
seen in Fig. 9. This failure mode showed delamination
near the end tabs where the fibers attempted to realign
themselves along the principal axis of loading. The
matrix was also considerably cracked throughout the
length of the pipe with broken fibers near the rupture
site. The nature of this loading ratio is such that the end
tabs create stress concentrations in their vicinity. The
strain parallel to the fibers and in the transverse direc-
tion was very close for all environmental conditions, Fig. 10. [3 hoop:1 axial] Failure mode of ½603 T glass fiber epoxy resin
both dry and wet. The stress–strain curves for the tubular specimens immersed in distilled water at 20 °C.
specimens immersed in water at 20 and 50 °C, Fig. 4,
were also very close in terms of elastic modulus and
fracture strengths, Tables 1–4.
In the [3 hoop:1 axial] loading ratio, the final failure
of specimens immersed in water at 20 °C occurred in the
fiber direction along the length of the specimen in a
helical manner as if uncoiling the tube, see Fig. 10. The
failure of the fibers themselves was mostly brittle with
some fiber pull-out. In contrast, the tubes immersed in
water at 50 °C failed by a localized rupture near the end
tab, see Fig. 11, similar to the failure mode of specimens
loaded in a [2 hoop:1 axial] ratio, cf. Figs. 9 and 11. A
localized burst near the end tab is indicative of stress

Fig. 11. [3 hoop:1 axial] Failure mode of ½603 T glass fiber epoxy resin
tubular specimens immersed in distilled water at 50 °C.

concentrations introduced by the end tabs. Unlike other

loading ratios, more fiber pull-out (brush-like failure)
was observed for the tubes in water at 50 °C, which is
different from the typical brittle failure observed at such
a temperature, cf. Figs. 9 and 11. Also, the failure of
specimens immersed in the [3 hoop:1 axial] loading ratio
experienced greater strain than their room temperature
counterparts. This could be, in part, due to the loss of
stiffness coupled with an increase in ductility in the
matrix, allowing for the possibility of greater strains.
Pure axial loading yielded helical uncoiling of the
specimen in the fibers direction along the length of the
Fig. 9. [2 hoop:1 axial] Loading failure mode of ½603 T glass fiber pipe for both water temperatures as shown in Fig. 12.
epoxy resin tubular specimens immersed in distilled water at two Brush-like failure was also observed for both groups. In
temperatures, left 20 °C, right 50 °C. pure axial loading, once the matrix has failed and soft-
 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874 1871

Fig. 12. Axial loading failure mode of ½603 T glass fiber epoxy resin
tubular specimens immersed in distilled water at 20 °C (right) and
50 °C (left, center).

ened sufficiently, the fibers attempt to realign themselves Fig. 13. Representative scanning electron micrograph of [3 hoop:1
along the principal axis of the loading. This could ex- axial] failure mode of dry ½603 T glass fiber epoxy resin tubular
plain the brush-like fracture surface observed at final specimen showing resin adhesion to the fibers (Mag: 750 Acc Volt:
5 kV).
failure as the fibers are no longer held in place and
aligned by the matrix. From the stress–strain curve
of axially loaded specimens, Fig. 3, the decrease in
axial strain with an increase in temperature is clearly
noticeable.

11. Microscopic observations

The matrix and the fiber–matrix interface show ap-

preciable damage, while the fibers appear unchanged in
the scanning electron micrographs (SEM), Figs. 13–15,
taken of the fracture surfaces of the tubular specimens
subjected to monotonic loading. In all three environ-
ments, fiber fracture was brittle, as was matrix failure.
E-glass fibers are known for their resistance to
leaching [15] due to various surface treatments, however
the results of the specimens immersed in water at 50 °C
temperature indicated a noticeable decrease in final
failure strength with increased temperature. The loca-
tion of the failures near the end tabs make a direct
comparison rather difficult in absolute values. From the
SEM micrographs, it was noted that the fiber fracture Fig. 14. Representative scanning electron micrograph of [3 hoop:1
surfaces were increasingly brittle with an increase in axial] failure mode of ½603 T glass fiber epoxy resin tubular specimen
temperature. Fiber fracture was generally along flat in water at 20 °C showing poor resin adhesion to the fibers (Mag:
plane surfaces characteristic of brittle fracture. The 850, Acc Volt: 5 kV).
failure initiation site is often visible, as are the paths of
the cracks. The traces of crack propagation can be de- specimens [1], the osmotic pressure of the water created
scribed as a riverbed fiber fracture surface. pores and channels in the matrix. With prolonged ex-
In terms of the matrix, damage occurs through the posure to high temperatures, it is possible that the same
diffusion of water into the specimen. Uneven swelling of degradation of the matrix could be observed in the tu-
the matrix in different lamina and resin rich areas can bular specimens. From Figs. 16 and 17, showing the
also lead to microcracking [6]. In the cases of the coupon matrix near the fracture surface of the specimens, it
1872 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874

