Fang Et Al. - 2018 - Investigation On Mechanical Properties of Fibreglass Reinforced Flexible Pipes Under Torsion
Fang Et Al. - 2018 - Investigation On Mechanical Properties of Fibreglass Reinforced Flexible Pipes Under Torsion
OMAE2018-77354
ABSTRACT bonded pipes, it has a higher density and shows better subsea
Fibreglass reinforced flexible pipe (FRFP) is regarded as a great mechanical behavior.
alternative to many bonded flexible pipes in the field of oil or FRFP studied in this paper, shown in (Fig.1), consists of a
gas transportation in shallow water. This paper describes an polyethylene liner, eight layers of reinforced tape made of
analysis of the mechanical behavior of FRFP under torsion. The polyethylene and fibreglass wrapping around the liner and an
mechanical behavior of FRFP subjected to pure torsion was outer polyethylene coating. The inner liner pipe is ultra high
investigated by experimental, analytical and numerical methods. molecular weight polyethylene (UHMWPE) and the outer
Firstly, this paper presents experimental studies of three 10-layer coating pipe is high density polyethylene (HDPE), while the
FRFP subjected to torsional load. Torque-torsion angle relations reinforced tape is made of HDPE and fibreglass.
were recorded during this test. Then, a theoretical model based
on three-dimensional (3D) anisotropic elasticity theory was
proposed to study the mechanical behavior of FRFP. In
addition, a finite element model (FEM) including reinforced
layers and PE layers was used to simulate the torsional load
condition in ABAQUS. Torque-torsion angle relations obtained
from these three methods agree well with each other, which
illustrates the accuracy and reliability of the analytical model
and FEM. The impact of fibreglass winding angle, thickness of
Figure 1. Structure of FRFP
reinforced layers and radius-thickness ratio were also studied.
Conclusions obtained from this research may be of great
Flexible pipes under various loads have been studied by
practicality to manufacturing engineers.
many scholars, however, most of the research is confined to
unbonded flexible pipes. On the numerical side, A. Bahtui et al.
KEY WORDS: Torque; Flexible pipe; Numerical solution;
[1,2] modeled an unbonded flexible riser under various loads in
Experiment; Torsional rigidity
which all layers are separated with contact interfaces between
each layer. Hector E. M. Merino et al. [3] created a FEM under
INTRODUCTION
torsion considering the friction and adhesion between layers,
Recently, fibreglass reinforced flexible pipe (FRFP) has
the results show that finite element estimations agree quite well
been widely used in offshore transportation due to its high
with the experimental measurements. Then M. S. Liu et al. [4]
corrosion resistance, light weight characteristic and relatively
imported a model into ABAQUS to simulate the riser’s
low fabrication and installation cost. Compared to other types of
mechanical behavior under torsion. This model takes into
*Corresponding author.
E-mail address: [email protected]
(a) Specimen 1
(b) Specimen 2
E FB E PEVI
ET E r E PE (1 VI )
VFB VFB (2)
E PE E FB(1 )
VI VI
VFB
G FB G PE
VI VFB (3)
G LT G Lr G PE (1 )
G PEVI G FB(1 VI ) VI
G FB G PEVI
GTr G PE (1 VI )
VFB V (4)
G PE G FB(1 FB )
VI VI
LT Lr FBVFB PE(1 VFB ) (5)
Figure 7. Cylindrical coordinate system
E FB
FBVFB PE (V VFB )
, E PE I
Tr PE (1 VI ) (6)
VFB E V
FB (1 FB )
T Z VI E PE VI
θ
L Table 4. Elastic constants of reinforced layers
0 0 0
1
0 0
GTr C12 C 22 C 23 0 0 C 26
1 C
(k ) C13 C 23 C 33 0 0 C 36
(14)
0 0 0 0
G LT
0 0 0 0 C 44 C 45 0
1 0 0 0 C 45 C 55 0
0 0 0 0 0
G LT C16 C 26 C 36 0 0 C 66
Where the symbol S is on-axis flexibility matrix. The
superscript “-1” in Eq. (7) denotes the inverse matrix. Inner Layer and Outer Layer Stiffness Matrix (k=1, n)
By introducing the flexibility and stiffness transformation As a homogenous isotropic material, the off-axis stiffness
matrices T and T which are a function of the angle , the
(k )
matrixC of kth layer is given as
relationships between the off-axis and on-axis elastic constants (k )
C11 C12 C13 0 0 0
are expressed as C C C 0
S ( k ) T S ( k ) (T 21 22 23 0 0
(k ) (k ) T
) (9)
C C C 0 0 0
C ( k ) T C ( k ) (T ) , k 2,3n 1
(k ) (k ) T
(10) C ( k ) 31 32 33 (15)
Where 0 0 0 C44 0 0
(k ) 0 0 0 0 C55 0
m2 n2 0 0 0 mn
2 0 0 0 0 0 C66
n m2 0 0 0 mn
It can be calculated by
0 0 1 0 0 0 (k ) ( k ) 1
T (S ) , k 1, n.
