Composite Structures 31 (1997) 15.

5-164
0 1997 Published bv Elsevier Science Ltd
Printed in Great Britain. All rights reserved
0263-8223/97/$17.00

PII:SO263-8223(96)00008-l

A field-friendly rigid composite coupler for

GRP pipes
Gert Huysmans: * Jeanne-Fkaqoise Marsol,” Ignaas Verpoest: Guido De Roe&,’ Lucrhe
De Ridded & Jan Vansant”
“Department of Metalluqy and Materials Engineering, Katholieke Universiteit Leuven, De Croylaan 2, B-3001 Leuven, Belgium
hDepartment of Civil Engineering, Katholieke Universiteit Leuven, De Croylaan 2, B-3001 Leuven, Belgium
‘RAYCHEM N, x Ultratec Division, Diestsesteenweg 692, B-3010 Kessel-Lo, Belgium

GRP pipes are widely used in applications that impose special requirements
on the materials used, such as resistance against corrosive media and light
weight. Of special importance in the field of GRP pipes is the development
of a reliable and installation-friendly coupling method. A new adhesive type
joint is presented which eliminates the occurrence of certain installation
deficiencies common with existing coupler types. In this paper the different
stages in the design of a suitable coupler concept are presented. 0 1997
Published by ElseGer Science Ltd.

INTRODUCTION joint. Therefore, there is a need for a reliable

coupler system which is flexible, easy to install
During the last decade, GRP pipes have been and homogeneous in quality.
increasingly used for process piping and for In this paper the development of a new cou-
offshore applications. GRP offers several pler concept is presented which tries to
advantages over conventional materials such as eliminate the deficiencies common to the tradi-
specific stiffness and strength and corrosion tional hand lay-up methods. Following this
resistivity. Along with the expanding market of concept, the adhesive-type coupler is con-
GRP pipings, several techniques were structed using a prepreg tape which is
developed for joining GRP pipes. These joints subsequently wrapped around the external pipe
can be classified into purely mechanical and surface (Fig. 1). The prepreg tape is made from
adhesive joints. Currently, the still most widely a non-crimp glass fabric impregnated with a
applied method is the adhesive-type joint, con- toughened epoxy resin. The resin, developed by
structed using the established hand lay-up
technique. Flexibility and the ease of installa- L
Y u
tion are the big advantages of this method.
However, hand lay-up has some well-known
drawbacks such as the lack of reproducibility
and large installation times. In particular, the
installation times for large pipe diameters can
4
-ADi
..---.-
-.-----.-.-.-----.-.-.---.---.-.-.---.---
become quite extensive. The durability of the
joint is, to a large extent, ruled by the skills of
the installer. Concerning the reliability of GRP
joints, a Veritec [l] study on saltwater pipe sys-
tems pointed out that 70% of the pipe failures
occurred during commissioning. About 60% of
the reported failures concerned leakage of the
Fig. 1. Scheme of the adhesive joint with indications of
*Author to whom correspondence should be addressed. possible delamination sites (arrows).

155
156 G. Huysmans et al.

