Proceedings of the ASME 2013 Pressure Vessels & Piping Division Conference

PVP2013
July 14-18, 2013, Paris, France

PVP2013-97622

STRESS ANALYSIS
ANALYSIS OF PIPE
P IPE SUPPORT ATTACHMENTS:
ATTACHMENTS: A COMPARISON
COMPARISON OF
ANALYTICAL
ANALYTICAL METHODS AND FINITE ELEMENT ANALYSIS
ANALYSIS FOR CIRCULAR AND
NON-CIRCULAR ATTACHMENTS
TTACHMENTS

Anindya Bhattacharya
Technical Head,
H ead, Stress Analysis
CB&I, 40 East Bourne Terrace,
Terrace, London, W 2 6LG,
6LG,
United Kingdom.
Phone: +442070535668

ABSTRACT
4. Comparison of FEA, WRC 107, WRC297 and “Kellogg”
Despite the availability
availability of special purpose
purpose FE codes with
methods w.r.t the following parameters:
 post processing facilities as per rules of ASME SEC VIII
Division 2, use of simple analytical methods like ring loading
• Type of loading (Radial, Longitudinal, Circumferential)
applied in a “ stand-alone manner” is absence of pressure
around a circumference or more complex methods like
Welding Research council bulletins 107 and 297, will continue • D T  , t T  and d  D ratios
to be used in the industry
industry for a significant
significant period of time for for
• Combined loading including pressure
stress analysis of pipe support attachments. The reasons are
few: not all engineering companies have such custom made
• Different element types
FE codes, lack of trained personnel to work with general
NOMENCLATURE
 purpose FE codes, ease of implementation of the available
methods and their successful design history, cost and time ξ  , ϕ  - compo
componenents the ( ξ ,ϕ  ) coordinate of the main
nts in the
issues with FE analysis etc. In this paper these available shell
methods will be reviewed based on their theoretical r , R - mid surface
surface radii
radii of the branch
branch pipe
pipe and main
main shell
 background, their range of appli cability w.r.t
w.r.t the typical
t ypical design  E ,ν  - Young’s
oung’s modulus and Poisson
Poisson ratio respectiv
respectively
ely
 parameters and their comparison with FE analysis. More un - radial
radial displac
displaceme
ementnt
recent
recent analytical methods
methods based on mathematically
mathematically accurate
accurate
space curves of intersections for circular attachments will also ρ 0 - diam
diamete ratio = d 
eterr ratio
 D
 be discussed. This study will include both circular as well as T  - thickn
thickness
ess of main
main shel
shelll
non-circular attachments. This paper will highlight the φ  - Airy
Airy stre
stress
ss func
functi
tion
on
strengths and weaknesses of the conventionally used methods
- inte
interna
rnall press
pressure
ure
especially
especially with respect
respect to their mathematical
mathematical limitations
limitations to
make an analyst aware of the potential over conservatism and ρ θ  , z  - global cylin
cylindrical
drical coordin
coordinates
ates in 3D space
space
under conservatism of these analytical methods. Finite
 pξ , pϕ  - surface
surface force components
components in the directions
directions
element analysis models will be discussed in detail
specifically in relation to elements used, element parameters, w - vertica
verticall disp
displac
laceme
ement
nt
 boundary conditions and post processing.  ET 3
 H  - flexu
flexural
ral rigidi
rigidity
ty of shel
shelll =
INTRODUCTION 12( 1 − µ 2 )
 Z  - loadin
loading
g in vertica
verticall directio
directionn
In this paper, the subject matter has been structured in the
t  - thickn
thickness
ess of
of attache
attached d shell
shell
following manner:
1. Discussion of the available theoretical methods, from the  ET 
 K  - foun
founda
datio
tions
ns stif
stiffn
fnes
esss = 2
simplest to the advanced.  R
2. Brief overview of basic shell mathematical model.  x - direction
directio n of longitudinal
longitu dinal axis of
of cylinder 
cylinder 
3. Brief overview of available finite element options. S  - section
section modul
modulusus of the trunn
trunnion
ion pipe
pipe
 A - area of cross
cross section
section of trunnion
trunnion pipe
pipe

 1 Copyright © 2013 by ASME

 2. AVAILABLE THEORETICAL APPROACHES TO
- resultant
resultant applied
applied bending
bending moment
moment on trunnion
trunnion
THE PROBLEM
PROBLEM OF ANALYSING
ANALYSING A CYLINDRICAL
CYLINDRICAL
 P  - load per per unit of
of circumferenc
circumferencee (applied
(applied as a ring
SHELL
SHELL WITH CYLI
CYLINDR
NDRICAL
ICAL OR NON-CY
NON-CYLIN
LINDRIC
DRICAL
AL
load)
ATTACHMENTS.
r t  - radiu
radiuss of trunn
trunnion
ion
 F  - forc
forcee on trunn
trunnio
ion
n 2.1 Approach 1:
β 1 , β 2 - attachment
attachment parameters
parameters for rectangular
rectangular attachments
attachments = This approach
approach is popularly
popularly known as “Kellogg
“Kellogg”” method in
c1 c2 the piping industry
industry.. This approach
approach has been so named as it
& appeared
appeared for the first time in [4] and is based on ring loading
loading
 R  R
around a circular cylinder.
c1 , c2 - half dimensio
dimensions
ns of the rectang
rectangle
le along
along
circumferential and meridional directions Governing
Governing differential
differential equation [1]:
respectively For an axi-symmetric loading on a circular cylinder, the
∇ - Laplac
2
Laplacian
ian Operator, ∇ .∇ ( )
Operator, governing differential equation is the well known beam on
elastic foundation equation:
1. SHELL THEORIES:
THEORIES:
d 4w
There are various shell theories and each one has its own  H + Kw = Z  (1)
dx 4
 protagonist. Any shell theory has to be evaluated within the
d 4w ETw Z 
 postulates of Sanders-Koiter’s approach [12, 21] which can be
4
+ 2
= (2)
summarised as follows: dx HR  H 
1. The equations can be written in general tensor form. 3( 1 − µ 2 ) 1.28
2. The deformations
deformations areare described
described by six strain measures
measures,, Introducing β 4 = 2 2
, i.e. β  =
 = considering
 R T   RT 
three of which are components of the usual membrane strain
µ  = 0.3 .
tensor and the other three deviate from the components of the
geometrical
geometrical curvature
curvature change
change tensor only by terms that are we therefore get
 bilinear in the components of the curvature and membrane
strain tensor. d 4 w
3. The stresses are described by six stress measures that satisfy 4
+ 4β 4 w =  Z   (3)
dx  H 
the equations of equilibrium without approximation.
4. The theory has a principle
principle of virtual work that
that is exact for The solution of this differential equation and boundary
displacements obeying the Kirchoff hypothesis;
hypothesis; in conjunction
conjunction conditions are detailed in [1]
with approximate constitutive relations between the stress and Extending the above analysis to a case of bending of a
strain measures. Well-set boundary value problems can be cylindrical shell by a load uniformly distributed along a
formulated, and the usual minimum and reciprocal relations of circular section [1], we get:
structural mechanics hold good.  P 
Maximum Bending Moment = , where P= load per unit
5. The theory contains an exact static-geometric analogy.
analogy. This 4 β 
analogy can be formulated by replacing the static quantities by length of circumference.
corresponding geometrical quantities in homogeneous
equations of equilibrium and the resulting equations become Bending stress, σ bending  = 1.17  P 
1.5
 R
 (4)
identical with the compatibility conditions. T 
6. When applied to the symmetrical bending of shells of
revolution, the stress and strain measures agree with those
P, can be defined in terms of a local radial load,  P r  and local
generally
generally used.
used. They are consistent
consistent with those
those of the most moment,  M r  . This
This is necess
necessary
ary because
because P is a line
line load
load
simple curved beam theory.
theory. distributed around the circumference of the shell.
For the present purpose, we will discuss the issue of If a load  P r  is divided by the attachment
attachment perimeter it becomes
becomes
cylindrical pipes with circular (referred to as trunnion) as well
 P r 
as non-circular (referred to as pipe shoes) attachments. Hence for a nozzle of radius, r t  . or a moment  over section
there is no “puncture” in the header pipe. The mathematical 2π r t 
 problem of the main shell with cut-out is a boundary value  M r 
modulus of the attachment becomes, .
 problem of partial differential equation. It means that the π r t 2
cylindrical shell equation, whose general solutions have many
Flexural stresses are added to membrane longitudinal and
unknown constants, is suitable on the shell surface with or
hoop stresses to get total stress = membrane stress in direction
without cut-out. In order to determine the unknown constants
i + flexural stresses in direction i computed
computed by the expression
expression
the boundary
boundary conditions
conditions have to be used.
in eq-(4)

