Composites: Part A 36 (2005) 291–300

www.elsevier.com/locate/compositesa

Analysis of the mechanical properties of a reinforced

thermoplastic pipe (RTP)
M.P. Kruijer*, L.L. Warnet, R. Akkerman
Section of Applied Mechanics and Polymers, Department of Mechanical Engineering, University of Twente,
P.O. Box 217, 7500 AE, Enschede, The Netherlands

Abstract

This paper describes the analysis of a long length reinforced thermoplastic pipe. For this new class of pipe, which is constructed of a
polyethylene liner pipe over wrapped with two layers of non-impregnated twisted aramid cords, peculiar deformation behaviour was
observed when a pipe was pressurised. This behaviour was found to be a result of a difference in cord-slack between the two reinforcement
layers in conjunction with an unbalanced torsion moment generated by the two reinforcing layers. Cord slack is a certain surplus length of the
reinforcing cords relative to the pipe geometry. The cord-slack is estimated for the two different layers and first incorporated into an earlier
published model, based on a plane stress characterisation. As no substantial improvement has been achieved by this approach, a new model
based on a plane strain characterisation is introduced. This model shows good agreement with the experimentally determined strains for the
hydrostatic pressure load-case.
q 2004 Elsevier Ltd. All rights reserved.

Keywords: A. Thermoplastic resin; B. Mechanical properties; D. Mechanical testing; On-shore oil industry composite pipe

1. Introduction † An even number of helical windings of continuous

reinforcement to contain the applied inner pressure and
Composite pipes are increasingly used in the oil and gas other loads.
industry, instead of conventional carbon steel pipes. The † An outer cover layer to protect the fibre layers from
main reason is the better corrosion resistance of composite external damage.
pipes compared to steel pipes. The first composite alterna-
tives to steel pipelines consisted of composite pipes based on As mentioned in Ref. [2], the continuous reinforcement
thermoset matrices. The properties of thermoset composites material usually consists of carbon, aramid or glass fibres,
are well known and documented [1]. However, the embedded in a thermoplastic matrix of polyethylene (PE),
increasing demands of the oil and gas industry, involving polypropylene (PP), polyamide (PA) or polyvinylidene-
high temperature resistance and the need for damage fluoride (PVDF). The fibres can be either single filaments,
tolerance and flexibility, often exceed the capabilities of yarns or braided. As discussed in Ref. [3], RTPs can be
thermosets. As thermoplastics can satisfy these demands, manufactured using various types of helical winding
several types of continuous fibre reinforced thermoplastic processes. One of these is the helical tape wrapping method.
pipe (RTP) systems are currently under development. This process, which will be discussed in more detail in
The principal components of an RTP are: Section 2.2, has the potential to be continuous (as opposed
to thermoplastic filament winding), which enables the
† A thermoplastic liner, the major function being to production of substantially long pipes.
provide a leak free and corrosion resistant containment
The pipes considered here are produced by the helical
for the transported fluid.
tape wrapping method, using prepreg tapes in which
non-impregnated twisted aramid yarns are embedded in a
* Corresponding author. Fax: C31-228-355540. high-density PE matrix. Since the twisted aramid yarns are
E-mail address: [email protected] (M.P. Kruijer). not impregnated, there is no bonding between the fibres
1359-835X/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesa.2004.06.018
292 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300

Nomenclature
Af total cross-section surface of the 311, 322, 312 strains in local directions (i.e. cord and
reinforcement tape matrix directions)
Am cross-section surface of the matrix srr, sqq, szz, szq stresses in global directions
material s11, s22, s12 stresses in local directions (i.e. cord and
Ac cross-section surface of the cords matrix directions)
Vf fibre volume fraction A1, B1, C1 unknown constants in the plane stress
Vm matrix volume fraction model
Ef fibre Young’s modulus A, B, C, D unknown constants in the multi-layer
Em matrix Young’s modulus model
E1 stiffness in cord direction q1 fibre angle of inner tape layer
nm matrix Poisson’s ratio q2 fibre angle of outer tape layer
rc linear cord density q0 initial fibre angle
rf fibre density [Q] stiffness matrix for unidirectional ply
hf thickness fibre layer used in model [Q*] stiffness matrix for pipe wall or trans-
P internal pressure formed reduced stiffness matrix
[T] transformation matrix
3rr, 3qq, 3zz, gzq strains in global directions
[R] Reuter’s matrix

