Sei sulla pagina 1di 17

Proceedings of the ASME 2013 Pressure Vessels & Piping Division Conference

PVP2013

July 14-18, 2013, Paris, France

PVP2013-97622

STRESS ANALYSIS OF PIPE SUPPORT ATTACHMENTS: A COMPARISON OF ANALYTICAL METHODS AND FINITE ELEMENT ANALYSIS FOR CIRCULAR AND NON -CIRCUL AR ATTACHMENTS

Anindya Bhattacharya Te chnical Head , Stress Analysis CB&I , 40 East Bourne Te rrace, London, W 2 6LG, United Kingdom. Phone: +442070535668

ABSTRACT

Despite the availability of special purpose FE codes with post processing facilities as per rules of ASME SEC VIII Division 2, use of simple analytical methods like ring loading around a circumference or more complex methods like Welding Research council bulletins 107 and 297, will continue to be used in the industry for a significant period of time for stress analysis of pipe support attachments. The reasons are few: not all engineering companies have such custom made FE codes, lack of trained personnel to work with general purpose FE codes, ease of implementation of the available methods and their successful design history, cost and time issues with FE analysis etc. In this paper these available methods will be reviewed based on their theoretical background, their range of applicability w.r.t the typical design parameters and their comparison with FE analysis. More recent analytical methods based on mathematically accurate space curves of intersections for circular attachments will also be discussed. This study will include both circular as well as non-circular attachments. This paper will highlight the strengths and weaknesses of the conventionally used methods especially with respect to their mathematical limitations to make an analyst aware of the potential over conservatism and under conservatism of these analytical methods. Finite element analysis models will be discussed in detail specifically in relation to elements used, element parameters, boundary conditions and post processing.

INTRODUCTION

In this paper, the subject matter has been structured in the following manner:

1. Discussion of the available theoretical methods, from the simplest to the advanced.

2. Brief overview of basic shell mathematical model.

3. Brief overview of available finite element options.

4. Comparison of FEA, WRC 107, WRC297 and “Kellogg”

methods w.r.t the following parameters:

Type of loading (Radial, Longitudinal, Circumferential) applied in a “ stand-alone manner” is absence of pressure

in a “ stand-alone manner” is absence of pressure D T , t T and d
in a “ stand-alone manner” is absence of pressure D T , t T and d
in a “ stand-alone manner” is absence of pressure D T , t T and d

D T , t T and d D ratios

Combined loading including pressure

Different element types

NOMENCLATURE

x

, j

r,R

E,n

-

shell - mid surface radii of the branch pipe and main shell - Young’s modulus and Poisson ratio respectively

- radial displacement

components in the (x ,j )

coordinate of the main

u n

r 0

T - thickness of main shell

of the main u n r 0 T - thickness of main shell - diameter ratio

- diameter ratio = d D

f

p

r,q ,z

- Airy stress function

- internal pressure - global cylindrical coordinates in 3D space

p

x

, p

j - surface force components in the

in 3D space p x , p j - surface force components in the w -

w

- vertical displacement

H

-

flexural rigidity of shell =

ET

3

12 1

(

- m

2

)

directions

Z - loading in vertical direction t - thickness of attached shell

K

- foundations stiffness =

ET

 

R

2

x

- direction of longitudinal axis of cylinder

S

- section modulus of the trunnion pipe

A

- area of cross section of trunnion pipe

b

1

,

M

P - load per unit of circumference (applied as a ring

- resultant applied bending moment on trunnion

load)

r

t

F

b

2

- radius of trunnion

- force on trunnion

- attachment parameters for rectangular attachments =

c , c

1

2

2

-

c R &

1

c

2

R

dimensions

half

circumferential

respectively

of

and

the

meridional

rectangle

along

directions

- Laplacian Operator, .( )

1. SHELL THEORIES:

There are various shell theories and each one has its own

protagonist. Any shell theory has to be evaluated within the postulates of Sanders-Koiter’s approach [12, 21] which can be summarised as follows:

1. The equations can be written in general tensor form.

2. The deformations are described by six strain measures,

three of which are components of the usual membrane strain tensor and the other three deviate from the components of the

geometrical curvature change tensor only by terms that are bilinear in the components of the curvature and membrane strain tensor.

3. The stresses are described by six stress measures that satisfy

the equations of equilibrium without approximation.

4. The theory has a principle of virtual work that is exact for

displacements obeying the Kirchoff hypothesis; in conjunction with approximate constitutive relations between the stress and strain measures. Well-set boundary value problems can be formulated, and the usual minimum and reciprocal relations of structural mechanics hold good.

5. The theory contains an exact static-geometric analogy. This

analogy can be formulated by replacing the static quantities by corresponding geometrical quantities in homogeneous equations of equilibrium and the resulting equations become

identical with the compatibility conditions.

6. When applied to the symmetrical bending of shells of

revolution, the stress and strain measures agree with those generally used. They are consistent with those of the most simple curved beam theory. For the present purpose, we will discuss the issue of cylindrical pipes with circular (referred to as trunnion) as well as non-circular (referred to as pipe shoes) attachments. Hence there is no “puncture” in the header pipe. The mathematical problem of the main shell with cut-out is a boundary value problem of partial differential equation. It means that the cylindrical shell equation, whose general solutions have many unknown constants, is suitable on the shell surface with or without cut-out. In order to determine the unknown constants

the boundary conditions have to be used.

2. AVAI LABLE THEORETICAL APP RO ACH ES TO

THE PROBLEM OF ANALYSING A CYLINDRICAL SHELL WITH CYLINDRICAL OR NON-CYLINDRICAL ATTACHMENTS.

2.1 Approach 1:

This approach is popularly known as “Kellogg” method in the piping industry. This approach has been so named as it appeared for the first time in [4] and is based on ring loading around a circular cylinder.

Governing differential equation [1]:

For an axi-symmetric loading on a circular cylinder, the governing differential equation is the well known beam on elastic foundation equation:

H

d

4

w

dx

4

+

Kw

=

Z

d

4

w

ETw

Z

dx

4

HR

2

H

+

=

2

)

Introducing

b

4

m = 0.3.

we therefore get

4

d

w +

dx

4

4b

4

w

=

3(1

=

Z

H

-

m

2

R T

2

, i.e.

b =

1.28

RT
RT

considering

The solution of this differential equation and boundary conditions are detailed in [1] Extending the above analysis to a case of bending of a cylindrical shell by a load uniformly distributed along a circular section [1], we get:

Maximum Bending Moment

 

P

=

, where P= load per unit

 

4b

length of circumference.

Bending stress,

s

bending

=

1 17 . P R 1 5 . T
1 17
.
P
R
1 5
.
T

(1)

(2)

(3)

(4)

P, can be defined in terms of a local radial load,

moment,

distributed around the circumference of the shell.

If a load

P

r

and local

M

r . This is necessary because P is a line load

P

r

is divided by the attachment perimeter it becomes

P

r

2p r

t

for a nozzle of radius,

r

t

. or a moment over section

modulus of the attachment becomes,

M

r

p r

t

2

.