the Raychem company, was designed to meet longer shelf life (more than 2 years at room
the demands for increased shelf life. After the temperature).
wrapping procedure, the resin is cured using a The neat blend was cured following different
specially designed heating and pressurization conditions of temperature, isothermal curing
system. Human intervention is principally time and heat-up rate. Subsequently, mechani-
limited to the wrapping action of the prepreg cal tests and morphological inspections were
tape. In combination with a very reactive resin performed on the cured blend.
system, major reductions in installation times The mechanical tests comprised bending
can be obtained. Moreover, the use of prepregs tests, following the ASTM Standard D790-86,
eliminates direct contact with possible toxic and fracture toughness tests using compact ten-
components in the resin and ensures a uniform sion specimens following the February 1990
resin composition and good impregnation of the J-R curve test protocol on plastics (Welding
reinforcement. Institute, Cambridge, UK). The neat blend does
In the following the different steps in the not exhibit a measurable rigidity when it is
development of the concept are outlined. First, cured at temperatures lower than or equal to
a parametric study of the mechanical properties 120°C during a short time. From DSC tests, the
of the raw materials (neat and reinforced resin), time to initiate curing at temperatures below
developed within this research programme, 130°C was found to be very large. Therefore, it
allows to make a preliminary selection of a suit- can be concluded that these specimens were not
able composite coupler construction. As the even partially cured. Between 130 and 200°C
composite coupler will be wrapped over pipes, when the blend is cured for a sufficiently long
filament wound at +54”, a perfect knowledge of time, the bending modulus follows the
the interface quality in terms of adhesion behaviour indicated in Fig. 2. The heat-up rate
strength between the composite coupler and the did not show any significant influence on the
pipe has to be ensured. Together with a mechanical properties.
numerical investigation of the performance of The rigidity of the neat polymer blend is
the system as a whole, this leads to a final highest under the conditions of 130°C for
material reinforcement choice and a suitable 45 min, while the maximum toughness (best
sizing of the coupler as a function of the pipe crack resistance) is obtained for high curing
size. In addition, the installation process was temperatures independent of the curing time
simulated in detail to optimize the heating (Fig. 3).
device and curing parameters. This will, how- The differences in fracture toughness for high
ever, not be discussed here. Finally, the and low curing temperature are related to the
proposed concept was verified experimentally morphology change of the polymer. The frac-
by static burst pressure tests and a fatigue ture surface of the specimens broken after
programme. immersion in liquid nitrogen and specimens
etched with chemical treatment [2] were
observed by SEM. These observations revealed
THERMO-CHEMICAL AND MECHANICAL two distinct types of morphology, depending on
CHARACTERIZATION OF THE NEAT AND
REINFORCED RESIN
5.0
As a preliminary step in the design process, z
g 4.0
the parameters that affect the adhesion 5
strength between the coupler and the pipe were z 3.0
characterized.
8 2.0
Of prime importance are the physical and F 0 5 min.
chemical characteristics of the neat polymer F 1.0
m 45 min.
blend constituting the matrix component of 2
the prepreg composite coupler. This polymer 0.0
120 140 160 180 200 220
blend is a thermoplastic-thermoset blend, in
particular a PEI toughened epoxy. This specific Cure Temperature (“C)
blend has been developed by the Raychem com- Fig. 2. E modulus of the neat blend as a function of the
pany with the aim of getting a resin with a cure temperature and curing time.
 A rigid composite coupler for GRP pipes 157

600
g 500 ..__
5 400 T
g 300
200 , P-- 0 5 min.
100 +c+--i+- H 45 min.
??
Ol
120 140 160 180 200 220
Cure Temperature (“C)

Fig. 3. Fracture toughness G of the neat blend as a func-

tion of the cure temperature and curing time.

the curing temperature. At low curing tempera- Fig. 5. Morphology of the blend at higher cure temperatures.
tures the polymer blend is a three-phase system
composed of a continuous thermoplastic phase
and isolated thermoset phases (composed of big embrittled by the embedded, more scarcely dis-
and small nodules, respectively). At high curing persed nodules. As a result, the fracture
temperatures, a two co-continuous phase system toughness approaches the high toughness values
is formed: large hard nodules (thermoset phase) of the thermoplastic phase.
are embedded in a continuous soft phase The understanding of the influence of the
(thermoplastic phase). These two distinct mor- process parameters upon the final morpho-
phologies can be directly related to the logical structure of the blend and upon the
toughness behaviour of the neat polymer blend related mechanical properties has led to a pro-
[3]. Actually the first morphology (see Fig. 4) posal of an ‘optimal’ (suitable) curing cycle for
leads to a low fracture toughness as the two the installation of the coupler. Clearly, the
homogeneously dispersed thermoset phases highest toughness is obtained at the most ele-
tend to embrittle the thermoplastic phase. How- vated temperatures. However, the occurrence of
ever, in the second morphology (see Fig. 5) the local exotherms during curing, which can lead
crack path must follow the continuous thermo- to material degradation, prevents curing tem-
plastic phase, which acts on its own, and is not peratures higher than 175°C. For the same
reason, the heating rate had to be limited. A
best compromise was made between an accept-
able total installation time and amplitude of
local exotherms (thermal spiking). Increased
curing times did not show any improvement in
either toughness or stiffness properties. The
cure process simulation, using a thermo-chemi-
cal cure model based on DSC data, ensures that
for moderate heating rates and for a cure tem-
perature of 175°C the neat blend can be
considered as fully cured after a dwell time of
5 min. Therefore an isothermal curing time of
5 min was recommended.
After the complete characterization of the
matrix material, a suitable reinforcement archi-
tecture (fibre orientations, stacking sequence)
was designed for the composite coupler. For
economical reasons, E-glass fibres were the
Fig. 4. Morphology of the blend at lower cure temperatures. most obvious choice. For the selection of the
fibre architecture two aspects were taken into
158 G. Huysmans et al.