Fig. 15. Representative scanning electron micrograph of [3 hoop:1 Fig. 17. Representative scanning electron micrograph of [3 hoop:1
axial] failure mode of ½603 T glass fiber epoxy resin tubular specimen axial] failure mode of ½603 T glass fiber epoxy resin tubular specimen
in water at 50 °C showing little or no resin adhesion to the fibers (Mag: in water at 50 °C showing greater matrix cracking and damage near the
850 Acc Volt: 5 kV). fracture surface (Mag: 500 Acc Volt: 5 kV).

absorption where diffused water molecules attach

themselves to the polymer matrix by hydrogen bonding
[1].
Three specimens were not tested in order to examine
internal damage under the SEM. There was the issue of
how much damage would be introduced when fracturing
an untested specimen and how far would this damage
extend into the specimen. Are the cracks present from
the water or from the impact suffered upon fracture?
Therefore polishing was felt to be the least destructive
method on which observations could be based. At a
magnification of approximately 50, there is an increase
in the number of microvoids on the surface of the pol-
ished specimens. The voids could be present due to the
diffusion of water and ion exchange – damage that oc-
curred prior to polishing. Another possibility is that the
matrix of the specimens immersed in water have lost
some strength and are more prone to damage and the
formation of microvoids during polishing. Upon in-
spection of SEM figures, the specimens immersed in 50
Fig. 16. Representative scanning electron micrograph of [3 hoop:1 °C water show considerably more damage to the fibers
axial] failure mode of dry ½603 T glass fiber epoxy resin tubular in the polishing direction, which is not present other-
specimen showing some matrix cracking near the fracture surface wise. Possibly the samples were polished differently, but
(Mag: 650 Acc Volt: 5 kV). a more likely explanation is that the glass fibers of the
specimens immersed in water at 50 °C were far more
could be noted that pitting, in addition to surface susceptible to damage in the form of cracking, scratches
cracking, occurs at higher temperatures, whereas and seemingly some fiber–matrix debonding. In glass-
cracking only is observed in the specimens immersed in epoxy coupon specimens immersed in high temperature
water at room temperature. The notable yellowing in the water [1] cracks propagated from the specimen surface
matrix is indicative of matrix degradation. The optical to the microvoids on the interior of the specimen. The
difference in the matrix coloring is due to the water cracks and microvoids are a likely sign of water damage
 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874 1873