2 Copyright © 2013 by AS
A SM E
 compute P, these were the steps followed:
To compute
Computation of loads in longitudinal and circumferential
directions
directions by use of the following expressions:
expressions:
• longitudinal force = (longitudinal force x moment
arm)/ π r t 2
• circumferential force = (circumferential force x moment
2
arm)/ π r t 
• radial force = radial load/ 2π r t 
• equivalent
equivalent circumferentia
circumferentiall force
force = 2 x circ.
circ. force + 1.5 x
radial force
• equivalent longitudinal force = 1.5 x radial force +
longitudinal force
The above forces are used as P in eq-(4)
The reason behind the use of the factors 1.5 and 2.0 is
attributed
attributed to higher flexibilities
flexibilities in these directions.
directions.
The flexural stresses in longitudinal and circumferential
directions are then computed using these “equivalent” forces
and the membrane pressure stresses are then added to compute
the total stresses. Stress in the trunnion attachment is
 F   M 
computed as + .
 A S 
2.1.1. The case
case of Pipe shoes:
shoes:
Schematic arrangements for some pipe shoes are shown in
fig-(1). Dimension B stands
stands for shoe/gusset
shoe/gusset width,
width, G =
number of gussets, L = gusset spacing (this depends
depends on the
design
design),
), S = number
number of spines
spines,, M = spine spac
spacing
ing,, and A=
shoe length.
The approach taken for analysis of pipe shoes is similar to
that of trunnion
trunnion type attachments.
attachments. The computatio
computations
ns of section
section
 properties (few examples) are cited.
1. Pipe shoe with no gusset:
3
longitudinal moment of inertia =  A
longitudinal moment
12
distance longitudinally =  A
distance to centroid, longitudinally
2
moment of inertia, circumferential =  A
12
distance to centroid =  B
2
2. Pipe shoe
shoe with 4 gussets:
gussets:
3
longitudinal moment of inertia =  A
12
+ 5 9  BL2
distance centroid =  A
distance to centroid
2
3
moment of inertia, circumferentiall =  B
inertia, circumferentia
3
distance centroid =  B
distance to centroid
2

3 Copyright © 2013 by AS
A SM E
 Fig-(2) [31, 25] 
25] 

Bijlaard applied the radial force system, qn instead of

vertical force system, q z  . When subjected to a radial force
system the resultants not only include moments,  xb
or  yb

 but also force,  F  yb . This

This will
will be stat
statica
ically
lly equi
equiva
valen
lentt to
Fig-(1). Schem
 Schematic de
design of
of so
some com
common pipe shoe
shoe typ
types. exte
extern
rnal
al load
load Z invo
involv
lvin
ing
g force
force,,  F  zb , transverse bending

3. Saddle: moment,  xb

and longitudinal
longitudinal bending
bending moment,
moment,  yb
. The

moment of inertia, longitudinal =  A

3
+ BL
2 figure below shows the transverse bending moment,  M  xb case.
6 12
In fig-
fig-3(a
3(a),
), the linea
linearly
rly distri
distribu
buted
ted force
force sy
system,
stem, q z  is
distance to centroid =  A
2 statically
statically equivalent
equivalent to the
the transvers
transversee bending
bending moment,
moment,  xb
3 2
moment of inertia, circumferential =  B
6
+ AB 2
 but in fig-3(b), the linearly distributed force system, qn is
statically
statically equivalent
equivalent to the transverse
transverse bending
bending moment,
centroid =  B
 xb
distance
distance to centroid
2 and force,  F  as in fig-3(d).
2.2 Appro
Approach
ach 2:
2: The
The WRC-107
WRC-107 appr
approac
oach
h based
based on
the work of
of Bijlaard [34, 26, 2]
Bijlaard derived a theoretical solution based on
Timoshenko
Timoshenko equation
equationss [1, 26] for a cylindrical
cylindrical shell on end
supports under a force system, qn linearly distributed over a
ρ 0
square region defined by ξ  ≤ c , ϕ  ≤ c , where c = in
2
the developed
developed surface.
surface. In deriving the equation
equationss it has been
assumed that ε 0 = 0 (circumferential strain).
The force/moment
force/moment system
system is shown in fig-(2) below

4 Copyright © 2013 by AS
A SM E
 Fig-(3) [25] 

ϕ  ϕ  •Circumferential Moment ( taken from A.3.3.2 and Table

q z = qn cos ϕ = qn 0 cos ϕ  , q y = qn sin ϕ = qn 0 sin ϕ   (5) A-4 of WRC 107)
ϕ 0 ϕ 0
For the thin walled model, the measured circumferential
ϕ0 ϕ 0 ϕ 
 x = 2c ∫−ϕ qz Rc sin ϕ d ϕ = 2c 2 R ∫ −ϕ  qn 0 cos ϕ sin ϕ d ϕ    and longitudinal stresses were both higher than the computed
0 0 ϕ 0 values. Modifications were then done to Bijlaard's original
(6) work for both longitudinal and circumferential stresses for the
ϕ0 ϕ 0 ϕ   bending components (and for the circumferential stress, for the
 F y = 2c ∫−ϕ q yc d ϕ = 2c 2 ∫ −ϕ  qn 0 sin ϕ d ϕ   (7) membrane component also) but for longitudinal stresses there
0 0 ϕ 0
was minimal requirement for correction of the membrane
Bijlaard used Double Fourier Series to represent loads and component. Correction factor used was around 2.7 for bending
displacements. component of circumferential stress + 20-25% for membrane
component and correction factor of 2.72 was used for bending
The limitations of Bijlaard’s approach are the d   ≤ 0.3
 D component of longitudinal stress (no correction for membrane
and the
the usage
usage of Double
Double Fourier
Fourier series method which maymay not component).
component). It is stated in WRC-107 that the modified curves
curves
show converge
convergence
nce for certain boundary
boundary conditions
conditions [32, 26]. may be more conservative than the original work.
The limitation of d  is due to the use of
 D
of radial force instead
instead • Longitudinal Moment ( taken from A.3.3.2 and Table A-4
of WRC 107)
of vertical force. This results in significant error outside
For the thin walled model, the measured circumferential
aforementioned d  limit. The stresses are computed in 8 and longitudinal stresses were both higher than the computed
 D
specific locations around the intersection. The maximum values. Corrections were made to both the Membrane and
stresses need not be at these locations! Additionally stresses in Bending components. For the Longitudinal stress, no
Trunnion/Shoes cannot be computed. For rectangular correction was required for the Membrane component. Higher
attachments, the limitation is β 1 & β 2 < 0.5 . modification was required for membrane component (30%)
compared to bending component (18%) for circumferential
It is to be noted that
that WRC-107
WRC-107 is not only
only based on
stress. Correction factor used was 6.75 in bending component
Bijlaard’s theoretical work but also experimental works by
of longitudinal stress (no correction for membrane component)
Mehrson,
Mehrson, Wichman
Wichman and Hopper
Hopper [34]. The bulletin
bulletin shows a
comparison
comparison of the calculated and measured
measured stresses
stresses for both • Radial Load ( taken from A.3.3.4 and Table A-6 of WRC-
thick walled models and thin walled models. Following were 107)
the main
main issues
issues between
between the Experimental
Experimental works and Results agreed well on the transverse axis but the
Bijlaard’s work for thin shells. theoretical results were conservative by factors as high as 2.0
on the longitudinal axis.

5 Copyright © 2013 by AS
A SM E
on the longitudinal axis. Results for transverse axis agreed Based on this line of reasoning in-order to extend the
dm Dm applicability of the thin shell theoretical solution for
well for cases that are restricted by ≤2 cylindrical
cylindrical shells with cut-out,
cut-out, Xue et al
al adopted
adopted Morley’s
Morley’s
 Dm T 
equation [9], which has the same order of magnitude of
2.3
2.3 Appr
Approa
oach
ch 3: Post
Post WRC-1
WRC-107
07 appro
approac
ache
hes
s – WRC-
WRC- accuracy
accuracy as the general thin shell theory,
 R (
theory, i.e. O T  , instead )
297 and works of Morley, Simmonds and Hwang et
of Donnell’s.
Donnell’s. Morley’s
Morley’s equation
equation is expressed
expressed in complex-
complex-
al. [5, 33,
33, 11,
11, 16, 25]
valued form by Simmonds [11] as follows:
follows:
Theory
Theory of thin
thin ela
elast
stic
ic she
shells
lls,, in whi
which
ch T/R<
T/R<<1
<1 is
is
insignificant in magnitude is derived on the basis of Love-  4 2 ∂ 2  
 ∇ + ∇ − µ 2
χ  = P ( pξ , pϕ  , p )
∂ξ 2  
4 i  (11)
Kirchhoff assumptions. A generally accepted fact is this   
approach
approach has an
an error of
of order of magnitude O T  .
of magnitude
 R ( ) where,  χ  and µ  are the same as in eq-(10) and (9). The right
When a solution is derived by omitting some terms, which hand side of eq-(11) is a load function dependent on the
has order of magnitude larger than O T ( )
 R
(such as surface force components acting on the shell.
The cylindrical thin shell equations derived by
O ( T 
 R ) for shallow
shallow shell equations
equations),
), the accuracy
accuracy of the Goldenveizer, Morley, Simmonds and Timoshenko (which
was used by Bijlaard) have the same inherent error in order of
solution is bound to be lower. The detailed analysis of the
above-mentioned concept can be found in well-known ( )
magnitude O T  . The solution has the order of accuracy
 R
literature
literature and textbooks of thin shell
shell theory [6].
The “exact” equations for thin elastic cylindrical shells are ( )
O T  . WRC-297, which is based on Steele’s work on
 R
very complicated.
complicated. For a problem of cylindrical
cylindrical shell with cut- shallow shell equations covers a range of only
out
out [25],
[25], Donn
Donnelelll [8] prese
present
nted
ed an appapprox
roxim
imat
atee equ
equat
ation
ion
 R 6 ( )
r  ≤  sin π  < 0.5 .
(omitting terms of order of magnitude O ( T 
 R )
) from
For detailed analys
analysis
is of the approach taken by Xue et al
Flügge’s equation [13]. This equation is quite simple and can refer [17, 31].
 be expressed in complex-valued displacement-stress function In essence, the approach taken is to use compatibility
form (Lekkerkerk
(Lekkerkerker
er [15] and Steele
Steele [3]) as follows:
follows: conditions enforced on the geometrically correct curve of
intersection as opposed to an assumed curve of intersection
 4 ∂ 2  
 ∇ − 4 µ 2
i
∂ξ 2  
χ  = 0  (8) and using theories which are of order O T 
 R (
whic
which )
h may or
   may not involve using different
different shell theories
theories for intersecting
R cylinders.
4µ 2 = 12 (1 −ν 2 ) (9)
T  To summarize,
summarize, different
different cylindrical
cylindrical shell equations
equations are
4 µ 2 suitable to different ranges of the developed surface. Fig-(4)
 χ = un + i φ   (10)  below shows the different ranges of developed surface [25].
 ETR
where, u n , is radial displacement and φ  , Airy stress function
function..
Eq-(8) can be decomposed into two second-order partial
differential equations and is easy to solve in polar coordinate
system for the problem of cylindrical shell with cut-out [25].
However,
However, as pointed
pointed out by Koiter,
Koiter, eq-(8) can only
only be applied
for shallow
shallow shells.
shells. Koiter
Koiter [7] had
had written,
written, “It has
has been
been noted
noted
[9,14] that Donnell’s approximation is sometimes inaccurate”
and “the generalization of Donnell’s approximation is
applicable in the case of shallow shells in which the wave
length L of the deformatio
deformation n pattern
pattern on the middle surface is
always small compared with the minimum principal radius of
curvature
curvature R”. Based on fig-6.14
fig-6.14 in Donnell’s
Donnell’s book [10], the
the
applicable range of shallow shell equation for the problem of
cylindrical shell with opening is only r  ≤  sin π  = 0. 5 .
 R 6 ( )
The edge effects of general cylindrical shells and shallow
shells mathematically
mathematically differ.
differ.

