Michael A. Porter
Dynamic Analysis
Lenexa, Kansas
Dennis H. Martens
Black and Veatch Pritchard
Overland Park, Kansas
S. M. Caldwell
Eastman Chemical Company
Kingsport, Tennessee
ABSTRACT
A procedure for evaluating the results of a finite element analysis
employing shell/plate elements is proposed based on several previous
papers by the authors and a review of other related works. This
procedure relates the stress levels produced by the finite element
software to the provisions of ASME Section VIII, Division 2.
WHAT IS NEEDED AND WHY?
A procedure for the evaluation of 3D Finite element models
according to the ASME Section III and VIII Code requirements has
been published by the Pressure Vessel Research Council (PVRC)
(Hechmer and Hollinger, 1997). This document summarizes and
expands upon earlier papers by Hechmer and Hollinger (1987 and
1991) as well as incorporating the stress linearization concepts
published by Kroenke (1974) and Kroenke et al (1975). Together,
these works present a clear description of the process necessary to
evaluate 2D and 3D Solid element models as well as axisymmetric
element models.
The PVRC recommended evaluation procedure consists of
selecting an appropriate Stress Classification Line (SCL) in the model
and then linearizing the stresses along this line to identify the primary,
secondary and peak components of the stress for comparison with the
Code criteria. The selection and evaluation process for the SCL is
covered in great detail in the PVRC report. The process for linearizing
the stresses is also presented in a conceptual format. Currently, a task
force of the Design Subcommittee is working on the language
necessary to incorporate these procedures into the Code.
The PVRC work dealt exclusively with 3D Solid and
Axisymmetric type element models; no work using shell and/or plate
element models was included. In a significant portion of the industry
that is subject to the ASME Code, the vessels tend to be relatively
large diameter and thin wall (r/t >10) in configuration. Shell, plate
and 3D solid elements may all be used for applications where the
diametertowall thickness ratio is greater than 10. Due primarily to
economic and time constraints however, it is typically not practical to
employ 3D solid elements in the analysis of many general industry
vessels. The best tool that we can afford is more often the use of
shell and/or plate elements. Thus, the evaluation of shell and/or plate
element models in relationship to the Code should be addressed.
The stresses reported by shell elements are inherently linear. No
separate linearization process is required when evaluating a shell
element FE model. The only questions (assuming that the FE model is
itself constructed correctly) are which stresses to evaluate and where.
In a previous paper (Porter and Martens, 1996) the authors outlined a
procedure for such evaluation. While this earlier procedure does
provide some guidance for evaluation, it is not entirely consistent with
the PVRC recommendations. In the current paper we propose a
revised procedure which we believe to be more consistent with the
current PVRC document.
Notes and Limitations
Note: in the remainder of this paper these elements will be
referred to as simply shell elements. It is to be understood that this
term refers to both shell and thin plate type elements where
appropriate. The distinction between, and the proper employment of
the various element formulations and types, is the subject of another
paper (Porter, Martens, and Marcal, 1999).
The nozzle investigated in this paper is a 90degree nozzle in a
thin wall shell. This procedure may not be directly applicable to other
than 90degree junctions or to thick wall vessels. Additionally, the
results presented in this paper have been derived using linear thin plate
elements. Different numerical results will likely be obtained using
thick plate and/or shell elements.
Based on the material and temperatures examined in this model,
creep was not considered as a factor. For many common materials at
higher temperatures, the creep strength rather than the criteria cited in
this paper might be the limiting factor. The basis for establishing the
allowable stress based on creep is contained in ASME Section VIII
Division 1, Section II, Part D. From inspection, it is apparent that the
use of 1.1, 1.5 and 3 Sm values will effect the life expectancy for
Secondary stress that will be sustained throughout the creep
deformation of the nozzle structure. Similarly, consideration for life
expectancy due to cyclic applied loading must be addressed per the
Division 2 Appendix 5 methodology (Martens and Hsieh 1998).
