Beruflich Dokumente
Kultur Dokumente
www.elsevier.com/locate/ijpvp
Abstract
This study concerns the load interaction behaviour of 908 smooth piping elbows with circular cross-section and long straight tangent pipes.
The finite element method is used for stress analysis of elbows having a wide range of bend and pipe factors. The main aim of the study is to
establish the first yield interaction behaviour when an elbow is subjected to a combination loading of in-plane bending, torsion and internal
pressure. The study shows that load interaction is influenced by pipe factor, bend radius and load coupling effect, with thinner elbows being
affected to a larger degree.
q 2003 Elsevier Ltd. All rights reserved.
Keywords: Load interaction; First yield; Pipe elbow; Finite element method
1. Introduction bending and Kafka and Dunn [3] studied the influence of
internal pressure on in-plane bending. Rodabaugh and
This study looks at the main parameters that influence the George [4] made a serious attempt to analyse the effect of
performance of piping elbows, namely pipe dimensions and internal pressure on stresses when the elbow is being
the applied load combinations. The pipe factor l is chosen subjected to in-plane and out-of-plane moments, and after
to be between 0.0267 and 1.4 and the bend factor b ¼ 2; 3, 5 some modification for hoop membrane stresses by Gross
and 7. These values represent the range of piping commonly [5], the combined analyses form the basis of the stress
used in power plants. The main study deals with the first indices currently used in the nuclear piping code of ASME
yield behaviour of elbows when two loads act simul- Section III [6]. The BS 806 [7] code is based on studies
taneously, assuming linear behaviour and the absence of a carried out by Turner and Ford [8] and Smith [9]. They
load coupling effect. The resulting load interaction diagrams accounted for mid-wall strain and did not limit the value of
provide the means of assessing the level of conservatism in R=rm ; i.e. two of the limitations of the work by Rodabaugh
present design rules. A geometrically non-linear FE and George [4]. Many researches have been devoted to the
analysis is then carried out to investigate the extent of study of non-linear combination of pressure and in-plane
cross-sectional ovalization in two elbow models (thick and moment on elbows. Rodabaugh and George [4] and Spence
thin). The non-linear analysis is then extended to include
and Boyle [13] studied the influence of pressure on stresses
simultaneous application of in-plane bending moment and
produced by moment loads.
internal pressure to evaluate the coupling effect between the
For a combination of loads applied simultaneously, the
two loads.
ASME [6] and BS [7] formulation is largely based on the
The first theoretical treatment of elbows was carried out
maximum stress intensity (wherever it occurs) being equal
by von Karman [1] who developed stress formulae for in-
to, or less than, the sum of the maximum stress intensities
plane moment on simple elbows. Based on a similar energy
method, Vigness [2] studied the effect of out-of-plane due to the loads taken individually. This gives a con-
servative design if the maximum stresses from the
* Corresponding author. Tel.: þ 60-7-5534580; fax: þ 60-7-5566159. individual loads do not occur at the same location. In this
E-mail address: amran@fkm.utm.my (A.B. Ayob). study the maximum stress is obtained after the stresses from
0308-0161/$ - see front matter q 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2003.09.001
862 A.B. Ayob et al. / International Journal of Pressure Vessels and Piping 80 (2003) 861–869
the individual loads are summed at the corresponding pipe body motion of the structure, suppression of all displace-
locations. ments was applied at the other tangent pipe free end.
Following a linear stress analysis, a geometrically non-
linear analysis was carried out on two elbow models, EL01
2. The finite element models and EL07. Model EL01 represents a thin elbow with a small
bend radius and elbow model EL07 represents a moderate
The geometrical features of all 16 models, representing a elbow commonly used in industry. These two models were
wide range of b and l; are given in Table 1. The models subjected to separate loads of torque, in-plane moment and
have a bend angle of 908 and a mean diameter, dm ¼ 0:2 m: internal pressure to assess the extent of geometrical non-
The choice of dm is irrelevant because the results are linearity.
presented in terms of non-dimensional parameters. Finally, an analysis was carried out to determine the
The choice of mesh, geometric and load modelling was coupling effect between in-plane moment and internal
based on preliminary exercises performed successfully [10] pressure. Again, elbow models EL01 and EL07 were
on cylinders and branched pipes. A similar FE mesh was selected and they were subjected to internal pressure and
used for all the models so that the node numbering remains in-plane moment loads acting simultaneously.
unchanged in all models. This simplified the task of
extracting stress data from results files.
The torsional moment on the elbow was induced by 3. Linear analysis
applying tangential nodal forces at the outer edge of a
tangent pipe free end. The bending moment was induced by The FE model was generated using PATRAN [11]
graded axial forces applied on the nodes at the outer edge of software and the numerical stress analysis was performed
a tangent pipe free end. These loading techniques produced using the ASAS [12] finite element code. The stress is
consistent nominal stresses on the tangent pipe within a expressed in terms of an effective stress factor (ESF), which
distance of one pipe diameter from the loaded end. The is defined as the ratio of the maximum von Mises effective
methods of inducing torsional and in-plane bending loads by stress to the nominal maximum stress given by simple
means of nodal forces are shown in Fig. 1. To prevent rigid theory for an equivalent straight pipe. For pressure loading,
Table 1 the nominal stress is the mean diameter hoop stress, Pdm =2t;
Geometric dimensions of elbow models and for torque and in-plane moment loading, the nominal
Model b l dm =t Model b l dm =t
stress is the maximum bending stress, Mro =I: The ESFs will 4. Non-linear analysis
be referred to as stress factors in the following text.