and degradation of the matrix. Therefore, one could and the fibers could have experienced some initial
assume that the increasing number of microvoids seen in damage. Conversely, if the sizing is hydrophobic, it
the polished micrograph specimens with increasing could repel the moisture and protect the fibers [17,18].
temperature is most likely due to water damage. The nature of the sizing could have a great impact on
debonding at the fiber–matrix interface and the amount
of water actually coming in contact with the fibers. This
12. Discussion could possibly explain the decrease in final failure
strengths in the hoop direction with increasing temper-
It is possible that water reaching the fiber–matrix ature. It is possible that a small degree of leaching has
interface softened or destroyed the adhesion between the occurred and would become observable under the SEM
two phases. The weakening of the bonding is demon- with prolonged exposure to such a harsh environment.
strated by greater resin separation from the fibers with When comparing the dry specimens with those in an
increasing temperature as shown in Figs. 13–15. Further aqueous environment at loading ratios of [3 hoop:1
support for the hypothesis of decreasing interfacial bond axial] and [2 hoop:1 axial], higher functional failure
strength as a function of increasing temperature is strengths are seen in the tubes immersed in water. This
demonstrated by the observation of the amount of resin increase could have been attributed to the compressive
coating remaining on the surface of the fibers after forces experienced by the matrix. It is a commonly ac-
fracture. In Fig. 13, which is representative of all fibers cepted theory that the E-glass fibers are resistant to
in the fracture surface, resin is clearly well adhered to the moisture absorption and therefore should remain di-
fiber and in large clusters. However, with immersion in mensionally stable in water. The fibers are tensioned
distilled water and increasing temperature in Figs. 14 during filament winding and fixed in place by the ther-
and 15, the resin coating and clusters on the surface of moset matrix, therefore their length, position and size
the fibers decrease in numbers, thickness and size. should remain largely unchanged throughout the water
Even though the physical appearance of the matrix absorption process in the composite. However, it is
and the fibers seemed unchanged in the presence of possible that due to the water ingress and softening of
moisture and temperature, most of the property changes the matrix some minor fiber movement may occur. The
are on a chemical level, only observable microscopically matrix itself is prone to swelling and water damage in
with exposure to a specific environment over an ex- addition to thermal expansion. As the matrix expands,
tended period of time. As reported in Jones [4] and the fibers are tensioned which could serve to relieve or
Schutte [7], hydrolysis of the bisphenol-A epoxy resin by increase the previously introduced stresses discussed
hot water is a form of chemical degradation of the above. During testing, the internal oil pressure acts to
matrix. When a resin is hydrolized, the ester bonds are compress the swollen matrix against the structural fabric
destroyed. With less bonding between and within the of glass fibers. Therefore the swollen lamina in the water
polymer chains, the chains will slide past each other with immersed tubular specimens are in greater compression
greater ease; thus, inelastic deformation is achieved at than their dry counterparts. Thus there is an increase in
smaller loads and a loss in stiffness results. the energy needed for cracks to propagate through the
The test specimens experience the greatest shear for- thickness of the tubulars (weepage).
ces in pure axial and pure hoop loading. The matrix From the observation of Figs. 10 and 11, the differ-
dominated properties are the most sensitive to hygro- ences in the fracture modes of [3 hoop:1 axial] loading
thermal conditions; thus, the failure of the material in between the specimens in water at 20 °C and at 50 °C is
shear is responsive to moisture absorption and is highly quite striking. The specimen immersed in water at 50 °C
affected [4]. This lends itself to support the conclusion fractured close to the end tab, like the dry specimens
that the fibers have experienced little degradation while the immersion in water at 20 °C resulted in frac-
themselves, but rather the decrease in strength at final ture within the gauge length. Thus in the case of the dry
failure in hoop and axial loading can be attributed to the specimens and those in water at 50 °C, failure was due to
high shear forces at the interfacial bonds. The interfacial the stress concentrations at the constrained ends, so the
bonds and the matrix are weaker and thus more sus- ultimate strength of the pipe was not achieved. For the
ceptible to shear failure. specimen in water at 20 °C, its ultimate strength was
The sizing used on the fibers is another consideration. realized, perhaps explaining part of its significant in-
Manufacturers do not reveal what the chemical make-up crease in strength.
and properties of the sizing are. Some sizings are hy-
drophobic and others are hydrophilic. Bazhenov [16] 13. Conclusions
found that unsized fibers exposed to boiling water lost
81% of their fracture toughness, while the fracture The mechanical properties of glass-fiber epoxy resin
toughness of sized fibers was unaffected. Therefore if the composites were generally negatively impacted by the
sizing is hydrophilic, it could be dissolved in the water environmental effects of moisture and high temperature.
1874 F. Ellyin, R. Maser / Composites Science and Technology 64 (2004) 1863–1874