WHERE TO EVALUATE
The first issue to be considered is where the stresses will be
evaluated. Here we have specific guidance from the PVRC document.
Figure 1 illustrates a cross section through a typical nozzle/shell
intersection. In this case, the nozzle is vertical with a wall thickness t
while the shell is horizontal with a wall thickness of T. Also indicated
is a reenforcement pad with a thickness of P (typically P =T on thin
wall vessels). A fillet weld having a radius of r is also indicated at the
intersection. This fillet is required by ASME Section VIII, Division 1,
UW16, 1998 to have a minimum throat thickness of 0.7 times the
minimum thickness of two joining parts or , whichever is less. The
fillet usually does not have a concave radius and the legs of the typical
fillet are, for practical proposes, equal to throat minimum thickness
divided by 0.707. Therefore the minimum fillet leg length is
approximately equal to the minimum thickness of the two joining parts
or , whichever is less. Typically, this intersection would be modeled
using shell elements along the lines indicated. With shell elements it
is difficult, if not impossible to actually model the fillet, so it is
typically omitted.
The area that is crosshatched on Figure 1 is referred to as the
juncturering (or sometimes simply the ring) in the PVRC
document. Stresses (except for fatigue purposes) are not to be
evaluated in the juncturering according to the PVRC document.
Thus, the closest points to the intersection of the nozzle and shell
where stresses are to be evaluated are indicated by the distances a and
b on the shell and nozzle respectively.
From a practical modeling standpoint, if the shell elements are
constructed such that a line of nodes is created at distances a and b
from the intersection, evaluation of the stresses at these locations is not
at all difficult. With most commercial PCbased FE products, this is
not a difficult task. The minimum required distances a and b are
related to the joint dimensions by a =t/2 +t or 1 t and b =P +t.
Additionally, it is normally possible to eliminate elements during the
postprocessing stage such that the stresses on the eliminated elements
are not reported. Thus, evaluating the stresses at distances a and b
away from the intersection is a relatively straightforward procedure.
Figure 2 illustrates the configuration of a model of a nozzle/shell
intersection. The intersection modeled consists of a 24 diameter,
wall nozzle in a 96 diameter, wall shell. A thick, 42 diameter
reinforcement pad is modeled on the shell. Using these dimensions, a
=+= and b =+=1. The model was loaded as
follows:
Internal Pressure 165 PSI
Nozzle Force Z: 6,480 LB
Nozzle Moment X: 25,500 FTLB
Nozzle Moment Y: 33,160 FTLB
Nozzle Moment Z: 33,160 FTLB
INDICATED STRESSES
Figure 3 illustrates the indicated Stress Intensities in the nozzle
under all loads. Note that the elements within the junctionring
region have been hidden. As expected, the highest stresses are
indicated on the line of nodes adjacent to the hidden juncturering.
The stresses indicated in this illustration represent the combination of
both primary and secondary stresses on the nozzle (PL +Q) and are to
be compared with 3Sm. Figure 4 illustrates the membrane stresses in
the same region. If we consider these to be the local membrane
stresses (PL), they are to be compared with 1.5Sm. Note that this
model was refined several times to ensure convergence. The finished
model had 192 elements around the circumference of the nozzle. This
is a greater number of elements than the 96 elements that Ha (1995)
indicates is the minimum required for convergence with 90degree
pipe nozzle intersections.
While pictorial plots of the stresses such as Figures 3 and 4 are
informative, and perhaps all that is necessary to evaluate this nozzle,
there are other presentations of the stresses that are also useful. Figure
5 illustrates the stresses in the nozzle region as a function of the
distance away from the intersection between the nozzle and the shell.
These stresses have been assessed along a radial line passing through
the center of the nozzle and passing through the point of indicated
highest stress. The indicated stresses in Figure 5 are due to the
application of pressure, force and moment loading on the nozzle.