and the matrix. This results in a low flexural stiffness of the pressure. First, the materials used and the production
pipe, so that the pipe can be reeled. The liner pipe and cover process are outlined. As it is observed that the amount of
layer are of the same generic type of thermoplastic material, cord-slack will be different for both layers, a procedure to
heat bonded together providing a sturdy monolithic measure the amount of cord-slack for the two different
construction. There is about 200 km of this long length layers is described. After that, a model for the torsional
RTP, hereafter called long length reinforced thermoplastic behaviour is developed by incorporating the different slack
pipe (LLRTP), in operation now. values into a modified version of the existing plane stress
In order to design and engineer LLRTPs, it is very model [2,4]. Finally, an updated model for LLRTP based on
important to have a clear understanding of the mechanical a plain strain characterisation will be introduced. In the
behaviour of LLRTPs. For this purpose, a procedure to subsequent sections of this paper this model will be referred
predict the resulting stresses and strains as a result of a to as the multi-layer model.
particular loading case is needed. Mathematical models to
predict the stresses and strains for LLRTPs were already
reported in Refs. [2,4]. However, presented in a different 2. LLRTP manufacturing method
way, the models in Refs. [2,4] are essentially the same. Both
models assume a plane stress state, with no load take up in 2.1. Pipe materials
the direction perpendicular to the fibres. The measured and
calculated strains in axial and hoop direction were presented The thermoplastic resin, which is used for the reinforce-
as a function of the internal pressure. The apparent ment tapes, the liner and the cover, is a pipe grade HDPE, type
mismatch between the calculated and the measured hoop PE100 (Eltex TUB 121), which is hereafter called PE. The
strain at low pressures was found to be a result of the cord- main reasons for choosing this material are its long-standing
slack. Cord-slack is a certain surplus length of the cords record in the oil and gas industry. Some technical details on
relative to the pipe geometry. This phenomenon will be
Table 1
discussed in more detail later on. The hoop strain for higher
Physical and mechanical properties of PE100
pressures was well described by both models. The models
did not show an adequate agreement with the experimen- Property Symbol Standard Unit Value
3
tally determined axial strain. A discussion on the torsional Density rm ISO/R 1183 g/cm 0.9600–0.9609
behaviour was absent in both cases. However, peculiar (compound)
Yield stress sy ISO/DIS MPa O25
torsional behaviour was observed for the current LLRTP
(100 mm/min) 6259,2
when pressurised, as the rotation direction sudden changes Young’s- Em ISO 527-2 MPa 1205
sign at about 35 bar. This behaviour needs to be analysed modulus
and described. The current study can be seen as an extension Poisson’s ratio nm 0.4
of the models mentioned above. The focus is set on Elongation 3max ISO/DIS % O600
at break 6259,2
predicting the torsional behaviour as well as the tangential
(100 mm/min)
and axial strains of the LLRTP when loaded by hydrostatic
 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300 293

advantage of this process is that it is continuous. First,

the HDPE liner is extruded and cooled immediately. In the
second stage, the first tape layer is wrapped over the
extruded liner. The resulting winding angle is fixed to 548 by
the outer diameter of the liner and the tape width. As
Fig. 1. Outline of the aramid reinforced tape with the definition of the local indicated in Fig. 2, the winding tension involved is balanced
co-ordinate system used for the theoretical model. Af, Am and Ac represent
the cross-sections of the tape, the matrix material and the aramid cords,
by wrapping two tapes per pitch at the same time. The
respectively. procedure for wrapping the second layer will be exactly the
same, only the wrapping angle will be K548. It must be
PE100 are given in Table 1. The stress–strain relationship for noted that because of logistic reasons the same tape width is
PE is non-linear and depends on the strain and the strain-rate. used for both layers, resulting in a slight increase (about
In this study, the PE is modelled as a linear elastic material, 1.38) of the winding angle. The three components are heat
using a secant modulus of 400 MPa. This value is based on bonded together after wrapping the two tape layers. The
the loading speed, temperature and average strain during the structure is cooled rapidly after the heat bonding procedure
LLRTP pressurisation experiments. For the reinforced tapes and a cover layer is extruded over it. Finally the LLRTP is
the same PE is used. The cords are embedded in, but not cooled again, printed and reeled.
impregnated by, the PE. The Twaronw-aramid cords itself are
twisted in order to avoid filament buckling when the cords are 2.3. Cord-slack and the influences on the mechanical
subjected to a negative strain. For the model, a measured properties
stiffness of 80 GPa is used for the cords. The reinforcing
tapes, from which an outline is shown in Fig. 1, are produced As already described in Refs. [2,4], the cords will not
by an extrusion process. In Fig. 1, Af, Am and Ac are the total immediately contribute to the stiffness of the LLRTP. A
cross-section area, the cross-section area of the matrix and the certain strain, which is called the ‘take-up strain’, is needed
cross-section area of the cords, respectively. The fibre to pull taut the twisted aramid cords. This phenomenon of
direction and the in-plane direction perpendicular to the the cords having a certain surplus length to the pipe
fibres, denoted the ‘1’ and ‘2’ direction, respectively, are geometry is called cord-slack. The total amount of cord-
called the principal directions. The use of non-impregnated slack for the cords in the pipe is the result of two separate
cords in a tape is not conventional and probably viable only mechanisms. The first part is the construction strain of the
with aramid, because of its high abrasion resistance. twisted aramid cords. This amount of construction strain is
equal for both layers. The second part can be imputed to
2.2. Helical tape wrapping method shrinkage related geometrical changes during and after the
production process. Unlike to the first mechanism, the effect
Commercially available LLRTP is commonly produced of this shrinkage related part is not equal for both fibre
by the helical tape wrapping method. An important layers, since these are wrapped under a slightly different