(k )
(11) C (16)
0 0 0 m n 0 1 k k
(k )
0 0 0 n m 0 E Ek Ek
0 0 0
k
k 1 k
2mn 2mn 0 0 0 m 2 n 2 E Ek
Ek
0 0 0
k
(k ) k k
m 2 n 2 0 0 0 2mn
1
0 0 0
2 S
(k )
Ek Ek Ek
n m2 0 0 0 2mn 1 (17)
0 0 0 0 0
0 0 1 0 0 0 Gk
T
(k )
(12)
0 0 0 0
1
0
0 0 0 m n 0 Gk
0 0 0 n m 0 1
0
0 0 0 0
Gk
mn mn 0 0 0 m 2 n 2
r z
r
r (k ) (k )
u( k ) u( k ) u( k ) u ( k )
(20) ( k ) C 22 / C 33 (26a)
zr 0, r r
(k ) (k )
, z(k ) 0 r
r r r r z (k ) (k )
k C12 C13
Where, 0 is twist of pipe per unit length which is 1 (k ) (k )
(26b)
C 33 C 22
independent of z. The axial deformation of all layers is equal to a (k ) (k )
k C 26 2C 36
constant 0. 2 (k ) (k )
(26c)
In the absence of body forces, equilibrium equations in 4C 33 C 22
cylindrical coordinates can be described as: For the anisotropic materials (reinforced layers) used in
r( k ) 1 ( kr ) zr( k ) r( k ) ( k ) this study, there exists
0 (21a)
r r z r (k ) (k ) (k ) (k )
C 22 / C 33 0 and C 22 / C 33 1 .
( kr ) 1 ( k ) z(k ) 2 ( kr )
0 (21b) Using transformation matrix T
(k )
and T
(k )
, strains and
r r z r
zr( k ) 1 z(k ) z( k ) zr( k ) stresses in local material coordinates can be obtained:
r
r
z
r
0 (21c) ( k ) T( k ) ( k )
(27)
( k ) T( k ) (k )
Reinforced Layers Analysis ( k 2,3(n 1) )
The stress-strain relations of the k-th layer are given as Inner Layer and Outer Layer Analysis ( k 1, n )
C
(k ) (k ) (k )
(22) For isotropic materials, the stress -strain relations of the kth
layer are given as
C ( k )
(k ) (k )
(28)
0 C 44 z
k 1
0 0 0 0 Torsion condition
Substituting Eq. (20) and Eq. (29) into equilibrium equation n
2 z r r 2 dr T
rk k
Eq. (21), the equilibrium equations lead to: rk 1
(35)
k 1
r( k ) ( k ) ( k ) Using Eq. (32b), Eq. (33d) and Eq. (33e) can derive that the
r 0 (30a)
r r integration constant, A( k ) 0 . Therefore, the hoop
( kr ) 2 ( kr ) displacement can be expressed as follow:
0 (30b) u 0 rz
r r (36)
(k ) (k )
As an isotropic material, there exits C / C 1 , 22 33
For N-layered FGRFP, there are 2n+2 unknown constants
C C13 and C 26 C 36 0 , which leads to 1 ,
(k )
12
(k ) (k ) (k ) (k )
of integration, D k , E k k 1, 2 n , and 0 , 0 . Eqs.
k
1
(k )
2 0 ( k 1, n ) in Eq. (26) (32a), (33a), (33c), (34), and (35) can give 2n+2 equations to
determine these unknown constants.
Hence, the solutions of ur , u and u z in the inner and The integration constants for the multi-layered FGRFP can
outer layers can be expressed by: be obtained from the solution of the simultaneous equation as
u( k ) 0 rz A( k ) Eq. (37). Elements in the matrix M can be obtained through
transformations of 2n+2 equations.
uz( k ) 0 z B( k ) (31) D 1
E
k
2 0
u (k )
D r (k )
D 0
r
r
where, A( k ) , B ( k ) , D( k ) , E are unknown integration
k
D n
constants to be determined. 0
E 1
0
M 2 n 2,2 n 2 (37)
Boundary Conditions E
2
0
The unknown integration constants in Eqs. (25) and (31)
can be determined by substituting these equations into E n 0
boundary conditions and solving the algebraic equations. 0
0
T / 2 * pi
The traction-free condition at the inner surface and outer
2 n 2,1
surface are written as: 0
r(1) r0 0, r( n ) rn 0 (32a) Once their values are determined, the strains and
displacements are thus obtained from Eq. (20), Eq. (25) and Eq.