Figure 6 illustrates the membrane stress due to the combined loading
and due to pressure loading alone. It should be noted that the line used
for the membrane stress plot was not the same as for the stress
intensity, because the location of the maximum varied depending on
the type of stress being evaluated. Similarly, the location of the
highest stressed area changes between the pressure only and the
pressure plus moments loading conditions.
DISCUSSION OF STRESSES
The evaluation of stresses when complex configurations and
loadings are involved may be accomplished using the methodology in
Division 2. The code allowable stress must be based on the Division to
which the design and fabrication is conducted. With the indicated
stresses now plotted, the task becomes deciding which stress to
compare with which code criteria. Referring to Figure 4130.1 of
ASME Section VIII, Division 2 Appendix 4, there are essentially three
levels of stress used for compliance analysis: kSm, 1.5kSm and 3Sm.
For this discussion, we will assume that k=1 and we will leave
discussion of the peak stress and the associated Sa for another paper.
General Primary Membrane  Sm
It is clear that well away from the nozzle the membrane stresses
in both the shell and nozzle are below Sm. As a measure of how close
to the nozzle we can be before the membrane stress is allowed to
exceed Sm, we can take guidance from ASME Section VIII, Division
2 Appendix 4, 4112, (I), the definition for Local Primary Stress which
states:
A stress region may be considered as local if the distance over
which the stress intensity exceeds 1.1 Sm does not extend in the
meridional direction more than 1.0
Rt
where R is the midsurface
radius of curvature measured normal to the surface of the axis of
rotation and t is the minimum thickness of the region considered.
Using this criteria, the Stress Intensity should not exceed 22.0 ksi
(1.1Sm for A 51670 up to 500 F per Division 1) at a distance of
approximately 2.5 and 6.9 fromthe nozzle/shell intersection in the
nozzle and shell directions respectively. The definition cited above
was included in Appendix 4 as a guideline based on the observation of
the stress decay with distance in a number of examples. The important
aspect of this evaluation is to be assured that the stresses are trending
toward Sm when they reach this distance.
This nozzle meets the criteria in both directions. At a somewhat
greater distance, the membrane stresses are well below the Sm
allowable for the primary membrane stress. Thus, the Sm criteria
would appear to be met. It should be noted that, in general, FE is not
needed to evaluate compliance with the Primary membrane criteria.
The familiar =pr/t formula for the hoop stress in a cylinder is the
preferred means of checking compliance.
=
max
ave
Primary + Secondary 3Sm
It is clear from Figures 5 and 6 that the stresses tend to increase
as the distance from the intersection between the nozzle and shell is
decreased. The maximum indicated stress in all cases is at the node
adjacent to the junctionring. Comparing the maximum indicated
Stress Intensity (either from Figure 3 or Figure 5) with 3Sm to assess
compliance would be appropriate. Assuming that the material is SA
51670 (@ 500 deg), Division 1 Sm is 20.0 ksi. The maximum
indicated stress intensity is less than 58 ksi, so that compliance with
the 3Sm criteria for Primary plus Secondary (PL +Q) is indicated.
Note that by definition, PB does not apply in the nozzle region.
Local Primary Membrane 1.5Sm
The Local Membrane criterion requires that the membrane stress
not exceed 1.5Sm, in this case 30.7 ksi. From Figure 6, it is clear that
this stress level is met on both the nozzle and shell at the nodal SCL
line adjacent to the juncturering when the loading is due to pressure
only. When the piping load is applied, the local membrane stress at
the junction ring exceeds the 1.5 Sm criterion.
It is interesting to note that for many years the primary method
for computing the stress in nozzles that have a mechanical loading has
been WRC107. The stresses computed with this procedure were then
compared with the Code allowables. WRC107 does not classify the
stresses. General practice in many industries has been to compare
only the Primary plus Secondary (P +Q) Stress Intensity due to the
mechanical loading to 1.5Sm. While the membrane stresses are
computed, they are not always compared separately with the 1.5Sm
criterion.