Flexural stresses are added to membrane longitudinal and hoop stresses to get total stress = membrane stress in direction i + flexural stresses in direction i computed by the expression in eq-(4)

To compute P, these were the steps followed:

Computation of loads in longitudinal and circumferential directions by use of the following expressions:

longitudinal force = (longitudinal force x moment

arm)/

circumferential force = (circumferential force x moment

arm)/

radial force = radial load/

equivalent circumferential force = 2 x circ. force + 1.5 x radial force equivalent longitudinal force = 1.5 x radial force + longitudinal force

pr

t

pr

t

2

2

2pr

t

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 to higher flexibilities in these 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

computed as

F +

M

.

A S

2.1.1. The case of Pipe shoes:

Schematic arrangements for some pipe shoes are shown in fig-(1). Dimension B stands for shoe/gusset width, G = number of gussets, L = gusset spacing (this depends on the design), S = number of spines, M = spine spacing, and A= shoe length.

The approach taken for analysis of pipe shoes is similar to that of trunnion type attachments. The computations of section properties (few examples) are cited.

1. Pipe shoe with no gusset:

longitudinal moment of inertia =

3 12 A 2 A 12
3
12
A
2
A
12

A

distance to centroid, longitudinally =

moment of inertia, circumferential =

distance to centroid =

B 2
B
2

2. Pipe shoe with 4 gussets:

longitudinal moment of inertia =

3 A 5 2 + BL 12 9
3
A
5
2
+
BL
12
9

distance to centroid =

A 2
A
2

moment of inertia, circumferential =

3 3
3
3

B

distance to centroid =

B 2
B
2
to centroid = A 2 moment of inertia, circumferential = 3 3 B distance to centroid
Fig-(1). Schematic design of some common pipe shoe types. 3. Saddle: moment of inertia, longitudinal

Fig-(1). Schematic design of some common pipe shoe types.

3. Saddle:

moment of inertia, longitudinal =

3 6
3
6

A

+ BL

2 12
2
12

distance to centroid =

A 2
A
2

moment of inertia, circumferential =

distance to centroid =

B 2
B
2
3 6
3
6

B

+ AB

2 2
2
2

2.2 Approach 2: The WRC-107 approach based on the work of Bijlaard [34, 26, 2]

Bijlaard derived a theoretical solution based on Timoshenko equations [1, 26] for a cylindrical shell on end

n linearly distributed over a

in

square region defined by

supports under a force system,

x

£

q

c ,

j

£

c , where

r 0 2
r
0
2

c =

the developed surface. In deriving the equations it has been

assumed that

The force/moment system is shown in fig-(2) below

e

0

= 0 (circumferential strain).

in fig-(2) below e 0 = 0 (circumferential strain). Fig-(2) [31, 25] Bijlaard applied the radial

Fig-(2) [31, 25]

Bijlaard applied the radial force system,

vertical force system,

system the resultants not only include moments,

but also force,

external load Z involving force,

moment,

figure below shows the transverse bending moment,

In fig-3(a), the linearly distributed force system,

statically equivalent to the transverse bending moment,

but in fig-3(b), the linearly distributed force system,

statically equivalent to the transverse bending moment,

and force,

or

q

instead of

n

q . When subjected to a radial force

z

M

xb

M

yb

. This will be statically equivalent to

F

zb

,

transverse

M

yb

M xb

bending

. The

case.

is

q

z

M xb

q

n

is

M xb

F yb

M

xb

and longitudinal bending moment,

F

y

as in fig-3(d).

q z M (6) F y x = q cos n j = q n

q

z

M

(6)

F

y

x

=

q cos

n

j

=

q

n 0

j

j

0

cos

j

,

=

2

c

j

0

Ú

-

j

=

2

c

j

0

Ú

-

j

0

0

q Rc sin

z

j

qcd

y

j =

2

d

2

c

j

=

j

0

Ú

-

j

0

q

y

2

c

=

2 R

q

n

j

0

Ú

-

j

sin

j

0

q

n 0

q

n 0

j

j

0

sin

j

=

j

j

d

0

j

Fig-(3) [25]

q

n 0

j

j

0

cos

j

sin

sin

j

j

d

(5)

j

(7)

Bijlaard used Double Fourier Series to represent loads and displacements.

The limitations of Bijlaard’s approach are the

d
d

D £

0 3

.

and the usage of Double Fourier series method which may not show convergence for certain boundary conditions [32, 26].

show convergence for certain boundary conditions [32, 26]. The limitation of d D is due to

The limitation of d D is due to the use of radial force instead

of vertical force. This results in significant error outside

of vertical force. This results in significant error outside aforementioned d D limit. The stresses are

aforementioned d D limit. The stresses are computed in 8

specific locations around the intersection. The maximum stresses need not be at these locations! Additionally stresses in Trunnion/Shoes cannot be computed. For rectangular

attachments, the limitation is

It is to be noted that WRC-107 is not only based on Bijlaard’s theoretical work but also experimental works by Mehrson, Wichman and Hopper [34]. The bulletin shows a comparison of the calculated and measured stresses for both thick walled models and thin walled models. Following were the main issues between the Experimental works and Bijlaard’s work for thin shells.

b

1

&

b

2

< 0.5 .

Circumferential Moment ( taken from A.3.3.2 and Table

A-4 of WRC 107)

For the thin walled model, the measured circumferential

and longitudinal stresses were both higher than the computed

values. Modifications were then done to Bijlaard's original work for both longitudinal and circumferential stresses for the

bending components (and for the circumferential stress, for the

membrane component also) but for longitudinal stresses there

was minimal requirement for correction of the membrane component. Correction factor used was around 2.7 for bending component of circumferential stress + 20-25% for membrane

component and correction factor of 2.72 was used for bending

component of longitudinal stress (no correction for membrane component). It is stated in WRC-107 that the modified curves may be more conservative than the original work.

of WRC 107) For the thin walled model, the measured circumferential and longitudinal stresses were both higher than the computed values. Corrections were made to both the Membrane and Bending components. For the Longitudinal stress, no correction was required for the Membrane component. Higher

modification was required for membrane component (30%)

compared to bending component (18%) for circumferential stress. Correction factor used was 6.75 in bending component of longitudinal stress (no correction for membrane component)

Radial Load ( taken from A.3.3.4 and Table A-6 of WRC-

Longitudinal Moment ( taken from A.3.3.2 and Table A-4

107)

Results agreed well on the transverse axis but the theoretical results were conservative by factors as high as 2.0 on the longitudinal axis.

on the longitudinal axis. Results for transverse axis agreed

well for cases that are restricted by

d

m

D

m

D m T
D
m
T

£ 2

2.3 Approach 3: Post WRC-107 approaches – WRC- 297 and works of Morley, Simmonds and Hwang et al. [5, 33, 11, 16, 25]

Theory of thin elastic shells, in which T/R<<1 is insignificant in magnitude is derived on the basis of Love- Kirchhoff assumptions. A generally accepted fact is this

approach has an error of order of magnitude

When a solution is derived by omitting some terms, which

has order of magnitude larger than

O

( T ) R
(
T
)
R

.

