Professional Documents
Culture Documents
10.1.1.659.8414 Elbow31 Vs Elbow32, Modes, Shell Etc
10.1.1.659.8414 Elbow31 Vs Elbow32, Modes, Shell Etc
PVP2014
July 20-24, 2014, Anaheim, California, USA
PVP2014-28920
1
Corresponding author.
Contact info: L Zeng, Nuclear Technology, ÅF-Industry AB, Email. Lingfu.zeng@afconsult.com. Tel. +46 (10) 5053264.
2
Temporary engineer.
Pipe axis
y y i a i ( r 0 u r ) r u t t y 3a 3 (1) r°
Crown
where r a1 cos a2 sin and t a1 sin a2 cos.
tn a3 a3 .tt tt (3) a2
r°
a3 S
The vector n is written as a1
1
n n(n n) 2 (4) x
where
x
n a3 3tn (5)
Fig. 1 The undeformed configuration of a pipe and relevant
notations used for formulation of ABAQUS elbow
with 3 3 ( , S ) . elements
This definition ensures that n x / 0 , in addition, a penalty
term can be introduced later to ensure that n x / S is small. where H n ( S ) is a polynomial interpolation function of order
(N-1). The same function is used to interpolate a rotation
With such a definition of the pipe geometry, the interpolation is triplet,
made for the deformed pipe section in the following way: The
N n
current position of the pipe axis, x , and the direction vectors, ( S ) H n ( S ) (7)
n 1
a , are first interpolated as functions of S and, thereafter, the
offset of the mid-wall point, y , and the outward-normal
This gives a C ( ) A , where C is the rigid rotation matrix.
direction vector, n , interpolated as functions of S and .
x n
M P 3
H m ( S )Q asym ( p )( 0 ) mp (out-of-plane) b ( , S ) sym (16)
m 1 p 1
(12)
where I is the unit matrix. Choose now (1) Only the even numbered Fourier terms are used in the
approximation (23), which is identical to that introduced
x originally by von Karman. This implies that these
n 0 (21) displacements are symmetric about the crown of the elbow.
This is satisfactory for cases where the pipe radius is small
compared to the elbow torus radius, but is likely to be
and impose a penalty to make n x / S small. In the inadequate when the pipe radius is of the same order as the
computation of the curvatures, b , we torus radius.
approximate n x / S 0 , so that
Tab. 1 Relative deflections and rotations at the cantilever tip
3. BENCHMARK STUDIES computed by ELBOW31 and ELBOW32 elements
1.65
k 11.19
h
The stress intensification factor (SIF) is defined as the ratio of Integration points
through thickness
the maximum stress of the elbow to that determined by beam
theory. The analytical solution of the stress intensification
factor for this pipe bend under in-plane bending is [22]:
0.84 0.84
Axial stress: iA 2/3
3.015 Pipe data (Bathe & Almeida, 1980):
h 0.147 2 / 3 Wall‐thickness (t) = 0.374 in (9.525 mm)
1.80 1.80 Outer radius of the pipe (ro) =20.8t
Hoop stress: iH 2 / 3 =6.462 Radius of the pipe bend curvature (R) = 3.07r
h 0.147 2 / 3 Length of the long straight pipe (L1) = R=3.07r
Length of the short straight pipe (L2) = 0.77r
These SIF-values are adopted by the ASME Boiler & Pressure Young’s modulus (E) = 24.1 x106 psi (169.74 GPa)
Vessel Code for Class 1 nuclear piping. For Class 2/3 nuclear Posion’s ratio () =0.3
power piping, the SIF values are one-half of the above values.
Fig. 3 A pipe bend subjected to in‐plane bending
It may be worthwhile to note that beam-theory based software
for piping analysis. such as PIPESTRESS [5] or other similar
[4], is developed in such an ad hoc manner: Pipelines are all Computations are conducted using the ELBOW31 and
modeled by standard beam elements of circular pipe section, ELBOW32 elements, using 6 ovalization modes on two meshes
with bending stiffness terms for curved pipes “modified” as specified in Tab. 2 for modeling details, one relatively
through a so-called “flexibility knockdown factor”, and after coarse and one mesh with doubled number of elements, and
the finite element analysis the computed stresses in the curved using ABAQUS shell elements (4-node fully integrated shell
where M is the bending moment, E the elasticity modulus and I (2) Hoop stresses, at the inside layer of Section A-A,
the moment of inertia of the pipe section. Notice that the computed by ELBOW32, using 6 ovalization modes and
(3) The peak value of the axial stresses at the outside layer of
Section A-A, computed by ELBOW31 and ELBOW32,
agree well with the peak stress value predicted by
ABAQUS shell elements. But there is a notable difference
for axial stresses at the middle layer and at inside layers (b) Axial stress, Section A‐A, Inside layer
between the elbow and shell elements of ABAQUS. See
Fig. 9. Moreover, the locations where the peak stress taken
place predicted by ELBOW31 and ELBOW32 are at about
72 counting from the extrados, which are close to the
results by others, in particular, the analytic solution by
Clark & Reissner [22] and the experimental observation by
Smith & Ford [23]. However, the location predicted by the
shell elements is almost at 90 counting from the extrados,
which is probably due to that the mesh used for the
analysis, Fig. 4, is still too coarse.