Fig. 2. The wrapping process for fibre layer 1. (picture Krauss Maffei Kunsstofftechnik).
294 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300

the simple fact, the fibre layer will tend to unwind itself, the
jump in torsion angle is caused by the outer fibre layer.
Because of the slack difference, the inner fibre layer does not
directly resist the torsion moment generated by the outer
fibre layer. At G35 bar the slack in the inner fibre layer is
also consumed due to the torsion angle and the increasing
tangential and axial strain. This results in equal torsion
moments generated by the two fibre layers at this point.
When the pressure is further increased, the LLRTP shows
rotation in the opposite direction. This is a result of the
unbalanced torsion moments generated by the two fibre
layers, which will be treated in more detail later in this paper.

3. Experimental procedures

3.1. Pressure experiments

Fig. 3. Typical behaviour of a 4 00 pressurised LLRTP. The circled numbers
1, 2 and 3 represent the three different values for the tangential compliance
This study is concerned with one type of test, in
indicated when subjecting an RTP to internal pressure. The non-circled
numbers 1 and 2, which are used for the cord-slack estimations procedure, which the pipe is only subjected to an internal pressure.
are defined by the intersections of line 1 and 2 and line 2 and 3, respectively. The LLRTP samples were 2 m long and terminated at both
ends by specially designed end fittings [5]. In each case, the
angle. A ‘slack difference’ between the two layers is pressure loading rate was 15 bar/min (G5%). A dedicated
therefore induced. An example of the results of deformation pipe testing machine (Flexi-rig) was used for the pressure
measurements with a 4 00 (inner diameter) pressurised experiments, as shown in Fig. 4. The Flexi-rig can subject
LLRTP are shown Fig. 3. The influence of the cord-slack pipes to arbitrary combinations of internal pressure, torsion
is most obviously presented by the tangential strain. Within moment and axial forces. The resulting displacements in
the pressure range of 0–150 bar three different values of axial direction and the torsion angle can be measured
tangential compliance can be derived. They are indicated in simultaneously. A specially designed linear variable
Fig. 3 by the circled numbers 1, 2 and 3. From 0 to G30 bar differential transformer (LVDT) device was used [6] for
the two fibre layers are not contributing to any load take up. the hoop displacement and an extra measurement of the
Within this range the LLRTP behaves as a plain PE pipe. At axial displacement directly on the pipe. The LLRTP should
G30 bar, the slack in one of the fibre layers is consumed, be free to deform during the pressure experiment, therefore
and the fibre layer starts to take up load. This is confirmed by force control was used by continuously driving the resulting
the torsional behaviour of the pipe, which starts at the same axial forces and torsion moments to zero. The resulting axial
time. It can be demonstrated by laminate theory or by displacement w and the torsion angle q of the movable head

Fig. 4. LLRTP fixed in the Flexi-rig during a hydrostatic pressure experiment.

 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300 295

take-up strain needed in the cord direction to load the fibres

are given in Table 2.

4. Modelling

In this section, the existing plane stress model will first

be extended by incorporating the non-linear behaviour as
a result of the change in fibre angles and internal diameter
during pressurisation. After that a new model for LLRTP,
based on a plane strain characterisation will be introduced.