account. First, the use of non-crimp fabrics tion of these three different load cases was also
leads to a large design flexibility concerning the studied.
laminate lay-up. For the ease of manufacturing, Then GENLAM (GENeral purpose LAMi-
the different fibres within a single-layer group nate) [4], a through-the-thickness point-stress
are stitched together before prepregging. analysis computer program for composite lami-
Second, the non-crimp fabrics exhibit the best nates, was used to analyse the behaviour of this
behaviour, compared with woven fabrics, from a plate subjected to the several load cases. The
stiffness and strength point of view. Moreover, program computes the stiffness and the strength
during the wrapping action, problems of fibre of asymmetric hybrid laminates subjected to
wrinkling and buckling will arise and the non- complex in-plane loads and bending moments
crimp fabrics can overcome these problems. applied to the edges of the plate.
Woven fabrics already contain an initial out-of- The theoretical computations identified three
plane bending of the fibres due to the weaving types of multi-layered laminates that behave the
process. best under the different load cases of internal
Directly related to the reinforcement orienta- pressure, bending and torque to which a pipe
tion, the different mechanical loads that could can be subjected. These non-crimp fabrics are
possibly act on an in-service pipe had to be the [ &45”], [00/&45”], [oO/90°] laminate types,
identified. Pipes are subjected to an internal where the O*-direction refers to the pipe axis.
pressure as the main loading case combined Likewise, the quasi-isotropic type [45”/90’/
with bending and, to a lesser extent, with -45”/0”] has been added to the selection
torque, depending upon the application. As a because it is recognized as behaving the best
preliminary study, some simplified computations under a combination of differently oriented
were done in order to restrict the number of loads.
laminate types that were to be investigated The theoretical predictions of strength and
experimentally. In this approach the pipe and stiffness were verified experimentally. For this
the coupler were modelled as a flat plate sub- purpose the four different reinforcements were
jected to different load cases (Fig. 6). The impregnated with the polymer blend and cured
stacking sequence of the GRP pipes made out in an autoclave at 175°C for 5 min under a pres-
of glass-epoxy layers consists of a repetition of sure of 7 bar and vacuum conditions. As the
(&54”) oriented plies. The number of these real load cases would require complex biaxial
repetitions obviously depends on the pipe tests, the experimental verification was limited
thickness. to uniaxial tensile loading. Five tensile samples
The stacking sequence of the composite cou- were cut from each plate of reinforced blend.
pler had to be designed taking into account the The results of the tensile tests, according to
load cases detailed below. An internal pressure ASTM D3039-82, showed that the multi-layered
of 140 bar resulting in a biaxal loading was [0”/+45”] type behaves best from a tensile stiff-
taken as critical load case (N, = 0.37 MN/m and ness and strength point of view (Fig. 7). From
N, = 0.74 MN/m) for the design of 4-inch Fig. 7, it can also be noted that the GENLAM
(10 cm) diameter pipes. A maximal bending predictions coincide quite well with the experi-
load on the pipe was simulated by a constant mental results, both for strength and stiffness,
maximal tension (N, = 0.31 MN/m) while a making a theoretical approach valuable.
maximum torque was simulated by a constant
maximal shear (N, = 0.43 MN/m). A combina-
THE ADHESION BETWEEN THE PIPE
MATERIAL AND THE COUPLER
Internal Pressure Bending Torsion
MATERIAL

Using classical fracture toughness testing, the

adhesion between the pipe and the four pos-
sible couplers were subsequently investigated in
terms of interface fracture toughness. The inter-
face denotes here the bimaterial interface
Fig. 6. Plate model of the adhesive joint with indications between the pipe (glass-epoxy composite fila-
of possible load cases. ment wound at [ +54”]) and the first ply of the
 A rigid composite coupler for GRP pipes 159

25

Fig. 7. Comparison of experimental and predicted stiffness (left) and strength (right) for the different laminate types.