(4) The axial stresses predicted for Section B-B by the
ELBOW31 and ELBOW32, using 6 ovalization modes,
agree well with those by the ABAQUS shell elements. See
(c) Axial stress, Section A‐A, Outside layer
Fig.10.
(5) The ovalization predicted by the ELBOW31 and
ELBOW32 seem differ surprisingly much from that by
Dodge & Moore and by Sobel using software MARC. The
difference may possibly be caused by inconsistent
specification of input data and a careful verification of the
original data seems to be required.
(6) A significant difference in results using both ELBOW31
and ELBOW32 for choosing 3 and 6 ovalization modes
can be noticed. The minimum number of ovalization
modes for such a 90-degreed pipe bend requires at least 6
modes for a good prediction of peak stress values.
(7) It can be observed that the ADINA results, reproduced
directly from the original paper by Bathe and Almeida Fig. 5 Axial stresses computed in three layers at Section A‐A of
[12], seem to be better close to the analytical [22] and the pipe bend using ELBOW31 and ElBOW32 with 3 and
experimental results [23], although the computation was 6 ovalization modes. (a) Mid‐layer, (b) Inside layer,
made on a coarse mesh with 3 uniform elements used in (c) Outside layer.
the pipe bend.
(b) Hoop stress, Section A‐A, Inside layer (b) Axial stress, Section B‐B, Inside layer
(c) Hoop stress, Section A‐A, Outside layer (c) Axial stress, Section B‐B, Outside layer
Fig. 6 Hoop stresses computed in three layers at Section A‐A of Fig. 7 Axial stresses computed in three layers at Section B‐B of
the pipe bend using ELBOW31 and ElBOW32 with 3 and the pipe bend using ELBOW31 and ElBOW32 with 3 and
6 ovalization modes. (a) Mid‐layer, (b) Inside layer, 6 ovalization modes. (a) Mid‐layer, (b) Inside layer,
(c) Outside layer. (c) Outside layer.
(b) Hoop stresses, Section B‐B, Inside layer (b) Axial stresses at Section A‐A (ABAQUS/ELBOW32)
(c) Hoop stresses, Section B‐B, Outside layer
(c) Axial stresses at Section A‐A (ABAQUS/SHELL S4)
Fig. 8 Hoop stresses computed in three layers at Section B‐B of Fig. 9 Axial stresses at Section A‐A of the pipe bend computed
the pipe bend using ELBOW31 and ElBOW32 with 3 and using ELBOW31 and ELBOW32, with 6 ovalization modes,
6 ovalization modes. (a) Mid‐layer, (b) Inside layer, and SHELL S4 of ABAQUS. (a) ELBOW31; (b) ELBOW 32,
(c) Outside layer. (c) SHELL S4.
(b) Axial stresses at Section B‐B (ABAQUS/ELBOW 32) (b) Ovalization at Section A‐A
(Warping/ovalization not restrained at the free pipe‐end)
(c) Axial stresses at Section B‐B (ABAQUS/SHELL S4) Fig. 11 Ovalization computed at Section A‐A of the pipe bend
with different restraining conditions at the free pipe‐
end: (a) Warping and ovalization at the free end not
restained, (b) Warping and ovalization at the free end
fully restained.
4. CONCLUDING REMARKS
In this paper, elbow elements in ABAQUS are overviewed
and two most commonly used elements, ELBOW31 and
ELBOW32, are tested for two Benchmark examples. It is
shown from the tests that: (i) these elements predict a good
peak stress solution with a reasonably coarse mesh and 6
ovalization modes; (ii) the ovalization and the distribution of
stresses predicted around the pipe section show, though using a
Fig. 10 Axial stresses at Section B‐B of the pipe bend computed relatively dense mesh, a notable difference as compared to
using ELBOW31 and ELBOW32, with 6 ovalization solutions computed by ABAQUS shell elements; (iii) the
modes, and SHELL S4 of ABAQUS. (a) ELBOW31; ADINA elbow element seems to provide, though using a very
(b) ELBOW 32, (c) SHELL S4. coarse mesh, a solution closest to analytic and experimental
results. Moreover, it is observed that computations using these