The symbols P; T and M are used to denote non- The analysis of load coupling between internal pressure
dimensionalised pressure, torsion and bending moment and moment is complex. It is well known that especially in
loads. If PY is the pressure required to cause first yield in an thin elbows, an in-plane or torsional moment causes severe
elbow and P is the applied pressure, then P is defined as: ovalization of the pipe cross-section. In Section 3, stresses
from moment load are algebraically summed with those
P ¼ P=PY
from pressure load. In reality, if the two loads act
In the present study where the maximum stress is limited to simultaneously, internal pressure stiffens the pipe such
the yield stress, the maximum value of applied pressure is that ovalization due to the moment load is severely
PY and consequently, the maximum value of P is 1. The restricted. This results in a non-linear coupling between
other non-dimensionalised loads T and M are similarly the two loads, and consequently in a reduction of flexibility
defined. and, hence, stresses.
A general load interaction would be a situation when all The stiffening effect of pressure on combined loading is
the three loads act simultaneously on the elbow resulting in demonstrated here by comparing results from two further
a three dimensional P : T : M interaction diagram. The studies. In the first study, individual stresses that are
diagrams would then be presented on a three-orthogonal- obtained from geometrically non-linear FEA are combined
axes system with each axis representing a load. However, algebraically and in the second study, the loads are made to
only two-load interactions are studied here because of ease act simultaneously in a geometrically non-linear FEA. The
of presentation and discussion. second study shows the actual effect of load coupling. The
Figs. 2 –4 show the first yield interaction for load pairs two elbow models used in non-linear analysis are EL01 and
M : T;
T : P and M : P;
which were generated from linear EL07 because they represent ‘thin’ and ‘thick’ pipes,
superposition of individual load effects. The curves in each respectively. The limited choice of models is used merely to
diagram show the effect of d=t ratio on first yield. In M : T emphasize the presence of geometrically non-linear and
interaction, from Fig. 2(a)– (d), the pipe thickness does not coupling effects and not to determine the overall behaviour
affect first yield in a particular trend nor does it result in an of the effects.
appreciable increase or decrease of the maximum ESF. The The results show the stiffening effect of internal pressure
bend radius slightly affects the maximum ESF—the larger on in-plane moment, especially on the thinner model EL01.
the b ratio, the lower the ESF. The interaction is best A combined loading of pressure and opening moment can
represented by a circular relation, M 2 þ T 2 ¼ 1: reduce stress by as much as 51% while for a loading of
Similarly, the T : P load interactions in Fig. 3(a)– (d) pressure and closing moment the reduction is about 33%.
are not much affected by d=t and b except in short radius The thicker model EL07 shows smaller corresponding stress
bends, where (e.g. when b ¼ 2) the stresses are lower reductions of 9.3 and 3.4%.
when internal pressure dominates, giving a slightly larger From geometric non-linear analysis on individual loads,
yield locus. The interaction is best represented by a linear an in-plane closing moment causes stresses at the crown to
relation, T þ P ¼ 1: increase progressively with load but an in-plane opening
The interaction between pressure and in-plane bending moment causes a progressive reduction of the stress
moment is different from the two previous interactions. The increase. Although small, the non-linear effect in model
maximum combined stress can be lower than an individual EL01 is twice that of EL07. However, pressure loading
stress. From Fig. 4(a) – (d), the bend radius does not showed no signs of a geometric non-linear effect.
influence maximum stresses as much as the d=t ratio. When the results from both geometric non-linearity and
Different d=t ratios result in load capacities being increased load coupling analyses are examined closely, the trend
beyond their first yield individual values. In thin elbows, becomes apparent. In geometric non-linear analysis, the
where d=t $ 80; the first yield opening moment capacity is stress increase is due to the tensile stresses from the closing
increased by at least 22% when P < 0:4: In thick elbows, moment being added to the tensile pressure stresses at the
where d=t # 10; the closing moment capacity is increased outer surface of the crown. An opening moment load
by as much as 9% when P < 0:15: In addition, thick elbows produces compressive stresses at the same location, hence
ðd=t # 20Þ show a 15% increase in first yield pressure providing a reducing effect to the pressure stresses. When
capacity when an opening moment of M < 0:2 is present. the load coupling effect is taken into account, the stiffening
Taking into consideration the behaviour of the wide range of effect from internal pressure reduces ovalization and has a
elbows being studied, the best interaction is a linear relation, cancelling effect on the stress produced by geometric
M þ P ¼ 1: deformation. This effect of load coupling is large in the
In reality the load interaction behaviour is more complex thinner elbow model.