6 Copyright © 2013 by AS
A SM E
 methods like FEM as currently an engineer in an industry does
not have an easy tool to compare the FE results against some
 published benchmarks for d  >0.5. In other words, as long as
 D
we do not have analytical tools which are easily
implementable and which will address the problems to be
analyzed without having significant restrictions on geometry
and loading conditions, FE analysis should be the preferred
tool for analysis. The objective of this paper is to make an
analyst aware of the potential over conservatism and under
conservatism
conservatism in the available
available and
and widely
widely used methods
methods if an
analyst is constrained to use them.

4.0 FINITE ELEMENT ANALYSIS

ANALYSIS APPROACH TO THE
PROBLEM
[33] provides an excellent discussion on the issues
involving
involving conflicts between
between shell
shell theory and finite element
element
analysis of shells. To
To briefly summarize them:

• Ill conditioning due to significantly different

different strain
energies between membrane and bending modes.

Fig-(4) [25] 
• Use of low degree polynomial trial functions in the
displacement
displacement finite element method generally
generally leads to
Donnel [10] showed that his shallow shell equations could overstiffness in the response to bending actions.
 be suitable to the range of − π 6 < ϕ  < π  6 i.e. ρ 0 < 0.5 . In • Difficulty
Difficulty in deriving trial functions for in-extensiona
in-extensionall
[15] Lekkerkerker showed that the shallow shell equations  bending.
could be applied to the range of ρ 0 ≤ 0.25 . The different
Many authors [33,[33, 18, 30] have
have recommende
recommended d use of
applicable ranges adopted by different authors are dependant hybrid elements. In this paper, however, we have used only
on different allowable intrinsic errors. displacement based finite element method.
3.0 DIFFICULTIES IN IMPLEMENTATION
IMPLEMENTATION OF Finite elements available for shell analysis can be broadly
ANALYTICAL
ANALYTICAL SOLUTIONS:
SO LUTIONS: classified into the following groups:

Analytical solutions (rather analytical solutions backed by 1. Dege

Degene
nera
rated
ted solid
solid elemen
elementsts..
2. Element
Elementss based
based on
on basic
basic shell
shell mathe
mathemati
matical
cal mode
model.
l.
experimental findings like WRC 107/297 methods) are
3. Elemen
Elements
ts based
based on
on combin
combinatioation
n of plate
plate and
and membran
membranee
extremely useful in addressing stress
stress analysis issues of pipe elements.
support attachments as they are available in almost all
For a detailed discussion on type-1, refer [19]. The main
commercial pipe stress codes and methods like “Kellogg
feature of these elements is the number and variety of adhoc
method” can easily be developed into spreadsheets. The
assumptions made to accommodate the standard procedures of
difficulty is of course
course the limited range of applicability of finite element formulation. The variation of strain through
these methods specially in relation to d  ratio and for the thickness isn’t linear. Assumptions regarding dependence of
 D determinant of Jacobian Matrix in the direction of thickness
Kellog method , its main drawback is its mathematical can lead to violation of rigid body properties
properties [19].
oversimplification of a problem, an issue which is not Type-2
Type-2 elements are usually not available in commercial
commercial
negligible
negligible when the predominant
predominant form of the loading is FE codes.
codes. They suffer from
from rigid body motion problems
problems [18].
Radial. More advanced approaches as explained in section 2.3 The element S8R is ABAQUS is however close to these
elements as discussed in [18].
of this paper have solved the problem up to d  =0.8, but To explain the meaning of the term Basic Shell Mathematical
 D
these methods are yet not available in commercial pipe stress mode, we briefly describe the derivation of the governing shell
codes or as WRC bulletins and it will be a while before they equations using the tensor approach which involves the
will be available as handy tools for engineering applications. following steps [22]:
Such methods can be used to validate numerical analysis

7 Copyright © 2013 by AS
A SM E
Fundamental assumption of the shell theory based on Love-
STRI3 - Small Strain
Strain Triangular
Triangular Element
Element with 3 nodes
nodes and
Kirchoff hypothesis and zero strain in the through thickness quadratic variation of rotation (accurate representation of plate
direction.  bending because of linear curvature variation) and analytical
Expressing the base vectors of a surface located “off middle implementation
implementation of Kirchoff
Kirchoff constraint
constraint at locations
locations (DKT or
surface” i.e. a general surface in terms of the base vectors of Discrete Kirchoff element).
the middle surface (both covariant and contra-variant STRI65 - Small Strain Triangul
Triangular
ar Element with 6 nodes
nodes and
versions). Kirchoff
Kirchoff constraint imposed numerically
numerically at points.
• Expressing the metric tensor of a surface located “off
S8R - 8-node
8-node reduced
reduced integration
integration element
element for
for small strain
strain
middle surface” in terms of the metric tensor (both formulation.
formulation. This element
element has similarity with the Basic Shell
covariant and contra-variant versions) of the middle mathematical
mathematical model as described
described in [18],
[18], although they
they are
surface. not the same, the main difference being the use of Mindlin
• Expressing
Expressing the rotation vector.
vector. hypothesis. This element is susceptible to element distortion.
• Expressing the Cristoffel symbols and permutation The Hexagonal element used is a 20-node reduced integration
tensors (Levi-Cevita tensors) of the surface located “off
“off element. The method of analysis is Linear Elastic following
the middle surface” in terms of the corresponding tensors the Elastic Stress Classification Route of [28].
of the middle surface. The issue of classification of the FE computed stresses on
• Expressing the strain tensors of a surface located “off the lines of [28] has
has been dealt with
with in numerous
numerous papers
papers and
middle surface” in terms of the strain tensor of the middle will not be repeate
repeated
d here. In a nutshell,
nutshell, local
local membrane
membrane
surface (both covariant and contra-variant versions). stresses are designated as  Pl  , primary
primary + secondary
secondary stresses
stresses as
Strain tensors are expressed as the difference between  Pl + Pb + Q and peak stresses as  Pl + Pb + Q + F  in line with
metric tensors and curvature tensors in the deformed and [28]. Primary stresses
stresses develop to maintain
maintain equilibrium
equilibrium with
un-deformed states external loads, secondary stresses to maintain compatibility of
deformation (global) and peak stresses to maintain
• Writing expression for stress and moment resultants.
compati
compatibili
bility
ty of local
local deforma
deformation
tion..  Pl  stands for local
• Using appropriate constitutive relations.  primary stress,  Pb for primary
primary bending stress, Q for
bending stress,
In the Basic Shell Mathematical model version of Finite second
secondary
ary stress
stress and  F  for peak stress. Peak stresses are
Element implementation, the interpolation of the shell significant
significant only from the standpoint
standpoint of fatigue failure.
failure.
geometry is accomplished using the Iso-parametric procedure. FE convergence theorems are in  L2 or  H 1 norms which are
Covariant and Contra-variant base vectors of the interpolated difficult to implement when the exact solution is not shown
surface are computed using the usual finite element and in this presentation no attempt has been made to evaluate
interpolation procedures and the First Fundamental form, the the convergence using these norms. For checking the
Second Fundamental form and the Christoffel symbols are convergence of an FE model percentage change in stress is
then computed from these base vectors. In the Type-2 considered
considered from a model
model with very fine mesh to gradually
gradually
elements as described in [18], the normal vector is calculated  becoming cruder. Stresses are checked at Gauss points for
normal to the interpolated middle surface, although the normal accuracy
accuracy and un-avera
un-averaged.
ged. For conver
convergence
gence,, monotonic
monotonic
vectors
vectors at the nodal points areare exactly normal to the middle  behavior is checked with a maximum
m aximum permissible variation in
surface. stress taken as 5%.The mesh size around the intersection is
For a discussion
discussion on type-3
type-3 elements
elements any standard
standard text
taken as 0.3 rt  with progressive mesh grading away from it.
 book on FEM can be referred [19].
For continuum four elements have been used through the
The FE code used for the analysis is ABAQUS ver. 6.9-1.
thickness at and close to intersections. The objective of the FE
The ABAQUS element library [20] for shells is divided into
analysis wasn’t to catch the peak stresses which are used for
three categories consisting of general-purpose, thin, and thick
shell elements. Thin shell elements provide solutions to shell fatigue evaluation, because once the  Pl + Pb + Q stresses are
 problems that are adequately described by classical computed,
computed, the fatigue stresses
stresses can easily be computed
computed using
(Kirchhoff) shell theory; thick shell elements yield solutions Fatigue Strength Reduction Factors (FSRF) [28]. The results
for structures that are best modeled by shear flexible (Mindlin) of the analysis can then be extended to compute
shell theory; and general purpose shell elements can provide  Pl + Pb + Q + F  in a straightforward manner. [27] shows that
solutions to both thin and thick shell problems. All these modeling of welds to properly simulate joint stiffness does not
elements use bending strain measures that are approximations have serious impact on the computed stresses and hence,
of those of Koiter-Sanders version of shell theory [12]. welds are not part of the models. FSRF can be avoided if
For stress analysis, the following elements from Dong’s method [28] is used. However, this requires special
ABAQUS library
library have been
been used.  post processing ability of the FE Code. If welds are modeled,