4.1. Updated ‘plane stress’ model

Fig. 5. Plane stress cross-section of an LLRTP with the definition of For this simple approach, the stresses and strain do not
the global (r, q, z) and the local (1,2) coordinate system. The cord angle a, depend on co-ordinates r, z and q and thus the stresses szz,
is defined by the angle between the (local) 1-direction and the (global)
sqq and szq (disregarding any radial stresses) within one
z-direction. 3qq represents the tangential strain, which is here assumed to be
homogeneous through the thickness. layer are uniform through the thickness. First the stress–
strain relationships will be presented, which follow from
were measured, together with the tangential and axial general laminate theory presented in Ref. [7]. After that,
displacements measured directly on the pipe by the LVDT- these relationships are applied in a modified form for the
device. solution of this particular problem. The stresses and strains
for the isotropic and the fibre layers are denoted by
3.2. Cord-slack estimation superscript ‘m’ and ‘f’, respectively throughout this section.
When plane stress is assumed the relation between the
In this procedure, the amount of the take-up strain is stresses and strains for an isotropic layer can be represented
estimated, which is needed to pull taut the cords in the fibre by
layers, for two models further to be specified. This is done 0 m1 2 3 0 m1
sqq Q11 Q12 0 3qq
by examining the strains in the global directions r, z and q B mC 6 7 B mC
(see Fig. 5) at critical points 1 and 2, which are defined by @ szz A Z 4 Q12 Q22 0 5,@ 3zz A; (2)
the intersections of line 1 and 2 and line 2 and 3, sm
zq 0 0 Q66 3m
zq
respectively (Fig. 3). For the plane stress model, the strains
in the global directions in the fibre layers are equal to the where the stiffness matrix [Q] contains the elastic constants
strains measured at the outside of the cylinder. The global of the isotropic material. The definitions of the different
strains can be translated to the strain in the cord direction in elements of the [Q] matrix can be found in Ref. [7].
the fibre layers by Eq. (1). When the global strains at critical For the unidirectional tape layers, loaded in-plane along
points 1 and 2 (Fig. 3) are substituted in Eq. (1), the amount its principal directions, the relation between the stresses and
of cord-slack is found for fibre layers 2 and 1, respectively strains can be represented by
0 f 1 2 3 0 3f 1
s1 Q11 Q12 0
31 Z 3qq cos2 a C 3zz sin2 a C gzq sin a cos a; (1) B f C 6 B f C
1
B s C Z 4 Q12 Q22 7
@ 2A 0 5,B @ 32 A
C (3)
where 31, the strain in the cord direction, is the amount of
cord-slack in this case and a is the angle between the fibre sf12 0 0 Q66 3f12
direction and the z-axis. For the multi-layer model, where It must be noted that although the relation’s (2) and (3) are
the strains (except for 3zz) are a function of the pipe radius r, identical, the elements of the [Q] matrix in Eq. (3) are
the strains measured at the outside of the cylinder need first defined in another way [7] and have different values in
to be translated to the strains at the fibre layers, before
different directions. When the unidirectional tape layer is
applying Eq. (1). The resulting cord-slack parameters or
stressed in the global directions (z, q), which are at an angle
Table 2 a to the principal directions (1,2), the resulting strains can
Estimated cord-slack values be evaluated by the following relation
0 f 1 0 f 1
Plane stress Multi-layer sqq 3qq
model (%) model (%) B f C B C
B szz C Z ½Q ,B 3fzz C; (4)
Fibre layer 1 1.6 1.75 @ A @ A
Fibre layer 2 1.45 1.6
sfzq gfzq
296 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300