coupler. The four selected types of multi- mode II fracture testing were used. The ELS
layered reinforcements actually lead to three specimen is preferred to the end notched flex-
different interfaces; the ply of the coupler which ure (ENF) because the crack growth is more
is in contact with the pipe, can have three dif- stable and normally allows the determination of
ferent orientations: a complete crack resistance curve.
For both test types, the specimen has a rect-
(1) +45” or - 45”: applicable to the [ f 457, angular shape and is composed of two different
[45”/0”] and [45”/90”/- 45”/0”] types of materials (pipe material and coupler material).
coupler; The specimens are 160 mm long and 20 mm
(2) 0”: applicable to the [0”/90”], [0”/+45”] wide. A starter crack of 35 mm long is formed
and [45”/90”/-45”/0”] types of coupler; by an aluminium foil inserted at the bimaterial
(3) 90”: applicable to the [0”/90”] and interface. The foil is treated three times with
[45”/90”/ - 45”/0”] types of coupler. release agent (QZ-5111 spray from Ciba-
Geigy). In additional, in the ELS specimen, a
For each interface type the critical value of mode I precrack a few millimetres long is
the strain-energy release rate, corresponding to created to avoid an overestimation of the mode
a mode I (opening mode) and mode II (sliding II toughness [5,6]. The thickness of each adher-
mode) crack propagation, were determined. ent (pipe and coupler) was determined in order
The crack grows in the O”-direction correspond- to obtain approximately equal flexural rigidity in
ing to the pipe axis. the two beams (calculations performed using
The double cantilever beam (DCB; Fig. 8) the GENLAM program and FE modelling). In
specimen for mode I fracture testing and the order to avoid mixed-mode crack propagation
end loaded split (ELS; Fig. 9) specimen for under single-mode loading conditions, the two

Specimen Width B
Fig. 8. DCB test configuration. Fig. 9. ELS test configuration.
160 G. Huysmans et al.

materials need to have the same, or as close as Table 1. Mode I and mode II fracture toughness of the
bimaterial interfaces according to the different coupler
possible, flexural rigidity.
types
Using a classical tensile machine, the load is
Interface type
introduced via the end-tabs and the specimen is
(J%, (%)
loaded continuously in displacement control
Pipe/O” 1800 1000
with a cross-head speed of 2 mm/min. Then for Pipel90” 1800 1000
each crack length a, measured using a travelling Pipe/&” 2500 800
microscope, the load is recorded as well as the
crack opening displacement for the DCB speci-
men and the deflection for the ELS specimen. pipe/45” the crack rather jumped from the inter-
For bimaterial composites, a pure mode I face pipe/45” into the upper plies (45”/0”) of the
condition is never satisfied exactly during DCB coupler, meaning that the interface between the
testing as a result of mechanical incompatibility pipe and the coupler has a stronger fracture
between the two materials. Equal flexural toughness. Therefore, the value of 2500 J/m*
rigidity in the two beams causes the same represents a lower bound of this interface frac-
strains at both sides of the interface upon load- ture toughness.
ing. Because of the different modulus of the The mode II fracture toughness values were
beams, shear stresses will arise to restore equili- not easy to obtain. Problems such as formation
brium; this leads in turn to a small amount of of multi-cracks and unstable crack propagation
mode II loading. In addition, classical data were encountered. However, the obtained mode
reduction methods for both DCB and ELS test- II fracture toughness values have been evalu-
ing based on beam theory are not adapted for ated to similar values (800-1000 J/m2) for the
asymmetrical specimen geometries and bimater- three different interface types.
ial interfaces. Therefore, toughness values were
computed using FEM simulations of both tests.
Three-dimensional models were created for GEOMETRICAL DESIGN
both specimen types with incremental crack
lengths. The pipe material was modelled using A rough optimization of the laminate stacking
orthotropic brick elements while layered com- sequence of the coupler material for an internal
posite elements were used for the prepreg to pressure loading was performed using classical
account for the different prepreg lay-ups. The laminate theory. However, such an approach is
crack tip was modelled using three-dimensional not effective for the final selection of a suitable
quarter-point singularity elements. For the ELS laminate for the coupler as a whole due to the
specimens, contact elements were inserted complex stress state near the pipe and the cou-
between the opposite crack faces to prevent pler edges. Indeed, delamination and material
overlap. It was assumed that the contact was debonding starting from either the pipe or the
frictionless. coupler edge are the main failure mechanisms
For each crack increment, the experimentally for GRP joints.
determined load was applied to the model. The Therefore, a more profound fracture mechan-
models were validated by comparing computed ical approach was adopted in the design of the
and measured load-point displacements. The coupler to study the possibility of crack propa-
mode I and mode II strain-energy release rates, gation and debonding leading to ultimate
and hence fracture toughness values, were com- failure of the joint. The performance of the
puted based on the nodal forces and nodal coupler is measured in terms of the strain-
displacements at the crack tip following Irwin’s energy release rate (SERR) evaluated at
crack closure integral method [7]. The results (virtual or pre-existing) starter cracks located at
are shown in Table 1. the critical delamination sites.
The interfaces pipe/O” and pipe/90” reach the An extensive parametric study was carried
same level of the mode I fracture toughness, out allowing the optimal dimensioning and
while a higher mode I level can be noticed for shaping of the coupler. In addition, a selection
the interface pipe/45”. However, it must be of the proper laminate lay-up can be made.
noticed here that for the interfaces pipe/O” and All computations were performed on a three-
pipe/90” the crack propagated in a stable way dimensional finite-element model of the
along the desired interface. For the interface coupler. For composite pipes, an axisymmetric
 A rigid composite coupler for GRP pipes 161