The WRC107 procedure computes the stress only at four points
(0, 90, 180 and 270 degrees), none of which is necessarily the point of
maximum stress. When the moment (or force) loads are nearly equal
for the two axes perpendicular to the axis of the nozzle, the highest
stress occurs at other than the points where stress is computed by
WRC107. Thus, the use of FE analysis for computing nozzle stresses
can be considerably more conservative than WRC107.
As documented in an earlier paper (Porter and Martens, 1996),
this nozzle meets the Code criteria when the WRC107 stress
computation procedure is used. We can see from Figure 6 that the
local membrane criterion level of 30.75 ksi is closely approached in
the nozzle under pressure loading only. Typical bending loads
encountered in real life situations push the membrane stress over the
limit at the junctionring. The authors have found this to be the case in
a (yet to be published) wide range of typical nozzle and loading cases.
In fact, it is the exception rather than the rule when a conventionally
designed nozzle will meet the local membrane criterion at the
junctionring if the bending loads are included.
Koves and Sanger (1996) have indicated that a better method of
evaluating the membrane stress would be to average the stress within
RT 78 . 0
of the intersection. In a newly published paper, Koves
(1999) presents an equation that relates the average stress in this area
to the maximum indicated stress. Using this relationship, we can
compute the average local membrane stress based on the maximum
indicated stress as illustrated in Equation 1.
Equation (1)
Where:
ave
=Average local membrane stress within RT 78 . 0 of the
intersection
max
=Maximum local membrane stress within
RT 78 . 0
of the
intersection
L =
RT
R =Mean radius of shell
T =Thickness of shell
a =Mean radius of nozzle
Equation 1 is based on the analysis of numerous actual burst tests
of nozzles and is considered to be a conservative estimate of the stress.
Thus, if we divide the allowable local membrane stress (1.5Sm) by the
right hand side of Equation 1, we get an allowable stress level that can
be compared directly with the maximum indicated stress from the FE
model. Using this procedure, the allowable local membrane stress
becomes:
Equation (2)
With these facts in mind, it is the authors opinion that using the
Primary plus Secondary (PL +Q) Stress Intensity <3Sm criterion
only procedures is sufficient for evaluating the stress in a nozzle
analyzed using FE and is just as conservative as using the current
WRC107 stress computation procedure. It is possible that the local
membrane stress need not be evaluated at the juncturering. However,
pending further investigation, the average stress computation as
developed by Koves (1999) would seem to be an appropriate way to
evaluate the local membrane stress. The 1.1Sm criteria is applicable at
the distance from the nozzle/shell as stated in Section VIII, Division 2
Appendix 4, 4112, (I).
ACCEPTANCE CRITERIA
Using the Sm, 1.1Sm and 3Sm values and positions described, a
stress acceptance line can be constructed on Figure 5 (shown as Figure
7). The stress variance from the 1.1Sm location to the 3Sm location is
assumed to be linear for acceptance. However, on inspection, the FE
generated stress is a curve. This presentation is of value to the engineer
for assuring that the stress contour is smooth and consistent. The
presentation can easily be adapted to a post processor to present a
graphic solution to compare the code allowable stresses versus FE
developed stresses and the stress profile within the joint. The authors
submit that a visual graphic presentation facilitates the engineers
review of the stress data and code compliance.
Figure 7 addresses the Appendix 4 Primary plus Secondary stress.
The Sm evaluation location is placed at
rt 5 . 2
, the 1.1 Sm location is
placed at
rt
, and the 3Sm location is placed at the SCL. These
values are compared to the Stress Intensity (Tresca * 2) stress values
from the analysis.
0.4
a
L
0.4
0.3
1
L
a
0.1
1
L
a
3
.
1.5Sm
.
0.4
a
L
0.4
0.3
1
L
a
0.1
1
L
a
3
.