than just superposition of individual stresses. Sometimes the Fig. 5 shows first yield loci from linear and load coupling
coupling effect between the two interacting loads plays a analyses. The yield points at selected load combinations are
significant role in rendering linear analysis to be invalid. plotted to emphasise the effect of load coupling effect on first
864 A.B. Ayob et al. / International Journal of Pressure Vessels and Piping 80 (2003) 861–869
yield at various yield stresses of 200, 300 and 400 MPa. The of these closing and opening moments, the maximum
load coupling analysis results in lower maximum stresses and stresses are nearly identical and the non-linear analysis
hence the yield envelopes are larger. Elbows with higher gives slightly lower maximum stresses. The linear analysis
yield points have larger yield envelopes because the stresses first yield envelope is shown to be only slightly more
become progressively reduced as the moment is increased. conservative than the non-linear envelope.
The M : P interaction of the thicker model EL07, shown Table 2 shows the maximum increase in the first yield
in Fig. 6, does not exhibit much load coupling effect. Unlike opening moment carrying capacity of the elbows due to both
the thinner model EL01, the maximum effective stress in the non-linear and load coupling effects. The first yield closing
thicker model occurs at different elbow locations for moments ðþMÞ also show increases but not as high as
the closing and opening moments. For the same magnitude the opening moments ð2MÞ: For the thinner model EL01
A.B. Ayob et al. / International Journal of Pressure Vessels and Piping 80 (2003) 861–869 865
the largest increase in first yield moment occurs when P is interaction, i.e. the reduction of maximum stress. In the
between 0:37PY and 0:5PY ; giving a corresponding increase thinner model EL01, the torque load shows large non-
in moment capacity of between 160 and 88%. For the linear and load coupling effects. From Fig. 7(a) the non-
thicker elbow model EL07, there is no significant increase linear yield envelope of the thinner model is found to be
in the moment carrying capacity. However, the elbow very conservative compared to the thicker model
models do not exhibit any significant increase in pressure (Fig. 7(b)). In model EL01 since the effects of geometric
carrying capacity. and load coupling become increasingly large at high
The phenomenon of geometric non-linearity and load yield stress, the torque-carrying capacity of the elbow
coupling effect is similarly encountered in the P : T increases with stress limit. At all stress limits, the torque
866 A.B. Ayob et al. / International Journal of Pressure Vessels and Piping 80 (2003) 861–869
Fig. 5. Moment/pressure interaction, model EL01, linear and load coupling analyses.
Fig. 6. Moment/pressure interaction, model EL07, linear and load coupling analyses.
Table 2
Maximum increase in moment carrying capacity of elbows EL01 and EL07—load coupling analysis
Magnitude of pressure and % Increase in first yield Magnitude of pressure and % Increase in first yield
moment moment capacity moment moment capacity
Note: PY denotes the pressure to cause first yield in the respective elbows.
868 A.B. Ayob et al. / International Journal of Pressure Vessels and Piping 80 (2003) 861–869
Fig. 7. (a) and (b) Linear and non-linear Torsion/Pressure interaction (key to markers in Figs. 5 and 6 applies here).
is increased beyond TY when P # 0:5PY ; as can be seen an upper bound on deformations. Neglecting thermal
from Fig. 7(a). effects, Eq. (10) of ASME is given by:
Po range of service pressure In the BS and ASME Codes the maximum hoop stresses due
t nominal wall thickness to the individual loads are summed, regardless of their
C1 ; C2 secondary component stress indices locations on the pipe. In the present method, stresses at
Do outside pipe diameter every nodal point are determined for every load combi-
Mi section bending moment nation and subsequently the maximum stress is located.
Sm allowable stress Hence in this method the maximum stress may not be at the
Mx ; My ; Mz orthogonal moment components locations of maximum stress arising from individual loads.
In the ASME code, the interactions are similar to the finite
Eq. (10) of subsection NB-3653.1 ASME III [6] code is element method (FEM) but in the BS code, the circular T : P
meant to control primary plus secondary loads so as to place interaction is different from the FEM.
Table 3
Exact and equivalent load interaction relations
Exact relation ðpd=2t þ 0:5pÞ2 þ 4½Mt ðd þ 2tÞ=4I2 ¼ Y 2 ðpd=2t þ 0:5p þ rMi FTi =IÞ2 ¼ Y 2 ðrMi FTi =IÞ2 þ 4½Mt ðd þ 2tÞ=4I2 ¼ Y 2
Equivalent relation T 2 þ P 2 ¼ 1 M þ P ¼ 1 M 2 þ T 2 ¼ 1
Note the BS806 symbols are used in ‘Exact relation’. They are defined as: d; mean diameter; t; mean thickness; p; design pressure; Y; yield stress; Mt ;
torsional moment; Mi ; maximum in-plane moment; and FTi ; in-plane transverse stress intensification factor.
A.B. Ayob et al. / International Journal of Pressure Vessels and Piping 80 (2003) 861–869 869