8 Copyright © 2013 by AS
A SM E
  Pl + Pb + Q can be evaluated at the weld toe directly by (even Table-1 (Contd)
(Contd)
though it is a singularity) linearization at the stress Loading Type
Radial Longitudinal Circumferential
classification
classification line
line (SCL) as explained
explained by Kalnins [29].
[29]. The Force force force

only issue with this procedure

procedure is the through thickness
thickness stress FEA shell element
8 6 3
component. (STRI3) Cylinder 

To avoid end effect, the location of the trunnion has been FEA shell element
13 4 5
(STRI3) Trunnion
taken as 5D [24] i.e. five times the Outside Diameter of the
Header Pipe with respect to the end of the header. The worst FEA shell element
9 6 3
(STRI65) Cylinder 
aspect ratio around the intersection (HEX elements) was 6.0,
average aspect ratio 2.0. One end of the header was fixed in all FEA shell element
13 4 5
six DOFs and the other end is fixed in five DOF’s. The DOF (STRI65) Trunnion

along the longitudinal axis of the header was kept free to FEA continuum
continuum
9 5 3
element Cylinder 
generate longitudinal pressure stress (for models where
 pressure was applied). Linear and full integration elements FEA continuum
continuum
12 3 6
element Trunnion
were not selected in the quadrilateral and brick versions to
avoid shear locking. References [35, 24] provide excellent
guideline on modeling of Large Diameter Cylinder
intersections. Table-2

5.0 RESUL
RESULTS
TS 36 inch header
header,, 30 inch trunnion,
trunnion, wall thickness
thickness = 9.52 mm for
 both. Magnitude of Force = 10KN, length of trunnion = 100
The stresses shown in the tables below belong to the
mm, d   = 0.84 , t   = 1
 Pb + Pl + Q category 2
category and are in N/mm . Only maximum Von  D T 
Mises equivalent stress values are shown. For continuum Radial Longitudinal Circumferential
Loading Type
elements, stresses have been Linearized using [28] as a Force force force

guideline. For tables 1-5 the applied loadings are at the end of
WRC 107 Cylinder 45 2 12
the Trunnion which makes it a Shear Force + Bending
Moment at the Shell-Nozzle interface for the Longitudinal and
WRC 107 Trunnion NA NA NA
Circumferential Force applications.. Pressure is not a part of
the loadings in Tables 1-5. For WRC-107 and WRC-297
WRC 297 Cylinder 51 5 16
computations, code FE-107 has been used.
Table-1 WRC 297 Trunnion 56 4 16

30 inch header,
header, 24 inch trunnion, wall thickness = 9.52 mm for
Kellogg Cylinder 6 2 4
 both. Magnitude of Force = 10KN, length of trunnion = 100
mm, d   = 0.8 , t   = 1 : Kellogg Trunnion 0.5 0.2 0.2
 D T 
Radial Longitudinal Circumferential FEA shell element
Loading Type 21 7 5
Force force force (S8R) Cylinder 

FEA shell element

WRC 107 Cylinder 45 3 16 15 5 4
(S8R) Trunnion

FEA shell element

WRC 107 Trunnion NA NA NA 17 6 5
(STRI3) Cylinder 

FEA shell element

WRC 297 Cylinder 50 6 22 11 4 3
(STRI3) Trunnion

FEA shell element

WRC 297 Trunnion 54 6 20 20 7 5
(STRI65) Cylinder 

FEA shell element

Kellogg Cylinder 6 3 5 14 4 3
(STRI65) Trunnion

FEA continuum
continuum
Kellogg Trunnion 0 .6 0.4 0.4 19 6 4
element Cylinder 

FEA shell element FEA continuum

continuum
10 6 3 15 5 6
(S8R) Cylinder  element Trunnion

FEA shell element

15 5 6
(S8R) Trunnion

9 Copyright © 2013 by AS
A SM E
Table 3 Table 4 (Contd.)
(Contd.)
36 inch header,
header, 12 inch trunnion,
trunnion, and wall thickness
thickness = 9.52 Radial Longitudinal Circumferential
Loading Type
Force force force
mm for header;
header; and 6.35 mm for trunnion.
trunnion. Magnitude
Magnitude of Force
WRC 297
= 10KN
10KN,, length
length of
of trunn
trunnio
ion
n =100 mm, d   = 0.34 ,
=100 mm, Trunnion
90 13 44
 D
t   = 0.67 : Kellogg Cylinder 7 4 3
T 
Radial Longitudinal Circumferential Kellogg Trunnion 1 1 1
Loading Type
Force force force

FEA shell element

WRC 107 Cylinder 48 10 31 19 10 6
(S8R) Cylinder 

FEA shell element

WRC 107 Trunnion NA NA NA 20 9 7
(S8R) Trunnion

FEA shell element

WRC 297 Cylinder 54 30 41 17 8 5
(STRI3) Cylinder 

FEA shell element

WRC 297 Trunnion 103 30 75 19 7 6
(STRI3) Trunnion

FEA shell element

Kellogg Cylinder 15 11 22 19 10 6
(STRI65) Cylinder 

FEA shell element

Kellogg Trunnion 2 2 2 20 7 7
(STRI65) Trunnion

FEA shell element FEA continuum

continuum
46 16 29 17 11 6
(S8R) Cylinder  element Cylinder 

FEA shell element FEA continuum

continuum
48 16 31 19 8 6
(S8R) Trunnion element Trunnion

FEA shell element

42 14 26
(STRI3) Cylinder 

FEA shell element

43 15 27 Table 5
(STRI3) Trunnion

FEA shell element

24 inch header,
header, 8 inch trunnion, and wall thickness = 9.52 mm
45 16 28
(STRI65) Cylinder  for header
header and 8.18
8.18 mm
mm for trunnion.
trunnion. Magnitude
Magnitude ofof Force
Force =
FEA shell element length of trunnion = 100 mm, d   = 0.36 , t   = 0.86 :
10KN, length
(STRI65) Trunnion
47 15 30  D T 
FEA continuum
continuum Radial Longitudinal Circumferential
44 13 27 Loading Type
element Cylinder  Force force force

FEA continuum
continuum WRC 107 Cylinder 47 21 53
46 14 29
element Trunnion

WRC 107 Trunnion NA NA NA

Table 4 WRC 297 Cylinder 69 31 77

24 inch header,
header, 20 inch trunnion,
trunnion, and wall thickness
thickness = 9.52 WRC 297 Trunnion 74 34 78
mm for header
header and 6.35
6.35 mm for trunnion.
trunnion. Magnitude
Magnitude of Force
Force
= 10KN
10KN,, lenlengt
gth
h of
of tru
trunn
nnio
ion
n = 100 mm, d   = 0.84 ,
100 mm, Kellogg Cylinder 16 20 40
 D
t   = 0.67 : Kellogg Trunnion 2 4 4
T 
Radial Longitudinal Circumferential FEA shell element
Loading Type 48 26 46
Force force force (S8R) Cylinder 