where [Q*] is a fully populated [3!3] matrix, relating the As the isotropic and the fibre layers are heat welded
stresses to the strains in global directions for an together, it further applies that
unidirectional fibre layer under in-plane off-axis loading.
0 1 0 1
The [Q*] matrix is defined by 3m
qq 3fqq
B mC B C
½Q  Z ½TðaÞK1 ,½Q,½R,½TðaÞ,½RK1 ; (5) @ 3zz A Z B f C
@ 3zz A (9)
where 3m
zq gfzq
0 1
cos2 a sin2 a 2 sin a cos a
B C The stresses and strains in the global directions for the
½TðaÞ Z B
@ sin2 a cos2 a K2 sin a cos a C
A isotropic and the fibre layers can be written as a function of
the unknown A1, B1 and C1 by Eqs. (2)–(9). The three
Ksin a cos a sin a cos a cos2 a K sin2 a
unknowns A1, B1 and C1 in Eq. (8) are finally solved by
evaluating the following three equations [4].
and
0 1 X
1 0 0 (1) Equilibrium of axial forces Fzz Z 0 (10)
B C P
½R Z @ 0 1 0A
(2) Equilibrium of tangential
P forces Fqq Z 0
0 0 2 (3) Torque equilibrium MZ 0
For the case considered here, the stress–strain relations for
The plane stress model is further extended by taking into
the unidirectional reinforced tape can be simplified. For the
account the geometrical non-linearities. As a result of the
model, the height of the reinforced tape is taken equal to
deformations, which are considerable because of the fibre
the diameter of the aramid cords. When it is assumed that
slack, the change of the internal volume as well as the
the cords are perfectly cylindrical the height of the
change of the fibre angles are no longer negligible. The fibre
unidirectional fibre layer hf is defined as
sffiffiffiffiffiffiffiffi angle can be found to vary according to
4rc
hf Z ; (6) a Z a0 C ðK3zz C 3qq Þsin a0 cos a0 C gzq cos2 a0 ; (11)
prf

where rc is the cord linear density, commonly expressed in with a0 as the fibre angle in the unloaded state. Substituting
dtex and rf is the fibre density. The PE of the reinforced expression (11) into Eq. (5) leads to a non-linear system.
tape, which is located under and above the cords, indicated The cord-slack leads to further non-linearities. Details can
by the dotted area in Fig. 6 (right), is attributed to the be found in Section 4.2. The resulting system is solved
isotropic layers. As there is no adhesive bonding between iteratively, using a Newton–Raphson method [8].
the PE and the cords, no stresses can be transferred in the
principal 2-direction of the reinforced tape. As a result, there
is only a connection between s1 and 31 in Eq. (3). The values 4.2. Multi-layer ‘generalised plane strain’ model
for Q12, Q22 and Q66 in Eq. (3) are equal to zero. To
determine Q11, the stiffness in the 1-direction of the The previous assumption of uniform stresses through the
unidirectional fibre layer is needed, which is defined by thickness of one layer is only valid for thin walled tubes.
the rule of mixtures and is equal to The ratio between the axial and the tangential wall stresses
as a result of the hydrostatic pressure is 1:2 in that case.
E1 Z Ef ,Vf C Em ð1 K Vf Þ; (7) Based on the dimensions the LLRTP it can be demonstrated
where Ef is the fibre modulus and Vf is the fibre volume that this ratio will be equal to 1:2.25 rather than 1:2, and the
fraction. stresses can be non-uniform through the thickness. For this
As plane stress is assumed, the stresses for the isotropic reason, the LLRTP will be now considered as a thick
layers (i.e. liner and cover) can be represented by three cylinder and the stress situation will be characterised as
unknown constants generalised plane strain. The deformation in the z direction
is assumed to be uniform. The stresses and strains (except
sm
qq Z A1 sm
zz Z B1 sm
zq Z C1 (8) for 3zz and szz) will be no longer uniform, but they will
depend on the r-co-ordinate. First, the stress functions will
be deduced from general mechanics for this new approach.
This is presented by Eqs. (12)–(18). After that, the unknown
Fig. 6. Cross-section of the tape (left) and the modification used in the
constants are solved by formulating boundary conditions for
model (right). The circles represent the aramid cords. For the modification,
the tape height is set equal to the diameter of the cords. The remainder, this particular pipe problem.
(dotted area) consisting of plain PE is attributed to the isotopic PE layers in The stresses must satisfy the equilibrium equations in
the pipe cross-section (i.e. liner and cover). cylindrical co-ordinates, in the absence of body forces
 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300 297