O ( T ) (such as R
O
(
T
)
(such as
R

O

( T ) R
(
T
)
R

for shallow shell equations), the accuracy of the

solution is bound to be lower. The detailed analysis of the above-mentioned concept can be found in well-known literature and textbooks of thin shell theory [6]. The “exact” equations for thin elastic cylindrical shells are very complicated. For a problem of cylindrical shell with cut- out [25], Donnell [8] presented an approximate equation

(omitting terms of order of magnitude

from

Flügge’s equation [13]. This equation is quite simple and can be expressed in complex-valued displacement-stress function form (Lekkerkerker [15] and Steele [3]) as follows:

O

form (Lekkerkerker [15] and Steele [3]) as follows: O ) 2 ˆ Ê 4 2 —

)

2 ˆ Ê 4 2 — - 4m i c = 0 (8) Á ˜
2
ˆ
Ê 4
2
— -
4m i
c
= 0
(8)
Á
˜
2
x
Ë
¯
R
2
4 m
=
12 1
(
2
-n
)
(9)
T
2
4
m
c =
u
+
f
(10)
n
i ETR
where,
u
n , is radial displacement and f , Airy stress 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, as pointed out by Koiter, eq-(8) can only be applied for shallow shells. Koiter [7] had written, “It has been 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 deformation pattern on the middle surface is always small compared with the minimum principal radius of curvature R”. Based on fig-6.14 in Donnell’s book [10], the applicable range of shallow shell equation for the problem of

cylindrical shell with opening is only

The edge effects of general cylindrical shells and shallow shells mathematically differ.

r R
r
R
( p ) 6
(
p
)
6

£ sin

= 0 5

.

.

Based on this line of reasoning in-order to extend the applicability of the thin shell theoretical solution for

cylindrical shells with cut-out, Xue et al adopted Morley’s

equation [9], which has the same order of magnitude of

accuracy as the general thin shell theory, i.e.

, instead

of Donnell’s. Morley’s equation is expressed in complex- valued form by Simmonds [11] as follows:

O

( T ) R
(
T
)
R

Ê 4

Á

Ë

— + — -

2

4 m

2

i

2

x

2

ˆ

˜

¯

c

=

P

(

p

x

, p

j

, p

)

(11)

where, c and m are the same as in eq-(10) and (9). The right hand side of eq-(11) is a load function dependent on the

surface force components acting on the shell. The cylindrical thin shell equations derived by Goldenveizer, Morley, Simmonds and Timoshenko (which was used by Bijlaard) have the same inherent error in order of

( T magnitude O ) . The solution has the order of accuracy R (
(
T
magnitude
O
)
. The solution has the order of accuracy
R
(
T
O
)
. WRC-297, which is based on Steele’s work on
R
shallow
shell equations covers a range of only
r
p
£ sin
(
)
<
0 5
.
.
R
6

For detailed analysis of the approach taken by Xue et al refer [17, 31]. In essence, the approach taken is to use compatibility conditions enforced on the geometrically correct curve of intersection as opposed to an assumed curve of intersection

and using theories which are of order

which may or

may not involve using different shell theories for intersecting cylinders. To summarize, different cylindrical shell equations are suitable to different ranges of the developed surface. Fig-(4) below shows the different ranges of developed surface [25].

O

( T ) R
(
T
)
R
Fig-(4) [25] Donnel [10] showed that his shallow shell equations could be suitable to the

Fig-(4) [25]

Donnel [10] showed that his shallow shell equations could

be suitable to the range of

[15] Lekkerkerker showed that the shallow shell equations

could be applied to the range of

applicable ranges adopted by different authors are dependant on different allowable intrinsic errors.

3.0 DIFFICULTIES IN IMPLEMENTATION OF ANALYTICAL SOLUTIONS:

Analytical solutions (rather analytical solutions backed by experimental findings like WRC 107/297 methods) are extremely useful in addressing stress analysis issues of pipe support attachments as they are available in almost all commercial pipe stress codes and methods like “Kellogg method” can easily be developed into spreadsheets. The difficulty is of course the limited range of applicability of

0 £ 0.25 . The different

0 < 0.5 . In

-

p 6
p
6

< j <

r

p 6
p
6

i.e.

r

0 < 0 . 5 . In - p 6 < j < r p 6

these methods specially in relation to d D ratio and for the

Kellog method , its main drawback is its mathematical oversimplification of a problem, an issue which is not negligible when the predominant form of the loading is Radial. More advanced approaches as explained in section 2.3

Radial. More advanced approaches as explained in section 2.3 of this paper have solved the problem

of this paper have solved the problem up to d D =0.8, but

these methods are yet not available in commercial pipe stress codes or as WRC bulletins and it will be a while before they will be available as handy tools for engineering applications. Such methods can be used to validate numerical analysis

methods like FEM as currently an engineer in an industry does not have an easy tool to compare the FE results against some

not have an easy tool to compare the FE results against some published benchmarks for d

published benchmarks for d D >0.5. In other words, as long as

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 in the available and widely used methods if an analyst is constrained to use them.

4.0 FINITE ELEMENT ANALYSIS APPRO ACH TO THE PROBLEM

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

Ill conditioning due to significantly different strain

energies between membrane and bending modes. Use of low degree polynomial trial functions in the displacement finite element method generally leads to overstiffness in the response to bending actions.

Difficulty in deriving trial functions for in-extensional bending.

Many authors [33, 18, 30] have recommended use of hybrid elements. In this paper, however, we have used only displacement based finite element method. Finite elements available for shell analysis can be broadly classified into the following groups:

1. Degenerated solid elements.

2. Elements based on basic shell mathematical model.

3. Elements based on combination of plate and membrane elements.

For a detailed discussion on type-1, refer [19]. The main feature of these elements is the number and variety of adhoc assumptions made to accommodate the standard procedures of finite element formulation. The variation of strain through thickness isn’t linear. Assumptions regarding dependence of determinant of Jacobian Matrix in the direction of thickness can lead to violation of rigid body properties [19]. Type-2 elements are usually not available in commercial FE codes. They suffer from rigid body motion problems [18]. The element S8R is ABAQUS is however close to these elements as discussed in [18]. To explain the meaning of the term Basic Shell Mathematical mode, we briefly describe the derivation of the governing shell equations using the tensor approach which involves the following steps [22]:

Fundamental assumption of the shell theory based on Love- Kirchoff hypothesis and zero strain in the through thickness direction. Expressing the base vectors of a surface located “off middle surface” i.e. a general surface in terms of the base vectors of the middle surface (both covariant and contra-variant versions).

Expressing the metric tensor of a surface located “off

middle surface” in terms of the metric tensor (both covariant and contra-variant versions) of the middle surface. Expressing the rotation vector.

Expressing the Cristoffel symbols and permutation

tensors (Levi-Cevita tensors) of the surface located “off the middle surface” in terms of the corresponding tensors of the middle surface. Expressing the strain tensors of a surface located “off

middle surface” in terms of the strain tensor of the middle surface (both covariant and contra-variant versions). Strain tensors are expressed as the difference between metric tensors and curvature tensors in the deformed and un-deformed states Writing expression for stress and moment resultants.

Using appropriate constitutive relations.