FEA shell element

WRC 107 Cylinder 44 5 20 43 21 43
(S8R) Trunnion

WRC 107 FEA shell element

 NA NA NA 44 22 40
Trunnion (STRI3) Cylinder 

FEA shell element

WRC 297 Cylinder 44 7 23 38 19 36
(STRI3) Trunnion

10 Copyright © 2013 by AS
A SME
Table 5 (Contd)
(Contd) Table 6 (Contd)
(Contd)
24 inch header, 8 inch trunnion, and wall thickness = 9.52 mm    )
   l
   t
  n
  a   e    l
for header
header and 8.18
8.18 mm for trunnion.
trunnion. Magnitude
Magnitude of Force =    )    i
   t   m   a
   i
  e    l
  a   n   o    l    t
  p   e   e   i
  n   e
  e   r   a   n
  y   c
  r   c   c   e    M   n   e
  r
10KN, length trunnion = 100 mm, d   = 0.36 , t   = 0.86 :
length of trunnion    T
  g
  o
   F
  r    d
  o   t
  u
  r
  o   f    l
  a
   i
   d   t
  u
  e   t
   f
 D T    n
   i    l    F   i   m
   F   u   n
  o    i   n
   t   m  n
   d   a
   i   r   g
  a   n   r   c
  a   r    i
  s   g   e
  m
  u   e
  c   m
  a    d   e   o   e   i   r   n   o   r   o
  o   a    h   L    h   C   o   o    i
Radial Longitudinal Circumferential    L    R    S   (    S   (    T    L   m    C  m
Loading Type
Force force force
FEA shell
FEA shell element element
44 22 40 47 6 10 20 106 361
(STRI65) Cylinder  (STRI65)
Cylinder 
FEA shell element
39 19 37 FEA shell
(STRI65) Trunnion
element
45 6 9 23 104 402
FEA continuum
continuum (STRI65)
46 24 44
element Cylinder  Trunnion

FEA continuum
continuum FEA
41 19 43
element Trunnion continuum
44 6 5 18 106 360
element
Cylinder 

Table 6 is to reflect the effect of applying the Forces and FEA

moments at the Shell-Nozzle Interface as opposed to at the continuum
47 7 9 20 104 398
element
end of the Trunnion
Trunnion in Tables
Tables 1-5. Pressure is not a part of the Trunnion
Loading.
Table 6 For tables, 7-9, applied load in longitudinal, circumferential
36 inch header,
header, 12 inch
inch trunnion,
trunnion, wall thickne
thicknesses
sses 9.52mm and radial directions = 10KN (applied together), pressure =
and 6.35mm
6.35mm for header
header and trunnion
trunnion respectively
respectively.. Loads 18.9Barg.
applied at shell nozzle interface, Moment=10KN-m and For the WRC-107 analysis, pressure loading has NOT been
Force=10KN. d   = 0.34 , t   = 0.67 added as a radial load at the trunnion attachment.
 D T 
Table 7
   t
   )
   l   n
  a   e    l
  e    )
   l
   i
   t   m   a
   i
   t
30 inch header, 24 inch trunnion, wall thickness =9.52 mm for
  p   e   a   n   o    l   n
  e   i   e
  e   r   a
  y
   T
  c
  r
  o   r
  n
  c   d   c   e
  r
   M
   l
  n
   i
  e
  r  both (results shown for maximum  Pb + Pl + Q in MPa)
  g    F   o   t
  u   o   f   a    d   t
  u
  e   t
   f
   F   i   m
   F   u   n
  n
   i    l
  a   r   g   r   o
   i    i   n
   t   e   m  n
  u   e
   d    i   a   n   c
  a   r   s   g   m   c   m WR C 107 C yl
yli nd
nder 258
  a    d   e   o   e   i   r   n   o   r
  o   a    h   L    h   C   o   o    i   o
   L    R    S   (    S   (    T    L   m    C  m
WRC 10
107 Tr
Trunnion NA
WRC 107
48 4 4 13 99 310
Cylinder  Kellogg Cylinder 87

WRC 107 Kellogg Trunnion 1

 NA NA NA NA NA NA
Trunnion
FEAshe
FEAshell
ll ele
eleme
ment
nt (S8
(S8R)
R) Cyl
Cylin
inde
derr 121
121
WRC 297
54 4 4 13 153 413 FEAshe
FEAshell
ll elem
elemen
entt (S8R
(S8R)) Trun
Trunni
nion
on 63
Cylinder 

WRC 297 FEAshe

FEAshell
ll elem
elemen
entt (STR
(STRI3
I3)) Cyli
Cylind
nder
er 120
120
103 6 6 19 295 752
Trunnion
FEAshe
FEAshell
ll elem
elemen
entt (ST
(STRI
RI3)
3) Trun
Trunni
nion
on 59
FEA shell
FEA
FEA shel
shelll eleme
element
nt (STRI
(STRI65
65)) Cylin
Cylinde
derr 125
125
element (S8R) 46 6 10 21 108 363
Cylinder  FEAshe
FEAshell
ll elem
elemen
entt (ST
(STRI
RI65
65)) Tru
Trunn
nnio
ion
n 66
FEA shell FEAconti
FEAcontinuu
nuum
m elem
element
ent (Shel
(Shell)
l) Cylin
Cylinde
derr 126
126
element (S8R) 48 7 9 23 105 401
Trunnion FEA
FEA cont
contin
inuu
uum
m elem
elemen
entt Trun
Trunni
nion
on 70

FEA shell
element
46 6 9 21 106 359
(STRI3)
Cylinder 

FEA shell
element
46 7 8 24 103 403
(STRI3)
Trunnion

11 Copyright © 2013 by AS
A SME
Table 8
Table 10
36 inch header, 30 inch trunnion, wall thickness =9.52 mm for
36” pipe, wall thickness = 9.52 mm. Shoe design corresponds
 both (results shown for maximum  Pb + Pl + Q in MPa)
to 3-gusset,
3-gusset, A=450,
A=450, B=500, shoe plate thickness
thickness = 10 mm, L =
WRC 107 Cyli nd
nder 307 350 mm (refer fig-1); magnitude of load = 40KN. Pressure is
WRC 10
107 Tr
Trunnion NA
not applied. β 1 = 0.56 , β 2 = 0.49
Kell og
ogg Cy
Cyl in
inder 100 Radial Longitudinal Circumferential
Loading Type
Force Force Force
Kellogg Trunni on
on 0.8
WRC 107 Cylinder 181 18 61
FEAshe
FEAshell
ll ele
eleme
ment
nt (S8R
(S8R)) Cyli
Cylind
nder
er 146
146
Kellogg Cylinder 30 18 40
FEAshe
FEAshell
ll elem
elemen
entt (S8R
(S8R)) Trun
Trunni
nion
on 75
Ke llogg Shoe 2 9 3
FEAshe
FEAshell
ll elem
elemen
entt (STR
(STRI3
I3)) Cyli
Cylind
nder
er 143
143
FEA Shell element
element (S8R)
FEAshe
FEAshell
ll elem
elemen
entt (ST
(STRI
RI3)
3) Trun
Trunni
nion
on 76 75 20 36
Cylinder 

FEA
FEA shel
shelll elem
elemen
entt (STRI
(STRI65
65)) Cylind
Cylinder
er 145
145 FEA Shell ele ment ( S8R) Shoe 77 30 32

FEA
FEA shel
shelll ele
eleme
ment
nt (STR
(STRI6
I65)
5) Trun
Trunni
nion
on 76 FEA Shell element (STRI3)
(STRI3)
82 18 35
Cylinder 
FEAconti
FEAcontinuu
nuum
m elem
element
ent (She
(Shell
ll)) Cylin
Cylinde
derr 148
148
FEA Shell element (STRI3)
(STRI3)
FEAcon
FEAconti
tinu
nuum
um elem
elemen
entt Trun
Trunni
nion
on 74 82 18 35
Shoe

FEA Shell element (STRI65)

(STRI65)
Table 9 Cylinder 
75 22 37

36 inch header, 12 inch trunnion, wall thickness = 9.52 mm for  FEA Shell element (STRI65)
(STRI65)
80 30 33
Shoe
header and 6.35 mm for Trunnion (results shown for
maximum  Pb + Pl + Q in Mpa) FEA Continuum
Continuum element
element
81 24 33
Cylinder 
WRC 107 Cyli nd
nder 321 FEA Continuum element Shoe 78 35 32
WRC 10
107 Tr
Trunnion NA

Kell og
ogg Cy
Cyl in
inder 100
Table 11
Kellogg Trunni on
on 0.8 30” Pipe, wall thickness 9.52 mm, Shoe design corresponds to
FEAshe
FEAshell
ll ele
eleme
ment
nt (S8R
(S8R)) Cyli
Cylind
nder
er 157
157 3 Gusset, A=450, B=500, Shoe plate thickness=10 mm, L=350
mm (refer fig-1), Magnitude of load=40KN. Pressure is not
FEAshe
FEAshell
ll ele
eleme
ment
nt (S8
(S8R)
R) Tru
Trunn
nnio
ion
n 108
108
applied. β 1 = 0.67 , β 2 = 0.60
FEAshe
FEAshell
ll elem
elemen
entt (STR
(STRI3
I3)) Cyli
Cylind
nder
er 155
155
Radial Longitudinal Circumferential
FEA
FEA shel
shelll elem
elemen
entt (STR
(STRI3
I3)) Trunn
Trunnio
ion
n 102
102 Loading Type
Force Force Force
FEA
FEA shel
shelll elem
elemen
entt (STRI
(STRI65
65)) Cylind
Cylinder
er 159
159 WRC 107 Cylinder 173 21 63
FEA
FEA shel
shelll ele
eleme
ment
nt (STR
(STRI6
I65)
5) Trun
Trunni
nion
on 98 Kellogg Cylinder 27 17 38
FEAconti
FEAcontinuu
nuum
m elem
element
ent (She
(Shell
ll)) Cylin
Cylinde
derr 154
154 Ke llogg Shoe 2 9 3
FEAcon
FEAconti
tinu
nuum
um ele
eleme
ment
nt Tru
Trunn
nnio
ion
n 103
103 FEA Shell element
element (S8R)
60 15 18
Cylinder 