In a similar manner, the local primary membrane stress can be
evaluated in accordance with the criteria as indicated in Figure 8,
which addresses the Appendix 4 Primary stresses resulting from
design loads of pressure and weight. The same locations used for the
Primary plus Secondary stress evaluation are used here, but the
maximum allowable stress is given by Equation (2) rather than 3Sm.
The criteria values are compared to the computed membrane stress.
When the piping loads are added to the pressure and weight loads, the
membrane stress exceeds the 1.5 Sm allowable, but not the criteria
based on the Koves (1999) paper. As earlier stated, the authors have
found this to be more the rule than the exception on a significant
number of thin wall nozzles that have been examined. We believe that
this is an indication that the application of the 1.5 Sm limit for the
local membrane stress at the junction ring may be overly conservative
and that the Koves value is more realistic.
RECOMMENDED ASSESSMENT PROCEDURE
In order to assess the stress in a thin wall (r/t and R/T >10)
nozzle/shell junction using FE and shell elements, the authors
recommend that the following procedure be followed:
1. The nozzle should be modeled using shell elements with a
minimum of 96 elements around the circumference of the
nozzle. This assumes that linear plate elements are being
employed. If higher order shell elements are used, a lesser
number of elements may be required. A convergence
analysis or some other verifiable check must be employed to
assess the element behavior if fewer elements are employed.
2. The model should be constructed to ensure that a row of
nodes is located at a distance of 1.5t from the junction on the
shell side and P+t on the nozzle side.
3. The elements used should have a lengthtowidth aspect
ratio less than 2.0.
4. No transition in element size should be made within the pad
area and/or within rt (or 10 times the thickness, whichever
is greater) of the intersection on the nozzle and shell
respectively.
5. All operating loads including gravity, pressure, thermal and
external forces and moments, should be applied to the model
for the surface Stress Intensity (3Sm) check. The Local
Membrane under combined loading should also be evaluated
at the junctionring.
6. The computed Stress Intensity on the lines of elements 1.5t
on the shell and P+t on the nozzle shall be compared with
the 3Sm criterion. If the indicated Stress Intensities do not
exceed the criterion level, the nozzle meets the code
requirements. The local membrane stress at the junction
ring should be checked against the Equation 2 criteria.
Note: The 3Sm criterion is based on a range of stresses. In the
case of thermal loading it may be necessary to consider several loading
conditions to evaluate the proper range.
The above procedure assumes that the engineer will employ the
proper elements and verification methods to ensure the validity of the
FE model used for the analysis.
References
ASME (1998), "ASME Boiler and Pressure Vessel Code",
Section VIII, Division 2, The American Society of Mechanical
Engineers, New York, NY.
Ha, J .L, Sun, B.C., and Koplik, B., 1995, Local Stress Factors of
a PipeNozzle Under Internal Pressure, Nuclear Engineering and
Design, 157, Elsevier Science S.A., Lausanne, Switzerland, pp. 8191.
Hechmer, J . L. and Hollinger, G. L., 1987, Three Dimensional
Stress Criteria Application of Code Rules, PVP Vol. 120, Design
and Analysis of Piping, Pressure Vessels and Components, W. E.
Short II, et al., ed., ASME, New York, NY, pp. 189196.
Hechmer, J . L. and Hollinger, G. L., 1989, Code Evaluation of
3D Stresses on a Plane, PVP Vol. 161, Codes and Standards
Applications for Design and Analysis of Piping, Pressure Vessels and
Piping Components, J . P. Breen, et al., ed., ASME, New York, NY,
pp. 3345.
Hechmer, J . L. and Hollinger, G. L., 1991, Three Dimensional
Stress Criteria, PVP Vol. 210.2, Codes and Standards Applications
for Design and Analysis of Piping, Pressure Vessels and Piping
Components, ASME, New York, NY, pp. 181191.
Hechmer, J . L. and Hollinger, G. L., 1991, The ASME Code and
3D Stress Evaluation, Journal of Pressure Vessel Technology, Vol.
113, ASME, New York, NY, pp. 481487.