In the Basic Shell Mathematical model version of Finite Element implementation, the interpolation of the shell geometry is accomplished using the Iso-parametric procedure. Covariant and Contra-variant base vectors of the interpolated surface are computed using the usual finite element interpolation procedures and the First Fundamental form, the Second Fundamental form and the Christoffel symbols are then computed from these base vectors. In the Type-2 elements as described in [18], the normal vector is calculated normal to the interpolated middle surface, although the normal vectors at the nodal points are exactly normal to the middle surface. For a discussion on type-3 elements any standard text book on FEM can be referred [19]. The FE code used for the analysis is ABAQUS ver. 6.9-1. The ABAQUS element library [20] for shells is divided into three categories consisting of general-purpose, thin, and thick shell elements. Thin shell elements provide solutions to shell problems that are adequately described by classical (Kirchhoff) shell theory; thick shell elements yield solutions for structures that are best modeled by shear flexible (Mindlin) shell theory; and general purpose shell elements can provide solutions to both thin and thick shell problems. All these elements use bending strain measures that are approximations of those of Koiter-Sanders version of shell theory [12]. For stress analysis, the following elements from ABAQUS library have been used.

STRI3 - Small Strain Triangular Element with 3 nodes and quadratic variation of rotation (accurate representation of plate bending because of linear curvature variation) and analytical implementation of Kirchoff constraint at locations (DKT or Discrete Kirchoff element).

STRI65 - Small Strain Triangular Element with 6 nodes and Kirchoff constraint imposed numerically at points.

S8R- 8-node reduced integration element for small strain formulation. This element has similarity with the Basic Shell mathematical model as described in [18], although they are not the same, the main difference being the use of Mindlin hypothesis. This element is susceptible to element distortion.

The Hexagonal element used is a 20-node reduced integration element. The method of analysis is Linear Elastic following the Elastic Stress Classification Route of [28].

The issue of classification of the FE computed stresses on the lines of [28] has been dealt with in numerous papers and will not be repeated here. In a nutshell, local membrane stresses are designated as Pl , primary + secondary stresses as Pl + Pb + Q and peak stresses as Pl + Pb + Q + F in line with [28]. Primary stresses develop to maintain equilibrium with external loads, secondary stresses to maintain compatibility of deformation (global) and peak stresses to maintain compatibility of local deformation. Pl stands for local primary stress, Pb for primary bending stress, Q for secondary stress and F for peak stress. Peak stresses are significant only from the standpoint of fatigue failure.

FE convergence theorems are in

H 1 norms which are

difficult to implement when the exact solution is not shown and in this presentation no attempt has been made to evaluate the convergence using these norms. For checking the convergence of an FE model percentage change in stress is considered from a model with very fine mesh to gradually becoming cruder. Stresses are checked at Gauss points for accuracy and un-averaged. For convergence, monotonic behavior is checked with a maximum permissible variation in stress taken as 5%.The mesh size around the intersection is

L

2

or

taken as 5%.The mesh size around the intersection is L 2 or taken as 0 .

taken as 0.3 rt with progressive mesh grading away from it. For continuum four elements have been used through the thickness at and close to intersections. The objective of the FE analysis wasn’t to catch the peak stresses which are used for fatigue evaluation, because once the Pl + Pb + Q stresses are computed, the fatigue stresses can easily be computed using Fatigue Strength Reduction Factors (FSRF) [28]. The results of the analysis can then be extended to compute Pl + Pb + Q + F in a straightforward manner. [27] shows that modeling of welds to properly simulate joint stiffness does not have serious impact on the computed stresses and hence, welds are not part of the models. FSRF can be avoided if Dong’s method [28] is used. However, this requires special post processing ability of the FE Code. If welds are modeled,

Pl + Pb + Q can be evaluated at the weld toe directly by (even though it is a singularity) linearization at the stress classification line (SCL) as explained by Kalnins [29]. The only issue with this procedure is the through thickness stress component. To avoid end effect, the location of the trunnion has been 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 aspect ratio around the intersection (HEX elements) was 6.0, average aspect ratio 2.0. One end of the header was fixed in all six DOFs and the other end is fixed in five DOF’s. The DOF along the longitudinal axis of the header was kept free to generate longitudinal pressure stress (for models where pressure was applied). Linear and full integration elements 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.

5.0 RESULTS

The stresses shown in the tables below belong to the Pb + Pl + Q category and are in N/mm 2 . Only maximum Von

Mises equivalent stress values are shown. For continuum elements, stresses have been Linearized using [28] as a guideline. For tables 1-5 the applied loadings are at the end of the Trunnion which makes it a Shear Force + Bending Moment at the Shell-Nozzle interface for the Longitudinal and

Pressure is not a part of

Circumferential Force applications

the loadings in Tables 1-5. For WRC-107 and WRC-297

computations, code FE-107 has been used.

Table-1

30 inch header, 24 inch trunnion, wall thickness = 9.52 mm for both. Magnitude of Force = 10KN, length of trunnion = 100

mm,

d
d

D =

.

0 8

,

of Force = 10KN, length of trunnion = 100 mm, d D = . 0 8

t T =

1

:

Loading Type

Radial

Longitudinal

Circumferential

Force

force

force

WRC 107 Cylinder

45

3

16

WRC 107 Trunnion

NA

NA

NA

WRC 297 Cylinder

50

6

22

WRC 297 Trunnion

54

6

20

Kellogg Cylinder

6

3

5

Kellogg Trunnion

0.6

0.4

0.4

FEA shell element (S8R) Cylinder

10

6

3

FEA shell element (S8R) Trunnion

15

5

6

Table-1 (Contd)

Loading Type

Radial

Longitudinal

Circumferential

Force

force

force

FEA shell element (STRI3) Cylinder

8

6

3

FEA shell element (STRI3) Trunnion

13

4

5

FEA shell element (STRI65) Cylinder

9

6

3

FEA shell element (STRI65) Trunnion

13

4

5

FEA continuum

     

element Cylinder

9

5

3

FEA continuum

     

element Trunnion

12

3

6

Table-2

36 inch header, 30 inch trunnion, wall thickness = 9.52 mm for both. Magnitude of Force = 10KN, length of trunnion = 100

mm,

d
d

D =

0 84

.

,

Force = 10KN, length of trunnion = 100 mm, d D = 0 84 . ,

t T =

1

Loading Type

Radial

Longitudinal

Circumferential

Force

force

force

WRC 107 Cylinder

45

2

12

WRC 107 Trunnion

NA

NA

NA

WRC 297 Cylinder

51

5

16

WRC 297 Trunnion

56

4

16

Kellogg Cylinder

6

2

4

Kellogg Trunnion

0.5

0.2

0.2

FEA shell element (S8R) Cylinder

21

7

5

FEA shell element

     

(S8R) Trunnion

15

5

4

FEA shell element (STRI3) Cylinder

17

6

5

FEA shell element

     

(STRI3) Trunnion

11

4

3

FEA shell element

     

(STRI65) Cylinder

20

7

5

FEA shell element

     

(STRI65) Trunnion

14

4

3

FEA continuum

     

element Cylinder

19

6

4

FEA continuum

     

element Trunnion

15

5

6

Table 3

Table 4 (Contd.)

36 inch header, 12 inch trunnion, and wall thickness = 9.52

mm for header; and 6.35 mm for trunnion. Magnitude of Force

= 10KN,

,

length

of

trunnion

=100

Magnitude of Force = 10KN, , length of trunnion =100 mm, = 0 34 . T

mm,

=

0 34

.

of Force = 10KN, , length of trunnion =100 mm, = 0 34 . T =

T =

0 67

.