Results
Results for Pipe
Pipe Shoes: (Stres
(Stresses
ses at locations
locations of FEA Shell ele ment ( S8R) Shoe 80 22 12
singularities
singularities have not been considered)
considered) FEA Shell element (STRI3)
(STRI3)
62 14 18
Cylinder 
 Note: WRC-107 method has been used even though in most
FEA Shell element (STRI3)
(STRI3)
cases β 1 , β 2 are above the allowable limit. So far Pipe Shoes Shoe
75 22 13

are concerned, the typically used dimensions render them

FEA Shell element (STRI65)
(STRI65)
unsuitable for use of WRC-107. Despite this fact, the author in 60 14 20
Cylinder 
his experience has seen its usage for computation of local FEA Shell element (STRI65)
(STRI65)
stresses
stresses at Shoe Attachments
Attachments and its use
use is mostly
mostly due to 82 22 14
Shoe
availability
availability of this module in in most common
common pipe stress
stress FEA Continuum
Continuum element
element
63 18 25
 programmes. For the WRC 107 computation of Pipe Shoes, Cylinder 
the geometry
geometry of the attachment
attachment has been conside
considered
red as FEA Continuum element Shoe 82 20 16
Rectangular solid. Pipe Stress Program CAESAR II Version
5.2 has been used for this purpose. For Tables 10, 11 and 12,
2c1=500 mm and 2c2=450 mm.

12 Copyright © 2013 by AS
A SME
Table 12 Table 14 (results shown for maximum  Pb + Pl + Q in MPa)
24” pipe, wall thickness = 9.52 mm. Shoe design corresponds 30” header, wall thickness = 9.52 mm
to 3-gusset, A=450, B=500, shoe plate thickness = 10 mm, L =
WR C 107 Cyl in
inder 312
350 mm (refer fig-1); magnitude of load = 40KN. Pressure is
not applied. β 1 = 0.84 , β 2 = 0.75 WRC 107 Shoe NA

Kel lo
logg C yl
ylinder 161
Radial Longitudinal Circumferential
Loading Type
Force Force Force Kellogg Shoe 12

WRC 107 Cylinder 174 23 65 FEAshe

FEAshell
ll ele
eleme
ment
nt (S8R
(S8R)) Cyli
Cylind
nder
er 126
126

Kellogg Cylinder 24 15 34 FEAshe

FEAshell
ll elem
elemen
entt (S8
(S8R)
R) Shoe
Shoe 115
115

Ke llogg Shoe 2 9 3 FEA

FEA shel
shelll elem
elemen
entt (STR
(STRI3
I3)) Cyli
Cylind
nder
er 131
131

FEA Shell element

element (S8R) FEAshe
FEAshell
ll elem
elemen
entt (ST
(STRI
RI3)
3) Shoe
Shoe 118
118
35 12 22
Cylinder 
FEAshel
FEAshelll eleme
element
nt (STRI
(STRI65
65)) Cylin
Cylinde
derr 128
128
FEA Shell element ( S8R) Shoe 73 22 10
FEAshe
FEAshell
ll elem
elemen
entt (ST
(STRI
RI65
65)) Sho
Shoee 113
113
FEA Shell element (STRI3)
(STRI3)
34 12 22
Cylinder  FEA
FEA conti
continuu
nuum
m ele
eleme
ment
nt (Sh
(Shell
ell)) Cylin
Cylinde
derr 132
132

FEA Shell element (STRI3)

(STRI3) FEA
FEA cont
contiinuum
nuum elem
elemen
entt Shoe
Shoe 119
119
50 22 10
Shoe

FEA Shell element (STRI65)

(STRI65)
35 13 22 Table 15 (results shown for maximum  Pb + Pl + Q in MPa)
Cylinder 

FEA Shell element (STRI65)

(STRI65) 24” header, wall thickness = 9.52 mm
53 17 10
Shoe
WR C 107 Cyl in
inder 298
FEA Continuum
Continuum element
39 15 24 WRC 107 Shoe NA
Cylinder 

FEA Continuum ele ment Shoe 57 21 13 Kel lo

logg C yl
ylinder 136

Kellogg Shoe 12
For Tables 13-15, applied load in longitudinal, circumferential FEAshe
FEAshell
ll elem
elemen
entt (S8R
(S8R)) Cyli
Cylind
nder
er 80
and radial directions = 40KN( applied together), pressure =
FE A s he
he ll
ll el
ele me
me nt
nt (S
(S 8R
8R ) S ho
hoe 85
18.9 barg.
barg. Pressure has been applied
applied but not as radial thrust
thrust
load. FEA
FEA shel
shelll ele
eleme
ment
nt (STR
(STRI3
I3)) Cyl
Cylin
inde
derr 84

F EA
EA s he
he ll
ll el
el em
em een
n t ( ST
ST RI
RI 3)
3) Sh
Sh oe
oe 89
Table 13 (results shown for maximum Pb + Pl + Q in MPa)
FEAshe
FEAshell
ll elem
elemen
entt (ST
(STRI
RI65
65)) Cyl
Cylin
inde
derr 82
36” header, wall thickness = 9.52 mm FEAshe
FEAshell
ll elem
elemen
entt (STR
(STRI6
I65)
5) Shoe
Shoe 83
WR C 107 C yl
ylinder 330 FEAcon
FEAconti
tinu
nuum
um elem
elemen
entt (She
(Shell
ll)) Cylin
Cylinde
derr 88
WRC 107 Shoe NA F EA
EA c o nt
nt in
in uu
uu m e le
le me
me nt
nt Sh
Sh oe
oe 92
K el
ellogg Cy
Cyli nd
nder 186

Ke llogg Shoe 12

FEAshe
FEAshell
ll ele
eleme
ment
nt (S8
(S8R)
R) Cyl
Cylin
inde
derr 180
180

FEAshe
FEAshell
ll elem
elemen
entt (S8
(S8R)
R) Shoe
Shoe 155
155

FEA
FEA shel
shelll elem
elemen
entt (STR
(STRI3
I3)) Cyli
Cylind
nder
er 184
184

FEAshe
FEAshell
ll elem
elemen
entt (ST
(STRI
RI3)
3) Shoe
Shoe 156
156

FEAshel
FEAshelll eleme
element
nt (STRI
(STRI65
65)) Cylin
Cylinde
derr 182
182

FEAshe
FEAshell
ll elem
elemen
entt (ST
(STRI
RI65
65)) Sho
Shoee 153
153

FEAconti
FEAcontinu
nuum
um elem
elemen
entt (She
(Shell
ll)) Cylin
Cylinde
derr 188
188