vsrr 1 vsqr vszr srr K sqq

C , C C Z0
vr r vq vz r
vsrq 1 vsqq vszq 2,sqr
C , C C Z0
vr r vq vz r
vsrz 1 vsqz vszz srz
C , C C Z0 (12)
vr r vq vz r
Assuming axi-symmetry and generalised plane strain, the
equations in Eq. (12) can be simplified to
vsrr srr K sqq vsrq 2,srq
C Z0 C Z0 (13)
vr r vr r
The strain–displacement relations are in this case
vu vv v
3rr Z grq Z gqr Z K
vr vr r Fig. 7. Dimensions used for multi-layer model.
u vv
3qq Z gqz Z gzq Z (14)
r vz It is assumed, as in the previous described model that
vw the strains in the fibre layers are equal to the strains in
3zz Z Z Const: gzr Z grz Z 0
vz the isotropic material. For each of the three layers, the
For isotropic linear elastic material, the constitutive unknown constants A, B, C and D have to be solved,
equations are given by which brings the total amount of unknown constants to
12. The constants can be determined from conditions on
srr Z lð3rr C 3qq C 3zz Þ C 2,m,3rr the interfaces of the different layers supplemented with
sqq Z lð3rr C 3qq C 3zz Þ C 2,m,3qq similar equilibrium considerations as stated in Eq. (10).
First of all, continuity of the circumferential displacement
szz Z lð3rr C 3qq C 3zz Þ C 2,m,3zz szq Z m,gzq (15) v, which is derived in Appendix A, implies that the
constant B has to be equal for each layer. This reduces
where l and m are the Lame constants. When the strains the amount of unknown constants to 10. The necessary
(14) are substituted into the constitutive equations (15) conditions to solve the remaining 10 constants for
and the resulting stresses into the second simplified stress pressures larger than the second critical pressure point
equilibrium equation (13) it can be found (see Appendix are given below.
A) that
– Interface conditions for the radial strains
szq Z B,r; (16)
where B is an unknown constant. With the use of Eqs. 3l1 l2
rr Z 3rr for r Z Rf1 and
(13)–(15) it can be found that the stresses in radial and (20)
tangential directions can be expressed as 3l2 l3
rr Z 3rr for r Z Rf2 ;

KA A where Rf1 and Rf2 are the radii for the position of fibre
srr Z CC sqq Z C C; (17) layer 1 and 2, respectively (see also Fig. 7).
r2 r2
– Interface conditions for the axial strains
where A and C are unknown constants. By substituting
Eq. (17) into Eq. (15), it can be found that the stress in 3l1 l2
zz Z 3zz for r Z R1 to R5 and
the z-direction does not depend on r and z (21)
3l2 l3
zz Z 3zz for r Z R1 to R5
szz Z D; (18)
where D is an unknown constant to be determined. Since
the cords are very thin compared to the pipe it is – Edge stress conditions
assumed that the stresses and strains can be modelled as
uniform through the thickness and thus plane stress is sl1
rr Z KP for r Z R0 and
assumed for the fibre layers. As indicated in Fig. 7, it is (22)
sl3
rr Z 0 for r Z R5
assumed that the LLRTP consists of three layers of
isotropic linear elastic PE, which are separated by the
two infinitely thin fibre layers. The three isotropic layers The infinitely thin fibre layers will cause discontinuities
are defined by in the radial and the tangential stress from one isotropic
layer to another. The discontinuities in the radial stresses
li Z R2iK1 K R2iK2 for i Z 1 to 3 (19) should balance the tangential stresses in the cords by
298 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300

the following expressions

2pR2f1 Af
,ðsl2 l1 f1
rr K srr Þ Z sqq ,
s
K Fqq1 and
tanðq1 Þ sinðq1 Þ
2pR2f2 Af
,ðsl3 l2 f2
rr K srr Þ Z sqq ,
s
K Fqq2 ; ð23Þ
tanðq2 Þ sinðq2 Þ
where the terms with superscript s are the slack terms and F
indicates the force exerted by the cords. The slack terms are
defined and treated in more detail below.

– Torque equilibrium
X
2 3 ð R2iK1
X
Af Rfi sfi zq s
K Fzq C rszq 2pr dr Z 0; (24)
iZ1
cosðqi Þ iZ1 R2iK2

where szq is given in Eq. (16).

– Equilibrium of axial forces
Fig. 8. Comparison of experimental and predicted (M) values for
X
3 X2
Af sfizz hydrostatic pressure load-case. For the predicted values the updated plane
pðR22iK2 K R22iK1 Þ,slizz C K Fzzs Z PpR20 stress model is used. The dotted and the continuous lines refer to the
iZ1 iZ1
cosðq i Þ calculated values and the experimental data, respectively.
(25)