:

Loading Type

Radial

Longitudinal

Circumferential

Force

force

force

WRC 107 Cylinder

48

10

31

WRC 107 Trunnion

NA

NA

NA

WRC 297 Cylinder

54

30

41

WRC 297 Trunnion

103

30

75

Kellogg Cylinder

15

11

22

Kellogg Trunnion

2

2

2

FEA shell element (S8R) Cylinder

46

16

29

FEA shell element (S8R) Trunnion

48

16

31

FEA shell element (STRI3) Cylinder

42

14

26

FEA shell element (STRI3) Trunnion

43

15

27

FEA shell element (STRI65) Cylinder

45

16

28

FEA shell element (STRI65) Trunnion

47

15

30

FEA continuum

     

element Cylinder

44

13

27

FEA continuum

     

element Trunnion

46

14

29

Table 4

24 inch header, 20 inch trunnion, and wall thickness = 9.52

mm for header and 6.35 mm for trunnion. Magnitude of Force

,

= 10KN,

length

of

trunnion

100

Magnitude of Force , = 10KN, length of trunnion 100 = mm, = 0 84 .

=

mm,

= 0 84

.

of Force , = 10KN, length of trunnion 100 = mm, = 0 84 . T

T =

0 67

.

:

Loading Type

Radial

Longitudinal

Circumferential

Force

force

force

WRC 107 Cylinder

44

5

20

WRC 107

     

Trunnion

NA

NA

NA

WRC 297 Cylinder

44

7

23

Loading Type

Radial

Longitudinal

Circumferential

Force

force

force

WRC 297

     

Trunnion

90

13

44

Kellogg Cylinder

7

4

3

Kellogg Trunnion

1

1

1

FEA shell element

     

(S8R) Cylinder

19

10

6

FEA shell element (S8R) Trunnion

20

9

7

FEA shell element

     

(STRI3) Cylinder

17

8

5

FEA shell element

     

(STRI3) Trunnion

19

7

6

FEA shell element

     

(STRI65) Cylinder

19

10

6

FEA shell element

     

(STRI65) Trunnion

20

7

7

FEA continuum

     

element Cylinder

17

11

6

FEA continuum

     

element Trunnion

19

8

6

Table 5

24 inch header, 8 inch trunnion, and wall thickness = 9.52 mm

for header and 8.18 mm for trunnion. Magnitude of Force =

10KN, length of trunnion = 100 mm,

d
d

D =

0 36

.

,

t
t

T =

0 86

.

:

Loading Type

Radial

Longitudinal

Circumferential

Force

force

force

WRC 107 Cylinder

47

21

53

WRC 107 Trunnion

NA

NA

NA

WRC 297 Cylinder

69

31

77

WRC 297 Trunnion

74

34

78

Kellogg Cylinder

16

20

40

Kellogg Trunnion

2

4

4

FEA shell element (S8R) Cylinder

48

26

46

FEA shell element

     

(S8R) Trunnion

43

21

43

FEA shell element (STRI3) Cylinder

44

22

40

FEA shell element

     

(STRI3) Trunnion

38

19

36

Table 5 (Contd)

24 inch header, 8 inch trunnion, and wall thickness = 9.52 mm

for header and 8.18 mm for trunnion. Magnitude of Force =

10KN, length of trunnion = 100 mm,

d
d

D =

0 36

.

,

t
t

T =

0 86

.

:

Loading Type

Radial

Longitudinal

Circumferential

Force

force

force

FEA shell element (STRI65) Cylinder

44

22

40

FEA shell element (STRI65) Trunnion

39

19

37

FEA continuum

     

element Cylinder

46

24

44

FEA continuum

     

element Trunnion

41

19

43

Table 6 is to reflect the effect of applying the Forces and moments at the Shell-Nozzle Interface as opposed to at the end of the Trunnion in Tables 1-5. Pressure is not a part of the Loading.

Table 6

36 inch header, 12 inch trunnion, wall thicknesses 9.52mm

and 6.35mm for header and trunnion respectively. Loads

applied at shell nozzle interface, Moment=10KN-m and

Force=10KN.

T =

d
d

D =

0 34

.

,

t
t

0 67

.

Loading Type

Radial Force

Shear Force

(Longitudinal)

Shear Force

(Circumferential)

Torsional Moment

Longitudinal

moment

Circumferential

moment

WRC 107

           

Cylinder

48

4

 

4

13

99

310

WRC 107

           

Trunnion

NA

NA

NA

NA

NA

NA

WRC 297

           

Cylinder

54

4

 

4

13

153

413

WRC 297

           

Trunnion

103

 

6

 

6

19

295

752

FEA shell

           

element (S8R)

46

 

6

10

21

108

363

Cylinder

 

FEA shell

           

element (S8R)

48

7

 

9

23

105

401

Trunnion

 

FEA shell

           

element

(STRI3)

46

 

6

 

9

21

106

359

Cylinder

   

FEA shell

           

element

(STRI3)

46

7

 

8

24

103

403

Trunnion

 

Table 6 (Contd)

Loading Type

Radial Force

Shear Force

(Longitudinal)

Shear Force

(Circumferential)

Torsional Moment

Longitudinal

moment

Circumferential

moment

FEA shell

           

element

(STRI65)

47

 

6

10

20

106

361

Cylinder

 

FEA shell

           

element

(STRI65)

45

 

6

 

9

23

104

402

Trunnion

   

FEA

           

continuum

element

44

 

6

 

5

18

106

360

Cylinder

   

FEA

           

continuum

element

47

7

 

9

20

104

398

Trunnion

 

For tables, 7-9, applied load in longitudinal, circumferential and radial directions = 10KN (applied together), pressure =

18.9Barg.

For the WRC-107 analysis, pressure loading has NOT been

added as a radial load at the trunnion attachment.

Table 7

30 inch header, 24 inch trunnion, wall thickness =9.52 mm for both (results shown for maximum Pb + Pl + Q in MPa)

WRC 107 Cylinder

258

WRC 107 Trunnion

NA

Kellogg Cylinder

87

Kellogg Trunnion

1

FEA shell element (S8R) Cylinder

121

FEA shell element (S8R) Trunnion

63

FEA shell element (STRI3) Cylinder

120

FEA shell element (STRI3) Trunnion

59

FEA shell element (STRI65) Cylinder

125

FEA shell element (STRI65) Trunnion

66

FEA continuum element (Shell) Cylinder

126

FEA continuum element Trunnion

70

Table 8

36 inch header, 30 inch trunnion, wall thickness =9.52 mm for

both (results shown for maximum Pb + Pl + Q in MPa)

WRC 107 Cylinder

307

WRC 107 Trunnion

NA

Kellogg Cylinder

100

Kellogg Trunnion

0.8

FEA shell element (S8R) Cylinder

146

FEA shell element (S8R) Trunnion

75

FEA shell element (STRI3) Cylinder

143

FEA shell element (STRI3) Trunnion

76

FEA shell element (STRI65) Cylinder

145

FEA shell element (STRI65) Trunnion

76

FEA continuum element (Shell) Cylinder

148

FEA continuum element Trunnion

74

Table 9

36 inch header, 12 inch trunnion, wall thickness = 9.52 mm for

header and 6.35 mm for Trunnion (results shown for

maximum Pb + Pl + Q in Mpa)

WRC 107 Cylinder

321

WRC 107 Trunnion

NA

Kellogg Cylinder

100

Kellogg Trunnion

0.8

FEA shell element (S8R) Cylinder

157

FEA shell element (S8R) Trunnion

108

FEA shell element (STRI3) Cylinder

155

FEA shell element (STRI3) Trunnion

102

FEA shell element (STRI65) Cylinder

159

FEA shell element (STRI65) Trunnion

98

FEA continuum element (Shell) Cylinder

154

FEA continuum element Trunnion

103

Results for Pipe Shoes: (Stresses at locations of singularities have not been considered)

Note: WRC-107 method has been used even though in most

cases

are concerned, the typically used dimensions render them unsuitable for use of WRC-107. Despite this fact, the author in his experience has seen its usage for computation of local stresses at Shoe Attachments and its use is mostly due to availability of this module in most common pipe stress programmes. For the WRC 107 computation of Pipe Shoes, the geometry of the attachment has been considered as 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.