FEAcon
FEAconti
tinu
nuum
um elem
elemen
entt Shoe
Shoe 159
159

13 Copyright © 2013 by AS
A SME
6.0 DISCUSSION OF RESULTS AND SCOPE FOR the pipe shoe can be seen as analogous to this parameter.
FUTURE WORK Significant differences exist for the Circumferential loading
case also. For both Trunnion
Trunnion and Pipe Shoes, for some casescases ,
Tables 1,2,4
1,2,4 show that WRC 107 and WRC 297 results results
specially
specially for the Radial Load scenario,scenario, stresses
stresses in the
show significant differences with respect to FE results for the
shoes/Trun
shoes/Trunnion
nion elements
elements exceed
exceed stresses
stresses in the cylinder
cylinder
radial load case. This is because of the high d  ratio and which clearly shows the risk of using
using the Kellogg
Kellogg method for
 D
radial as opposed to vertical load representation of the same in computing stresses in the Pipe support attachments. A point to
WRC 107 as explained in section 2.2 of this paper. Tables 3, 5 note is that, the method of computing stresses in the Pipe
and 6 show that the results are comparable (even for the supports
supports cannot
cannot be technicall
technically
y stated as “Kellogg
“Kellogg Method”
Method” as
[4] only discusses computation of local stresses in the
Radial load case) indicating the criticality of the d  factor in
case) indicating Cylinder. The context of using the term “ Kellogg method” for
 D
WRC 107/297
107/297 approac
approaches.
hes. For
For the Kellogg
Kellogg Method, the the method of computing stresses in attachments is due the
significant difference is for the radial load case. This is fact that this computation based
based on elementary beam theory is
 because of t he basis of the method being axi-symmetri cal ring an essent
essential
ial feature
feature ( in author’s
author’s experience
experience)) of the
loading which significantly deviates from the actual spreadsheets
spreadsheets which use the Kellogg method to compute Local Local
mathem
mathematic
atical
al model
model in the radial
radial load
load situat
situation.
ion. The
The Kellogg
Kellogg stresses in the Cylinder. Hence the caution is using elementary
method also underestimates the stresses in the Trunnion. This  beam theory analysis for computation of local stresses in
is due to the use of simple
simple beam theory
theory as opposed
opposed to shell Attachments. Significant differences in results have not been
theory
theory and the non-co non-consi
nsidera
deration
tion of the compat
compatibil
ibility
ity seen in Finite element approach using different element types.
requirement between the header pipe and the Trunnion in this This however should not be taken as a blanket statement as the
method. Kellogg
Kellogg method also in most most (but not
not all) cases
cases models had proper
proper mesh
mesh grading
grading with
with adequate
adequatelyly small
 predicts lower magnitude of stresses in the Longitudinal and element size and the element distortion control was well
Circumferential Force applications. However the allowable within the recommended limits of the FE code. For improper
stresses
stresses in the
the Kellogg
Kellogg method as as long as
as they are specified
specified as mesh grading, element size, significantly distorted elements
the [28] allowable
allowable for local
local primary stress, the
the error will not in and improper integration methods, significant differences in
general make the analysis non-conservative except for the results can be seen between the elements, especially for
Radial Load scenario. For Tables 7, 8 and 9 which are for the Triangular elements which suffer from geometric anisotropy.
combined load scenario, WRC 107 results show significantly The stress
stress analyst
analyst should
should carefully
carefully study the theory
theory manual of
the FE code which he/she should be using with respect to
higher magnitudes of  Pb + Pl + Q with respect to FEA. Even
applicability, element distortion and integration rules.
though the Pressure loading has not been modeled as a Radial
Radial
loading for these Tables, which would have resulted in even The present analysis has to be extended for different load
higher magnitudes of  Pb + Pl + Q if the direction of this load combinations
combinations with varying
varying magnitudes
magnitudes of the individual load
vectors to quantify the degree of over or under conservatism
would have been in the same direction as the additive radial
of the available analytical methods. The present analysis
load, but the simplistic way of computing pressure stresses
mainly focuses
focuses on the stand alone
alone effect of individual
individual load
load
also (as in Tables 7, 8 and 9) induces higher stresses in the
vectors (although
(although Tables
Tables 7-9 as well as Tables
Tables 13-15
13-15 does
WRC 107 type type of analysis.
analysis. Pressure
Pressure induced
induced loading at at a
address combinations but more tests need to be done with
cylinder
cylinder to cylinder
cylinder interface with or without
without other external
external
varying magnitudes of the individual load vectors) . Effect of
loadings is complicated and WRC 107 analysis which
variance
variance in mesh grading and element size should be checked
checked
considers the loading on the cylindrical surface as a
to assist an analyst in selection of the “best element” for these
rectangular
rectangular loading cannot
cannot predict
predict the stresses
stresses correctly
correctly and
applications
applications,, if an analyst
analyst so desires.
desires. In the present scope
scope of
will err on the conservative side for most cases. WRC 107
work, the use of proper mesh grading, element size and
/297 analysis has shown lower magnitudes of Stress for Shear
integration rules have ironed out significant differences
Forces and Torsion
Torsion moments (Table
(Table 6 where the loadings have
 between the individual elements. Hence, the take away
 been applied at the Shell-Nozzle Interface) with respect to
message for an analyst with respect to individual element
FEA. However, these loadings, in general are not the
types is, as long as mesh grading , element size, distortion
govern
governing
ing loads
loads in piping
piping applica
application
tions.
s. When
When using
using WRC-
control and integration rules are properly used, there are no
107/297 modules of a Pipe Stress Program, an analyst should
 preferred element s , although
alt hough the analyst should carefully read
review the program document to see how pressure is modeled
the Theory manual of the FE code which he/she intends to use.
in these modules.
7.0 CONCLUSIONS:
Results in tables 10, 11 and 12 again show that the pattern
of variance in results between FEA and WRC 107/297 is most 1. Use of a particular shell theory requires an understanding of
of
significant with respect to radial loads. The reason can still be the order of magnitude
magnitude of error inherent in that theory
theory and its
attributed to the d  even though in case of the shoe
 D
dimension
dimension “d” is strictly not applicable
applicable but the dimension
dimension of

14 Copyright © 2013 by AS
A SME
applicability
applicability vis-à-vis
vis-à-vis the problem to be analyzed
analyzed specially
specially 107/297 based analysis is also not correct as the branch is not
with reference to d  and  D ratios.  pressurized for Pipe support applications where by “branch”
 D T  we mean the Trunnion.
2. Shell theories should be evaluated on the basis of Sander- 8. Finite elements for shell analysis have different approaches
Koiter postulates.  based on the theoretical considerations that form the basis of
3. The use of an axi-symmetric loading model (which in this their developments, with elements based on basic shell
 paper has been referenced to as “Kellogg method”) has been mathematical model being least popular because of the
historically
historically the most
most popular
popular method for analyzin
analyzing
g both  problem of addressing rigid body motion. Commercial FE
cylindrical and non-cylindrical attachments. codes should be having Hybrid elements in the element library
for shell applications.
applications.
4. WRC-107 method which is based on Timoshenko equations
9. Degenerated solid elements have used adhoc assumptions
has
has the
the same
same errerror (
or O T 
 R ) as Morley, Simmonds and on shell theory to work within the constraint
constraint of finite element
element
Goldenveize
Goldenveizerr equations.
equations. WRC-107
WRC-107 results may be more or formulation. Assumptions regarding the mathematical form of
less conservative than FE results. Results are
are generally overly dependence of the determinant of the Jacobian Matrix on the
thickness direction coordinate can lead to violation of rigid
conservative for d   > 0.5 . The analysis results show that in
 D  body properties.
some cases but not all (generally computation as per Kellogg
10. Not much difference has been found in results using 8-
method has shown lower magnitude of stresses with respect to
node reduced integration shell element developed on the line
WRC 107 or FE analys analysis),
is), Kellogg
Kellogg method significantly
significantly
of Mindlin hypothesis,
hypothesis, triangular
triangular elements
elements based on discrete
discrete
underestimates the stresses in Trunnion
Trunnion and Pipe Shoes.
Shoes. Hence
Kirchoff constraints (imposed analytically or numerically) and
it is recommended
recommended that this method
method should not
not be used and
and
use of solid elements for circular attachments. Stresses at
hence should not be used for evaluating stresses in Pipe
locations
locations of singularity
singularity have to be carefully addresse
addressed
d [29].
supports.
supports. A point to note is, the method as at appears
appears in [4]
The pattern of results i.e. relative invariance with respect to
addresses
addresses only the local stresses
stresses at the cylinder
cylinder,, so the
element types need not be always correct depending on the
evaluation of stresses in attachments cannot technically be
addressed
addressed as “Kellogg
“Kellogg method”,
method”, rather calculation
calculation based
based on  D ratio, element distortion, element size and use of
T 
elementary
elementary beam theory
theory. It is against this later
later which, the alternate numerical integration rules. In general, as long as
author in his
his experience
experience has seenseen as widely
widely used in the thin shell theory is valid and reduced integration rule is used
Industry as part of of the spreadshe
spreadsheets
ets based
based on “Kellog
“Kellogg g for shear flexible elements, with proper mesh grading and
method” is what this caution is directed at. keeping the element size at the intersection region
5. If an analyst
analyst is constrained
constrained to use Kellogg
Kellogg method for for significantly less than rt  , type of element is usually not a
analysis of local stresses on the pipe at support locations, the significant parameter. Stress Analyst should carefully review
allowable stress should not be exceeded beyond the allowable the Technical Manual(s) of FE Code for the capabilities and
for local primary stresses
stresses as per [28]. limitations of the available elements from the element library.
6. WRC-297 method is based on shallow shell theory and the
11. Analytical methods with d  as high as 1.0 with ease of
order of magnitude in error is due to omission of some terms  D
implementation
implementation is required
required not only because
because the available
available
which are of the order O ( T 
 R ) and has shown overly methods like WRC107/297 etc are inadequate for such
conservative
conservative behavio
behavior,r, specially
specially for the Trunnion
Trunnion stresses
stresses for applications
applications but also as a tool to properly benchmark
benchmark the FE
most of the cases analyzed. Use of WRC 297 for Pipe support results. Till such time, FE models will continue to be
attachments is not recommended.  benchmarked against WRC 107 type of analysis for similar
loading within the limits of the applicable geometry.
geometry.
7. When comparing results between an analytical and FE
approach, it is best to check the model on a component by 12. Additional tests need to be done for Pipe Shoes for varying
compon
componentent basis
basis i.e. the model
model isis loade
loadedd with
with only
only one effects of  D and combined
combined loadings.
loadings. In author’s
author’s opinion
opinion it
T 
force/moment
force/moment component
component in the absence
absence of pressure.
pressure. This
is futile to expect usability of WRC-107 for shoe attachments,
check will show
show stresses
stresses because
because of which components
components are
as based on typical dimensions of Pipe Shoes, these geometric
over/under represented in the final results. Since WRC-
 parameters will in most cases be not satisf ied.
107/297
107/297 does not not have a provision
provision for checking
checking pressur
pressuree
loadin
loading,
g, simulati
simulating
ng the same by a modified
modified radial
radial load
load (=
(= 13.
13. WRC 107 /297 analysis has shown lower magnitudes of
applied
applied radial load
load + pressure
pressure times area) or superpo
superposing
sing the Stress for Shear Forces and
and Torsion
Torsion moments (Table
(Table 6 where
results
results with the usual membrane
membrane stresses
stresses in the header
header pipe the loadings have been applied at the Shell-Nozzle Interface)
due to to pressur
pressuree genera
generally
lly makes
makes the anal analys
ysis
is over-
over- with respect to FEA. However, these loadings, in general are
conservative
conservative.. The modification
modification of the radial load in a WRC- not the governing factors in piping applications.
applications.