Hechmer, J . L. and Hollinger, G. L., 1997, 3D Stress Criteria:
Guidelines for Application, PVRC Grant 9114 Final Report, ASME,
New York, NY.
Koves, W. J . and Sanger, R. J ., 1996, Evaluation of Pressure
Design Criteria for Nozzles, International Conference on Pressure
Vessel Technology, Vol. 2, ASME, New York, NY, pp. 271279.
Koves, W. J ., 1999, Evaluation of Pressure Design Criteria For
Nozzles (II), to be presented at PVP99, Boston, MA, ASME, New
York, NY.
Kroenke, W. C., 1974, Classification of Finite Element Stresses
According to ASME Section III Stress Criteria, Pressure Vessels and
Piping, Analysis and Computers, ASME, New York, NY, pp. 107140.
Kroenke, W. C., Addicott, G.W., and Hinton, B.M., 1975,
Interpretation of Finite Element Stresses According to ASME Section
III, 75PVP63, ASME, New York, NY, pp. 112.
Martens, D. H., and Hsieh, C. S., 1998, Finite Element
Investigation of a CBA Reactor for the Effects of Thermal Loading,
PVP Vol. 368, ASME, New York, NY, pp. 139146.
Porter, M. A. and Martens, D. H., 1996, "A Comparison of the
Stress Results from Several Commercial Finite Element Codes with
ASME Section VIII, Division 2 Requirements," PVP Vol. 336, ASME,
New York, NY., pp. 341346.
Porter, M. A., Martens, D. H., and Hsieh, C. S., 1997, "A
Comparison Finite Element Codes and Recommended Investigation
Methodology," PVP Vol. 359, ASME, New York, NY., pp. 241246.
Porter, M. A. and Martens, D. H., 1998, Stress Evaluation of a
Typical Vessel Nozzle Using PVRC 3D Stress Criteria: Guidelines for
Application," PVP Vol. 368, ASME, New York, NY., pp. 297301.
t
b
a
T
P
Centerline of plate element
r
Pad
Shell
Figure 1 Cross section of nozzle/shell junction Figure 2 Nozzle/shell intersection model
Figure 3 Stress Intensity in nozzle/shell intersection model Figure 4 Membrane stress in nozzle/shell intersection model
Stress Intensity Profile  Operating Loads
0
10000
20000
30000
40000
50000
60000
12 10 8 6 4 2 0 2 4 6 8
Di stance from Intersecti on  In
S
t
r
e
s
s

P
S
I
Stress Intensity  Inside
Stress Intensity  Outside
Nozzl e  Shel l
Membrane Stress Profiles
0
10000
20000
30000
40000
50000
60000
12 10 8 6 4 2 0 2 4 6 8
Di stance From Intersecti on  In
S
S
t
r
e
s
s

P
S
I
Presure +Piping Loads
Pressure only
Nozzl e  Shel l
Figure 5 Stress Intensity Figure 6 Membrane Stress
Stress Profile  Design (Pressure) Load
0
10000
20000
30000
40000
50000
60000
12 10 8 6 4 2 0 2 4 6 8
Di stance From Intersecti on  In
S
S
t
r
e
s
s

P
S
I
Limit Curve
Pressure +Piping loads
Pressure Only
Nozzl e  Shel l
(rt)^0.5
1.1 Sm
(rt)^0.5
1.1 Sm
2.5(rt)^0.5
Sm
1.5Sm
Stress Profile  All Loads
0
10000
20000
30000
40000
50000
60000
12 10 8 6 4 2 0 2 4 6 8
Distance from Intersection  In
Limit Curve
Stress Intensity  Inside
Stress Intensity  Outside
Nozzle  Shell
(rt)^0.5
1.1 Sm
(rt)^0.5
1.1 Sm
2.5(rt)^0.5
Sm
3Sm
Figure 7 Stress Intensity evaluation Figure 8 Membrane Stress evaluation