1r r0 zr1 r0 0, nr rn zrn rn 0 (32b) (31), respectively. Then the stresses can be calculated as
Where r0 and rn are the inner and outer radius, products of strains and stiffness matrices.
respectively.
Geometric Nonlinearity
As a result of the deformation, the change of cross -section
Interface Conditions dimension and the fiber angles are no longer negligible.
Assuming that the interfaces between the core and skin Nonlinearity should be taken into account in the analysis. The
layers are perfectly bound, continuity conditions for the change of cross-section dimension can be obtained from the
displacements and stresses at the interfaces lead to variable radius displacements. The change of fiber angles in the
ur k rk =ur k 1 rk (33a) reinforced layers can be calculated from the following equation
proposed by Kruijer (2005)[1]:
u rk =u rk
k k 1
(33b) k 0 k ( z )sin 0 k cos 0 k z cos2 0 k (38)
rk r k 1 rk
k
r (33c) Where, 0 k is the initial wound angle of the k-th
reinforced layer. Change of the wound angle leads to a non-
NUMERICAL SIMULATIONS
In this part, a finite element model (FEM) is presented
studying the mechanical behavior of FRFP by using
ABAQUS/Standard nonlinear finite element analysis tool. The
geometrical dimensions of the FEM are in accordance with
those from manufacturing.
As shown in Fig.9 and Fig.10, the FRFP consists of 10
layers, an inner layer, an outer layer and 8 reinforced layers. It is Figure 10. Side view of FRFP
assumed that the interfaces between layers are connected
closely. Extrusion and partition commands were used to
separate the reinforced layer into 8 layers. The 8-node linear
brick reduced integration element C3D8R was used to mesh the
pipe. HDPE was considered isotropic and fibreglass was
considered transverse isotropic, therefore, the orientation of
reinforced layers was assigned according to different layers. A
global rectangular system was defined and X, Y, Z represents
three directions. An orientation discrete field defining a spatially
varying orientation was selected, on which a cylindrical Figure 11. Discrete field of one layer before aligning orientation
coordinate system was defined and 1, 2, 3 denotes the axial,
hoop and radial direction, respectively. For example, one layer of
the model before aligning orientation is shown in Fig.11. An
additional rotation of 55°was assigned to this layer along the
radial direction 3, which is shown in Fig.12.
Since the ends of FRFP were fixed with end-fittings in the
test machine, in order to make the loading conditions similar to
the ones from test specimens, one of the ends was totally fixed,
and the other end was considered only rotated along the Z axis.
On one end, the cross-section was coupled to a reference
point RP2 at which point the displacement of u1, u2, u3, ur1, ur2,
Figure 12. Discrete field of one layer after aligning
ur3 were constrained to zero. The other end was coupled to
orientation
another reference point RP1, which was located at the center of
the cross section. At last, only a displacement loading was
applied in the ur3 direction at the reference point RP1 to
RESULTS AND DISCUSSIONS
simulate the torsion of FRFP.
In order to investigate the accuracy and reliability of the
results obtained from the analytical solution and finite element
model, the torque-torsion angle curve from these two methods
were compared with the curve from experiment. The theory
process was solved by MATLAB. In MATLAB the total torque
applied on the pipe was 3,000N·m, which was added evenly over
500 steps, so 6N·m torque was applied on the pipe every step.
While in the FEM, on one of the reference point, UR3 was
rotated 0.4 rad. The deformation and von Mises stress
distribution of the model is shown in Fig.13. The torque-torsion
angle relationship of these three methods agree with each other
well enough in the elastic phase, however, there are still some
differences among these three methods mainly because of the
Figure 9. Front view of FRFP material properties. The average torsional rigidity of these three
methods remains about 8333N·m/rad, as shown in Fig.14.
PARAMETRIC ANALYSIS
ACKNOWLEDGMENTS
The authors would like to thank Shang Hai Fei Zhou Bo
Yuan Petroleum Equipment Company and OPR (Offshore
Pipelines and Risers Inc.) for providing the specimens and
laboratory equipment. Also, we appreciate that Kevin Conroy
seriously help us to examine the grammar of this paper.