The cord-slack 3s is the estimated slack in the cord as a linear elastic material. The prediction of the axial strain
direction. The resulting stress in the cords is equal to the is observed to be less accurate. The RTP is modelled too
cord stiffness times the strain minus the cord-slack, which is stiff in the axial direction, which cannot be caused by a
expressed as a compressive strain. This is done by mismatch of the fibre angles, as the tangential strain is very
subtracting the terms: Fqq s
; Fzq
s
and Fzzs from the resulting well described. The torsion angle is not very well predicted
fibre layer forces. These three terms are defined as either. When the estimated values for the fibre slack are
used, the jump in torsion angle is indeed in the right
s
Fqqi Z Vf Af Ef 3sfi sinðqi Þ for i Z 1; 2 direction but far larger than the experimentally measured
jump in torsion angle. For pressures higher than the second
s
X
2
critical pressure point, the model predicts a further increase
Fzq Z Vf Af Ef 3sfi sinðqi ÞRfi
of the torsion angle, while the experiments show the
iZ1
opposite trend.
X
2 The results for the multi-layer model are presented in
Fzzs Z Vf Af Ef 3sfi cosðqi Þ (26) Fig. 9. It is obvious from Fig. 9 that the strains are far better
iZ1
described by the multi-layer model than for the plane stress
Again the same non-linearities as described for the plane model. For the axial strain, the largest deviation appears at
stress model are taken into account. The constants are found pressures before the critical pressures, where hardly any
solving the equations using a Newton–Raphson method, axial strain was measured. The model however predicts a
after which all strains, stresses and cord forces can be linearly increasing axial strain for this pressure range. It is
calculated. very likely that this deviation can be attributed to the visco-
elastic behaviour of the PE. The jump in torsion angle of
about 18/m is very well described by the model when the
estimated cord-slack difference of 0.15% is used in the
5. Results multi-layer model. When the slack is consumed for both
layers, the model also predicts a reverse of the torsion angle.
The results of the experiment and the plane stress model, It is likely that the more non-linear character of
are presented in Fig. 8. As already mentioned, the pipe is
only subjected to an internal pressure and is free to deform Table 3
in all directions. The input data, which is used for the model Input data used in model for a 4 00 LLRTP
can be found in Table 3. The tangential strain is well R0 50 mm R5 62.5 mm
predicted for pressures larger than 35 bar. The deviations R1 56.54 mm Rf1 57.15 mm
between the measured and the modelled tangential strains at R2 57.76 mm Rf2 58.95 mm
R3 58.34 mm Vf 0.46
lower pressures can be explained by the non-linear R4 59.56 mm hf 1.22 m
behaviour of the PE, whereas the model describes the PE
 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300 299

account in the multi-layer model. This behaviour is made

possible by the relatively low elasticity modulus of PE.
The RTP is therefore sensitive to a small unbalance of the
torsion moments caused by the two different fibre layers.
It is of great importance to take into account the change
of the fibre angles during deformation. The average initial
fibre angle is equal to 54.78. This is based on thin cylinder
theory, from which it follows that the ratio between the
tangential and axial stresses is 2:1. When it is assumed that
the matrix transfers no loads, one can find:
sqq sin2 ðqÞ,s1
Z Z tan2 ðqÞ Z 20 q Z 54:738 (27)
szz cos2 ðqÞ,s1
Based on the initial geometry of the LLRTP, it can be found
that the ratio between the average tangential and axial
stresses is in the order of 2.25:1. With the use of Eq. (27) it
can be found that the average initial angle should be 56.38. It
Fig. 9. Comparison of experimental and predicted (M) values for can be concluded that with the average initial angle of 54.78,
hydrostatic pressure load-case. For the predicted values the multi-layer too much stiffness is applied in the axial direction and as a
model is used. The dotted and the continuous lines refer to the calculated
values and the experimental data, respectively.
result too little stiffness is applied in the tangential direction.
This is partially corrected by the cord-slack. Because the
cord-slack is consumed at about 2% tangential strain of the
the measured reverse torsion angle is also a result of the
LLRTP (without any axial strain), the fibre average fibre
visco-elastic behaviour of the PE.
angle is equal to 55.248 when it starts contributing to resist
the inner pressure. It was also found that by incorporating
6. Discussion the change of the fibre angles as a function of the strains
in the model, the average fibre angle converges to 56.38 by
Although the inner and outer fibre layers generate a increasing the inner pressure. Including the change of the
positive and negative torsion moment respectively, they fibre angles during deformation will be even of greater
are not balanced as indicated by the negative torsion angle importance for a good prediction of the resulting strains in
at values of the inner pressure larger than the second case of off axis load-cases.
critical pressure point. The resulting torsion moment In most cases axial deformation is unwanted. A perfect
induced by one layer is defined as the resulting shear LLRTP should not increase in length and should not show
force times the distance to the centre of the RTP. Three any torsional behaviour when pressurised. As the mechan-
mechanisms are responsible for the unbalance in torsion ical behaviour is now well understood, it is now possible
moments: with the use of the model to optimise the initial angles for
the two fibre layers in order to keep the axial strain and the
1) The two layers have different fibre angles
torsion angle both equal to zero at a certain pressure level.
q1 sq2 0 tfzq1 stfzq2 0 Fzq
f1 f2
sFzq