2 are above the allowable limit. So far Pipe Shoes

b

1

,

b

Table 10

36” pipe, wall thickness = 9.52 mm. Shoe design corresponds to 3-gusset, A=450, B=500, shoe plate thickness = 10 mm, L = 350 mm (refer fig-1); magnitude of load = 40KN. Pressure is

not applied.

b

1

= 0.56 ,

b

2

= 0.49

Loading Type

Radial

Longitudinal

Circumferential

Force

Force

Force

WRC 107 Cylinder

181

18

61

Kellogg Cylinder

30

18

40

Kellogg Shoe

2

9

3

FEA Shell element (S8R) Cylinder

75

20

36

FEA Shell element (S8R) Shoe

77

30

32

FEA Shell element (STRI3) Cylinder

82

18

35

FEA Shell element (STRI3) Shoe

82

18

35

FEA Shell element (STRI65) Cylinder

75

22

37

FEA Shell element (STRI65) Shoe

80

30

33

FEA Continuum element Cylinder

81

24

33

FEA Continuum element Shoe

78

35

32

Table 11

30” Pipe, wall thickness 9.52 mm, Shoe design corresponds to 3 Gusset, A=450, B=500, Shoe plate thickness=10 mm, L=350 mm (refer fig-1), Magnitude of load=40KN. Pressure is not

applied.

b

1

= 0.67 ,

b

2

= 0.60

Loading Type

Radial

Longitudinal

Circumferential

Force

Force

Force

WRC 107 Cylinder

173

21

63

Kellogg Cylinder

27

17

38

Kellogg Shoe

2

9

3

FEA Shell element (S8R) Cylinder

60

15

18

FEA Shell element (S8R) Shoe

80

22

12

FEA Shell element (STRI3) Cylinder

62

14

18

FEA Shell element (STRI3) Shoe

75

22

13

FEA Shell element (STRI65) Cylinder

60

14

20

FEA Shell element (STRI65) Shoe

82

22

14

FEA Continuum element Cylinder

63

18

25

FEA Continuum element Shoe

82

20

16

Table 12

Table 14 (results shown for maximum Pb + Pl + Q in MPa)

24” pipe, wall thickness = 9.52 mm. Shoe design corresponds to 3-gusset, A=450, B=500, shoe plate thickness = 10 mm, L = 350 mm (refer fig-1); magnitude of load = 40KN. Pressure is

not applied.

b

1

= 0.84 ,

b

2

= 0.75

Loading Type

Radial

Longitudinal

Circumferential

Force

Force

Force

WRC 107 Cylinder

174

23

65

Kellogg Cylinder

24

15

34

Kellogg Shoe

2

9

3

FEA Shell element (S8R) Cylinder

35

12

22

FEA Shell element (S8R) Shoe

73

22

10

FEA Shell element (STRI3) Cylinder

34

12

22

FEA Shell element (STRI3) Shoe

50

22

10

FEA Shell element (STRI65) Cylinder

35

13

22

FEA Shell element (STRI65) Shoe

53

17

10

FEA Continuum element Cylinder

39

15

24

FEA Continuum element Shoe

57

21

13

For Tables 13-15, applied load in longitudinal, circumferential and radial directions = 40KN( applied together), pressure = 18.9 barg. Pressure has been applied but not as radial thrust load.

Table 13 (results shown for maximum Pb + Pl + Q in MPa)

36” header, wall thickness = 9.52 mm

WRC 107 Cylinder

330

WRC 107 Shoe

NA

Kellogg Cylinder

186

Kellogg Shoe

12

FEA shell element (S8R) Cylinder

180

FEA shell element (S8R) Shoe

155

FEA shell element (STRI3) Cylinder

184

FEA shell element (STRI3) Shoe

156

FEA shell element (STRI65) Cylinder

182

FEA shell element (STRI65) Shoe

153

FEA continuum element (Shell) Cylinder

188

FEA continuum element Shoe

159

30” header, wall thickness = 9.52 mm

WRC 107 Cylinder

312

WRC 107 Shoe

NA

Kellogg Cylinder

161

Kellogg Shoe

12

FEA shell element (S8R) Cylinder

126

FEA shell element (S8R) Shoe

115

FEA shell element (STRI3) Cylinder

131

FEA shell element (STRI3) Shoe

118

FEA shell element (STRI65) Cylinder

128

FEA shell element (STRI65) Shoe

113

FEA continuum element (Shell) Cylinder

132

FEA continuum element Shoe

119

Table 15 (results shown for maximum Pb + Pl + Q in MPa)

24” header, wall thickness = 9.52 mm

WRC 107 Cylinder

298

WRC 107 Shoe

NA

Kellogg Cylinder

136

Kellogg Shoe

12

FEA shell element (S8R) Cylinder

80

FEA shell element (S8R) Shoe

85

FEA shell element (STRI3) Cylinder

84

FEA shell element (STRI3) Shoe

89

FEA shell element (STRI65) Cylinder

82

FEA shell element (STRI65) Shoe

83

FEA continuum element (Shell) Cylinder

88

FEA continuum element Shoe

92

6.0 DISCUSSION OF RESULTS AND SCOPE FOR FUTURE WORK

Tables 1,2,4 show that WRC 107 and WRC 297 results show significant differences with respect to FE results for the

significant differences with respect to FE results for the radial load case. This is because of

radial load case. This is because of the high d D ratio and

radial as opposed to vertical load representation of the same in WRC 107 as explained in section 2.2 of this paper. Tables 3, 5 and 6 show that the results are comparable (even for the

5 and 6 show that the results are comparable (even for the Radial load case) indicating

Radial load case) indicating the criticality of the d D factor in

WRC 107/297 approaches. For the Kellogg Method, the significant difference is for the radial load case. This is because of the basis of the method being axi-symmetrical ring loading which significantly deviates from the actual mathematical model in the radial load situation. The Kellogg method also underestimates the stresses in the Trunnion. This is due to the use of simple beam theory as opposed to shell theory and the non-consideration of the compatibility requirement between the header pipe and the Trunnion in this method. Kellogg method also in most (but not all) cases predicts lower magnitude of stresses in the Longitudinal and Circumferential Force applications. However the allowable stresses in the Kellogg method as long as they are specified as the [28] allowable for local primary stress, the error will not in general make the analysis non-conservative except for the Radial Load scenario. For Tables 7, 8 and 9 which are for the combined load scenario, WRC 107 results show significantly higher magnitudes of Pb + Pl + Q with respect to FEA. Even though the Pressure loading has not been modeled as a Radial loading for these Tables, which would have resulted in even higher magnitudes of Pb + Pl + Q if the direction of this load would have been in the same direction as the additive radial load, but the simplistic way of computing pressure stresses also (as in Tables 7, 8 and 9) induces higher stresses in the WRC 107 type of analysis. Pressure induced loading at a cylinder to cylinder interface with or without other external loadings is complicated and WRC 107 analysis which considers the loading on the cylindrical surface as a rectangular loading cannot predict the stresses correctly and will err on the conservative side for most cases. WRC 107 /297 analysis has shown lower magnitudes of Stress for Shear Forces and Torsion moments (Table 6 where the loadings have been applied at the Shell-Nozzle Interface) with respect to FEA. However, these loadings, in general are not the governing loads in piping applications. When using WRC- 107/297 modules of a Pipe Stress Program, an analyst should review the program document to see how pressure is modeled in these modules.