15 Copyright © 2013 by AS
A SME
 Shells Subjected to External Forces on Nozzle, ASME
Journal of Pressure Vessel Technology, Vol. 128.
ACKNOWLEDGMENTS
[17] Ming-De
Ming-De Xue, Xue, Qing-
Qing-Hai
Hai Du,
Du, Keh-Ch
Keh-Chih ih Hwang
Hwang,, Zhi-
The author wishes to acknowledge Professor M.D.Xue of Hai Xiang, 2010, An analytical method for cylindrical
Tsinghua University, Department of Engineering Mechanics shells with nozzles due to internal pressure and external
for providing some valuable suggestions and document load
loadss – Part
Part I• Theoretical foundation; ASME Journal
references and for answeri ng some questions on her paper. The of Pressure Vessel Technology Vol 132
author also wants to thank Dr.Subrata Saha of Reliance [18] Bathe,
Bathe, K.J.,Le
K.J.,Lee,e, P.S.,
P.S., 2005,
2005, Insig
Insight
ht into
into finite
finite elemen
elementt
Industries Ltd India, Mr. Suraj Kunder of Costain UK and ex- shell discretizations
discretizations by use of the the ‘‘basic
‘‘basic shell
colleague and friend Mr.Arijit Chatterjee for providing mathematical model’’ Computers and Structures Vol. Vol. 83
valuable guideline and suggestions. [19] Mac Neal
Neal,, R.H.,
R.H., Finite
Finite Elemen
Elements ts their
their desig
designn and
 performance, Marcel-Dekker, NY. NY.
REFERENCES
[20]
[20] Theo
Theoryry Manu
Manual al ABAQUS
ABAQUS,, a produ product ct of Dass
Dassau ault
lt
[1] Timos
Timoshen
henko,ko, S:
S: 1940,
1940, Theory
Theory of plateplatess and shel
shells,
ls, MC-
MC- Systems, SA.
Graw Hill, New York  York  [21] Mollman
Mollman,, M, 1981,1981, Introdu
Introductio
ctionn to the Theory
Theory of Thin
[2] Bijlaard,
Bijlaard, P.PP.P.,
., 1955,
1955, “Stresse
“Stressess from
from Radial
Radial loads
loads and Shells, Wiley-Blackwell.
External moments in Cylindrical pressure vessels”, [22] Flugge
Flugge,, W.,
W., Tensor
ensor analy
analysis
sis and
and Continu
Continuum um Mechan
Mechanics ics,,
Welding Journal, M iami, FL, US Vol. Vol. 34 Springer, Berlin.
[3] Steele
Steele,, C.R and
and Steele,
Steele, M.L.M
M.L.M 19831983,, “Stres
“Stresss analys
analysisis of [23] Chen.,
Chen., M.W.L.
M.W.L.Y Y., Li, J.G.,
J.G., 2000,
2000, A twotwo step
step approac
approach h of
 Nozzle in Cylindrical vessels with external Load, Stress Classification and Primary Structure Method.
ASME Journal of Pressure Vessel Technology, Vol. 105 Trans
Trans ASME Vol. Vol. 122
[4] M.W.Kel
M.W.Kellog loggg Company
Company,20,201111,“D
,“Desig
esignn of Piping [24] Xue,L.,
Xue,L.,Wid
Widera
era,, G.E.O
G.E.O.,Sa
.,Sang,
ng, Zhifu.,
Zhifu., Feb. Feb. 2006,
2006,
Systems” Flexibi
Flexibility
lity Factors
Factors forfor Branch
Branch Pipe Conne Connectio
ctions
ns
[5] Mehrso
Mehrson, n, J.L,
J.L, Mokhtari
Mokhtarian,an, K.,K., Ranjan,
Ranjan, G.V G.V and and subjected to in-plane and out of plane moments ,Journal
Rodabaugh, E.C., 1984, “Local Stresses in Cylindrical of Pressur e Vessel Technology
Technology
Shells due to External Loadings on Nozzle Supplement [25]
[25] Priv
Privat
atee comcommu muni
nica
cati
tion
on,, ProProfe
fess
ssor
or M.D.M.D.Xu Xuee of
to WRC bulletin 107”WRC Bulletin No. 297 Tsinghua
Tsinghua University,
University, People’s Republic of China.
[6] Golden
Goldenveiveizer
zer A.L.,
A.L., 1961,
1961, Theo
Theory ry of Elastic
Elastic Thin shells,
shells, [26]
[26] Bhat
Bhattac
tacha
hary
ryaa A, 201 2012,2, A CompCompariariso
sonn of of Simp
Simplele
Pergamon, Oxford, Analytical Methods for evaluating local Stresses at Pipe
[7] Koiter
Koiter,, W.T
W.T,, 1959.
1959. “A Consist
Consistent
ent First
First Approx
Approximaimation
tion Supports with Finite Element Analysis results
in the General Theory of Elastic Shells”, Proceedings of  NAFEMS UK Conference 2012.
the Symposium on the Theory of Thin Elastic Shells, [27]
[27] Bhat
Bhattac
tacha
hary
ryaa A, 2011
2011,, A Finite
Finite Elem
Elemen entt based
based Stud
Study y on
Delft, the Netherlands, W.T.Koiter ed., North-Holland, Stress Intensification Factors (SIF) for reinforced
Amsterdam. Fabricated Tees NAFEMS World Congress Boston
[8] Donnel
Donnell, l, L.H.,
L.H., 1933
1933,, “Stabi
“Stability
lity of Thin
Thin Walled
alled Tube
Tubess 2011
under Torsion”,
Torsion”, NACA Report No. 479 [28] ASME Boiler and Pressure
Pressure Vessel code, Sec VIII Div2 Div2
[9] Morley
Morley,, L.S.D.,
L.S.D., 1959,
1959, “An Impro
Improvem vementent on Donne
Donnell’ ll’ss (2007 Edition), ASME Publication
Approximations for Thin walled Circular Cylinders”, [29] Kalnin
Kalnins,s, A, 2008,
2008, Stres
Stresss Class
Classific
ificatio
ation
n Lines
Lines Straig
Straightht
Q.J.Mech., Appl. M ath., Vol. Vol. 12 through Singularities ASME PVP Conference, Chicago,
[10] Donnel
Donnell, l, L.H,
L.H, 1976,
1976, Beams
Beams,, Plates
Plates and and Shells,
Shells, Mc-Mc- ILL, USA
Graw -Hill, New York, York, Chapter 6 [30] Bathe,
Bathe, K.J.,
K.J., Chap
Chapelle
elle,, D., 2010,
2010, FinitFinitee Elemen
Elementt
[11] Simmonds
Simmonds , J.G.,J.G., 1966,
1966, A Set of Accurate
Accurate Equation
Equationss forfor Analysis of Shells Fundamentals, Springer.
Circular Cylindrical Elastic Shells, “,Int. J. Solids [31]
[31] Xue,
Xue, M.D.
M.D.,, Li, D.F
D.F.,
., Hwa
Hwang ng,, K.C,
K.C, 200
2005, 5, A Thin
Thin She
Shellll
Structure., Vol. 2, Theoretical Solution for Two Intersecting Cylindrical
[12] Budian
Budianskysky,B
,B and Sand
Sanders
ers , J.L, 1963,
1963, “ On the the best
best first
first Shells Due to External Branch Pipe Moments ASME
First Order Linear Shell Theory”, Progress in Applied Journal of PVT Vol. 127
Mechan
Mechanics ics (The Prager Prager Anniver
Anniversarysary Volume),
olume), [32] Wong, F.M.G,
.M.G, 1984,
1984, Stresses
Stresses and Flexibili
Flexibilities
ties for
for
Macmillan, London. Pressure Vessel Attachments. Dissertation thesis for
[13] Flugge,
Flugge, W.1967,
W.1967, Stresses
Stresses in Shells,
Shells, Springer,
Springer, Berlin. Master of Science in Nuclear Engineering.
[14] Hoff, N.J., 1955, The accuracy
accuracy of of Donnell’
Donnell’ss equatio
equations ns [33] Morley
Morley,, L.S.D, Morris,
Morris, A.J.1
A.J.1978
978,, Conflict
Conflict betwee
between n
J.Appl. Mech Vol. 22 Finite Elements and Shell Theory, Paper presented at
[15] Lekkerkerker
Lekkerkerker,, J.G.,
J.G., 1972.
1972. The
The Determina
Determinationtion of Elastic
Elastic Second World Congress on Finite Element Methods,
Stresses near Cylinder-to-Cylinder Intersection. Nuclear Bournemouth, UK.
Eng. Des., Volume 20. [34] Wichman,
Wichman, K.R., Hooper Hooper,, A.G.,
A.G., Mehrson,
Mehrson, J.L., Stresses
Stresses
[16] Xue,
Xue, M.D., Du, Du, Q.H.,
Q.H., Li, D.F
D.F. and Hwang
Hwang,, K.C.,
K.C., 2006.
2006. in Spherical Shells due to External Loadings, Welding
Theoretical Stress Analysis of Intersecting Cylindrical Research Council Bulletin No. 107.

16 Copyright © 2013 by AS
A SME
[35] Koves,
Koves, W.,
W., Mokhtarian
Mokhtarian,, K., Rodabaugh,
Rodabaugh, E., Widera,
Widera, W.,
2006, Large Diameter Shell Intersections, Welding
Research Council Bulletin No. 497.

17 Copyright © 2013 by AS
A SME