7. Conclusions and future work

2) The two layers have different fibre radii
The mechanical behaviour, and in particular the torsional
behaviour of a LLRTP was analysed and explained. It
Rf 1 sRf 2 appeared that the complex torsional behaviour is a result of
the cord-slack difference between the two fibre layers.
The cord-slack values for the two fibre layers were
3) The strains are non-uniform trough the thickness estimated from the strain–internal pressure behaviour. The
existing, earlier published plane stress model was modified
to take into account the change in cord angle and the change
3fqq1 s3fqq2 0 Fzq
f1 f2
sFzq
of the geometry. When the estimated cord-slack values were
The first two mechanisms are taken into account in the incorporated into the model, the model did not show good
plane stress model and resulted in a monotonously agreement with most of the experimental data.
increasing torsion angle. A reduction of the torsion The model was further extended to take into account the
angle is only found by taking the third mechanism into non-uniform strain distribution through the thickness.
300 M.P. Kruijer et al. / Composites: Part A 36 (2005) 291–300

The updated model shows good agreement with the And hence
experimental data. To further improve the model it will be
necessary to take into account the visco-elastic behaviour of E 00 ðzÞ Z C1 en F 00 ðzÞ Z C2 ; (A5)
the PE, which is the objective for the ongoing research in leading to
this field.
EðzÞ Z C1 ,z C D1 en FðzÞ Z C2 ,z C D2 (A6)
When Eq. (A6) is substituted into Eq. (A4), this yields
Acknowledgements z  
1
v Z C1 C D1 C C2 ,z,r C D2 ,r (A7)
Pipelife Nederland B.V., producer of Soluforcew LLRTP r r
made this research possible by making available their test After applying the boundary conditions for the axial
equipment and LLRTP specimens. The authors wish to symmetric pipe it turns out that
express their thanks to the company and most especially to
vðr; zÞ Z C2 ,z,r (A8)
Dr L.G.P. Dalmolen and Dr A. Fahrer for their helpful
discussions. With the use of Eqs. (14) and (15) it can be finally found that
szq Z B,r (A9)
Appendix A. Derivation of the stress function for szq

When the strains (14) are substituted in the constitutive References

equations (15) and the resulting stresses into the second
simplified stress equilibrium equation, it can be found that [1] Frost SR. Glass fibre-reinforced epoxy matrix filament-wound pipes for
use in the oil industry. Compos Manufact 1994;5(2):73–82.
vgrq 2,grq [2] Frost SR. The development of RTP for use in the oil industry. In:
C Z0 (A1) Composite materials for offshore operations-2. Houston: American
vr r Bureau of Shipping; 1999 p. 341–60.
which can also be written as [3] Fisher EH, Gibson AG. Continuous fibre reinforced thermoplastic pipes
for transport and distribution of fluids for the oil and gas industries.
v 2 Plast, Rubber Compos 1999;27(10):447–51.
ðr grq Þ Z 0; with solution0 r 2 grq Z f ðzÞ (A2)
vr [4] Chapman BJ, Evans JT, Frost SR, Gibson AG. Mechanical testing of
continuous fibre reinforced thermoplastic pipes. In: Proceedings of
It further implies that 11th International Conference on Composites Materials, Australia
vv v v v p. V517–26.
grq Z K Z r (A3) [5] Soluforcew reinforced thermoplastic pipesystem. Commercial
vr r vr r brochure, Pipelife Nederland B.V.
Substitution of Eq. (A3) into Eq. (A2) leads to [6] Gibson AG, Hicks C, Wright PNH, Fahrer A. Development of glass
fibre reinforced polyethylene pipes for pressure applications. Plast,
v f ðzÞ EðzÞ Rubber Compos 2000;29(10):509–19.
ZK C gðzÞ or vZ C FðzÞ,r (A4)
r 2,r 2 r [7] Powell PC. In: Engineering with fibre-polymer laminates. London:
Chapman & Hall; 1994.
gzq Z vv=vz is only a function of r from which follows that [8] Verbeke J. The Newton–Raphson method. Int J Math Educ Sci Technol
E 00 ðzÞ=rC F 00 ðzÞ,r can only be a function of r. 1995;26(2):177–94.