loading can cause rotational displacements as a computed for the three opening modes using
result of the coupling effect between in plane Irwin’s crack closure integral method for
loading and out-of-plane deformations charac- quarter-point singularity elements [7].
teristic of asymmetric lay-ups. Therefore, a full Computations were performed on different
three-dimensional analysis is required. pipe sizes from the same pressure class ranging
The pipe was internally loaded with a con- from 3- to 24-inch pipes. Four laminate types
stant pressure. End cap effects were were investigated which were retained, as has
incorporated by axial tensile stresses which are been discussed earlier, because of their favour-
related to the internal pressure, pipe diameter able strength and toughness properties. These
and structural wall thickness. are the [O”/900], [ +45”], [0“/+45”] and the
Two coupler types were investigated. The [90’/+45”] laminates. The coupler is built up by
simplest one is the butt-to-butt type coupler. stacking two layers into a tape using skew sym-
Here an artificial crack is formed by the pipe metry. The tape is then wrapped around the
edges, with a length equal to the pipe wall pipe until the desired coupling thickness has
thickness. For large pipe sizes this configuration been reached.
is not recommended as the large precrack can All pipes investigated belong to the same
initiate premature failure of the joint. There- pressure class, which means that they have the
fore, in the second model, the pipes are tapered same diameter to wall thickness ratio (approxi-
and the space in-between is filled with a gap- mately 35). The pressure used in the FEM
filler material (neat or reinforced blend) (see calculations is the nominal pressure augmented
also Fig. 1). This way, the crack length is with a safety factor 2.4.
reduced and the load-transfer area increased. The coupling thickness was chosen to
Failure can now occur through delamination approximate the pipe wall thickness unless
along the pipe-gap filler interface or by crack otherwise stated. As will be shown, this thick-
propagation into the gap filler. ness is close to the optimal coupler thickness.
Another possible delamination area is located The bond length of the coupler is expressed in
at the coupler edge. The stress concentration at terms of the pipe diameter. This way, size
this location can lead to material debonding. effects can be computed based upon the pipe
Tapering the coupler edge provides a smoother diameter as one single parameter. Coupler sys-
transition for the load transfer between pipe tems with proportional dimensions in each
and coupler material, and hence reduces the direction will be referred to as proportionally
risk of delamination. scaled couplers.
Which one of the two delamination sites will Within the same pressure class the computed
be the most critical depends on pipe size and strain-energy release rates for proportionally
coupler thickness. Failure occurs by a crack scaled couplers exhibit a similar scale effect.
starting at the most critical edge and growing This is shown for the [900/+45”] lay-up in Figs
towards the other side until ultimate failure 10 and 11 for mode II where the SERR is
(Fig. 1). To study this possibility of crack propa- divided by the pipe diameter and plotted as a
gation, artificial starter cracks of 2 mm length function of the ratio total bond length/pipe
were introduced at both edges.
At the coupler edge, the crack is located at
the interface between the pipe and the coupler.
For the pipe edge, a starter crack is situated at
the pipe-coupler interface for the butt-to-butt
joint and at the pipe-gap filler interface for the
tapered pipes.
While the loading at the pipe edge crack is
merely mixed mode, mode II is dominant at the
coupler edge under influence of local axial
bending stresses and shear stress. The bending
effect increases with coupler thickness because 0 0.2 0.4 0.6 0.8 1 1.2
the coupler acts as a local stiffener for the pipe.
The strain-energy release rates, which are a Fig. 10.Normalized GII for a 2 mm pipe edge crack;
measure of the possibility of crack growth, were [90”/45”/-457 coupler without gap filler and taperings.
 162 G. Huysmans et al.