The rate of moisture uptake was found to be propor- [2] Baizeau R, Davies P, Choqueuse D, LeBras J. Evaluation of
integrity of composite tubes. In: Cardon AH, Fukuda H,
tional to the temperature of the distilled water envi-
Reifsnider K, editors. Progress in durability analysis of composite
ronment. The final moisture content of the tubular systems. Rotterdam: Balkema; 1996. p. 225–32.
specimens based on weight gain was 0.23% and 0.29% [3] Perreux D, Suri C. A study of the coupling between the
for water temperatures of 20 and 50 °C, respectively. phenomena of water absorption and damage in glass/epoxy
Matrix swelling can lead to the relief and creation of composite pipes. Compos Sci Technol 1997;57:1403–13.
[4] Jones FR. Chapter 3 – Durability of reinforced plastics in liquid
residual stresses. During the mechanical loading, the
environments. In: Pritchard G, editor. Reinforced plastics dura-
moisture absorption has a positive effect on the advent bility. Abington Hall, Abington Cambridge, England: Woohead
of functional failure due to the compressive forces en- Publishing Limited; 1999.
gendered by the swollen matrix; hence leading to matrix [5] Brinson LC, Gates TS. Viscoelasticity and aging of polymer
crack closure. The matrix also undergoes plasticization, matrix composites. In: Kelly A, Zweben C, editors. Comprehen-
sive composite materials, vol. 2. Amsterdam: Elsevier Science
hydrolysis and a lowering of its glass transition tem-
Ltd.; 1999. Internet ed.
perature ðTg Þ; resulting in reduced stiffness. [6] Sims GD, Broughton WR. Glass fiber reinforced plastics-proper-
The cohesion at the fiber/matrix interface is reduced ties. In: Kelly A, Zweben C, editors. Comprehensive composite
by the presence of moisture, increasing in the incidence materials, vol. 2. Amsterdam: Elsevier Science Ltd.; 2000. Internet
of fiber pull-out at the fracture surfaces was observed. ed.
[7] Schutte CL. Environmental durability of glass-fiber composites.
Although no visible damage was incurred to the fibers,
Mater Sci Eng 1994;R13:265–324.
the microscopic fractographs point out that in a distilled [8] Kumosa L, Armentrout D, Kumosa M. An evaluation of the
water environment with high temperature leaching critical conditions for the initiation of stress corrosion cracking in
could occur. A further observation is that a considerably unidirectional E-glass/polymer composites. Compos Sci Technol
weakened matrix would support the fibers differently 2001;61:615–23.
[9] Mertiny P, Ellyin F. Influence of the filament winding tension on
during multi-axial testing; thus, affecting their ultimate
physical and mechanical properties of reinforced composites.
strength and distributing the load less effectively. It was Composites: Part A 2002;33:1615–22.
noted that fiber and matrix fracture was increasingly [10] Ellyin F, Wolodko J. Testing facility for multiaxial loading of
brittle with higher temperatures. In distilled water at tubular specimens. In: Kulluri S, Conacuse P, editors. Multiaxial
50 °C, strengths at functional and final failure were re- fatigue and deformation testing techniques, ASTM STP, vol.
1280. American Society of Testings and Materials; 1997. p.
duced; and to a lesser extent, the same was observed in
237–48.
the case of tubular specimens immersed in distilled water [11] Ely T, Kumosa M. The stress corrosion experiments on an E-
at 20 °C. glass/epoxy unidirectional composite. J Compos Mater
2000;34:841–78.
[12] Kajorncheappunngam S, Gupta RK, GangaRao HVS. Effect of
aging environment on degradation of glass-reinforced epoxy. J
Acknowledgements Compos Construct 2002;6:61–9.
[13] Mengel M, Kumosa L, Ely T, Armentrout D, Kumosa M.
Initiation of stress-corrosion cracking in unidirectional glass/
This work is part of a research program on the long polymer composite materials. Compos Sci Technol 2001;61:
term durability of polymer matrix composites. The 231–46.
program is supported, in part, by the Natural Sciences [14] Springer GS. Chapter 1 – Environmental effects. In: Springer GS,
and Engineering Research Council (NSERC) of Can- editor. Environmental effects on composite materials, vol. 3.
Lancaster, PA: Technomic Pub. Co.; 1981.
ada. The authors wish to acknowledge the technical
[15] Compos Mater Des. Available: http://callisto.my.mtu.edu/
assistance and contribution of Bernie Faulkner to this MY472/.
project. They would also like to extend thanks to Pierre [16] Bazhenov SL. Strong bending in the DCB interlaminar test of
Mertiny, Cecilia Grufman and Garret Meijer of the thin, E-glass woven-fabric-reinforced laminates. Composites
ACME Group for comments on the manuscript. R.M. 1991;22:275–80.
[17] Park S-J, Kim T-J. Studies on surface energetics of glass fabrics in
thank all the members of the ACME Group who were
an unsaturated polyester matrix system: effect of sizing treatment
extremely supportive and helpful during her stay. on glass fabrics. J Appl Polym Sci 2001;80:1439–45.
[18] Mader E, Grundke K, Jacobasch H-J, Wachinger G. Surface,
interphase and composite property relations in fiber-reinforced
References polymers. Composites 1994;25:739–44.
[19] Wolodko JD. Biaxial Fatigue and Leakage Characteristics of
[1] Ellyin F, Rohrbacher C. Effect of aqueous environment and Fiber Reinforced Composite Tubes, Ph.D. Thesis. Edmonton,
temperature on glass-fiber epoxy resin composites. J Reinf Plast AB, Canada: Department of Mechanical Engineering, University
Compos 2000;19:1405–27. of Alberta; 1999.