Results in tables 10, 11 and 12 again show that the pattern of variance in results between FEA and WRC 107/297 is most significant with respect to radial loads. The reason can still be

with respect to radial loads. The reason can still be attributed to the d D even

attributed to the d D even though in case of the shoe

dimension “d” is strictly not applicable but the dimension of

the pipe shoe can be seen as analogous to this parameter. Significant differences exist for the Circumferential loading case also. For both Trunnion and Pipe Shoes, for some cases , specially for the Radial Load scenario, stresses in the shoes/Trunnion elements exceed stresses in the cylinder which clearly shows the risk of using the Kellogg method for computing stresses in the Pipe support attachments. A point to note is that, the method of computing stresses in the Pipe supports cannot be technically stated as “Kellogg Method” as [4] only discusses computation of local stresses in the Cylinder. The context of using the term “ Kellogg method” for the method of computing stresses in attachments is due the fact that this computation based on elementary beam theory is an essential feature ( in author’s experience) of the spreadsheets which use the Kellogg method to compute Local stresses in the Cylinder. Hence the caution is using elementary beam theory analysis for computation of local stresses in Attachments. Significant differences in results have not been seen in Finite element approach using different element types. This however should not be taken as a blanket statement as the models had proper mesh grading with adequately small element size and the element distortion control was well within the recommended limits of the FE code. For improper mesh grading, element size, significantly distorted elements and improper integration methods, significant differences in results can be seen between the elements, especially for Triangular elements which suffer from geometric anisotropy. The stress analyst should carefully study the theory manual of the FE code which he/she should be using with respect to applicability, element distortion and integration rules.

The present analysis has to be extended for different load combinations with varying magnitudes of the individual load vectors to quantify the degree of over or under conservatism of the available analytical methods. The present analysis mainly focuses on the stand alone effect of individual load vectors (although Tables 7-9 as well as Tables 13-15 does address combinations but more tests need to be done with varying magnitudes of the individual load vectors) . Effect of variance in mesh grading and element size should be checked to assist an analyst in selection of the “best element” for these applications, if an analyst so desires. In the present scope of work, the use of proper mesh grading, element size and integration rules have ironed out significant differences between the individual elements. Hence, the take away message for an analyst with respect to individual element types is, as long as mesh grading , element size, distortion control and integration rules are properly used, there are no preferred elements , although the analyst should carefully read the Theory manual of the FE code which he/she intends to use.

7.0 CONCLUSIONS:

1. Use of a particular shell theory requires an understanding of the order of magnitude of error inherent in that theory and its

applicability vis-à-vis the problem to be analyzed specially

applicability vis-à-vis the problem to be analyzed specially with reference to d D and D T
applicability vis-à-vis the problem to be analyzed specially with reference to d D and D T

with reference to d D and D T ratios.

2. Shell theories should be evaluated on the basis of Sander-

Koiter postulates.

3. The use of an axi-symmetric loading model (which in this

paper has been referenced to as “Kellogg method”) has been

historically the most popular method for analyzing both cylindrical and non-cylindrical attachments.

4. WRC-107 method which is based on Timoshenko equations

as Morley, Simmonds and

Goldenveizer equations. WRC-107 results may be more or less conservative than FE results. Results are generally overly

. The analysis results show that in

some cases but not all (generally computation as per Kellogg method has shown lower magnitude of stresses with respect to WRC 107 or FE analysis), Kellogg method significantly underestimates the stresses in Trunnion and Pipe Shoes. Hence it is recommended that this method should not be used and hence should not be used for evaluating stresses in Pipe supports. A point to note is, the method as at appears in [4] addresses only the local stresses at the cylinder, so the evaluation of stresses in attachments cannot technically be addressed as “Kellogg method”, rather calculation based on elementary beam theory. It is against this later which, the author in his experience has seen as widely used in the Industry as part of the spreadsheets based on “Kellogg method” is what this caution is directed at.

conservative for

has the same error

O

( T ) R
(
T
)
R

.

0 5

at. conservative for has the same error O ( T ) R . 0 5 D

D >

5. If an analyst is constrained to use Kellogg method for

analysis of local stresses on the pipe at support locations, the

allowable stress should not be exceeded beyond the allowable for local primary stresses as per [28].

6. WRC-297 method is based on shallow shell theory and the

order of magnitude in error is due to omission of some terms

which are of the order

conservative behavior, specially for the Trunnion stresses for most of the cases analyzed. Use of WRC 297 for Pipe support attachments is not recommended.

7. When comparing results between an analytical and FE

approach, it is best to check the model on a component by component basis i.e. the model is loaded with only one force/moment component in the absence of pressure. This check will show stresses because of which components are over/under represented in the final results. Since WRC- 107/297 does not have a provision for checking pressure loading, simulating the same by a modified radial load (= applied radial load + pressure times area) or superposing the results with the usual membrane stresses in the header pipe due to pressure generally makes the analysis over- conservative. The modification of the radial load in a WRC-

and has shown overly

O

( T ) R
(
T
)
R

107/297 based analysis is also not correct as the branch is not pressurized for Pipe support applications where by “branch” we mean the Trunnion.

8. Finite elements for shell analysis have different approaches

based on the theoretical considerations that form the basis of

their developments, with elements based on basic shell mathematical model being least popular because of the problem of addressing rigid body motion. Commercial FE codes should be having Hybrid elements in the element library for shell applications.

9. Degenerated solid elements have used adhoc assumptions

on shell theory to work within the constraint of finite element

formulation. Assumptions regarding the mathematical form of dependence of the determinant of the Jacobian Matrix on the thickness direction coordinate can lead to violation of rigid body properties.

10. Not much difference has been found in results using 8-

node reduced integration shell element developed on the line of Mindlin hypothesis, triangular elements based on discrete Kirchoff constraints (imposed analytically or numerically) and use of solid elements for circular attachments. Stresses at locations of singularity have to be carefully addressed [29]. The pattern of results i.e. relative invariance with respect to element types need not be always correct depending on the

to element types need not be always correct depending on the D T ratio, element distortion,

D T ratio, element distortion, element size and use of

alternate numerical integration rules. In general, as long as thin shell theory is valid and reduced integration rule is used for shear flexible elements, with proper mesh grading and keeping the element size at the intersection region

and keeping the element size at the intersection region significantly less than r t , type

significantly less than rt , type of element is usually not a significant parameter. Stress Analyst should carefully review the Technical Manual(s) of FE Code for the capabilities and limitations of the available elements from the element library.