6-

5 --
F :
J 4-t
E
cl 3--

22
0 0.2 0.4 0.6 0.8 1 1.2
L/D1lmmhrml

Fig. 11. Normalized G,, for a 2 mm coupler edge crack;

[90“/45”/ - 4.57 coupler without gap filler and taperings.

diameter. The solid curve is obtained for a Fig. 14. Test set-up for internal pressure and torsion
experiments on GRP pipes.
hypothetical pipe with an average diameter to
thickness ratio of 35. The scatter around this
curve is caused by a slightly varying real diam- eter to thickness ratio of the different pipe sizes
and the slight variation in thickness ratio
between pipe and coupler. Similar curves are
2mnCo@erEdgeCradc(fvlcdeff) obtained for mode I. At the pipe edge a mixed

111.20
!

.__&___

lF.--t---
0 0.2 0.4 0.6 0.8 1 1.2

LDl (mruhnrJ

Fig.12. Normalized G,, master curves for a 2 mm coupler

edge crack, as a function of the normalized adhesion
length LIDi.

Fig. 15. Detail of the gripping device for GRP pipes.

2mnPipeEd&Jecra&&lodeff)

7ooi

70

60

s 50

; 40

$
7 30

t 20
oJ’*” ) ” ” ~ “‘I 1
10 / I I I I I
0.0 5.0 10.0 15.0
Laminate 1 hkknrr (mm) 0
0 12 3 4 5 6 7
Fig. 13. Grr at the pipe edge as a function of the coupler Log (Number of Cycles)
thickness; 1Zinch pipe (structural wall thickness 10 mm);
no gap filler and taperings. Fig. 16. P-log (N) plot of the investigated couplers.
 A rigid composite coupler for GRP pipes 163

mode exists, while at the coupler edge mode II PERFORMANCE EVALUATION OF THE
is dominant. As the singular stress fields at both FINAL PRODUCT
edges move away from each other, as the bond
length increases, the SERR drops and reaches Several prototypes were produced for the vali-
an asymptotic value at a bond length of 0.7 dation of the design. Three-inch diameter pipes
times the pipe diameter. In this way, a mini- were joined by the new coupler concept and the
mum required bond length can be derived for complete systems were tested under static pres-
any pipe within the same pressure class. The sure loading. Following the procedure described
same trends are observed for other lay-ups and in ASTM D1599-88, a value of 70 bar was
varying coupler thicknesses. However, clearly obtained as an average value of the burst pres-
the SERR becomes extremely large for very sure of such a system. As the working pressure
large pipes. Although there will be other pheno- of the 3-inch diameter pipe considered is
mena, such as local plastic deformations, which 32 bar, the experimental burst pressure presents
make the linear elastic fracture mechanics a safety factor of 2.2.
applied here invalid, it is obvious that thicker Next, in order to evaluate the life-time per-
laminates are more sensitive to delamination. formance, a fatigue programme was set up and
Using a gap filler at the pipe edges and tapering systems were cycled under an internal pressure
the coupler at the coupler edge becomes almost as described in ASTM D2992-87 procedure A.
necessary for reducing the SERR for large A special gripping device for GRP pipes was
pipes. This will not, however, be discussed here. designed and developed in the framework of
Figure 12 shows the mode II normalized the project in order for the fatigue of pipes or
SERR master curves at the coupler edge as a piping systems under an internal pressure, pos-
function of the normalized bond length for dif- sibly combined with a torque (Fig. 14), to be
ferent stacking sequences. At the pipe edge followed. As described in Fig. 15, the specimen
similar curves were obtained for G, and G,,. has specially designed conical end-tabs in order
Here, the [W/f 45”] shows the best behaviour, to prevent failure occurring in the grips.
followed by the cross-ply [0”/90”]. At the cou-
As prescribed in the ASTM standard men-
pler edge the cross-ply is preferred, followed by
tioned above, the frequency of the test was
the [90’/_+45”]. From these results clearly the
chosen to be 25 cycles/min and, for technical
design of the coupler based upon simple lami-
reasons, a sinusoidal wave form was adopted
nate theory can give contradictory results
with R = 0.1. Three samples were tested at,
because end effects are ignored. These end
respectively, 80, 60 and 40% of the burst pres-
effects are, however, most important as they are
sure value and the obtained results, noted in
responsible for the failure of the joint.
Fig. 16, seem promising. The specimens tested
Figure 13 shows the mode II SERR at the
at the two highest levels of pressure (80 and
pipe edge for a 1Zinch pipe (structural wall
60% of the burst pressure) broke within a
thickness 10 mm) for different coupler thicknes-
ses (i.e. an increasing number of wraps). reasonable time span. The specimen tested at
Although there is no well-defined optimum, 40% of the burst pressure did not break but it
increasing the coupler thickness above the pipe was decided to stop the test. Actually, owing to
thickness does not add significantly to the the low frequency used, the test is very time
strength of the joint. The only improvement consuming (lo6 cycles require about 1 month to
comes from the pipe edge. A larger coupler be reached).
thickness decreases the stiffness and cross-sec- Of course, these results still have to be con-
tion transitions at the pipe edge, and hence the firmed using a larger number of test samples.
stress singularity decreases. At the coupler edge
the SERR is much less dependent upon the
coupler thickness. The load always has to be CONCLUSIONS
transferred from the stiffer coupler to the
smaller pipe cross-section, therefore the stress The design of a new concept of composite cou-
singularity at the coupler edge remains. The pler has been successfully achieved thanks to a
same conclusions were found for the other pipe complete thermochemical, mechanical and
sizes. geometrical study.
164 G. Huysmans et al.

An optimal curing cycle has been proposed REFERENCES

and precise geometrical requirements were
1. Proceedings of the Conference on GRP-pipes, Bond voor
recommended. Nieuwe Materialen, Arnhem (the Netherlands), 1994.
A good correlation between the experimental 2. Blundell, D. J., Spherulitic morphology of the matrix of
results and the different modelling involved in thermoplastic PEEK/carbon fibre aromatic polymer
composites. J. Matel: Sci. 1989, 24, 2057-2064.
this rather extensive study was obtained, result- 3. Ochi, M., Marsol, J. F. and Verpoest, I., Relation
ing in a reliable design of this new concept of between the morphology and the toughness in a PEI
composite coupler. modified system as a function of the curing conditions.
The fatigue performance, although very Internal report, Department of MTM, Katholieke Uni-
versiteit Leuven, 1993
promising, must still be completed. 4. Tsai, S. W., Composites Design, 4th edn. Think Compo-
sites, 1988.
ACKNOWLEDGEMENTS 5. Albertsen, H., Ivens, J., Peters, J., Wevers, M. and
Verpoest, I., Interlaminar fracture toughness of CFRP
influenced by fibre surface treatment, Part 1: Experi-
The authors gratefully acknowledge the Flemish
mental results. Comp. Sci. Technol. 1995, 54, 133-145.
Government for their support of this research 6. Durmel, D. J.-P., Szpicak, J.-A., Singh, S. and Part-
in the IWT-project ‘A field-friendly rigid com- ridge, I. K., Crack initiators and pre-cracking
posite coupler for GRP pipes’. Further we techniques for fracture testing of polymer matrix com-
posites. In Proc. 3rd Int. Conf on Deformation and
would like to thank the Belgian Government for Fracture of Composites, Guildford, UK, 1995, pp.
the support to the Department of Metallurgy 190-194.
and Materials Engineering (K. U. Leuven) 7. De Roeck, G. and Abdel Wahab, M. M., Strain energy
release rate formulae for 3D finite elements. J. Engng
through the ‘Interuniversitaire Attractiepolen’
Fract. Mech. 1995, 50, (4), 569-580.
action.