11.

of the available elements from the element library. 11. Analytical methods with d D as high

Analytical methods with d D as high as 1.0 with ease of

implementation is required not only because the available

methods like WRC107/297 etc are inadequate for such applications but also as a tool to properly benchmark the FE results. Till such time, FE models will continue to be benchmarked against WRC 107 type of analysis for similar loading within the limits of the applicable geometry.

12. Additional tests need to be done for Pipe Shoes for varying

Additional tests need to be done for Pipe Shoes for varying effects of D T and

effects of D T and combined loadings. In author ’s opinion it

is futile to expect usability of WRC-107 for shoe attachments, as based on typical dimensions of Pipe Shoes, these geometric parameters will in most cases be not satisfied.

13. WRC 107 /297 analysis has shown lower magnitudes of

Stress for Shear Forces and Torsion moments (Table 6 where the loadings have been applied at the Shell-Nozzle Interface) with respect to FEA. However, these loadings, in general are not the governing factors in piping applications.

ACKNOWLEDGMENTS

The author wishes to acknowledge Professor M.D.Xue of Tsinghua University, Department of Engineering Mechanics for providing some valuable suggestions and document references and for answering some questions on her paper. The author also wants to thank Dr.Subrata Saha of Reliance Industries Ltd India, Mr. Suraj Kunder of Costain UK and ex- colleague and friend Mr.Arijit Chatterjee for providing valuable guideline and suggestions.

REFERENCES

Shells Subjected to External Forces on Nozzle, ASME Journal of Pressure Vessel Technology, Vol. 128. [17] Ming-De Xue, Qing-Hai Du, Keh-Chih Hwang, Zhi- Hai Xiang, 2010, An analytical method for cylindrical shells with nozzles due to internal pressure and external loads – Part ITheoretical foundation; ASME Journal of Pressure Vessel Technology Vol 132

[18]

Bathe, K.J.,Lee, P.S., 2005, Insight into finite element shell discretizations by use of the ‘‘basic shell mathematical model’’ Computers and Structures Vol. 83

[19] Mac Neal, R.H., Finite Elements their design and performance, Marcel-Dekker, NY. [20] Theory Manual ABAQUS, a product of Dassault

[1]

Timoshenko, S: 1940, Theory of plates and shells, MC-

Systems, SA.

Graw Hill, New York

[21]

Mollman, M, 1981, Introduction to the Theory of Thin

[2] Bijlaard, P.P., 1955, “Stresses from Radial loads and External moments in Cylindrical pressure vessels”, Welding Journal, Miami, FL, US Vol. 34

[22]

Shells, Wiley-Blackwell. Flugge, W., Tensor analysis and Continuum Mechanics, Springer, Berlin.

[3]

[23]

Chen., M.W.L.Y., Li, J.G., 2000, A two step approach of

Steele, C.R and Steele, M.L.M 1983, “Stress analysis of Nozzle in Cylindrical vessels with external Load, ASME Journal of Pressure Vessel Technology, Vol. 105

[4] M.W.Kellogg Company,2011,“Design of Piping

Stress Classification and Primary Structure Method. Trans ASME Vol. 122 [24] Xue,L.,Widera, G.E.O.,Sang, Zhifu., Feb. 2006,

[25] Private communication, Professor M.D.Xue of

[5]

Systems” Mehrson, J.L, Mokhtarian, K., Ranjan, G.V and

in the General Theory of Elastic Shells”, Proceedings of

Flexibility Factors for Branch Pipe Connections subjected to in-plane and out of plane moments ,Journal

[6]

Rodabaugh, E.C., 1984, “Local Stresses in Cylindrical Shells due to External Loadings on Nozzle Supplement to WRC bulletin 107”WRC Bulletin No. 297 Goldenveizer A.L., 1961, Theory of Elastic Thin shells,

of Pressure Vessel Technology

Tsinghua Universit y, People’s Republic of China. [26] Bhattacharya A, 2012, A Comparison of Simple

[7]

Pergamon, Oxford, Koiter, W.T, 1959. “A Consistent First Approximation

Analytical Methods for evaluating local Stresses at Pipe Supports with Finite Element Analysis results NAFEMS UK Conference 2012.

the Symposium on the Theory of Thin Elastic Shells, Delft, the Netherlands, W.T.Koiter ed., North-Holland, Amsterdam.

[27]

Bhattacharya A, 2011, A Finite Element based Study on Stress Intensification Factors (SIF) for reinforced Fabricated Tees NAFEMS World Congress Boston

[8]

Donnell, L.H., 1933, “Stability of Thin Walled Tubes

2011

 

[28]

under Torsion”, NACA Report No. 479 Morley, L.S.D., 1959, “An Improvement on Donnell’s Approximations for Thin walled Circular Cylinders”, Q.J.Mech., Appl. Math., Vol. 12

[10] Donnell, L.H, 1976, Beams, Plates and Shells, Mc-

[9]

ASME Boiler and Pressure Ve ssel code, Sec VIII Div2 (2007 Edition), ASME Publication

[29] Kalnins, A, 2008, Stress Classification Lines Straight through Singularities ASME PVP Conference, Chicago, ILL, USA

 

Graw -Hill, New York, Chapter 6

[30]

Bathe, K.J., Chapelle, D., 2010, Finite Element

[11]

Simmonds , J.G., 1966, A Set of Accurate Equations for

Analysis of Shells Fundamentals, Springer.

Circular Cylindrical Elastic Shells, “,Int. J. Solids

[31]

Xue, M.D., Li, D.F., Hwang, K.C, 2005, A Thin Shell

[12]

Structure., Vol. 2, Budiansky,B and Sanders , J.L, 1963, “ On the best first First Order Linear Shell Theory”, Progress in Applied Mechanics (The Prager Anniversary Volume), Macmillan, London.

Theoretical Solution for Two Intersecting Cylindrical Shells Due to External Branch Pipe Moments ASME Journal of PVT Vol. 127 [32] Wong, F.M.G, 1984, Stresses and Flexibilities for Pressure Vessel Attachments. Dissertation thesis for

[13]

Flugge, W.1967, Stresses in Shells, Springer, Berlin.

Master of Science in Nuclear Engineering.

[14]

Hoff, N.J., 1955, The accuracy of Donnell’s equations

[33] Morley, L.S.D, Morris, A.J.1978, Conflict between

[15]

J.Appl. Mech Vol. 22 Lekkerkerker, J.G., 1972. The Determination of Elastic Stresses near Cylinder-to-Cylinder Intersection. Nuclear

Finite Elements and Shell Theory, Paper presented at Second World Congress on Finite Element Methods, Bournemouth, UK.

[16]

Eng. Des., Volume 20. Xue, M.D., Du, Q.H., Li, D.F. and Hwang, K.C., 2006. Theoretical Stress Analysis of Intersecting Cylindrical

[34]

Wichman, K.R., Hooper, A.G., Mehrson, J.L., Stresses in Spherical Shells due to External Loadings, Welding Research Council Bulletin No. 107.

[35] Koves, W., Mokhtarian, K., Rodabaugh, E., Widera, W., 2006, Large Diameter Shell Intersections, Welding Research Council Bulletin No. 497.